CRANFIELD UNIVERSITY

Paul J Bradley

Fundamental Study into the Governing

Conditions of Rotor Thermal Bows in
Hydrodynamic Bearings

School of Engineering

PhD Thesis
Academic Year :2011 - 2012

Supervisor : Professor David Mba

March 2012

© Cranfield University, 2012. All rights reserved. No part of this publication may be
reproduced without the written permission of the copyright holder.
Fundamental Study into the Governing
Conditions of Rotor Thermal Bows in
Hydrodynamic Bearings

School of Engineering

Paul J Bradley

CRANFIELD UNIVERSITY

6th March 2012

Declaration

The work contained in this thesis has not been submitted elsewhere for any
other degree or qualification, and unless otherwise referenced it is the authors
own work.
Abstract
This thesis presents an experimental investigation into the conditions
generated within a high speed bearing system where thermally generated
shaft bows can arise from differential journal heating, and under certain
condtions this effect has been observed to result in unstable shaft vibration -
the ‘Morton Effect’. This thesis documents the development of a simplified
analytical procedure for evaluating the thermal activity within an orbiting
journal when running in hydrodynamic bearings. The aim of this work was to
generate controlled experimental data regarding journal differential heating
effects to support the understanding and development of appropriate
modelling and predictive techniques.

A high speed rotor test rig, running in 50mm diameter bearings of fixed and
variable geometry configuration, was used to obtain directly measured
temperature distributions within the rotor when running under varying speed
and unbalance response conditions.

Two separate rotor designs were used. The first is designed as a rigid rotor
where no structurally influenced rotor dynamic phenomena are present within
the running ranges. The second is a flexible rotor designed to operate in a
super critical condition where conditions are replicated to provide a rotor that
is sensitive to the variables required for thermal bow development and the
rotordyanmic conditions promotional of associated instabilities. Existing
theoretical models, in combination with operationally observed characteristics,
were used to develop a design predicted to become unstable within the test
running range.

Journal temperature measurements were obtained for rigid and flexible rotors
over a range of speed conditions in intentionally introduced mechanical
unbalance conditions. Journal temperature differentials were obtained with a
clear correlation between journal orbit size and journal temperature
differential. The flexible rotor was operated for prolonged periods of time in
the predicted unstable region but instability was never initiated for any test
condition. Peak journal temperature differentials were measured as 1.7 oC.

A new analytical model for the bearing oil film and journal thermal
developments is presented which has reasonable correlation to other
published literature.
Dedication

For Dad
Acknowledgements

I am most grateful to my supervisor, Professor David Mba for all his advice
and support over the duration of the work. I am most thankful for the
commitment he has shown to helping me reach the point of completion of the
work, as without his help the project would never have been accomplished.

I am profoundly grateful to the large number of people who have helped

throughout the duration of the work, right through from the test rig
manufacture to the writing up of the thesis. Whilst the list is too long to
mention everybody I would like to specifically thank the following for their
extra special help and support. Bob Orgy, Nick Glover and Scott Morgan for
all their help and advice with the test rig setup and instrumentation. Lee
Partel, Carmine Pette and Martin Powers for their advice and assistance in
fabricating various stands, brackets and widgets for the test rig at short notice.
Particular thanks go to Colin Sims and Gordon Sims for their dedication in
helping me on the manufacture on many of the test rig components, without
their help the testing would not have been possible. I am also grateful to
David Wilson for taking time out of his weekends and evenings to open up the
factory, come rain or shine, to allow me to run the test equipment.

I would also like to thank the Whitworth Awards society for their support, both
financially and as a society. Without the funding from the Whitworth Awards
panel then the test rig construction would not have been possible.

I am extremely grateful to Allen Gears for providing the test facilities and data
acquisition equipment to allow the test rig to be operated over the duration of
the research. My utmost appreciation also goes to Professor Keogh for all of
his input, support and advice on the theretical modelling.

Finally I would like to thank those people who without I could not have
completed this research. Thanks go to all my friends for all their support and
understanding over the last few years. In particular I would like to thank Alan
Williams and Mark Saunders for giving up their time to help proof the thesis
and analytical work. To my family, I am eternally grateful for all their patience
and unfaltering faith in me, and in particular to my loving wife and new born
daughter without whose support I could not have seen the project through.
Contents

Contents

1. Introduction ................................................................................................... 1
2. Literature Review .......................................................................................... 6
2.1 Rotor Rubs (Newkirk Effect) ................................................................... 6
2.2 Spiral Vibrations ..................................................................................... 9
2.3 Rotor Differential Heating ..................................................................... 13
2.4 Morton Effect ........................................................................................ 15
2.5 Hydrodynamic Bearing Theory ............................................................. 31
3. Analytical Review of Bearing Journal Differential Heating .......................... 32
3.1 Introduction........................................................................................... 32
3.2 Hydrodynamic Lubrication in Bearings ................................................ 33
3.2.1 Brief Overview of Hydrodynamic Lubrication ................................. 33
3.2.2 Equations Governing Hydrodynamic Lubrication in Journal
Bearings ................................................................................................. 34
3.2.3 Solution of the Reynolds Equations for Journal Bearings .............. 35
3.3 Keogh et al [9] Procedure ..................................................................... 37
3.3.1 Lubricant Flow Terms .................................................................... 38
3.3.2 Lubricant Energy Terms................................................................. 39
3.3.3 Lubricant Film Thickness ............................................................... 41
3.3.4 Lubricant Temperature Oscillations ............................................... 41
3.3.5 Bearing and Journal Conduction.................................................... 43
3.3.6 Lubricant Temperature Development ............................................ 44
3.3.7 Prediction of Rotor Bend ................................................................ 48
3.3.8 Bend Influence on Rotor Dynamic Systems................................... 50
3.4 Balbahadur [19] Procedure................................................................... 52
3.4.1 Relationships for film thickness...................................................... 52
3.4.2 Journal Position and Perturbation (Solution of Reynolds
equation) ................................................................................................. 59
3.4.3 Energy Equations and Journal Temperature ................................. 64
3.4.4 Thermal Unbalance and Stability Evaluation ................................. 68
4. Benchmark Examples ................................................................................. 72
4.1 Introduction........................................................................................... 72
4.2 Benchmark example cases .................................................................. 72
4.2.1 Rotordynamic principles applied in XLRotor [51] ........................... 74
4.2.2 Case Study 1 – deJong and Morton [11]........................................ 76
4.2.3 Case Study 2 – Faulkner, Strong and Kirk [12,13] ......................... 88
4.2.4 Case Study 3 – High Speed Overhung Compressor
(HSOC) ................................................................................................... 94
4.3 Discussion .......................................................................................... 104
Contents

5. Design of Experiment - Experimental Test Rig ......................................... 109

5.1 Introduction......................................................................................... 109
5.2 Experimental Rig Design Requirements ............................................. 110
5.3 Measurement, Instrumentation and Control ....................................... 111
5.4 Experimental Test Rotor and Bearing Design .................................... 113
5.4.1 Bearing Design and Selection...................................................... 113
5.4.2 Rotor Dynamic Analysis ............................................................... 120
5.4.3 Synchronous Thermal Instability (Morton Effect) Analysis ........... 143
5.5 Detail Design ...................................................................................... 150
5.5.1 Main Test Rotor and Bearings ..................................................... 150
5.5.2 Drive System ............................................................................... 152
5.5.3 Transmission ............................................................................... 152
5.5.4 Lubrication System ...................................................................... 154
5.5.5 Instrumentation ............................................................................ 155
5.5.6 Monitoring, Data Logging and Data Management ....................... 169
6. Experimental Rig Commissioning and Test Procedure ............................. 173
6.1 Introduction......................................................................................... 173
6.2 Test Rig Preparation and Assembly ................................................... 173
6.2.1 Rotor Balancing ........................................................................... 177
6.2.2 Test Rig Alignment ...................................................................... 178
6.2.3 Mechanical – Electrical Runout Measurement ............................. 179
6.2.4 Bearing Geometry Measurement ................................................. 179
6.2.5 Lubrication System ...................................................................... 182
6.3 Instrumentation Calibration and Validation ......................................... 183
6.3.1 Proximity Probes .......................................................................... 183
6.3.2 Shaft Temperature Sensors ......................................................... 184
6.3.3 General Sensor Calibration.......................................................... 186
6.4 Running Commissioning .................................................................... 186
6.4.1 Bearings....................................................................................... 187
6.4.2 Shaft Journal Temperature Sensors ............................................ 188
6.4.3 Rotor Vibration Response ............................................................ 189
6.5 Rotor Commissioning Summary ......................................................... 191
6.6 Test Procedure ................................................................................... 191
7. Experimental Results ................................................................................ 194
7.1 Introduction......................................................................................... 194
7.2 Data Conditioning and Post Processing ............................................. 194
7.2.1 Shaft Vibration and Position......................................................... 194
7.2.2 Bearing Temperatures and Lube Oil/Air Temperatures ............... 206
7.2.3 Shaft Temperatures ..................................................................... 206
7.3 Test Results ....................................................................................... 208
Contents

7.3.1 Shaft Vibrations ........................................................................... 209

7.3.2 Shaft Temperatures ..................................................................... 218
8. Analytical Evaluation and Development of Journal Temperature
Distribution .................................................................................................... 236
8.1 Introduction......................................................................................... 236
8.2 Finite Difference Heat Transfer Model................................................ 237
8.2.1 The Heat Transfer Model ............................................................. 238
8.2.2 The finite Difference Technique ................................................... 239
8.3 Bearing Oil Film Temperature Distribution Model ............................... 248
8.3.1 The Oil Film Function ................................................................... 249
8.3.2. Bearing Temperature Distribution Model .................................... 260
8.4 Cavatitive Region ............................................................................... 269
8.5 Oil Inlet boundary conditions .............................................................. 272
8.5.1 Side leakage ................................................................................ 272
8.5.2 Re-circulation flow ....................................................................... 273
8.5.3 Velocity flow ................................................................................. 273
8.5.4 Pressure Flow .............................................................................. 274
8.6 Lubricant Thermal Perturbations ........................................................ 276
8.7 Shaft Bearing Journal Temperature Distribution Model ...................... 285
8.7.1 Finite Difference Grid ................................................................... 285
8.7.2 Initialisation and Boundary Conditions ......................................... 286
8.7.3 Time Domain Solution.................................................................. 286
9. Analysis of Results and Discussion .......................................................... 289
9.1 Introduction......................................................................................... 289
9.2 Rotordynamic Performance ................................................................ 289
9.3 Shaft Journal Temperature Measurements ........................................ 291
9.4 Comments on Existing Theoretical Models ........................................ 315
9.5 Shaft Journal Temperature Theoretical Model Results ...................... 318
9.5.1 Oil Film Energy Model and Temperature Equation ...................... 318
9.5.2 Oil Film Temperature Perturbations ............................................. 321
10. Conclusions ............................................................................................ 324
11. Further Work ........................................................................................... 330
References ................................................................................................... 333
Bibliography .................................................................................................. 339
Appendix A ................................................................................................... 340
Appendix B ................................................................................................... 344
Appendix C ................................................................................................... 349
List of Figures

List of Figures

Figure 1 Diagram of Newkirk Effect Above and Below 1st Critical

Speed [19] ........................................................................................................ 7
Figure 2 Diagrammatic Representation of the Backward Bending of
Shaft Running Above Critical Speed [1] ............................................................ 8
Figure 3 Vibration for a Rotor Displaying the Newkirk Effect [2] ..................... 10
Figure 5 Typical Experimental Results Obtained by Kroon and
Williams [2]. Solid Lines are True Exponential Curves, Datapoints
were Obtained by Test [2] ............................................................................... 11
Figure 6 Fundamental Component (1x) of Relative Shaft Vibration [11] ......... 17
Figure 7 Typical Temperature Distribution with Heat Barrier Sleeve
[14] .................................................................................................................. 19
Figure 8 4150 RPM Constant Speed Factory Test Proximity Probe 1x
Vibration Amplitude vs Time [37] .................................................................... 28
Figure 9 Principle of Hydrodynamic Lubrication Between Two Inclined
Surfaces – figure taken from [44] .................................................................... 34
Figure 11 Plain Cylindrical Bearing Film Thickness Function ......................... 53
Figure 12 Tilting Pad Bearing Geometric Relationships.................................. 54
Figure 13 Tilting Pad Bearing Film Thickness Function Comparison 5
Pad, Rj=25 mm, Cd=0.125 mm, Pre-load=0.375, Load angle = 90o ................ 58
Figure 14 Geometric Relationships of Elliptical Synchronous Orbit ................ 60
Figure 19 Rotordynamic Model of de Jongh et al [11] Rotor ........................... 77
Figure 20 Calculated Critical Speed Map of de Jongh et al [11] Rotor
Including Comparative Data Presented within the Literature .......................... 78
Figure 23 Calculated Unbalance Response of de Jongh et al [11]
Rotor - using approximate stiffness coefficients from [11]............................... 79
Figure 24 Curve fitting of calculated support bearing stiffness and
damping coefficients ....................................................................................... 81
Figure 25 Calculated forward whirl damped natural frequencies for
de Jongh et al [11] rotor .................................................................................. 82
Figure 26 Calculated damped mode shape for first critical speed for de
Jongh et al [11] rotor ....................................................................................... 83
Figure 27 Calculated damped mode shape for second critical speed
for de Jongh et al [11] rotor ............................................................................. 83
Figure 28 Calculated unbalance response for the de Jongh et al [11]
rotor with 100 g.mm unbalance at the NDE overhung mass ........................... 84
Figure 30 Calculated rotor deflection shape of de Jongh et al [11] rotor
at location of 1st critical speed ........................................................................ 84
List of Figures

Figure 31 Calculated rotor deflection shape of de Jongh et al [11] rotor

at location of 2nd critical speed of 12200 rpm ................................................. 85
Figure 32 Calculated rotor deflection shape of de Jongh et al [11] rotor
at observed thermal stability threshold speed of 11500 rpm ........................... 85
Figure 35 Calculated unbalance response of de Jongh et al [11] rotor
where centre only imbalance introduced......................................................... 88
Figure 36 Undamped critical speed map for Turbocharger rotor
presented by Faulkner, Strong and Kirk [12,13] – figure taken from
[12,13] ............................................................................................................. 89
Figure 37 Calculated undamped critical speed map of Faulkner et al
[12,13] rotor .................................................................................................... 89
Figure 38 Rotordynamic model of Faulkner et al [12,13] rotor ........................ 90
Figure 40 Calculated unbalance response for Faulkner et al [12,13]
rotor compressor bearing ................................................................................ 92
Figure 41 Calculated unbalance response for Faulkner et al [12,13]
rotor turbine bearing........................................................................................ 92
Figure 42 Calculated unbalance response for Faulkner et al [12,13]
rotor turbine disc ............................................................................................. 93
Figure 43 Predicted rotor response for turbocharger rotor presented by
Faulkner, Strong and Kirk [12,13] ................................................................... 93
Figure 44 Calculated operating rotor response for Faulkner et al
[12,13] rotor .................................................................................................... 93
Figure 45 Calculated operating rotor response for Faulkner et al
[12,13] rotor .................................................................................................... 94
Figure 49 Calculated undmaped critical speed map for HSOC
compressor rotor ............................................................................................. 97
Figure 50 Calculated damped natural frequency map for HSOC
compressor rotor ............................................................................................. 98
Figure 52 Calculated response for NDE bearing location of HSOC
compressor rotor (normal viscosity) .............................................................. 100
Figure 53 Calculated response for DE bearing location of HSOC
compressor rotor (normal viscosity) .............................................................. 101
Figure 54 Calculated response for NDE bearing location of HSOC
compressor rotor (higher viscosity) ............................................................... 101
Figure 55 Calculated response for DE bearing location of HSOC
compressor rotor (higher viscosity) ............................................................... 101
Figure 56 Calculated operating deflection shape for HSOC
compressor rotor at endurance running speed of 11695 rpm ....................... 102
Figure 57 Calculated operating deflection shape for HSOC
compressor rotor at maximum running speed of 13400 rpm......................... 102
List of Figures

Figure 58 Calculated operating deflection shape for HSOC

compressor rotor at endurance running speed with low bearing pad
inlet viscosity ................................................................................................. 103
Figure 64 Un-damped Critical Speed Map for 750mm Bearing Span
With Added Mass at Shaft Centre ................................................................. 123
Figure 65 Rotordynamic Model of Flexible Rotor – with Added Mass ........... 125
Figure 66 Undamped critical speed map for flexible rotor without
added mass .................................................................................................. 126
Figure 67 Damped natural frequency map for rotor in fixed geometry
bearings, mean clearance, normal inlet temperature with no added
mass ............................................................................................................. 126
Figure 69 Damped natural frequency map for final rotor design with
fixed geometry bearings, mean clearance and 3.5 kg added overhung
mass ............................................................................................................. 129
Figure 70 Damped mode shape for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at 1107 rpm critical (bearing rock mode) with log decrement of 3.743 .......... 130
Figure 71 Damped mode shape for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at 1120 rpm critical (bearing rock mode) with log decrement of 3.646 .......... 130
Figure 72 Damped mode shape for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at 6111 rpm critical with log decrement of 0.214 ........................................... 131
Figure 74 Damped response plot for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at NDE bearing ............................................................................................. 132
Figure 75 Damped response plot for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at NDE flange ............................................................................................... 132
Figure 76 Damped response plot for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at centre mass .............................................................................................. 133
Figure 77 Damped response plot for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass
at DE bearing ................................................................................................ 133
Figure 78 Damped response operating deflection shape and NDE
bearing whirl orbit for final rotor design with fixed geometry bearings,
mean clearance and 3.5 kg added overhung mass at DE bearing at
6200 rpm. ...................................................................................................... 134
List of Figures

Figure 79 Damped response operating deflection shape and NDE

bearing whirl orbit for final rotor design with fixed geometry bearings,
mean clearance and 3.5 kg added overhung mass at DE bearing at
10000 rpm. .................................................................................................... 135
Figure 80 Damped response operating deflection shape and NDE
bearing whirl orbit for final rotor design with fixed geometry bearings,
mean clearance and 3.5 kg added overhung mass at DE bearing at
12600 rpm. .................................................................................................... 135
Figure 81 Damped natural frequency map for final rotor design with
variable geometry bearings, mean clearance and 3.5 kg added
overhung mass ............................................................................................. 136
Figure 82 Damped mode shape for final rotor design with variable
geometry bearings, mean clearance and 3.5 kg added overhung mass
at 5972 rpm critical with log dec of 0.139 ...................................................... 136
Figure 83 Damped mode shape for final rotor design with variable
geometry bearings, mean clearance and 3.5 kg added overhung mass
at 8940 rpm critical with log dec of 1.044 ...................................................... 137
Figure 84 Damped response plot for final rotor design with variable
geometry bearings, mean clearance and 3.5 kg added overhung mass
at NDE bearing ............................................................................................. 137
Figure 85 Damped response operating deflection shape and NDE
bearing whirl orbit for final rotor design with variable geometry
bearings, mean clearance and 3.5 kg added overhung mass at NDE
bearing at 9600 rpm. ..................................................................................... 138
Figure 86 Rigid rotor dynamic model ............................................................ 140
Figure 87 Rigid rotor undamped critical speed map...................................... 141
Figure 88 Damped natural frequency map for rigid rotor design with
fixed geometry bearings, mean clearance .................................................... 141
Figure 89 Damped mode shapes for rigid rotor design with fixed
geometry bearings and mean clearance (a) ‘rocking’ mode at 1104
rpm with log dec of 3.762 (b) ‘bounce’ mode at 1120 rpm with log dec
of 3.647 ......................................................................................................... 141
Figure 90 Damped response plot for rigid rotor design with fixed
geometry bearings and mean clearance 12.7 g.mm at NDE flange .............. 142
Figure 91 Damped response operating deflection shape and NDE
bearing whirl orbit for rigid rotor design with variable geometry
bearings and mean clearance at 12500 rpm................................................. 142
Figure 92 Shaft orbit showing hot and cold spots on rotor journal for
flexible rotor in fixed geometry bearings ....................................................... 145
List of Figures

Figure 93 Morton stability analysis of fixed geometry bearings with 20

g.mm unbalance at overhung flange, 3.5 kg added mass and mean
bearing clearance ......................................................................................... 146
Figure 94 Morton stability analysis of fixed geometry bearings with 50
g.mm unbalance at overhung flange, 3.5 kg added mass and mean
bearing clearance ......................................................................................... 146
Figure 95 Journal temperature differential fixed geometry bearing at
NDE journal with 50 g.mm unbalance, 3.5 kg added mass and mean
bearing clearance ......................................................................................... 147
Figure 96 Morton stability analysis of variable geometry bearings with
20 g.mm unbalance at overhung flange, 3.5 kg added mass and
maximum bearing clearance ......................................................................... 147
Figure 97 Journal temperature differential variable geometry bearing at
NDE journal with 20 g.mm unbalance, 3.5 kg added mass and
maximum bearing clearance ......................................................................... 148
Figure 98 Morton stability analysis of variable geometry bearings with
20 g.mm unbalance at overhung flange, 3.5 kg added mass and mean
bearing clearance ......................................................................................... 148
Figure 99 Journal temperature differential variable geometry bearing at
NDE journal with 20 g.mm unbalance, 3.5 kg added mass and mean
bearing clearance ......................................................................................... 149
Figure 100 Main test (flexible) rotor assembly schematic ............................. 152
Figure 101 Rigid test rotor assembly schematic ........................................... 152
Figure 102 Test rotor and transmission assembly ........................................ 153
Figure 103 Test rotor shaft proximity probe locations ................................... 155
Figure 104 Push-Pull Test Bearing Proximity Probe Arrangement ............... 156
Figure 107 Thermistor sensor assembly ....................................................... 158
Figure 110 As supplied thermistor accuracy due to resistance variance ...... 160
Figure 111 Typical thermistor temperature-resistance curve ........................ 162
Figure 112 Heat sink block used for thermistor sensor calibration ................ 163
Figure 113 Time-temperature curve for thermistor calibration ...................... 164
Figure 114 Thermistor sensor error measurement (Green sensor) .............. 165
Figure 115 Thermistor sensor error measurement (Blue sensor) ................. 166
Figure 116 Thermistor sensor error measurement (Red sensor) .................. 166
Figure 117 Thermistor sensor error measurement (White sensor) ............... 166
Figure 118 Original slip ring coupling signal wire failure ............................... 168
Figure 119 Schematic of final slip ring instrumentation design ..................... 168
Figure 120 Final slip ring instrumentation design .......................................... 168
Figure 121 ‘Potting’ of signal wire soldered joints to prevent fatigue
failures .......................................................................................................... 169
List of Figures

Figure 122 Pipework and instumentation layout for test rig .......................... 171
Figure 123 Test Rig General Arrangement ................................................... 174
Figure 124 Test Rig Basic Equipment (1) ..................................................... 174
Figure 125 Test Rig Basic Equipment (2) ..................................................... 175
Figure 126 Test rig with main rotor housing and guards fitted ...................... 176
Figure 127 Test rig with main flexible rotor ................................................... 176
Figure 128 Test rig transmission to test rotor alignment readings ................ 178
Figure 129 Electrical Mechanical Runout for Rigid Rotor.............................. 179
Figure 130 Offset Half bearing geometric measurement relationships ......... 180
Figure 131 Shaft proximity probe calibration................................................. 184
Figure 132 Measured variation in temperature measurement from
shaft thermistors during slow roll gradual oil temperature increase .............. 185
Figure 133 Slow roll glitch for rigid rotor NDE ............................................... 190
Figure 134 Synchronous (1X) vibration amplitude and phase response
(Bode Plot) of rigid rotor in fixed geometry bearings ..................................... 190
Figure 136 Coordinate system for clocking angle of radial proximity
probes ........................................................................................................... 203
Figure 137 NDE Bearing filtered shaft vibration DR0301 .............................. 209
Figure 138 NDE Bearing filtered and unfiltered shaft vibration DR0321 ....... 210
Figure 139 NDE 1X filtered orbit 8000 rpm DR0321 ..................................... 210
Figure 140 Harmonic Analysis of NDE Vibration for DR0301 ....................... 211
Figure 141 NDE Bearing filtered shaft vibration DR0401 – Run-up Test ...... 212
Figure 142 NDE Bearing Filtered Vibration DR0407 – Run-up Test ............. 213
Figure 143 Synchronous (1X) flexible rotor shaft orbits at 6250 rpm -
DR0407......................................................................................................... 213
Figure 144 Synchronous (1X) flexible rotor shaft orbits at 9000 rpm -
DR0407......................................................................................................... 213
Figure 145 NDE Bearing Filtered Vibration DR0428 – Run-up Test ............. 214
Figure 146 NDE Synchronous (1X) flexible rotor shaft orbits - DR0428 ....... 214
Figure 147 NDE Bearing Filtered Vibration DR0453 – Run-up Test ............. 215
Figure 148 NDE Synchronous (1X) flexible rotor shaft orbits - DR0453 ....... 215
Figure 149 NDE Bearing Filtered Vibration DR0651 – Run-up Test ............. 216
Figure 150 NDE Bearing Filtered Vibration DR0673 – Run-up Test ............. 217
Figure 151 NDE Synchronous (1X) flexible rotor shaft orbits - DR0673 ....... 217
Figure 152 NDE Test Bearing Temperature Measurements
Throughout Test DR0301 ............................................................................. 218
Figure 153 NDE circumferential bearing shell temperature
measurements DR0301 ................................................................................ 219
Figure 154 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0301......................................................................................................... 219
List of Figures

Figure 155 Shaft journal temperature differential measurements for

DR0301......................................................................................................... 220
Figure 156 Journal Circumferential Temperatures for DR0301 .................... 220
Figure 157 NDE Test Bearing Temperature Measurements
Throughout Test DR0321 ............................................................................. 221
Figure 158 Shaft journal temperature differential measurements for
DR0321......................................................................................................... 221
Figure 159 Journal Circumferential Temperatures for DR0321 .................... 222
Figure 160 NDE Test Bearing Temperature Measurements
Throughout Test DR0321 ............................................................................. 222
Figure 161 Shaft journal temperature differential measurements for
DR0399......................................................................................................... 223
Figure 162 Journal Circumferential Temperatures for DR0399 .................... 223
Figure 163 NDE Test Bearing Temperature Measurements
Throughout Test DR0321 ............................................................................. 224
Figure 164 NDE circumferential bearing shell temperature
measurements DR0401 ................................................................................ 224
Figure 165 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0401......................................................................................................... 224
Figure 166 Shaft journal temperature differential measurements for
DR0401......................................................................................................... 225
Figure 167 Journal Circumferential Temperatures for DR0401 .................... 225
Figure 168 Shaft journal temperature differential measurements for
DR0407......................................................................................................... 226
Figure 169 Journal Circumferential Temperatures for DR0401 ................... 226
Figure 170 NDE Test Bearing Temperature Measurements
Throughout Test DR0427 ............................................................................. 227
Figure 171 NDE circumferential bearing shell temperature
measurements DR0427 ................................................................................ 227
Figure 172 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0427......................................................................................................... 227
Figure 173 Shaft journal temperature differential measurements for
DR0427......................................................................................................... 228
Figure 174 Journal Circumferential Temperatures for DR0427 .................... 228
Figure 175 Shaft journal temperature differential measurements for
DR0428......................................................................................................... 229
Figure 176 Journal Circumferential Temperatures for DR0428 .................... 229
Figure 177 NDE Test Bearing Temperature Measurements
Throughout Test DR0453 ............................................................................. 230
List of Figures

Figure 178 NDE circumferential bearing shell temperature

measurements DR0453 ................................................................................ 230
Figure 179 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0453......................................................................................................... 230
Figure 180 Shaft journal temperature differential measurements for
DR0453......................................................................................................... 231
Figure 181 Journal Circumferential Temperatures for DR0453 .................... 231
Figure 182 NDE Test Bearing Temperature Measurements
Throughout Test DR0651 ............................................................................. 232
Figure 183 NDE bearing pad temperature measurements DR0651 ............. 232
Figure 184 Shaft Journal and NDE Bearing Temperatures for DR0651 ....... 232
Figure 185 Shaft journal temperature differential measurements for
DR0651......................................................................................................... 233
Figure 186 Journal Circumferential Temperatures for DR0651 .................... 233
Figure 187 NDE Test Bearing Temperature Measurements
Throughout Test DR0673 ............................................................................. 234
Figure 188 NDE bearing pad temperature measurements DR0673 ............. 234
Figure 189 Shaft journal temperature differential measurements for
DR0673......................................................................................................... 235
Figure 190 Journal Circumferential Temperatures for DR0673 .................... 235
Figure 191 Discretisation of the analytical derivative solution [60] ................ 240
Figure 194 Small grid discreatisation effect around centre node in
polar coordinates .......................................................................................... 248
Figure 195 Basic coordinate system used in the bearing geometry .............. 250
Figure 197 Triangles from Figure 196 ........................................................... 251
Figure 199 Forward and backward whirl orbits from vibration ellipse............ 259
Figure 200 Fixed Geometry Offset Half Dynamic Orbit Relationships .......... 259
Figure 203 Cavitation region in plain bearing with oil streamers ................... 270
Figure 204 Axially averaged oil film temperature for a single oil inlet
bearing .......................................................................................................... 271
Figure 205 Oil flow schematic for a single inlet bearing ................................ 274
Figure 206 Journal Circumferential Temperature Distribution for
‘Steady State’ Assumption Boundary Conditions .......................................... 277
Figure 207Journal Circumferential Harmonic Temperature Distribution
for ‘Steady State’ Assumption Boundary Conditions ..................................... 278
Figure 208 Convective boundary layer concept with simplification of
flow modelling ............................................................................................... 280
∂p
Figure 209 Couette and Poiseuille Flow Due to Pressure Gradient ....... 281
∂θ
Figure 210 Plot of a converged heat lubricant-journal heat flux .................... 284
List of Figures

Figure 211 Structured 2D Polar Finite Difference Grid.................................. 286

Figure 212 Finite Difference Grid Converged Journal Temperature
Distribution 50mm diameter bearing, 5000 rpm, ε =0.3, σ 1 =0.2,
forward orbit, 2 oil inlets ................................................................................ 287
Figure 213 Typical Convergence Trend of Central Node Temperature ........ 288
Figure 214 Measured and Calculated Rotordynamic Response for
DR0428 (rotor residual mechanical unbalance removed) ............................. 290
Figure 215 Measured and Calculated Rotordynamic Response for
DR0673 (rotor residual mechanical unbalance removed) ............................. 290
Figure 216 Measured 1st Order Shaft Journal Temperature Differential
for Rigid Rotor ............................................................................................... 292
Figure 217 Log-Log Plot of DR03 Tests for 100, 200.................................... 293
Figure 218 Analytical Curves for Rigid Rotor Unbalance Loads for 100,
200 and 300 g.mm. ∆T = AT .N B ...................................................................... 294
T

Figure 219 1st Harmonic journal temperature differential and phase

angle with respect to introduced mechanical unbalance ............................... 294
Figure 220 1st Harmonic journal temperature differential against orbit
size ............................................................................................................... 295
Figure 221 Rigid Rotor 1st Harmonic Temperature Phase Lag with
Respect to Rotor Response at 8000 rpm ...................................................... 296
Figure 223 Peak-Peak rotor response with respect to whirl orbit ellipse
ratio ............................................................................................................... 297
Figure 224 1st Harmonic Temperature Differential Phase Angle for 300
g.mm Unbalance of Rigid Rotor .................................................................... 298
Figure 226 1st Harmonic Journal Temperature Differential for Flexible
Rotor with 1.5 kg OHM with Respect to Orbit Size at 9000 rpm ................... 301
Figure 227 1st Harmonic Journal Temperature Differential Phase Lag
for Flexible Rotor with 1.5 kg OHM with Respect to Orbit Size at 9000
rpm................................................................................................................ 301
Figure 228 1st Harmonic Journal Temperature Differential and Phase
Angle against Speed for a 100 g.mm Unbalance on Flexible Rotor with
1.5 kg OHM ................................................................................................... 302
Figure 229 1st Harmonic Journal Temperature Differential against
Speed for a 40 g.mm Unbalance on Flexible Rotor with 5.0 kg OHM ........... 303
Figure 230 Calculated Phase Angles for NDE Bearing Centre and
Measurement Point Locations Flexible Rotor Fixed Geometry Bearings
5kg OHM ....................................................................................................... 304
Figure 231 Measured and Calculated Phase Angles for NDE Bearing
Flexible Rotor Fixed Geometry Bearings 5.0 kg OHM – DR0428 ................. 305
List of Figures

Figure 232 1st Harmonic Journal Temperature Differential Phase Angle

for Flexible Rotor in Fixed Geometry Bearings – Corrected by
Calculated Rotordynamic Phase Shift........................................................... 305
Figure 233 1st Harmonic Journal Temperature Differential for Flexible ......... 306
Figure 234 1st Harmonic Journal Temperature Differential Phase Angle
for Flexible Rotor in Variable Geometry Bearings with 5.0 kg OHM.............. 307
Figure 235 Shaft Whirl Orbit for the Flexible Rotor in Variable
Geometry Bearings with 5 kg OHM............................................................... 308
Figure 236 1st Harmonic Journal Differential Temperature with
Respect to Response Modulus for Different Arrangements and
Conditions ..................................................................................................... 308
Figure 238 1st Harmonic Journal Temperature Differential Running in
Fixed Geometry Bearings with 150 g.mm of Imbalance Located at the
Rotor Mid-span ............................................................................................. 310
Figure 239 1st Harmonic Journal Temperature Differential Phase
Running in Fixed Geometry Bearings with 150 g.mm of Imbalance
Located at the ............................................................................................... 310
Figure 242 Measured and Calculated [19] Journal Temperature
Differential Flexible Rotor in Fixed Geometry Bearings with 40 g.mm
Unbalance ..................................................................................................... 313
Figure 243 Measured and Calculated [19] Journal Temperature
Differential Flexible Rotor in Variable Geometry Bearings with 40
g.mm Unbalance ........................................................................................... 314
Figure 244 Keogh et al [9] Bearing Case ...................................................... 319
Figure 245 Circumferential Temperature Distribution for Steady
Eccentricity Comparison Between Literature and Research ......................... 320
Figure 246 Calculated Journal Temperature Distribution with a
Forward Whirl Orbit ....................................................................................... 321
Figure 247 Journal Temperature Distribution Harmonic Analysis ................. 322
Figure 248 Calculated Journal Temperature Distribution with a ................... 322
Figure 249 Journal Temperature Distribution Harmonic Analysis ................. 323
List of Tables

List of Tables

Table 1 Breakdown of Various Rotor Types Identified as Experiencing

the Morton Effect [33]...................................................................................... 24
Table 2 Calculated support bearing basic performance data .......................... 80
Table 3 Calculated support bearing oil film stiffness and damping
coefficients ...................................................................................................... 81
Table 4 Bearing oil film coefficients across speed range for NDE
bearing ............................................................................................................ 98
Table 5 Bearing oil film coefficients across speed range for DE bearing ........ 98
Table 6 Fixed Geometry Bearing Types ....................................................... 115
Table 7 Fixed Geometry Bearing Designs .................................................... 116
Table 9 Summary of Variable Geometry Bearing Parameters Evaluated ..... 119
Table 10 Natural Frequency of Plain Rotors of Varying Bearing Span ......... 123
Table 11 Summary of Test Rotor Critical Speeds and Mode Shapes ........... 127
Table 12 Fixed geometry bearing basic design specification ........................ 128
Table 13 Fixed geometry bearing oil film stiffness and damping
coefficients for mean clearance and normal oil inlet temperature ................. 128
Table 14 Variable geometry bearing basic design specification ................... 129
Table 15 Variable geometry bearing oil film stiffness and damping
coefficients for mean clearance and normal oil inlet temperature ................. 129
Table 16 Lubrication Oil Summary ................................................................ 154
Table 17 Thermistor sensors Steinhart-Hart constants ................................. 164
Table 18 Individual sensor calibrations/accuracy for flexible and rigid
rotor .............................................................................................................. 167
Table 19 Test rig sensor summary................................................................ 169
Table 20 Data acquisition rates for all mesasurement equipment and
channels ....................................................................................................... 171
Table 21 Fixed Geometry Offset Half Bearing Manufactured
Measurements .............................................................................................. 181
Table 22 Specific test activity streams .......................................................... 192
Table 23 Rigid rotor test programme DR03 – fixed geometry bearings ........ 192
Table 24 Flexible rotor test programme DR04 – fixed geometry bearings .... 193
Table 25 Flexible rotor test programme DR06 – variable geometry
bearings ........................................................................................................ 193
Table 26 Results summary of vibration data for DR03 test programme........ 211
Nomenclature

Nomenclature

Chapter 3 – Section 3.3

BX, BY Coefficients relating rotor bend to vibration amplitudes

c Bearing radial clearance
cB, cL, cJ Bearing, lubricant, journal material specific heat capacities
d Fractional film width
D Bearing diameter
G Gain parameter for bend development
gmn, hmn Journal bearing thermal coefficients
h Dimensional film thickness
hB, hJ Bearing, journal heat transfer coefficients
H Dimensionless film thickness, h/c
I Journal section second moment of area
kB, kL, kJ Bearing, lubricant, journal material thermal conductivities
KT Rotor stiffness parameter
lb, lD Rotor length parameters
L, LJ Bearing, journal section lengths
mb, md Rotor mass parameters
NB, NJ Dimensionless bearing, journal heat transfer coefficients,
RhB/kB, RhJ/kJ
p Lubricant dimensional pressure
P Lubricant dimensionless pressure, p(c/R)2/6
Pe Peclet number,
r Dimensionless radial coordinate
ro Dimensionless outer bearing radius, Ro/R
R Inner bearing radius
Ro Outer bearing radius
t Time
T Dimensionless temperature
To, T1, T2 Dimensionless zero orbit and orbit perturbed lubricant
temperatures
TB, TJ, TL Dimensionless axially averaged bearing, journal, lubricant
temperatures
TA, TS Dimensionless ambient, lubricant supply temperatures
u, v, w Flow velocity components
U, V, W Dimensionless flow velocity components
Wb Bearing steady load
zXb, zYb Complex vibration amplitudes at bearing location
α Journal material thermal expansion coefficient
β Circumferential angular coordinate referred to fixed axes
βc Circumferential cavitation angular coordinate
γ Orbit orientation parameter
γ B ,γ J Dimensionless parameters, kB c/kLR, kJc/kLR
∆ Lubricant inlet temperature parameter
∈o Zero orbit eccentricity ratio
Nomenclature

∈1 ,∈2 Backward, forward whirl circular orbit dimensionless radii

∈L Bearing parameter, (L/D)2
η Dimensionless cross-film coordinate, y/cH with 0<y<h
θ Circumferential coordinate referred to journal rotating axes,
Θ Dimensional temperature
Θf Reference temperature
κB ,κ J Bearing, journal diffusivities
µ Average lubricant viscosity
ξ Dimensionless axial coordinate, 2z/L
ξJ Journal section dimensionless axial length, LJ/L
ρ Radial coordinate
ρB , ρL , ρJ Bearing, lubricant, journal material densities
τ Dimensionless time
φ0 Zero orbit attitude angle
ψ x ,ψ y Rotor bend angles referred to rotating axes
ψ C ,ψ Ci ,ψ Ct Complex rotor bend angles

Chapter 3 – Section 3.4

Ax , Ay Amplitude parameters for elliptic orbit

cl Lubricant specific heat capacity
Cb radial bearing clearance
Cp radial pad clearance
e eccentricity
E Young’s Modulus
Eb Rate of energy transfer to bearing
Ej Rate of energy transfer to journal
Elub Rate of energy storage in lubricant
Evisc Rate of visc. energy dissipation
fp Fractional angular position of pad pivot
h Film thickness
H Heat transfer coefficient
I Area moment of inertia
L Bearing length
Ld Disk overhang length
m Preload factor
md Mass of overhung disk
Npads Number of pads
Ob Bearing center
Oj Journal center
Op Pad center of curvature
P Hydrodynamic pressure within the bearing
Po Lubricant supply pressure
Rb Bearing radius
Nomenclature

Rj Journal radius
Rp Inner radius of curvature of pad
Re Reynolds’ number
Rec Critical Reynolds’ number (Re > Rec => turbulent)
tp Pad thickness
T Journal circumferential temperature
T0 Lubricant supply temperature
Tamb Ambient temperature
uj Surface velocity of journal
U Resultant unbalance
Um Mechanical unbalance
Ut Thermal unbalance
Uthr Threshold unbalance
W Rotor weight
Wb Bearing load
x, y, z Local coordinate system with origin on journal surface
X, Y, Z Fixe coordinate system with origin at bearing center
yd Thermal deflection of disk
z Axial dimension
β Thermoviscosity coefficient
δ Pad tilt angle
∆p Angular dimension of pad
ε Eccentricity ratio
θ Circumferential angle
θc Angle for line-of-centers
θj Angle to journal center
θp Pivot angle for pad
µ Lubricant viscosity
µ0 Lubricant supply viscosity
ζ Angle used in PJB film thickness expression
ρl Lubricant density
τ Lubricant shear stress
φ Angle between thermal and mechanical unbalances
φx, φy Phase parameters for elliptical orbit
ψb Thermal bend angle
ξ Attitude angle
ω Angular journal speed

Chapter 5

C A , CB , CC Steinhart-Hart thermistor sensor constants

Cd Diametral clearance
Cr Radial clearance
pR Bearing pad pre-load
Ri Fixed resistor resistance
Nomenclature

Rtherm Thermistor sensor resistance

T Temperature
VG Voltage across gauge
VS Supply voltage

Chapter 6

χr Bearing pad offset

φm Bearing pad cant angle

Chapter 7

Ao , Am , Ag Vibration amplitudes for glitch, measured and glitch

extracted
k Harmonic order
N Number of data points
t Time
tK Non harmonic residual of shaft journal temperature
measurement
tn Temperature measurement of nth sensor
Tk Measured journal 1st harmonic temperature amplitude
xp , y p X and Y probe gap measurement
φTk Measured journal1st harmonic temperature phase angle
φo , φm , φg Vibration phase angles for glitch, measured and glitch
extracted
ω Shaft rotation speed

Chapter 8

a1,2 Circular whirl orbit radii (1= forward, 2=backward)

c Specific heat capacity
cL Specific heat capacity lubricant
cr Radial clearance
e Journal eccentricity
e′ Journal effective dynamic eccentricity
E& Energy transport
fL Oil film convective flow fraction
FE Bearing power loss coefficient
h Oil film thickness / heat transfer coefficient
H Non dimensional oil film thickness
Ih Fractional film coefficient
Nomenclature

k Thermal conductivity
kL Lubricant thermal conductivity
Linlet Oil inlet grove length
Lb Length of bearing
pinlet Pressure at oil feed groove
q& Heat flux
q& j Heat flux from lubricant to journal
Q& Heat flow rate
Qv , Q p , Qr Lubricant flow rates, velocity, pressure, recirculation
r Radial point
Rs Journal radii
Rb Bearing radii
Tm Finite difference node temperature
T∞
TL Lubricant temperature
To1,2 Bearing pad lubricant inlet temperature
U Journal velocity
α Thermal diffusivity
β angle from oil inlet to maximum film thickness
ε Journal eccentricity ratio
ε′ Journal effective dynamic eccentricity ratio
φ Angle around bearing
φ∆T Journal 1st harmonic temperature differential phase angle
γ′ Angle between rotation vector and minimum film thickness
µ Lubricant viscosity
θ Angle around journal
θ′ Attitude angle of effective eccentricity
ρ Density
σ 1,2 Non dimensional orbit radii (1= forward, 2=backward)
τ Lubricant shear stress
∆T Journal 1st harmonic temperature differential
Nomenclature
Chapter 1 - Introduction

Chapter 1

Introduction

1. Introduction

Rotating machinery has been used as a means of providing labour saving

devices to man for many hundreds of years. It is however in the last century,
following the industrial revolution, that the most significant developments in
rotating machinery have taken place and with it the development of their
science.

The development of the theory behind the dynamics of rotating machines has
been driven forward by the ever increasing power demands required for the
various types of rotating machinery, such as steam turbines, gas turbines,
electrical generators, pumps and compressors to name but a few. The
requirement for modern machinery to be capable of delivering, absorbing and
transmitting multi megawatt powers has driven the operating speeds of the
rotors to higher and higher levels to enable practical designs to be realised.
With these high operating speeds come a number of design challenges in that
the physical limitations of the rotor materials restrict the diameters of the
machine because of centrifugal loading and rotors need to become longer and
consequently more flexible. This results in many of today’s pieces of turbo-
machinery operating above their first, second and sometimes third natural
frequencies, defining the need for accurate and reliable methods for predicting
the characteristics, response and stability of such super critical machines.

It is common in these rotating machines to experience vibration related

problems, the causes of which can be diverse in nature. The most common
cause of vibration in rotating machinery is synchronous to the rotor shaft
speed and is most often due to the mechanical unbalance of the rotor. Well
performed multi-plane balancing of the rotors helps to produce machines that
run smoothly across their entire running range. However this process can be
complex due to the number of influencing parameters (such as hydrodynamic
and aerodynamic features etc) present within most rotating machinery and
their variable nature during operation, specifically machines where the rotor is
required to operate at super critical speeds. Even with these tools to control
the vibration response of the rotors, other aspects and features of a rotor-

1
Chapter 1 - Introduction

bearing system can result in unforeseen dynamic excitation and response

yielding ‘vibration’ related problems.

At a top level vibration can be categorised into two types - those that are
stable in nature and those that are unstable. Stable vibrations, although
possibly high in magnitude, will remain at a constant level while operating in a
steady state condition. Unstable vibrations however, while running at certain
speeds, will increase in magnitude with time whilst operating at constant
rotational speed. These unstable vibrations by their very nature have the
potential to be most damaging and ultimately destructive, often preventing
machines from achieving their required operating speeds and powers.
Additionally the time dependant nature of these types of stability issues make
them more complex to model, identify and predict at the design stage,
presenting themselves as a significant problem to the rotordynamic engineer.

A significant amount of effort has been directed toward research and theory
development into the prediction of rotating machinery performance over the
last 100 or so years. Aside from the founding pioneers, such as Jeffcott [25]
and Reynolds [26] the most significant developments have taken place in the
last 30-40 years following the need for a large number of large high speed
turbo-generators, super critical compressors and the introduction of the power
gas turbine engine in the late 1960’s and early 1970’s.

In the majority of cases, high speed large industrial machines operate on fluid
film bearings due to their relative insensitivity to high operating speeds and
loads on large diameter shafts whilst providing non-finite life operating
principles. As a result, a significant proportion of the work associated with
rotating machinery has been directed toward the development and
understanding of hydrodynamic bearing theory and measurement.

A fluid film bearing uses a liquid media (normally oil) to provide a

hydrodynamic pressure field between a rotating shaft and stationary bearing
resulting in a non contacting support for the shaft and its loads. Two principal
types of hydrodynamic bearing are used in rotating machinery - journal
bearings and thrust bearings. Journal bearings are used to support the radial
loads of a shaft, and thrust bearings are required to accommodate any axial
loads or location requirements of the rotating member. The bearing which is
focused on within this work is the journal bearing and its interactions with the
rotor and its’ dynamics. The journal bearing in its simplest form consists of a
cylindrical rotating shaft and a stationary bearing shell of a slightly larger

2
Chapter 1 - Introduction

diameter. The clearance between the two components results in a

converging-diverging geometry which, when filled with the lubricant,
generates the necessary hydrodynamic pressures required to support the
shaft and produces, whilst it does so, a series of relatively complex dynamic
conditions between the oil film and rotor.

The analytical, computational and diagnostic techniques involved with rotating

machinery analysis have developed significantly in the last 80 years with the
most significant theoretical advances being made in the 1970’s and 1980’s
with respect to both bearing and rotordynamic performance evaluation, such
as the work of Lund [27,28] etc. The exponential increase in computational
power available to implement some of the earlier developed analytical
methods now allows for large numerical representations of complex rotor
systems to be characterised in relatively short computation times of a few
minutes, as opposed to several hours or even days that 1980’s technology
allowed. This, coupled with the more advanced sensor technology and
diagnostic tools such as the dynamic FFT analyser and digital data recorders,
allows for the highly dynamic conditions present within high speed rotating
machinery to become better understood and more advanced evaluation
techniques developed.

The specific topic of this research is unstable synchronous vibration which is

thermal in nature. The vibration has been observed on large and medium
sized high speed rotating machines [11], resulting in the need for significant
modifications to overcome the problem. Measurements taken during
acceptance testing of one of the machines has shown that at constant speed,
the vibration magnitude increases with a corresponding change in shaft
vibration phase angle. Subsequent inspections have ruled out the possibility
of a seal rub being the cause of the vibration.

The phenomenon described, often referred to as the 'Morton Effect' [13] after
the author of a paper describing such a vibration, is attributed to the variation
of viscous oil shear occurring within a hydrodynamic journal bearing oil film
due to synchronous orbit of the shaft rotor. The change in viscous shear in the
oil film gives rise to thermal fluctuation, which results in the formation of a
temperature differential across the rotor journal. A static bow in the rotor forms
as a function of the differential shaft temperature, causing an increase in the
unbalance of the shaft and hence orbit magnitude. The increase in journal
orbit widens the temperature differential, increasing the unbalance, hence

3
Chapter 1 - Introduction

providing a positive feedback loop where, under certain conditions, unstable

spiral vibrations occur.

It is the intention of this research to perform a series of experimental tests to

provide actual data for a rotor specifically engineered to be susceptible to
experiencing the synchronous thermal Instability due to differential bearing
journal heating. The aim is to quantify the impact and influence
characteristics and various operating parameters and practical features.

The principal focus of the research is to design, produce and operate a rotor
test rig suitable for the investigative research into high speed rotor
performance with respect to bearing journal thermal differential heating, more
specifically the association spiral vibration instabilities. The test rig should be
suitably instrumented to enable quantification of the governing physical
properties which are understood to influence hydrodynamic bearing thermally
induced rotor instability. A more detailed breakdown of the specific objectives
is provided in the following list:

1. Provide a thorough review of the published literature in the relevant

subject areas linked to synchronous thermal rotor bows, thermal
instabilities and spiral vibrations.

2. Perform a detailed analytical review of the relevant theories, analytical

procedures and modelling techniques, benchmarking analysis
techniques where appropriate.

3. Using well documented practical reference cases perform a detailed

review of the rotor-bearing systems design by means of rotordynamic
and appropriate thermal modelling techniques. This will enable the
common characteristics to be assessed, in rotors where bearing
induced thermal rotor bows and unstable vibrations have been
observed, giving a qualitative, and where possible, quantitative account
of the apparent governing terms. Summarising this information to
establish qualitative observations as to a rotor’s sensitivity to thermal
rotor bow and thermal instability, and use of it to assist in the
experimental rotor design.

4. Specify, analyse and design an appropriate test rig for the evaluation of
journal bearing induced thermal rotor bow, with the ability to capture
both comprehensive dynamic response and vibration modes, but more

4
Chapter 1 - Introduction

significantly obtain live shaft journal temperature measurement. Use

available data from known thermally sensitive or thermally unstable
rotating machines to develop appropriate design guidance and
evaluation procedures.

5. Build and commission a test rig with appropriate physical

characterisation and measurement validation procedures.

6. Carry out an extensive programme of experimental testing to

characterise rotor journal temperature differential measurement for a
range of operating parameters and configurations, such as; lubrication
conditions, operating speeds, rotor geometry, bearing type, introduced
mechanical unbalance.

7. Obtain experimental test data for a rotor operating under conditions

identified as being sensitive to journal bearing induced thermal rotor
bow, and capture rotor performance data across areas of high rotor
bow sensitivity and possible thermal instability.

8. Audit existing analytical procedures against generated test data and

suitable practical case studies. Introducing qualitative, and where
practical quantitative, evaluation techniques. Also, where beneficial,
applying or adapting appropriate existing analytical procedures.

9. Provide state of the art evaluation of journal bearing induced

synchronous thermal rotor bow and instability phenomenon in rotating
machines, covering design considerations and detailing appropriate
evaluation techniques with guidance as to application, applicability and
limitation.

5
Chapter 2 – Literature Review

Chapter 2

Literature Review

2. Literature Review

2.1 Rotor Rubs (Newkirk Effect)

In 1925, in an article from the magazine of the association of German

engineers, Dr D. Thoma [29] described a vibration problem for a four journal
bearing, single thrust bearing vertical rotor of a water-wheel generator that
had an operating speed of 600 rpm. The unit experienced a high vibration
problem which developed within a few minutes, occurring after the machine
had been operating satisfactorily for a prolonged period. The machine was
shut down and restarted. Once again after months of satisfactory operation
the vibration was experienced resulting in the machine being shut down.
Observations showed that immediately after the shut down a bow was present
in one end of the shaft which disappeared over the next few minutes. It had
been observed during operation that the bearing at this location had been
running warmer than the others. Dr. Thoma documented that the vibration
was due to a point on the bearing journal being heated along one element
which resulted in a thermal bow, which in turn caused the journal to move
closer still to the bearing, resulting in more heat and shaft bow, and so on.
The net result of this was an unstable vibration of the rotor.

Where Dr. Thoma [29] had experienced and documented this type of vibration
phenomena, giving some physical description, it was Burt Newkirk [1] who
went some way further to understanding the problem of unstable vibrations in
rotors due to thermal bowing of a rotor. Newkirk [1] published an article titled
'Shaft Rubbing' in 1926 giving a qualitative analysis of a thermally-induced
phenomena that would describe Dr. Thoma’s [29] observations. It was
supported by a series of experimental observations.

Newkirk [1] described how a rotor operating below its critical speed can have
a ‘heavy spot’ (residual unbalance) which causes this point in the shaft to
distort and move radially outward. If this point then comes into contact with a
stationary component causing a rub, heat is generated. This heat results in
the material around the ‘heavy spot’ expanding and thus causes the rotor to

6
Chapter 2 – Literature Review

bow out about this location. The rotor proceeds to dig in harder at the rub
location generating more heat, which causes the feedback cycle described
above to take place.

Newkirk [1] then went on to state that rotors operating above their critical
speed are not subject to this cycle and resulting instability. In rotors that
operate above their critical speed, the ‘heavy spot’ is no longer located on the
outside of the rotor, but the inner. Any heating of the shaft due to a rub will
cause the rotor to bow in a direction which moves the ‘heavy spot’ inward and
toward the shaft centre line. This reduction in unbalance means that there is
no positive feedback aspect to the rub and thermal bow as the shaft moves
inward, away and clear of rub producing stable operation of the machine.
However, shafts which are in perfect balance or rub at a point due to a raised
section of geometry of the shaft, are an exception to the above. For rotors
where the rub occurs at raised points along the shaft, although the heavy spot
may not be at the same location, the light rubbing will bow the shaft about this
location and very soon becomes the heavy spot and starts to respond as
previously described.

Figure 1 Diagram of Newkirk Effect Above and Below 1st Critical Speed [19]

Newkirk [1] performed a number of experiments in conjunction with Mr. H.D.

Taylor [22] at the General Electric Company. The first were performed with a

7
Chapter 2 – Literature Review

carefully balanced rotor to which an out of balance of known mass and

location was introduced. The rotor was run at various speeds and with a slight
rub added onto the shaft surface. For the rotor operating below its first critical
speed, the vibration increased rapidly until severe rubbing occurred with high
levels of vibration. Whereas when the operating speed was above the critical
of the rotor and light rubbing was applied, vibration did not increase and
rubbing did not build up with the rotor eventually clearing itself completely
from the rub.

Figure 2 Diagrammatic Representation of the Backward Bending of Shaft

Running Above Critical Speed [1]

Marking blue that had been put onto the shafts before the tests showed that
the rubbing had occurred at the location of the out-of-balance below the
critical speed and on the opposite side to the unbalance when above the
critical speed.

A second type of test was performed by introducing a bow into the rotor shaft
by means of elastically deflecting the shaft in the static state. The mechanical
out of balance was now produced by the mechanical bow in the shaft. The
rotor was run above and below its critical speed. It was again observed that
below its critical, the out of balance (and thus bow) were outward facing and
above the critical inward facing. Although no rubbing tests were performed,

8
Chapter 2 – Literature Review

this confirmed that the same phenomena would occur in rotors where an initial
bow causes the mechanical unbalance.

2.2 Spiral Vibrations

In 1939 Kroon and Williams [2] later went on to give a more qualitative
analysis of what was now known as the Newkirk Effect.

They describe a mathematical model in which a rotor experiences a steady

increase in vibration amplitude with a continuous change in phase angle
relative to time. A method was proposed that allowed the identification for the
cause of 'unstable' increases in synchronous vibrations with time. Kroon and
Williams [2] indicated how the problem could be categorised into a change of
system state or a change in the unbalance of the rotor. A change in system
state is where either or both of the stiffness and damping characteristics are
changing with time. This is identified by the modification in resonant
characteristics displayed, particularly when running close to a resonate
region. In general this sort of change in amplitude in one direction would have
little effect in the orthogonal direction, resulting in small changes in phase
angle, if any.

However, where there is a change in unbalance due to a distortion etc. then

the shift in phase angle is essentially unlimited. Such a change in balance
results from an alteration in shaft distortion due to a thermal effect caused by
such things as rubs, i.e. the Newkirk effect. These vibrations, when plotted in
a polar form, have the appearance as shown in Figure 3 and hence go to
explain the title of 'Spiral Vibrations' for such phenomenon. The large changes
in phase and steady increase in amplitude show consistent geometrical
relations for vibrations in opposite directions which is consistent with a change
in unbalance. The time taken for the large changes in phase (several times
360o) were observed by Kroon and Williams [2] as being as short as 16
minutes to as long as 5 hours. These points support the model of a thermally
distorting rotor giving rise to a continually changing magnitude and vector of
unbalance.

The mathematical model produced by Kroon and Williams [2] explains how for
a shaft with an out of balance force (due to a static shaft eccentricity vector 'x')
has a dynamic unbalance vector 'y' with a lag of φ, see figure 4a. If rubbing
occurs then the contact will be initially made at the point of vector 'y'. This rub
will cause asymmetric heating to be generated in the rotor, causing a bow in

9
Chapter 2 – Literature Review

the direction of vector 'y'. The increase in bow causes an additional unbalance
vector ∆x which when combined with the original static vector x’ gives rise to a
new instantaneous unbalance vector x1' and hence dynamic unbalance vector
y1. This results in a corresponding increase in phase angle, see Figure 4 (b).
With the magnitude of rotor displacement increased, contact now occurs with
greater force and at a position further round the shaft circumference, causing
the cycle described above to repeat giving rise to an unstable vibration.

Figure 3 Vibration for a Rotor Displaying the Newkirk Effect [2]

Kroon and Williams [2] performed a number of simple tests on specifically

simplified rotors where the distortions resulting from heating a fixed spot were
measured. A quantitative assessment was produced using dimensional
analysis where it was found that the rotor could be accurately modelled using
two equations for the heating and cooling curves. See Figure 5.

10
Chapter 2 – Literature Review

Figure 4 (a) Relationship

(a) Between Static and Dynamic Bow (b)
(b) Development of
Spiral [2]

A fully analytical model was developed for the case of an infinitely long thin
walled cylinder subjected to local heating. Solutions were provided for heat
distribution and resulting deflection. The analysis was based upon some
significant assumptions in the definitions of the applied boundary conditions.

Figure 5 Typical Experimental Results Obtained by Kroon and Williams [2].

Solid Lines are True Exponential Curves, Datapoints were Obtained by Test
[2]

11
Chapter 2 – Literature Review

The work does not provide an accurate or exact method of solution but offers
a method of diagnosis as to the potential cause of a particular vibration
problem by means of the measured 'spiral vibrations'.

A full quantitative analysis of the thermal state in a rubbing rotor was not
tackled until 1973 by Dimarogonas [3,4]. A full analytical solution was
developed for a rotor in which the static bow due to an arbitrary heat input
could be determined as a function of a source bow and heat function [5]. The
dynamic response of the system was used to calculate the corresponding
dynamic bow. Dimarogonas [3,4] found that the heat generation and heat
function resulting from the dynamic bow could be determined as well as
stability / response of the rotor with respect to time. The rotor system was
modelled using two non-linear differential equations solved by the Runge-
Kutta method, using a digital computer program to calculate the system
response with respect to time.

The analysis established three separate modes associated with the Newkirk
effect which describe the stability of the system. These modes are spiralling,
oscillating and constant modes. Spiralling is an unstable mode, constant is a
stable mode and oscillating is a transitional state between the former two. A
number of experimental results compared well with the solutions for the
spiralling, constant and osolating modes. No analytical solution was available
for the oscillating mode [5].

The fundamental conclusions from the work are that the main factor
determining the mode in which the system will respond is the phase angle
between the static and corresponding dynamic bow. It also shows that the
mode is indirectly affected by the critical speed of the rotor, but is not the
principle influencing feature [5].

12
Chapter 2 – Literature Review

2.3 Rotor Differential Heating

A quantitative analysis of the Newkirk [1] effect was performed by

Kellenberger [6] with respect to Turbogenerator seal rings causing rotor
differential heating around the circumference. Unlike Dimarogonas [4] a
number of simplifying assumptions are made and more significantly the
system equations remain linear. The analysis is dominated by two parameters
'p' and 'q' expressing the power entering and being removed from the rotor
respectively. The stability of the system can be shown to be governed by the
ratio of p/q, where above some threshold, unstable increasing spiral vibrations
will exist. Stability is defined as the point at which the amplitude and phase,
when plotted in polar form, produce a circle.

The qualitative analysis compared well with published results - even with the
inclusion of the simplifying elements used in calculation. Spirals were
identified as normally opposing the rotation direction of shaft but could be
reversed if sufficiently high damping or inertia is present within the seal ring
(stator).

Kellenberger [6] concludes that the time taken to complete one 360o phase
shift is governed by the power entering the rotor only. Also the shaft critical
speed is not a significant factor in the production of spiral vibrations produced
in this manner. This is an important point as this is in some contradiction to
the rubbing mechanism identified and quantified by Newkirk [1]. The location
of the running frequency with respect to the critical speeds and shaft mode
shapes can determine the stabilising or destabilising effects under such
conditions.

To the author's knowledge, the first published identification for the problem of
differential heating of the journal due to synchronous vibrations within the
bearing was made by Ericsson [7]. The work was an entirely analytical study
which observed the temperature distribution within a bearing pad (more
specifically a tilting pad bearing) due to synchronous vibrations. The work was
split into two specific aspects, the 'static' solution and the 'perturbation'
solution (vibrating).

A series of idealising and simplifying assumptions are made which allow the
static and perturbation equations for pressure, temperature and viscosity to be
uncoupled. A number of numerical solution procedures are provided for in the

13
Chapter 2 – Literature Review

solution of the equations for static and perturbation hydrodynamic conditions

within the oil film by means of direct integration, difference methods and finite
element method.

The effect of the variable parameters were investigated in non dimensional

form along with some numerical examples based on specific assumptions
related to bearing vibration magnitude, vibration orbit and pad oil inlet
temperature. The solutions are considered to be adiabatic and provide no
account for heat transfer into either of the bearing pad or journal.

The analysis provides a method of solution (by means of application of

numerical methods using a digital computer) for establishing the temperature
variation around an infinitely long journal bearing (specifically tilting pad)
surface and within the oil film due to a harmonically varying film thickness.
However, the calculations rely heavily upon a series of assumptions -
specifically oil pad inlet temperatures, oil film viscosity and oil conductivity. It
shows that assumptions regarding the oil, (particularly viscosity and fluid
conductivity) and frequency of vibration have significant effects on the
temperature distribution around the slider surface. Ericsson's [7] analysis
method does not numerically or specifically address the heat transfer between
the fluid and rotor, rotor bow nor any feedback resulting from changes in
dynamic operating conditions due to harmonic vibration amplitude.

The potential for hydrodynamic bearings to produce a thermal differential

heating of the rotor was again recognised by Schmied [8] who also identified
slip rings as a potential source for spiral vibrations. Schmied [8] developed a
method for the solution of spiral vibrations using finite elements. The model
uses the linear relationship between thermal and mechanical effects and
solution procedure presented by Kellenberger [6] for seal rings. Although
recognising the importance of bearing hydrodynamics in the subject area the
analysis procedure is approximate and does not address the thermodynamic
and heat transfer effects present in initiating spiral vibrations. Also the
operating parameters of the bearing are not specifically incorporated, in that
orbits are assumed circular and of a known radius.

Schmied's [8] model provides a generic solution procedure using finite

elements for multi bearing and slip ring machines. The work of Kellenberger
[6] is extended slightly in that an attempt is made to equate the proportionality
factors p, p* and q (identified by Kellenberger) as opposed to estimation from

14
Chapter 2 – Literature Review

real machine data. However, the results are approximate and practically
indicative in application.

2.4 Morton Effect

It was Keogh and Morton [9] who produced the first fully qualitative and
quantitative analysis of a spiral vibrations in a rotor system due to thermal
effects within a hydrodynamic journal bearing. It is due to the work conducted
by Keogh et al [9,10] that from that point onward spiral vibration problems
associated with rotor bearings are referred to as the 'Morton Effect'.

The initial published work of Keogh et al [9,10] provides an analytical

assessment of the thermohydrodynamic process that occurs within a plain
journal bearing due to synchronous perpetrations in the oil film with respect to
time. Morton [9] gives reference to some industrial applications where spiral
vibrations have been observed where the cause was found not to be
attributed to rubs or seal rings, notably in machines with overhung sections.

It is an important feature that the analysis looks at the bearing solution

thermodynamically, as all previous analyses had not provided a thorough
account for the thermal conditions within the bearing. The qualitative analysis
of the 'Morton Effect' for a rotor running in hydrodynamically lubricated
bearings, shows that it will undergo a synchronous orbit due to unbalance.
This orbit will cause one specific circumferential section of the shaft to
experience slightly smaller clearances at the position of the minimum film
thickness than the section of shaft 180 degrees opposed. Generally, smaller
film thickness are associated with higher rates of viscous shear stresses, thus
a difference in film operating temperature will occur at shaft rotation
frequency. The result of this is to generates a 'hot spot' on the shaft journal
approximately at the position of the minimum film thickness resulting in a
temperature differential across the shaft and producing a thermal bow. This
mechanism will then, under certain conditions, continue to follow the process
of feedback already described.

A time dependant analysis is provided, calculating the thermal bend by

coupling the heat transfer effects with the rotor dynamic calculations. The
analysis developed by Dimarogonas [3] is used to determine the rotor bend
associated specifically with the thermal influences. By converting the thermal
bend response into the frequency domain, a positive feedback loop is used to

15
Chapter 2 – Literature Review

assess the system stability by means of Nyquist plots and Eigenvalues with
respect to shaft speed.

A simple model of a single bearing overhung shaft section was used as an

example model. The solution of this predicted a speed range of instability
which gave Eigenvalue results producing a system gain where G>1.

The analysis showed that temperature differences of 4o-10o C can develop

under certain synchronous vibration orbits. Agreement was made with
Kellenberger's [6] observations that unstable vibrations result in a 'spiral'
progressing in the direction of rotation, whereas stabilising vibrations rotate in
the opposite direction.

Analysis also showed that conditions for stability are speed dependent and
are usually in the vicinity of a critical speed due to the level of dynamic
magnification of rotor unbalance response. A significant feature identified is
that the position of maximum shaft heating is not coincident with the position
of minimum film, but with a significant lag due to convective heat transfer
within the lubricant film. A lag of approximately 50-55 degrees was calculated
in the numerical example performed by Keogh et al [9].

A practical example of the Morton effect was investigated by de Jongh and

Morton [11] when a synchronous instability, caused problems during API
(American Petroleum Institute) testing of a centrifugal compressor on the
manufacturers test bed. The compressor rotor, having a mass of 450 kg, was
mounted on two tilting pad hydrodynamic journal bearings. It was found on
test that the rotor could not achieve the full running speed of 11,947 rpm due
to high synchronous vibrations although the closest critical was at 14,500 rpm.
It was observed that at a speed below approximately 9000 rpm vibrations
were stable, but as speed was increased in incremental steps neither the
vibration magnitude nor phase stabilised. At 11,400 rpm the vibrations began
to grow rapidly and the run was reduced back down to 10,000 rpm. However
the vibrations did not decrease immediately - after a short dwell the vibrations
returned to about normal, see Figure 6. From this it was deduced that the
problem was thermal in nature and the diagnosis of the Newkirk effect was
identified and attributed to a seal rub.

16
Chapter 2 – Literature Review

Figure 6 Fundamental Component (1x) of Relative Shaft Vibration [11]

As a result the test was repeated with the labyrinth seals removed, however
the vibration problems again demonstrated the same pattern, ruling out the
Newkirk Effect. To enable the compressor to get through test, the overhung
mass of the rotor was reduced by use of alternative materials. The resolution
of the problem, although not fully understood, pointed towards the Morton
effect being the prime cause.

To enable a fuller understanding of the problem, a test rotor was constructed

having effectively identical dynamic characteristics as the actual compressor
rotor. After some prior testing, small Resistance Temperature Detectors
(RTD's) were located within the rotor bearing journal. Four sensors were used
to allow the full temperature profile to be obtained based on the assumption of
harmonic distribution. Glue and slight interference fits were used to ensure
good thermal contact and the electrical signals were passed through a
slipringless transmitter to the stationary measuring equipment. Temperature in
the bearing pads was also monitored as well as the shaft displacement at
various positions along its length. All were referenced to a once per revolution
pulse marker.

Results from the test rig showed that at speeds above 10,500 rpm, significant
temperature differentials across the journal developed - from 3o C and
upward. It should be noted at this point that the magnitude at which the

17
Chapter 2 – Literature Review

temperature differential will become significant with relation to rotor stability is

very closely linked to the rotor and bearing geometry. For a rotor which has a
high thermal bow gain, temperature differentials as low as 1oC could be
sufficient to trigger unstable vibration growth.

As the rotor speed was increased spiral vibrations occurred, in the same
direction as shaft rotation, with large unstable amplitudes occurring at
approximately 11,500 rpm. At constant speed it was observed that the
vibration levels were increasing at approximately 15% of the bearing
clearance per minute which coincided with continually increasing differential
journal temperatures. Further testing showed that the location of the onset of
instability could be moved by around 900 rpm by quite small changes in
overhung mass, demonstrating the sensitivity of the problem to overhung
rotors.

Further examples of the Morton effect were encountered by Falkner, Strong

and Kirk [12,13] during testing of an overhung turbocharger rotor. After some
initial problems with loose fitting rotors, another problem was experienced
further up the speed range. The problem demonstrated the characteristics of
the Morton effect and was identified as such. Although phase measurements
were not made, the time dependent vibration amplitude increases and
vibration hysterisis loops pointed toward the phenomenon. In this case the
problem was overcome by modifications to the bearing, producing an increase
in operating eccentricity ratio and resulting in a less centered orbit which
allowed increased cooling of the shaft 'hot spot'.

De Jongh and van der Hoeven [14] describe another example of the Morton
effect in which a compressor experienced unacceptable levels of synchronous
vibrations during operation, after successfully completing a full API shop test.
After some examination and further testing, the seal rub was eliminated and
the Morton effect identified as the cause in one of the machine bearings. Due
to machine specific constraints, modification of the bearing was not possible
and an alternative method of overcoming the problem was required. A heat
barrier sleeve was incorporated to introduce a discontinuity in the heat
transfer in the shaft journal, see Figure 7. The result of this was to reduce the
amount of temperature variation around the journal and hence the rotor
thermal bend. Additional testing confirmed the success of the modification and
stable operation of the machine was experienced in the field.

18
Chapter 2 – Literature Review

Figure 7 Typical Temperature Distribution with Heat Barrier Sleeve [14]

More recently some further analytical research has been conducted

associated with the Morton effect. Larson [15,16] extends the thermal
modelling of Keogh and Morton [9,10] by the inclusion of tilting pad bearings
whilst making some simplifications in the solution procedure. Tucker and
Keogh [17] and Gomiciaga and Keogh [18] have employed CFD techniques to
analytically determine the temperature distribution around and across a
rotating bearing journal with a perturbating orbit. Dynamic cavitation effects
are determined (rather than initially estimated) in an attempt to obtain more
accurate temperature differentials. The results of the analysis demonstrate
that the journal does not behave in a isothermal fashion and can have
significant static temperature differentials across the rotating shaft journal.

Possibly the most extensive amount of research conducted into the Morton
effect since Keogh and Morton [9,10] has been presented by Balbahadur [19]
and Balbahadur and Kirk [20,21,23,24]. Balbahadur [19,20,24] developed a
simplified model in which a steady state analysis is performed with the
objective of determining the onset of the Morton effect. The analysis is based
on determining an unbalance threshold force which is a function of the total
rotor mass. Account is taken for both plain and tilting pad journal bearings and
a suite of software was developed to perform the analysis determining a

19
Chapter 2 – Literature Review

stability speed threshold. The criteria used to establish the thermal unbalance
force thereshold is to take the product of a constant factor and the rotor total
mass.

Balbahadur and Kirk [21,23] use a number of published examples of the

Morton instability [11,13,14] as case studies for comparison with the results of
the technique developed, to identify the onset of instability based on an
unbalance force threshold. The results give reasonable agreement between
the experimental and theoretical model. Balbahadur and Kirk [19,20] identify
the significance of the orbit centre position and orbit magnitude on the onset
of Morton instabilities. The analysis indicates that rotors operating with small
eccentricity ratios and large circular orbits are more susceptible to Morton
instabilities than those of large operating eccentricities. Some discrepancy is
apparent as to the factor used to determine the unbalance force threshold. In
Kirk and Balbahadur [24] a factor of 35% or rotor weight is reported to provide
good correlation of a instability speed threshold whereas in later work
[19,20,21,23] this factor is reduced to 15%. The precise causes for this
change is not apparent in the available literature, however it is the author’s
assumption that when a wider field of case studies was evaluated [19] then a
better aggregate correlation was observed with a factor of 15%, which is
consistent with the results presented in [19].

Balbahadur [19] performed some preliminary experimental testing of a rotor

intended to demonstrate the Morton effect, the results of this testing however
did not culminate in the presence of the Morton effect. The experimental rig
monitored shaft vibration; however it did not have the ability to measure rotor
journal temperatures. Little meaningful data was obtained from the rig as a
result of other unresolved dynamic issues within the test rotor system.

The analysis presented by Balbahadur [19] does not provide a full account for
all of the variables involved in the Morton effect, but does provide a tool which
appears to serve as a useful guide on the susceptibility of a rotor to instability.
It is also a useful simplified analytical method to model the impact of the
bearing operating characteristics, such as shaft orbit, on the stability of a
rotor.

Kirk and Guo [30] go onto present some minor developments of Balbahadur’s
[19] work, but principally the same analytical and computational process is
applied. The only significant apparent analytical modification to the procedure
is the method by which the considered journal hot and cold spot locations are

20
Chapter 2 – Literature Review

determined. Balbahadur [19] defined the journal hot and cold spots by static
geometrical means where the hot spot is defined as the point which is closest
to the bearing sleeve at a time constant t=0, identified by Kirk et al [30] as
method #1. Kirk et al [30] propose two further approaches that can be
considered to define these locations. The same static geometrical approach is
applied but now the hot spot can also be defined as either the closest point to
the bearing sleeve at time constant t=tm, method #2 or the hot spot is defined
as the point at which the journal is found to have the maximum temperature
found in one orbit cycle, method #3. Kirk et al [30] do not provide an account
as to which method provides a more accurate representation of the physical
situation but later go on to conclude that, based on the two case studies
presented, it is recommended that method #2 be applied for both plain and
tilting pad bearing approaches. This is substantiated by the principle that
method #2 produced better correlation to the case studies considered for tiling
pads and the plain bearing solution is relatively insensitive to the method
based on the analytical procedure being employed.

Kirk et al [30] conclude that, in addition to the key influencing criteria identified
by Balbahadur [19] on sensitivity to thermal instability (low bearing eccentricity
and circular obits), the relative phase angle between the initial mechanical
unbalance and the thermal unbalance is also of significant importance. This
intuitive statement relates to the fact that at large phase angles, (around 180
degrees) the thermal and mechanical unbalances would combine to result in a
lower effective mechanical unbalance condition, favouring stability. As part of
the tilting pad case study Kirk et al [30] investigate the effects of bearing
clearance and pre-load on the sensitivity and location of thermally induced
instability. It is stated that the increasing of bearing clearance has the greatest
effect on improving instability risk, while changes in pad pre-load seem to
have no serious effect on the thermal instability when considered within the
same clearance range. The small impact made by the changes in pre-load
were stated to provide some improvements to instability for a relative
decrease in pad pre-load, however these impacts were relatively limited.

A further case study was published by Marscher and Illis [31,65] where the
unexplained cyclic vibration of an integrally geared compressor supported in
tilting pad bearings, was attributed to a form of the Morton effect. Following
installation of an improved oil cooling system the compressor, which had been
in service for some years, began to experience high vibration which cycled
over approximately a six minute period. The vibration at both support bearings
was seen to increase from ~18 microns pk-pk to 31 microns pk-pk over a

21
Chapter 2 – Literature Review

specific period in time during which any change in vibration amplitude was
accompanied by an associated phase change. The phase change during
cycling equated to approximately +60o for both support bearings. The problem
was confirmed to not be associated with any critical speeds of the rotor, the
fundamental of which was 195 Hz and the rotor operating speed was 255 Hz.
It was also observed that the thermal cycling was in some way coupled
between each end of the shaft, with the two bearing vibration amplitudes
maintaining different phase angles which reduced and increased with time.

The Newkirk effect was ruled out by the authors as it was concluded that if a
rotor-stator/seal rub was the principle cause the rub would ‘burn’ itself out and
clear over a given period of time. A significant element which pointed toward a
thermal influence in the rotor system was the fact that the cyclic increase in
vibration was initially triggered when the oil inlet temperature was dropped
below a specific threshold temperature, which was found to be ~52 degrees
C. Further reductions in inlet temperature below this level only served to
further increase the magnitude of the high vibration cycles. Whilst not fitting
the traditional model of the Morton effect amplitude-phase relationship,
several of the conditions within this reference do indicate some characteristics
symptomatic of the Morton effect. In particular the potential for interaction
between opposing bearings raises some further complex coupled thermal-
rotordynamic aspects not yet discussed within the literature.

Schmied et al [32] present another example of a physical machine

experiencing a synchronous vibration problem. The machine was a double
overhung turbo-expander for the cryogenic industry running in 5-pad tilting
pad bearings operating at a top end running speed of ~18000 rpm. The
problem clearly demonstrated the forward spiral vibration characteristics
associated with a thermal instability at both support bearings and it is stated
by Schmied et al [32] that all other possibilities were eliminated and a journal
bearing hot spot condition was considered the principle cause. The rotor was
a short stiff rotor and as such was running in a sub critical condition. A
number of practical changes were attempted on the test bench such as
changing oil inlet conditions and bearing clearance - all of which introduced
minor changes but did not overcome the problem. A form of hot spot analysis
was performed based on the thermal stability sensitivity procedure presented
by Schmied [32] and incorporated the hot spot model of Kellenburger [6]. The
procedure uses a coupled rotordynamic matrix solution similar to that given by
Keogh et al [10] where the thermal conditions are based on [8]. The stability
assessment procedure compares the added heat (heat generated within the

22
Chapter 2 – Literature Review

oil film) to the eliminated heat (heat lost to the rotor/bearing structure) and is
p.Ω
expressed as where p = the proportionality factor for added heat, q =
q
proportionality factor for eliminated heat, and Ω = shaft speed. The
application of the method used is independent of bearing geometry and
therefore it is stated by Schmied et al [32] that this simplifying assumption will
affect the accuracy of the heat input terms. As such, to determine sensitivity,
only 50% of the calculated values are used in the analysis. The result of the
hot spot evaluation predicted that the rotor would be sensitive to thermal
instability therefore some proposed changes were evaluated in an attempt to
determine their impact. The physical changes that were implemented to the
rotor, which resulted in an acceptable dynamic performance on the test
bench, was to increase the diameter of the centre section of the rotor and to
reduce the support bearing width. The effect of which increases the stiffness
of the shaft in the overhung deflection mode and the reduction in bearing
width reduces the thermal bend sensitivity, due to a reduced level of
proportional thermal strain in the rotor. In addition to the geometrical changes
the lubricant viscosity was also reduced from ISO VG46 to VG 32 to reduce
the amount of heat being generated due to oil film shearing.

Schmied et al [32] also states that spiral vibration problems have been
experienced on highly load bearings with highly elliptical orbits. This is in
some contradiction to the proposal by Balbahadur [19] who comments that
centred shafts with circular orbits are most sensitive. Schmied et al [32]
however, does not include any evidence or references of such examples.

In 2008 Frits de Jongh [33] published a technical paper in which a collective

summary of the published work and field examples linked to the Morton effect
is presented. de Jongh [33] details the principle areas of activity that have
been undertaken with regard to the phenomenon giving examples of practical
cases where the Morton Effect has been attributed to rotating machinery
vibration problems. The details of the work are consistent with the references
already outlined earlier in this chapter. However, de Jongh [33] also gives
commentary on some as yet unpublished work based on internal works
conducted by Morton conducted at GEC in the UK. De Jongh [33] states that
Morton performed tests on a large low speed test rig where circumferential
temperature measurements were taken by means of 12 thermocouples
placed on the rotating shaft journal. The test rig uses a bearing diameter of
approximately 28 inches with an operating speed of 1800 rpm. No specific

23
Chapter 2 – Literature Review

details are provided of the detailed bearing geometry, operating clearances or

lubricant type/viscosity.

It was found during the tests, that significant temperature differentials were
observed even for relatively small shaft orbits (“only a few % of bearing radial
clearance” [34]). It was also stated that a near linear relationship between
shaft orbit size and journal temperature differential was observed. This is also
consistent with the work presented by Larson [15,16] detailed earlier in this
chapter, where a linear response between journal differential heating and
vibration amount is stated. However, it should be noted at this point that no
detail is provided as to the size of the vibration orbits as a percentage of radial
clearance over which a near linear response with temperature differential was
observed. Also no comment is made as the temperature distribution profile
around the shaft journal circumference, i.e. harmonic and non harmonic
components.

De Jongh [33] groups the various examples of rotating machinery vibration

issues that have been reported as a result of the Morton effect. De Jongh [33]
identifies that in all of the cases presented they share some common
characteristics, in that they are all supported by plain hydrodynamic bearings
and to a greater or lesser degree have relatively large overhung moments. De
Jongh [33] makes a useful summarisation that whilst all the rotors have some
common elements and features, the actual rotors do have different types of
configuration and a breakdown of these is presented in Table 1. (It should be
noted that de Jongh’s [33] table is no longer current due to the addition of two
further case studies published since this work).

Table 1 Breakdown of Various Rotor Types Identified as Experiencing the

Morton Effect [33]

24
Chapter 2 – Literature Review

de Jongh [33] provides details the approach taken by de Jongh and Van
derhooven [14] to establish the sensitivity of a rotor set to synchronous
thermal instability due to shaft bending, was introduced as standard to the
rotordynamic assessment of a rotor assembly. De Jongh [33] states that the
procedure consisted of determining the rotor mechanical unbalance for unit
thermal bending at the bearing location. From this condition, the rotor
response at the bearing locations is determined. Using some empirical data,
the non uniform temperature distribution is estimated based on the operating
conditions under evaluation. The estimated temperature differential is then
used to determine the resulting shaft thermal bend. The modulus of the
complex thermal bend angle is determined as a ratio of the unit input and
output complex bend angle and it is this ratio which is used as a criterion for
stability.

De Jongh [33] notes some possible practical corrective actions for machines
that are considered to be subject to the Morton Effect instability.

• Limit the design speed

• Reduce overhung moments
• Change bearing clearances
• Reduce bearing length
• Change bearing type or geometry
• Heat barrier sleeve
• Increase specific bearing loading and eccentricity
• Change shaft material
• Change lubrication oil viscosity
• Increase inlet oil flow

It is also noted that the Morton effect not only applies to new machinery but
can also be experienced on established machinery in the field, such as
compressors, where it is common to upgrade certain components to obtain
increased performance or capacity, as observed by Faulkerner et al [13].

Childs and Saha [35] present an iterative algorithm for the synchronous
response analysis for analysing the Morton effect. The procedure is
dependent on a suitable thermal model being applied to solve for the journal
thermal conditions and thermal sensitivity to shaft orbit size. Childs et al [35]
state that a procedure similar to that presented by Keogh et al [9] is required

25
Chapter 2 – Literature Review

to obtain the necessary journal temperature differentials and phase angles,

such that these can be applied to the rotordynamic model. The procedure
uses the results of a complex thermal analysis to calculate the shaft
temperature differentials and thermal phase angle for 5 orbit radii conditions
for a given bearing operating condition. These results are then interpolated
between data points when applied to the rotordyanmic model. The algorithm
in principle then involves performing an iterative procedure of rotor response
(the initial condition of response is determined for the shaft mechanical
unbalance). From this, the resulting degenerated forward and backward whirl
amplitudes are extracted. Applying these amplitudes to the thermal analysis
results, appropriate forward and backward whirl journal temperature
differentials and phase angles can be interpolated as appropriate. Childs et al
[35] then determine a resultant maximum journal temperature differential and
phase angle by:

(
∆T1 = ∆T f2 + ∆Tb2 + 2.∆T f .∆Tb .cos 2.β * + ϕTf + ϕTb ) Eq. 2.4.1

( ) (
 ∆T f .sin β * + ϕTf − ∆Tb .sin β * + ϕTb
ϕT = tan 
−1 )  Eq. 2.4.2
( ) (
 ∆T f .cos β * + ϕTf − ∆Tb .cos β * + ϕTb
 ) 

Where ∆T f , ϕTf , ∆Tb , ϕTb are the temperature differential and phase for forward
and backward whirl respectively, computed from the complex bearing energy
model. β * is the shaft bend angle present from the previous iteration
calculation.

The resulting shaft bend angle is then calculated from Dimoragonas [8] and
importantly, this is used to determine the moments at the bearing location
which are then directly applied to the rotordynamic model to generate the
excitation (as opposed to other methods which determine an equivalent
mechanical unbalance force due to the eccentric overhung mass).

The unbalance components are combined and a new thermal unbalance

solution is performed, this procedure is repeated until a converged solution for
the shaft orbit is obtained. The gain factor following each iteration is stated by
Childs et al [35] as:

Fi + Bi Eq. 2.4.3
GaT =
Fi −1 + Bi −1

26
Chapter 2 – Literature Review

Where Fi and Bi are the forward and backward whirl amplitudes respectively.

Childs et al [35] state that the analysis is heavily dependent on the availability
of appropriate data from a bearing thermal model of some sort (which
provides both temperature and phase data for complex shaft orbits). To date
the only models available - Keogh et al [9] or CFD (Computational Fluid
Dynamics) solutions - are prohibitively complex and computationally/time
intensive for practical application. Childs et al [35] also state that if a
computational method is employed, where data is interpolated between a set
of given conditions from the bearing thermal analysis, then it is possible to
encounter a numerically unstable condition if too few orbit data points are
available for interpolation, resulting in an oscillating result which tends to
diverge.

Most recently Murphy and Lorenz [36,37] have published several works
addressing rotor thermal hot spot development and corresponding spiral
instabilities. Murphy et al [37] present an industrial case study where a
variable speed electrical machine was found to exhibit unstable time
dependant shaft vibrations when running at its maximum design speed of
4175 rpm for prolonged periods (~2 hours). The rotor was supported in partial
arc plain bearings and was designed to run through several critical speeds,
the fourth of which was predicted to be at a frequency close to running speed,
with a calculated frequency of 4000 cpm. The natural mode at this speed was
associated with high vibration activity exhibited by the overhung sections of
the rotor, although no specific qualification of predicted stability/damping for
this mode was quoted in the literature. Although, it is noted by Murphy et al
that this mode is sensitive to unbalance at the overhung shaft section.

The unstable vibration was observed to occur whilst running at operating

speeds over 4000 rpm for some 100+ minutes before excessive exponential
growth was encountered. A significant aspect was the time history
characteristic of vibration which showed symptoms of a harmonic function
with an exponential growth curve, suggesting an oscillatory nature, before
instability becomes onset (see Figure 8). An interesting feature of this case
study is that several units of identical design were produced with each
machine displaying slightly different characteristics with respect to any Morton
effect instability. Murphy et al [37] state that of four machines produced, one
machine showed no clear signs of instability during shop testing, another
showed an oscillating mode but no instability and another machine showed
very rapid unstable growth. Murphy et al [37] identify that from the test data

27
Chapter 2 – Literature Review

presented, it is clear that the time taken for an instability to show itself, or be
present, are very sensitive to the physics involved in the Morton effect
phenomenon; such as variations of bearing clearance, bearing alignment
ambient temperature, machine operating temperature, lubricant properties,
overhung mass, support stiffness, etc.

Figure 8 4150 RPM Constant Speed Factory Test Proximity Probe 1x

Vibration Amplitude vs Time [37]

For the each of the tests performed over the four different machines Murphy
et al [37] present the relative growth factors and time constants experienced
for the vibration oscillations and divergent cases. Murphy et al [37] document
shaft vibration growth rates ranging from 7 %/min to 177 %/min. Differences
of up to 160 %/min were published for the same machine between different
runs.

It was observed by Murphy et al [37] that, for the machines that showed the
least resistance to experiencing the Morton effect, synchronous response of
the machine when first reaching running speed was lower than that of the
other machines. It is postulated by Murphy et al [37] that this effect of lower
synchronous response close to the Morton instability threshold could be
indicative of a less stable configuration. The qualification for this was that the
thermal bow condition could result in a lower state of mechanical unbalance
prior to the instability threshold being reached or instability taking hold.

The spiral vibration problem in the machine described above was stated as
being overcome by the introduction of an alternative bearing design. The
original partial arc bearing arrangement was replaced with a 4-lobe plain
bearing configuration. The introduction of this bearing design was seen to

28
Chapter 2 – Literature Review

eliminate the presence of the Morton effect instability. The proposed

justification by Murphy et al [37] for the 4-lobe bearing eliminating the Morton
effect, was given to be the increased cooling effect introduce by the 4-lobe
bearing.

Murphy et al [37] outline a procedure to evaluate the sensitivity of a rotor

thermal instability on the Morton effect. The procedure was applied to assist in
evaluating the case study presented in [37]. It outlines how the application of
a number of sensitivity terms can be combined to establish how sensitive a
system is to instability. The sensitivity terms relate to shaft temperature
differential as a function of vibration magnitude, thermal bend unbalance
sensitivity as a function of thermal differential and other more typical rotor
dynamic elements. The product of these terms yields a first order differential
equation where the solution results in a single complex eigenvalue s . Stability
of the system is defined by Murphy et al [37] as any result where the real
component of the eigenvalue is < 1. Murphy et al [37] identify 3 possible
modes of stability: (1) Re s < 1 stable-converging spiral vibration (2) Re s = 1
stability threshold-oscillating but not divergent (3) Re s > 1 unstable-divergent
spiral vibration. These modes of stability reflect those identified by Dimigronas
[5] when associated with rotor rubs.

A significant area highlighted by Murphy et al [36,37] is that the requirement

for an accurate thermal model of the bearing is critical for accurate predictions
of rotor stability to be achieved. The principle suggested by Murphy [36] to
establish the temperature distribution around the journal uses similar
principles as that applied by Balbahadur [19] with the exception that a full
circumferential temperature distribution is extracted for each orbit location
instead of just the hot and cold spot locations. It is noted by Murphy et al [36]
that accurate definition of the temperature differential is difficult to achieve and
whilst the complex procedure employed by Keogh et al [9,10] provides the
necessary data, it is not practical to implement these complex procedures at
an industrial level. In particular Muphy et al [36] indicates that the phase
angle, which cannot be calculated with the presented approach, of the hot
spot around the journal is of critical importance when defining the sensitivity of
a system’s stability performance. No approach or solution for solving the
journal thermal problem is proposed or presented within the literature. Any
account for the phase angles involved is included by means of introduction of
assumed phase lags based on some associated data or intuitive estimates;
although Murphy et al [36] does state that there is the need for experimental

29
Chapter 2 – Literature Review

verification and/or data of the thermal temperature differential and phase

angles.

Lorenz [38] presents a study in which a propriety CFD modelling package is

used to attempt to model the thermal conditions associated within the bearing
oil film, shaft journal and bearing housing. The principle objective of the work
was to be able to determine the necessary thermal data points required for
implementation in any thermal stability rotordynamic analysis such as that of
Childs et al [35] and Murphy et al [36]. The research investigates some
practical aspects of the CFD modelling techniques used to represent the fluid-
film bearing model such as mesh density, boundary conditions and solution
methods. A periodic iterative approach is adopted over a full transient solution
technique in an attempt to overcome the associated physical and
computational difficulties that would be involved in the latter.

The conclusion of the work of Lorenz [38] was that it was possible to establish
appreciable temperature differentials across the shaft journal, however the
magnitude did not agree with that of other referenced literature such as [9,11].
The magnitudes predicted by Lorenz [38] were systematically higher than
those of Keogh et al [9]. Although Lorenz [38] states the results did compare
well to that presented by Dowson [39].

Another key element within the research was the determination of the phase
angle of the hot spot on the journal. The work of Lorenz [38] showed not to be
in agreement with that of Keogh et al [9]. For a circular forward whirl orbit, the
phase angle was predicted to lag the minimum film thickness point by ~60o,
whereas the results of Lorenz [38] showed that the phase led by
approximately 42o degrees. In addition to the numerical results, it is of notable
observation that for the single point solution cases undertaken in the
research, the solution times required are expensive both computationally and
in time. Typical solution times for a single orbit converged solution were given
by Lorenz [38] as 16 hours with significant further time and effort required to
extract the solved results and post process.

30
Chapter 2 – Literature Review

2.5 Hydrodynamic Bearing Theory

When considering the influences taking place in rotors exhibiting thermal

instability due to shaft bow, an appropriate appreciation of the hydrodynamic
effects and analytical modeling techniques involved is necessary. It is not the
intention to cover in detail here the very numerous sources of material
covering this subject matter, however, the principle aspects relating
hydrodynamic theory are described in the landmark paper by Reynolds [26].
Here the governing equations relating the hydrodynamic principles in fluid
bearings are described and presented. From this founding work many
hundreds of other significant works are available addressing the many
important hydrodynamic features and effects present in rotating machinery
and in particular plain bearings. Significant contributions in recent times have
been the works of Sommerfeld [40], Ovkirk [41], Lund [27,28], Garner [42] and
Cameron [43] to name but a few who have all contributed significantly to the
development of hydrodynamic theory relating to journal bearings for rotating
machinery. The references covered here, whilst far from exhaustive, provide
sufficient account to cover the generalised hydrodynamic conditions which are
of interest within the scope of this research.

31
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Chapter 3

Analytical Review of Bearing Journal

Differential Heating

3. Analytical Review of Bearing Journal Differential Heating

3.1 Introduction

This chapter presents the analytical methods applied in the evaluation of

bearing journal differential heating and more specifically, the assessment of
thermally induced rotor bend stability evaluation by several published
methods. Two specific methods evaluated in detail are: (1) those presented
by Keogh and Morton et al. in their papers on ‘Journal bearing differential
heating evaluation with influence on rotor dynamic behaviour’ and ‘The
dynamic nature of rotor thermal bending due to unsteady lubricant shearing
within a bearing’ [9,10], (2) that presented by Balbahadur and Kirk in several
papers [20,21,23,24] and in more detail in the thesis ‘A
thermoelastichydrodynamic model of the Morton effect operating in overhung
rotors supported by plain or tilting pad bearings’ [19]. In recent times several
other works have been presented on this topic as outlined in the previous
section, however these are effectively subtle variations on the principles
summarised in this chapter with the founding work of Keogh et al and
Balbahadur et al being the most comprehensive works available today in the
published domain.

Much of the founding analytical procedures involved the thermal bend

analysis procedures relate to well published quantitative techniques for the
assessment of hydrodynamic bearings and lateral rotordynamic
behaviour/response. As such only a brief analytical summary is provided here
for the above two methods by way of an introduction to the thermal bend
stability theory developed in each case.

32
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

3.2 Hydrodynamic Lubrication in Bearings

3.2.1 Brief Overview of Hydrodynamic Lubrication

Lubrication in some form or another is a fundamental principle in the working

of all machines where relative motion is present between the moving parts.
Without the presence of lubrication the losses and forces introduced would
result in the workings of all machines, be they rotating or linear, to become
impractical.

The mechanism of lubrication is the principle where two surfaces that have
some form of relative movement are separated by a film of material which can
be sheared in some way without causing unacceptable levels of damage or
wear to the associated working surfaces. Principally four generic regimes of
lubrication can be considered and, dependant on the functional requirements
of a given mechanism, the type of lubrication experienced can consist of one
or several of the conditions given below across its operating range.

Four principle lubrication regimes :

1. Full or Hydrodynamic Lubrication: two surfaces are completely
separated by a fluid film with applied loads supported by the lubricating
media.
2. Elastohydrodynamic Lubrication: two surfaces are separated by a
very thin fluid film operating under high pressures where significant
interaction with the elastic deformations of the sliding surfaces exist.
3. Mixed Lubrication: two surfaces are partly separated by a film and
partly in contact (usually by asperities in the sliding surfaces surface
roughness)
4. Boundary Lubrication: two surfaces mostly are in contact where a
fluid or gas lubricant may be present but the lubricating properties are
the result of the physical interaction between the two sliding surfaces.

For the purposes of this work, the principles and theories associated with full
hydrodynamic lubrication are those which are relevant to the application of
plain bearings for high speed rotating machinery. It is the interaction between
these lubricating principles and the rotordynamic characteristics of a rotor that
interact to produce the thermal instability of interest.

The advantageous principles of hydrodynamic lubrication have been

employed for over two centuries. Hydrodynamic lubrication was given

33
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

scientific context in the mid 1800’s where several works presented qualitative
accounts of the principles of pressure generation in fluid films of two sliding
bodies. However, it was not until the work of Reynolds [26] in 1886 that a
detailed analytical analysis was presented showing how a converging wedge
shaped film was required to generate a pressure profile in the lubricating fluid
capable of supporting load. It is this keystone paper that has provided the
foundation on which all subsequent hydrodynamic analysis principles have
been based.

Figure 9 Principle of Hydrodynamic Lubrication Between Two Inclined

Surfaces – figure taken from [44]

3.2.2 Equations Governing Hydrodynamic Lubrication in Journal

Bearings

The work conducted by Reynolds [26] provided a mathematical description of

the hydrodynamic pressure profile that is generated between two converging
surfaces of differential sliding speeds separated by a fluid. The pressure in the
lubricant film is generated by the action of the viscous fluid being stationary at
the boundaries of the two working surfaces and, due to the relative velocities,
the lubricant is dragged into the converging gap. Working on a principle of
virtual incompressibility of the lubricant, a pressure field is generated which is
capable of supporting a load nominally normal to the direction of sliding.
Reynolds derived a differential equation which related the influences of
lubricant viscosity, sliding velocity and pressure, which was based upon a
given geometric description of the gap between the two working surfaces.

The partial differential equation presented by Reynolds in three dimensions is:

34
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

∂  h3 ∂p  ∂  h3 ∂p   ∂ ∂ ∂h 
 .  +  .  = 6  (U1 + U 2 ) h + (W1 + W2 ) h + 2  (Eq 3.2.1)
∂x  η ∂x  ∂z  η ∂z   ∂x ∂z ∂t 

The Reynolds partial differential equation in its full three dimensional form is
generally too complex to solve practically. To overcome this degree of
complexity several methods have been developed to simplify it to a form
which can more readily be solved.

∂  h3 ∂p  ∂  h3 ∂p   ∂h ∂h 
 .  +  .  = 6 U + 2  (Eq. 3.2.2)
∂x  η ∂x  ∂z  η ∂z   ∂x ∂t 

Equation 3.2.2 shows the Reynolds equation with several of these

simplifications applied to equation 3.2.1, such as the recognition that generally
no relative sliding occurs in the axial direction and the velocity is constant
∂ ∂h
throughout the film, i.e (Uh ) ≡ U . A further simplification can be made for
∂x ∂x
bearings running at steady conditions without perturbation of the oil film where
∂h
the squeeze film pressure term can be equated to zero = 0 and yields the
∂t
Reynolds equation as typically written:

∂  h3 ∂p  ∂  h3 ∂p  ∂h
 .  +  .  = 6U . (Eq. 3.2.3)
∂x  η ∂x  ∂z  η ∂z  ∂x

3.2.3 Solution of the Reynolds Equations for Journal Bearings

As the Reynolds second order partial differential equation cannot be solved

directly, it is necessary to implement further assumption and simplification to
enable discrete analytical solutions to be obtained. The two principle
simplifications applied to journal bearings are typically known as the Infinitely
Long Bearing and the Infinitely Short Bearing. The Infinitely Long Bearing
theory was introduced by Sommerfeld [40] whereby the bearing length is
assumed infinite and hence the pressure variation in the axial direction is
∂p
zero. This allows the term to be set to zero which results in the Reynolds
∂z
equation being reduced to one dimension as given by equation 3.2.4. The
application of this approach means that it can only be considered as
representative of the centre portion of the bearing.

35
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

∂  h3 ∂p  ∂h
 .  = 6U . (Eq. 3.2.4)
∂x  η ∂x  ∂x

Infinitely Short Bearing theory conversely considers the bearing to be of

infinitely short length. It was presented by Ocvirk [41] and results in a
pressure profile where the gradient in the z direction is much greater than that
in the x direction. This enables the first term from equation 3.2.3 to be
∂  h3 ∂p 
considered as zero,  .  = 0 , and results in Ocvirk’s Reynolds equation
∂x  η ∂x 
:

∂  h3 ∂p  ∂h
 .  = 6U . (Eq. 3.2.5)
∂z  η ∂z  ∂x

Ocvirk’s equation provides a method to solve the differential equation by

practical means for relatively simple film thickness relationships and, as it
retains two dimensions, generally results in a more accurate result than the
Infinitely Long Bearing approach for length to diameter ratios between 0.3 and
1. For this reason it is more widely used for the solution of practical plain
journal bearings which most commonly have L/D ratios of ≤ 1.

Where more accurate results are being sought, or with more complex film
thickness relationships such as journal misalignment or use of variable
geometry pads, a form of numerical solution needs to be applied. These
techniques employ the same fundamental principles and arrangements of the
Reynolds equation but reach solutions of the Reynolds equations by means of
approximations. This yields results within predefined or acceptable tolerance
ranges. Many numerical techniques are available for the solution of such
mathematical problems with the likes of finite element and finite difference
techniques providing common means of obtaining a solution to equation 3.2.3
(2D Reynolds equation).

The solution of the Reynolds equation is of significant importance to the topic

of synchronous thermal instabilities as it provides the basis around which the
shaft journal quasi-static and dynamic bearing positions within the bearing are
determined. From this, the corresponding pressure profiles and film thickness
shear rates can be established by some relationship which in turn provides a
means for extraction of journal temperature variations to be established. The

36
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

basic principles of the above Reynolds theory is in some form applied to both
of the analytical procedures reviewed here as presented firstly by Keogh and
Morton [9] and then Balbahadur [19]. Details of which are outlined in the
remained of this chapter.

3.3 Keogh et al [9] Procedure

Keogh and Morton presented several works [9,10] which evaluate the
principles associated with the differential heating of a bearing journal when
running in hydrodynamically lubricated bearings. A summarised review of the
work is contained in the literature review in chapter 2, and a detailed overview
of the analytical procedures is now covered in the following sub section. Of
the works reviewed, the work given in reference [9] is covered in detail here,
which addresses the models used to describe both the thermal conditions at
the bearing-oil-journal interfaces and the coupling interaction with the
rotordyanmic behaviour and response due to the induced thermal bend. The
work reviewed in [10] goes to build on the work given in [9] to introduce a time
dependant control logic which establishes a measure of system stability and
the associated time constants involved in the thermal oil film developments
and stability development. The review focus is placed on the theory
surrounding the actual modelling of the thermal bend development and
sensitivity, without account for the analytical procedure.

The principle approach of the method presented by Keogh et al [9] is to tackle

the problem by means of development of a full thermohydrodynamic model of
the oil film, bearing and shaft journal. The interrelations within the bearing are
evaluated for the thermal and spatial conditions present, with a plain bearing
handling the dynamic and steady conductive heat flux terms between the
lubricant and physical elements. The technique seeks to provide a holistic
approach to coupling the complex energy, hydrodynamic, thermal and
structural aspects of the dynamic thermal bow problem.

The principle of this section is to present and discuss the governing equations
derived and presented by Keogh and Morton [9]. Where possible the
derivation and expression developments are followed through in this text,
however in some areas the analytical developments and resulting expressions
are presented without proof. This is a reflection of the complex nature of the
analytical procedure and the intention of this paper is provide a sufficient
account of the methodology and expressions used to model the dynamic

37
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

thermal conditions, only to enable suitable observations and comparisons

made to the scope of work embodied with the author’s research.

The analytical account of the work of Keogh et al [9] is presented by

addressing the following aspects of the analytical model in turn. Keogh et al
[9] breaks the problem down as per the list below to develop the necessary
expressions and boundary conditions required to solve and couple the various
dynamic and steady state conditions of interest.

• Lubricant flow terms

• Lubricant energy terms
• Lubricant film thickness
• Lubricant temperature and temperature perturbations
• Bearing and journal conduction
• Lubricant temperature development
• Prediction of rotor bend
• Thermal bend influence on rotordynamics

3.3.1 Lubricant Flow Terms

First, the hydrodynamic formulations and lubricant flow terms are expressed
for the thermohydrodynamic oil film model. Keogh et al [9] presents the
general Reynolds equation as given previously in equation 3.2.2, in non
dimensional form (equation 3.3.1). This non dimensional equation is then
reduced by means of applying Ovkirk’s equation for short bearing theory but
now following through the non dimensional approach, equation 3.3.2.

∂  3 ∂P  1 ∂  3 ∂P  ∂H ∂H
H + . H = + 2. Eq. 3.3.1
∂β  ∂β  ∈L ∂ξ  ∂ξ  ∂β ∂τ

1  ∂H ∂H  2
P= .
3 
2.H  ∂β
+ 2.
∂τ 
( )
 . ξ − 1 . ∈L Eq. 3.3.2

The common assumptions for the short bearing theory are applied here in that
only positive pressure (P >= 0) can exist, negatively calculated pressure are
taken as 0.

38
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

The accompanying non dimensional velocity components for the full film
region can be expressed for the three principle directions as given in equation
3.3.3.

U =η (Eq. 3.3.3-a)

c ∂H ∂H 
V =   . η 2 + η 2 . (1 − η ) .
R  ∂β
( )
+ 3.η 2 − 2.η 3 .  (Eq. 3.3.3 -b)
∂τ 

3.η . (1 − η )  ∂H ∂H 
W =− . + 2.  .ξ . ∈L (Eq. 3.3.3-c)
H  ∂β ∂τ 

Flow continuity considerations need to be accounted for due the divergent

section of the oil film beyond the minimum film thickness point. Beyond this
location the oil film enters a cavitatived state where the oil breaks into a
number of streamers [45]. It is considered by Keogh et al [9] that the velocity
terms described in equation 3.3.3 are maintained but are now local to the
various oil streamers in the divergent section with reference to any axial
ordinate terms.

3.3.2 Lubricant Energy Terms

Whilst, to the author’s knowledge, no commonly adopted form of the non

dimensional energy equation for journal bearings is commonly applied, Keogh
et al [9] presents a non dimensional form of the energy balance relationship
by applying the full dimensional energy equation as given in equation 3.3.4
and introducing the non dimensional flow terms of equation 3.3.3 to result in
the dimensionless energy equation for short bearings given in equation 3.3.5.

 ∂Θ ∂Θ ∂Θ ∂Θ   ∂ 2Θ ∂ 2 Θ ∂ 2 Θ 
ρ L .cL .  u. + v. + w. +  L  2 + 2 + 2 +Φ
= k . (Eq. 3.3.4)
 ∂x ∂y ∂z ∂t   ∂x ∂y ∂z 

∂T η . (1 − η )  ∂H ∂H  ∂T
η. + . η . + ( 2.η − 1) .  . .....
∂β H  ∂β ∂τ  ∂η
(Eq. 3.3.5)
3.η . (1 − η )  ∂H ∂H  ∂T ∂T 1 ∂ 2T 1
− . + 2.  .ξ . + = 2 2
+ 2
H  ∂β ∂τ  ∂ξ ∂τ Pe .H ∂η H

39
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

ρ L .cL .Ω.c 2
Where Pe is the Peclet number Pe = which is a non dimensional
kL
term which links the rate of lubricant advectation with the diffusion flow rate.
Θ f is a specific reference temperature which is defined by Keogh et al [9] as
2
µ .Ω.  R 
Θ= . which is then used to non dimensionalise the temperature
ρ L .cL  c 
Θ
terms by T = . The presented equation 3.3.5 does not provide terms of
Θf
axial and circumferential conduction within the lubricant and the heat source
1
term is independent of axial coordinate. A further reduction of equation
H2
3.3.5 is achieved by introducing an axially averaged temperature across the
1
wetted bearing width where TL = ∫ T .dξ = T , removing the axial dependant
0

∂T
terms giving equation 3.3.6.
∂ξ

∂ 2TL Pe .η . (1 − η )  ∂H ∂H  ∂TL ∂T ∂T
2
− . η . + ( 2.η − 1) . . − Pe .η .H 2 . L − Pe .H 2 . L = − Pe
∂η 2  ∂β ∂τ  ∂η ∂β ∂τ
(Eq. 3.3.6)

To account for continuity of heat flux at the boundaries between the fluid film
and the bearing/journal, use of an axial fractional film function is introduced
d ( β ,τ ) to account for the cavative areas around the bearing circumference.
This allows consistency to be maintained for the axially averaged temperature
and heat flux in the bearing and journal surfaces.

The boundary conditions on the lubricant for the bearing and journal
interfaces are now written as :

∂TL H ( β ,τ ) ∂TB
TL ( 0, β ,τ ) = TB (1, β ,τ ) ( 0, β ,τ ) = −γ B . . (1, β ,τ ) (Eq. 3.3.7)
∂η d ( β ,τ ) ∂r

∂TL H ( β ,τ ) ∂TJ
TL ( 0, β ,τ ) = TB (1, β ,τ ) (1, β ,τ ) = −γ J . . (1,θ ,τ ) (Eq. 3.3.8)
∂η d ( β ,τ ) ∂r

40
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

ρ
Where r is non dimensionalised bearing ordinate, r = , and θ is a
R
circumferential coordinate but now referred to a fixed axis on the rotating shaft
journal.

3.3.3 Lubricant Film Thickness

The definition of the lubricant film thickness function is expressed in non

dimensional form in equation 3.3.8. The film thickness function H consists of
both the steady film thickness Ho as defined by the steady eccentricity ratio
∈o , and the film thickness perturbations resulting from the shaft elliptical orbit.
The elliptical orbit is represented by decomposed circular forward and
backward whirl orbit components. It is possible to analytically represent an
elliptical orbit with two cylindrical orbits of opposing rotation. This method of
representation is described in a little more detail later in chapter 8.

H = H o + ∈1 .H1 + ∈2 .H 2 (Eq. 3.3.9)

where the non-dimensional form for h ( β ) is given by H o ( β ) = 1+ ∈o .cos ( β )

for the plain cylindrical bearing under consideration and the film thickness
components due forward and backward whirl are given by:

H1 ( β ,τ ) = sin ( β + τ + φo − γ ) H 2 ( β ,τ ) = sin ( β − τ + φo − γ ) (Eq. 3.3.10)

3.3.4 Lubricant Temperature Oscillations

Keogh et al [9] considers the perturbation terms ∈1 ,∈2 as small parameters

and linearises the energy balance relationships which then lead to the
temperature perturbed relationships as expressed in equation 3.3.11.

TL = To + ∈1 .T1 + ∈2 .T2
TB = TBo + ∈1 .TB1 + ∈2 .TB 2 (Eq. 3.3.11)
TJ = TJo + ∈1 .TJ 1 + ∈2 .TJ 2

The energy equations for the zero and perturbed orbits are now written :

41
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Zero orbit energy equation :

∂ 2To Pe .η (1 − η ) ∂H o ∂To
2
∂T ∂T
2
− . . − Pe .η .H o2 . o − Pe .H o2 o = − Pe (Eq. 3.3.12)
∂η 2 ∂β ∂η ∂β ∂τ

Perturbed orbit energy equation :

∂ 2T j ∂H o ∂To ∂T ∂T
− Pe .η 2 . (1 − η ) . . − Pe .η .H o2 j − Pe .H o2 j = .....
∂η 2 ∂β ∂η ∂β ∂τ
 ∂ ∂  ∂T ∂T ∂T
Pe .η . (1 − η ) . η . ( H o H j ) + ( 2.η − 1) . ( H o H j )  . o + 2.Pe .η .H o .H j o + 2.Pe .H o .H j . o
 ∂β ∂τ  ∂η ∂β ∂τ
(Eq. 3.3.13)

The following equation can be established to involve the oil film thickness H :

H ( β ,τ )
= Ao ( β ) + ∈1 . A1 ( β ,τ ) + ∈2 . A2 ( β ,τ ) Eq. 3.3.14
d ( β ,τ )

Where:
H0 ( β ) 0≤ β ≤π

Ao ( β ) =  H o ( β )
H π π ≤ β ≤ 2.π
 o( )

0 0≤ β ≤π

A j ( β , τ ) =  H j ( β , τ ) .H o ( π ) − H o ( β ) .H j ( π , τ )
 2
π ≤ β ≤ 2.π
 H o (π )

Therefore the boundary conditions can now be expressed for the zero orbit
and the perturbed orbit by equations 3.3.15 and 3.3.16. The formulation of the
boundary conditions require an understanding of the thermal conditions in the
bearing and journal, and the orbit perturbed temperatures are functions of the
zero orbit temperature.

Zero orbit boundary conditions:

42
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

∂To ∂T
To ( 0, β ,τ ) = TBo (1, β ,τ ) ( 0, β ,τ ) = −γ B . Ao ( β ) . Bo (1, β ,τ )
∂η ∂r
(Eq. 3.3.15)
∂To ∂T
To (1, β ,τ ) = TJo (1,θ ,τ ) (1, β ,τ ) = −γ J . Ao ( β ) . Jo (1,θ ,τ )
∂η ∂r

Perturbed orbit boundary conditions:


T j ( 0, β ,τ ) = TBj (1, β ,τ ) 


∂T j ∂TBj ∂TBo 
( 0, β ,τ ) + γ B . Ao ( β ) . (1, β ,τ ) = −γ B . Aj ( β ,τ ) . (1, β ,τ ) 
∂η ∂r ∂r 

 (Eq. 3.3.16)


T j (1, β ,τ ) = TJj (1,θ ,τ ) 


∂T j ∂TJj ∂TJo 
(1, β ,τ ) + γ J . Ao ( β ) . (1,θ ,τ ) = −γ J . Aj ( β ,τ ) . (1,θ ,τ ) 
∂η ∂r ∂r 

3.3.5 Bearing and Journal Conduction

As the model presented by Keogh et al [9] provides a full thermodynamic

account of the oil film bearing and journal interactions, the relationships
governing this interaction are developed and presented. The equations,
coefficients and boundary conditions developed by Keogh et al [9] for the
conduction terms are presented here without proof in equations 3.3.17 and
3.3.18.

It is shown by Keogh et al [9] that the periodic heat inputs via the lubricant film
for the axially averaged bearing and journal and journal temperatures at the
lubricant interface can be expressed in the following non dimensional form:

∞ ∞
qmn i( nβ + mτ )
TB (1, β ,τ ) = TA + ∑∑ .e
−∞ −∞ hmn
(Eq. 3.3.17)

43
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

∞ ∞
∂TB
(1, β ,τ ) = ∑∑ qmn .ei( nβ + mτ )
∂r −∞ −∞

∞ ∞
pmn i( nθ + mτ )
TJ (1,θ ,τ ) = TA + ∑∑ .e
−∞ −∞ g mn
(Eq. 3.3.18)
∞ ∞
TJ
(1,θ ,τ ) = TA + ∑∑ pmn .ei( nθ + mτ )
∂r −∞ −∞

The terms pmn and g mn are thermal coefficients of the axially averaged
temperatures and heat fluxes which relate the time and special Fourier
components at the lubricant interfaces for the journal and bearing. Keogh et al
[9] states that for typical material, geometric and operating parameters, the
time harmonic coefficients have significantly higher magnitudes than those of
the stationary coefficients (where m=0). These coefficients are used to
implement the boundary conditions on the lubricant at the shaft journal and
bearing interfaces. The appropriate form for the feature of interest, the shaft
journal temperature is given by equation 3.3.19. As the elements which
dictate the steady rotor bend are primarily of interest and it is stated by Keogh
that it is shown that the thermal shaft bend is governed by the magnitude of
the thermal coefficient g mn where m=0 and n=-1.

∞ ∞
pmn i( nθ +( m− n )τ )
TJ (1,θ ,τ ) = TA + ∑∑ .e
−∞ −∞ g mn
(Eq. 3.3.19)
∞ ∞
TJ
(1,θ ,τ ) = ∑∑ pmn .ei( nβ +( m−n )τ )
∂r −∞ −∞

Details of the methods used to determine and solve for pmn and g mn are given
in appendix B of reference [9].

3.3.6 Lubricant Temperature Development

The next step is to consider expressions for the development of the lubricant
around the bearing in both space and time. The expressions below are for
steady state conditions of orbit whirl where the temperatures are seen to
oscillate in time periodically with the shaft rotation, and where transient effects
are ignored. From the assumptions and evaluations previously taken, it can

44
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

be considered that the zero orbit temperatures are steady and the orbit
temperatures vary steadily with time, consisting of fundamental harmonic
components only. The following expressions are used to describe the
temperature conditions for the zero and perturbed orbit components :

To (η , β ,τ ) = TA + So (η , β ) (Eq. 3.3.20)

T j (η , β ,τ ) = S j (η , β ) .eiτ + S j (η , β ) .e − iτ (Eq. 3.3.21)

Equation 3.3.20 requires term So to be determined. Due to the discontinuity in

temperature distribution at the oil inlet, a Fourier series in isolation cannot be
applied to obtain a solution and some account for the ‘step’ change that takes
place at the inlet is required. The discontinuity at the inlet is as a result of the
re-circulating oil within the bearing being refilled with fresh supply oil to
maintain the continuity of flow around the bearing. Keogh et al [9] presents a
form of expression where for a bearing with a single inlet, the discontinuity
can be described by means of two terms Vo and U o . The orbit temperature is
then written :

Vo (η ) .β 2 ∞
So (η , β ) = U o (η ) .β + + ∑ Wok (η ).eik β (Eq. 3.3.22)
2 −∞

As the oil film is being considered in two dimensions some knowledge of the
form of the discontinuity across the oil film is required, which will be governed
by the mixing conditions around the oil inlet. Due to the complexity of this
problem Keogh et al [9] defines expressions for the approximation of Vo and
U o by developing and applying the following conditions and relationships :

So = so = 0 at β = 0
∂So
= 2.π .Vo = 0
∂β
∂ 2 So
= &&
so
∂η 2

so (η = 0 ) = so (η = 1) = 0 Vo (η = 0 ) = Vo (η = 1) = 0

45
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Keogh et al [9] proposes the approximation given by equation 3.3.23 to allow

a solution for the temperature due to the unknown mixing conditions present
at the inlet.

so (η ) = ∆.η 2 . (1 − η ) (Eq. 3.3.23)

&&

Where Vo and so would now be given:

∆.η . (1 − η )
Vo (η ) = 2
(Eq. 3.3.24)
2.π .Pe .H o ( 0 )

 1 η3 η4 
so (η ) = ∆.η .  − + −  (Eq. 3.3.25)
 30 12 20 

The parameter ∆ defines the magnitude of the temperature change at the

inlet due to the mixing of the ‘hot’ re-circulating oil and ‘cool’ supply oil. To do
this a mean cross film temperature is used where the inlet step change is
given by :

1 99
[TM ]β =0 = 2.∫0 η.so (η ).dη = − .∆ (Eq. 3.3.26)
11340

This value of ∆ can by adjusted until the desired pad inlet conditions are
obtained.

Keogh et al [9] applies a refill law at the inlet, where it is assumed the supply
oil is sufficient to replenish the oil flow lost around the bearing, giving equation
3.3.27 which calculates the temperature of the pad oil inlet temperature.

1
TS = .[TM ]β = 0 + TM ( 2.π ) (Eq. 3.3.27)
 H (π ) 
1−  o
 H ( 2.π ) 
 o 

With the procedure and expressions defined for determining the steady orbit
lubricant pad inlet temperature and inlet discontinuity conditions, attention can
now be given the development of the expressions used to solve for the infinite
Fourier series terms W0,k . For the steady zero orbit the application of equation

46
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

3.3.22 to the energy equations for the zero orbit conditions given in equations
3.3.12 and 3.3.15 , without proof, is given by Keogh et al [9] as the ordinary
differential equation having the form :

∞
&& + P .∑ ( b .W&& − ik .η .c .W ) = r
W (Eq. 3.3.28)
0, n e n,− k 0, k n ,− k 0, k n
−∞

Where the boundary conditions for the lubricant-bearing interface (η = 0 ) is

defined as :

 2 & & 
−π .  3 .π .Vo + U o  , n = 0
∞
  
W&0,n + γ B .∑ an ,− k .h0,k .W0, k = (Eq. 3.3.29)
−∞ − V&o + π .V&o + U& o , n ≠ 0
 n 2 i.n

And at the lubricant-journal interface (η = 1) is defined as :

 2 & & 
−π .  3 .π .Vo + U o  , n = 0
∞
  
W&0,n + γ J .∑ an, − k .g k , k .W0,k = (Eq. 3.3.30)
−∞ − V&o + π .V&o + U& o , n ≠ 0
 n 2 i.n

For the non zero orbit temperature solutions the same principle can be
applied the orbit perturbed conditions. In this case however the solution is
considered purely periodic with no discontinuity at the inlet location. As such
the expression can be developed to consider only the fundamental harmonic
components only. Therefore the non zero orbit temperature can be expressed
by the following equation :

∞
S j (η , β ) = ∑ W j ,k (η ) .ei.k .β (Eq. 3.3.31)
−∞

And similarly to the zero obit condition the ordinary differential equation to be
solved for the perturbed condition is given, again without proof, by Keogh et al
[9] as :

47
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

∞
&& + P .∑ ( b .W& − ik .η .c .W − ic .W ) = r
W (Eq. 3.3.32)
j ,n e n ,− k j ,k n ,− k j ,k n ,− k j ,k j ,n
−∞

With the applied boundary conditions at the lubricant interfaces of :

At the lubricant-bearing interface (η = 0 ) :

∞
W& j ,n + γ B .∑ an ,− k .h1,k .W j , k = s j ,n ,η = 0 (Eq. 3.3.33)
−∞

At the lubricant-journal interface (η = 1) :

∞
W& j ,n + γ J .∑ an, − k .g1+ k ,k .W j ,k = t j ,n ,η = 1 (Eq. 3.3.34)
−∞

Expressions for the coefficients s j ,n and t j ,n are presented in appendix A of

[9]. Where they are expressed in terms of the steady state zero orbit
condition. The infinite system of resulting differential equations describing the
lubricant film temperatures is solved by truncation of the infinite series using
Chebyshev collocation techniques [46]. The zero orbit condition is first solved
which then allows the orbit perturbed solutions to be determined, with the
procedures for both the zero and perturbed orbit solutions following the same
procedure. It is stated by Keogh et al [9] that the infinite series were typically
truncated to 10 harmonics.

3.3.7 Prediction of Rotor Bend

The shaft bend can be expressed in terms of bend angles in the referenced
transverse axis of the journal, ψ x ,ψ y as shown in Figure 10. The methods
presented by Dimogronas [3] for the description of shaft bend due to thermal
influences, is applied for the analysis of the shaft bend. Applying this analysis
process, the shaft bend angle is expressed in complex form by:

 π L
2.α iφo + 2  R 2.π 2J
ψ C = ψ x + iψ y = .e .∫ ∫ ∫ Θ.ρ 2 .eiθ .dz.dθ .d ρ (Eq. 3.3.35)
I 0 0 0

48
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Y
Fixed Axis
y
Rotating Axis x

τ
X

Figure 10 Axis About which Rotor Bend Angles are Defined

It can be concluded that the complex bend angle can take the form of
equation 3.3.35. As the procedure involves the integration of the temperature
distribution across the entire journal cross section (assuming an axially
averaged temperature distribution), it can be seen that the only steady bend
conditions in the rotor are produced as a result of the fundamental perturbed
orbit harmonic. Other higher order harmonics will only result in small
osolations during the shaft rotation, due principally to the thermal inertia of the
shaft when running at operating conditions and speeds typical for such
machines. Therefore as the synchronous rotor dynamic response is governed
only by the steady rotor bend, it can be written that the bend angle can be
expressed by equation 3.3.37.

ψ C = ψ C 0 + ∈1 .ψ C1 + ∈2 .ψ C 2 (Eq. 3.3.36)

ψ C =∈1 .ψ C1,0 + ∈2 .ψ C 2,0 + small oscillating terms (Eq. 3.3.37)

Where ψ Cj ,0 = iP0,−1.W j ,−1 (1) .eiφ

o
(Eq. 3.3.38)

49
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

3.3.8 Bend Influence on Rotor Dynamic Systems

To now handle the dynamic response of the shaft, the bend angle ψ C needs
to be translated into dynamic rotor displacement characteristics capable of
being modelled using typical rotordynamic evaluation techniques. Initially the
bend angle must be described and is given by Keogh et al [9] as :

ψ C = BX .z Xb + BY .zYb (Eq. 3.3.39)

Where, when expressed in terms of radial clearance, the dimensionless terms

BX and BY are calculated in terms of the non zero orbit conditions by :

ψ C1,0 +ψ C 2,0
BX =
2
(Eq. 3.3.40)
i (ψ C1,0 + ψ C 2,0 )
BY =
2

The degenerated bend angle ψ C needs now to be incorporated in the

equations of motion governing the rotordynamic system. To do this, using
typical solution methods, the system degree of freedom matrix χ needs to
embody not only the generalised deflections in the two principle axis but also
the components resulting from the bend angle χ B , where the total system
degree of freedom matrix is given by :
χT = χ + χ B (Eq. 3.3.41)

Where if the influence of any external forces are ignored, the equation of
motion for a rotordynamic system can now be expressed by :

M R .χ&&T + ( Cb + Ω.I R ) .χ&T + ( K R + K b ) .χT = K R .χ B (Eq. 3.3.42)

Where M R , I R and K R are the rotor mass, gyroscopic and stiffness matrices
respectively. The bearing performance characteristics are included into the
equation of motion by the Cb and K b matrices which define the bearing
stiffness and damping oil film coefficients.

50
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Keogh et al [9] partitions the degree of freedom vector into the X and Y fixed
coordinate directions where the bend angle is then expressed as :

χ B = Re ( c.ψ Ci .U B .eiΩ.t ) (E1. 3.3.43)

Where U B is of the form :

U 
UB =  
 −iU 

The complex synchronous response Z is then solved by means of the steady

state particular solution of the equation of motion (3.3.42) which satisfies
equation 3.3.44, D is given by equation 3.3.45 and describes the constant
elements of the equation of motion:

χT = Re ( c.Z .eiΩ.t ) (Eq. 3.3.44)

D.Z = ψ Ci .K R .U B (Eq. 3.3.45)

D = −Ω 2 .M R + iΩ. ( Cb + Ω.I R ) + K R + K b (Eq. 3.3.46)

Once the system of linear equations has been solved for the steady state
synchronous response the complex bend angles can be extracted at the
bearing location and the resulting thermal bend as a result of the system
response determined using equation 3.3.46 to obtain z Xb and zYb .

In order to identify some means of qualification of the system sensitivity to

thermal rotor bending, Keogh et al [9] presents a procedure whereby the
resulting thermal bend is determined for an initial static shaft bend and the
gain of the bend development can be expressed by equation 3.3.48.

ψ Ct = BX .z Xb + BY .zYb (Eq. 3.3.47)

ψ Ct
G= (Eq. 3.3.48)
ψ Ci

51
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Keogh et al [9] clarifies that the analysis gives no indication of time dependant
development of the thermal bend due to the steady state assumption for the
thermal conditions within the lubricant for both zero and perturbed forms.
However, it is stated by Keogh et al [9] that where thermal bend development
is considered significantly slower than the dynamic response, Im G > 0 then
rotor vibration will spiral with the direction of rotation and when Im G < 0 the it
will spiral in an opposing direction. With regards to commentary as to a
systems stability Keogh et al [9] summarises that if follows that if Im G = 0
then Re G > 1 would result in unstable development of the thermal bend for
synchronous orbit excitations.

3.4 Balbahadur [19] Procedure

A method of analysis is presented by the work of Balbahadur [19] where a

simplified analytical evaluation is presented. Here a thermal bend
development and stability assessment criteria is developed. The procedure is
principally concerned with developing a technique in which the film thickness
equation is established for both plain and tilting pad journal bearings, where
an energy equation is then used to establish a temperature condition at
specific locations in the bearing. Finally this is used to establish a thermally
developed bend which is compared to a defined stability threshold. The
analytical procedures applied and presented by Balbahadur [19] are reworked
here with the associated derivations where relevant.

3.4.1 Relationships for film thickness

To enable the Reynolds conditions and subsequent thermal conditions within

the bearing to be evaluated, two sets of equations are presented to describe
the oil film thickness around the bearing. The two conditions described by
Balbahadur [19] were for the simple plain cylindrical bearing and the central
pivot tilting pad bearing.

Plain Bearing

The oil film thickness function derived by Balbahadur [19] is consistent with
that presented in most literature for a plain cylindrical bearing with no
misalignment present and is stated (without proof) by equation 3.4.1.

52
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

h = Cr (1 + ε .cos β ) (Eq. 3.4.1)

Figure 11 shows the geometric relationships for film thickness in a plain

cylindrical bearing.

Rb

Rj

Ob β
e
Oj

Figure 11 Plain Cylindrical Bearing Film Thickness Function

e
Where ε =
Cr

Tilting Pad Bearing

The analytical description of the oil film function for the titling pad bearing
requires a slightly more complex derivation and a specifically derived film
function is presented in [19]. Figure 12 shows a schematic diagram of the
geometry relations present in a tilting pad bearing. The figure shows a bearing
with 5 pads but bearings with any practical number of pads can be employed.

53
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Figure 12 Tilting Pad Bearing Geometric Relationships

Y Y Y

J M N

o δ λ
Rb + t p Ob J + h
Rp + t p Rp
Rj
π + θ − θc θp θ
X X X
Ob Ob γ Ob γ
e l r ξ r κ

Oj Op Op

(a) (b) (c)

Geometric relationships from Figure 12

From Figure 12 it is possible to determine :

Rj e OJ
= = b (Eq. 3.4.2)
sin [π − θc + θ ] sin(o) sin(ι )

Solving this equation for o, ι and Ob J gives :

54
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

 e.sin(π − θ c + θ 
o = sin −1   (Eq. 3.4.3)
 Rj 

 e.sin (π − θc + θ ) 
ι = π − [π − θc + θ + o ] = θ c − θ − sin −1   (Eq. 3.4.4)
 Rj 

R j .sin(ι )
Ob J = (Eq. 3.4.5)
sin(π − θ c + θ )

Substituting equation 3.4.4 in to 3.4.5:

Rj   e.sin (π − θc + θ )  
Ob J = .sin π − θc + θ − sin −1   (Eq. 3.4.6)
sin(π − θ c + θ )   Rj  

As the amount of eccentricity can be considered to be small relative to the

bearing journal radius for practical bearings it follows that the sin-1 term will
approximate to be equal to is argument.

 e.sin (π − θ c + θ )   e.sin (π − θ c + θ ) 
sin −1  ≈ 
 Rj   Rj 

Hence
Rj   e.sin (π − θc + θ )  
Ob J = .sin π − θc + θ −   (Eq. 3.4.7)
sin(π − θ c + θ )   Rj  

Applying trigonomic identities and expanding gives :

  e.sin (π − θc + θ )  
sin (π − θc + θ ) .cos   − ........
Rj   Rj  
Ob J = .  (Eq. 3.4.8)
sin(π − θ c + θ )   e.sin (π − θ c + θ )  
.......cos (π − θc + θ ) .sin   
 
 R j  

By the application of small angle approximation for cos and sin terms, the
significant analytical simplification results for equation 3.4.8 gives :

55
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Ob J = R j − e.cos (π − θ c + θ ) (Eq. 3.4.9)

Now applying the sine rule to Figure 11 it can be shown that :

2 2
r= (R p + t p ) + ( Rb + t p ) − 2. ( R p + t p ) . ( R p + t p ) .cos(δ ) (Eq. 3.4.10)

From Figure 11the following relationship can also be established:

Rp r ON
= = b (Eq. 3.4.11)
sin(θ + γ ) sin(λ ) sin(κ )

Solving for λ :

 r.sin (θ + γ ) 
λ = sin −1   (Eq. 3.4.12)
 Rp 

For practical bearings the ratio of r to Rp will be much less than 1 which allows
equation 3.4.12 to be simplified to :

 r.sin (θ + γ ) 
λ≈  (Eq. 3.4. 13)
 Rp 

Equation 3.4.11 is then further rearranged and with the application trigonomic
identities yields:

Rp   r.sin (θ + γ )   r.sin (θ + γ )  
Ob N = . sin (θ + γ ) .cos   + cos (θ + γ ) .sin   
sin (θ + γ )    
 Rp   Rp  
(Eq. 3.4.14)

Again significant simplification is achieved by applying the principle of r<<Rp,

and small angle assumptions resulting in:

Ob N = R p _ r.cos (θ + γ ) (Eq. 3.4.15)

Taking the following relationships from Figure 11

56
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

γ = π − θ p − δ − ξ (Eq. 3.4.16)

r.sin (ξ ) = ( Rb + t p ) .sin (δ ) (Eq. 3.4.17)

Substituting equation 3.4.16 into into equation 3.4.15 yields:

Ob N = R p − r.cos (δ + ξ ) .cos (θ − θ p ) − r.sin (δ + ξ ) .sin (θ − θ p ) (Eq. 3.4.18)

It is then shown that :

2
( )
r.cos (ξ ) = r. 1 − sin 2 (ξ ) = r 2 − ( Rb + t p ) .sin 2 (δ ) (Eq. 3.4.19)

Substituting for r into equation 3.4.19 gives :

r.cos (ξ ) = ( R p + t p ) − ( Rb + t p ) .cos (δ ) (Eq. 3.4.20)

From equations 3.4.17 and 3.4.20 and applying trig identities :

r.cos (δ + ξ ) = r.cos (δ ) .sin (ξ ) = ( R p + t p ) .cos (δ ) − ( Rb + t p ) (Eq. 3.4.21)

r.sin (δ + ξ ) = r.sin (δ ) .cos (ξ ) + r.cos (δ ) .sin (ξ ) = ( R p + t p ) .sin (δ ) (Eq. 3.4.22)

A simplification is again presented where δ (pad angle) is considered to be

very small and as such small angles would yield the following reduced
equation :

r.cos (δ + ξ ) = R p − Rb (Eq. 3.4.23)

and
r.sin (δ + ξ ) = δ . ( R p + t p ) (Eq. 3.4.24)

Substituting equations 3.4.23 and 3.4.24 into equation 3.4.18 :

Ob N = R p − ( R p − Rb ) .cos (θ − θ p ) − δ . ( R p + t p ) .sin (θ − θ p ) (Eq. 3.4.25)

57
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Using Figure 12 it can be seen that film thickness, h, can be expressed as:

h = Ob N − Ob J (Eq. 3.4.26)

The full film thickness function is now assembled by substituting for equations
3.4.9 and 3.4.25:

h = ( R p − R j ) + e.cos (θ c − θ ) − ( R p − Rb ) .cos (θ − θ p ) − δ . ( R p + t p ) .sin (θ − θ p )

(Eq. 3.4.27)

This gives the fluid film function for tilting pad journal bearings as derived by
Balbahadur [19].

A comparison is presented with another published film function algorithm for

tilting pads bearings. The comparison is presented in Figure 13 is for the
method presented by Balbahadur [19] and that of San Andres [47].

1.5 1.9
Non Dimensional Film Thickness (h/Cr)

Non Dimensional Film Thickness (h/Cr)

1.7
1.3
1.5
1.1
1.3

0.9 1.1

0.9
0.7
0.7
0.5
San Andres 0.5 San Andres
Balbahadur Balbahadur
0.3 0.3
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
Angle Around Bearing (deg) Angle Around Bearing (deg)

δ = 0.05, t = 9, ε = 0 δ = 0.05, t = 9, ε = 0.5

1.9
Non Dimensional Film Thickness (h/Cr)

1.7

1.5

1.3

1.1

0.9

0.7

0.5 San Andres

Balbahadur
0.3
0 45 90 135 180 225 270 315 360
Angle Around Bearing (deg)

δ = 0.05, t = 0, ε = 0.5
Figure 13 Tilting Pad Bearing Film Thickness Function Comparison 5 Pad,
Rj=25 mm, Cd=0.125 mm, Pre-load=0.375, Load angle = 90o

58
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

As can be seen from Figure 13 the method derived here as presented by

Balbahadur [19] generally compares well with that of San Andres [47] and
other published data, although some discrepancies are apparent at the pad
leading and trailing edges under certain conditions. Balbahadur [19] generally
results in thinner oil films at the trailing edges of the pad. As can be seen from
the Figure 13, the principle difference between the two methods is the
influence of the account for the pad thickness. When the pad thickness is
neglected the two methods produce the same result.

3.4.2 Journal Position and Perturbation (Solution of Reynolds equation)

To establish the static position of the bearing journal and dynamic translations
of the shaft journal it is necessary to solve the Reynolds equation and
determine the dynamic stiffness and damping bearing characteristics.

3.4.2.1 Plain Bearings

The solution of the Reynolds equation (Eq. 3.2.4), whilst subject to some
transposition, is as presented by Cameron [43] for a plain cylindrical bearing.
It is used to determine the steady state eccentricity ratio and attitude angle for
a given journal load. Detailed derivations are not included as this is presented
in detail by Cameron [43] and is the solution of the Reynolds equation
applying short bearing theory and Ocvick and DuBois [41] boundary theory
applying the half Somerfeld conditions. As such, no specific details of the
derivation are presented here.

Wb .Cb2 π 2 .ε 02 ε 04
= 3
+ 4
(Eq. 3.4.28)
µ .ω.R j .L3 (
16. 1 − ε 02 ) ( 1 − ε 02 )

and for attitude angle :

π . 1 − ε 02
tan (ψ a 0 ) = (Eq. 3.4.29)
4.ε 0

To solve equation 3.4.29 for a specific load, it is necessary to know the

viscosity conditions around the bearing which is a function of the oil film
thickness and hence eccentricity ratio. This interdependency presents a

59
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

problem in the solution. The approach applied by Balbahadur [19] is that of an

effective viscosity which now allows equation 3.4.28 to be solved for ε o .
Balabahadur [19] uses a curve fitted 8th order polynomial describing the right
hand side (RHS) of equation 3.4.28 which is used to determine ε o as a
function of the left hand side (LHS) of equation 3.4.28. The polynomial
solution presented by Balbahadur [19] can be found in appendix A of [19],
although it should be noted that ε o can be solved by using standard iterative
techniques also.

The synchronous perturbation of the shaft journal is described as an elliptical

orbit around the steady state journal centre Oj. Figure 14 shows the
geometrical relationships of the elliptical orbit experienced by the
synchronously orbiting journal. The locus of any point P on the journal surface
is mathematically described by equation 3.4.30. The elliptical orbit geometric
descriptions presented in Figure 14 can be linked to the geometric
relationships used in the oil film thickness equations by equation 3.4.31.

Figure 14 Geometric Relationships of Elliptical Synchronous Orbit

x = e0 .cos (θ j 0 ) + Ax .cos (ω.t + φx ) + R j .cos (ω.t + λ ) (Eq. 3.4.30 – a)

y = e0 .sin (θ j 0 ) + Ay .sin (ω.t + φ y ) + R j .sin (ω.t + λ ) (Eq. 3.4.30 – b)

60
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

ξ = π − (ω.t + λ − θ j ) (Eq. 3.4.31)

Where
 e0 .sin (θ j 0 ) + Ay .sin (ω.t + φ y ) 
θ j = tan −1   (Eq. 3.4.32)
 e0 .cos (θ j 0 ) + Ax .cos (ω.t + φx ) 

3.π
and θ j0 = +ψ a 0 (Eq. 3.4.33)
2

The hot spot on the shaft due to oil film shearing temperatures is defined by
Balbahadur [19] as the point on the journal surface which is closest to the
bearing surface at time (t)=0. Which when applied to Figure 14 yields the
following at t=0:

 Ay .sin (φ y ) 
γ = γ 0 = tan −1  =λ (Eq. 3.4.34)
 Ax .cos (φx ) 
 

Where the assumption is that γ = ω.t + λ which results in the points Oj0, Oj and
point P aligning on a common straight line.

Taking the definition for γ 0 and applying this to equations 3.4.30 (a) and (b)
allows the hot and cold spots to be defined for any point on the elliptical orbit
locus, i.e. t > 0 .

xH = e0 .cos (θ j 0 ) + Ax .cos (ω.t + φx ) + R j .cos (ω.t + γ 0 )

(Eq. 3.4.35)
yH = e0 .sin (θ j 0 ) + Ay .sin (ω.t + φ y ) + R j .sin (ω.t + γ 0 )

xC = e0 .cos (θ j 0 ) + Ax .cos (ω.t + φx ) + R j .cos (ω.t + γ 0 + π )

(Eq. 3.4.36)
yC = e0 .sin (θ j 0 ) + Ay .sin (ω.t + φ y ) + R j .sin (ω.t + γ 0 + π )

In the polar coordinate system of reference the following results :

61
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

ξ H = π − (ω.t + γ 0 − θ j ) (Eq. 3.4.37)

ξC = θ j − ω.t − γ 0 (Eq. 3.4.38)

For the solution of the perturbation equations it is a pre-requisite that the

amplitude and phase terms in the X and Y axis ( AX , AY , φX , φY ) are known.
Balbahadur [19] extracts these parameters by means of a rotordynamic
software package which solves the rotor system under consideration for a
forced response condition by means of the Transfer Matrix method - although
any suitable rotordyanmic analysis software or procedure can be used to
establish this information.
The stiffness and damping coefficients associated with the bearing are
determined by the procedures presented by Vance [48]. The coefficients for
stiffness Kxx, Kxy, Kyx, Kyy and damping Cxx, Cxy, Cyx, Cyy are then computed for
the specific operating condition and used by the computer programme solving
for the mechanical unbalance response characteristics. Details of the
analytical procedures used to determine the stiffness and damping
characteristics of plain journal bearings are presented in several literature
sources and as such details of the derivations are not presented here.

3.4.2.2 Tilting Pad

To solve the more geometrically complex tilting pad bearing accounting for the
introduced angular degrees of freedom, a numerical procedure is necessary
to determine a solution. In the work presented by Balbahadur [19] a finite
element method (FEM) is applied to establish the journal static equilibrium
position within the bearing. As with the plain bearing, a method of solution of
the Reynolds equation is required to determine the pressure distribution, and
based on the calculated pad pressure distribution, the resulting forces and
moment must be resolved to establish a point of equilibrium for each pad and
collectively between all pads.

Balbahadur’s [19] finite element formulation for the pressure distribution

calculations is based around the solution of the reduced Reynolds equation
for two dimensional flow and uses the equation 3.2.3 as presented earlier in
this chapter.

62
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

A brief overview of the analytical model is presented here, however specific

details of the finite element formulations and matrices assemblies is not
repeated here as it does not directly related to the topic of this paper.

Balbahadur [19] solves the Reynolds equation problem by means of a

discreatised domain of the pad surface as shown in Figure 15. The FEM
modelling techniques applied incorporate the use of ‘h’ type linear
quadrilateral elements where mesh refinement was used to obtain acceptable
levels of convergence for the pad pressure distributions.

P=P0 at all 4 Node

boundaries Quadrilateral
Elements

Figure 15 Discreatised FEM mesh of a single tilting pad

After some work the finite element formulation of the Reynolds approximation
is given as:
NPE

∑ KE
j =1
i, j .Pj = FEi (Eq. 3.4.39)

Where KEi , j is the element stiffness matrix and FEi , j is the vector of the
element loads and these are given by :

h3  ∂φi ∂φ j ∂φi ∂φ j 
KEi , j = ∫∫
element
µ
.
 ∂x
.
∂x
+
∂z
.
∂z
 .dA (Eq. 3.4.40)


 ∂h 
FEi = − ∫∫
element
φi . 6.u j .
 ∂x 
.dA (Eq. 3.4.41)

Using equation 3.4.39 the associated global stiffness and force matrices are
assembled using linear basis functions and their associated derivatives. The
boundary conditions are then applied to the global matrices and the system of
equations solved for the nodal pressure values. Balbahadur [19] details a

63
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

computer based programme to conducted the procedure of building and

solving the associated finite element formulations which are included in
appendix B of ref [19]. The boundary conditions applied to the pad by
Balbahadur [19] are that the pressure P, at the four boundaries of the pad
under evaluation, are set equal to the lubricant supply pressure P0.

The converged equilibrium solution is obtained by determining the moment on

each pad based on the approximated pad pressure distributions, and for
equilibrium to be present, the pad moment must be equal to zero. The
resulting forces on the journal are also evaluated. By movement of the shaft it
is possible to establish the forces in the X and Y directions, the resultant of
which has to balance the radial load applied to the bearing Wb . The balanced
solution is achieved for a given set of operating parameters by varying the
individual pad angles δ and the journal eccentricity eo. An iterative technique
based on Newton-Raphson is employed to solve the finite element
formulations to a specified level of force convergence between the X, Y loads
on the bearing journal and the applied radial load.

The method for defining the synchronous shaft orbit uses the same method as
that adopted for the plain bearing problem (I.e. the response is calculated by a
third party rotordynamic package for a given set of bearing film stiffness and
damping coefficients and rotor unbalance condition). Balbahadur determines
the stiffness and damping coefficient by means of the same finite element
∂h
formulation but now also introducing the time dependant term in to the
∂t
Reynolds equation. The bearing pad tilt angle velocity is considered to be
∂δ
= 0 . The stiffness and damping terms are then used to solve the
∂t
rotordyanmic response shaft orbits across the desired speed range for
synchronous unbalance.

3.4.3 Energy Equations and Journal Temperature

A key feature of the thermal instability model is to establish the temperature

distribution around the circumference of the shaft journal. To enable a
quantities assessment to be undertaken the development and application of
some form of energy relationship is required which considers both the oil film
and journal as a minimum.

64
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

A simple energy model is presented by Balbahadur [19] which considers the

energy flow in a plain journal bearing which consists of a control box
encompassing the oil film and a section of the bearing and journal surfaces.
The model assumes steady state conditions within the control box, no axial
heat flow and significantly that the net journal temperature is obtained by
averaging the distribution determined for each dynamic location within the
orbit.

Rotor Ej

Journal ω .R j

Elub e ( x + dx )
h Elub e ( x )
Evisc

Bearing Control
Eb
Volume
dx

Fig 15b Simple Energy Model Presented by Balbahadur [19]

To maintain the conservation of energy and hold true to the steady state
assumption, the following energy equation is presented, where the energy
accumulation rate is zero.

E& visc = ( E& lub( x + dx ) − E& lub( x ) + E& j + E& b ) (Eq. 3.4.41)

Of the energy generated due to the viscose shear stress in an element (dx,dz)
of the oil film for a specific point around the bearing, the energy lost to the
journal and bearing is written by Balbahadur [19] as :

E& j + E& b = f .H .dx.dz. (T − Tamb ) + (1 − f ) .H .dx.dz. (T − Tamb ) (Eq. 3.4.42)

Where H is the heat transfer coefficient for the bearing and journal between
the lubricant oil film, f is a fraction which defines the proportion of energy
that is transferred to the journal with the remainder being dispersed into the
bearing shell, housing and surroundings.

65
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Applying substitutions into equation 3.4.41 and introducing appropriate terms

to describe the heat source in the oil film, the energy present within the oil
flowing into and out of the element dx gives equation 3.4.43 as presented by
Balbahadur [19]:

ω.R j .c1 dT
ω.R j .τ = ρ1.h. . + H . (T − Tamb ) (Eq. 3.4.43)
2 dx

A key assumption to note at this point is that the total energy loss equates to
H .dx.dz. (T − Tamb ) where Tamb = mean ambient journal temperature = mean
ambient bearing housing temperature.

Applying a linear shear stress profile simplification, expressing dx in a polar

domain and simplifying results in the following governing differential equation
as presented by Balbahadur [19] :

2
dT 2.H 2.µ .ω.R j
+ .T − =0 (Eq. 3.4.44)
dξ ρ1.c1.ω.h ρ1.c1.h 2

Where T = (T − Tamb ) and is the temperature rise of the lubricant above

ambient.

To enable the differential equation 3.4.44 to be solved, it is necessary to have

some prior knowledge of the operating viscosity around the bearing.
Balbahadur [19] selects to apply an ISO viscous model where an effective
viscosity approach is applied. The effective viscosity is established by setting
the bearing eccentricity to 0, setting h = cr and calculating the temperature
rise over 180 degrees of the bearing for an oil inlet temperature and viscosity
equal to the lubricant supply conditions. Using this temperature rise an
effective viscosity µ0 is calculated using Reynolds viscosity equation [43]. The
application of the effective viscosity allows for equation 3.4.44 to be solved
numerically for circumferential temperature distribution. The method of
solution of the differential equation used by Balbahadur [19] was to apply a
proprietary ordinary differential equation solver within the Matlab [49] software
programming system.

66
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Tilting Pad Bearings

The energy equation used for the temperature distribution in the tilting pad
bearings uses the same fundamental expressions as defined above for the
plain bearing problem, using equation 3.4.41. However due to the more
complex film function descriptions, it is stated by Balbahadur [19] that a
numerical solution method is required to solve the energy equation.
Balbahadur employs a finite difference (FD) technique to solve the differential
equation across a 1D discreatised grid, using Euler’s method, and presents
the FD solution to the energy equation as:

  2.H    2.H .Tamb 2.µ .ω.R 2j 

Ti +1 = 1 − ∆θ .    .Ti + ∆θ .  + 2  
(Eq. 3.4.45)
  ρ1.c1.ω.h i   ρ1.c1.ω.h ρ1.c1.h i

To provide the required lubricant inlet temperature conditions Balbahadur [19]

assumes a relationship between the lubricant supply temperature, average
temperature rise (estimated) and ratio of oil film thickness for the pad under
consideration. This relationship is defined as :

hmin
Tin = T0 + .∆Tmean (Eq. 3.4.46)
hin
Where ∆Tmean = estimated average temperature rise
The purpose of the above equation is to provide some account for the
principle that higher entry film thickness to a pad will result in lower pad inlet
temperatures.

Different to the plain bearing, a specific relationship is used to determine the

heat transfer coefficient H (into the bearing and journal) in equation 3.4.45 for
the bearing based on work presented by Ettels [50] for the convective energy
loss in variable geometry bearing pads, which is given by Ettels [50] as :

−0.4
H = 25.5.u 0.7 −0.2
j .µ0 . ( R j .∆ p ) (Eq. 3.4.47)

Where R j .∆ p = pad angular length

The effects of lubricant mixing, different heat transfer effects (journal to

bearing) and the convective heat transfer in the bearing pockets were
neglected and not considered in the analysis.

67
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

For both the plain and tilting pad bearing the temperature differential ∆T
across the shaft journal is obtained by straight forward averaging of the
temperature at point P for a number of positions around the shaft orbit. In the
example case provided by Balbahadur [19], ten or more positions around the
orbit for each condition are typically employed.

3.4.4 Thermal Unbalance and Stability Evaluation

With a model applied to determine the temperature conditions present within

the bearing, it is now necessary to develop a relationship to this temperature
distribution and the thermally induced synchronous forcing function and
corresponding stability measure.

Balbahadur [19] uses a simple relationship equating a quasi-static bending

moment to the shaft due to a linear transverse thermal strain. This strain is the
then used to determine the bend angle of the shaft line at the non drive side of
the bearing sleeve. The resulting angle is then used to calculate an
eccentricity at the free end of the overhung shaft and the corresponding
unbalance force is determined from the eccentricity, lumped overhung mass
and operating shaft speed.

Figure 16 Rotor Thermal Bend Model

68
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

The following relationships are established to determine the unbalance force

due to the overhung mass and thermal shaft bend:

Using Figure 16 the bending moment in the shaft can simply be described by:

dψ
M = E.I . (Eq. 3.4.48)
dz
Where the stress is
R j .M
σ max = = E.α .∆T (Eq. 3.4.49)
I

and Balbahadur [19] introduces a relationship between the thermal expansion

and ∆T , where ∆T is as defined in the previous energy equation solutions.
∆T is assumed to be constant across the bearing width.

Substituting the above yields:

dψ α .∆T
= (Eq. 3.4.50)
dz Rj

Where integration with respect to dz and applying boundary conditions 0 and

L gives :

α .∆T .L
ψb = (Eq. 3.4.51)
Rj
dy
Noting that ψ = and further integrating with respect to dz yields the centre
dz
line deflection at L as :

α .∆T .L2
yd = (Eq. 3.4.52)
2.R j
Where yd = yb + ym

Applying the assumption of small angles :

(
yd ≈ Ld − L
2 ) .ψ b (Eq. 3.4.53)

69
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

Combining equations 2.34 and 2.35 :

α .∆T .L.Ld
yd = (Eq. 3.4.54)
Rj

With thermal unbalance being defined as :

U t = md . yd (Eq. 3.4.55)

With a relationship for determining the magnitude of thermally induced

unbalance presented, it is necessary to define a means of establishing some
form of stability criterion.

The method presented by Balbahadur [19] vectorially combines the forces

resulting from the thermally induced bend with the residual mechanical
unbalance loads to produce a net unbalance force. These loads are combined
applying the relative angular phase conditions of the hot spot and the residual
mechanical unbalance. This can be analytically described as:

U = U t2 + U m2 − 2.U t .U m .cos (ω.t − θ CH ) (Eq. 3.4.56)

The method by which Balbahadur [19] assesses the stability of the bearing
rotor system is to define a constant threshold unbalance force which is then
compared to the calculated resultant unbalance force for a given speed and
response condition. If the thermal unbalance force is greater than the
specified threshold the rotor is considered to be thermally unstable.

f thr .W
U thr = (Eq. 3.4.57)
ω2

Where f is a factor relating to a fraction of the rotor weight.

The value proposed by Balbahadur [19] to define the point at which instability
would occur is when the value of f thr = 0.15 , i.e. 15% of the rotor weight. This
level is suggested by Balbahadur [19] based on that it provides reasonable
correlation with the case studies presented in Keogh and Morton [10],
Faulkernker et al [12, 13] and de Jongh et al [11] rotor models within [19]. The

70
Chapter 3 – Analytical Review of Bearing Journal Differential Heating

procedure applied to determine stability is to graphically plot U and U thr for a

specified speed range. Where the plots intersect, the instability threshold point
is determined.

71
Chapter 4 – Benchmark Examples

Chapter 4

Benchmark Examples

4. Benchmark Examples

4.1 Introduction

To enable a suitable test rig design to be produced it is necessary to obtain a

suitable understanding of the coupled rotordynamic and hydrodynamic
conditions present in rotors that have been observed to be susceptible to the
Morton effect. Whilst little data is available for machines known to have
experienced the phenomenon several pieces of published work are available
which have be reported to have been subject to synchronous instabilities
matching the characteristics of the Morton effect, see [11,13,14,31,32,37]. As
it is well known that the rotordynamic conditions of the bearing system play
such a significant part in the sensitivity of a rotor to thermally induced type
instabilities, it has been the approach in this research that each of the cases is
evaluated in detail rotordynamically to provide a summarised account of the
specific conditions and performance conditions under which thermal
instabilities are observed to be prevalent. The deliverable from this element of
work is to provide a qualitative assessment and if possible identify some
quantitative parameters that can be used, to guide as to whether a systems
qualities make it susceptible to bearing induced thermal instability. These
assessments are then taken to provide the principles on which the design
rational for the experimental test rig can be based.

Detailed rotordynamic analysis is performed as part of this research for some

selected published examples and the results and analysis of this work are
presented here.

4.2 Benchmark example cases

The cases selected to provide the benchmark studies were obtained from the
published literature and a single unpublished example. In total three rotor
bearing systems were evaluated, each with subtly different operating and

72
Chapter 4 – Benchmark Examples

physical parameters, but in all cases it was concluded that the instabilities
observed were due to bearing thermally induced rotor bends.

The three cases taken that were studied:

1. deJong and Morton Rotor [11]

2. Faulkner, Strong and Kirk [12,13]
3. High Speed Overhung Compressor (HSOC)

Case studies 1 and 2 are covered in some detail in chapter 2 whereas case
study 3 is an industrial example provided by a large OEM in support of this
research. Case study 1 is considered the most informative benchmark case
due to the greater amount of machine operation data being available, and also
the inclusion of some temperature measurements. Cases 2 and 3 were
industrial machines that were seen to observe instability characteristics
consistent with the Morton effect, where only limited measurements were
taken before practical in service solutions were implemented. The information
available in all three cases is sufficient to enable definition of rotordynamic
models to be created with the requirement for application of some relatively
limited assumptions. The most significant area of assumption is around the
specific detail of some of the bearing profiles and clearances; however the
effects of small variations to these bearing properties can be tested and was
performed in the cases shown here.

In each of these cases some basic rotordynamic data is provided in the

published literature, however, to enable a suitable level of detail to be
evaluated for this research it was considered necessary to conduct a separate
rotordynamic analysis of each of the rotor-bearing systems during the
research. The method of rotordynamic evaluation was conducted using a
number of different systems during the initial research and design phases.
Early work was carried out using a custom developed code using MatLab [49]
applying a relatively simple transfer matrix modelling technique, however,
whilst suitable initial characterisation of the rotordynamic performance was
obtained, a more detailed and flexible rotordynamic approach was required.
The final rotordyanmic analysis work was conducted using a proprietary
package called Xlrotor [51] provided by RMA Inc, and all rotordynamic results
presented within the thesis have been evaluated using this software package.
Xlrotor [51] is a comprehensive rotordynamic evaluation suite which allows all
necessary analytical procedures to be conducted, such as damped

73
Chapter 4 – Benchmark Examples

eigenvalue extraction, rotor mode shapes, unbalance response and the use of
unsymmetrical bearing stiffness-damping matrices.

All of the rotordynamic analysis and results data presented in this chapter
hereafter has been performed, processed, assessed and presented as part of
this research.

4.2.1 Rotordynamic principles applied in XLRotor [51]

Xlrotor [51] is a rotordynamic software package produced by Rotating

Machinery Analysis, Inc (RMA) of the USA. The programme is a set of tools
that are embeded within a proprietary spreadsheet environment, it allows the
undamped and damped natural frequencies of rotor bearing systems to be
computed. The software capability covers many rotordynamic procedures, not
only lateral but also torsional and branched rotordynamic systems.

The features used for the work presented in this thesis are related to the
lateral rotordynamic analysis features only. More specifically the undamaped
natural frequency, damped eigenvalue, unbalance response and rotor
deflected shapes analysis procedures have been employed.

The rotor dynamic model is based on a lumped mass model where the rotor is
split into a number of elements or sections. Each section is represented by an
elastic beam section where the beam mass for that section is split
proportionally between each end of the beam element. The beam formulation
used in the software is the Euler beam analytical model, where for longer
beam sections (with L/D ratios > ~0.6) a correction factor is required to
account for the additional moment inertia that results from the lumped mass
distributions. Additional mass and/or inertias can be applied at each station by
explicitly specifying the magnitudes required at a given station or location.

Bearings or supports are introduced as linear stiffness and damping

components at a given station in the rotor model. As many support locations
as desired can be introduced into the rotordynamic model. The mathematical
model allows for unsymmetrical stiffness matrices and uses 8 specified oil film
linearised stiffness and damping coefficients, i.e. both the direct and cross
coupled coefficients Kxx, Kxy, Kyx, Kyy and Cxx, Cxy, Cyx, Cyy.. The bearing data is
specified for a number of speed conditions and the stiffness and damping data
is curve fitted across the required speed range and then calculated for each
speed step in the solution process.

74
Chapter 4 – Benchmark Examples

The principal theory governing the solutions of the dynamic models are by
application of the governing equations of motion. Where the free-free damped
natural frequency of a rotor-bearing system is governed by equation 4.2.1,
and the unbalance response of the rotor system is similarly determined by
application of the non trivial solution of the forced response relationship
presented in equation 4.2.2.

Principal Equations of Motion

d 2u du
m. 2 + c. + k .u = 0 Eq. 4.2.1
dt dt

Forced harmonic excitation

d 2u du
m. 2
+ c. + k .u = f .cos (ω.t ) Eq. 4.2.2
dt dt

when applied to a discrete rotor system and expressed in matrix form is

written :

M .U&& + ( D + G ) .U& + K .U = F (ω , t ) Eq. 4.2.3

where m = mass M = mass matrix

c = damping C = damping matrix
k = stiffness G=gyroscopic force matrix
u = displacement K = stiffness matrix
ω = angular velocity U = displacement vector
f = excitation force F = time dependant force vector

The stiffness K and damping C matrices are unsymmetrical due to the cross
coupling effects introduced by the journal bearing oil film interactions.

The numerical solution method used to solve for the eigen analysis and
response analysis can be of the transfer matrix or FEM method. The transfer
matrix method provides a quick and efficient solution but is restricted to non
branched systems. The FEM method allows for the solution of branched
systems but requires longer solution times. As the analysis required within the
work presented in this thesis was limited to a single shaft line, then the
Transfer Matrix solution was used exclusively for all of the rotordynamic work
presented. The transfer matrix technique consists of considering the rotor as a

75
Chapter 4 – Benchmark Examples

series of mass/inertia stations and beam/stiffness elements. Each station and

element is represented by a pair of stiffness and mass matrices describing the
analytical relationships for the principle degrees of freedom. The product of
these matrices is then used to form the system matrix which can be solved
iteratively for the associated eigenvalues of the system, where the iteration
variable is rotational velocity ω . Any specific model will have as many eigen
values as there are numbers of stations in the system. The associated eigen
vector for a given eigenvalue allows for the rotor mode shapes to be
expressed for any speed condition.

4.2.2 Case Study 1 – deJong and Morton [11]

In the work presented by de Jongh and Morton [11] two rotors are referenced
which display the spiralling instability phenomenon associated with the Morton
effect (see Figure 18). The second rotor is a slightly smaller rotordynamically
scaled version of the first and it is this second rotor that is modelled because a
more compressive set of data is available for its operation and measurement.
A more detailed account of this example is given in the review of literature in
chapter 2 and [11].

The data for the rotor was taken from that presented in de Jongh et al [11].

It was observed by de Jongh and Morton [11] that the rotor was very sensitive
to unbalance in the overhung section of the rotor and, as part of the
production testing, several overhung masses were evaluated. During the
running of the rotor the vibration and phase data were recorded and plotted
relative to time. It was observed that at a specific running speed of 11000 rpm
the shaft vibration became unstable and steadily rose whilst at steady state
speed and inlet temperature. This characteristic showed a hysteresis effect
during the run up and run down phases. Here the rotordynamic conditions
present during the running range observed are modelled. The following model
looks at the original rotor overhung mass and also the impact of the reduced
shaft end mass.

Figure 19 shows the discretised rotordynamic model used for the analysis of
the rotor-bearing model. As the test compressor used in the work of de Jongh
et al [11] consisted of a number of shrunk on discs, account was taken for
these in the model by considering the additional stiffening effect of these
components and the slight damping properties that such fits offer in flexible

76
Chapter 4 – Benchmark Examples

rotors of this type. By means of comparison evaluation of the undmaped

critical speeds is made with those presented by de Jongh et al [11], Figure 17.
The results of this analysis are presented in Figure 20 by means of a critical
speed map.

Figure 18 de Jongh and Morton [11] Figure 17 Undamped Critical Speed

Compressor Rotors Map for Test Compressor Rotor

DeJong and Motor Scaled Compressor Rotor

0.4

0.3
Shaft Radius, m eters

0.2

0.1
5 10 15 20
25 27
0

-0.1

-0.2

-0.3
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Axial Location, meters

Figure 19 Rotordynamic Model of de Jongh et al [11] Rotor

77
Chapter 4 – Benchmark Examples

Undamped Critical Speed Map

de Jongh and Morton Test Compressor Rotor
20000
Data from de Jongh etal [11]
18000
Calculated
Critical Speed, rpm 16000
14000

12000
10000

8000

6000
4000

2000

0
0.E+00 1.E+07 2.E+07 3.E+07 4.E+07 5.E+07 6.E+07 7.E+07 8.E+07
Effective Support Stiffness, N/m

Figure 20 Calculated Critical Speed Map of de Jongh et al [11] Rotor Including

Comparative Data Presented within the Literature

To compare the rotordynamic model used here to that presented in the paper
by de Jongh et al [11] a response analysis was performed using the support
stiffness data provided by de Jongh et al [11]. This was based on the stiffness
and damping conditions in the ‘XX’ and ‘YY’ planes being identical with no
cross coupling terms. Figure 21 to Figure 23 show the bearing stiffness and
rotordynamic response predictions of de Jongh et al [11] and that of the model
used within this thesis. No bearing damping data was provided in the paper
and values appropriate for this size and type of bearing were calculated and
used in the model. Performed in this research, the response model and
analysis was conducted with a 100 g.mm unbalance mass located at the non
drive end (NDE) flange of the rotor with a 0o phase lag, with the response
evaluated at the NDE bearing centre location.

Figure 21 Unbalance response prediction presented by de Jongh et al [11]

78
Chapter 4 – Benchmark Examples

Bearing Performance Data - Stiffness

7.00E+07

6.00E+07

Bearing Stiffness Coefficient (N/m)

5.00E+07

4.00E+07 Kxx Data

Kxy Data
3.00E+07 Kyx Data
Kyy Data
2.00E+07 Kxx fitted
Kxy_fitted
1.00E+07 Kyx_fitted
Kyy_fitedt
0.00E+00
0 5000 10000 15000 20000 25000

Shaft Rotational Speed (rpm)

Figure 22 Bearing stiffness data used in rotordynamic model of de

Jongh et al [11] rotor

Rotordynam ic Response Plot

DeJong and Morton Test Compressor Rotor
Sta. No. 5: NDE Brg
10 360

9 270
180
8 Major Amp
(based 120 microns Cd)

90
7 Horz Amp
Response, % Cd

0
6 Vert Amp
-90
5 1st critical Horz Phs
-180
Speed Vert Phs
4 -270
4000 rpm
-360
3
-450
2 2nd critical
Speed -540
1 -630
13800 rpm
0 -720
0 5000 10000 15000 20000 25000
Rotor Speed, rpm

Figure 23 Calculated Unbalance Response of de Jongh et al [11] Rotor -

using approximate stiffness coefficients from [11]

From Figure 21 and Figure 23 the results of the initial analysis can be seen to
correlate well with those presented by de Jongh et al [11] for the undamped
critical speed map and simple response evaluation, Figure 20 and Figure 23
respectively (where %Cd is pk-pk vibration as a % of bearing diametral
clearance). The critical speed map frequencies tie up well across the running
speed range with relation to support stiffness, with the 4th vibration mode
presenting the biggest difference with this being approximately 200-300 rpm
higher than that predicted by de Jongh et al [11]. This could in some part be
attributed to the slight differences in mass distribution and account for
additional stiffness presented by the shrunk on discs. However for a model of
this type, the level of correlation is suitable for the purposes of this
characterisation.

79
Chapter 4 – Benchmark Examples

The simple response analysis (Figure 23) also correlates well that presented
in the published work (Figure 21). The location of the natural modes and
general characteristic of the response are well positioned with relation to the
model results presented by de Jongh et al [11]. The model presented by de
Jongh et al [11] gives a slightly more pronounced ‘ramp’ toward the second
bending mode but this aspect would be influenced by the level of damping
used in the support bearings for the analysis. For this there will be some
difference between the two models. A significant feature for the subject of
interest is the phase relationship, and it can be seen that this also correlates
well with the conditions around both the first and second critical speeds
showing the same characteristics and closely matched magnitudes.

In summary, the initial evaluation shows that sufficient correlation exists in this
model for it to be applied for the more detailed characterisation required.

To allow the full rotordynamic characterisation of the test compressor shaft

rotordynamic performance, the actual dynamic bearing oil film stiffness and
damping coefficients need to be predicted. The bearing performance was
evaluated across a range covering the operating range (0-12500 rpm) of
interest. The bearing conditions were solved for 5 operating speeds within the
running range and then curve fitted using a 3rd order polynomial to provide
speed dependant stiffness and damping data for use in the rotordynamic
model. Table 2 gives a summary of the calculated bearing conditions, and
Table 3 and Figure 24 provide the bearing oil film characteristics calculated.

Basic Bearing Geometry and Performance Data

Bearing diameter (mm) 100
Bearing pad length (mm) 50
Number of pads 5
Load direction Load on pad (LOP)
Angle of Pad (degrees) 60 degrees
Diametral clearance (microns) 120
Specific load (kPa) 450
Pivot type Centre pivot
Shaft speed (rpm) 5000 10000 15000 20000
Eccentricity ratio 0.2055 0.1323 0.0856 0.0550
Minimum film thickness (microns) 35 37 36 34
Table 2 Calculated support bearing basic performance data

80
Chapter 4 – Benchmark Examples

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm N/m N/m N/m N/m N.s/m N.s/m N.s/m N.s/m
5000 208617 192 257 238498 219011 0 0 235090
10000 338786 -86 -196 356838 169415 0 0 174363
15000 526039 -104 -241 537587 159286 0 0 161067
20000 813713 -89 -264 821619 160538 0 0 161212
Table 3 Calculated support bearing oil film stiffness and damping coefficients

900000 250000

800000
Kxx Cxx
700000 200000
Kxy Cxy
600000
Stiffness (N/m)

Damping (N.s/m)
Kyx Cyx
500000 Kyy 150000 Cyy
400000 Kxx_fit Cxx_fit
300000 Kxy_fit Cxy_fit
100000
200000 Kyx_fit Cyx_fit
Kyy_fit Cyy_fit
100000
50000
0

-100000
0 5000 10000 15000 20000 25000 0
0 5000 10000 15000 20000 25000
Shaft Speed (rpm)
Shaft Speed (rpm)

Figure 24 Curve fitting of calculated support bearing stiffness and damping

coefficients

Hydrodynamic bearing solutions for the tilting pad bearings were performed
using a custom code ‘Vickers – JPU Tilting Pad Journal Bearing Performance
Program’ V3.21’ [52]. The program is able to evaluate either fully flooded or
direct fed lubricated bearings with central or offset pivots. Performance data
for oil film thickness, shaft eccentricity ratio, operating temperatures and
stiffness and damping characteristics is calculated within the code. The pad
minimum film thickness typically occurs at the point of the trailing edge of the
pads, where the film thickness is a function of both the shaft eccentricity and
pad tilt angle.

The rotordynamic model was then solved to extract the eignvalues for the
rotor with relation to operating speed and the Campbell chart in Figure 25
shows the locations of the predicted natural frequencies. From Figure 25 it
can be seen that the predicted locations for the first and second critical
speeds are at ~4400 rpm and ~12300 rpm respectively. The associated
calculated logarithmic decrements for the first and second critical speeds are
0.673 and 0.227 respectively. The bearing performance data used for the
analysis presented here was based on the lower bearing diametral clearance
of 120 microns.

81
Chapter 4 – Benchmark Examples

Damped Natural Frequency Map - Campbell Diagram

DeJong and Morton Test Compressor Rotor
20000
18000

Natural Frequency (rpm)

16000
14000
12000
10000
8000
6000
4000
2000 1X
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Rotor Speed (rpm)

Figure 25 Calculated forward whirl damped natural frequencies for

de Jongh et al [11] rotor

In order to gain a good understanding and characterisation of the

rotrodynamic conditions within the rotor the modes shapes, response at the
non drive end bearing and operating deflection shapes were extracted. Figure
26 to Figure 32 show the calculated mode shapes and responses. The
response analysis was performed using a mechanical unbalance force vector
of 100 g.mm at a 0 degrees phase angle positioned at the NDE overhung
rotor flange. During the rotor analysis an investigation was performed as to
the sensitivity of the rotor to unbalance elsewhere on the shaft. It was
observed that the response at the NDE bearing was relatively insensitive to
imbalance elsewhere in the centre of the rotor. The effect of this can be seen
in Figure 35 where an imbalance load was applied at the rotor bearing mid-
span (equal to 50% of the API [53] limit) and the response evaluated. The
imbalance mass was rotated into several phase positions to determine not
only the sensitivity to imbalance magnitude elsewhere on the rotor, but also to
the relative phase. The above activity cannot fully establish how the rotor will
respond to the distribution of balance elsewhere on the rotor due to the near
infinite number of combinations possible. It does however show that the
overhung section characteristics do not appear to be particularly sensitive to
these effects, allowing the characteristics analysed to be considered as
relatively general characteristics for the rotor.

As is the case for any practical machines, a certain level of residual unbalance
would also reside elsewhere in the rotor. To provide for this a representation
was included for unbalance on the rotor between the bearing span. The phase
of this unbalance was opposed to that of the overhang unbalance by 180
degrees. The magnitude of this unbalance was 50 g.mm which is based on

82
Chapter 4 – Benchmark Examples

25% of the calculated API [53] limit for compressor rotors of this size and
operating speed.

Damped Eigenvalue Mode Shape Plot

de Jongh and Morton Test Compressor Rotor

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Im(y)
-1 f=3976.6 cpm
-1.5 d=.118 logd
N=4000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

de Jongh and Morton Test Compressor Rotor

forward
backward
f=3976.6 cpm
d=.118 logd
N=4000 rpm

Figure 26 Calculated damped mode shape for first critical speed for de Jongh
et al [11] rotor

Damped Eigenvalue Mode Shape Plot

de Jongh and Morton Test Compressor Rotor

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Im(y)
-1 f=12170.4 cpm
-1.5 d=.2698 logd
N=12000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

de Jong and Morton Test Compressor Rotor

forward
backward
f=12170.4 cpm
d=.2698 logd
N=12000 rpm

Figure 27 Calculated damped mode shape for second critical speed for de
Jongh et al [11] rotor

83
Chapter 4 – Benchmark Examples

Rotordynam ic Response Plot

de Jongh and Morton Test Compressor Rotor
Sta. No. 5: NDE Brg
20 360

18 270
180

(based on 120 microns Cd)

16
90 Major Amp
14
Response, % Cd
0
Horz Amp
12 -90
Vert Amp
10 -180
-270 Horz Phs
8
-360 Vert Phs
6
-450
4
-540
2 -630
0 -720
0 5000 10000 15000 20000
Rotor Speed, rpm

Figure 28 Calculated unbalance response for the de Jongh et al [11] rotor with
100 g.mm unbalance at the NDE overhung mass

Rotordynam ic Response Plot Rotordynam ic Response Plot

de Jong and Morton Test Compressor Rotor de Jong and Morton Test Compressor Rotor
Sta. No. 5: NDE Brg Sta. No. 5: NDE Brg
20 360
20 360
18 270 270
18
180
(based on 120 microns Cd)

180
(based on 120 microns Cd)

16 16
90 90
14 Major Amp
14
Response, % Cd

Response, % Cd

0 0
Horz Amp
12 -90 12 -90
Vert Amp
10 -180 10 -180
-270 -270
Horz Phs
8 8
-360 -360 Vert Phs
6 6
-450 -450
4 4
-540 -540
2 -630 2 -630
0 -720 0 -720
0 5000 10000 15000 20000 25000 0 5000 10000 15000 20000 25000
Rotor Speed, rpm Rotor Speed, rpm

(a) (b)
Figure 29 Calculated unbalance response for the de Jongh et al [11] rotor with
100 g.mm unbalance at the NDE overhung mass and 100 g.mm at the bearing
mid-span at (a) 0o phase lag and (b) 180o phase lag

Rotordynamic Deflected Shape Plot

de Jongh and Morton Test Compressor Rotor
Deflected Shape at 4400 rpm
0.006
Major Amp
0.005
Horz Amp
Response Amplitude, mm pk

0.004
Vert Amp
0.003
0.002
0.001
0
-0.001 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

-0.002
-0.003
-0.004
-0.005
Axial Location, m eters

Figure 30 Calculated rotor deflection shape of de Jongh et al [11] rotor at

location of 1st critical speed

84
Chapter 4 – Benchmark Examples

Rotordynamic Deflected Shape Plot

de Jong and Morton Test Compressor Rotor
Deflected Shape at 12200 rpm
0.08
Major Amp
0.06
Horz Amp

Response Amplitude, mm pk
0.04 Vert Amp

0.02

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
-0.02

-0.04

-0.06

-0.08
Axial Location, m eters

Figure 31 Calculated rotor deflection shape of de Jongh et al [11] rotor at

location of 2nd critical speed of 12200 rpm

Rotordynamic Deflected Shape Plot

de Jong and Morton Test Compressor Rotor
Deflected Shape at 11500 rpm
0.04
Major Amp
0.035
Response Amplitude, mm pk

Horz Amp
0.03
Vert Amp
0.025
0.02

0.015

0.01

0.005

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
-0.005
Axial Location, m eters

Figure 32 Calculated rotor deflection shape of de Jongh et al [11] rotor at

observed thermal stability threshold speed of 11500 rpm

As a further check to the correlation of model predictions of the actual rotor

performance when under test, the response of the rotor due to unbalance
located at the rotor centre and overhung mass were performed. These results
can then be directly compared for characteristics with the physical
measurements presented by de Jongh et al [11].

85
Chapter 4 – Benchmark Examples

Calculated Shaft Orbit Calculated Shaft Orbit

2.5 5

2 4

Vertical Displacement (microns)

3
Vertical Displacement (microns)
1.5

1 2

0.5 1

0 0

-0.5 -1

-1 -2

-1.5 -3

-2 -4

-2.5 -5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -5 -4 -3 -2 -1 0 1 2 3 4 5

Horizontal Displacem ent (microns) Horizontal Displacement (m icrons)

(a) (b)
Figure 33 Predicted shaft whirl orbits at NDE bearing of de Jongh et al [11]
rotor at (a) 4500 rpm – location of 1st critical and (b) 11500 rpm – thermal
instability threshold speed

The specific magnitudes cannot be readily compared due the particular

amounts of introduced unbalance not being available via the literature [11],
but as the authors analysis is based on a linear response model, the
characteristics can be taken as representative for any scalable unbalance
condition where the relative proportions or phase of unbalance are nominally
maintained and linear bearing response characteristics remain valid. This can
be a significant assumption where larger vibration amplitudes are present.

Figure 34 Measured 1X response data for test compressor rotor with

introduced centre unbalance presented by de Jongh et al [11] – figure taken
from [11]

An additional test was performed by de Jongh et al [11] where an imbalance

load was introduced into the centre of the test compressor rotor. The vibration

86
Chapter 4 – Benchmark Examples

response of which are given in Figure 34. The purpose of the evaluation was
to observe if any significant temperature differential or instability was
observed. It was observed [11] that an appreciable temperature differential
could be identified when crossing the critical speed but the rotor remained
stable at all times.

To capture this condition in the rotordynamic work performed by the author, a

case which reflects this was evaluated. The results of this can be seen in
Figure 35. It was identified that to obtain the level of response observed in [11]
that a significant unbalance load was required (many times that of the API [53]
limit). It is the opinion of the author that this would be to ensure a sufficient
level of vibration takes place at the NDE bearing this close to a nodal point for
the 1st critical mode to generate conditions where a measurable temperature
difference would be generated. This principle is also supported by the fact the
rotor was only run up to approximately 60% of full speed as the vibration
levels beyond this would be unacceptable as a result of the much higher
imbalance loads.

Comparing the results given in Figure 35 to those presented by de Jongh et al

[11] in Figure 34 shows good correlation of the critical speed and the general
response characteristic is maintained. One notable observation however, is
that the quality factor Q, see Blake [67], between the two responses is much
lower on that of the calculated data. The appearance of the response in
Figure 35 would suggest that more damping is present with the actual test
rotor than is being predicted in the author’s rotordynamic model. Possible
sources for this discrepancy are small variations on the bearing performance,
but a more probable source would be the additional damping introduced by
the rotor structure itself and that of the contribution of the main bearing
support housings for which damping is not considered within the mathematical
model.

The work conducted on the de Jongh and Morton [11] rotor has shown that
the rotordynamic models appear to satisfactorily predict the rotordynamic
conditions observed within the super critical rotors, with the fundamental
rotordynamic characteristics comparing well with the presented data. Further
in-depth analysis of the model then allows for the more detailed understanding
of the rotor conditions during running and instability phases to support in the
design of the test rotor for the research of this thesis.

87
Chapter 4 – Benchmark Examples

Rotordynam ic Response Plot

de Jong and Morton Test Compressor Rotor
Sta. No. 4: NDE Brg
30 360
270
25 180

(based on 120 microns Cd)

90 Major Amp
Response, % Cd
20 0
Horz Amp
-90
Vert Amp
15 -180
-270 Horz Phs
10 -360 Vert Phs
-450
5 -540
-630
0 -720
0 2000 4000 6000 8000 10000
Rotor Speed, rpm

Figure 35 Calculated unbalance response of de Jongh et al [11] rotor where

centre only imbalance introduced

The same modelling principles are now applied by the author to the two
further case studies to extract the same rotordynamic characteristic
information.

4.2.3 Case Study 2 – Faulkner, Strong and Kirk [12,13]

The rotor used in this case study is based on several papers presented by
Faulkner, Kirk and Strong [12,13]. Here a high speed overhung turbocharger
was seen to display unstable operating vibrations when running at constant
speed. Following an initial design modification due to a loose impellor rotor
when at speed, increasing vibrations were observed with time at constant
speed with an associated continually shifting phase. The machine was
observed to show signs of instability when a running speed of 9800 rpm was
reached. Some operational vibration measurements were taken and on
exhausting other root causes thermal synchronous instability ‘Morton Effect’
was attributed. The machine was fitted with modified bearings of the pocket
type which provided increased eccentricity in the bearing and resulted in the
machine no longer demonstrating the phenomenon. This enabled full
operating speed to be achieved indefinitely. The following analysis looks at
the rotor and bearing conditions prior to the introduction of the pocket bearing
assemblies.

An undamped critical speed map has been produced and this used as a
measure of the rotor response in line with the data presented for turbocharger
[12,13]. The analysis shows that three critical speeds are present below

88
Chapter 4 – Benchmark Examples

40,000 rpm as is presented in Figure 37, with the associated rotordynamic

model compiled for the analysis shown in Figure 38. The undamped critical
speed map determined by Faulkner et al [12,13] is shown in Figure 36.

Figure 36 Undamped critical speed map for Turbocharger rotor presented by

Faulkner, Strong and Kirk [12,13] – figure taken from [12,13]

Undamped Critical Speed Map

Falkner, Strong and Kirk Turbcharger
100000
Critical Speed, rpm

10000

1000
1.E+06 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Bearing Stiffness, N/m

Figure 37 Calculated undamped critical speed map of Faulkner et al [12,13]

rotor

89
Chapter 4 – Benchmark Examples

Faulkner, Strong and Kirk Turbocharger Rotor

0.3

0.2

Shaft Radius, meters

0.1
10
5 15 20 25

-0.1

-0.2

-0.3
0 0.2 0.4 0.6 0.8 1
Axial Location, meters

Figure 38 Rotordynamic model of Faulkner et al [12,13] rotor

The bearing types used for the analysis are the 3 lobe type which is consistent
with those present within the machine prior to the bearing modifications
conducted at the turbine bearing end [12]. The bearing performance data was
calculated across the running speed range and the stiffness and damping
data used for the damped eigenvalue extraction and response analysis.

The mode shapes at the rotor bearing assembly critical speeds were
extracted for a running speed up to 16000 rpm. The first two modes are
strongly linked to the bearing stiffness characteristics and can be
characterised as the rocking and bounce modes respectively. The second
‘bounce’ mode does however include some component of rotor shaft bending
associated with the overhung masses of the rotor, and this point has some
significance with relation to the process of shaft thermal bend development.
The third mode is principally a 1st shaft flexural bending mode and sits
approximately 4000 rpm above the max running speed of the rotor. Figure 39
a, b and c show the rotor mode shapes for the first, second and third damped
natural frequencies respectively.

90
Chapter 4 – Benchmark Examples

Undamped C.S. Mode Shape Plot Undamped C.S. Mode Shape Plot
Faulkner, Strong and Kirk Turbocharger Rotor Faulkner, Strong and Kirk Turbocharger Rotor

1.5 1.5

1 1
Re(x) Re(x)
0.5 0.5
Im(x) Im(x)
0 0
Re(y) Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2 1.4 -0.5 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Im(y)
Im(y)
-1 -1 f=6567.9 cpm
f=5329.1 cpm
-1.5 K=52556550 N/m -1.5 K=52556550 N/m

Axial Location, meters Axial Location, m eters

Undamped C.S. Mode Shape Plot Undamped C.S. Mode Shape Plot
Faulkner, Strong and Kirk Turbocharger Rotor Faulkner, Strong and Kirk Turbocharger Rotor

forward forward
backward backward
f=5329.1 cpm f=6567.9 cpm
K=52556550 N/m K=52556550 N/m

(a) Bearing associated ‘rocking’ (b) Bearing and shaft associated

mode ‘bounce’ mode
Undamped C.S. Mode Shape Plot
Faulkner, Strong and Kirk Turbocharger Rotor

1.5
1
Re(x)
0.5
Im(x)
0
Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Im(y)
-1 f=16448.1 cpm
-1.5 K=52556550 N/m
Axial Location, m eters

Undamped C.S. Mode Shape Plot

Faulkner, Strong and Kirk Turbocharger Rotor

forward
backward
f=16448.1 cpm
K=52556550 N/m

(c) Shaft 1st flexural bending mode

Figure 39 Calculated damped modes shapes for Faulkner et al [12,13] Turbocharger

rotor

The rotor response analysis was performed by the author through introducing
an unbalance level of 350 g.mm at the turbine rotor shaft end. The magnitude
of unbalance is based on 0.5 oz.in as presented by Faulkner et al [12,13]. The
response data is presented in Figure 40, Figure 41 and Figure 42 for the
compressor bearing, turbine bearing and turbine disc locations respectively.
The deflected rotor shapes have been extracted for the locations of the

91
Chapter 4 – Benchmark Examples

second critical speed and the instability running speed of ~10000 rpm. These
calculated operating deflection shapes are provided in Figure 44 and Figure
45.

Rotordynam ic Response Plot

Faulkner, Strong and Kirk Turbocharger Rotor
Sta. No. 6: Comp Brg
30 360
270
25 180
Response, microns pk-pk

90
Major Amp
20 0
Horz Amp
-90
Vert Amp
15 -180
-270 Horz Phs
10 -360 Vert Phs
-450
5 -540
-630
0 -720
0 5000 10000 15000 20000
Rotor Speed, rpm

Figure 40 Calculated unbalance response for Faulkner et al [12,13] rotor

compressor bearing

Rotordynam ic Response Plot

Faulkner, Strong and Kirk Turbocharger Rotor
Sta. No. 16: Turb Brg
40 360
270
35
180
Response, microns pk-pk

30 90 Major Amp
0
25 Horz Amp
-90
Vert Amp
20 -180
-270 Horz Phs
15
-360 Vert Phs
10 -450
-540
5
-630
0 -720
0 5000 10000 15000 20000
Rotor Speed, rpm

Figure 41 Calculated unbalance response for Faulkner et al [12,13] rotor

turbine bearing

92
Chapter 4 – Benchmark Examples

Rotordynam ic Response Plot

Faulkner, Strong and Kirk Turbocharger Rotor
Sta. No. 21: Turb
70 360
270
60 180

Response, microns pk-pk

90 Major Amp
50
0
Horz Amp
40 -90
Vert Amp
-180
30 -270 Horz Phs
-360 Vert Phs
20
-450

10 -540
-630
0 -720
0 5000 10000 15000 20000
Rotor Speed, rpm

Figure 42 Calculated unbalance response for Faulkner et al [12,13] rotor

turbine disc

Figure 43 Predicted rotor response for turbocharger rotor presented by

Faulkner, Strong and Kirk [12,13]

Rotordynamic Deflected Shape Plot

Faulkner, Strong and Kirk Turbocharger Rotor
Deflected Shape at 6200 rpm
0.03

Major Amp
Response Amplitude, mm pk

0.02
Horz Amp
0.01 Vert Amp

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
-0.01

-0.02

-0.03
Axial Location, m eters

Figure 44 Calculated operating rotor response for Faulkner et al [12,13] rotor

93
Chapter 4 – Benchmark Examples

Rotordynamic Deflected Shape Plot

Faulkner, Strong and Kirk Turbocharger Rotor
Deflected Shape at 9500 rpm
0.02

Major Amp
Response Amplitude, mm pk
0.015
Horz Amp
0.01 Vert Amp

0.005

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
-0.005

-0.01
Axial Location, m eters

Figure 45 Calculated operating rotor response for Faulkner et al [12,13] rotor

The predicted shaft obits at the machine running speed of 9500 rpm are
shown in Figure 46 (a) and (b) for the compressor and turbine end bearing
respectively.

Calculated Shaft Orbit Calculated Shaft Orbit

4 10

8
3
Vertical Displacement (microns)

Vertical Displacement (microns)

6
2
4
1 2

0 0

-2
-1
-4
-2
-6
-3 -8

-4 -10
-4 -3 -2 -1 0 1 2 3 4 -10 -8 -6 -4 -2 0 2 4 6 8 10
Horizontal Displacement (m icrons) Horizontal Displacem ent (m icrons)

(a) Compressor bearing (b) Turbine bearing

Figure 46 Predicted bearing whirl orbits for Faulkner et al [12,13]

rotor

4.2.4 Case Study 3 – High Speed Overhung Compressor (HSOC)

The final case study is based on an unpublished example where a high speed
single impellor compressor rotor was seen to experience high and short term
unstable vibrations whilst undergoing performance evaluation on the test bed.
The rotor was designed and manufactured by a large multinational Original

94
Chapter 4 – Benchmark Examples

Equipment Manufacturer (OEM) with extensive experience in the design,

manufacture and operation of high speed rotating machinery. The rotor design
is a radial flow CF high speed compressor driven by a gas turbine for use in
the Oil and Gas industry. The bearing assembly used to support the rotor was
by means of two 5 pad tilting pad journal bearings. The bearings at the Drive
End (DE) and Non-Drive End (NDE) are of different diameters and length, with
the larger diameter, longer bearing being present at the non drive compressor
rotor end.

Whilst undergoing testing the machine was seen to operate with typical
vibration response characteristics within normal levels when running at normal
operating temperatures. However, when the machine underwent a cold start
up, a condition of high vibration was experienced when run up to full speed
(~14250 rpm). The vibration level was seen to increase slowly with time and
when the unit was slowed back down to its steady running speed (11695
rpm), the vibration was seen to remain at a high level and then slowly reduced
back to its ‘normal’ levels. Associated with the changing vibration was the
characteristic change in phase observed when a thermal stability event is
taking place. The phase shift in this case was not particularly significant
compared to other examples, but is clearly present. The lower phase shift
could be related to the shorter time constants involved when running at the
maximum speed before slowing back down to the steady running speed. The
characteristic hysteresis effect can be seen in the Bode plot shown in Figure
48.

The high vibration levels were observed in only one of the two support
bearings. The high vibrations were seen at the DE bearing with levels at the
NDE being significantly lower and not displaying the same steady increase
with vibration and only very small movement in phase taking place. A
significant observation of this particular case is that the unstable vibration is
observed at the bearing location opposite to the end where the large shaft
overhung section is positioned, i.e. the DE bearing (Figure 47). This
significance is that in previous examples, the unstable vibration characteristic
has been principally associated with the bearing located at the same end as
the overhung section or, in the case of overhung rotors both bearings see
increased vibration.

Following some initial investigative work by the OEM the possibility of thermal
instability was considered and a basic evaluation of the rotor sensitivity to this
issue was investigated by Dr G Kirk of Virginia Tech University, USA. Based

95
Chapter 4 – Benchmark Examples

on this work it was concluded that the most probable cause was due to
differential rotor heating. Also the conditions under which the vibration was
observed was as a result of the higher viscous shearing effects seen when
operating with the bearing pad lower temperature inlet oil at higher viscosity,
this is present when the machine is starting up before full thermal soak has
occurred in the rotor and housings.
HSOC Rotor

0.3

0.2
Shaft Radius, meters

0.1
10
5 15 18

-0.1
Coupling
Impeller NDE DE
-0.2
Bearing Bearing
-0.3
0 0.2 0.4 0.6 0.8 1
Axial Location, meters

Figure 47 Rotor dynamic model of HSOC compressor rotor

The solution to the presence of the problem was a modification to the bearing
by the introduction of increased bearing clearance which resulted in a slight
increase in oil film thickness and reduction in circumferential viscous shear
rate. The qualitative assumption of this is to result in a lower temperature
differential being generated at the bearing journal location and reducing the
thermal bend feedback into the unbalance cycle.

Figure 48 Measured shaft vibration for (a) DE bearing location and (b) NDE
bearing location of HSOC compressor rotor

96
Chapter 4 – Benchmark Examples

As with the previous two case studies, the author has evaluated the rotor to
obtain an understanding of the rotordynamic and hydrodynamic
characteristics present across the running range. In addition to the principle
rotordynamic evaluation a brief evaluation of the effect of the bearing
characteristics on the rotordynamic performance has been included so as to
understand the difference in conditions when running at lower bearing pad
lubricant inlet viscosities and how this could influence the conditions at the
rotor with respect to thermal bend stability sensitivity.

Undamped Critical Speed Map

HSOC Rotor

20000
18000
Critical Speed, cpm

16000
14000
12000
10000
8000
6000
4000
Mode 1
2000
Mode 2
0
1.00E+06 1.00E+07 1.00E+08 1.00E+09 1.00E+10

Bearing Stiffness, N/m

Figure 49 Calculated undmaped critical speed map for HSOC compressor

rotor

The undamped critical speed map was calculated as is shown in Figure 49.
Within the speed range of interest, two natural modes were identified and both
these showed significant sensitivity to support stiffness with reference to the
mode frequency location. There would also be a notable change in associated
mode shape within the support stiffness range considered.

To allow the full rotordynamic characterisation of the HSOC compressor shaft,

the dynamic bearing oil film stiffness and damping coefficients were
calculated. The bearing performance was evaluated across the operating
range of interest and split into 4 operating speed sets. As described previously
the data was curve fitted to yield speed dependant stiffness and damping data
for use in the rotordynamic model. Table 4 and Table 5 summarise the
bearing oil film characteristics calculated. It was found when calculating the
critical speeds of the rotor, that to match the observed dynamic response it
was necessary to use a DE support bearing with lower support stiffness. This
was incorporated by using a bearing clearance which would be at the top
clearance for a bearing of this size.

97
Chapter 4 – Benchmark Examples

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm N/m N/m N/m N/m N.s/m N.s/m N.s/m N.s/m
5000 366720000 123972 -328728 371823000 154557 63403 -63662 156486
10000 597312000 -19671 -51126 600675000 120520 47555 -47590 121190
15000 885927000 -52809 -83859 888141000 112390 42136 -42145 112699
20000 1272729000 -58026 -86625 1274235000 111145 38768 -38767 111311

Table 4 Bearing oil film coefficients across speed range for NDE bearing

Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy

rpm N/m N/m N/m N/m N.s/m N.s/m N.s/m N.s/m
5000 6204400 919.3 -3945.4 6233500 77415 33578 -33627 77946
10000 10005700 1675.2 -4028.3 10014500 60789 25555 -25568 60945
15000 13477200 -472.8 -344 13492200 52527 21347 -21345 52634
20000 18363000 -748 -555 18373400 51232 19889 -19887 51297

Table 5 Bearing oil film coefficients across speed range for DE bearing

Resolving the model with the calculated bearing support stiffness and
damping characteristics the damped natural frequencies shown in Figure 50
were obtained. The Campbell diagram shown in Figure 50 suggests that two
critical speeds will be experienced within the running range. However in
practice the lower red curve in Figure 50 is associated with a reverse
precession whirl mode and under typical operating conditions this would not
be excited by simple unbalance response. The results predict that the rotor
will experience a critical speed at 14000 rpm. The mode shape associated
with this point is given in Figure 51. It can be seen that the mode is a
combination of bearing and shaft governed elements, with significant rotor
deflection taking place in the area of the NDE overhung section with relatively
low bearing activity. At the DE the displacement at the bearing is
comparatively high but there is little rotor deflection taking place at this point of
the shaft.

Damped Natural Frequency Map - Campbell Diagram

HSOC Rotor

20000

17500

15000
Natural Frequency (rpm)

12500

10000
Mode 1 (forward)
7500
Mode 2 (forward)
5000 Mode 3 (reverse)
1X
2500

0
0 2500 5000 7500 10000 12500 15000 17500 20000

Rotor Speed (rpm)

Figure 50 Calculated damped natural frequency map for HSOC compressor

rotor

98
Chapter 4 – Benchmark Examples

Damped Eigenvalue Mode Shape Plot

HSOC Rotor

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Im(y)
-1 f=13941.4 cpm
-1.5 d=1.1841 logd
N=14000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Of fset Half Bearings - Overhung Mass

forward
backward
f=13941.4 cpm
d=1.1841 logd
N=14000 rpm

Figure 51 Calculated damped mode shape for natural frequency

at 13941 cpm

The response of the rotor was evaluated by introducing an imbalance force

equating to 150 g.mm (0o phase) located at the impeller rotor location. Figure
52 and Figure 53 show the response of the rotor at the NDE and DE bearing
locations respectively. It can be seen that the general characteristics of the
response between ~5000 rpm and the full speed ~1400 rpm compare well to
the measured data shown in Figure 48. The response in the lower frequency
range however does not agree quite so well. It is unclear as to the cause of
the relatively high vibration observed in the lower frequency range as it does
not show any common traits with typical rotor dynamic issues, here the level
of vibration is highest at very low speeds where dynamic balance forces are at
their lowest. Also it can be seen that there is also a significant amount of non
synchronous vibration activity in this area which can be seen from the dashed
line in Figure 48. It is not uncommon to observe these types of characteristics
in rotating machines where other external effects can give rise to observed
moderate vibration levels at lower running speeds where, in this case,
aerodynamic effects, coupling loads or even shaft measurement glitch can
influence the actual or perceived vibration levels. For the purposes of this
evaluation little focus is placed on this apparent discrepancy between the
predictive tool and observed vibration as it is not considered that these
aspects are significant to the thermal problem in the rotor, due to the lower
levels of heat generation in the bearings at these speeds.

99
Chapter 4 – Benchmark Examples

The NDE bearing shows significantly lower response than the DE bearing by
a factor of ~7 which is consistent the observed measurements, these show an
~5 times increase at the DE bearing relative to the NDE. The impact on
rotordynamic performance of a higher lubricant viscosity in the bearings is
also considered in an attempt to understand if the thermally induced effect is
driven by the rotordynamic as well as the increased oil shear heat flux when
running with higher oil viscosity. The impact of the higher viscosity
rotordynamically is to result in nominally higher stiffness and damping values,
which in this case is a factor of some 50% when at ~12000 rpm. Figure 54
and Figure 55 present the response results for the higher viscosity condition.
It can be observed that predicted impact of the higher viscosity on the
rotordynamic performance conditions is limited. A very slight reduction in
amplitude is seen at the NDE bearing and there is a shift to bring the peak
critical speed down closer to running speed. Conversely the amplitude at the
DE bearing reduces by a slightly higher amount by ~25% even though the
critical speed has been pulled down closer to the running speed.

Rotordynam ic Response Plot

HSOC Rotor
Sta. No. 8: NDE Brg
50 360

45 270
180
40
90 Major Amp
Response, mm p-p

35
0
Horz Amp
30 -90
Vert Amp
25 -180
-270 Horz Phs
20
-360 Vert Phs
15
-450
10 -540
5 -630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 52 Calculated response for NDE bearing location of HSOC compressor

rotor (normal viscosity)

100
Chapter 4 – Benchmark Examples

Rotordynam ic Response Plot

HSOC Rotor
Sta. No. 14: DE Brg
70 360
270
60 180

Response, microns pk-pk

90
50 Major Amp
0
Horz Amp
40 -90
Vert Amp
-180
30 -270 Horz Phs
-360 Vert Phs
20
-450
-540
10
-630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 53 Calculated response for DE bearing location of HSOC compressor

rotor (normal viscosity)
Rotordynam ic Response Plot
HSOC Rotor
Sta. No. 8: NDE Brg
50 360

45 270
180
40
90
Major Amp
Response, mm p-p

35
0
30 -90
Horz Amp

25 -180 Vert Amp

20 -270 Horz Phs

-360 Vert Phs
15
-450
10
-540
5 -630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 54 Calculated response for NDE bearing location of HSOC compressor

rotor (higher viscosity)
Rotordynam ic Response Plot
HSOC Rotor
Sta. No. 14: DE Brg
70 360
270
60 180
90 Major Amp
50
Response, mm p-p

0
Horz Amp
40 -90
Vert Amp
-180
30 -270 Horz Phs
-360 Vert Phs
20
-450
-540
10
-630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 55 Calculated response for DE bearing location of HSOC compressor

rotor (higher viscosity)

101
Chapter 4 – Benchmark Examples

The associated rotor deflection shapes at the steady running speed of 11695
rpm and the maximum running speed of 13400 rpm are given in Figure 56 and
Figure 57. The deflected rotor shape is also provided for the high inlet
viscosity condition when running at 11695 rpm in Figure 58. Predicted shaft
whirl orbits are shown in Figure 59 and Figure 60 for the NDE and DE
bearings. The observed shaft whirl orbits from the measured HSOC
compressor rotor are shown in Figure 61. These have been re-synthesised
from the synchronous amplitude and phase data presented in the HSOC test
data. A notable difference in the calculated vs measured shaft orbits is the
highly circular form of the predicted orbits, whereas the DE bearing for the
measured data displays a significant ellipse.

Rotordynamic Deflected Shape Plot

HSOC Rotor
Deflected Shape at 11695 rpm
0.025
Response Amplitude, mm pk

0.02

0.015 Major Amp

Horz Amp

0.01 Vert Amp

0.005

0
0 0.2 0.4 0.6 0.8 1
Axial Location, m eters

Figure 56 Calculated operating deflection shape for HSOC compressor rotor

at endurance running speed of 11695 rpm

Rotordynamic Deflected Shape Plot

HSOC Rotor
Deflected Shape at 13400 rpm
0.035

0.03
Response Amplitude, mm pk

0.025
Major Amp
0.02
Horz Amp
0.015 Vert Amp

0.01

0.005

0
0 0.2 0.4 0.6 0.8 1
Axial Location, m eters

Figure 57 Calculated operating deflection shape for HSOC compressor rotor

at maximum running speed of 13400 rpm

102
Chapter 4 – Benchmark Examples

Rotordynamic Deflected Shape Plot

HSOC Rotor
Deflected Shape at 11695 rpm
0.03

Response Amplitude, mm pk
0.025

0.02
Major Amp
0.015 Horz Amp
Vert Amp
0.01

0.005

0
0 0.2 0.4 0.6 0.8 1
Axial Location, meters

Figure 58 Calculated operating deflection shape for HSOC compressor rotor

at endurance running speed with low bearing pad inlet viscosity

Calculated Shaft Orbit Calculated Shaft Orbit

4 16
3.5 14
3 12
Vertical Displacement (microns)
Vertical Displacement (microns)

2.5 10
2 8
1.5 6
1 4
0.5 2
0 0
-0.5 -2
-1 -4
-1.5 -6
-2 -8
-2.5 -10
-3 -12
-3.5 -14
-4 -16
-4 -3 -3 -2 -2 -1 -1 -0 0 0. 1 1. 2 2. 3 3. 4 -1 -1 -1 -1 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
.5 .5 .5 .5 5 5 5 5 6 4 2 0

Horizontal Displacem ent (microns) Horizontal Displacem ent (microns)

Figure 59 Predicted whirl orbits for HSOC compressor rotor for

NDE and DE end bearings rotating at endurance running speed
11695 rpm

Calculated Shaft Orbit Calculated Shaft Orbit

5 20

4 16
Vertical Displacement (microns)
Vertical Displacement (microns)

3 12

2 8

1 4

0 0

-1 -4

-2 -8

-3 -12

-4 -16

-5 -20
-5 -4 -3 -2 -1 0 1 2 3 4 5 -20 -16 -12 -8 -4 0 4 8 12 16 20

Horizontal Displacement (m icrons) Horizontal Displacement (m icrons)

Figure 60 Predicted whirl orbits for HSOC compressor rotor for

NDE and DE end bearings rotating at maximum running speed
13400 rpm

103
Chapter 4 – Benchmark Examples

Calculated Shaft Orbit Calculated Shaft Orbit

80 30

60
20

Vertical Displacement (microns)

Vertical Displacement (microns)

40
10
20

0 0

-20
-10
-40
-20
-60

-80 -30
-80 -60 -40 -20 0 20 40 60 80 -30 -20 -10 0 10 20 30
Horizontal Displacem ent (m icrons) Horizontal Displacem ent (m icrons)

Figure 61 Generated shaft whirl orbits from measured HSOC

compressor rotor test data for NDE and DE bearing locations
running at 13400 rpm

4.3 Discussion

Taking the author’s rotordynamic work presented here in combination with

physical data presented in the literature, the principle observations, conditions
and relationships are now briefly discussed. The author concludes with a short
summary of the principal rotordynamic and bearing performance
characteristics that can be associated with the Morton effect instability for use
in predictive sensitivity assessments and moreover the design of the
experimental test rig.

Of the cases considered in this chapter all machines showed the time
dependant hysteresis effect (as shown in Figure 34 and Figure 48) on shaft
vibration whereby, once the machine had reached a point of instability and
was slowed, the vibration level did not immediately reduced to its initial level
but remained higher during the run down, it only approached its original run-
up levels after a period of time. This effect was accompanied in all cases by
some shift in phase angle, where this was observed to be 70-80o, 160-180o
and 130-150o, for the de Jongh et al [11], Faulkner et al [12,13] and HSOC
rotors respectively. Of the cases evaluated, the time constant involved for the
unstable nature of the vibration to take hold was between 1 and 5 minutes.
With the de Jongh et al [11] rotor increasing from 15%Cd microns to 30%Cd
microns in 60 seconds, Faulkner et al [12,13] rotor stability cyclic period taking
~90 seconds and steadily increasing in amplitude over a loner time base of

104
Chapter 4 – Benchmark Examples

15-20 minutes. Finally the HSOC rotor increasing from ~21%Cd to ~33%Cd in
~400 seconds.

Bearing loadings for all bearings were relatively low, with specific loadings
always below 0.3 MPa. This was accompanied by lower levels of eccentricity
ratio within the bearing (ranging from 0.1-0.3) which is directly associated with
the low specific loadings. Both fixed geometry and variable geometry bearings
have been seen to initiate thermal instability conditions. It is also observed
that the shaft whirl orbit shapes in the region of instability are very circular,
although the re-synthesised orbits generated from the measured data of the
HSOC rotor show higher elliptical components than the predictions.

In the cases examined not all rotors were operating at, or necessarily close to,
a region of a predicted rotordynamic critical. The de Jongh et al [11] and
HSOC rotors were both operating close to critical speeds. However, in the
case of the Faulkner et al [12,13] rotor, it was operating in a super critical
region but away from the next shaft bending mode by a suitable separation
margin > 30%. In all cases the shaft critical modes are associated with some
shaft rotor bending. In the case of the de Jongh et al [11] rotor the first simple
support bending mode is traversed and the rotor is operating closely to the
second shaft bending mode, whereas in the Faulkner et al [12,13] and HSOC
rotor the modes are coupled shaft overhung-bearing modes. In all the cases
examined the calculated rotor operating deflection shape at the unstable
operating speed is associated with a rotor bend point at the bearing location
where the shaft overhang is present. From these observations it is possible to
state that the operating vibration mode shape will tend to be sympathetic to a
rotor bow of an overhung section, it does not necessarily need to be
associated with a region of amplified dynamic response by operation close to
or at a critical speed, which is in slight contradiction to that presented by de
Jongh et al [11].

From the rotordynamic work presented here and the information presented in
the literature, it is clear that one of the most significant aspects in the thermal
stability of the rotors is the phase conditions present at the overhung rotor
bearing section. The deflected rotor response shapes are consistent with a
‘typical’ shaft bend mode being present at the overhung section of the rotor.
However the calculated mode shapes for the rotor critical speeds give rise to
conditions where the journal surface on the inside of the bow is closest to the
bearing surface (i.e. minimum film thickness and hence heat flux). Therefore
from a qualitative perspective, this condition would tend to generate a thermal

105
Chapter 4 – Benchmark Examples

differential such that the thermal bow, as a result a predominately forward

whirl orbital, would result in a stabilising effect with respect to the mechanical
unbalance. However the speeds at which instability have been observed to
initiate are where the calculated phase lag begins to shift to a point of
~between 80 and 150 degrees or greater. This condition would then result in a
vectorial contribution which will begin to compliment or add to the level of
mechanical unbalance as a result of any thermal bow. The observation that
the phase difference is less than 90 degrees would require that a further lag
between the minimum film thickness and the thermal bend formation be
present to provide a condition where a positive feedback loop is created to
result in a system gain >1. This condition of phase lag between the shaft hot
spot and rotor peak response was identified by Keogh et al [9] in the plain
bearing THD model for a simple shaft model and also by deJongh et al [11].
The rotordynamic work presented here indicates that a significant thermal
phase lag would be needed in addition to the presence of rotordynamic phase
lag to be present to provide a suitably strong positive gain via the shaft
thermal bow. This would provide the necessary threshold point to initiate
instability. Typically the phase lag would need to be somewhere in the region
of 15+ degrees in the fully unstable cases and between 0 and 40 degrees for
the marginal cases based on the calculated rotordynamic cases considered
here.

The unbalance response characteristics of the rotors considered is predicted

to be quite different in each case. The de Jongh et al [11] rotor was found to
be very sensitive to unbalance (which correlates with the predicted low level of
log decrement seen for the rotor second critical speed) at the overhung mass
with a vibration to unbalance ratio of 0.025 %Cd/g.mm at 10000 rpm with this
increasing to 0.052 %Cd/g.mm at the instability speed of 11500 rpm. The
Faulkner et al [12,13] and HSOC rotor were found to have lower sensitivities
at 0.028 %Cd/g.mm at 10000 rpm for Faulkner et al [12,13] and 0.03
%Cd/g.mm for the HSOC rotor. The unbalance sensitivity conditions correlate
with the more aggressive thermal instability response observed on the de
Jongh et al [11] rotor when compared to the less aggressive conditions seen
on the Faulkner et al [12,13] and HSOC cases. It should be noted however,
that the unbalance distribution in the rotor elsewhere was seen to significantly
modify the response characteristics at the bearings and also the operating
deflection shape. This could therefore modify the response sensitivity and
phase conditions at the overhung mass bearing in terms of altering the rotor’s
response to thermal instability and the location of the stability speed threshold.

106
Chapter 4 – Benchmark Examples

A closer inspection of the results for the HSOC rotor would suggest that the
change in rotordynamic conditions, due to an increased inlet viscosity, were
associated with the cooler start up conditions, is very slight. The more
influential aspect that drives the instability threshold is the increased viscous
shearing in the oil and corresponding heat generation.

In the Faulkner et al [12,13] rotor, whilst the unstable threshold speed is not
apparently closely associated with a critical speed, it does run in fairly
responsive rotor-bearing mode in the high speed range. During this,
significant and progressive phase change is predicted in the analysis. This
phase change (if linked with an additional thermal phase shift), when coupled
to the predicted shaft rotor deflection shape, would provide a positive gain to
the magnitude of rotor unbalance response at the turbine shaft end.

An interesting feature of note from both the Faulkner et al [12,13] and HSOC
rotor is the interaction between the DE and NDE bearings where significant
influence is seen between the two locations. From which it is apparent that
this interaction is significant in driving the instability. This coupled relationship
is one that would require additional consideration when linking any
rotordynamic and thermal instabilities, where existing attempts at thermal
bowing analysis have generally been limited to sub-modelling a particular
section of the dynamic system focused around the bearing and overhung
shaft sections.

It is clear from the conditions outlined above, that the combination of various
operating parameters and the methods in which they react make the precise
analysis of the thermal stability condition a very complex one. It should also
be noted that, whilst rotordynamic and hydrodynamic analyses have
developed to a level of accuracy once thought impossible, there remain a
number idealisations and assumptions in the analytical and numerical
procedures which result in a residual level of error in the performance
predictions. As such it can be understood how the precise evaluation of
thermal instabilities is a very difficult one where no prior knowledge of the
thermal or rotordynamic conditions of the machine are available.

107
Chapter 4 – Benchmark Examples

The principle observations and conditions concluded from the examined

cases are:

• Regions of thermally induced instability are associated with areas of

rotordynamic shaft bending at or around the bearing location.
• Fundamental modes of the vibration result in high spots being on the
inner bend point therefore suitable phase lag is required to generate a
positive feedback condition. I.e. any instability will occur at some point
below or above a natural frequency where the level of separation is
heavily governed by the system damping.
• The combined rotordynamic and phase conditions are key to the
conditions promotional for unstable conditions within a system.
• Orbit shape is typically circular and centred in the bearing (low
eccentricity ratios)
• All shafts are running above or in close proximity to a critical speed.
• Both fixed and variable geometries have shown the phenomenon.
• All rotors have significant overhung sections on the rotor.
• Sensitivities to overhung section imbalance range from 0.025 to 0.052
%Cd/g.mm. Rotors with higher sensitivities show more aggressive
instability conditions.
• All published cases have been seen to occur on relatively high speed
machines operating between 9000 rpm and 14000 rpm.

108
Chapter 5 – Design of Experiment

Chapter 5

Design of Experiment
- Experimental Test Rig

5. Design of Experiment - Experimental Test Rig

5.1 Introduction

This chapter describes the methods and process used to specify, analyse and
latterly design the experimental apparatus used to obtain the empirical data
presented within this thesis. Whilst some prior studies had been undertaken
into this topic area within the Cranfield rotating machinery department no
suitable test rigs were in existence at the outset of the research, hence the
necessity for the complete design and construction of suitable test equipment
to be undertaken.

The principle need for the creation of the test equipment was driven firstly by
the requirement to generate some controlled experimental data for a high
speed rotor specifically engineered to be prone to the Morton effect. The
generation of this data is needed due to a lack of quality data being available
in the field. In particular the need for coupled rotordynamic measurement and
rotor journal temperature is an area, that to the author’s knowledge, no
dedicated empirical data has been published to date with the exception of the
one industrial example presented de Jongh et al [11]. A secondary function of
the rotor test rig was to provide a means of practically benchmarking the
current analytical procedures presented in published work to provide
practicable means of assessing the susceptibility of high speed rotors to
bearing initiated synchronous thermal instability, such as that presented by
Balbahadur [19].

109
Chapter 5 – Design of Experiment

5.2 Experimental Rig Design Requirements

At the outset it is critical to ensure that the requirement of the test equipment
are well defined and achievable. The principle for the objective of the testing
work been generally outlined earlier in this thesis but a more prescriptive set
of requirements is outlined in the following points.

The experimental test apparatus should be specified and designed with

capability to address the following functional criteria:

• Rotor-bearing system rotordynamic design to be sympathetic to the

conditions considered to be promotional of the Morton effect as defined in
chapter 4 of this thesis.

• Rotor is to operate above the first shaft bending critical speed and within
an appropriate proximity of second bending critical.

• Operating speed is to be representative of appropriate industrial machines

where thermal instability has been observed. A minimum peak operating
speed of 10000 rpm should be applied.

• Dynamic shaft journal temperature measurement of the bearing journal at

the non-drive overhung mass section of the rotor is required. Sensors are
to be suitable to capture a fundamental harmonic temperature distribution
at an accuracy level of +/- 0.15 degrees of peak absolute operating
temperature. The selected measurement solution must not significantly
impact on rotordynamics or thermal distribution of the rotor.

• A clean, constant and controlled lubrication oil supply system is required,

allowing for the adjustment of supply pressures and temperatures to be
made. Oil inlet temperatures should be representative of appropriate
industrial machines, i.e. 40 oC.

• Rotor speed should be able to operate at any speed between 0 rpm and
full speed (i.e. variable speed operation) by means of an independent
drive system.

• Test rotor journal bearing temperature shall be measurable to enable

suitable operating temperature profiles to be obtained from the test
bearing and for safety purposes on non test bearing locations.

• Dynamic shaft vibration measurements are required, capable of capturing

both phase and amplitude of vibration.

110
Chapter 5 – Design of Experiment

• Steady state shaft position measurement of test bearing is required.

• Vibration measurements along the rotor length to enable basic operating

mode shapes to be interpreted.

• Data acquisition equipment must be suitable of capturing dynamic data at

a sampling rate which allows for accurate frequency analysis and high
speed shaft spatial movements to be adequately expressed. All low
frequency thermal data must be capable of being captured and logged at
a sampling frequency of no less than 10 samples per minute.

• Rotor must be capable of being operated in a safe and controllable

manner for extended periods of time by a single operator.

5.3 Measurement, Instrumentation and Control

As part of the experimental activity, a key element of the work is to obtain

suitable measurements of sufficient accuracy to fully characterise and
observe the principal features of interest relating to the thermal instability.

The rotordynamic aspects are clearly a fundamental feature that requires

close monitoring and measurement. To physically observe this sufficient
instrumentation is needed to qualify the dynamic response of the rotor at key
points along the rotor’s length. This should be achieved by means of dynamic
displacement measurements at the bearing locations and at several
significant points on the rotor to enable the operating deflection shape of the
shaft to be measured. Suitable means of phase referencing is needed to
enable a full planar spatial description at each location and correlation
between various points on the rotor.

Thermal measurement is again a fundamental parameter of interest and this

is both in the bearings, shaft journal and lubrication oil. The thermal conditions
in and around the bearing need to be measured sufficiently to obtain either a
suitable means of benchmarking theoretical predications or an as measured
map of the temperature conditions within and around the bearing. The
accuracy of instruments used should be such that the resolution and absolute
measurement obtained is several orders less than the expected temperature
variations around the bearing shell. Whilst the temperature conditions at both
bearings is of interest, a fuller understanding at the test bearing is necessary
as it is the circumferential thermal distribution in the bearing which dictates the
resulting temperature distribution within the shaft journal.

111
Chapter 5 – Design of Experiment

Of all the parameters involved in system instability, the temperature

distribution around the shaft journal provides the most significant condition
required to drive the time based response of the system and interaction with
the rotor geometric conditions. This parameter, whilst very significant, is the
one feature which is least analytically described and also to date very little
practical measurement has been reported or published. As part of this
research, direct measurement of the shaft journal will be needed to provide a
direct observation of the thermal conditions present within the shaft rotor and
how they develop across the dynamic operating range of the rotor. The
measurement of this area of the machine is more challenging due to the
nature of the rotating high speed shaft combined with the relatively high
resolution of temperature measurement needed to sufficiently describe the
temperature conditions for the expected temperature differentials around
shaft. Based on the case studies presented and the work discussed by Keogh
et al [9] then temperature variations across the shaft journal could range from
~2 to 10 degrees at the point of instability. This requires that, to ensure
suitable measurement under stable conditions is obtained, absolute
measurement accuracies should in the range of +/- 0.15 oC or better.

Several other parameters associated with the lubrication supply to the

bearings requires monitoring and control. The two principle parameters being
the supply viscosity, which can be monitored and controlled by means of inlet
temperature, and also the supply pressure to the bearing which will influence
the rate of bearing flow and hence cooling conditions. These parameters need
to be carefully controlled and monitored so as to ensure a good level of
repeatability in the running conditions at the bearings for the range of test
conditions evaluated. The potential sensitivity of the thermal conditions within
the bearing and shaft due to the oil inlet viscosity are inferred in the example
case study 3 summarised in chapter 4 where the instability was only incurred
when operating at lower oil inlet conditions.

All instrument outputs need to be suitably logged and acquired to enable the
necessary post processing activities. For dynamic signals such as the shaft
vibration and any seismic measurement, this should be carried out at a
suitable frequency to enable both harmonic analysis and real time description
of the shaft movements when considered with relation the rotor running speed
(i.e. sufficient points are acquired per shaft rotation to geometrically describe
the shaft’s position and loci). Temperature measurements for the purposes of
this study can be considered as quasi-static due to the steady state or slowly
moving nature of the thermal differentials and thermal inertias involved. The

112
Chapter 5 – Design of Experiment

changes of temperature with time have been observed to be slow with relation
to shaft rotation frequencies in the work presented by Keogh et al [10]. As
such the sampling frequency for the temperature measurements can be as
low as 0.25 Hz without influencing the quality of measurement.

5.4 Experimental Test Rotor and Bearing Design

5.4.1 Bearing Design and Selection

Two types of bearing have been initially identified for use within the rig. The
selection consists of one fixed geometry bearing and one variable geometry
bearing. The design and selection procedure for each of the bearings is
described in the following sub sections.

5.4.1.1 Fixed Geometry Bearing

A fixed geometry bearing is, as the name suggests, one in which the profile of
the bearing bore that is produced by its manufacturing process to the required
design specifications and remains consistent during operation. A number of
established fixed geometry bearings are available, each providing slightly
different operating characteristics. The basic necessity for different bearing
types being required is to ensure that, within the operating speed ranges shaft
stability can be maintained and the occurrence of a potentially aggressive
vibration termed as 'Half Speed Whirl' or 'Shaft Whirl' be avoided. Whilst the
specific limiting ranges (such as speed or load) for bearing type is dependent
on the actual rotor-bearing arrangement, Table 6 provides a brief summary of
the general performance measures for different fixed geometry bearing types.

For the experimental test rig two types of fixed geometry bearing were
considered, the taper dam and the offset half. Each of these bearings offers
some resistance to whirl by introducing pre-load into the oil film using a form
of geometrically forced converging oil film. The taper dam is a simpler bearing
to manufacture and as such less costly, so this would have been the preferred
option if a workable solution could be identified. However due to the shaft
speeds and static shaft loads involved it was felt unlikely that this type of
geometry would provide a suitably stable design. It was evaluated for
completeness.

113
Chapter 5 – Design of Experiment

The fundamental design constraint with reference to the bearings was the
amount of available drive power as dictated by the test drive motor and
inverter. As such it was necessary to ensure that the motor/inverter torque
speed characteristic was matched to the bearing speed power loss
characteristics. At full speed the max motor output shaft power was limited to
4.5 kW. After calculating and evaluating some approximate bearing power
losses for a range of bearing journal sizes and comparing these against the
available motor torque, a nominal bearing journal diameter of 50mm with a
length to diameter ratio of 0.7 was selected as the limiting constraint. With the
nominal bearing size determined, a number of bearing designs were analysed
and the various relevant operating parameters extracted for comparison.
Table 7 presents the data for the taper dam and offset half bearing designs for
a range of assembled bearing diametral clearances. Stiffness and damping
values are not included in the table due to the quantity of data generated but a
measure of stability is included by means of the critical mass and whirl
frequency. It should be noted that the stability requirement is based on a
critical mass being more than half the shaft mass, i.e. approx. 15 kg, and to
provide a sufficient margin for calculation error, a value of +25% was applied
to his half mass criteria. The bearing performance calculations were
performed using the Franklin Institute HYBJRN AGSL custom Fortran routine
[54].

From the data in Table 8 it can be seen that the taper dam bearing variants do
not provide a stable bearing at all running speeds and also bearing
temperatures can be seen to exceed 100 deg C. For satisfactory operation
and to ensure no unwanted dynamic or functional performance complications
are present which could cloud any evaluations relating to the presence of the
'Morton Effect' the bearings must operate in a fully stable range with respect
to bearing whirl and operate within peak metal temperatures of 95 degrees C.

114
Chapter 5 – Design of Experiment

Table 6 Fixed Geometry Bearing Types

It has been suggested, in several of the published works relating to Morton

Effect instabilities, that hydrodynamic bearing conditions where the rotor is
sensitive to instability [10,19,23], the quasi-static shaft running position
remains relatively central with respect to the bearing centre, having
eccentricity ratios ε o generally operating at < 0.3. Additional to this is the
requirement that a reasonably circular orbit with moderate amplitude is

115
Chapter 5 – Design of Experiment

attained. For this to occur the bearing must not have too higher values for
stiffness and damping or this could produce an over stable system with low
response to synchronous unbalance. It is the balance between having a
stable system but also a responsive system that requires careful and
considered selection of the bearing arrangement and geometries.

Bearing Bearing Diametral Clearance Cd Bearing Pre-load

Design Type (mm) pR
minimum maximum maximum Minimum
2 Offset Half 0.102 0.143 0.5 0.3
3 Offset Half 0.127 0.168 0.5 0.3
4 Offset Half 0.153 0.193 0.5 0.3
5 Offset Half 0.102 0.143 0.3 0.3
6 Offset Half 0.127 0.168 0.3 0.3
7 Offset Half 0.153 0.193 0.3 0.3
8 Taper 0.127 0.168 0.1 0.16
Dam
9 Taper 0.153 0.193 0.1 0.16
Dam
10 Taper 0.178 0.219 0.1 0.16
Dam
11 Taper 0.203 0.244 0.1 0.16
Dam
12 Plain Cyln 0.102 0.143 0.1 0.16
Table 7 Fixed Geometry Bearing Designs

Using the evaluations outlined above the fixed geometry bearing type
selected was an offset half type. Considering the specific operating
parameters of oil film thickness, stability, operating temperature, stiffness and
damping the following designs 2, 3, 4 and 5 provide acceptable functional
solutions, see Table 8. For the final selection of the bearing geometry and
clearances the rotordynamics and bearing dynamics need to be considered in
combination. This aspect is covered in the following sections rotordynamic
analysis.

116
Chapter 5 – Design of Experiment

Table 8 Summary of Fixed Geometry Bearing Calculations

117
Chapter 5 – Design of Experiment

5.4.1.2 Variable Geometry Bearing

A variable geometry bearing is a partial arc design configuration of a plain

bearing. The configuration is made of multiple individual bearing pads which
are allowed to pivot or tilt to conform with the dynamic loads from the lubricant
and shaft, generate a converging oil film wedge capable of supporting radial
loads. The principle benefit of tilting pads bearings is that they are capable of
operating at high speeds under light load conditions whilst maintaining a high
degree of stability and resistance to whirl. A tilting pad bearing however
generally has a lower load carrying capacity than a plain bearing and also,
due to its increased complexity, carries a higher cost.

The pads are produced with a greater radii of curvature than the radii at which
they are mounted relative to the bearing centre. This condition introduces the
pre-load effect which helps to generate the oil film wedge and as such directly
influences the bearing load carrying capacity and specifically the dynamic
properties of the bearing. The bearing can be configured with a varying
number of pads, typically 4 or 5, which are either configured with a central
pivot or an offset pivot which allows bi-directional and uni-directional operation
respectively. Figure 62 shows a schematic of a typical tilting pad bearing
configuration.

Split Loose
Housing Pads
Shell

Rocker or
Pivot

Figure 62 Tilting pad bearing schematic.

The design of the tilting pad bearing for the experimental rig was based
around a 5 pad central pivot configuration. The bearing diameter was pre-
determined at 50mm by the previous work conducted on the fixed geometry
bearing selection. It was originally intended that an existing bearing design
would be procured from an established bearing manufacturer. However, due

118
Chapter 5 – Design of Experiment

to the cost constraints of the project and the high cost of the bearings, it was
decided that the bearings would be specifically designed and manufactured.

The detailed parameters of the bearing design were determined such that the
fixed geometry bearing and tilting pad bearing would be fully interchangeable.
Clearance and pre-load ranges for the bearing were selected to be consistent
with the internal geometry used in industrial variable geometry bearings when
running under light loads. Table 9 gives a summary of the tilting pad bearing
specifications.

Minimum Maximum
Bearing Diametral Shaft
Eccentricity Film Bearing
Design Clearance Cd Speed
Ratio Thickness Temperature
Ref (microns) (rpm)
(mm) (deg C)
70 5000 0.099 0.025 64.0
10000 0.063 0.025 77.0
13
15000 0.050 0.025 87.0
20000 0.041 0.024 95.2
90 5000 0.157 0.032 60.9
10000 0.105 0.033 71.6
14
15000 0.082 0.033 79.5
20000 0.068 0.033 86.4
100 5000 0.190 0.035 59.9
10000 0.127 0.036 69.7
15
15000 0.098 0.037 76.9
20000 0.084 0.037 83.6
125 5000 0.270 0.041 57.6
10000 0.186 0.044 66.5
16
15000 0.146 0.045 72.9
20000 0.123 0.046 78.7
140 5000 0.312 0.044 56.5
10000 0.221 0.048 64.9
17
15000 0.176 0.05 71.4
20000 0.150 0.051 76.9
160 5000 0.365 0.047 55.4
10000 0.261 0.053 62.8
18
15000 0.214 0.055 69.2
20000 0.186 0.057 75.1

Table 9 Summary of Variable Geometry Bearing Parameters Evaluated

Bearing performance predictions were calculated for range of operating speed

conditions and clearances. The limiting bearing operational constraints were
consistent with those defined for the fixed geometry bearing design. Table 9
provides a summary of the bearing performance data for the running speeds
evaluated. The fixed geometry bearing calculations were performed using a
custom code referenced as HYDJRN [54], which is custom code developed
by Allen Gears in conjunction with the Franklin Institute hydrodynamic bearing
codes. The computational procedure uses a 2D finite difference grid to solve
the Reynolds equations and flow conditions within the bearing oil film. Bearing
performance criteria for lubricant temperatures, pressures, film thickness and
stiffness and damping data are produced by the analysis.

119
Chapter 5 – Design of Experiment

5.4.2 Rotor Dynamic Analysis

From the observations made in chapter 4 and from comments presented by

both Keogh et al [10] and Balbahadur et al [19], it is suggested that a
significant element in creating conditions promotional for thermal instability to
occur is the presence of a critical speed within, or of close proximity to, shaft
running speeds. From the work conducted in chapter 4 this has not been the
situation in all industrial cases, however a ‘lively’ rotor is required. As such it
was defined that that the experimental rotor should be designed to operate
above or close to a suitable natural frequency. However care must be taken to
ensure that the high vibrations experienced at and around critical speeds are
within acceptable and controllable levels. This to some extent can be
managed through the stiffness and damping properties in the bearings.

To aid in producing a rotor design that would be susceptible to displaying the

Morton Effect, an approach was taken whereby the experimental test rotor
bearing system would be designed so as to produce similar dynamic
characteristics as defined in chapter 4 section 4.3. The rotor dynamic
characteristics were intended to follow those observed in the de Jongh et al
[11] rotor, where the rotor was seen to be very sensitive to the thermal
instability experienced. This methodology dictates that the rotor is capable of
running supercritical, i.e. the rotor runs above its first shaft bending natural
frequency.

The specification for the rotor was targeted to have a running range above the
first bending natural frequency and up to or very close to that of a second
shaft bending mode of the rotor with the maximum rotation speed limited to
10000 rpm. This criterion presents two beneficial features. Firstly, the impact
of the first bending natural frequency as well as the second can now be
investigated, and secondly, conditions similar to those presented for the
compressor rotor produced by de Jongh et al [11], which was shown to
display the 'Morton Effect' thermal instability, will be introduced.

A speed limitation of 15000 rpm was established to ensure that some margin
remained for the total power consumption of the test rig, which would be
limited by the drive motor and frequency inverter capacity (~4.5 kW). Also at
speeds beyond 15000 rpm it becomes very problematic and costly to provide
methods of extracting the shaft temperature sensor signals from the rotor. The
method selected for the extraction of the temperature sensor signals was by
means of electrical slip ring. To operate at speeds much above 15000 rpm

120
Chapter 5 – Design of Experiment

without significant signal distortion or excessive wear would require very

specialist instrumentation and slip ring technology which would beyond the
budget constraints of this research.

5.4.2.1 Preliminary Lateral Rotordynamics

With the basic experimental design criteria established, as a starting position

to determine preliminary shaft rotor dimensions, some basic rotor critical
speeds were calculated. The shaft diameter was kept fixed nominally at
50mm in line with the bearing dimensions and the bearing span altered to
produce a range of shaft natural frequencies. As the shaft overhang has been
identified as a key feature in the thermal instability process a nominal shaft
overhang length of 200mm was kept constant for all bearing spans. The
bearing support stiffnesses were varied across a range to provide indication of
the impact of the support effects on the natural frequencies and mode shapes.

The location the natural frequencies is not the only element that needs to be
considered at this stage. The associated mode shapes are of great
significance especially where they have been attributed to being one of the
necessary driving elements observed to provide conditions susceptible to the
manifestation of when thermal instabilities can occur [10,11]. As such nominal
support stiffnesses were taken from the bearing calculations that had already
been performed, and this used to develop a series of un-damped critical
speed maps. As the stiffness and damping effects in the bearing are complex
and consist of not only directional stiffnesses but also cross-coupling effects,
only the Kxx and Kyy stiffness data was considered for simplicity at this point.

The undamped critical speed map was evaluated for the support stiffness
range of 1e+5 to 1e+10 N/m. See Table 10. The associated critical speeds for
the rotor configurations were established for bearing oil film stiffness range
Kxx and Kyy. The resulting natural frequencies that were calculated were
assessed and the mode shape categorised in line with the modes shapes as
shown in Figure 63. It is important the mode shapes for the natural
frequencies encountered within the test rig speed range are suitably
described and that they satisfy the previously established criterion with
relation to shaft bending modes etc. The mode shapes shown in Figure 63
provide very distinct and separate deflected vibration modes, however in
reality the interaction of bearing and rotor stiffness increases the complexity of
the modes and in some cases combine to give more complex deflected

121
Chapter 5 – Design of Experiment

shapes. Where this occurs the assigned mode shapes in Table 10 are
classified by the combination of the two most relevant mode shapes.

Bearing Modes Shaft Bending Modes

a) Bounce
c) 1st Bending

b) Rocking d) 2nd Bending

Figure 63 Rotor Mode Shapes Supported in Two Hydrodynamic Mode Shapes

Table 10 shows each of the calculated natural frequencies for the bearing
spans considered and assigns the associated mode shapes for each
frequency. It can be seen from the data that some of the natural frequencies,
although within the speed range, are not specifically shaft bending modes but
essentially bearing dominated modes. As such these are not the dynamic
conditions that have been defined as those desired for creating conditions
sensitive to the Morton Effect as has been specified for the experimental test
rotor design. Also damping levels for these bearing modes are invariably high
due to being governed by the oil film specifically. For this reason the rotors of
625mm and 500mm bearing span were disregarded as the associated shaft
stiffnesses were too high to introduce the desired mode shapes.

Other more practical considerations in the rotor design were aspects such as
physical size, manufacture capability and ability to balance etc. Due to these
points the preference was to have as short a bearing span rotor as possible
whilst providing the desired rotordynamic conditions. This approach identified
the 750mm bearing span as being the most preferred arrangement. However,
the natural frequency locations for neither the nominal support stiffness or the
rigid supports fulfilled the specification for operating speed range. In an
attempt to modify the location of the natural frequencies without radically
altering the mode shapes, a larger diameter section of 150mm was added to
the centre of the shaft for a length of 100mm. The objective of introducing this
feature was to add some significant mass to the rotor centre without
increasing the shaft stiffness by a similar order of magnitude and pushing
natural frequencies up. For comparative purposes all of the rotor bearing

122
Chapter 5 – Design of Experiment

spans were again assessed with the change in rotor geometry. Figure 64
presents the results for the 750mm bearing span in an un-damped critical
speed map with approximated nominal support stiffness Kxx and Kyy.

Rotor Bearing Support Natural Mode

Span (mm) Stiffness (N/m) Frequency (cpm) Shape
1000 3.2E+07 5695 C
1000 3.2E+07 14684 D
1000 1.00E+07 5793 C
1000 1.00E+07 17584 D
875 3.2E+07 7222 C
875 3.2E+07 14907 D
875 1.00E+07 7381 c
875 1.00E+07 19013 d
750 3.2E+07 9342 c
750 3.2E+07 15169 a/d
750 1.00E+07 9595 c
750 1.00E+07 20401 d
625 3.2E+07 12077 c
625 3.2E+07 15855 a
625 1.00E+07 12544 c
625 1.00E+07 22448 a/d
500.0 3.2E+07 13931.0 c
500.0 3.2E+07 19029.0 a
500.0 1.0E+07 15665.0 c
500.0 1.0E+07 26997.0 a
Table 10 Natural Frequency of Plain Rotors of Varying Bearing Span

Undamped Critical Speed Map

Allen Gears Morton Effect Test Rotor
750mm Bearing Span
40000

35000
Critical Speed, cpm

30000

25000

20000

15000

10000

5000

0
1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09 1.00E+10 1.00E+11

Bearing Stiffness, N/m

Figure 64 Un-damped Critical Speed Map for 750mm Bearing Span With
Added Mass at Shaft Centre

123
Chapter 5 – Design of Experiment

The resulting natural frequency ranges of 6565-6876 and 14649-19087 cpm

for the first and second natural frequencies respectively are now close enough
to the specified running range that, as part of the detailed rotor design and
analysis with specific bearing coefficient data, the desired rotordynamic
properties can be obtained. Specifically the introduction of the oil film damping
effects onto the rotor will cause a reduction in the natural frequency locations
to take place, bringing down the second bending mode closer to the proposed
running range capability of the experimental apparatus.

5.4.2.2 Detailed Lateral Rotordynamics

With the nominal rotor dimensions determined, the detailed rotordynamics

were performed. In this stage of the design an increased number of operating
parameters were considered, such as rotor stability, response, orbit size and
phase, unbalance location and fundamentally the damped natural (eigen)
frequencies of the system.

In the process of analysing the rotor, a number of bearing designs as

established in section 5.4.1 were considered (based upon fixed and variable
geometry). For each design the maximum and minimum clearance conditions
were considered in combination with different oil inlet conditions. The number
of various rotor-bearing combinations generated significant amounts of data
and only a summary of those results is presented here. However, it can be
said that the controlling feature in the design evaluation and selection of the
rotor-bearing arrangement was the system rotordynamic stability.

For the majority of bearing designs considered (essentially clearance and pre-
load changes), the natural frequency locations were acceptable through the
implementation of some minor geometry changes. It was found that for many
of the system combinations the logarithmic decrement (log dec), which is a
measure of the system stability (see chapter 4), was low at the location of the
first shaft bending natural frequency. It is normal for a system to be
considered stable by means of analysis when the log dec value is 0.2-0.3 or
greater. Many of the designs produced negative values which is a strong
indication that system would become fully unstable and therefore practically
inoperable.

There are two principle reasons for the instability. One is that due to
excessive clearance and insufficient preload in the fixed geometry bearing

124
Chapter 5 – Design of Experiment

'Half Speed' or 'Shaft' whirl was present. The second was that it appeared that
if the bearings were too stiff there was little movement at the bearing supports
and hence little damping contribution from the hydrodynamic bearings. As it is
also important to have a bearing design where a moderate orbit due to
mechanical unbalance is present, then a compromise was struck between the
amount of clearance in the bearing, the rotor system stability and the bearing
performance.

The calculated natural frequencies and associated mode shapes were

extracted for each combination of bearing clearance and oil inlet temperature.
The specific natural frequencies were established by means of a Campbell
diagram for each case the results of which is presented in Table 11. A
Campbell diagram for the nominal bearing clearances is provided in Figure 67
with the associated mode shapes for the 1st and 2nd critical speed presented
in Figure 68. The data and associated hydrodynamic and rotordynamic
models are based on final dimensions produced following the detail design of
the experimental test rotor and components such as couplings etc. An
overview of the detailed design is provided later in this chapter. It should be
noted at this point that the initial design specification was engineered around
the use of fixed geometry bearings as the use of variable geometry bearings
was only introduced later in the research programme.

Flexible Rotor Model - Offset Half Bearings - Overhung Mass

0.3

0.2
Shaft Radius, meters

0.1

3031
5 10 15 20 25
0

-0.1

-0.2

-0.3
0 0.2 0.4 0.6 0.8 1
Axial Location, meters

Figure 65 Rotordynamic Model of Flexible Rotor – with Added Mass

125
Chapter 5 – Design of Experiment

Undamped Critical Speed Map

Flexible Rotor Model - Offset Half Bearings

100000

Critical Speed, rpm

10000

1000
1.00E+06 1.00E+07 1.00E+08 1.00E+09 1.00E+10

Bearing Stiffness, N/m

Figure 66 Undamped critical speed map for flexible rotor without added mass

In addition to the model for the plain rotor a number of further calculations
were performed where an added mass was included on the overhung section
of the rotor. This additional mass allows for modification of the rotordynamic
characteristics and also provides a mechanical gain function to the conditions
influencing the mechanism of the Morton Effect.

Damped Natural Frequency Map - Campbell Diagram

Flexible Rotor Model - Offset Half Bearings
18000 0.400
16000 0.350

Logrithmic Decriment
Natural Frequency

14000 0.300
12000
0.250
rpm

10000
0.200
8000
0.150
6000
4000 0.100
Log Dec at 6500
2000 0.050
Log Dec at 15700
0 0.000
0 2000 4000 6000 8000 10000 12000 14000 16000 18000

Rotor Speed, rpm

Figure 67 Damped natural frequency map for rotor in fixed geometry bearings,
mean clearance, normal inlet temperature with no added mass

126
Chapter 5 – Design of Experiment

Damped Eigenvalue Mode Shape Plot Damped Eigenvalue Mode Shape Plot
Flexible Rotor Model - Offset Half Bearings Flexible Rotor Model - Offset Half Bearings

1.5 1.5
1 1
Re(x) Re(x)
0.5 Im(x) 0.5 Im(x)
0 Re(y) 0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y) -0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y)
-1 f=6495.3 cpm -1 f=17177.3 cpm
-1.5 d=.2919 logd -1.5 d=3.1739 logd
N=6500 rpm N=16000 rpm
Axial Location, m eters Axial Location, m eters

Damped Eigenvalue Mode Shape Plot Damped Eigenvalue Mode Shape Plot
Flexible Rotor Model - Offset Half Bearings Flexible Rotor Model - Offset Half Bearings

forward forward
backward backward
f=6495.3 cpm f=17177.3 cpm
d=.2919 logd d=3.1739 logd
N=6500 rpm N=16000 rpm

(a) (b)
Figure 68 Mode shapes for rotor in fixed geometry bearings, mean clearance, normal
inlet temperature with no added mass (a) is first critical at 6500 rpm and (b) is second
critical at 15750 rpm

The effect of the overhung mass is to provide a rotordynamic gain function

whereby any thermal differential generated at the NDE rotor journal results in
the corresponding bow. This bow then in effect generates further unbalance
excitation due to the effectively eccentric mass of the overhung section of
shaft and results in a corresponding movement on the journal due to the
increased unbalance forces. This is a basic feedback loop mechanism and if
the overhung mass amount is increased this results in a greater amplification
for any unit thermal bend at the bearing journal.

Critical Speed 1 Critical Speed 2 Critical Speed 3

Clearance Frequency Frequency Frequency
Bearing Type OHM Log dec Mode Type Log dec Mode Type Log dec Mode Type
range (cps) (cps) (cps)
Fixed Geometry minimum 1.3 836 3.994 a/b 6793 0.361 c 20253 1.544 d
mean 1.3 1105 3.754 a/b 6464 0.302 c 17177 3.174 d
maximum 1.3 1743 3.644 a/b 6682 0.59 c 24108 1.811 d
Fixed Geometry minimum 5.0 839 3.969 a/b 6128 0.173 c 12486 3.168 d
mean 5.0 1120 3.647 a/b 6111 0.214 c 13971 3.647 d
maximum 5.0 1447 3.673 a/b 6088 0.282 c 18253 3.743 d
Variable Geometry minimum 5.0 6060 0.0861 c 10339 0.7866 d - - -
mean 5.0 5972 0.1391 c 8940 1.0447 d - - -
maximum 5.0 5951 0.1517 c 8557 1.0819 d - - -

Table 11 Summary of Test Rotor Critical Speeds and Mode Shapes

The results of this work are also summarised in Table 11, with some of the
critical speed maps and modes included in Figure 69 through to Figure 73. It
should be noted however that the introduction of a larger mass at the shaft
overhang does not necessarily always result in a higher gain function for the

127
Chapter 5 – Design of Experiment

thermal instability due to the complex relationship between rotor shaft

deflection, shaft phase and unbalance response. Conditions can occur where
increased mass results in a higher gain which tends to a stabilising thermal
condition as opposed to an instability. In some conditions the increase in
mass modifies the rotor response characteristics sufficiently that the net effect
is to provide a more thermally stable rotor with regards to the Morton effect.
This is considered more commonly to be due to reduced response at the
bearing as opposed to any phase changes but this is not exclusively the case
and is specifically dependant on the rotor geometry and bearing performance
characteristics.

Fixed Geometry ‘Offset Half’ Bearing Design Specification

Bearing Diameter 50 mm
Bearing Length 35 mm
Minimum Diametral Clearance 0.142 mm
Minimum Diametral Clearance 0.168 mm
Maximum Pre-load 0.3
Minimum Pre-load 0.3
Bearing Radial Load 135 N
Bearing Specific Pressure 0.075 MPa
o
Oil Inlet Temperature 40 C
Oil Type (Grade) ISO VG 46
Table 12 Fixed geometry bearing basic design specification

Stiffness Oil Film Coeficients (N/m) Damping Oil Film Coefficients (N.s/m)
Rotational Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy
5000 29693312 13350590 -48663248 15089423 90485 -57024 -56849 145932
10000 45732257 21006240 -73393558 22610441 70426 -47126 -43972 114486
15000 59531680 27282193 -95433935 29193587 60966 -42308 -38016 99157
20000 70499356 32270247 -112950771 34425066 54133 -34600 -33636 88032

Table 13 Fixed geometry bearing oil film stiffness and damping coefficients for
mean clearance and normal oil inlet temperature

128
Chapter 5 – Design of Experiment

Variable Geometry Tilting Pad Bearing Design Specification

Bearing Diameter 50 mm
Bearing Pad Length 20 mm
Minimum Diametral Clearance 0.100 mm
Minimum Diametral Clearance 0.125 mm
Maximum Pad Pre-load 0.6
Minimum Pad Pre-load 0.4
Number of Pads 5
o
Length of Pad 60 C
Pivot Location Centre
Radial Load Direction
Load on pad (LOP) LBP
Load between pad (LBP)
Bearing Radial Load 135 N
Bearing Specific Pressure 0.135 MPa
o
Oil Inlet Temperature 40 C
Oil Type (Grade) ISO VG 46
Table 14 Variable geometry bearing basic design specification

Stiffness Oil Film Coeficients (N/m) Damping Oil Film Coefficients (N.s/m)
Rotational Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy
5000 15771000 5254 -3894 17930000 19516 9644 -10184 21399
10000 23397000 20060 12480 24922500 14684 7204 -7366 15346
15000 29565500 7134 -11882 30788000 12352 6008 -6092 12694
20000 34900000 -10647 -24629 35944500 10815 5189 -5254 11031

Table 15 Variable geometry bearing oil film stiffness and damping coefficients
for mean clearance and normal oil inlet temperature

Damped Natural Frequency Map - Campbell Diagram

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass

18000 4.000
16000 3.500
14000
Logrithmic Decriment

3.000
Natural Frequency

12000
2.500
(cpm)

10000
2.000
8000
1.500
6000
4000 1.000
1X
Log Dec at 6111 rpm 0.500
2000 Log Deg at 13730 rpm
0 0.000
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Rotor Speed, rpm

Figure 69 Damped natural frequency map for final rotor design with fixed
geometry bearings, mean clearance and 3.5 kg added overhung mass

129
Chapter 5 – Design of Experiment

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass

forward
backward
f=1107.1 cpm
d=3.7436 logd
N=1000 rpm

Figure 70 Damped mode shape for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at 1107 rpm
critical (bearing rock mode) with log decrement of 3.743

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass

forward
backward
f=1120.2 cpm
d=3.6466 logd
N=1000 rpm

Figure 71 Damped mode shape for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at 1120 rpm
critical (bearing rock mode) with log decrement of 3.646

130
Chapter 5 – Design of Experiment

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Offset Half Bearings - Overhung Mass

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y)
-1 f=6111.8 cpm
-1.5 d=.2144 logd
N=6000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Offset Half Bearings - Overhung Mass

forward
backward
f=6111.8 cpm
d=.2144 logd
N=6000 rpm

Figure 72 Damped mode shape for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at 6111 rpm
critical with log decrement of 0.214
Damped Eigenvalue Mode Shape Plot
Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y)
-1 f=13971.7 cpm
-1.5 d=3.6477 logd
N=13500 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Offset Half Bearings - Overhung Mass

forward
backward
f=13971.7 cpm
d=3.6477 logd
N=13500 rpm

Figure 73 Damped mode shape for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at 13971 rpm
critical (bearing rock mode) with log decrement of 3.647

131
Chapter 5 – Design of Experiment

To establish predictions for the rotors response across the operating speed
range, several response analyses were performed for the various
combinations of bearing, lubrication and overhung mass. The results of this
analysis were used to understand the anticipated conditions of displacement
that would be expected across the rotor length. The level of unbalance used
to evaluate the rotor response was consistent with an unbalance mass of 750
mg at a radii of 50mm at station 1 (the overhung non drive end flange) of the
model. This level of imbalance is in accordance with ISO 1940 [55] grade
G2.5 for a rotor of this weight. (ISO 1940 is a balancing standard which is
used to determine permissible levels of residual unbalance in rotating
machinery and grade G2.5 is typical for machines such as industrial
compressors of gas turbines).

Rotordynam ic Response Plot

Flexible Rotor Model - Offset Half Bearings - 3.5kg Overhung Mass
Sta. No. 8: NDE Brg
14 360
270
12 180
Response, microns pk-pk

90
10 Major Amp
0
Horz Amp
8 -90
-180
Vert Amp
6 -270 Horz Phs
-360 Vert Phs
4
-450
-540
2
-630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 74 Damped response plot for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at NDE bearing

Rotordynam ic Response Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass
Sta. No. 1: Flange
200 360
270
180
180
160
Response, microns pk-pk

90
Major Amp
140 0
Horz Amp
120 -90
Vert Amp
100 -180
-270 Horz Phs
80
-360 Vert Phs
60
-450
40
-540
20 -630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 75 Damped response plot for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at NDE flange

132
Chapter 5 – Design of Experiment

Rotordynam ic Response Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass
Sta. No. 17: Ctr
160 360
270
140
180

Response, microns pk-pk

120 90 Major Amp
0
100 Horz Amp
-90
Vert Amp
80 -180
-270 Horz Phs
60
-360 Vert Phs
40 -450
-540
20
-630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 76 Damped response plot for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at centre mass

Rotordynam ic Response Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass
Sta. No. 26: DE Brg
14 360
270
12 180
Response, microns pk-pk

90
10 Major Amp
0
Horz Amp
8 -90
-180
Vert Amp
6 -270 Horz Phs
-360 Vert Phs
4
-450
-540
2
-630
0 -720
0 5000 10000 15000 20000 25000
Rotor Speed, rpm

Figure 77 Damped response plot for final rotor design with fixed geometry
bearings, mean clearance and 3.5 kg added overhung mass at DE bearing

In addition to evaluating the impact of imbalance at the overhung flange of the

rotor, imbalance was also included at the shaft centre. Combinations of
imbalance at both locations simultaneously were also assessed, considering
both in and out phase conditions. Figure 74 to Figure 77 present the
unbalance response for the rotor at the stations 1, 8, 17 and 26 which
correspond to the NDE flange, NDE end bearing, rotor centre and DE bearing
respectively. The response data presented in Figure 78 to Figure 80 is with a
3.5 kg added overhung mass and a residual unbalance values of 37.5 g.mm
and 12.5 g.mm applied to the NDE flange and shaft centre mass respectively,
0 degrees phase lag was applied between the unbalance vectors located at
the NDE flange and rotor centre. The 12.5 g.mm applied to the rotor centre is
to account for a proportion of the unbalance mass to be distributed at this

133
Chapter 5 – Design of Experiment

location as would be in place in a practical component (i.e. no matter how well

balanced, practically a proportion of the unbalance mass will be located at the
shaft centre due to its high proportion of weight of the total shaft). It should be
noted that only synchronous excitation was applied as is consistent with
imbalance forces in rotating machinery. No account for higher order
excitations were considered in the rotor response analysis.

For each of the response conditions the deflected rotor shapes was extracted.
These deflected rotor shapes provide an understanding of the conditions that
will be experienced at the bearings when considered in combination with
phase. With a means of quantifying the conditions at the bearings across the
running speed range it is possible to qualitatively determine that the
conditions are consistent with the criteria defined as promotional to the
initiation of thermal instability, as per chapter 4. Orbit shapes were also
extracted at the NDE bearing location at or around the rotor critical speeds.
Figure 78 to Figure 80 present the deflected shapes for the rotor with and
without the overhung mass for the critical speed locations.

Rotordynamic Deflected Shape Plot

Flexible Rotor Model - Offset Half Bearings - 3.5 kjg Overhung Mass
Calculated Shaft Orbit
Deflected Shape at 6200 rpm
4
0.08
3
0.06
Vertical Displacement (microns)
Response Am plitude, m m pk

2
0.04
1
0.02
0
0
-1
0 0.2 0.4 0.6 0.8 1 1.2
-0.02
-2
-0.04 Major Amp
-3
Horz Amp
-0.06
Vert Amp -4
-0.08 -4 -3 -2 -1 0 1 2 3 4
Horizontal Displacem ent (m icrons )
Axial Location, m eters

Figure 78 Damped response operating deflection shape and NDE bearing

whirl orbit for final rotor design with fixed geometry bearings, mean clearance
and 3.5 kg added overhung mass at DE bearing at 6200 rpm.

134
Chapter 5 – Design of Experiment

Rotordynamic Deflected Shape Plot Calculated Shaft Orbit

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass
5
Deflected Shape at 10000 rpm
4
0.012

Vertical Displacement (microns)

Major Amp 3
Response Amplitude, mm pk

0.01 Horz Amp 2

Vert Amp 1
0.008
0

-1
0.006
-2
0.004 -3

-4
0.002
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5
0
Horizontal Displacem ent (m icrons)
0 0.2 0.4 0.6 0.8 1 1.2
Axial Location, m eters

Figure 79 Damped response operating deflection shape and NDE bearing

whirl orbit for final rotor design with fixed geometry bearings, mean clearance
and 3.5 kg added overhung mass at DE bearing at 10000 rpm.

Rotordynamic Deflected Shape Plot Calculated Shaft Orbit

Flexible Rotor Model - Offset Half Bearings - 3.5 kg Overhung Mass
8
Deflected Shape at 12600 rpm
0.025 6

Vertical Displacement (microns)

Major Amp
0.02 4
Response Amplitude, mm pk

Horz Amp
0.015
2
Vert Amp
0.01
0
0.005
-2
0
-0.005 0 0.2 0.4 0.6 0.8 1 1.2 -4

-0.01 -6

-0.015 -8
-8 -6 -4 -2 0 2 4 6 8
-0.02
Horizontal Displacem ent (m icrons)
-0.025
Axial Location, m eters

Figure 80 Damped response operating deflection shape and NDE bearing

whirl orbit for final rotor design with fixed geometry bearings, mean clearance
and 3.5 kg added overhung mass at DE bearing at 12600 rpm.

The rotordynamic data presented in the preceding figures relates to the fixed
geometry offset half type bearing configuration. The same procedure was
applied for the evaluation of the flexible rotor for the tilting pad (TP) bearings.
Figure 81 to Figure 83 present the damped critical speed map (Campbell
diagram), rotor damped mode shapes, rotor response at the NDE bearing and
the rotor deflected shape at 10000 rpm for the nominal TP bearing clearance
condition with normal oil inlet temperature.

135
Chapter 5 – Design of Experiment

Damped Natural Frequency Map - Campbell Diagram

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass
30000 1.200
1X
Log Dec at 5972 rpm
Natural Frequency, rpm
25000 Log Dec at 8940 rpm 1.000

Logrithmic Decriment
20000 0.800

15000 0.600

10000 0.400

5000 0.200

0 0.000
0 2000 4000 6000 8000 10000 12000 14000 16000
Rotor Speed, rpm

Figure 81 Damped natural frequency map for final rotor design with variable
geometry bearings, mean clearance and 3.5 kg added overhung mass

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y)
-1 f=5972.0 cpm
-1.5 d=.139 logd
N=6000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass

forward
backward
f=5972.0 cpm
d=.139 logd
N=6000 rpm

Figure 82 Damped mode shape for final rotor design with variable geometry
bearings, mean clearance and 3.5 kg added overhung mass at 5972 rpm
critical with log dec of 0.139

136
Chapter 5 – Design of Experiment

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass

1.5
1
Re(x)
0.5 Im(x)
0 Re(y)
-0.5 0 0.2 0.4 0.6 0.8 1 1.2
Im(y)
-1 f=8940.7 cpm
-1.5 d=1.0447 logd
N=9000 rpm
Axial Location, m eters

Damped Eigenvalue Mode Shape Plot

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass

forward
backward
f=8940.7 cpm
d=1.0447 logd
N=9000 rpm

Figure 83 Damped mode shape for final rotor design with variable geometry
bearings, mean clearance and 3.5 kg added overhung mass at 8940 rpm
critical with log dec of 1.044

Rotordynam ic Response Plot

Flexible Rotor Model - Tilting Pad Bearings - 3.5 kg Overhung Mass
Sta. No. 8: NDE Brg
18 360
270
16
180
Response, microns pk-pk

14 90 Major Amp
12 0
Horz Amp
-90
10 Vert Amp
-180
8 -270 Horz Phs
6 -360 Vert Phs
-450
4
-540
2 -630
0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 84 Damped response plot for final rotor design with variable geometry
bearings, mean clearance and 3.5 kg added overhung mass at NDE bearing

137
Chapter 5 – Design of Experiment

Rotordynamic Deflected Shape Plot Calculated Shaft Orbit

Flexible Rotor Model - Tilting Pad Bearings -3.5 kg Overhung Mass 8
Deflected Shape at 9600 rpm
0.02 6

Vertical Displacement (microns)

Major Amp
0.015 4
Horz Amp
Response Amplitude, mm pk

2
0.01 Vert Amp
0
0.005
-2
0
0 0.2 0.4 0.6 0.8 1 1.2 -4
-0.005
-6
-0.01
-8
-0.015 -8 -6 -4 -2 0 2 4 6 8
Horizontal Displacem ent (m icrons)
-0.02
Axial Location, m eters

Figure 85 Damped response operating deflection shape and NDE bearing

whirl orbit for final rotor design with variable geometry bearings, mean
clearance and 3.5 kg added overhung mass at NDE bearing at 9600 rpm.

Considering the results of the rotordynamic design and analysis of the test rig,
rotor and bearing configuration with respect to the observations taken from
the literature and those highlighted in chapter 4, the following comments can
be made.

The test rotor design encounters several areas of significant unbalance

response within the running range which are coupled with rotor bending
modes. The rotor is capable of operating above the first critical bending mode
under all bearing and rotor configurations, and runs close to, at or above a
second shaft bending mode. The second shaft bending mode is largely
associated with the overhung section and NDE bearing. The response at the
NDE bearing remains active across the entire range, showing a similar
characteristic to the displayed in the de Jongh et al [11] rotor. This similarity in
characteristics to the de Jongh et al [11] rotor is also apparent in the predicted
mode shapes.

Also in line with observations made from case studies in chapter 4, the
operating deflection shape is associated with a rotordynamic bend at the NDE
bearing location when approaching the second bending mode, where the
inner bend surface is closest to bearing wall when a zero degree phase lag
exists. Therefore the requirement for phase change is present as was
observed in the case study rotors. The rotordynamic predictions of the test
rotor are such that a significant phase change takes place in the high speed
region (~8-12 krpm) which, when coupled with a thermal lag of 20-0 degrees
would give a phase range of over 90 degrees. This should provide a

138
Chapter 5 – Design of Experiment

speed/phase band suitable to traverse a condition where the thermal and

mechanical unbalance components would combine (although this condition
alone does not guarantee instability as suitable conditions for the rate of
thermal differential heating and coupled unbalance response are also
required, See Figure 84 and Figure 85.

The bearing operating eccentricity ratios are predicted to run at low levels of
typically <0.15 for speeds above ~3000 rpm, with the NDE shaft obits in the
regions of high response or critical speed locations having significant circular
form. Also the NDE bearing response is shown to be sensitive to unbalance
located at both the shaft centre, and more significantly, the overhung NDE
flange.

In summary, it has been possible to rotordynamically design a test rotor-

bearing system which is predicted to display the principle governing
characteristics and conditions, as established in chapter 4, that can be
typically associated with machines having been reported as suffering journal
differential heating instability. Both the fixed and variable geometry bearings
show the characteristics sympathetic to being sensitive to the instability. One
feature where some difference between the case study rotors and that of test
rotor exists is in that the surface journal velocities of the test rotor are
somewhat lower than those cases considered in chapter 4. The test rotor
peak journal velocity is ~35 m/s and the case study bearing velocities being
40 to 66 m/s. This effect however is somewhat countered by the fact that
lower clearances employed in the bearings result in similar operating
temperatures within the oil films and also the lower stiffness and damping
properties of the system make it comparative in sensitivity to unbalance than
the larger rotors in the case studies.

5.4.2.3 Rigid Rotor

In addition to the main experimental test rotor, which has been specified as a
flexible rotor in that it runs in a super critical condition, a smaller rigid rotor
was also designed. The concept behind the creation of a shorter rigid rotor
was to allow for operation of the experimental rig using the main test bearings
without any significant influence or feedback mechanism from the
rotordynamic conditions within the rotor shaft. The use of a rigid rotor, which
does not introduce any significant rotordynamic complexities, provides a
means for characterising the thermal journal conditions within the bearing and
any associated relationships with unbalance and shaft journal displacements.

139
Chapter 5 – Design of Experiment

The absence of any significant phase changes also helps to provide

conditions which provide more stable operating conditions on which to found a
baseline. The application of this type of rotor also allows for the journal
temperature measurement instrumentation to be tested and verified.

The rigid rotor is designed such that the rotor journal static load on each
bearing is very close to that of the main test rotor, replicating the basic
conditions within the bearings as for the flexible rotor. The rigid rotor geometry
has been defined such that the rotor does not encounter, with suitable
separation, any shaft bending natural modes within the experimental rig
operating speed range. This is principally achieved by means of reducing the
rotor bearing span, increasing shaft diameter in the centre of the rotor and
reducing the overhung section of the NDE for the rotor. Figure 86 shows the
rotordynamic model of the rigid rotor configuration. It can be seen in Figure 86
that the rotor bearing span has been reduced from 750mm to 430mm.
Rigid Rotor Model - Offset Half Bearings

0.16

0.12

0.08
Shaft Radius, meters

15
0.04
10

-0.04

-0.08

-0.12

-0.16
0 0.1 0.2 0.3 0.4 0.5 0.6
Axial Location, meters

Figure 86 Rigid rotor dynamic model

Rotordynamic analysis of the rigid rotor was again performed in line with the
procedures as outlined for the flexible rotor. Full details of this analysis are not
presented here but the general rotor response characteristics are summarised
in Figure 87 to Figure 91 providing the undamped critical speed map,
Campbell diagram and rotor response to unbalance.

140
Chapter 5 – Design of Experiment

Undamped Critical Speed Map

Rigid Rotor Model - Offset Half Bearings

100000

Critical Speed, rpm

10000

1000

100
1000000. 10000000. 100000000. 1000000000. 10000000000.

Bearing Stiffness, N/m

Figure 87 Rigid rotor undamped critical speed map

Damped Natural Frequency Map - Campbell Diagram

Rigid Rotor Model - Offset Half Bearings
18000 4.000
16000 3.500
Natural Frequency, cpm

14000

Logrithmic Decriment
3.000
12000
2.500
10000
2.000
8000
1.500
6000
1.000
4000
1X
2000 0.500
Log Dec at 1104 rpm
0 0.000
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
Rotor Speed, rpm

Figure 88 Damped natural frequency map for rigid rotor design with fixed
geometry bearings, mean clearance

Damped Eigenvalue Mode Shape Plot Damped Eigenvalue Mode Shape Plot
Rigid Rotor Model - Offset Half Bearings Rigid Rotor Model - Offset Half Bearings

forward forward
backward backward
f=1104.6 cpm f=1120.0 cpm
d=3.762 logd d=3.6471 logd
N=1000 rpm N=1000 rpm

Figure 89 Damped mode shapes for rigid rotor design with fixed geometry
bearings and mean clearance (a) ‘rocking’ mode at 1104 rpm with log dec of
3.762 (b) ‘bounce’ mode at 1120 rpm with log dec of 3.647

141
Chapter 5 – Design of Experiment

Rotordynamic Response Plot

Rigid Rotor Model - Offset Half Bearings
Sta. No. 4: NDE Brg
10 360

270
9
180
8
90
Response, microns pk-pk 7
0
Major Amp

6 Horz Amp
-90
Vert Amp
5 -180
Horz Phs
-270
4 Vert Phs
-360
3
-450
2
-540
1 -630

0 -720
0 4000 8000 12000 16000 20000
Rotor Speed, rpm

Figure 90 Damped response plot for rigid rotor design with fixed geometry
bearings and mean clearance 12.7 g.mm at NDE flange

Rotordynamic Deflected Shape Plot

Rigid Rotor Model - Offset Half Bearings Calculated Shaft Orbit
Deflected Shape at 12500 rpm 5
0.0012
4
Major Amp

Vertical Displacement (microns)

0.001 3
Horz Amp
Response Amplitude, mm pk

2
0.0008 Vert Amp
1
0.0006 0

0.0004 -1

-2
0.0002
-3

0 -4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 -5
-0.0002 -5 -4 -3 -2 -1 0 1 2 3 4 5
Horizontal Displacem ent (m icrons)
-0.0004
Axial Location, m eters

Figure 91 Damped response operating deflection shape and NDE bearing

whirl orbit for rigid rotor design with fixed geometry bearings and mean
clearance at 12500 rpm.

142
Chapter 5 – Design of Experiment

5.4.3 Synchronous Thermal Instability (Morton Effect) Analysis

Having evaluated the rotordynamics with respect to thermal sensitivity of the

rotor, some consideration is given to the thermal evaluation of the rotor. This
consideration is addressed using two approaches. The first was to evaluate
the rotor design in line with the physical, rotordynamic and hydrodynamic
observations taken from the case studies presented in chapter 4. The second
was to apply the appropriate analytical techniques presented in the published
literature. This element of the design evaluation is an important one, in that
not only is it being used as a tool on which consider the design criteria with
regard to the promotion of instability, but it will also provide a benchmark as to
the capability of the existing analytical methods to predict the thermal and
stability response of a real rotor with respect to journal differential heating.

The analytical procedure applied is that as presented by Balbahdur [19] in his

PhD thesis as opposed to the analytical CFD approach applied by Keogh et al
[9]. The principle reason behind this is that the Keogh et al [9] method is a
highly analytically intense process which, without significant mathematical and
computational development, cannot be readily applied to non cylindrical
bearings such as those considered in this work. Also, as a practical
rotordynamic tool it would prove too computationally and time intensive.
Conversely the method presented by Balbadur [19], whilst having some
significant idealisations and assumptions, is more readily applicable. It also
allows for an experimental validation of the presented theory in [19] to be
made both from a dynamic response and journal thermal response
consideration.

Whilst the Balabadur [19] analytical procedure was used as a tool to assess
the detail design of the rotor to thermal instability, the principle design
considerations were based on producing physical conditions in the rotor
operational performance which is consistent or representative with the
observations identified in section 4, which resulted in the rotordynamic design
criteria for the shaft as detailed in section 5.4.2.2.

The definitive criteria in the Balabahadur [19] thermal stability modelling

method for calculating the points of instability, is the resultant level of
unbalance generated by the rotor due to the thermal bow about the bearing
journal. This is represented as a fraction of the rotor mass. Fifteen percent
was the level used by Balbahadur [19] in his research which was based upon
running a number of practical examples through the analysis procedure. For

143
Chapter 5 – Design of Experiment

the purposes of consistency (and in the absence of any other credible data
with which to determine a suitable thermal unbalance threshold level by the
application of this technique) the same 15% threshold was applied but in
combination with a band pass type approach where an upper threshold
criteria of 30% was also applied. This allowed for some indication of how
sensitive to the calculated and input data the predicted stability of the rotor
would be.

5.4.3.1 Fixed Geometry Bearings

System information from the analysis according the Balbahadur [19] method
for the amplitude and phase data for the non-drive end bearing journal was
extracted from the rotordynamic analysis performed in section 5.4.2.2. The
phase data was manipulated to be in a form that could be used in the Morton
Effect stability analysis and other system information such as bearing
dimensions etc included. The analysis was performed using a number of
unbalance values for all of the rotor system configurations, e.g. bearing
clearances added mass etc.

The results of the analysis showed that for the bearing and test rotor design
used, dependant on the level of initial residual unbalanced, all of the rotor
bearing configurations were predicted to go unstable at some point within the
operating speed range. As would be expected, the point at which the analysis
suggested the rotor would go unstable was directly related to areas of high
response. When the unbalance condition is linked to the rotor centre location,
the point of instability can be seen to associate with the first critical speed at
around 6000 rpm. If the unbalance level is significantly reduced, or more
notably moved to the overhung flange, the analysis shows that instability is
reached at some point between the first and second natural frequencies. The
movement in the instability location is not only due to the change in unbalance
location but also the change in rotordynamic conditions used as inputs for the
analysis, i.e. amplitude and phase responses at the NDE bearing location.

The results of some of the analysis predictions are shown in Figure 92

through to Figure 95. Figure 93 and Figure 94 show the stability predictions
for the rotor in fixed geometry bearings (mean clearance) with 20 g.mm and
50 g.mm unbalance respectively located at the NDE flange, with 3.5 kg of
added overhung mass. The rotor unbalance curve is derived from the
calculated thermal bend unbalance for any given speed condition and the
unbalance threshold curve is based on a specified % of the total rotor mass.

144
Chapter 5 – Design of Experiment

Instability is defined as occurring where the rotor unbalance and threshold

curves intersect. It can be seen that instability is predicted between 11000
and 12635 rpm for the 20 g.mm case and 8862 and 10820 rpm for the 50
g.mm case.

The predicted journal temperature differential with respect to speed can be

seen in Figure 95.

Similar results are shown for the variable geometry conditions in Figure 96 to
Figure 99. Figure 97 shows the stability prediction for the tilting pad bearings
running with maximum clearance and a 20 g.mm unbalance force. The
stability range was calculated as 8148 to 8940 rpm. With the Figure 98
showing the same unbalance condition but now for mean clearance
conditions, with instability predicted between 8292 and 9288 rpm. Predicted
journal temperature differentials are given in Figure 96 and Figure 97 for the
maximum and mean clearance conditions respectively.

It should be noted here that for reasons explained in literature review in

chapter 2 this analysis is not exhaustive and remains essentially unproven
with some areas having some significant simplifications and assumptions. As
such the analysis has been used only as an indicative guide in the specific
design of the test rig rotor and bearings, with the actual design criteria also
being strongly based on more practical and operational requirements such as
rotor mode shapes, response phase, shaft orbit geometry and bearing
characteristics as identified by de Jongh and Morton [11].

Figure 92 Shaft orbit showing hot and cold spots on rotor journal for flexible
rotor in fixed geometry bearings

145
Chapter 5 – Design of Experiment

Stability Plot - Fixed Geometry Bearings

Balbahadur Procedure [19]

200 Low er Instability Threshold

180 Rotor Unbalance
Upper Instability Threshold
160
Unbalance (g.mm)

140
120
100
80
12635
60
40 11029

20
0
0 2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 93 Morton stability analysis of fixed geometry bearings with 20 g.mm

unbalance at overhung flange, 3.5 kg added mass and mean bearing
clearance

Stability Plot - Fixed Geometry Bearings

Balbahadur Procedure [19]

200 Low er Instability Threshold

180 Rotor Unbalance
Upper Instability Threshold
160
Unbalance (g.mm)

140
120
100
10820
80
8862
60
40
20
0
0 2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 94 Morton stability analysis of fixed geometry bearings with 50 g.mm

unbalance at overhung flange, 3.5 kg added mass and mean bearing
clearance

146
Chapter 5 – Design of Experiment

Journal Temperature Differential

7

Journal Temperature Difference

6

4
(deg C)
3

0
2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 95 Journal temperature differential fixed geometry bearing at NDE

journal with 50 g.mm unbalance, 3.5 kg added mass and mean bearing
clearance

5.4.3.2 Variable Geometry Tilting Pad Bearings

Thermal Stability Plot - Variable Geometry Bearings

Balbahadur Procedure [19]

200 Low er Instability Threshold

180 Rotor Unbalance
Upper Instability Threshold
160
Unbalance (g.mm)

140
8940
120
100
80
8148
60
40
20
0
0 2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 96 Morton stability analysis of variable geometry bearings with 20

g.mm unbalance at overhung flange, 3.5 kg added mass and maximum
bearing clearance

147
Chapter 5 – Design of Experiment

Journal Temperature Differential

12

Journal Temperature Difference

10

(deg C) 6

0
2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 97 Journal temperature differential variable geometry bearing at NDE

journal with 20 g.mm unbalance, 3.5 kg added mass and maximum bearing
clearance

Thermal Stability Plot - Variable Geometry Bearings

Balbahadur Procedure [19]

200 Low er Instability Threshold

Rotor Unbalance
180
Upper Instability Threshold
160
Unbalance (g.mm)

140
120 9288

100
80
8292
60
40
20
0
0 2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 98 Morton stability analysis of variable geometry bearings with 20

g.mm unbalance at overhung flange, 3.5 kg added mass and mean bearing
clearance

148
Chapter 5 – Design of Experiment

Journal Temperature Differential

10

Journal Temperature Difference

9
8
7
6

(deg C) 5
4
3
2
1
0
2500 5000 7500 10000 12500 15000
Shaft Speed (rpm)

Figure 99 Journal temperature differential variable geometry bearing at NDE

journal with 20 g.mm unbalance, 3.5 kg added mass and mean bearing
clearance

From the results of the thermal instability analysis, in accordance with [19], it
was observed that the rotor would be considered as sensitive to instability and
expected to go unstable within the running range for a moderate level of
mechanical unbalance. The fixed geometry bearing was predicted as having a
higher instability threshold with notably lower journal temperature differentials
being predicted than for the tilting pad. The tilting pad configuration showed a
journal temperature differential prediction of up to 10 deg C. The predicted
temperature differentials for the tilting pad, in the opinion of the author, would
seem somewhat high for the conditions under analysis, however even if
overestimated, the procedure would predict a rotor that is sensitive to the
Morton effect instability.

An interesting point of note is that the rotor without the added mass is shown
to have a lower stability threshold than that with the larger overhung mass.
Whilst this might immediately seem counter intuitive as the mechanical gain
has been increased by the increased overhung mass, the implication is that
the rotordynamic conditions at the NDE bearing are notably different for the
two cases. It is these complex rotordynamic relationships which could result in
the introduction of the overhung mass actually pushing up the threshold
speed by means of altering the mode shape, frequency and response at the
bearing of interest.

It should also be noted that the polynomial expression for steady shaft
eccentricity ratio as a function of applied load was found to be sensitive to

149
Chapter 5 – Design of Experiment

load magnitude. In particular at low loads of specific bearing pressures some

cases would not resolve due to the low journal bearing specific pressures in
the test rotor. In these cases this was overcome by artificially increasing the
load by ~50% (still only a small load). The effect of this is not highly significant
on the steady bearing conditions for higher speeds, although it does result in
a slightly higher journal eccentricity ratio. In the case of running speed at 8000
rpm, this increased the eccentricity ratio from ~0.25 to ~0.3 (max clearance
condition) which will have some impact on the bearing temperature
calculations.

5.5 Detail Design

Following the analytical assessment and key design criteria definition the
practical and detail design aspects of the test rig have to be considered. The
complete test rig configuration has been broken down into the following six
sections:

1. Main Test Rotor and Bearings

2. Drive System
3. Transmission
4. Lubrication System
5. Instrumentation
6. Monitoring, Data Logging and Data Management

5.5.1 Main Test Rotor and Bearings

The principle design features and criteria for the test rotor and bearings have
been covered in the preceding sections in this chapter. However, a number of
design features worthy of mention and are summarised here.

In the design of the rotor, care was taken to ensure that the manufacturing
specifications were consistent with maintaining a good state of balance across
the running range and provide good quality surfaces from which to measure
shaft movements. To provide these conditions the shaft was precision ground
on all diameters to provide both good dimensional control and runout errors.
The shaft was also thermally stress relived at several stages throughout the
manufacturing cycle to provide good shaft stability to residual stress
distortions during finishing machining and also whilst in service.

150
Chapter 5 – Design of Experiment

The location points for the shaft temperature measurements were achieved by
means of providing small axial holes drilled just below the shaft surface into
the centre of the shaft journal. The holes were 4.75 mm diameter and
positioned such that a wall thickness of 2.0 mm was produced from the
journal surface.

The number of sensors to include under the journal surface was an area of
investigation during the early stages of the research. Several publications [11]
indicate that only 4 measurement points are required as the expected journal
temperature distribution would be harmonic as a result of the synchronous
unbalance and associated oil film thickness function. It is only the first
harmonic component that will cause the shaft to develop a steady state
thermal bow. Other higher order harmonic distributions would serve to
produce polar equilibrium stresses which do not result in a geometric
displacement of the shaft overhang mass. Whilst these assumptions hold true
for the analytical problem described, it is quite possible that when using more
complex bearing geometries combined with complex vibration orbits
temperature distributions other than harmonic are quite plausible. To fully
understand the generation of the thermal state in the bearing and rotor journal
a greater number of measurement points could offer increased understanding
of this feature. A downside of considering the inclusion of more than 4
sensors on the shaft diameter is that an increase in sensors results in a
distortion of the thermal mass/inertia of the rotor journal and also the stiffness
of the shaft. Also, whilst with 4 sensors only the first harmonic amplitude can
be reliability attained, it will be possible to observe the level of non
fundamental temperature distribution around the shaft.

Based on the potential practical issues that would be introduced with high
sensor numbers due to the risk to the measurement and modelling accuracy,
also that the necessary fundamental harmonic thermal distribution is the
principle area of interest, the number of sensors included in the shaft journal
was specified as 4.

Detailed manufacturing specifications for the main test rotor and rigid shaft
rotor were designed and produced applying the same philosophy and
processes. Full manufacturing specifications for the test rotors and key test rig
components are provided in appendix B.

151
Chapter 5 – Design of Experiment

Figure 100 Main test (flexible) rotor assembly schematic

Figure 101 Rigid test rotor assembly schematic

5.5.2 Drive System

As the rotor is required to run at a range of speeds some sort of variable

speed drive is required. The drive is to be supplied by electric motor and the
variable speed aspect can be achieved by one of two means. One is by using
a DC motor and thyristor type control, the other is to use an AC induction type
motor and a frequency inverter. The preferable method was to use an AC
induction motor and frequency inverter drive due to the low cost and ready
availability of AC induction motors. However, frequency inverters can be quite
costly for powers of around 7.5 kW. However, due to a fortunate
circumstance, a Jaguar CDII750 7.5 kW frequency inverter became available
and was donated to the project and a 5.75 kW ABB M2AA 112 M4 Aluminium
body electric motor was selected for the prime mover. With the correctly
designed transmission this provided a more than sufficient power supply to
the test rig, covering the required operating speed range.

5.5.3 Transmission

In order to obtain the required maximum rotor running speed of 15000 rpm, a
means of step up transmission was required to increase the motor output

152
Chapter 5 – Design of Experiment

speed from 2160 rpm. Two options were considered to achieve this, one
being the use of a set up gearbox and the other use of a belt drive
arrangement. For reasons of reduced cost, complexity and lead time a belt
drive system was selected. The operating speed and powers are somewhat
outside the normal operating bounds for most belt drives. As such designing
and sourcing suitable components presented some problems.

One supplier of high performance belts (Conti-Tech) was, after some

discussion, able to offer a solution with a pulley wheel ratio of 9:1, achieved in
one stage. The running conditions of the belts was outside normal operating
limits but for the purposes of research purposes the predicted belt life of the
belts was sufficient to meet the estimated testing requirements.

The belt arrangement selected was a multi ribbed V belt ContiTech 1143/540
J12. The arrangement consists of two drive wheels one of 30mm diameter
and another of 270mm diameter. The smaller wheel had to be specially
designed due the high speeds, high ratio and bearing arrangement required
which was performed by the author and manufactured by Allen Gears.

The drive transmission is coupled to the main rotor via a flexible element
coupling. The coupling is a multi-disc pack type with one set of flexible
elements at each end, capable of accommodating both offset and axial
misalignment errors up to 2.5 and 1.0 mm respectively. For practical reasons
the coupling size had to be 80mm diameter and as such was capable
transmitting much higher torques than required.

Transmission
Arrangement

Figure 102 Test rotor and transmission assembly

153
Chapter 5 – Design of Experiment

5.5.4 Lubrication System

The lubrication system consists of a relatively straight forward arrangement. It

comprises of an electric motor driven pump, filter, oil heater, cooling system
and pressure control valve. The supply pump was an existing piece of
equipment available through Allen Gears with a pumping flow capacity of 70
l/min with a peak discharge pressure of 6.0 bar.

The filter is a 10µm PAL cylinder cartage filter with a 150 l/min flow capacity
and is again an existing piece of equipment available at Allen Gears.

The oil heater is a Chromaloy low power 4.0 kW immersion heater with a
variable thermostat, and is located in the base of the lubricating oil test stand
sump.

Oil cooling capability is provided by means of a Boman 1425-2 shell-tube

cooler with a 8 kW(max) heat rejection capacity, which uses pressure fed
fresh water cooling media. It was specifically selected and procured
specifically for use with the experimental test apparatus.

The oil tank forms part of the main test stand base and has a maximum oil
capacity of 70 litres. The oil type used for all of the commissioning and
experimental activity was an ISO VG46 viscosity mineral oil with properties as
given in Table 16. This particular viscosity grade of oil was selected as it is a
commonly used oil type in the types of high speed turbo machinery observed
to be sensitive to the Morton Effect, such as compressors, steam turbines,
turbochargers gearboxes etc.

Manufacturer : Castrol
Brand : Hyspin AWS 46 – Hydraulic Mineral Oil

Kinematic Viscosity at 40 Deg C 46.0 Cst

2
Dynamic Viscosity at 40 Deg C 3.96E-2 N.s/m
Relative Density at 40 Deg C 0.857
Specific Heat Capacity at 40 Deg C 1.947 kJ/kg Deg C
Kinematic Viscosity at 100 Deg C 6.94 Cst
2
Dynamic Viscosity at 100 Deg C 5.683E-3 N.s/m
Relative Density at 100 Deg C 0.819
Specific Heat Capacity at 100 Deg C 2.165 kJ/kg Deg C

Table 16 Lubrication Oil Summary

154
Chapter 5 – Design of Experiment

5.5.5 Instrumentation

5.5.5.1 Shaft Vibration

The method of shaft vibration and displacement measurement was achieved

by means of non contacting eddy current proximity probes. The specific probe
types selected were 5.0mm tip diameter SKF M65 probes and extension
leads used in conjunction with SKF osolator demodulator proximiter probe
drivers. The output of the proximity probes is a voltage signature which
consists of an ‘AC’ component (the shaft vibration) and a DC component
(shaft position or gap) which can be directly input to the DAQ equipment. A
total of ten proximity probes were specified to measure four different positions
along the test rotor length ; DE bearing, shaft centre, NDE bearing and NDE
flange (Figure 103).

Figure 103 Test rotor shaft proximity probe locations

The bearing journal probes were mounted directly to the bearing shells to
minimise the effects of relative thermal growth. At the NDE test bearing, four
probes were configured in a ‘push – pull’ arrangement to minimise the errors
in shaft position measurement introduced by relative thermal growth of the
shaft and bearing/housing. The application of the push pull arrangement
allows for direct measurement of the relative change in bearing operating
clearance and provides an approximate measurement of the operating oil film
thickness. The probes were mounted as close as practical to the bore of the
bearing to reduce the effects of thermal dimensional changes on the gap
measurement due to bearing shell temperature. Figure 104 shows a
schematic of the probe arrangement on the test bearing.

155
Chapter 5 – Design of Experiment

Figure 104 Push-Pull Test Bearing Proximity Probe Arrangement

(Note: the diagram shows the fixed geometry bearing but same principle applied on tilting pad bearings)

5.5.5.2 Bearing Temperature Measurement

Bearing temperature measurement was made by using simplex PT100

Resistance Temperature Detectors (RTD’s), 3 wire type. These sensors are
located axially in centre of the bearing at a radial position of 2mm below the
surface of the babbitt. The location holes for the RTD canisters are produced
to provide a tight fit on the diameter of the sensors to aid heat transfer in
combination with a heat transfer compound during installation. Both the test
bearing and the drive end bearing have provision for 10 sensors position
circumferentially around the bearing. The NDE test bearing was fitted with
RTD sensors at all 10 locations and the DE bearing being fitted with only four,
see Figure 105.

RTD Sensor
locations

NDE Test Bearing DE Bearing

Figure 105 Bearing Temperature RTD Sensor Locations

156
Chapter 5 – Design of Experiment

The RTD sensors fitted to the tilting pad bearing were again axially located at
2mm below the babbitt surface at the leading and trailing edges of the pad,
with 10 (2/pad) sensors again fitted to the test bearing and four in the DE
bearing. Sensors were located at the leading and trailing edges of each pad.

5.5.5.3 Shaft Temperature Measurement

The most challenging measurement to be made was that of the shaft journal
temperature distribution due to the requirement to provide power to and
receive a signal from some kind of temperature measurement sensor in the
high speed rotating shaft. A number of options were explored to achieve this
objective. The initial choice would have been to use the PT100 RTD sensors
as used in the bearing, as these have a good inherent absolute accuracy and
an adequate operational temperature range. However the use of these
sensors requires a low noise, low resistance method of electrical coupling to
the sensor due to the very low inherent resistance levels and sensitivity of
measurement to resistance change. To practically achieve this for the
accuracy levels required, it would necessitate the use of a wireless option due
to the concern that surrounded the use of an electrical slip ring with RTD’s
(i.e. slip ring noise and resistance). The possibility of such a telemetric option
was explored with several specialist instrumentation providers, the conclusion
of which was that due to the high operating speed requirements, whilst
achievable, the associated equipment and signal conditioning costs would be
high and beyond the restrictive budget of this work.

Alternative options as to how to achieve the required measurements without

compromising on the required measurement accuracy capability were
continued, with focus being around a solution which allowed the use of a high
speed low noise electrical slip ring. The conclusion of the work was that the
use of high quality ceramic thermistors could provide a suitable solution to
conduct the required journal measurements. The selection of thermistors was
made due to the high resistance and low sensitivity to signal noise
characteristics. This allowed the concerns relating to the electrical slip ring to
be negated, with typical slip/ thermistor ring resistance ratio of approximately
1.35x10-4 without any requirement for significant on rotor signal conditioning.

A principal disadvantage of selecting thermistors is that a suitable usable

sensor construction is not readily available on the market. As such the design
and construction of the sensors had to be carried out specifically for the
project.

157
Chapter 5 – Design of Experiment

The sensor construction used miniature high quality 100kΩ @ 20oC

NTC(Negative Temperature Coefficient - reducing resistance with
temperature) thermistors with minimum absolute accuracy levels of +/-0.2oC,
and sensor repeatability within <0.01oC within a temperature range of 20oC to
100oC. Small brass canisters were manufactured to house the sensors with
the sensor ‘potted’ in the canister with a high integrity epoxy resin. Special
attention was applied during the ‘potting’ process to ensure no air pockets
were present in the epoxy so as not to effect the heat transfer within the
sensor. PTFE high temperature oil resistant signal sheathing was used to
insulate the signal wires and then sealed in high temperature oil resistant heat
shrink sleeve. Figure 106 shows the basic thermistor sensor construction.

Brass Sensor
+ signal
Canister Insulation
(insulated)
Sleeve

Thermistor Epoxy Common

Material Resin (insulated)

Figure 106 Thermistor sensor basic construction

Figure 107 Thermistor sensor assembly

Several problems are presented with the use of thermistors as the method of
temperature measurement. The principle difficulties are the non-linear
relationship with temperature and the second is how to obtain a suitable
output voltage or current than can be recorded.

To enable a suitable output signal relating to the thermistor resistance value

to be obtained, a simple signal conditioning circuit was designed and

158
Chapter 5 – Design of Experiment

constructed. The circuit was based on a traditional Wheatstone bridge circuit

consisting of 3 fixed high accuracy resistors. The bridge was configured such
that the thermistor acted as a variable resistor in the forth leg of the circuit,
see Figure 108. Each thermistor had a dedicated bridge circuit with a
common power supply rail. The signal conditioning board was configured
such that removal links allowed for direct measurement of the fixed resistors
to enable accurate quantification of the resistance of each leg of the bridge,
minimising the residual errors measured for the thermistor resistance. The
bridge circuit shown in Figure 109 was used for each thermistor, and the
measured resistance of the thermistor for any given temperature condition
was establish by the application of the relationship in equation 5.5.1.

Removable Links to
enable accurate
individual resistor
mΩ measurements

Figure 108 Thermistor Wheatstone bridge signal circuit

(signal conditioning unit not shown)

R1 R2

VG

R4
R3

- +
Vs
Figure 109 Thermistor Wheatstone bridge configuration

159
Chapter 5 – Design of Experiment

 VG   R2 .R3 
  .R3 +  
 VS   R1 + R2 
Rtherm = (Eq. 5.5.1)
 R2   VG 
1−  − 
 R1 + R2   VS 

where Rtherm = unknown thermistor sensor

R1 , R2 , R3 = Fixed resistors

The gauge voltage from Figure 109 was taken as the input signal to the data
acquisition as a measure of the thermistor resistance.

The second problem presented by the application of thermistors is in two

parts. The highly nonlinear response of the sensors resistance change to
temperature presents difficulties in practically interpreting the corresponding
bridge voltages to a temperature accurately without the need for cumbersome
or inaccurate discreatised data tables (Figure 110). Additionally the absolute
accuracy of the sensor across the temperature range of interest can result in
accuracies of +/- 0.2 degrees. This level of variability is outside the required
measurement accuracy as outlined earlier in this chapter.

As Supplied Thermistor Accuracy

1.50
1.25 + Error
Error Due to Resistance Tolerance

1.00 - Error
0.75
0.50
0.25
(Deg C)

0.00
-0.25
-0.50
-0.75
-1.00
-1.25
-1.50
20 30 40 50 60 70 80 90 100
Temperature (Deg C)

Figure 110 As supplied thermistor accuracy due to resistance variance

By using the inherent high degree of repeatability of the thermistor sensors, a

technique was used to analytically characterise the resistance-temperature

160
Chapter 5 – Design of Experiment

relationship which allowed for the direct determination of temperature for a

given measurement of resistance or equivalent voltage. The method applied
can also result in the ability to significantly increase the absolute accuracy of
the sensor by obtaining an accurate analytical description of the resistance
curve for each sensor. The technique used was that as presented by
Steinhart et al [56] and is commonly known as the Steinhart-Hart equations. It
states that the application of this relationship can increase the absolute
accuracy of any given thermistior to +/-0.01oC or better over a 200oC
temperature range. The principle equation presented by Steinhart et al is
derived from empirical curve fitting of the resistance-temperature
characteristic of thermistor devices and described using the natural log of
resistance relative to the inverse of temperature, where (1/T) is expressed in
terms of a polynomial expression as :

1 N
= A0 + A1 ( ln ( R ) ) + ........... + AN ( ln ( R ) ) (Eq. 5.5.2)
T

where T = temperature in degrees Kelvin

A0 to AN are polynomial coefficients

It is presented in the accepted literature [57] that a 3rd order polynomial is able
to accurately describe the nonlinear response for any specific thermistor and
equation 5.5.2 can be reduced to equation 5.5.3 and is known as the
Steinhart-Hart equation.

1 3
= C A + CB . ( ln ( Ri ) ) + CC . ( ln ( Ri ) ) (Eq. 5.5.3)
Ti

The coefficients C A , CB and CC are the Steinhart-Hart constants for the

specific thermistor which requires analytical modelling. The coefficients for
each specific thermistor need to be established and this can be achieved by
accurate determination or specification of the resistance of the thermistor
within the desired temperature measurement range. For example, for a given
temperature resistance curve such as that in Figure 111, the resistance for
three specific temperature conditions can be taken. With this data it is
possible to solve for the three unknown coefficients by constructing 3 sets of
linear simultaneous equations. To provide an efficient and direct solution,
expressions for the coefficients were developed within the thesis and following

161
Chapter 5 – Design of Experiment

some work the following equations 5.5.4, 5.5.5 and 5.5.6 are presented to
express each of the coefficients for a given thermistor sensor.

1 1 1 1
 −  . ( ln ( R3 ) − ln ( R2 ) ) −  −  . ( ln ( R2 ) − ln ( R1 ) )
CC =  T2 T1   T3 T2 
( ) ( )
( ln ( R3 ) − ln ( R2 ) ) . ln ( R2 ) − ln ( R1 ) − ln ( R3 )3 − ln ( R2 )3 .( ln ( R2 ) − ln ( R1 ) )
3 3

(Eq. 5.5.4)

1 1
( 3
 −  − CC . ln ( R2 ) − ln ( R1 )
T T1 
3
)
CB =  2 (Eq. 5.5.5)
( ln ( R2 ) − ln ( R1 ) )

1 3
CA = − CB .ln ( R1 ) − CC .ln ( R1 ) (Eq. 5.5.6)
T1

Temperature-Resistance Relationship for a Thermistor

10000
Resistance (kOhms)

1000

100

10

1
-50 -25 0 25 50 75 100 125 150
Temperature (deg C)

Figure 111 Typical thermistor temperature-resistance curve

With the three Steinhart-Hart coefficients it is possible to determine the

temperature for each thermistor accurately based on the measured
resistance.

162
Chapter 5 – Design of Experiment

Applying and solving the Steinhart-Hart equation for each thermistor allows
the associated coefficients to be determined. To be able to solve the
equations, the resistance of each sensor at three temperatures needs to be
accurately known. The method used to determine this data was to measure
the resistance response of each thermistor against the temperature
measurement of a ‘master’ sensor. The sensor used to provide the master
measurement for the purposes of calibration was a specially calibrated PT100
RTD which had been previously calibrated to an accuracy measurement of +/-
0.01 oC.

Calibration

A calibration block was constructed into which the four themistor sensors
were mounted along with the master RTD and two further PT100 RTD’s
Figure 122. The calibration block was a steel disc which was introduced to
provide a degree of thermal inertia to the measurement point. This was to
reduce any effects caused by air movement and forced convection within the
heating medium that could be experienced if the sensors were exposed
directly to the heating atmosphere. All four shaft sensors were calibrated
together in the same runs to enable a high degree of relative accuracy in that
any absolute errors in the master RTD were principally negated by the fact
that all thermistor sensors and the RTD were at the same temperature
reading, irrespective of the small absolute error in the master PT100.

Figure 112 Heat sink block used for thermistor sensor calibration

The calibration block was placed into an accurately controlled heat treatment
furnace. All sensors were monitored and recorded during a heating cycle in
which the furnace temperature was steadily raised to 100oC or just over. Once
at temperature the furnace was switched off and all forced circulation

163
Chapter 5 – Design of Experiment

disengaged. The furnace temperature allowed to cool slowly down to ambient.

The typical time required to heat the furnace and allowing to cool to ~25oC
took around 22 hours. Figure 113 shows a typical calibration cycle for the
thermistors, the temperature shown is that of the master sensor.

Furnace Heat Profile

Calibration Run

100
Reference
Sensor

80
Temperature (Deg C)

60

40

20

0
0.00 200.00 400.00 600.00 800.00 1000.00 1200.00 1400.00
Time (mins)

Figure 113 Time-temperature curve for thermistor calibration

From the measured calibration data the three thermistor resistances were
obtained and the Steinhart-Hart constants established for each sensor. The
constants were taken to give an operating temperature range of the sensor of
25oC to 100oC. Table 17 gives the three temperature constants obtained for
both sets of thermistor sensors for the rigid and flexible rotors.

Steinhart-Hart Constants
Sensor CA CB CC
G(1) 0.007744 -0.00444 4.3643E-07
B(1) 0.007635 -0.00043 4.0807E-07
R(1) 0.007739 -0.00044 4.3356E-07
W(1) 0.007691 -0.00043 4.2176E-07
G(2) 0.000884 0.00020 7.3626E-08
B(2) 0.000818 0.00021 4.6681E-08
R(2) 0.000901 0.00020 7.9879E-08
W(2) 0.000843 0.00021 5.9158E-08
Table 17 Thermistor sensors Steinhart-Hart constants

To establish the functional accuracy and repeatability of the thermistor

sensors when applying the Steinhart-Hart coefficient from Table 17, several
repeat heat cycles were carried out in the furnace and the measured
thermistor temperature computed with the previously determined Steinhart-
Hart coefficients. The computed thermistor temperatures were compared with

164
Chapter 5 – Design of Experiment

the master sensor (RTD PT100) temperature. Also the relative temperatures
of each of the thermistors sensors were compared to each other. The most
relevant measure of accuracy for the measurement of interest is the relative
accuracy of the thermistors sensors. A plot of the temperature measurement
error for each of the thermistors against the master sensor is presented in
Figure 114 to Figure 117. It can be seen that for the temperature range
between 30oC and 75oC, all sensors are within +/- 0.025oC whereas at the
higher temperatures the accuracy levels reduce slightly up 100 oC. The data
shown in Figure 114 to Figure 117 is for the sensors used in the flexible rotor.
However similar characteristics were obtained for the sensors used in the rigid
rotor and the accuracy ranges defined were consistent with those presented
in Table 18 between the temperatures of 25 and 80 oC. The temperature
measurement error is defined as Terror is defined as:

Terror = Tsensor − Tmaster

where Tsensor = the as measured temperature from the thermistor

Tmaster = the temperature as measured with the master sensor

Table 18 summarises the derived sensor accuracies when considered across

several temperature ranges. Each of the four different thermistor sensors is
identified by the colours green, blue, red and white.

Thermistor Sensor Accuracy

Green Thermistor Sensor
0.1

0.075

0.05
Temperature Error Terror Deg C

0.025

-0.025

-0.05

-0.075

-0.1
30 40 50 60 70 80 90
Temperature Deg C

Figure 114 Thermistor sensor error measurement (Green sensor)

165
Chapter 5 – Design of Experiment

Thermistor Sensor Accuracy

Blue Thermistor Sensor
0.1

0.075

0.05
Temperature Error Terror Deg C

0.025

-0.025

-0.05

-0.075

-0.1
30 40 50 60 70 80 90
Temperature Deg C

Figure 115 Thermistor sensor error measurement (Blue sensor)

Thermistor Sensor Accuracy

Red Thermistor Sensor
0.1

0.075

0.05
Temperature Error Terror Deg C

0.025

-0.025

-0.05

-0.075

-0.1
30 40 50 60 70 80 90
Temperature Deg C

Figure 116 Thermistor sensor error measurement (Red sensor)

Thermistor Sensor Accuracy

White Thermistor Sensor
0.1

0.075

0.05
Temperature Error Terror Deg C

0.025

-0.025

-0.05

-0.075

-0.1
30 40 50 60 70 80 90
Temperature Deg C

Figure 117 Thermistor sensor error measurement (White sensor)

166
Chapter 5 – Design of Experiment

The repeatability of the sensor accuracies was also verified by repeating the
furnace heating cycle two further times. In each of the following cases the
temperatures were maintained within the accuracy ranges identified in Table
18.

Temperature Range / Measurement Accuracy and

Repeatability
o
30-75 C 75-85 oC 85-95 oC
Thermistor + oC - oC + oC - oC + oC - oC
Green(1) 0.025 0.025 0.050 0.025 0.075 0.025
Blue (1) 0.025 0.025 0.025 0.050 0.025 0.075
Red (1) 0.025 0.025 0.050 0.025 0.075 0.025
White (1) 0.025 0.025 0.025 0.050 0.025 0.075
Table 18 Individual sensor calibrations/accuracy for flexible and rigid rotor

The thermistor signals were fed to the rotor via a 12 channel (24 way) high
speed IML-Moog slip ring. The arrangement in which the signal wires were
coupled between the slip ring and rotor underwent a number of design
iterations. Two key elements were, firstly the means by which the slip ring and
rotor were coupled and the second was the means of making the connections
between the signal wires to the slip ring. As it was a requirement to be able to
remove the rotor from the rig several times a relatively quick and reliable way
of joining the signal wires was important.

The coupling was achieved by a custom designed and manufactured coupling

which is based on a coil spring with coupling ferrules on each end. This
allowed for the necessary drive to be obtained, whilst accommodating any
static of dynamic misalignment between the two components. The means of
connecting the signal wires was more challenging with several problems
incurred during the commissioning and testing phases. Early designs were
based around the signals being made by means of soldering joints in the rig.
However, following a number of failures, it was concluded that a more robust
and quick release method of coupling the sensors was needed. The use of a
pin type connection was finally selected with a specialist connector and
aluminium housing assembly produced. The rotor side housing was made as
a split aluminium housing to both provide ease of access to the signal cables
during initial fitting and the use of aluminium limited the impact of the mass of
the instrumentation assembly. Figure 118 shows the failure of the initial
soldered arrangement and Figure 120 is the final design arrangement used
throughout the testing phase.

167
Chapter 5 – Design of Experiment

Figure 118 Original slip ring coupling signal wire failure

Figure 119 Schematic of final slip ring instrumentation design

Figure 120 Final slip ring instrumentation design

168
Chapter 5 – Design of Experiment

Figure 121 ‘Potting’ of signal wire soldered joints to prevent fatigue failures

5.5.5.4 General Measurements

A number of other sensors were also used across the rig predominantly
associated with the lubrication oil supply. Table 19 summarise the
instrumentation and sensors used across the test apparatus. The sensors
locations can be seen in Figure 122.

Table 19 Test rig sensor summary

5.5.6 Monitoring, Data Logging and Data Management

To enable the desired number of channels and sensor types to be monitored,

a number of different systems were used to perform the data logging and
monitoring. The most desirable situation would be to use a common data
logger to perform all the recordings of signals, however this would require
some 42 channels covering 8 sensor and signal types which was neither
available to the researcher or affordable within the research project. By using
multiple systems care had to be taken as a number of problems had to be

169
Chapter 5 – Design of Experiment

overcome to eliminate issues with interference and incompatibility between

different equipment. In total four different logging systems were employed.
These systems covered both low frequency logging and high frequency
vibration data.

The four systems used were :

• Newlyme Special Purpose DAQ System (NL)
• Anvil Instruments General Purpose DAQ System (AV)
• Sony EX Dynamic Signal Recorder (SX)
• OROS Dynamic Signal Analyser (OROS)

The Newlyme (NL) system is an existing data logging system at Allen Gears.
The system has a specific number of dedicated channels with built in signal
conditioning for certain sensor types. The system is used to provide the active
monitoring system by which the test rig is controlled and monitored. All
measured parameters (with the exception of the shaft temperature sensors)
are displayed on the system and a basic digital log is generated. Using the NL
system, peak-peak shaft vibration readings are acquired. No frequency data is
processed within NL.

The digital recordings of all the temperature sensors (including shaft journal
sensors) are fed into the Anvil instruments DAQ module. This is an existing
piece of equipment at Allen Gears. The system records all of the RTD and
Thermocouple readings digitally to the computer hard disk at a sampling
frequency of 1s/s. Signal conditioning for the RTD’s and Thermocouples is
managed within the instrument itself, making calibration a more
straightforward procedure. The shaft temperature measurements are supplied
as DC voltages to the instrument via the signal conditioning circuit outlined in
the previous section.

The high frequency data was acquired using the Sony Ex (SX) which is a 16
channel dynamic digital data recorder and analyser. This unit is used to give
basic frequency information whilst acquiring data and has limited amount of
analysis capability. The data recorded is later post processed using Matlab to
provide frequency spectra, shaft orbits and phase relationships etc. The
dynamic signals for the proximity probes, accelerometers and key phasor are
recorded with the SX.

170
Chapter 5 – Design of Experiment

Sampling Rate
DAQ System Channel Ref Sensor/Reading Type
(Hz)
Sony (SX) chan 1 - 10 Proximity Probe - vibration 10k
chan 11 - 14 Proximity Probe - DC gap 2k
chan 15 Accelerometer 25.6k
chan 16 Key phasor 10k
Oros (OROS) Chan 1 - 10 Proximity Probe - vibration 5.12k *
chan 11 Key phasor 5.12k
Newlyme (NL) chan 20 - 30 Proximity Probe - vibration 1 **
chan 1-4 Thermocouples - Lube sys 1
chan 32 Speed tachometer 1 ***
chan 60 Pressure - Lube sys 1
chan 70 - 84 RTD - Bearing temperature 1
Anvil (AV) chan 1 - 12 RTD - Bearing temperature 2
chan 13 - 16 Thermistors - Shaft temperature 2
chan 17 - 20 Thermocouples - Lube sys 2
chan 21 Pressure - Lube sys 2
chan 22 Speed tachometer 2

* Time domain filtered 1X is logged every 1s

** Signal is pre-conditioned at a sample rate of 5k Hz to determine pk-pk
** Tachometer unit acquires at higher rate to provide output DC voltage proportional to speed

Table 20 Data acquisition rates for all mesasurement equipment and channels

Figure 122 Pipework and instumentation layout for test rig

171
Chapter 5 – Design of Experiment

An OROS dynamic signal analyser was used to track and record the filtered
synchronous order of the shaft vibration. The shaft vibration 1st order was
filtered and tracked with the processed information presented both visually
live to the monitor screen and recorded digitally to disk. Post processing of
this data was later performed using Matlab and Microsoft Excel.

The high volume of data acquired during the testing was stored and
referenced within a structured file naming system and entered into an access
database for ease of future referencing.

172
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Chapter 6

Experimental Rig Commissioning

and Test Procedure

6. Experimental Rig Commissioning and Test Procedure

6.1 Introduction

This chapter describes the physical test apparatus used for the experimental
testing, and details the commissioning and system calibration verification
procedures applied. The test rig arrangement is consistent with the design
specifications presented in chapter 5. An overview of the testing programme
is also provided in this section which specifies the various test conditions and
variable parameters that were defined at the outset of the experimental
programme. Whilst testing was underway results were reviewed and the test
programme was modified. This is reflected in the information provided here.

The process of commissioning is defined as the activities of test rig

installation, instrumentation verification, measurement and sensor calibration
and finally principle characterisation of the operational equipment.
Characterisation is principally associated with demonstrating that the test
equipment performs as expected for fundamental operating parameters when
compared to the design assumptions and predictions.

6.2 Test Rig Preparation and Assembly

The experimental test arrangement consists principally of the main test rotor
support stool assembly, the transmission drive motor assembly, drive control
panel and data acquisition equipment. Figure 123 shows a general
arrangement of the test equipment key components. The fully assembled test
rig is shown in Figure 124 to Figure 127.

173
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Figure 123 Test Rig General Arrangement

During the initial commissioning of the test apparatus the rigid rotor was used
to verify the functionality of the instrumentation, drive system and lubrication
system. This was then directly followed by the rigid rotor test programme. The
flexible rotor commissioning took place later in the experimental work once the
rigid rotor-fixed geometry bearing work had been completed. Figure 124 to
Figure 127 show the test rig assembly with both the rigid and flexible rotor
assemblies.

1 2 3

Figure 124 Test Rig Basic Equipment (1)

174
Chapter 6 – Experimental Rig Commissioning and Test Procedure

4 5 6 7 8
9
10

14 15
11 12 13

Figure 125 Test Rig Basic Equipment (2)

1. Test Rotor Control Panel

2. Monitoring Stations
3. Data Acquisition and Instrumentation Equipment
4. Electric Drive Motor
5. Pulley Transmission
6. Test Rotor Labyrinth Seal
7. Test Rotor (Dummy)
8. Test Rotor Bearings and Pedestals
9. Slip Ring Assembly
10. Test Rotor Housing and Base Plate
11. Test Rotor Support Stool and Lube Oil Sump
12. Lube Oil Inlet and Pressure and Thermocouple Sensors
13. Lube Oil Inlet Pressure Control Valve
14. Electric Motor Pump Set
15. Lube Oil Element Filter

175
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Figure 126 Test rig with main rotor housing and guards fitted

Figure 127 Test rig with main flexible rotor

As part of the commissioning process, several notable activities were

performed prior to the final assembly or operation of the rig that could have a
significant influence on the rotor performance. A brief detail of how these
aspects were considered and controlled is covered in the next few sections.

176
Chapter 6 – Experimental Rig Commissioning and Test Procedure

To ensure that no significant parasitic and/or undesirable rotordynamic

features were introduced to the test rotors or rig, it was imperative that the
alignment and unbalance states were defined, controlled and understood.

6.2.1 Rotor Balancing

To provide the best possible initial dynamic conditions for both test rotors they
were dynamically balanced as discrete components using an independent
Shenk multi plane dynamic balancing machine to achieve the lowest practical
levels of residual mechanical unbalance. The rigid test rotor was twin plane
balanced with corrections made by removing material by grinding small
chamfers on the corners of the large outside diameters at two correction
planes. The rotor was balanced at a speed of 3000 rpm with the final residual
unbalance values of 0.97 g.mm and 2.1 g.mm at the two planes.

The main flexible rotor was initially single plane balanced to get the rotor to
such a state where the residual unbalance level permitted higher speed
balancing to be performed. The initial single plane balancing was focused on
the large diameter at the shaft centre where the largest level of unbalance
would be expected. The large centre diameter is also the point at which the
rotor is most sensitive to the 1st bending natural mode. By single plane
balancing at this location it became practical to run the rotor up the necessary
balancing speeds for complete balancing of the rotor.

Once the initial balancing was completed, the rotor was twin plane balanced
with corrections made in three planes in total. As the rotor was running above
its first natural bending mode it was necessary to balance at multiple locations
and this was achieved be means of an iterative balancing procedure. Again
balance corrections were made by means of material removal with chamfers
ground on the corners of the various shaft diameters. The final mechanical
residual unbalance levels at the bearing locations was 0.039 g.mm and 0.028
g.mm for the DE and NDE journal respectively and 0.033 g.mm when run at a
balance speed of 2500 rpm (below first bending critical).

Balancing of the flexible test rotor was repeated with the two overhang
weights fitted to shaft. For these cases, balancing was performed below the
predicted rotor critical speeds with correction material removed only for the
overhung mass discs by means of axially drilled holes.

177
Chapter 6 – Experimental Rig Commissioning and Test Procedure

6.2.2 Test Rig Alignment

As with the rotor unbalance the alignment of the test rotor to the drive
transmission is a critical feature that influences the rotor response and
excitation. To minimize any dynamic forces generated from the flexible
element coupling between the rotor and pulley pinion, alignment was tightly
maintained during the set up process. To further minimise the magnitude of
the dynamic forces presented to the test rotor, the number flexible element
blades used in the coupling was reduced. The flexible element membrane
blades were reduced to the practical minimum for the level of torque being
transmitted which gives a lower angular stiffness and hence lower parasitic
force transmission where operating misaligned. Care was taken when
selecting the final number of blades so as not to introduce any dynamic issues
relating to the axial vibration mode of the coupling.

The alignment procedure was performed with the test rotor as the fixed
component and the transmission and motor aligned to it. The alignment of the
transmission to the rotor was positioned to give both correct vertical,
horizontal and face alignments. The height adjustment was achieved by
grinding the transmission support feet to the desired dimensions. Allowance
was made for the rise in the bearing when operating and thermal growth of the
support pedestal due to the rotor operating temperature. The final alignment
figures were as shown in Figure 128.

Figure 128 Test rig transmission to test rotor alignment readings

A similar alignment procedure was repeated for the low speed pulley to motor
coupling; however it is noted that the alignment at this interface would have a
far less significant effect on the test rotor dynamic performance.

178
Chapter 6 – Experimental Rig Commissioning and Test Procedure

6.2.3 Mechanical – Electrical Runout Measurement

In the areas where shaft proximity probes observe the running surfaces of the
shaft a pre-assembly check was performed to measure the level of residual
mechanical and electrical runout or ‘glitch’ on the rotors (shaft ‘Glitch’ is
discussed in more detail in chapter 7). The rotors were mounted in a vee
block assembly and the level of mechanical runout was measured with a
traditional dial test indicator (finger type). Finally the combined electrical-
mechanical runout was measured using a master proximity probe, with the
results recorded on a paper trace as the shaft was slowly rotated. The initial
results showed that the residual electrical runout was over 8 microns, to
improve this, the observed surfaces were burnished in a grinding machine
using a diamond burnishing tool to a surface finish of ~0.4µm. The process of
burnishing improves the homogeneity of the magnetic signature from the shaft
surface and results in lower electrically measured variations by the proximity
probe.

Figure 129 Electrical Mechanical Runout for Rigid Rotor

NDE Observed Surface

The burnishing process was able to reduce the electrical-mechanical runout in

all cases but some magnitudes (although low by typical industrial standards)
remained higher than desired. To avoid any further machining operations it
was decided these residual errors would be compensated for when in the test
rig. Detail of the principle and procedure to accomplish this is provided later in
this and the following chapter.

6.2.4 Bearing Geometry Measurement

An important feature to verify was the as manufactured sizes and geometry of

the bearings. With both the fixed geometry offset half and tilting pad bearings,
the topology of the bearing geometry is of critical importantance with respect

179
Chapter 6 – Experimental Rig Commissioning and Test Procedure

to both the running responses of the rotor and the thermal conditions within
the bearing. It has been observed from the published literature and the case
studies presented in chapter 4 that the operational clearance of the bearing
can have a significant influence on the sensitivity to the thermal instability.

The fixed geometry bearings were measured by using a specific process to

ascertain the level of pre-load offset and actual joint and crown clearances
within the bearing. The procedure was based around obtaining a number of
clock readings around both the bore and outside diameter of the bearing and
using this data to curve-fit the results to obtain the cylindrical pad geometry
with relation to the bearing datums.

The as manufactured geometry of the bearing was obtained by means of

clocking the bearing on a rotary table and measuring the deviations of the
clock with respect to the datum of the bearing assembly. Analytically these
readings were then curve fitted using a least squares fit procedure to
determine the effective diameter of each pad, its magnitude of offset and
rotational error relative to the datum halving joint of the bearing. Figure 130
provides a schematic of the principal applied. The set of equations in matrix
form given in equation 6.2.1 are shown to express the least squares fit for the
top and bottom pads of the bearing. Details of the derivations are given in
appendix A.

Figure 130 Offset Half bearing geometric measurement relationships

180
Chapter 6 – Experimental Rig Commissioning and Test Procedure



n ∑ cos (θ ) ∑ sin (θ )  
 
Cr  
 
∑ CH 
 ∑ cos (θ ) ∑ cos (θ ) ∑ sin (θ ) .cos (θ ) .  χ r .cos (θ ) = ∑ CH .cos (θ )
2

 ∑ sin (θ ) ∑ sin (θ ) cos (θ ) ∑ sin 2 (θ )   χ r .sin (θ )   ∑ CH .sin (θ ) 


(Eq. 6.2.1)

where n = number of data points

From equation 6.2.1 the equations can now be solved for

Cr , χ r .cos (φm ) , χ r .sin (φm ) and χ r .sin (φm ) from which χ r and φm can be given
by 6.6.2 and 6.2.3.

( 2
χ r = ( χ r .cos (φm ) ) + ( χ r .sin (φm ) ) )
2 2
(Eq. 6.2.2)

 χ r .sin (φm ) 
φm = tan −1   Eq. 6.2.3)
 χ r .cos (φm ) 

χr
Where pad pre-load = m =
Cr

The results of the bearing measurements for the offset half bearings are
presented in Table 21 for both the DE and NDE bearing.

Pre-Load Cant Angle

Diametral
Top Half Bottom Half
Bearing Ref Top Half Bottom Half Clearence
(degrees) (degrees)
(microns)
A 0.388 0.304 -37.4 -16.3 139
B 0.44 0.319 -32.1 2.1 152
Table 21 Fixed Geometry Offset Half Bearing Manufactured Measurements

A similar procedure was applied for the measurement of the tilting pad bearing
assembly. In this case, due to the variable geometry arrangement, the
bearings are not measured as an assembled unit as this cannot practically
and reliably be achieved. Instead each of the components is measured
individually. The main bearing housing was measured using a micro vertex
instrument which provided a typical dimensional measurement but to a high
degree of accuracy (0.001 mm). The reading was taken at several points
around the bore of the bearing to ensure both the size of the bore was
consistent and that the form (i.e. roundness) of the feature was acceptable.

181
Chapter 6 – Experimental Rig Commissioning and Test Procedure

The pads were measured in two ways. The first was to establish the pad
thickness which ultimately will determine the operational bearing clearance
(when combined with the housing bore measurement). Each pad was
measured using the micro vertex with a spherical ruby tip. The second
measurement was to establish the curvature of the bearing pad which would
control the level of pre-load present within the pad. This measurement was
achieved by using a CMM to acquire special co-ordinates across the pads
length. These co-ordinates were then curve fitted to extract the effective
radius of the pad by the same analytical procedure describe above. The final
assembled diametrial clearances for the variable geometry tilting pads were
111 and 128 microns.

Using the as measured bearing data, all bearings were recalculated for their
operational performance predictions and used in the rotordynamic evaluations
to obtain data for the as manufactured conditions.

As a final practical check, in the case of both the offset halve and tilting pad
bearings, shaft lift checks were also performed in the fully assembled test rig.
This process involves taking a reading of the shaft position with it sat in the
bottom of the bearings and then a second reading with the shaft raised into
the top halve of the bearings. The results of which were consistent with the as
measured clearances for the bearings.

6.2.5 Lubrication System

The lubrication system was flushed without any of the test rotor bearing or
pedestals fitted for a period of 12 hours at 50 degrees C. The test rotor
pedestals were then fitted and the process repeated. Finally a third flush was
performed with the entire rig assembled without the test rotor fitted. This
process was performed to ensure that all remaining contamination and debris
had been flushed through the filter before running the rotor in its bearings.
This provided an oil quality and cleanliness that would not have any significant
impact on the integrity of performance of the test apparatus.

A number of other checks were conducted relating to alignments, lubrications

system, balancing and transmission components etc, most of which were
conducted on the test rig assembly. Whilst these features are necessary for
satisfactory performance of the test arrangement they do not directly affect the

182
Chapter 6 – Experimental Rig Commissioning and Test Procedure

test rotor and are not of specific interest to the presented topic. As such they
are not covered in detail here.

6.3 Instrumentation Calibration and Validation

6.3.1 Proximity Probes

The sensors used to measure the shaft vibration and position were eddy
current type proximity probes. The specific sensors used were SKF 6mm tip 8
mV/µm in conjunction with SKF osolator demodulator (ODM) probe drivers
which were separately housed in an instrumentation cabinet to minimise
background interference.

Each of the proximity probe sensors used on the test equipment was first
calibrated as individual sensors to a master ODD driver to obtain the
displacement voltage characterise for that specific sensor. This was
performed to ensure that each sensor was within acceptable full scale errors
before building up into the system. Similar checks were performed for each of
the ODM by using a master proximity probe sensor. The method of calibration
was to use a special micrometer stand Figure 131 which is used to accurately
record the voltage reading for a known set of gaps. The probes were
calibrated across a range of 0 to 2280 microns. The power supply was
provided by means of a calibrated and condtioned DC power supply system at
24v DC.

With all the sensors and ODM driver checked, the instruments were built up
into the test and data acquisition system. With a complete system in place it
was now necessary to perform the final calibration of the probes. This was
carried out in the complete DAQ system to minimise the errors in data
readings which result from the cumulative effect of multiple sensors,
instruments and cables on the overall system calibration. The calibration
procedure was performed using the same equipment and procedure as
outlined above. The ODM drivers were supplied by means of a steady 24v DC
power supply provided through the data logging equipment.

183
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Figure 131 Shaft proximity probe calibration

The specific calibrations for each probe location were then set in the data
acquisition equipment for each respective channel.

When the proximity probes were installed into the test rig it was necessary to
set the gap between the probe and measured surface to a dimension within
the useable probe range. For the sensors used, the distance at which the best
linear response is achieved was 10 volts which equates to a gap of ~1250
microns. Each of the sensors used was fitted to this nominal gap setting.

6.3.2 Shaft Temperature Sensors

A detailed description of the shaft sensor design, construction and calibration

methodology was presented in chapter 5. The sensor calibrations performed
were as discrete sensors using a common bridge circuit. Applying the same
approach as that taken for the proximity probes, an additional calibration
verification was undertaken of the shaft thermistors as a complete acquisition
system.

The purpose of the verification was to gain a measurement of the thermisor

accuracy when coupled through the entire DAQ system and, more
significantly, the slip ring assembly. An accurate means of verifying the
measurement accuracy of the thermistors is difficult to achieve in the dynamic
system, principally due to the fact that a rotating system is required. More
significantly, a means of establishing a well distributed, stable and defined
temperature on the rotor is required.

184
Chapter 6 – Experimental Rig Commissioning and Test Procedure

The dynamic calibration accuracy measurement was performed with the rotor
fully assembled in the test stand and connected to the ANVIL DAQ via the slip
ring assembly. The method for introducing a controlled temperature was to
run the rotor at low speed (~60 rpm) with the oil being circulated with inlet
conditions presented at air ambient temperatures. The oil sump heater was
then switched on and the oil inlet temperature allowed to rise steadily over a
period of time. As the rotor rotation was so slow the temperature generated at
the oil film due to shear was considered to be negligible. The very low levels
of shaft vibration would mean any temperature perturbation in the oil film
would also be very small. With negligible heating from the oil film the dominate
temperature source to the shaft and bearing was by means of conduction from
the lubricating oil supply. The bulk temperature measurements of the shaft
were then compared to provide an in situ measurement of shaft temperature
sensor relative accuracy.

The time taken for the shaft to increase from ~25 oC to ~75 oC was nominally
12 hours. As with the individual sensor calibrations, the process was repeated
two further times. The results from these checks gave an absolute sensor to
sensor variation error for the thermistors of 0.06oC. Repeatability was shown
to be consistent with a variation of 0.06oC being the largest absolute error
within the 3 runs.

Slow Roll Thermistor 'in rig' Calibration Check

0.2 25
Ambient Temperature
0.175 Max Pk-Pk T Delta
Ambient Temperature (deg C)

20
Pk-Pk Rotor Temperature

0.15
Difference (deg C)

0.125
15

0.1

10
0.075

0.05
5
0.025

0 0
30 35 40 45 50 55 60 65 70 75

Mean Rotor Temperature (deg C)

Figure 132 Measured variation in temperature measurement from shaft

thermistors during slow roll gradual oil temperature increase

The results of the slow roll calibration verification gave results that were in line
with those seen for the individual sensor calibration accuracies when the
additional contribution of slip ring noise and DAQ thermal instrument wander

185
Chapter 6 – Experimental Rig Commissioning and Test Procedure

(note changes in ambient temperature across test) are considered. The

measured error level of 0.06oC is considered to be acceptable for the journal
shaft temperature measurement requirements of the experimental work.

6.3.3 General Sensor Calibration

All other sensors used were calibrated within normal operating accuracies by
means of standardised calibration procedures. The calibration accuracies for
each of the sensor types is in line with specifications as defined in chapter 5.

6.4 Running Commissioning

Before undertaking the full testing program, several running commissioning

tests were performed. This activity involved running the test rig with primarily
the rigid test rotor to establish the integrity and performance of the test
apparatus, instrumentation, bearings and drive system.

Several test runs were carried out where the rotor was run up in incremental
speed steps to a maximum speed of 14000 rpm. Further run up and run down
tests were performed to characterise the rotordynamic behaviour of the rotors.
During the final commissioning runs, the rotor was run up in speed steps of
1000 rpm and the speed maintained until thermal stability in the bearings and
shaft had been reached. This processes tended to take approximately 10
minutes between each speed increment, dependant on operating speed.
During the commissioning runs the rotors were operated both with and without
intentional mechanical unbalance weights introduced. The introduction of the
unbalance mass was applied to establish that safe operating conditions were
achieved when running in a forced excitation condition.

Throughout the test runs all of the temperature and shaft vibration parameters
were recorded and monitored to ensure that the readings were being received
correctly and to ensure that the bearing temperatures were satisfactory and
generally in line with expectations. The shaft vibration peak-peak levels were
monitored to ensure that the dynamic performance was stable and within
acceptable operating levels.

186
Chapter 6 – Experimental Rig Commissioning and Test Procedure

6.4.1 Bearings

6.4.1.1 Fixed Geometry Bearings

The first commissioning runs were carried out with the rigid rotor in fixed
geometry offset half bearings. The bearing temperatures at both the drive and
non drive end bearings were recorded and monitored across the running
range of the rig. The maximum temperatures were measured to be well within
acceptable operating limits across the entire speed range of the rotor. The
maximum temperature measured in the DE and NDE bearings was measured
to be 70 and 67 oC respectively which occurred when running at 12500 rpm.
Comparing the predicted and measured peak temperatures shows a good
correlation across the entire speed range, with temperatures differences of ~6
o
C when running at 12500 rpm with the measured temperatures being the
lower.

The lower temperatures measured on the actual bearing would be expected

due to the fact the temperature measurement is taken a some point below the
surface, and the magnitude of the difference is acceptable within the
boundaries of the analytical prediction techniques available.

6.4.1.2 Tilting Pad Bearings

In the latter part of the test programme tilting pad bearings were fitted onto the
main test rotor. The rotor used for operating in these bearings was limited to
the flexible rotor only. As for the fixed geometry bearing arrangement a
commissioning procedure was undertaken to validate the functionality of the
new features introduced to the test set up which was principally limited to the
bearings themselves and the RTD sensors fitted to the bearing. The bearing
temperatures were again monitored across the running range of the rotor up
to a speed of 12500 rpm (a lower maximum speed was limited during
commissioning due to precautionary measures due to previous damage to
shaft temperature instrumentation). Bearing pad temperatures of 58 oC at
10000 rpm were observed. This compares to a calculated peak pad
temperature of 69 oC resulting in a –11 oC difference. This level of
discrepancy is slightly higher than expected but could be as a result of the
level of oil flow presented to the bearing pads via the radial oil spray orifices.
Unfortunately the actual flow into the bearings could not be determined on the
rig due to a manufacturing defect resulting in some oil leakage within the

187
Chapter 6 – Experimental Rig Commissioning and Test Procedure

bearing pedestal preventing an accurate reading being obtained. The

temperature difference, however, between the leading and trailing edge of the
bearing is consistent with that analytically predicted.

6.4.2 Shaft Journal Temperature Sensors

Commissioning checks on the thermistor sensors in the shaft journal were

conducted prior to the running tests. As the temperature measurements are a
fundamental feature for the experimental work, significant efforts were
focused on the verification and validation of the shaft temperature
measurements.

As the sensor supply voltage from the bridge circuits were provided by means
of a brush type slip ring, continuity checks were performed to verify the
electrical resistance to earth and indications of any cross interference
between channels. Each slip ring way was checked for resistance to earth
with all channels demonstrating resistances greater than 16MΩ and no
voltage leaks identified across ways for any of the slip ring channels. The
maximum resistance measured across any of thermistor channels when short
circuited was 74mΩ compared to a minimum thermistor resistance within the
expected operating range of 7.697kΩ at 90 oC. This equates to a nominal
error of +0.0003 oC at 90 oC .

Some problems were experienced with earth leakage of thermistor signals

within the slip ring later in the testing. This was found to be due to build up of
carbon deposits creating a short circuit path. The issue was resolved by
stripping and cleaning of the slip ring. For all future tests it was routine to
clean the slip ring inner housing following each test run.

Once the continuity checks and measurements had been made the thermistor
signals were recorded during the commissioning running tests and the signals
assessed for general quality and stability. The rotor was run up in speed
increments of 1000 rpm and held at a constant speed for a period of time.
During each dwell period the shaft temperatures were allowed to stabilise
before accelerating up to the next speed increment, to a maximum speed of
12500 rpm. The absolute temperature levels showed reasonable stability
within a short duration (<10s), with absolute temperature measurements
fluctuating by no more than 0.05oC pk-pk at 12500 rpm constant running
conditions. This suggests some level of electrical and slip ring noise but at a

188
Chapter 6 – Experimental Rig Commissioning and Test Procedure

manageable level. Typical durations to reach stable shaft temperatures were

between 5 and 15 minutes dependant on the running speed, with the
response of the sensors showing high sensitivity to change in operating
conditions with little sign of hysteresis with relation to shaft speed increases
and decreases.

To provide increased accuracy of the shaft temperature readings, the input

voltage to the bridge circuit from the DC power supply was also logged with
the DAQ system. The DC supply system was set to a nominal supply voltage
of 10 VDC but, due to instrument stability and ambient temperature variation,
the supply voltage could vary by ~80 mV. By recording the output from the
power supply this could then be used in the bridge resistance calculations at
each discreet data point and any associated error resulting from deviation
from the 10 VDC negated.

Repeatability checks were performed and the results showed that the stability
and consistency of results was maintained for each test run. A specific
measure cannot be meaningfully applied between each test run due to the
sensitivity of the shaft temperatures to supply lubricant temperature and
specific operating speeds.

6.4.3 Rotor Vibration Response

The rotordynamic response was evaluated by means of a steady run up test

and free rundown. The results were verified with the nominal predicted natural
frequency and response data generated at the test rotor design stage. For
both the rigid and flexible rotor, acceptable vibration responses were obtained
across the entire running range of 0 – 14000 rpm with pk-pk vibration
maximum of 5 microns and 8 microns for the rigid and flexible rotors
respectively. No imbalance loads were introduced onto the shafts for this test
verification.

To measure the level of ‘in rig’ electrical mechanical runout ‘glitch’ at each of
the observed surfaces, a slow roll check was performed at ~80 rpm (so as
limit any effects of dynamic movements within the bearing) during which the
‘phantom’ vibration signature was recorded. Figure 133 shows an example of
the shaft ‘glitch’ reading for the rigid rotor NDE bearing observed band. This
data was used during the vibration analysis by subtracting it from the
operational vibration data. The method used to establish and remove the shaft
glitch from the vibration signature is described in more detail in chapter 7.

189
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Figure 133 Slow roll glitch for rigid rotor NDE

Figure 134 shows the measured response data for the NDE bearing of the
flexible rotor in fixed geometry bearings with no unbalance introduced to the
shaft.

20 180
Y Vibration
18
135
Pk-Pk Shaft Vibration Filtered 1X (microns)

Phase
16
90
1X Vibration Phase (degrees)

14
45
12

10 0

8
-45
6
-90
4
-135
2

0 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

Figure 134 Synchronous (1X) vibration amplitude and phase response

(Bode Plot) of rigid rotor in fixed geometry bearings

190
Chapter 6 – Experimental Rig Commissioning and Test Procedure

6.5 Rotor Commissioning Summary

The commissioning activities outlined above provided satisfactory and safe

operating conditions across the intended operating range. Some changes in
the instrumentation configurations and equipment was required to obtain
reliable and acceptable data but the final result of the systems was to give
data that satisfied the design specification brief.

Measured performance conditions such as thermal and vibration

measurements were in line with those expected from theoretical predictions.
The largest discrepancy was observed in the tilting pad bearing temperatures,
but a significant element of this could be attributed to the offset introduced by
the location of the RTD’s in the pads being several mm below the babbit
surface and in an area of greater cooling.

In conclusion, the rig was deemed suitable to undertake the experimental

programme.

6.6 Test Procedure

A specific series of test programmes was compiled to cover the experimental

testing. The test programmes were added to and adapted as the work
progressed and as results became available. However the basic programmes
remained largely unchanged across the testing duration. The original
programme of work was focused around four testing work packages, DR01,
DR02, DR03 and DR04. These test programmes were focused around
commissioning and testing both the rigid and flexible rotors in fixed geometry
offset half bearings. A second batch of work was defined in principle to look at
a similar test programme for the rotors running in tilting pad bearings. It was
not originally envisaged that the tilting pad bearing work would make up part
of this research, but it was decided to include the work to provide a wider
range of data within the test programme. The additional test programme was
introduced for the running of the flexible rotor in tilting pad bearings and is
identified as DR06 work package. As such the number of tests and testing
time was increased significantly. Table 22 summarises each work package
conducted within the testing programme.

The following tables provide details of the test programmes for DR03, DR04
and DR06.

191
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Testing streams
DR01 Rigid Rotor Commissioning with Fixed Geometry Bearings
DR02 Flexible Rotor Commissioning with Fixed Geometry Bearings
DR03 Rigid Rotor with Fixed Geometry Bearings Test Programme
DR04 Flexible Rotor with Fixed Geometry Bearings Test Programme
DR05 Flexible Rotor Commissioning with Tilting Pad Bearings
DR06 Flexible Rotor with Tilting Pad Bearings Test Programme
Table 22 Specific test activity streams

For each testing programme, a number of specific tests and test conditions
were defined. For each test the basic test parameters and speed ranges were
specified.

Two separate speed ranges were defined for the test programme as high
speed and low speed tests. This was carried out due to the problems
experienced with the shaft temperature measurement instrumentation when
running at high speed. To reduce the risk of damaging the instrumentation at
high speed and prevent significant delay to the testing, the low speed tests
were generally performed at the begining of each specific phase, and the high
speed tests were performed at the end. A number of tests were added into the
programme and several conditions repeated a number of times to validate
repeatability of readings.

Table 23 to Table 25 provide a detailed breakdown of the tests conducted

under each test phase.

Added Unbalance
Inlet Speed
Test No. Test Description Unbalance Angular
Temp Range
Amount (g) Location
T3.1 No unbal 40 deg C Inlet (0-8000) 40 Low 0 -
T3.2 No unbal 30 deg C Inlet (0-8000) 30 Low 0 -
T3.3 -0.1g unbal 40 deg C Inlet (0-8000) 40 Low -0.1 180
T3.6 -1.00g unbal 40 deg C Inlet (0-8000) 40 Low -1 180
T3.7 -2.50g unbal 40 deg C Inlet (0-8000) 40 Low -2.5 180
T3.8 -5.00g unbal 40 deg C Inlet (0-8000) 40 Low -5 180
T3.11 No unbal 40 Deg C Inlet (8000-12500) 40 High 0 -
T3.21 -7.5g unbal 40 deg C Inlet (0-8000) 40 Low -7.5 180
T3.22 25.5g unbal (no bolt fitted) 40 deg C Inlet (0-4500) 40 Low -25.5 180
T3.24 -7.5 unbal 40 deg C Inlet (0-8000) 40 Low -7.5 0

Table 23 Rigid rotor test programme DR03 – fixed geometry bearings

192
Chapter 6 – Experimental Rig Commissioning and Test Procedure

Added Unbalance
Inlet Speed Dummy
Test No. Test Description Amount (g)
Temp Range Weight Fitted
(Unbalance angle)
T4.1 No unbal 40 deg C Inlet (0-10000) 40 Medium No 0
T4.6 -1.00g unbal 40 deg C Inlet (0-11000) 40 Medium No -1 (180)
T4.7 -2.50g unbal 40 deg C Inlet (0-10000) 40 Medium No -2.5 (180)
T4.27 No unbal 40 deg C Inlet Dummy Weight (0-10000) 40 Medium Yes 0
T4.28 -1.025g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-10000) 40 Medium Yes -1.025 (180)
T4.29 -2.021g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-10000) 40 Medium Yes -2.021 (180)
T4.51 -1.1g ubal @ ctr 0g unbal @ flg 50 deg C Inlet (0-10000) 40 Medium No -1.1 (C0)
T4.53 -2.5g ubal @ ctr 0g unbal @ flg 50 deg C Inlet (0-10000) 40 Medium No -2.5 (C0)

Table 24 Flexible rotor test programme DR04 – fixed geometry bearings

Added Unbalance
Inlet Speed Dummy
Test No. Test Description Amount (g)
Temp Range Weight Fitted
(Unbalance angle)
T6.1 No unbal 40 deg C Inlet (0-10000) 40 Medium No 0
T6.6 -1.00g unbal 40 deg C Inlet (0-11000) 40 Medium No -1.0 (180)
T6.7 -2.50g unbal 40 deg C Inlet (0-11000) 40 Medium No -2.5 (180)
T6.51 No unbal 40 deg C Inlet Dummy Weight (0-12500) 40 High Yes 0
T6.66 -1.1g ubal @ ctr 0g unbal @ flg 50 deg C Inlet (0-11000) 40 Medium No -1.1 (C0)
T6.67 -2.5g ubal @ ctr 0g unbal @ flg 50 deg C Inlet (0-11000) 40 Medium No -2.5 (C0)
T6.73 -1.025g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-11000) 40 Medium Yes -1.025 (180)
T6.74 -1.025g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-11000) 40 Medium Yes -1.01 (270)
T6.75 -1.025g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-11000) 40 Medium Yes -1.025 (0)
T6.76 -1.025g unbal @ flg 0g @ ctr 40 deg C Inlet Dummy Weight (0-11000) 40 Medium Yes -1.01 (90)

Table 25 Flexible rotor test programme DR06 – variable geometry bearings

193
Chapter 7 – Experimental Results

Chapter 7

Experimental Results

7. Experimental Results

7.1 Introduction

This chapter presents the experimental results obtained and the methods
used to extract and post processes the measurements and data captured
throughout the experimental testing of the rotor bearing rig. The results data
captured can be divided into three principle categories, these being mainly
associated to the various types of measurements taken during the testing as
outlined in the previous chapters. The test data result groups are (1) shaft
vibration and position, (2) journal bearing temperatures and (3) shaft journal
temperatures. A more detailed breakdown of the test data for each group is
provided in the following sections.

To enable meaningful interpretation of the measured data, an extensive

amount of post processing was required for the various data measurements
and a detailed overview of the procedures developed and applied is presented
within this chapter.

7.2 Data Conditioning and Post Processing

This subsection provides details of the techniques and procedures used to

extract results data from the raw data captured during the practical
experimental testing. Much of the data captured is in a raw format and it is
important that appropriate methods are used to condition the signals into
meaningful quantities without introducing significant errors and unwanted
numerical filtering.

7.2.1 Shaft Vibration and Position

The data for both the shaft vibration and position was captured by means of
the non contacting proximity probes as described in chapters 5 and 6. The
sensors were placed at various locations along the shaft length enabling the

194
Chapter 7 – Experimental Results

dynamic response of the rotor to be captured across the range of operating

conditions.
The data from the proximity probes was captured in the three ways using the
three different acquisition instruments. The DAQ instruments used captured
both low frequency sampling and high frequency acquisition data. Three of the
data capture instruments were used to obtain shaft vibration related data and
for convenience will be referred to as following for the remainder of this
chapter.

System I = Low frequency data logger (Newlyme - NY)

System II = High frequency digital data recorder (Sony - SO)
System III = High frequency signal analyzer (Oros - OROS)

A detailed description of the equipment used and the respective

configurations is provided in the preceding chapter 6.

Data captured by system I for the proximity probes was used to provide a
peak to peak total vibration level for each sensor position. The data was used
as a real time monitor throughout the testing with a data log reading being
captured once every 60 seconds. (Real time updates to the digital display
were once every second). The data captured with this system was
predominantly used for control and monitoring during the test operation. As
such the data is not processed further or presented here as the dynamic
recorded data is captured by system I and II.

System II was used to capture the raw data from each of the proximity probes
at various points throughout the testing sequence. The data captured by this
method is for the purpose of post processing following the test itself and has
been used to determine high resolution shaft vibration trends, direct
measurement of unfiltered shaft orbits and direct measurement of shaft
position. The data was captured at ‘high’ frequency to enable a suitable
frequency range to be evaluated and sufficient resolution of shaft orbits and
vibrations to be achieved. The sampling rates of the channels ranged from
2048 samples/s to 16384 samples/s. Due to the resulting size of the digital
files, data was captured for short bursts throughout the testing of the rotor.
Approximately ten second bursts were acquired at each running speed
condition. Where run-ups/run downs or conditions of transient activity were
being assessed then data was captured for the entire duration of the event.

195
Chapter 7 – Experimental Results

System III was used specifically to captured the filtered once per revolution
(1X) shaft vibration amplitude and associated phase. The rate at which the
data was sampled was again considered high at 4092 samples/s (19.6
s/rotation a 12500 rpm) to ensure good quality data capture. However the
output data was recorded as a simple synchronous filtered total vibration
component with the associated phase angle and was logged at 1 sample per
second. This information was then later post processed and presented to give
the rotor unbalance response characteristics and synthesized once per rev
filtered shaft orbits.

7.2.1.1 Shaft Glitch

In rotating machinery diagnostics and measurement, the term ‘shaft Glitch’ or

combined electrical mechanical run out is often used to describe the effects of
inherent residual errors presented in measurement of shaft vibration by
means of non contacting probes and in particular those of the eddy current
operating principle (as presented in chapter 6). The term shaft Glitch was
widely introduced by proximity probe manufacturer Bentley Nevada and has
been given good account in a number of publications, one of which was
presented by Biggs [58]. The feature gives specific account for the resulting
measurement error introduced on a rotating shaft due to the mechanical effect
of relative run out and inconsistent magnetic response around the proximity
probe observed surface.

When producing a shaft with eddy current observed surfaces special

precautions are made in the preparation of the monitored shaft diameters as
outlined in section 5.5. Careful manufacture of the shaft journals and observed
diameters is required in combination with a low level demagnetization and
burnishing of the on the observed diameter surface. This typically enables
combined residual mechanical-electrical run out errors to be limited to ~8
microns. However where accurate measurements are required, such as those
within this research, further account has to be taken of these residual errors.

To provide account for the residual error present within the shaft proximity
probe measurements a ‘slow roll’ compensation technique was applied. This
consisted of running the shaft in its bearings at as slower speed (~30 rpm) as
practical and recording the resulting proximity probe data for every location.
This data was captured simultaneously with a shaft once per rev key phasor
to provide visibility of the relative phase of the captured data to a point on the

196
 Chapter 7 – Experimental Results

test rotor. Once captured and processed this data can be used to subtract
from the running vibration data to ‘filter’ out the unwanted phantom vibration
components. Figure 135 gives a simple example of how the principle of the
slow roll glitch extraction technique was implemented and later applied.

1 Shaft Rotation
25
25
20 20
15 15
10 10
5 5
0 0
-5 -5
-10 -10
Datum point on shaft
-15 -15
-20 -20
-25 -25
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360

Measured vibration signal of running shaft Measured slow roll signal from
shaft (Glitch)

25
20
15
10
5
0
-5
-10
-15
-20
-25
0 30 60 90 120 150 180 210 240 270 300 330 360

Measured signal subtracting shaft glitch

Figure 135 Shaft ‘glitch’ extraction principle

using slow roll measurement

The slow roll data captured was taken at the beginning and end of every test.
This was carried out because, whilst the mechanical component of the shaft
glitch remains, constant the magnetic response component (due to the eddy
current field interacting with the observed surface microstructure etc.) can
vary as a result of proximity probe gap or precise axial position. As most of the
proximity probes had to be reset following each test due to disruption to the
bearings or unbalance weights, account for this potential source of
variation/error had to be accommodated.

Slow roll data was captured and compensated for in two ways. The first
method was based around extracting the first order components from the
captured data and applying this to the running data in the frequency domain.
The second more involved procedure was based around using the captured
raw signal data and conditioning this with the running data in the time domain.

197
Chapter 7 – Experimental Results

The two separate methods were applied to the data captured by System III
and System II respectively.

Glitch Data Extraction for system III

The vibration data is processed by the Oros software to extract the 1st order
synchronous vibration amplitude and phase lag angle. These 1st order
components are obtained for the slow roll ‘glitch’ ( A ,φ )
g g and measured

vibration ( Am , φm ) . Where, using these components, the time domain vibration

response can be described by :

xm = Am .cos (ω.t + φm )
Eq 7.2.1
xg = Ag .cos (ω.t + φ g )

Considering xm and xg in complex form :

xm = am .cos (ω.t ) + bm .sin (ω.t )

Eq 7.2.2
xg = ag .cos (ω.t ) + bg .sin (ω.t )

Following some work it can be shown that :

am = Am .cos (φm ) , ag = Ag .cos (φg )

Eq 7.2.3
bm = Am .sin (φm ) , bg = Ag .sin (φg )

Having obtained the complex factors for the measured and glitch data the
necessary relationships to extract the actual vibration are developed. Again,
following some work, the amplitude and phase lag angle of the ‘true’ or glitch
filtered measurement of the shaft vibration can be shown to be given by:

2 2
Ao = (a m − ag ) + ( bm − bg ) Eq. 7.2.4

 bm − bg 
φo = tan −1   Eq. 7.2.5
 am − a g
 

198
Chapter 7 – Experimental Results

Where Ao and φo are the amplitude and phase lag angle of the conditioned
shaft vibration. From these constants the synchronous response data can be
determined for the rotor with the ‘glitch’ effects removed.

Glitch Data Extraction for system II

• Import raw signal data into Matlab as a matrix

• Attribute associated variables
• Identify trigger point (0 deg phase location)
• Write data point increments into trigger matrix
• Modify number of points in real-time to match sample rate of trigger
blocks
• Extract blocks for each trigger increment
• Modify number of points in block to a standard number of samples(i.e.
1024)
• Write vector to vibration matrix
• Average vibration matrix over sample length to give single glitch vector

It should be noted that of the two methods described above, the latter is the
more informative as this gives a direct representation of the physical
measurements around the shaft observed surface and can be used as a true
time domain measurement.

The first method provides a quick and direct method extracting 1X shaft
vibration and orbit shape around which the current thermal instability theory is
principally developed, whereas the second procedure allows for more detailed
evaluation of actual shaft orbits to be evaluated and its effects on effective film
thickness etc around the bearing journal observed. The first method was
applied more generally as this can be more readily applied to the results data
in an effective way without requiring a significant amount of numerical
processing, whilst still retaining the key dynamic information of the shaft
synchronous vibration. Also, due to the analytical evaluation being conducted
before the data is logged to disk, it means manageable data files can be
obtained. This allows the records to be taken for the entire duration of the test
as opposed to the high frequency data captured as in the second procedure
which applies only of relatively short bursts as key points of the running. The
time domain glitch extraction was only applied where the high frequency raw
data was used for further analysis or investigation of the shaft orbits or
vibration at specific test conditions.

199
Chapter 7 – Experimental Results

To perform the numerical processing with the first method, the procedure was
applied to the data by means of a spreadsheet type environment due to the
relatively simple computational effort. The second time domain procedure was
performed by application of a specific code developed in Matlab [49] by the
author.

7.2.1.2 Shaft Vibration and Orbits

Shaft vibration data was captured through applying the same principles as
outlined in section 7.2.1.1 for the slow roll ‘glitch’ extraction. Synchronous
filtered shaft vibrations were captured using the system II acquisition system
and raw time domain data was captured by system III.

The method for data processing was similar to that outlined in section 7.2.1.1
in that the captured data was structured in synchronous blocks which were
then combined into a single matrix for each section of data. From each of
these blocks the electrical mechanical residual run out error was then
subtracted to leave only the contribution of the dynamic vibration to the as
measured data. This data was then used to establish the following key
parameters:

• Total Level Peak to Peak Shaft Vibration

• Time averaged synchronous vibration sample
• Harmonic content
• Direct unfiltered shaft orbits
• Shaft Running Position
• Filter 1X peak to peak shaft vibration and phase (Bode plots)
• Filtered 1X shaft orbits

Total Level Peak-Peak Vibration

Total level peak to peak vibration was extracted from the captured data by
means of a short MatLab routine to interrogate the data and identify the peak-
peak level identified form data block sizes of 0.1X filtered. The peak-peak
result for each block was averaged on a rolling consecutive 8 block basis to
give a stable measurement, removing any random elements in the acquired
signal.

200
Chapter 7 – Experimental Results

Time Averaging

Averaging is often used in the measurement and post processing of vibration

data. The use of averaging produces statistically more reliable and repeatable
data where significant variation from one data sample and the next is often
observed. The most popular method for averaging is carried out in the
frequency domain as is commonly applied in dynamic analysers where the
Fast Fourier Transform (FFT) is used to extract the harmonic distribution of a
given data sequence. The FFT for each data bock is then averaged in the
frequency domain to provided more statistically reliable results.

Time averaging was used when running at a constant speed to provide a

means of harmonic filtering to the data samples whilst retaining the actual
time domain signal for further analysis and post processing. This method of
averaging is useful as it retains all amplitude and phase data relating the
synchronous only vibration components of the rotor. This avoids converting to
the frequency domain which can often result is certain components of the data
being truncated or lost such as phase.

The time averaging process was conducted within the MatLab [49] routines
used to process the data samples.

It should be noted that handling data in the time domain can be

computationally demanding and result in significant processing time. Careful
thought is need as how to manipulate the data to make the process practical
and efficient. It is especially a problem when using high sampling rates and
running at relatively low speeds where synchronous block sizes can become
very large resulting in unwieldy vectors and matrix sizes that require numerical
truncation for further analysis. For this reason the time domain procedures
have not been routinely used on all test result sets and are only applied where
specific areas of interest are identified.

Harmonic Analysis

To provide a simple breakdown of the frequency components contained in the

data samples, a harmonic analysis was performed on the conditioned blocks
of synchronous data. The harmonic analysis was applied to the data blocks
following the subtraction of the residual mechanical-electrical run out.
Harmonic extraction was applied to 0.1X filtered data blocks. This was

201
Chapter 7 – Experimental Results

conducted so that both sub-synchronous and super synchronous components

were identified.

The harmonic components were extracted by application of a discrete Fourier

analysis to the data block samples. Averaging was performed in the frequency
domain to the block wise harmonic signatures. To show the level of variation
seen across a given data set the amplitude of preceding harmonic
components was tracked for the entire data sample. Suitable processing was
used to give block sizes on 2n that enable the Fast Fourier Transform (FFT)
algorithm to be applied for ease of post processing on some data sets.

The aforementioned MatLab routine was developed to perform the

computational procedure to apply the discrete Fourier transform to the data
samples.

Unfiltered Shaft Orbits

The unfiltered shaft obits were extracted by taking 1X post processed data
blocks of 1024 data points for a given test condition for a specific probe. 16
consecutive blocks were averaged to give a more statistically reliable result
(effectively a short duration time average). The resulting single (time
averaged) block for each probe the X and Y probe for a given location can
then be plotted against each other in a Cartesian coordinate system to give
the shaft orbit. As at each location the orientation of the probes was not the
same, the coordinates then had to be rotated by the appropriate clocking
angle to present them relative to a common datum. The coordinate
translations were performed by application of a rotation matrix of the form :

cos (θ p ) − sin (θ p )   X   X p 
 .  =  
 sin (θ p ) cos (θ p )  Y  Yp 
 

 X p   x1 x2 .... xn −1 xn 
 =
Yp   y1 y2 .... yn−1 yn 

X 
Where   is the cartesian coordinate orbit matrix were xi and yi are the
Y 
specific coordinates of a given data point i for the X and Y probe respectively
and θ p is the clocking angle of rotation required for the specific pair of probes
being considered.

202
Chapter 7 – Experimental Results

Figure 136 Coordinate system for clocking angle of radial proximity probes

Running Shaft Position

The process used to determine the shaft running position was to take the DC
component of the proximity probe signals. The DC component is the mean
value taken over a specified period of time (the length of time would be related
to the shaft speed to ensure integrity of the reading is not removed by over or
under averaging) of the vibrating signature from the proximity probe when the
effect of the steady state gap is not removed prior to processing.

The acquired DC proximity probe data was block-wise averaged in the time
domain in a similar manner to that carried out for the time averaged vibration
data. The specific relationship to the shaft phase trigger is less critical here
though as the dynamic vibration element of the signal is not of interest. For
each mean block-wise value this was written to a new vector which can then
be used as record of the steady state probe gap reading for a given operating
point or condition.

To obtain the gap, the understanding of some of the physical bearing

dimensions and the static starting position must be known. The bearing
dimensional data was obtained earlier by means of component measurements
and then also by a direct lift check measurement in the test rig. The static
shaft position measurement is needed to know what the static DC gap
readings are from the proximity probes observing the rotor. This data was
obtained by means of taking a static shaft position reading at the start of every

203
Chapter 7 – Experimental Results

test at ambient conditions prior to any oil being circulated around the rotor or
bearings (i.e. rotor temperature would nominally be at ambient conditions).
This static measurement can then be subtracted from the dynamic gap
measurement to give the change in gap and hence shaft positional change
relative to the bearing pedestal.

Another feature, which must also be considered as it can result in errors in

measurement if overlooked, is the effect of relative thermal expansion
between the observed surface and point of measuring instrument. In this case
the rotor journal thermal growth relative to the thermal growth of the probe
mounting point. This was addressed by means of the ‘push-pull’ probe
arrangement as shown in figure 103 chapter 5. This arrangement allows for
compensation to be taken for the reduction or increases of effective internal
bearing clearance as the simple addition in readings of diametrically opposed
probes. The residual of this summation should be 0 for no relative change due
to thermal growth and any positive value corresponding to an effective
reduction in bearing internal clearance and vice versa for a negative residual.
The shaft running position relative to the bearing centre point, datum O, for a
specific probe plane can then be obtained by :

−Cd . ( ∆X Gap + ∆YGap )

X Pos , YPos = + X T , YT (Eq. 7.2.6)
4

Where X T = X Dyn − X Static , YT = YDyn − YStatic and ∆X Gap , ∆YGap are the residual of
the push-pull probe measurements.

Once the X Pos , YPos vectors have been extracted for the required speed
conditions then, as with the orbit special vectors, the data points were
translated into the correct orientation by means of the rotation matrix.

Some care needs to be taken when no circular geometry profiles or variable

geometry profiles are being used in the bearings. In the case of the fixed
geometry bearings used here, no special consideration was required as,
although modified geometry is present, the bearing is split horizontally in the
pedestal. This means the diametrical clearance Cd used in equation 7.2.6 is
the same as the measured crown clearance for the bearing. In the case of
variable geometry bearings then a factor is applied which accounts for the
effect of the increased drop of the shaft in the bearings due to the load
between pad condition. This relationship for a five pad bearing is the total lift

204
Chapter 7 – Experimental Results

value multiplied by f lift = 0.894 . Where when considered for the running
position measurements the expected radial lift to the bearing centre would
1
become frlift = .
2. f lift − 1

Filtered Shaft Vibration

Filtered shaft vibration was obtained using the analysis routines provided in
the OROS frequency analyzer software. The procedure takes the high
frequency data from the proximity probes and performs a FFT on this signal.
Using a low number of frequency domain averages it extracts the amplitude
and phase for the frequency that corresponds to the shaft order of interest.
The shaft specific order is limited by the sampling frequency, number of
spectral lines and shaft running speed range. For the analysis performed here
the fundamental shaft order 1 was specified for extraction. The amplitude and
phase values are then logged at a much slower sampling rate to give a typical
Bode plot of peak vibration amplitude and phase with reference to either time
or shaft speed.

Filtered Shaft Orbits

The filtered shaft orbits were produced by using the data obtained from for the
filtered shaft vibration described above. From the amplitude and phase data
for a pair of corresponding probes, it is possible to synthesize the orbit in
Cartesian coordinates in the same manner as done for the unfiltered orbits,
but in this case only the 1X orbit component is retained. The resulting orbit
ellipse is rotated by the appropriate clocking angle by application of the
following trigonomic relationship:

x p (ω.t ) = x (ω.t ) .cos (θ p ) − y (ω.t ) .sin (θ p )

(Eq. 7.2.7)
y p (ω.t ) = x (ω.t ) .sin (θ p ) + y (ω.t ) .cos (θ p )

Where x (ω.t ) = Ax .cos (ω.t + φx ) and y (ω.t ) = Ay .cos (ω.t + φ y )

205
Chapter 7 – Experimental Results

7.2.2 Bearing Temperatures and Lube Oil/Air Temperatures

The bearing temperatures and lube oil/air temperatures were measured using
PT100 RTD’s and K type thermocouples respectively. In the case of both
types of sensor the signal conditioning and calibration aspects were managed
within the DAQ systems. The only requirement was to determine each specific
sensor sensitively (established as outline in chapter 6) and provide this to the
appropriate DAQ post processing software. The measured temperature is
then provided directly in the logged data files from the respective instrument.

7.2.3 Shaft Temperatures

The shaft journal temperature measurements were acquired as voltages by

the Anvil DAQ system and logged at a sample rate of 1 samples/s. These
voltages were then translated into temperature measurements by application
the Steinhart-Hart constants applying the methods presented in chapter 6
which are not repeated here. The non linear voltage responses acquired from
the thermistors bridge circuits were processing into actual temperature values
for further processing and diagnosis by means of spreadsheet program
created in Excel. This program used the individual constants, calibration
settings and associated data to determine the temperature readings with
reference to speed and/or time.

Having now obtained temperature data for the journal for all four shaft
sensors, it is necessary to further process these measurements to give a
more meaningful account of the temperature conditions present within and
around the rotor journal. The temperature measurements for each of these
sensors is ordered to reflect the sequence as they appear around the rotor
journal with respect to the shaft key phasor in the direction of rotation of the
test. Further processing and interrogation is then performed to extract the
temperature differential data relevant to the thermal bowing conditions and
how they associate/correlate to the vibration data.

Whilst the peak-peak temperature differential is of interest, the principle

features of interest for the journal temperatures are the form of the
temperature distribution and more so the magnitude and location of the
fundamental harmonic temperature distribution i.e. amplitude and phase of the
1st harmonic temperature differential. Of additional interest is the bulk

206
Chapter 7 – Experimental Results

operating temperature of the journal and also the level of non fundamental
harmonic components in the thermal distribution.

To extract the fundamental harmonic component from the measured data

points, a Discrete Fourier Transform (DFT) analysis was applied each data
record. The temperature distribution can be expressed by the following
Fourier description for a discreet set of data consisting of N data points. The
greater the number of data points the higher the number of frequency orders
that are incorporated. This gives a result which better matches the true result
for the complete periodic data set. Equation 7.2.8 gives the temperature
distribution T (θ ) in terms of the Fourier coefficients and angle theta around
the shaft journal.

to N /2
+ ∑ ak .cos ( k .θ ) + bk .sin ( k .θ ) (Eq. 7.2.8)
2.π
T (θ )0 =
2 k =1

The Fourier coefficients for the order of interest can be determined by:

∑t
n =1
n
to = (Eq. 7.2.9)
N

2 N  k .2.π . ( n − 1)  2 N  k .2.π . ( n − 1) 
ak = .∑ tn .cos   bk = .∑ tn .sin  
N n =1  N  N n=1  N 

The amplitude and phase data of the relevant order can now by determined
as follows:
 bk 
Tk = ak2 + bk2 φT = tan −1   (Eq. 7.2.10)
k
a
 k

For the measured temperatures of the shaft journal, 4 sensors were used
making N=4. The temperature distribution that will result in a thermal bend it is
the first order content that is of interest where k = 1 . This process allows for
the peak to peak amplitude ( 2.T1 ) and phase angle ( φT 1 ) to be established
from the measured temperatures. It should be noted that, whilst it is
numerically valid to be able to extract the second order harmonic content of
the temperature distribution (i.e. the limit N/2, where N=4), there are only 2
points to describe a single cycle of the second order content significant
amplitude and phase errors. For orders higher than the fundamental, a

207
Chapter 7 – Experimental Results

number of data points must equate to ≥ 2.56 per cycle to be statistically valid
for amplitude data (as is stated by Shannon and Nyquist [59]). In the case of
the data obtained here the signature content that is non harmonic can be
presented by taking the 1st harmonic component for a given angular location
from the measured value at this point. The residual of this measurement
t K can be considered as the non fundamental harmonic content and is
presented in the results data.

t K ( n ) (θ ) = tn − T1.cos (θ + φT 1 ) (Eq. 7.2.11)

The processing of the data in accordance with the method outlined above was
conducted within the spreadsheet program mentioned earlier. The above
procedure can be applied to the entire data captured during the testing
providing a thermal Bode plot with reference to time and/or shaft speed, which
also includes some information of the non harmonic content.

7.3 Test Results

This section presents the results data acquired from the experimental testing
of the three test programmes undertaken. Not all results are presented here
due to the large volume of information, but a number of significant test
conditions show the key aspects observed from the testing. This allows
qualitative and quantitative analysis to be made from the conditions observed
during running.

The results section is broken down into two principle sections of vibration
related data and thermal data. The vibration data looks at the measured shaft
dynamic and spatial features, whereas the thermal data is based around the
temperature conditions in the bearings - more specifically the shaft journal.

General data logs were taken during every test which captures all conditions
such as oil drain temperatures, oil supply temperatures, oil supply pressures,
motor power and ambient temperature etc. These are not presented in detail
here due to the repetitive nature and the data being of passive interest only.
Some aspects of the general test data are used in the following chapter when
analysing the results of interest with respect to the operating conditions.

208
Chapter 7 – Experimental Results

7.3.1 Shaft Vibrations

This section presents the rotor shaft vibration and spatial test results for the
test programmes DR03, DR04 and DR06.

A large amount of data was gather and processed for the shaft rotordynamics
during the testing. Only a summary of the key features is presented here to
enable the principle performance features to be understood. Results data for
filtered and unfiltered shaft vibration, whirl orbits and running shaft position
was evaluated for each of test conditions.

7.3.1.1 Rigid Rotor Test Programme DR03

This sub section covers some of the vibration results for the rigid rotor test
DR03 programme running fixed geometry bearings. The two test conditions
for which some vibration results are presented here are:

DR0301 No unbalance, oil inlet temp 40oC

DR0321 -7.5g (-300 g.mm) unbalance position 1, oil inlet temp 40oC

The filtered 1X vibration for the base line condition of test DR0301 for no
unbalance at the NDE bearing is given in Figure 137.

7 0
X Probe X Probe
Filtered 1X Vibration Phase (degrees)
Filtered 1X Vibration pk-pk (microns)

6 Y Probe -20 Y Probe

5 -40

4 -60

3 -80

2 -100

1 -120

0 -140
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Shaft Speed (rpm) Shaft Speed (rpm)

Figure 137 NDE Bearing filtered shaft vibration DR0301

The total level pk-pk vibration (glitch removed) and filtered 1X response
results the NDE bearing for test DR0321 is presented in Figure 138.

209
Chapter 7 – Experimental Results

35 0
X Probe 1X X Probe

Filtered 1X Vibration Phase (degrees)

30 Y Probe 1X -20 Y Probe
X Probe unfiltered
Vibration pk-pk (microns)

25 Y Probe unfiltered -40

20 -60

15 -80

10 -100

5 -120

0 -140
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Shaft Speed (rpm) Shaft Speed (rpm)

Figure 138 NDE Bearing filtered and unfiltered shaft vibration DR0321

The 1X filtered orbit at the NDE bearing is given in Figure 139 for the DR0321
test condition running at 8000 rpm.

40

30
Vertical Displacement (microns)

20

10

-10

-20

-30

-40
-40 -30 -20 -10 0 10 20 30 40
Horizontal Displacem ent (m icrons)

Figure 139 NDE 1X filtered orbit 8000 rpm DR0321

The shaft running positions were seen not to varying significantly between
different test conditions performed within the conditions (principally the
introduction of unbalance) covered within this work. It is noted that shaft
position is clearly influenced by running speed, as would be expected.

A harmonic frequency analysis of the NDE bearing time domain signal is

shown in Figure 140 for DR0301 and DR0321 respectively.

210
Chapter 7 – Experimental Results

4.5 180 25 180

Pk-Pk Displacement Pk-Pk Displacement
4 135 135
Displacement Phase Displacement Phase
Shaft Dynamic Displacement

Shaft Dynamic Displacement

Shaft Dynamic Displacement

Shaft Dynamic Displacement

Angle 20 Angle
3.5
90 90

Phase Angle (deg)

Phase Angle (deg)

3
(microns pk-pk)

(microns pk-pk)
45 45
15
2.5
0 0
2
10
-45 -45
1.5
-90 -90
1
5

0.5 -135 -135

0 -180 0 -180
1 2 3 4 5 6 1 2 3 4 5 6
Harmonic of Shaft Rotation Order Harmonic of Shaft Rotation Order

DR0301 DR0321
Figure 140 Harmonic Analysis of NDE Vibration for DR0301
and DR0321 at 8000 rpm

A summary of the key vibration characteristics for the rigid rotor is given in
Table 26.

Added Shaft X Probe 1X X Probe 1X Y Probe 1X Y Probe 1X X Probe Y Probe

Test Ref Unbalance Speed pk-pk Phase pk-pk Phase Raw pk-pk Raw pk-pk
Mass (g) (rpm) (microns) (degrees) (microns) (microns) (microns) (microns)
DR0301 0 8000 4.1 -112.32 8.02 -18.82 7.12 10.48
DR0303 0.1 8000 3.1 -131.16 8.51 -14.95 7.02 11.07
DR0306 1 8000 4.57 -115.92 10.12 -42.25 7.47 12.21
DR0307 2.5 8000 8.32 -110.69 14.02 -53.7 10.29 15.3
DR0308 5 8000 14.62 -94.81 20.94 -59.9 16.91 22.19
DR0321 7.5 8000 20.52 -90.9 27.71 -60.42 22.91 29.08
Table 26 Results summary of vibration data for DR03 test programme

7.3.1.2 Flexible Rotor Test Programme DR04

This sub section covers some of the vibration results for the flexible rotor test
programme running in fixed geometry bearing, DR04. Four test conditions for
which vibration results are presented here are :

DR0401 No added unbalance mass, oil inlet temp 40oC

DR0407 -2.5g (-100 g.mm) added unbalance position x, no added mass
oil inlet temp 40oC
DR0428 -1.02g (-41 g.mm) added unbalance position x, 3.5kg added
mass, oil inlet temp 40oC
DR0353 -2.5g (-150 g.mm) added unbalance position x shaft centre, oil
inlet temp 40oC

211
Chapter 7 – Experimental Results

Figure 141 to Figure 144 give results data for the flexible rotor without any
added overhung mass applied at the NDE flange.

7.00 180
1X Pk-Pk Vibration

6.00 1X Phase 135

90
5.00
Pk-Pk Vibration (microns)
45

Phase (degrees)
4.00
0
3.00
-45

2.00
-90

1.00 -135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

X Probe
7.00 180
1X Pk-Pk Vibration

1X Phase 135
6.00

90
5.00
Pk-Pk Vibration (microns)

45

Phase (degrees)
4.00
0
3.00
-45

2.00
-90

1.00
-135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

Y Probe
Figure 141 NDE Bearing filtered shaft vibration DR0401 – Run-up Test

25.00 180

1X Pk-Pk Vibration
135
1X Phase
20.00
90
Pk-Pk Vibration (microns)

45
Phase (degrees)

15.00

10.00
-45

-90
5.00

-135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

X Probe

212
Chapter 7 – Experimental Results

25.00 180
1X Pk-Pk Vibration
135
1X Phase
20.00
90

Pk-Pk Vibration (microns)

45

Phase (degrees)
15.00

10.00
-45

-90
5.00

-135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

Y Probe
Figure 142 NDE Bearing Filtered Vibration DR0407 – Run-up Test

15 15

12 12

9 9

6 6

3 3

0 0

-3 -3

-6 -6

-9 -9

-12 -12

-15 -15
-15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15

DE Bearing NDE Bearing

Figure 143 Synchronous (1X) flexible rotor shaft orbits at 6250 rpm - DR0407

15 15

12 12

9 9

6 6

3 3

0 0

-3 -3

-6 -6

-9 -9

-12 -12

-15 -15
-15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15

Figure 144 Synchronous (1X) flexible rotor shaft orbits at 9000 rpm - DR0407

213
Chapter 7 – Experimental Results

The following Figure 145 and Figure 146 are the same vibration
characteristics but now for the rotor with the 3.5 kg added mass applied to the
NDE flange end for test condition DR0428.

25.00 180

1X Pk-Pk Vibration

1X Phase 135

20.00
90
Pk-Pk Vibration (microns)

45
15.00

Phase (degrees)
0

10.00
-45

-90
5.00

-135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

X Probe
20.00 180

1X Pk-Pk Vibration
18.00
1X Phase 135

16.00
90
14.00
Pk-Pk Vibration (microns)

45
12.00

Phase (degrees)
10.00 0

8.00
-45

6.00
-90
4.00

-135
2.00

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

Y Probe
Figure 145 NDE Bearing Filtered Vibration DR0428 – Run-up Test

15 15

12 12

9 9

6 6

3 3

0 0

-3 -3

-6 -6

-9 -9

-12 -12

-15 -15
-15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15

6250 rpm 9000 rpm

Figure 146 NDE Synchronous (1X) flexible rotor shaft orbits - DR0428

214
Chapter 7 – Experimental Results

The final set of figures for the DR04 test programme vibration results are for
the test condition DR0453 where no unbalance is introduced at the NDE
flange but 150 g.mm is applied to the shaft rotor centre. Figure 147 and Figure
148 give the filtered shaft vibration response at the NDE bearing and shaft
orbits respectively.
18.00 180
1X Pk-Pk Vibration
16.00 135
1X Phase

14.00
90
Pk-Pk Vibration (microns)

12.00
45

Phase (degrees)
10.00
0
8.00
-45
6.00

-90
4.00

2.00 -135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

X Probe
18.00 180
1X Pk-Pk Vibration
16.00 135
1X Phase
14.00
90
Pk-Pk Vibration (microns)

12.00
45
10.00 Phase (degrees)
0
8.00
-45
6.00

-90
4.00

2.00 -135

0.00 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Shaft Speed (rpm)

Y Probe
Figure 147 NDE Bearing Filtered Vibration DR0453 – Run-up Test

15 15

12 12

9 9

6 6

3 3

0 0

-3 -3

-6 -6

-9 -9

-12 -12

-15 -15
-15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15

6250 rpm 9000 rpm

Figure 148 NDE Synchronous (1X) flexible rotor shaft orbits - DR0453

215
Chapter 7 – Experimental Results

7.3.1.3 Flexible Rotor Test Programme DR06

This sub section covers some of the vibration results for the flexible rotor test
programme running in variable geometry bearings, DR06. Two test conditions
for which vibration results are presented here are :

DR0651 -No added unbalance, no added mass, oil inlet temp 40oC
DR0473 -1.02 (-41 g.mm) unbalance position x, 3.5kg added mass, oil
inlet temp 40oC

10.00 180
1X Pk-Pk Vibration
9.00
1X Phase 135

8.00
90
7.00
Pk-Pk Vibration (microns)

45
6.00

Phase (degrees)
5.00 0

4.00
-45

3.00
-90
2.00

-135
1.00

0.00 -180
0 2000 4000 6000 8000 10000
Shaft Speed (rpm)

X Probe
10.00 180
1X Pk-Pk Vibration
9.00
1X Phase 135

8.00
90
7.00
Pk-Pk Vibration (microns)

45
Phase (degrees)

6.00

5.00 0

4.00
-45

3.00
-90
2.00

-135
1.00

0.00 -180
0 2000 4000 6000 8000 10000
Shaft Speed (rpm)

Y Probe
Figure 149 NDE Bearing Filtered Vibration DR0651 – Run-up Test

216
Chapter 7 – Experimental Results

25.00 180
1X Pk-Pk Vibration

1X Phase 135

20.00
90

Pk-Pk Vibration (microns)

45

Phase (degrees)
15.00

10.00
-45

-90
5.00

-135

0.00 -180
0 2000 4000 6000 8000 10000
Shaft Speed (rpm)

X Probe
14.00 180
1X Pk-Pk Vibration

12.00 1X Phase 135

90
10.00
Pk-Pk Vibration (microns)

45

Phase (degrees)
8.00

0
6.00
-45

4.00
-90

2.00 -135

0.00 -180
0 2000 4000 6000 8000 10000
Shaft Speed (rpm)

Y Probe
Figure 150 NDE Bearing Filtered Vibration DR0673 – Run-up Test

15 15 15

12 12 12

9 9 9

6 6 6

3 3 3

0 0 0

-3 -3 -3

-6 -6 -6

-9 -9 -9

-12 -12 -12

-15 -15 -15

-15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15 -15 -12 -9 -6 -3 0 3 6 9 12 15

6000 rpm 9500 rpm 10000 rpm

Figure 151 NDE Synchronous (1X) flexible rotor shaft orbits - DR0673

217
Chapter 7 – Experimental Results

7.3.2 Shaft Temperatures

This section presents temperature data for the test bearing and shaft journal
for a number of the test conditions covered by programmes DR03, DR04 and
DR06.

Results data for the test bearing temperatures is presented for all the NDE
measurement locations and referred to test time and operating speed. Also,
for several locations, the circumferential temperature distribution is presented.

Shaft journal temperature differential temperature is presented over the test

duration providing 1st harmonic pk-pk temperature differential and associated
phase. On the same presentation the as-measured pk-pk temperature
differential is also shown. For each steady running condition the journal
circumferential temperature distribution is displayed giving the as measured,
1st harmonic and non harmonic residual. Error bars are included on the as
measured data points to provide an indication of the expected accuracy range
of the sensor based on the individual sensor calibrations.

The results information presented in this section is a summarised selection of

the test data for ease of presentation.

7.3.2.1 Rigid Rotor Fixed Geometry Bearings (DR03 Tests)

DR0301 – No added unbalance 0-12500 rpm

Test (NDE) Bearing Shell Temperatures

70
12 RTD5

65 10 RTD6
Temperature (deg C)

Shaft Speed (1000's rpm)

RTD7
8
60 RTD8
6
RTD9

55 4 RTD10
2 RTD11
50 0 RTD12
RTD13
-2
45 RTD14
-4
Speed
40 -6
0 1000 2000 3000 4000 5000 6000
Time (seconds)

Figure 152 NDE Test Bearing Temperature Measurements Throughout Test

DR0301

218
Chapter 7 – Experimental Results

Bearing Shell Temperature (Oil Inlet 40 deg C) Bearing Shell Temperature (Oil Inlet 40 deg C)
50 70
49
48
65

Temperature (Deg C)
Temperature (Deg C)

47 60
46
45 55
44
43 50
42 Top Pad 45 Top Pad
41 Btm Pad Btm Pad
40 40
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Angle Around Bearing (De gree s) Angle Around Bearing (Degrees)

(a) 2500 rpm (b) 12500 rpm

Figure 153 NDE circumferential bearing shell temperature measurements
DR0301

Shaft and Bearing Temperatures

80 18000

70 16000

14000
60

12000
Temperature (deg C)

Shaft Speed (rpm)

50
10000
40
8000
30
G
6000
R
20 W 4000
B
10 RTD14 2000
Shaft Speed
0 0
0 10 20 30 40 50 60 70 80

% Time of Test

Figure 154 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0301

219
Chapter 7 – Experimental Results

% Time of Test Run

0 10 20 30 40 50 60 70 80
180
1st Harm Phase Ang

Phase Angle (degrees)

135
90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 13000
1st Harm T Diff
12000
1.6 Max Dif
Shaft Speed 11000
1.4
Pk-Pk Temperature Difference

10000
9000
1.2

Shaft Speed (rpm)

8000
1
(deg C)

7000

0.8 6000
5000
0.6
4000

0.4 3000
2000
0.2
1000
0 0
0 10 20 30 40 50 60 70 80

% Time of Test Run

Figure 155 Shaft journal temperature differential measurements for DR0301

Shaft Sensor Temperature

46.3

46.25
Temperature (deg C)

46.2

46.15

Measured

Calc 1st Harm

46.1
Residual

Full 1st Calc

46.05
0 1 2 3 4 5 6
Sensor Position

2500 rpm
Shaft Sensor Temperature

79
Measured
78.95
Calc 1st Harm
78.9
Residual
78.85 Full 1st Calc
Temperature (deg C)

78.8

78.75

78.7

78.65

78.6

78.55

78.5
0 1 2 3 4 5 6

Sensor Position

12500 rpm
Figure 156 Journal Circumferential Temperatures for DR0301

220
Chapter 7 – Experimental Results

DR0321 – 7.5 g unbalance 0-8000 rpm (position 4)

Test (NDE) Bearing Shell Temperatures

8 RTD5
60 RTD6

Shaft Speed (1000's rpm)

Temperature (deg C) 6 RTD7

RTD8
55
4 RTD9

RTD10
50 RTD11
2
RTD12

RTD13
45 0
RTD14
Speed
40 -2
0 1000 2000 3000 4000 5000 6000
Time (s)

Figure 157 NDE Test Bearing Temperature Measurements Throughout Test

DR0321

% Time of Test Run

0 10 20 30 40 50 60
180
135 1st Harm Phase Ang
Phase Angle (degrees)

90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 9000
1st Harm T Diff
1.6 Max Dif 8000
Shaft Speed
Pk-Pk Temperature Difference (deg

1.4 7000

1.2 6000
Shaft Speed (rpm)

1 5000
C)

0.8 4000

0.6 3000

0.4 2000

0.2 1000

0 0
0 10 20 30 40 50 60
% Time of Test Run

Figure 158 Shaft journal temperature differential measurements for DR0321

221
 Chapter 7 – Experimental Results

Shaft Sensor Temperature Shaft Sensor Temperature

43.26
47.95

43.24
47.9

43.22
47.85

Temperature (deg C)
Temperature (deg C)

43.2
47.8
43.18
47.75
43.16
47.7 Actual
Actual
43.14 Calc 1st Harm
Calc 1st Harm
47.65 Residual
43.12 Residual
Full 1st Calc
Full 1st Calc 47.6
43.1 0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position Sensor Position

1000 rpm 3000 rpm

Shaft Sensor Temperature

66

65.8

65.6

65.4
Temperature (deg C)

65.2

65

64.8

Actual
64.6
Calc 1st Harm

64.4 Residual
Full 1st Calc
64.2
0 1 2 3 4 5 6
Sensor Position

8000 rpm
Figure 159 Journal Circumferential Temperatures for DR0321

DR0399 - 7.5 g unbalance 0-8000 rpm (position 1)

Test (NDE) Bearing Shell Temperatures

8
RTD5
60 7
Shaft Speed (1000's rpm)

RTD6
Temperature (deg C)

6 RTD7

RTD8
55 5
RTD9
4
RTD10
50 3 RTD11

2 RTD12

RTD13
45 1
RTD14
0
Speed
40 -1
12900 13900 14900 15900 16900 17900 18900
Time (s)

Figure 160 NDE Test Bearing Temperature Measurements Throughout Test

DR0321

222
 Chapter 7 – Experimental Results

% Time of Test Run

62 67 72 77 82 87 92 97
180
1st Harm Phase Ang
135

Phase Angle (degrees)

90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 9000
1st Harm T Diff
Max Dif
1.6 8000
Shaft Speed
Pk-Pk Temperature Difference (deg C)

1.4 7000

1.2 6000

Shaft Speed (rpm)

1 5000

0.8 4000

0.6 3000

0.4 2000

0.2 1000

0 0
62 67 72 77 82 87 92 97
% Time of Test Run

Figure 161 Shaft journal temperature differential measurements for DR0399

Shaft Sensor Temperature Shaft Sensor Temperature

66.6
45.74 Measured
Calc 1st Harm
45.72
66.4 Residual
45.7 Full 1st Calc

66.2
Temperature (deg C)

45.68
Temperature (deg C)

45.66
66

45.64

65.8
45.62

Measured
45.6
Calc 1st Harm 65.6
45.58 Residual
Full 1st Calc
45.56 65.4
0 1 2 3 4 5 6 0 1 2 3 4 5 6

Sensor Position Sensor Position

1000 rpm 8000 rpm

Figure 162 Journal Circumferential Temperatures for DR0399

223
 Chapter 7 – Experimental Results

7.3.2.1 Flexible Rotor Fixed Geometry Bearings (DR04 Tests)

DR0401 – 0 g unbalance 0-9000 rpm 1.5 kg OHM

Test (NDE) Bearing Shell Temperatures

65 8 RTD5
RTD6

Shaft Speed (1000's rpm)

Temperature (deg C)

RTD7
60 6 RTD8
RTD9
55 4 RTD10
RTD11

50 2 RTD12
RTD13
RTD14
45 0
Inlet
Speed
40 -2
0 2000 4000 6000 8000
Time (s)

Figure 163 NDE Test Bearing Temperature Measurements Throughout Test

DR0321
Bearing Shell Temperature (Oil Inlet 40 deg C) Bearing Shell Temperature (Oil Inlet 40 deg C)
50
49 65

48
60
Temperature (Deg C)
Temperature (Deg C)

47
46
55
45
44 50
43
42 45
Top Pad Top Pad
41 Btm Pad
Btm Pad
40 40
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Angle Around Bearing (Degrees) Angle Around Bearing (Degrees)

(a) 3000 rpm (b) 9000 rpm

Figure 164 NDE circumferential bearing shell temperature measurements
DR0401
Shaft and Bearing Temperatures
80 10000
G

75 R 9000
B
8000
70 W
RTD14 7000
Temperature (deg C)

Shaft Speed (rpm)

65 Shaft Speed
6000
60 5000

55 4000

3000
50
2000
45
1000
40 0

35 -1000
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000
Time (s)

Figure 165 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0401

224
 Chapter 7 – Experimental Results

% Time of Test Run

0 10 20 30 40 50 60 70 80 90
180
135

Phase Angle (degrees)

90
45
0
-45
-90
-135 1st Harm Phase Ang

-180

Shaft Temperature Difference

1.8 9000
1st Harm T Diff
Max Dif
1.6 Shaft Speed 8000
Pk-Pk Temperature Difference (deg C)

1.4 7000

1.2 6000

Shaft Speed (rpm)

1 5000

0.8 4000

0.6 3000

0.4 2000

0.2 1000

0 0
0 10 20 30 40 50 60 70 80 90

% Time of Test Run

Figure 166 Shaft journal temperature differential measurements for DR0401

Shaft Sensor Temperature Shaft Sensor Temperature

58
47.96
Actual
Actual
Calc 1st Harm
47.94 Calc 1st Harm 57.95
Residual
Residual
Full 1st Calc
47.92 Full 1st Calc
Temperature (deg C)

57.9
Temperature (deg C)

47.9

57.85
47.88

57.8
47.86

47.84 57.75

47.82
0 1 2 3 4 5 6 57.7
0 1 2 3 4 5 6
Sensor Position
Sensor Position

3000 rpm 6000 rpm

Shaft Sensor Temperature
68.35

68.3

68.25
Temperature (deg C)

68.2

68.15

68.1

68.05 Actual
Calc 1st Harm
68 Residual
Full 1st Calc
67.95
0 1 2 3 4 5 6
Sensor Position

9000 rpm
Figure 167 Journal Circumferential Temperatures for DR0401

225
 Chapter 7 – Experimental Results

DR0407 - 2.5 g unbalance 0-9000 rpm (postion 4) 1.5 kg OHM

% Time of Test Run

0 10 20 30 40 50 60 70
180
135 1st Harm Phase Ang

Phase Angle (degrees)

90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 10000
1st Harm T Dif f
Max Dif 9000
1.6
Shaf t Speed

8000
Pk-Pk Temperature Difference (deg C)

1.4

7000
1.2

Shaft Speed (rpm)

6000
1
5000
0.8
4000
0.6
3000

0.4
2000

0.2 1000

0 0
0 10 20 30 40 50 60 70

% Time of Test Run

Figure 168 Shaft journal temperature differential measurements for DR0407

Shaft Sensor Temperature Shaft Sensor Temperature

60.5 68.6
Measured

60.4 Calc 1st Harm

68.4
Residual
60.3 68.2
Full 1st Calc
Temperature (deg C)

Temperature (deg C)

60.2 68

60.1 67.8

60 67.6

Measured
59.9 67.4
Calc 1st Harm

59.8 67.2 Residual

Full 1st Calc
59.7 67
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Sensor Position Sensor Position

6500 rpm 9000 rpm

Figure 169 Journal Circumferential Temperatures for DR0401

226
Chapter 7 – Experimental Results

DR0427 – 0 g unbalance 0-900 rpm @ flange 5 kg OHM

Test (NDE) Bearing Shell Temperatures

9
65 8
RTD5
RTD6

Shaft Speed (1000's rpm)

Temperature (deg C) 7 RTD7
60
6 RTD8
RTD9
5
55 RTD10
4
RTD11

50 3 RTD12

2 RTD13
RTD14
45 1
Inlet
0
Speed
40 -1
500 2500 4500 6500 8500 10500
Time (s)

Figure 170 NDE Test Bearing Temperature Measurements Throughout Test

DR0427

Bearing Shell Temperature (Oil Inlet 40 deg C)

65

60
Temperature (Deg C)

55

50

45 Top Pad
Btm Pad
40
0 60 120 180 240 300 360
Angle Around Bearing (Degrees)

9000
Figure 171 NDE circumferential bearing shell temperature measurements
DR0427
Shaft and Bearing Temperatures
80 10000
G

75 R 9000
B
8000
70 W
RTD14
7000
Temperature (deg C)

Shaft Speed (rpm)

65 Shaft Speed
6000
60
5000
55
4000
50
3000
45
2000

40 1000

35 0
0 2,000 4,000 6,000 8,000 10,000
Time (s)

Figure 172 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0427

227
 Chapter 7 – Experimental Results

% Time of Test Run

5 15 25 35 45 55 65 75 85
180
135 1st Harm Phase Ang

Phase Angle (degrees)

90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 10000
1st Harm T Diff

1.6 Max Dif 9000

Shaft Speed
Pk-Pk Temperature Difference (deg C)

8000
1.4
7000
1.2

Shaft Speed (rpm)

6000
1
5000
0.8
4000
0.6
3000
0.4
2000

0.2 1000

0 0
5 15 25 35 45 55 65 75 85

% Time of Test Run

Figure 173 Shaft journal temperature differential measurements for DR0427

Shaft Sensor Temperature Shaft Sensor Temperature

58.3
Measured 67.05

Calc 1st Harm

Residual 67
58.25
Full 1st Calc
66.95
Temperature (deg C)

Temperature (deg C)

58.2
66.9

58.15 66.85

66.8 Measured
58.1
Calc 1st Harm
66.75 Residual
Full 1st Calc
58.05
66.7
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position Sensor Position

6250 rpm 9000 rpm

Figure 174 Journal Circumferential Temperatures for DR0427

228
Chapter 7 – Experimental Results

DR0428 – 1g unbalance @ flange 0-9000 rpm (position 4) 5 kg OHM

% Time of Test Run
0 10 20 30 40 50 60 70 80 90 100
180
135

Phase Angle (degrees)

90
45
0
-45
-90
-135 1st Harm Phase Ang

-180

Shaft Temperature Difference

1.8 12000
1st Harm T Diff

1.6 Max Dif

Shaft Speed 10000
Pk-Pk Temperature Difference (deg C)

1.4

1.2 8000

Shaft Speed (rpm)

1
6000
0.8

0.6 4000

0.4
2000
0.2

0 0
5 15 25 35 45 55 65 75
% Time of Test Run

Figure 175 Shaft journal temperature differential measurements for DR0428

Shaft Sensor Temperature Shaft Sensor Temperature

58.4 69.7
Measured
69.6
Calc 1st Harm
58.35 69.5
Residual

Full 1st Calc 69.4

Temperature (deg C)

58.3
Temperature (deg C)

69.3

69.2
58.25
69.1

58.2 69 Measured

68.9 Calc 1st Harm

58.15 Residual
68.8
Full 1st Calc
68.7
58.1
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position Sensor Position

6250 rpm 10000 rpm

Shaft Sensor Temperature
73.6

73.4

73.2

73
Temperature (deg C)

72.8

72.6

72.4
Measured
72.2 Calc 1st Harm

Residual
72
Full 1st Calc
71.8
0 1 2 3 4 5 6
Sensor Position

11000 rpm
Figure 176 Journal Circumferential Temperatures for DR0428

229
Chapter 7 – Experimental Results

DR0453 – 2.5 g @ centre unbalance 0-9000 rpm (positon 1) 1.5 kg OHM

Test (NDE) Bearing Shell Temperatures

65 8 RTD5
RTD6

Shaft Speed (1000's rpm)

Temperature (deg C)
RTD7
60 6 RTD8
RTD9
55 4 RTD10

RTD11

50 2 RTD12
RTD13
RTD14
45 0
Inlet

Speed
40 -2
360 2360 4360 6360 8360
Time (s)

Figure 177 NDE Test Bearing Temperature Measurements Throughout Test

DR0453
Bearing Shell Temperature (Oil Inlet 40 deg C)

65

60
Temperature (Deg C)

55

50

45 Top Pad
Btm Pad
40
0 60 120 180 240 300 360
Angle Around Bearing (Degrees)

9000
Figure 178 NDE circumferential bearing shell temperature measurements
DR0453
Shaft and Bearing Temperatures
80 10000
G
R 9000
75
B
8000
W
70
RTD14 7000
Temperature (deg C)

Shaft Speed (rpm)

Shaft Speed
65 6000

5000
60
4000
55 3000

2000
50
1000
45
0

40 -1000
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
Time (s)

Figure 179 Shaft Journal and NDE Bearing (RDT14) Temperatures for
DR0453

230
Chapter 7 – Experimental Results

% Time of Test Run

5 15 25 35 45 55 65 75 85 95
180
135

Phase Angle (degrees)

90
45
0
-45
-90
-135 1st Harm Phase Ang

-180

Shaft Temperature Difference

1.8 10000
1st Harm T Diff
Max Dif
1.6 9000
Shaft Speed

8000
Pk-Pk Temperature Difference (deg C)

1.4

7000
1.2

Shaft Speed (rpm)

6000
1
5000
0.8
4000
0.6
3000

0.4
2000

0.2 1000

0 0
5 15 25 35 45 55 65 75 85 95
% Time of Test Run

Figure 180 Shaft journal temperature differential measurements for DR0453

Shaft Sensor Temperature Shaft Sensor Temperature

58.9
67
Actual
58.8 Calc 1st Harm 66.95
Residual
58.7 66.9
Full 1st Calc
Temperature (deg C)

58.6 66.85
Temperature (deg C)

66.8
58.5

66.75
58.4
66.7
58.3 Actual
66.65 Calc 1st Harm
58.2
Residual
66.6
Full 1st Calc
58.1
66.55
0 1 2 3 4 5 6
58
0 1 2 3 4 5 6 Sensor Position
Sensor Position

6450 rpm 9000 rpm

Figure 181 Journal Circumferential Temperatures for DR0453

231
 Chapter 7 – Experimental Results

7.3.2.1 Flexible Rotor Variable Geometry Bearings (DR06 Tests)

DR0651 – 0 g unbalance @ flange 0-9000 rpm 5 kg OHM

Test (NDE) Bearing Pad Temperatures

RTD5
60 8
RTD6

Shaft Speed (1000's rpm)

Temperature (deg C)

RTD7
6 RTD8
55
RTD9
4 RTD10

50 RTD11

2 RTD12
RTD13
45 RTD14
0
Inlet
Speed
40 -2
0 5000 10000 15000
Time (s)

Figure 182 NDE Test Bearing Temperature Measurements Throughout Test

DR0651

Bearing Pad Temperature (Oil Inlet 40 deg C) Bearing Pad Temperature (Oil Inlet 40 deg C)

65 Pad 1 Pad 2 65
Pad 3 Pad 4
Pad 5
60 60
Temperature (Deg C)

Temperature (Deg C)

∆T=1.15 ∆T=3.52 ∆T=4.46

55 55 ∆T=3.26
∆T=2.34

∆T=-0.52 ∆T=1.38 ∆T=2.26

50 50
∆T=1.43 ∆T=1.25
Pad 1 Pad 2
45 45
Pad 3 Pad 4
Pad 5
40 40
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Angle Around Bearing (Degrees) Angle Around Bearing (Degrees)

4000 rpm 9000 rpm

Figure 183 NDE bearing pad temperature measurements DR0651

Shaft and Bearing Temperatures

80 10000
G
75 R 9000
B
70 W 8000
RTD14
65 7000
Temperature (deg C)

Shaft Speed (rpm)

Shaft Speed
60 6000

55 5000

50 4000

45 3000

40 2000

35 1000

30 0
0 2,000 4,000 6,000 8,000 10,000 12,000
Time (s)

Figure 184 Shaft Journal and NDE Bearing Temperatures for DR0651

232
Chapter 7 – Experimental Results

% Time of Test Run

5 15 25 35 45 55 65 75 85 95
180
135 1st Harm Phase Ang

Phase Angle (degrees)

90
45
0
-45
-90
-135
-180

Shaft Temperature Difference

1.8 9000
1st Harm T Diff
1.6 Max Dif 8000
Shaft Speed
1.4 7000
Pk-Pk Temperature Difference

1.2 6000

Shaft Speed (rpm)

1 5000
(deg C)

0.8 4000

0.6 3000

0.4 2000

0.2 1000

0 0
5 15 25 35 45 55 65 75 85 95
% Time of Test Run

Figure 185 Shaft journal temperature differential measurements for DR0651

Shaft Sensor Temperature Shaft Sensor Temperature
43.94
54.6

43.92
54.5

43.9
54.4
Temperature (deg C)

Temperature (deg C)

43.88
54.3

43.86
54.2

43.84 Actual 54.1 Actual

Calc 1st Harm
Calc 1st Harm
43.82 Residual 54 Residual
Full 1st Calc
Full 1st Calc
43.8
53.9
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position
Sensor Position

1000 rpm 6250 rpm

Shaft Sensor Temperature

59.3

59.2

59.1
Temperature (deg C)

59

58.9

58.8

Actual
58.7
Calc 1st Harm

58.6 Residual
Full 1st Calc
58.5
0 1 2 3 4 5 6
Sensor Position

9000 rpm
Figure 186 Journal Circumferential Temperatures for DR0651

233
 Chapter 7 – Experimental Results

DR0673 - 1 g unbalance @ flange 0-9000 rpm (position 4) 5 kg OHM

Test (NDE) Bearing Pad Temperatures

60 9
RTD5
58 8
RTD6

Shaft Speed (1000's rpm)

Temperature (deg C) 56 7 RTD7
54 6 RTD8

RTD9
52 5
RTD10
50 4
RTD11
48 3 RTD12

46 2 RTD13

RTD14
44 1
Inlet
42 0
Speed
40 -1
0 5000 10000
Time (s)

Figure 187 NDE Test Bearing Temperature Measurements Throughout Test

DR0673

Bearing Pad Temperature (Oil Inlet 40 deg C) Bearing Pad Temperature (Oil Inlet 40 deg C)

65 Pad 1 Pad 2 65
Pad 3 Pad 4
60 Pad 5 60
Temperature (Deg C)

Temperature (Deg C)

∆T=4.38
∆T=5.34
55 55
∆T=4.23
∆T=1.67 ∆T=3.94 ∆T=2.9
∆T=2.81
50 ∆T=2.21 50
∆T=1.76 ∆T=1.42

Pad 1 Pad 2
45 45
Pad 3 Pad 4
Pad 5
40 40
0 60 120 180 240 300 360 0 60 120 180 240 300 360
Angle Around Bearing (Degrees) Angle Around Bearing (Degrees)

5000 rpm 10000 rpm

Figure 188 NDE bearing pad temperature measurements DR0673

234
Chapter 7 – Experimental Results

% Time of Test Run

0 10 20 30 40 50 60 70 80 90
180
135

Phase Angle (degrees)

90
45
0
-45
-90
-135 1st Harm Phase Ang

-180

Shaft Temperature Difference

1.8
1st Harm T Diff 10000
1.6 Max Dif
Shaft Speed
1.4 8000
Pk-Pk Temperature Difference

1.2

Shaft Speed (rpm)

1 6000
(deg C)

0.8
4000
0.6

0.4
2000
0.2

0 0
0 10 20 30 40 50 60 70 80 90
% Time of Test Run

Figure 189 Shaft journal temperature differential measurements for DR0673

Shaft Sensor Temperature Shaft Sensor Temperature
53.6
43.25
Measured
Calc 1st Harm
53.4
Residual
43.2
Full 1st Calc

53.2
Temperature (deg C)

Temperature (deg C)

43.15

53

43.1

52.8

43.05 Measured
52.6 Calc 1st Harm
Residual
Full 1st Calc
43
52.4
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position Sensor Position

2200 rpm 6250 rpm

Shaft Sensor Temperature

58.8

58.6

58.4
Temperature (deg C)

58.2

58

57.8

57.6 Measured
Calc 1st Harm
57.4 Residual
Full 1st Calc
57.2
0 1 2 3 4 5 6
Sensor Position

10000 rpm
Figure 190 Journal Circumferential Temperatures for DR0673

235
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Chapter 8

Analytical Evaluation and Development of

Journal Temperature Distribution

8. Analytical Evaluation and Development of Journal

Temperature Distribution

8.1 Introduction

This chapter presents some methods developed for the qualitative and
quantitative description of bearing journal temperature distribution. The
principal focus of the work presented here is to attempt to provide a means of
suitable approximation for the expected temperature distribution around the
circumference of an whirling journal without the requirement for complex CFD
type analysis. The intention is to present a method which can be developed
for use in conjunction with typical rotordynamic packages for the coupled
evaluation of the rotor response and differential journal heating stability. The
work presented here provides only an initial presentation of some new
methods for the modelling of the journal thermal conditions and further
development and development is intended to provide models suitable for
integration into existing rotordynamic procedures.

Two distinct aspects of evaluation are tackled. The first is associated with
introducing an approach that allows for the effects of more complex bearing
geometries to be evaluated for a specific vibration condition. Here the
temperature distribution can be described with suitable resolution for the
entire journal circumference. The intent here is not to develop complex
thermohydrodynamic analysis models, but to apply the fundamental principles
of the energy and temperature relationships within a hydrodynamic bearing
and develop expressions and/or numerical procedures to allow for the
application to journal temperature differential heating. The principle objective
of this modelling work is to develop a suitable means of establishing the
thermal boundary conditions that can be applied to a suitable model of the
bearing journal. In particular, specific focus will be placed on the development

236
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

of the 1st harmonic temperature differential and arguably more significantly its
phase relative to the point of minimum film thickness.

The second area of analytical assessment addresses the transfer of heat

energy into the journal from the oil film. The analysis procedure is based
around the development of a model of the shaft bearing journal where a pre-
described thermal boundary condition is applied to the shaft circumference.
The applied boundary condition is dynamic and is related to the associated
operating speed of the shaft. This allows for a description of the shaft thermal
distribution in the time domain to be developed. The method used to approach
this problem was by the application of the Finite Difference Method. The
results from the shaft temperature distribution are then used in a closed form
algorithm to determine the associated shaft thermal bend angle and resulting
shaft unbalance vector. A more detailed description of the analytical methods
considered and the technique used are provided later in this chapter.

8.2 Finite Difference Heat Transfer Model

To enable a measure of the time dependant nature of the thermal bow

development and corresponding dynamic interaction, a means of
understanding the temperature development within the shaft is needed. As
has been outlined in the literature, several methods for carrying out this type
of analysis have been presented. In all cases reviewed thus far the techniques
that have been employed are highly analytically intensive and either
computationally intensive such as the CFD approaches or analytically
complex to the point where practical implementation of real dynamic systems
becomes impractical.

The intention of the modelling work presented here is to provide a

computational technique which allows a practical compromise between the
computational complexity and the notably simplified time independent
methods as presented in the literature. When considering the problem the
various techniques that can be applied were considered with respect to the
level of analytical complexity and the probable computation requirements
needed to provide solutions of acceptable accuracy. Following some initial
evaluation, it was considered that a form of numerical solution was more
preferable to that of a detailed analytical description. The reason was to
maintain flexibility to enable application to real dynamic systems and features.
In addition, the application of an analytical approach would require significant

237
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

assumption with relation to the real features or bearing geometries present

within practical rotor systems, some of which would require some form of
numerical solution in any case due to complex features or interactions being
present. The implication of discontinuity features in analytical systems can
introduce significant complexity whilst still requiring significant assumptions
about those discontinuity features such as bearing oil inlets or the complex
boundary conditions created by the oil film temperature and energy states.
Whilst the accuracy or certainty around these features when incorporated into
a numerical analysis offer little if any gain over that of the analytical methods
the implication and control of such areas is much more readily implemented
and more importantly adjusted.

A number of numerical techniques are available for the solution of such

problems but it is the finite difference modelling technique which is adopted
here. The use of finite elements or finite volume techniques were considered
however, for the solution of the temperature distribution in the shaft journal,
the finite difference technique allows for the most effective and least complex
model to be developed. The use of the finite volume technique (in place of the
differencing technique) was considered as this would allow for the
thermohydrodynmic modelling of the oil film to be coupled with the shaft
journal. However an approach was taken to attempt to separate the solution
for the boundary conditions from the conduction heat transfer analysis of the
shaft. The benefit is maintaining a practically applicable and quick (less
computationally intensive) solution. It is important that if a practical approach
can be implemented, the results are generated with quick, efficient
computational procedures.

8.2.1 The Heat Transfer Model

The principle of the temperature distribution within the shaft journal is

governed by the transient heat flux at the journal surface and the flow of heat
within the shaft section. The one dimensional flow of heat within a body is
given by Fourier’s differential equation of heat conduction:

d 2T ρ .c
+ = 0 eq. 8.2.1
dx 2 k

238
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

The above differential equation (eq 8.2.1) provides a description of the steady
state heat transfer in a body. For the time dependent problem of the shaft
journal it is therefore necessary to introduce a time dependant relationship
with the rate of heat conduction known as the diffusion term. By introducing a
time dependant temperature term into equation 8.1 ∂T the one dimensional
∂t
equation for heat diffusivity can be written :

∂ 2T ρ .c ∂T
+ . = 0 eq. 8.2.2
∂x 2 k ∂t

where

ρ.c ∂T 1 ∂T
. = .
k ∂t α ∂t

k
α= is the thermal diffusivity of the body which is a measure of how
ρ .c
quickly a material can transfer heat from a higher energy location to a lower
one.

Whilst it is possible to derive analytical solutions to the heat diffusion

differential equation for specific bodies, the assumptions and simplifications
required, significantly limit the return on analytical effort when handling the
difficult boundary conditions presented by the whirling shaft in the
hydrodynamic bearing. The application of the finite difference technique
allows for closely approximated solutions to the heat transfer problem and the
definition of complex dynamic boundary conditions and discontinuity features
within the bearing.

8.2.2 The finite Difference Technique

As with all the numerical the solution techniques, the governing equations are
solved by means of representing the differential terms using an algebraic
relationship. The finite difference method uses this approach by substitution
the derivative terms with difference terms.

The principle of the finite difference method is based on approximating the

derivative equation with a linear interpolation between successive points. If we

239
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

take Figure 191 to represent a second order derivative (as presented in the
heat equations) then by dividing the X axis into a number of sections, it is
possible to establish an approximation of the first derivative by taking the
difference between the adjacent points along the curve. As the heat equation
is a second order derivative, the first order derivative in finite difference terms
must first be established.

Figure 191 Discretisation of the analytical derivative solution [60]

To solve the second order partial derivatives presented in equation 8.2.2 the
first and second order derivatives need to expressed in finite difference
formulation.

The 1st order approximation with central differencing is given by :

dT Tm − Tm−1
≅
dx m−
1 ∆x
2

And Eq. 8..2.3

dT Tm +1 − Tm
≅
dx m+
1 ∆x
2

And from this the formulation for the second order derivative can be
determined:

dT dT
− Tm +1 − Tm Tm − Tm−1
dx 1 dx 1 −
d 2T m+
2
m−
2 ∆ x ∆x
≅ = Eq. 8.2.4
dx 2 m−
1 ∆x ∆x
2

240
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Therefore

1
m+
d 2T 2 Tm +1 − 2.Tm + Tm−1
≅ Eq 8.2.5
dx 2 m−
1 ∆x 2
2

Where ∆x represents the dimensional grid spacing in the dimension of heat

transfer. Applying the above approximation for the second derivative to the
heat equation 8.2.2, gives the finite difference equation for heat transfer :

d 2T ρ .c Tm+1 − 2.Tm + Tm −1 ρ .c
+ ≅ + = 0 Eq. 8.2.6
dx 2 k ∆x 2 k

The application of the finite difference method requires the medium under
consideration to be discretised by a series of control volumes with an interior
node to which the heat fluxes are subjected. The discretisation of a domain
must typically be of uniform distribution across the generated grid to ensure
continuity across the differencing (although there are techniques that can be
applied to overcome this restriction such as Sundqvist et al [66]). A typical one
dimensional gird is shown in Figure 192 where it can be seen at the boundary
nodes consideration must be given to the control volume now being ½ of the
discretisation increment.

∆x ∆x
2 ∆x ∆x ∆x ∆x ∆x ∆x 2

J1 J2 J3 J4 J i −3 J i−2 J i −1 Ji

Figure 192 1 dimensional finite difference discreatisation grid with equal spacing

8.2.2.1 Boundary Conditions

Where external nodes exist, care must be taken as the finite difference
formulation needs to be considered differently at the these locations. Typically
external nodes are typically located at the boundary conditions. As such the
descriptions of the conditions are based on the specific type of boundary
condition being applied. The main principle behind the conditions required at
external nodes is that formulations are written such that they observe the
energy balance:

241
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

. .
∑
m +1, m −1, n + , n −1
Q + Q m,n = 0

The specific formulation of the conditions for external nodes is specific to the
type of boundary condition applied. For the purposes of the analysis of the
shaft journal, the boundary conditions considered are that of an insulated
boundary, a convective boundary(heat flux and temperature specified
boundary constraint. The specific temperature boundary condition is the
simplest of the boundary conditions as the formulation is based on
temperature values being defined or assigned specifically to an external node.
Where :

Tn = T∞ Eq. 8.2.7

The formulation of the convection and heat flux boundary conditions can be
constructed as follows for a 1 dimensional case :

.
(T − T )  ∆x 
0 = h. A.(T∞ − To ) + k . A 1 o + qo . A.   Eq. 8.2.8a
∆x  2 

.
(T − T )  ∆x 
0 = qo . A + k . A 1 o + qo . A.   Eq. 8.2.8b
∆x  2 

Where the associated differential equation is given by :

∂T
0 = k. + h.T − h.T∞ Eq 8.2.9
∂x

8.2.2.2 Discretisation Accuracy

Equations 8.2.5 and 8.2.6 are of the 1st order approximation and with this
there is an associated error in its approximation. It is possible to increase the
accuracy of the approximations of the finite difference equations by the
application of higher order formulations. A brief study of the relative
accuracies and computational requirements involved was made to establish
which formulation would provide the most suitable when applied to the

242
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

differencing model of the rotor. Second and third order formulations for the
first and second derivatives were also explored.

The results from the finite differencing analysis evaluation were assessed and
the relative accuracies compared for the three formulations of both 1st and 2nd
order derivatives, observation is made as to the level of increased complexity
and computational effort incurred with each solution. The level of increased
computational effort required for the higher order equations is significant as
when a finite difference grid is constructed, the calculations will be performed
many thousands of times which can have a profound effect on the associated
solution times even on a modern computer.

From the evaluation it was considered that sufficient accuracy can be

obtained with the application of the 1st order approximations when used with a
sufficiently refined grid. The use of the lower order approximation means that
the CPU intensity is kept to a minimum to speed up solution times with
acceptable impact on solution accuracy.

8.2.2.3 Two Dimensional Grid Formulation

The above equations have been presented for a 1 dimensional

representation. The analysis of the bearing journal requires that a two
dimensional grid be developed. The application of the above formulations can
be readily applied to a 2 dimensional grid construction. For a 2D grid with
equal spacing the 1st order formulation for a given node will be :

.
Tm +1,n − 2.Tm,n + Tm −1,n Tm,n +1 − 2.Tm ,n + Tm,n −1 q m ,n
+ + = 0 Eq. 8.2.10
∆x 2 ∆y 2 k

Writing equation 8.2.10 for dTm, n and introducing the implicit transient solution
using α gives :

Tm+1, n − 2.Tm, n + Tm−1, n Tm ,n +1 − 2.Tm,n + Tm ,n −1

dT = α .dt. 2
+ α .dt. Eq. 8.2.11
∆x ∆y 2

243
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Polar Coordinates

To suitably represent the 2 dimensional section or the shaft journal it is

necessary to develop the above finite difference equations in the polar
coordinate system. The control volume grid construction and coordinate
system as presented in Figure 193 is used.

qr

qθ
q r + ∆r qθ + ∆ θ

∆θ
∆r

Figure 193 Finite difference control volume and node

construction in polar coordinates

Applying the same formulation principles as for the Cartesian equations the
energy balance for an internal node m can be developed as :

qr .(r − 0.5.∆r ).∆θ − qr +∆r .(r + 0.5.∆r ).∆θ + qθ .∆r − qθ +∆θ .∆r.....
Eq. 8.2.12
+ qg .r.∆θ .∆r = 0

Where r , θ , ∆r and ∆θ are the polar terms used for describing the control
volume.

244
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Where the heat fluxes can be expressed by:

. Tm ,n −1 − Tm,n
qr = k 1 . (a)
m,n −
2
∆r

. Tm,n − Tm ,n +1
q r +∆r = k 1 . (b)
m,n+
2
∆r
Eq . 8.2.13
. Tm−1,n − Tm, n
qθ = k 1 . (c)
m − ,n
2
r.∆θ

. Tm,n − Tm,n +1
qθ +∆θ = k 1 . (d)
m,n +
2
r.∆θ

Combining equations 8.2.13 and substituting into 8.2.12 yields the finite
difference formulation for a 2D internal node in polar coordinates, also
substituting steady state heat transfer constant k for transient heat diffusion
α gives :

α .dt
dT = . Tm +1,n − 2.Tm,n + Tm−1, n  + α .dt. Tm ,n +1 − 2.Tm, n + Tm, n−1  + ....
r2 
Eq. 8.2.14
α .dt
. Tm, n+1 − Tm,n −1 
2.r 

The central node

A feature presented in the polar coordinate system not experienced in the

cartesian grid construction is the node at the centre of a solid shaft or disc.
The central node is a singularity and as such cannot be treated with the
previously presented equations for other internal nodes. The formulation of
the conditions for the central node derived here are consistent with those
presented by Muneer et al [61].

∆r
Considering the control volume of as shown in Figure 193 the energy
2
balance equation can be expressed as

245
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

2
. . . . ∆r.∆θ .  ∆r 
0 = −(q r1,2 + q r1,2 + q r1,2 + ...... + q r1,2 ). + q g .π .   Eq. 8.2.15
2  2 
Were i is the number of segments connected to the central node within the
central control volume boundary qi , j and is the radial heat flux between the
corresponding control volume boundary from nodes i and j. Where qij can be
expressed as

. T1,1 − Tm,2
q r .m,2 = k Eq. 8.2.16
∆r

Substituting into equation 8.2.14, introducing transient conditions and

simplifying yields the finite difference formulation for the central node with
respect to time :

4.α
dT ≅ .(T1,2 + T2,2 + T3,2 + ........ + Ti ,2 ) − T1,1 Eq. 8.2.17
i.∆r 2

It is also necessary to express the boundary conditions in the polar coordinate

system. For the convective boundary condition the finite difference equation
for transient heat transfer as :

α .dt α .2.(r − 0.5.∆r ).dt

dT = . Tm+1,n − 2.Tm,n + Tm−1,n  + . Tm ,n −1 − Tm,n  + .....
 ∆r  2  ∆r 
r −  .r.∆θ r −  ∆r
 4   4 
2.h.r.dt
........ . T − T 
 ∆r   ∞ m ,n 
ρ .cs .∆r.  r − 
 4 
Eq. 8.2.18

Where T∞ is the temperature of the fluid at the boundary with the solid
surface.

8.2.2.4 Numerical Stability

When undertaking transient finite difference analysis using an implicit method,

it is possible to result in unstable numerical isolations. In this case the time

246
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

steps involved between each iteration can become too large and result in and
oscillating divergent and unstable nodal calculations.

The conditions under which unstable characteristics are observed is when the
time step increment in relation to the grid spacing results in errors that are
either increasing in size or at a rate greater than the rate of convergence. The
errors are generally in the form of truncation or discretisation errors where the
error is a direct result of the error in the difference approximation, and round-
off errors where the numerical precision in the calculations is insufficient and
the effect accumulates with each increment or time step. The latter of the two
error sources is not typically a problem with modern high bit computers but the
grid discretisation is influential and is managed by means of reducing grid size
relative to time step ∆t . It is possible to determine a minimum time step/mesh
size to ensure stability is maintained across the difference mesh, and is given
by Li [62] for the heat equations as in equation 8.2.19 , where the minimum
time step required for a given ∆x is :

∨
∆x 2
dt ≤ Eq. 8.2.19
2.α

The problem with this time step stability requirement is that in transient
analysis it can result in much higher computational effort to evaluate a given
time period. The issue is that as the mesh density to improve accuracy is
increased a reduction in the time step is necessary, resulting in more solutions
per time step and increased number of time steps. This compounding effect
significantly increases the processing times required to solve the models. A
notable additional effect in the polar system is that, as the radii approaches
the centre, the arc length of the circumferential control volumes become very
small. This is not a problem typically encountered in Cartesian grids due to the
grid spacing typically being evenly spaced across the entire model.

247
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

High element
concentration

Figure 194 Small grid discreatisation effect around centre node in polar
coordinates

With the above conditions in mind, the finite difference model grid resolution
should be kept to as course grid as possible (within acceptable accuracy
limits) to provide the most effective model solution durations.

The application the finite difference relationships developed above now

provide a means by which a two dimensional representation of the shaft
journal section at the bearing location can be modelled in the time domain.
The remaining key element required to solve the thermal conditions is the
suitable description of the dynamic journal boundary conditions defined and
expressed at the grid boundaries (shaft outer surface).

8.3 Bearing Oil Film Temperature Distribution Model

To successfully create a model that in some way represents the dynamic

thermal conditions present with the bearing rotor system a suitably detailed
description of the steady state and transient thermal conditions present with
the bearing oil film are necessary. The work presented here has been
developed as part of the research to provide a calculation method and
procedure which predicts the thermal conditions present within the lubricant oil
film at the journal bearing boundaries. The equations and relationships have
been developed to attempt to reduce the highly involved computational
requirements typically needed for detailed modelling of the oil film energy
relationships present within the oil film. The intention is not to provide an
alternative procedure to give a complete description of the thermal and

248
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

dynamic properties present within the three oil film dimensions, but to present
appropriate energy and temperature conditions for application to shaft journal
heat transports from the lubricant oil film. In particular focus on representation
and modelling of non circular bearing geometries is presented as this is an
area which to date has not received any popular attention within the subject
area. At an industrial level this limits the applicability of the existing
procedures when attempts are made to apply procedures to practical high
speed machines.

8.3.1 The Oil Film Function

To gain a suitable understanding of the thermal conditions within the lubricant

film it is necessary to characterise the oil film thickness function around the
bearing for any given point in space and/or time, and from this
characterisation develop a means of understanding the heat generation
development around the bearing-journal surfaces.

8.3.1.1 Geometric Relationships

The first step in defining the oil film function h(θ ) is to develop a series of
equations and relationships that can be used to fully describe the oil film
spatial conditions between the bearing and journal surfaces. This geometrical
relationship is commonly referred to as the oil film function. This a
geometrically related function which is principally dependant on the bearing
surface topology and the operating position of the rotating shaft or journal
(when perfectly round bearing journal are assumed and local thermal
influences on the surface of the bearing are assumed negligible).

Many authors have derived and presented algorithms for the determination of
the oil film thickness for journal bearings. The mostly commonly encountered
is that of plain cylindrical bearing. It is the plain cylindrical bearing which will
be considered first here. The following derivations are based on the most
widely adopted analytical description which is compressively covered by
Cameron [43]. The development of these oil film functions will then be
adapted to describe the oil film functions for the non circular fixed and variable
bearing geometries used in the experimental research of this project.

Figure 195 shows the coordinate system adopted for the following equations
and derivations.

249
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

ω.t

R θ

Figure 195 Basic coordinate system used in the bearing geometry

and energy formulations

Plain Cylindrical Bearing

Under steady state conditions the normal section of the shaft and bearing
presented in Figure 196 is assumed. The geometrical conditions are dictated
by the level of eccentricity (e) present between the centre of bearing Ob and
the centre axis of the rotating shaft journal Os, i.e. e=Ob.Os. It is typical to
non dimensionalise this offset with reference to the bearing radial clearance
Cr to yield the factor known as the eccentricity ratio ε , (where 0 is gives a
fully centred shaft journal relative the bearing and eccentricity ratio of 1 is
where the shaft journal comes into contact with the bearing surface).

e
ε= where Cr = radial clearance
Cr

250
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Figure 196 Plain bearing geometric relationships

Examining the trigonomic arrangement from Figure 197 in more detail, the
following description in terms of triangles can be established.

Figure 197 Triangles from Figure 196

.

Using Figure 197:

Os B = Os A + AB = Rs + h Eq. 8.3.1

251
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

This in trigonometric terms is given by:

Os B = e.cos(θ ) + Rb .cos(α ) Eq. 8.3.2

Therefore

h = e.cos(θ ) + Rb .cos(α ) − Rs Eq. 8.3.3

Applying the sine rule and writing for α by:

e Rb
= Eq. 8.3.4
sin(α ) sin(θ )

where

e.sin(θ )
sin(α ) = Eq. 8.3.5
Rb
and applying the trigonomic identity cos 2 (α ) + sin 2 (α ) = 1 yields:

cos(α ) = 1 − sin 2 (α ) Eq. 8.3.6

substituting for sin(α )

2
 e 
cos(α ) = 1 −   .sin 2 (θ ) Eq. 8.3.7
 Rb 

At this point it should be noted that for practical bearings the ratio of the radial
clearance is generally of the order of 0.1-0.2% of Rb . This results in the angle
2
 e 
α being a very small value. Also   will be of the order 1 − 2 ×10−6 % of Rb
 Rb 
resulting in cos(α ) ≈ 1 . Substituting this into equation 8.3.3 yields :

h = e.cos(θ ) + Rb − Rs = e.cos(θ ) + Cr Eq. 8.3.8

252
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Re-writing with ε for eccentricity ratio gives :

h = Cr (1 + ε .cos(θ )) Eq. 8.3.9

Where equation 8.28 is the oil film function for plain cylindrical bearings and
for bearings with clearance ratios of <~0.5% yiels results with errors within
0.1% for h.

In real bearing applications the steady state eccentricity ratio and the angle at
which it acts relative to a datum need to established for any given load vector
and operating speed. These two features are calculated by means of solving
the Reynolds equation for a steady state operation. The principal derivations
and application of hydrodynamic lubrication and the Reynolds equation to
cylindrical bearings is summarised in chapter 3 and will not be re-addressed
here.

The solution of the Reynolds equation for various bearing types is extensively
covered in the available literature applying a wide range of solution
techniques. As the principal aspects of the work undertaken here are derived
from the geometric and operating conditions in the bearing, it is not necessary
to solve the pressure distribution within the bearing to derive the desired
relationships required to define the required thermal boundary conditions. The
information needed from the solution of the hydrodynamic problem is the
eccentricity ratio ε and attitude angle ϕ for a given operating condition. As
such it is not the intention of this work to address specifically the solution of
the Reynolds equation for a given bearing configuration, as the application of
one of the many existing techniques provides a satisfactory and effective
means of establish the necessary geometric data. For the purpose of the work
detailed in this section, it is assumed that the steady sate conditions for the
shaft journal position are known and used as an input to the analytical
evaluation.

Currently equation 8.3.8 is expressed with referenced to the point of maximum

film thickness where θ = 0 . In real bearing configurations the orientation of the
bearing is typically such that a physical datum is used with which to reference
the geometric and other performance data (see Figure 198). As such for a
plain cylindrical bearing equation 8.3.9 can be re-written :

253
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

h = Cr (1 + ε .cos(θ − β )) Eq. 8.3.10

where h can now be expressed with reference to any arbitrary point around
the bearing.

The film function defined above is consistent with the form presented by many
sources in the literature for the steady state condition in the bearing where the
shaft rotation axis remains at point Os. Generally when evaluating bearing
performance it is only the steady state condition which is of interest and
dynamic perturbations are considered to have only a minor and/or transient
effect on bearing oil film temperatures. However for the purposes of the
subject matter of the research, these transient conditions of direct interest and
importance with relation to the journal thermal conditions. For this reason it is
necessary to develop here the geometric conditions present during the
dynamic orbit with respect to time.

The shaft orbits can be functions of any order of shaft rotation and the
resulting shaft movements will, in almost all cases, be a combination of
multiple shaft orders. The most dominant and common order considered is
that of the synchronous orbit, whereby the shaft movement is directly related
to the shaft rotation frequency. During the synchronous orbit the shaft journal
centre Os will traverse a circular or elliptical path about the steady eccentricity
location Os.

To obtain the dynamic oil film function for any value of Ω.t it is necessary to
consider the geometric time based conditions presented by the shaft
synchronous orbit. Now consider the conditions shown if Figure 198 for a
cylindrical orbit.

When considering the dynamic shaft orbit it is now necessary to determine the
new relationships between the oil film thickness function and the dynamic
position of the shaft. Due to the additional time dependant geometric
component of the bearing–rotor the oil film function equation 8.3.10 cannot be
defined with respect to only one variable θ . Now also it is a function of time t .
The approach taken here is to determine an effective eccentricity ratio for any
given point of the shaft rotation cycle, this is denoted as e′ in Figure 198. This
dynamic eccentricity can then be used in direct substitution of e in equation
8.3.8 to obtain the oil film thickness for any given point in time of the rotating
shaft.

254
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

β
β'

Figure 198 Cylindrical whirl orbit geometric conditions in a plain bearing

However it has to be recognised that associated with the magnitude of the

new dynamic eccentricity there is an associated change in the instance at
which the direction of eccentricity occurs with respect to the shaft rotation
angle. Account must be taken of this angular variation γ when referencing it
to the datum point or point of interest, see Figure 198.

The corresponding angle θ ′ is established to define the dynamic location of

the minimum film thickness around the bearing with respect to the fixed
coordinate system at the bearing centre and the rotating shaft journal centre.

The introduction of a synchronous shaft orbit means that a single location for
the minimum film thickness with reference to the bearing shell is no longer
true. Account of how this changes with relation to a single revolution needs to
be considered when relating the film thickness to the rotating shaft journal. In
a steady state assumption, the point of minimum film thickness at any point of
rotation will follow the trajectory of the eccentricity ratio e. However with a
shaft orbit introduced, the location of the minimum film thickness on the

255
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

bearing surface will oscillate around the bearing circumference as the shaft
rotates and the orbit is traversed. As can be seen from Figure 198 for a
circular orbit when the points Ob, Os and O’s are in line, (i.e. at Ω.t = θ = θ + π ),
the relative angle that is generated between ObO’s and the minimum film
thickness for any point in the shaft rotation is denoted by γ . This angle is
important as this point has an influence on where the cavatitive region in the
bearing begins and how the thermal conditions generated in the oil film relate
to the journal rotational datum’s.

In the referenced literature [19,30] it has been taken that the point of minimum
film thickness is the hot spot for the rotating journal and this point tracks
around the shaft. The geometric relationships discussed above suggest that
this assumption may not always be appropriate as, due to the shift in
minimum film thickness location, other sections of the rotating journal may see
lower aggregate film thickness than the point at which Ω.t = θ . Also any time
dependant functions with regard to the heating of the lubricant or the heat
transfer to the journal will result in the location of the shaft hot spot being
moved to some other circumferential location on the journal, i.e. the time
derivative effects seen in the model of Keogh et al [9].

The equations relating the dynamic geometric conditions for the plain bearing
are now derived. From Figure 198 the dynamic effective eccentricity ratio can
be expressed as :

e '(t ) 2 = (e.cos(θ ) + a.cos(Ω.t )) 2 + (e.sin(θ ) + a.sin(Ω.t )) 2 Eq. 8.3.11

Expanding out equation 8.3.11 yields

e '(t ,θ ) 2 = e2 .cos 2 (θ ) + 2.a.cos(Ω.t ).e cos(θ ) + a 2 .cos 2 (Ω.t )......

e2 .sin 2 (θ ) + 2.a.sin(Ω.t ).e sin(θ ) + a 2 .sin 2 (Ω.t ) Eq. 8.3.12

Noting the trigonometric identify cos 2 (θ ) + sin 2 (θ ) = 1 then

e2 .cos 2 (θ ) + e 2 .sin 2 (θ ) = e2
Eq. 8.3.13
a 2 .cos 2 (Ω.t ) + a 2 .sin 2 (Ω.t ) = a 2

256
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Substituting into equation 8.3.2 with some work gives

e '(t ) 2 = e 2 + a 2 + 2.e.a.(cos(Ω.t ).cos(θ ) + sin(Ω.t ).sin(θ )) Eq. 8.3.14

Finally applying the trigonometric identity

cos(a ± b) = cos(a ).cos(b) m sin(a ).sin(b)

allows equation 8.3.3 to be reduced to

e '(t ) = e2 + a 2 + 2.e.a.cos(θ − Ω.t ) Eq. 8.3.15

Where equation 8.3.15 is the expression for determining the effective

eccentricity ratio for any point in a circular shaft orbit.

The angle of instance θ ′ of the dynamic eccentricity in the bearing is given by:

 e.cos(θ ) + a.cos(Ω.t ) 
θ '( t ) = tan −1   Eq. 8.3.16
 e.sin(θ ) + a.sin(Ω.t ) 
Equations 8.3.15 and 8.3.16 can be written in non dimensional from by
introducing Cr, where equation 8.3.17 and 8.3.18 are the non dimensional
eccentricity ratio and angle of instance respectively for any point around the
shaft orbit.

ε '(t ) = ε 2 + σ 2 + 2.ε .σ .cos(θ − Ω.t ) Eq. 8.3.17

 ε .cos(θ ) + σ .cos(Ω.t ) 
θ '( t ) = tan −1   Eq. 8.3.18
 ε .sin(θ ) + σ .sin(Ω.t ) 

a
where σ is non dimensional circular orbit radii
cr

From Figure 198 and the above equations it is now possible to describe the
angle γ (′t ) relating the minimum film thickness and the angle of rotation of the
shaft Ω.t .

257
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Where θ A = θ − θ ' , therefore

γ '(t ) = Ω.t − θ A − θ = Ω.t − 2.θ + θ ' Eq. 8.3.19

Equation 8.3.17, 8.3.18 and 8.3.19 can now be introduced to equation 8.2.10
to give the film thickness as a function of time t for any angle Ω.t around the
shaft orbit.

h( t ,θ ) = Cr (1 + ε '(t ) .cos(θ − β − γ (′t ) )) Eq. 8.3.20

The details necessary to establish the whirl orbit parameters need to be

determine from either experimental results or by means of a rotordyanmic
response analysis. The whirl orbit characteristics are heavily dependant on
the stiffness and damping conditions within a bearing oil film and also the
rotordynamic conditions in the rotor itself can also have a notable influence.
For the purposed of the work presented here the orbit data is assumed to be a
known set of data.

It should be noted at this point that in most bearings, a purely circular whirl
orbit rarely occurs and that some form of ellipse generally results. The same
principle of calculation can be applied to a shaft exhibiting an elliptical shaft
orbit as defined in chapter 2 Figure 14. However for the purposes of the oil
film perturbation analysis the orbit ellipse can be considered as separate
forward and backward circular whirl orbits. The effective journal centre can
then be determined for any point around the shaft whirl orbit at time t.

It was presented by Keogh et al [9] that it is possible to degenerate any elipse

into forward and backward whirl orbit components. The forward and backward
whirl components for a given elliptical shaft orbit can be established by :

a+b a−b
σ1 = , σ2 =
2 2

where : a and b are the ellipse major and minor axis

σ 1 and σ 2 are forward and backward non dimensional whirl radii
respectively
ψ is the ellipse rotation angle

258
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Figure 199 Forward and backward whirl orbits from vibration ellipse

Application to Non Cylindrical Offset Half Bearing

Figure 200 Fixed Geometry Offset Half Dynamic Orbit Relationships

259
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

In the same way the effective dynamic eccenticty e` for a time t was used for
the plain cylindrical bearing we can now define an effective eccentricity with
respect to either the top or pottom pads by introduing the pad offset sm. The
effective dynamic eccentricity for the top pad then becomes defined by
OOHTOj0. eOHT can then be used to for the necesary dynamic orbit calculations
as is outlined above. For the steaty running postion relative to each pad the
effective steady state eccentricity is given by OOHBOj and OOHTOj for bottom
and top pads respectivley.

The oil film and energy solutions are then applied to the top and bottom pad
for the appropriate θ range. In that when θ is in the top half pad then eOHT is
used and when in the bottom range then eOHB is applied. The conditions for
the energy solution are the same as that used for the two oil inlet bearing
case.

8.3.2. Bearing Temperature Distribution Model

To suitably represent the temperature conditions and distribution within the

bearing, it is necessary to considerer the coupled conditions of the
hydrodynamics and the thermodynamics. A number of models have been
presented in the literature to this end but generally what they all share in
common is that for a single load case the level of complexity associated with
both analytical or numerically building and then solving the model is high. In
all but a small few [9,10,17,18] the shaft journal is considered either
isothermal or of uniform temperature during the analysis. The intent of the
work covered here is to present a simplified model which can be applied and
computed with relative ease when applied to a range of bearing geometries.

The approach developed here is based on the development of a model which

considers the hydrodynamic conditions as interrelated but not analytically
coupled. This is attempted by considering a pre-determined solution to the
Reynolds equation to establish the geometric conditions governing the oil film
geometric characteristics. These geometric conditions are then used to
establish an energy development model which is used to calculate the
temperature distribution within the bearing. The hydrodynamics and
thermodynamics can be associated by an initial iterative routine which can be
used to determine an acceptably converged condition for the bearing
geometric running position. This is done by re-evaluating the Reynolds

260
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

equation for new ‘effective’ operating oil film temperatures. However, it must
be pointed out here that the focus of information presented in this section is
not associated with developing this process, it is principally intended as a
method to enable the necessary temperature boundary conditions to be
established for a dynamically vibrating shaft, based on initial steady state
running conditions and response characteristics when the Reynolds solution
problem has been converged.

To establish the desired temperature distribution within the bearing oil film a
form of energy balance model has to be established. As the purpose of the
work is to allow for a high number of calculations to be performed in a
cumulative procedure within the time domain, it is necessary to develop a
numerically effective analytical description of the temperature to provide the
necessary boundary conditions. To this end a number of assumptions have to
be introduced to allow acceptable solution simplifications, methods and
computation times to be achieved whilst maintaining, what the author
considers to be, an acceptable level of accuracy when modelling the thermally
dynamic physical system.

Firstly consideration of the oil film in the bearing is required and the energy
balance within the system. In practice we are dealing with a three dimensional
system with energy flows in each of these planes. To reduce the complexity of
the problem the model will be reduced to a two dimensional case. In this
simplification it assumed that the lubricant film temperature and heat flux to
the bearing shell and journal are constant in the z direction and are
represented as axially averaged values/functions. Now considering the
element dx shown in figure 201 and making the following principle
assumptions:

1. Temperature is constant across oil film (i.e. averaged cross film

temp)
2. Steady state conditions are assumed within the element and all
convective heat transfer has reached steady conditions.
3. The temperatures at the bounding surfaces of the bearing shell and
the journal are equal to the temperature of the oil film
4. Conduction terms within the oil film are not considered
5. ISO viscous conditions are assumed within the lubricant
6. Thermal effects are not considered to alter film function geometry
7. Bulk flow laminar conditions are assumed

261
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

It can be seen from the assumptions that (2) presents a slight contradiction as
the assumption of steady state conditions is incompatible with the transient
type analysis to be performed within the finite difference model. The problem
that this presents is explored and discussed later with justifications as to how
the assumption can be applied without significant impact of the accuracy of
the boundary conditions within the model presented.
.
Es

.
Eµ . h
E OIL ( x + dx ) E OIL ( x )
.
Eb
dx
Figure 201 Energy balance in an element of lubricant dx

Applying an energy balance to the model described above it is possible to

write that:

E&OIL( x + dx ) = E& µ + E& OIL( x ) − E& b − E& s Eq. 8.3.21

The energy generated in the element dx is produced as a result of the viscous

shear that takes place in the oil film between the bearing surface and the
rotating journal. The velocity profile in the oil film will vary between a linear
and parabolic type relationship due to Couette and Poiseuille flow terms. This
type of profile results in variable shear stress across the oil film, making
analytical description more difficult. To aid in simplification of the analytical
description a Petroff type of assumption is applied whereby a linear velocity
profile is assumed across the oil film, as has been commonly applied by
several of the literature sources [19,43].

The assumption relating to the cross film linear oil film profile for shear stress
and also for circumferential flow terms can be justified by the simplicity it
introduces and by the small effects it has for moderately loaded bearings. For
bearings where the eccentricity ratio and developed pressures are in the
region of <0.8 ε then the velocity profile will follow very closely to the
assumption. Only where high eccentricities and high pressure gradients are
present does the flow profile have a very significant effect on the net shear
forces.

262
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

du du
The shear stress in the oil film is given by τ = µ . where is the gradient
dy dy
of the velocity profile across the oil film. Applying the linear velocity profile
assumption this can now by written as :

U Ω.Rs
τ = µ. = Eq. 8.3.22
h h

Considering the element dx the energy in the element due to viscous shear is
given by :

dEµ = U .τ .du Eq. 8. 3.23

Substituting for τ and noting that du ≈ Rs .dφ into equation gives the following,
where h is dependent on φ so it is possible to substitute h for the oil film
function into equation 8.3.23 yielding :

Ω 2 .Rs 3 .µ .Lb Ω 2 .Rs 3 .µ .Lb

dEµ = .dφ = .dφ Eq. 8. 3.24
h(φ ) cr .(1 + ε .cos(φ ))

From equation 8.3.24 it is possible to describe the energy accumulation

around the bearing as the advection or circulating oil passes through the
bearing oil film gap. Solution of the above equation can, with some work, be
1
solved by direct integration with respect to φ . The term presents
(1 + ε .cos(φ ))
some analytical challenge but applying the following integration procedure
equation 8.3.30 results.

Separating out effective constants

Ω 2 .Rs 3 .µ .Lb dφ
Eµ = .∫ Eq. 8. 3.25
cr . (1 + cos(φ ))

φ 
Using substitutes for t = tan   where
2

263
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

(1 − t 2 )
1 + ε .cos(φ ) = 1 + ε . Eq. 8.3.26
(1 + t 2 )
With some work gives

dφ dt
∫ 1 + cos(φ ) = 2.∫ A − (1 + ε ).t
1
2
where A1 = 1 + ε Eq. 8.3.27

1
Using the standard integrals ∫ 1+ x 2
.dx = tan −1 ( x) + C

1 A  A.x 
and ∫ A+ x 2
.dx =
A
.tan −1   + C Eq. 8. 3.28
 A 

Equation 8.3.28 can be reduced to the following equation with subs for A1

dφ 2 1+ ε  1− ε 1+ ε φ 
∫ 1 + cos(φ ) = 1 + ε . 1− ε
. tan −1  . . tan    + C Eq. 8. 3.29
 1+ ε 1− ε  2 

Substituting into equation 8.3.25 gives

2.Ω 2 .Rs3 .µ .L  η φ 
Eµ = .tan −1 .  . tan    + C Eq. 8. 3.30
cr .(1 + ε ). η  η  2  


1− ε
where η =
1+ ε

Equation 8.3.30 allows direct calculation of the energy accumulation in the oil
around the bearing. The final step is to solve for the constant of integration
which is dependent on the oil inlet location relative to the minimum oil film
thickness point. In all information presented to date, and that provided in the
literature, the oil inlet is assumed to be 180 degrees around from the point of
minimum film thickness. Using this assumption the boundary conditions
become Eµ = 0 when φ = 0 and therefore C = 0 , resulting in equation 8.3.32
proving the complete solution. However in practical bearings this is rarely the
case. Just as in situations where the type of bearing has 2 or more inlets or
due to variation in the direction of applied loads the attitude angle changes. As

264
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

an example, in the bearings on the experimental test rig, the minimum oil film
thickness point was ~120 degrees from the inlet. To cover these more
probable conditions, the boundary conditions can be applied such that if we
say Eµ = 0 when φ = π − θ (See Figure 198) then the constant of integration
becomes:

2.Ω 2 .Rs3 .µ .L  η  β 
C=− . tan −1 .  .tan    Eq. 8.3.31
cr .(1 + ε ). η  η  2  


Substituting into equation 8.3.30 and simplifying gives

2.Ω 2 .Rs3 .µ .L  −1  η θ + β  −1
 η  β 
Eµ = .  tan  . tan    − tan  . tan     Eq. 8. 3.32
cr .(1 + ε ). η   η  2   2   
  η

Equation 8.3.32 describes the rate of heat generation accumulated around the
bearing and can be used to gain an understanding of the power loss and
amount of energy being distributed around the bearing. It is used later in the
finite difference model help establish the bearing inlet conditions. However, in
order to obtain the required temperature conditions of the oil film, the
continuity of flow conditions must be maintained whereby the volume of oil
available in the oil gap is considered. The above equation 8.3.32 does not
allow for continuity of flow around the bearing and as such film temperatures
cannot be established from this equation.

As the oil gap converges the volume of oil is reducing with the surplus
lubricant being expelled from the sides of the bearing, i.e. side leakage. This
side leakage takes some of the accumulated energy out of the bearing and oil
film, leaving only the remainder in the oil film to continue its route around the
bearing circumference. Therefore the application of equation 8.3.32 would
result in artificially high temperatures being progressively established around
the bearing. To account for the effects of continuity of flow it is necessary to
account for the cumulative rate of side leakage, as the lubricant moves around
the bearing, into equation 8.3.32. To do this in energy terms alone, as above,
becomes a more challenging with respect to the associated direct integration.
Therefore the same energy balance principle is applied to the model only this
time introducing temperature into the bounding equations. The energy
balance model can now be considered as below:

265
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

E& µ = E& OIL( x + dx ) − E&OIL( x ) + E& b + E& s Eq. 8. 3.33

where E&b and E& s are assumed to be 0.

The above base model construction is now similar to the one applied by
Balbahadur [19] but in this case the journal and bearing interfaces are
considered to be isothermal with respect to the oil film, and in effect the
energy dissipated from the film into the journal and bearing have been
neglected. This assumption significantly simplifies the resulting differential
equation. However the impact of this assumption must be validated for the
range of conditions that would be experienced in such bearings in application
to ensure this does not significantly impact the temperature result of the oil
film and journal.

Justification of Isothermal Assumption

Considering the viscous shear energy term and the energy loss term due to
heat transfer into the journal to be given by :
ω 2 .R 2j .µ .dx.dz
Eµ = ω.R j .τ .dx.dz = Eb = Es = H . (Tb, s − Tamb ) .dx.dz Eq. 8.3.34
h
The fraction of energy lost from the oil due to the viscous oil shear for a given
element can be determined.

Eb, s ω 2 .R 2j .µ
= Eq. 8.3.35
Eµ h.H . (Tb, s − Tamb )

Using typical values for a bearing running under fairly moderate conditions
where the journal velocity V j = ω.R j , these can be considered to be V j =15m/s,
operating viscosity µ = 0.05 N.s.m-1, film thickness h =75 µ m , ΔTj-b=50oC. A
heat transfer coefficient H =50 W/m2 is applied between the lubricant and
shaft/journal. This results in a fraction of heat lost from the film to the journal
of ~1.5%. This demonstrates that the impact on the energy state in the oil film
due to the heat loss to the journal can be neglected for typical operating
conditions without significant impact on the calculated lubricant temperature.
As the bearing speed increases, this error would also tend to reduce due to
the higher rate of energy generation and reduced ΔTj-b as the shaft
temperature increases.

266
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Removing the journal-bush energy transfer term from equation 8.3.33 results
in the simplified equation :

0 = Eµ − ( EOIL ( x + dx ) − EOIL ( x ) ) = Eµ − EOIL ( x + dx ) + E( x )

Eq. 8.3.36

Substituting the relevant terms :

ρ .h.Ω.Rs .cL .Lb ρ .h.Ω.Rs .cL .Lb

0 = Ω.Rs .τ .L.dx − .(T + dT ) + .T Eq. 8. 3.37
2 2

Where simplification gives:

ρ .h.Ω.Rs .cL .Lb

0 = Ω.Rs .τ .L.dx − .dT Eq. 8. 3.38
2

Re-writing and further simplifying the change in film temperature term dT can
be expressed by

2.τ .Lb
dT = .dx Eq. 8. 3.39
ρ .h.cL

Substituting for τ , h and dx = R.dφ

2.µ .Rs2 .Ω.Lb

dT = .dφ Eq. 8. 3.40
ρ .cL .cr2 .(1 + ε .cos(φ )) 2

Integrating to obtain T as a function of φ

2.µ .Rs2 .Ω.Lb dφ

T= 2
.∫ Eq. 8. 3.41
ρ .cL .cr (1 + ε .cos(φ ))2

Where it can be found that :

 φ   φ 
 tan    2.ε . tan  
dφ 2. η 2  ε  2
∫ (1 + ε .cos(φ ))2 = − (ε 2 − 1) . tan  η  . 1 + ε − 1 + 2
−1
+C
1 2 φ  
  (ε − 1).(1 + ε ).  .tan   + 1
  η 2 
Eq. 8.3.42

267
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

1+ ε
and η =
1− ε

Substituting equation 8.3.42 into equation 8.3.41

 φ  φ 
2  tan    2.ε .tan  
TL =
4.µ .R .Ω.Lb . η
s
.tan −1   2   .  ε − 1 + 2 +C
2 2
 
η  1 + ε   1 
ρ .cL .cr .(1 − ε ) 2 2 φ 
  (ε − 1).(1 + ε ).  .tan   + 1
  η 2 
Eq. 8.3.43

To obtain the constant of integration the boundary condition can be applied

that TL = 0 when φ = π − θ .

 β  β 
2  tan    2.ε .tan  
C=−
4.µ .R .Ω.Lb . η
s
.tan −1   2   .  ε − 1 − 2
2 2
ρ .cL .cr .(1 − ε )  
η  1 + ε   1 β  
  (ε 2 − 1).(1 + ε ).  .tan 2   + 1
  η 2 
Eq. 8.3.44

Substituting into equation 8.3.43 gives the general solution for oil film lubricant
temperature around the bearing.

  θ + β     β  
tan  tan  2    ε
4.µ.Rs2 .Ω.Lb . η  −1   2   
TL = TSUP + .  tan   − tan     . 
−1
−1 + ...........
ρ.cL .cr2 .(1 − ε 2 )   η   η  1 + ε 
    
     Eq. 8.3.45
  θ + β    β  
  tan    tan  2   
2.ε       
− 
2
....... 2 .
(ε −1).(1 + ε )   1 2 θ + β    1 2β  
  η .tan  2  + 1  η .tan  2  + 1 
       

Due to the fairly involved integration process full details are not provided here

for the solution of but details can be found in appendix C.

The application of the equation 8.3.45 results in the following oil film
temperature profile around the bearing for a single and dual oil inlet
arrangement respectively. The characteristic of the oil film temperature profile

268
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

will change significantly as the angle between the oil inlet and the minimum
film thickness is varied. This results in a more drastic increase temperature
occurring as the distance between the inlet and minimum film point is
decreased below θ < π . Figure 202 shows the temperature profile for the
π
same bearing geometry but now with two inlets and θ reduced from π to .
2
It should be noted here that the temperatures are non dimensional as an
arbitrary heating term has been used for convenience with an effective oil inlet
temperature of 0 to each pad. The data is presented only for comparative
purposes to demonstrate the temperature profiles generated by a particular
bearing with the eccentricity ratio and attitude angles applied.

Figure 202 Calculated lubricant temperature distributions for a single and

double inlet bearing

8.4 Cavatitive Region

As mentioned earlier in the section, once the oil film passes the minimum oil
film point the circulating oil breaks down and forms a cavative region within
the bearing due to continuity of flow. The oil in this region has been shown to
form a number of streamers or fingers whereby the remaining space is filled
by air drawn in through the sides of the bearing at atmospheric pressure (i.e.
no hydrodynamic pressure/lift). In some previous works this area has been
considered as an air oil mixture but it has been shown by testing using
transparent bearings full oil films are maintained but across a reduced portion
of the bearing width [45]. Figure 203 shows the generalised characteristics
observed in the cavative region.

269
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Generally the full film region of the bearing extends from the oil inlet to the
point of minimum film thickness hmin . Where several inlets are present within
the bearing there will be several areas where both full film and cavatitive
regions exist. In all of these cases, consideration must be paid to the effects
on the temperatures in the oil film and shaft journal due to this partial film or
cavatitive region. Additionally care must also be taken as it is possible,
although uncommon in practical bearings, under certain circumstances that
the cavitation region can be adjacent to an oil inlet. This condition is created
due to the eccentricity condition, resulting in a diverging section in the
direction of rotation at the location of an oil inlet.

Figure 203 Cavitation region in plain bearing with oil streamers

–figure taken from [44].

In practice the temperatures generated in the oil film will remain nominally the
same for streams that follow round into the partial film area of the bearing as
at a local level full films being maintained (although some heat cooling effects
will take place between the streamers and the entrained air). However the
amount of energy that is transmitted to the bearing in these locations will be
reduced to the wetted area and account for this needs to be provided in the
associated equations.

The method for introducing this effect here follows a similar approach as
presented by Keogh et al [9] where a fractional film coefficient is used in the
cavatitive region. The assumption within the oil film is that the temperatures

270
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

are axially averaged across the bearing width. The same principle can be
applied to the heat flux in that the cavatitive region the oil film temperature is
axially averaged by means of a fractional oil film coefficient. Therefore by
modifying equation 8.3.45 to include the fractional film coefficient such that:

TL = I h (θ ) .TL Eq. 8.4.1

Where I h ( Ω.t ) is the fractional film coefficient with boundary conditions:

I h ( Ω.t ) = 1 for 0 < θ < θ hmin

hmin( Ω.t )
I h ( Ω.t ) = for θ > θ hmin
h( Ω.t )

The effect of this on the axially averaged oil film temperature can be seen for
the single oil inlet example in figure Figure 204.

(a) No account for cavitation (b) with fractional film function

Figure 204 Axially averaged oil film temperature for a single oil inlet bearing

Care must be taken to define the converging and diverging sections of the
bearing when considering the perturbed orbit solutions as, during the shaft
orbit the location of minimum film thickness will oscillate about the peak
minimum film thickness location. This is governed by γ as defined earlier in
equation 8.3.19.

When considering tilting pad bearings there is no need for the fractional film
coefficient as due to the variable geometry and preload within the bearing the
pads always tend to result in a converging geometry and full oil film (subject to
sufficient oil supply being provide at the pad inlet). However some

271
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

consideration is need of the cooling effect at the oil galleries between the
operating pads.

The effect of the cavitation region is to lower the effective temperatures in the
journal where the net heat flux from the oil film is reduced to the reduction in
wetted area. This in turn has an effect on the magnitude of the temperature
distribution around the journal.

8.5 Oil Inlet boundary conditions

The final aspect relating to the shaft oil film boundary conditions is to
determine the oil inlet temperature to each of the pads. The lubricant supplied
to the pad is a mixture of the recirculation flow within the bearing and fresh
lubricant supplied via the oil inlets. The determination of the mixing of the oil at
the inlet is not a straight forward one and it is not the intention to conduct a
detailed evaluation here. However, an attempt is taken to derive an algorithm
suitable to provide approximations as to provide sufficient for the bearing oil
film model. Keogh et al [9] tackles the problem of the pad oil inlet temperature
and the mixing effects that take place, however the methods presented are
not applicable here due to the fact no cross film dimension is considered in
the thermodynamics.

Considering the fixed geometry bearings, a means of establishing the level of

mixing that takes place is required. In fixed geometry bearings the fresh oil
supplied to the bearing is principally related to the volumetric difference
created in the cavitation region due to the divergent section of the bearing. As
the oil inlet is a discontinuity in the temperature distribution model, an
algorithm that can determine the effective step temperature change is
required.

The oil flow into and around a bearing consists of several mechanisms, which
can be considered separately as components of velocity flow, pressure flow,
side leakage and recirculation flow.

8.5.1 Side leakage

The side leakage has been highlighted previously in this section and is
associated with the oil lost from the bearing during the divergent sections of
the bearing due to a reduction in volume of oil film gap. Account for this oil

272
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

flow loss is managed within the fundamental film temperature equations

presented earlier, Eq 8.3.45 and Eq. 8.4.1.

8.5.2 Re-circulation flow

The recirculation flow is the body of oil that remains following the end of the
positive pressure curve in the direction of rotation just after the minimum film
thickness point. The oil continues to flow around the bearing accumulating
heat until it reaches an oil inlet were it mixes with the fresh supply oil or a
further divergent section is encountered and further oil is lost by side leakage.
The former of these two conditions is most common in fixed geometry
bearings.

The volume flow rate of the re-circulating oil can be expressed by

U .h(θ ) .Lb
Qr = .I h (θ ) Eq. 8.5.1
2

8.5.3 Velocity flow

This is the oil that is ‘pumped’ into the bearing at the oil inlet by means of the
rotating shaft and oil velocity profile. The rate of flow is principally governed by
the surface velocity of the shaft and the size of the oil inlet. The velocity flow
forms the principle means of oil replenishment into the bearing. Taking the
Reynolds equation component for circumferential direction flow as given by
equation 8.5.2 and applying substitutions we get equation 8.5.3 for the
velocity flow component. The velocity flow is calculated from the start of the
pressure profile.

− h3 ∂p h
qx = . + (U1 + U 2 ) . Eq. 8.5.2
12.µ ∂x 2

ω.R j .h.Linlet  h 2 ∂P 
Qv = . 1 − 2
.  Eq. 8.5.3
2  6.Cr ∂θ 

Where Linlet = oil inlet groove length

273
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

8.5.4 Pressure Flow

The pressure flow is the flow that is associated with the oil expelled through
the sides of the bearing in the region of the oil inlet due to the oil supply
pressure. This flow effectively mixes with the re-circulating and velocity flow at
the inlet and contributes to the pad inlet temperature. It can have a significant
effect on the supply temperature where bearings are operating with low
eccentricity.

The amount of pressure flow can be estimated be equation 8.5.4 as presented

by Martin [63,64] by :

3
C3. p 1.25 − 0.25.Linlet  hθ31 − hθ32  2.R j θ 2 − θ1 θ =θ 2
h (θ )
Q p = d inlet . 1 
. + . .∑
8.µ 3
Cr    5 Cr3
 Lb 3  6.Lb . 1 − b
L

θ =θ1
6.Lb .  − 1  Linlet 
 Linlet 
Eq 8.5.4

Where θ1 and θ 2 are the angular extents of the oil inlet.

The above components of flow combine to provide the lubricating and cooling
oil within the bearing. Figure 205 presents a single bearing pad and shows
how the various components of oil enter and exit the bearing. For bearings
with several oil inlets a similar arrangement can be considered for each pad.

Qv + Q p − Qr
θh min

Qv Qr

Qp θ
Qv − Qr

Figure 205 Oil flow schematic for a single inlet bearing

274
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

The components of flow can be considered to combine as stated in Figure

205 where the velocity and pressure flow ‘top up’ the gap left by the cavitation
in the divergent section of the bearing.

The following expressions have been derived within this research to

determine the oil inlet temperatures to each bearing pad. If perfect mixing is
assumed at the oil inlet the pad inlet temperature can be described by
equations 8.5.5 and 8.5.6 for a bearing with two oil inlets.

Qr 2 .To 2 + Qr 2 .∆T2 + Qv1.TSUP + Qp1.TSUP

To1 = Eq. 8.5.5
Qr 2 + Qv1 + Q p1

Qr1.To1 + Qr1.∆T1 + Qv1.TSUP + Q p 2 .TSUP

To 2 = Eq. 8.5.6
Qr1 + Qv 2 + Q p 2

With some work, the above equations can be resolved to obtain To1 in terms
To 2 in equation 8.5.7 and hence solve for To 2 . Following which To1 can be
determined from equation 8.5.5.

 Qr1.Qr 2 
To 2 . 1 −  =
 (Qr 2 + Qv1 + Q p1 ).(Qr1 + Qv 2 + Qp 2 ) 
 
Eq. 8.5.7
Qr1.∆T1 + TSUP .(Qv 2 + Q p 2 ) Qr1.(Qr 2 .∆T2 + Qv1.TSUP + Q p1.TSUP )
+
(Qr1 + Qv 2 + Q p 2 ) (Qr1 + Qv 2 + Qp 2 ).(Qr 2 + Qv1 + Q p1 )

275
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Care needs to be taken when running in bearings with two oil inlets, as
dependant on the direction of eccentricity, it is possible for the oil to be
supplied into a divergent section with respect to the direction of sliding.
Therefore the oil supplied will be insufficient to fully replenish the oil film for a
certain distance around the bearing until a convergent section with oil gap
condtions matching the oil volume is again reached.

8.6 Lubricant Thermal Perturbations

Having established geometric and energy relationships for the bearing both
for the zero orbit and dynamic orbit cases, consideration must now be given to
the thermal developments resulting from the shaft synchronous orbit.

It would be a reasonable approach to use the expressions derived above to

calculate the energy/temperature conditions around the circumference of the
journal for a point around the dynamic whirl orbit and use these as the
boundary conditions for the journal finite difference model. The process would
then be repeated for a successive number of locations around the dynamic
orbit with an appropriate increment of ∆t until convergence is attained and the
steady temperature differential can be extracted. It was this process that was
initially developed and used to solve the journal time domain heating problem.

However this approach introduces several issues due to the physics of the
problem. As the fundamental assumption in this approach is the application of
steady state conditions for discrete dynamic positions around the whirl orbit,
i.e. ‘quasi’ dynamic conditions, this introduces erroneous energy distributions
within the oil film. Neglecting the time dependant function of the orbit
perturbed temperatures results in a condition where the heat accumulation
due to the advection of lubricant is over estimated, and more significantly,
results in incorrect dynamic temperature distributions with higher
temperatures being tended toward the outflow pad side.

The impact of this energy development discrepancy is a tendancy to over

predict the temperature differential present within the shaft journal, which in
the limited cases evaluated would appear to equate to approximately 30-40%.
Also, and possibly more significantly, is the effect on the circumferential
location of the hot spot of the temperature differential with respect to the point
of minimum film thickness, in effect the ‘hot spot’ phase angle. The resulting

276
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

‘hot spot’ phase angle with reference to the point of minimum film thickness on
the journal produced by the ‘steady state’ process described above, results in
a phase lead when the journal is subjected to a forward circular orbit. This
phase lead is in contradiction to the analytical predictions presented in [9] and
[18] where phase lag angles between 35o and 55o for plain cylindrical
bearings where predicted for the conditions assessed. Also the physical
observations presented by de Jongh et al [11] quoted a measured phase lag
angle of ~20o between the journal hot spot location and the point of minimum
film thickness for the rotor journal, which was running in tilting pad bearings.

The temperature distribution in Figure 206 shows the results of an analysis

performed following the above procedure for a bearing within eccentricity ratio
of 0.65, running at 10000 rpm with a 0.1Cd whirl orbit diameter. The resulting
converged ∆T and phase angle was 13.1oC and +61.8o respectively. The
temperature distribution was seen to consist of predominantly the fundamental
harmonic component, but some small non harmonic components are present,
seeFigure 207.

Direction of Rotation

Minimum Film
Thickness Point

Figure 206 Journal Circumferential Temperature Distribution for ‘Steady State’

Assumption Boundary Conditions

277
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

14 150
Delta T

Delta T 1st Harmonic Phase Angle

12 Delta T Phase 100

Shaft 1st Harmonic Delta T

10 50

(degrees C)
8 0

6 -50

4 -100

2 -150

0 -200
1 2 3 4 5 6
Harmonic Order

Figure 207Journal Circumferential Harmonic Temperature Distribution for

‘Steady State’ Assumption Boundary Conditions

As the phase angle is so influential for the rotordyanmic/thermal shaft bend

development and in particular the stability of the rotor, it is necessary for any
method of temperature differential prediction to introduce a means for
establishing this phase angle, of which the ‘hot spot’ lag contributes a
significant amount. The principle effect being overlooked with the procedure
described above is the way in which the time dependant energy function (by
which heat is generated in the lubricant film) is created and how the mass
dependant heat transfer then develops around the bearing, i.e. convective
heat transports. It is the convective heat transfer within the lubricant film which
results in conditions whereby the greatest thermal source at the journal
surface is found upstream of the direction of lubricant flow (i.e. opposite to the
direction of shaft rotation) for a forward whirl orbit. This observation of the
convective heat transfer being responsible for the hot spot phase lag was
briefly quoted by Keogh et al [9] when summarising the analytical results
presented in [9].

The evaluation of the transient thermal conditions and resulting convective

heat transports presents two notable challenges with respect to its evaluation.
Firstly, to model the convective heat transfer effects around the bearing an
understanding of the flow conditions around the bearing, and across the film is
generally required. It is this cross film aspect which presents the first
challenge. The analysis procedure being discussed here has been reduced to
a 1 dimensional heat flow problem within the oil film, i.e. no cross film ordinate
exists. As previously discussed, significant computational efforts are

278
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

introduced when looking in detail at the mass and heat flows within the oil film
itself which detracts from the purpose of this modified analysis procedure.

The second significant challenge is that direct simulation of the thermal

developments within an orbiting bearing introduces the need for long time
constants to allow for a converged ‘steady state’ periodic dynamic condition to
be reached. The solution of such a condition requires many rotations of the
shaft and as such many journal orbits which, dependant on the time step, can
result in a significant number of calculations. Therefore the use of more
complex modelling methods of the oil film become prohibitive and any method
developed here is required to result in short to moderate solution times.

The approach presented here to simulate the orbit perturbed oil film
temperatures is based around developing a method to provide account for the
convective heat transport whilst retaining the 1 dimensional thermal oil film
model (i.e. cross film averaged temperature conditions). The associated heat
fluxes between the lubricant film and shaft journal are then determined in the
time domain from the 1 dimensional film temperature distribution by
application of the 2 dimensional oil film/journal spatial conditions. This
approach allows for a less complex numerical and computational solution for
any given operating condition.

Convective Model

Solution of the time dependant lubrication film temperature due to the orbiting
shaft is approximated by considering the oil film to consist of two boundary
layers. One of the layers is associated with the static bearing shell and the
other is coupled to the rotating shaft surface. The mean cross film
temperature at any point in time is the volumetrically averaged temperature of
the two layers for any specific location around the bearing.

If an element as per Figure 208 with a specific velocity profile of the oil film is
considered then the proportion of lubricant mass affecting to the adjacent
element (in direction of sliding) will be a factor f of the element volume dx.h.
Where f relates to the proportion of mass flow transporting from one element
to another in the direction of sliding. The two bodies of lubricant are now
considered to mix together along with any additional heat variation introduced
to the element to develop a new bulk temperature for the element dx. As the

279
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

shaft rotates by the next time step ∆t the process repeats to develop a
modified temperature condition for dx. Hence the film can be considered to
consist of ‘static’ and dynamic lubricant elements both of which vary in
temperature over time as the shaft rotates and the orbit processes. The
process embodies/provides account of the spatial time derivative element of
∂T
the energy equation with respect to a fixed location around the bearing.
∂t

Advancing
Lubricant
Direction of Sliding
Journal

Oil Film

Bearing
Rj.dθ
Quasi-Stationary
Lubricant
(ref to element)

Figure 208 Convective boundary layer concept with simplification of flow

modelling

For a shaft exhibiting no whirl orbit then the steady state solution for this
condition can be established by equation 8.4.1 as the heat input to the oil film
for any angular position around the bearing remains independent of time.
However with the introduction of the orbit, the internal heat generation within
the lubricant film is periodic with time which results in a non stationary periodic
temperature state within the oil film. The convective heat transfer dictates the
time dependant distributions as the flow velocity of the lubricant varies across
the film from the journal velocity down to zero at the bearing shell boundary.
From this analogy it is clear to recognise that a significant number of orbit
perturbations will be required before a periodic steady state thermal condition
is established.

The convective model outlined above is simulated by numerical means where

the oil film is discreatised into a number of angular elements. For each of the
angular elements the temperature state of the lubricant TL will be a function of

280
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

the lubricant temperature entering that element by the dynamic boundary

layer, the temperature of the ‘static’ boundary layer for that angular position
plus the temperature perturbation at the that angular location due to the shaft
orbit with assumed Couette flow :

Tdx −1 + Tdx &

Tdx′ = + Tdx Eq. 8.6.1
2

However it is the case that for more heavily loaded bearings that the flow
consists of a nonlinear profile due to the Poiseuille flow component which is
dependant on the sliding velocity and dp/dθ (i.e. pressure gradient). Such
conditions will change the proportions of flow such that the level of advancing
fluid and mixing from element to element will change considerably as Figure
209 highlights.

Oil Film

Positive Pressure
Negative Pressure
Gradient Nonlinear Flow
Gradient
Velocity Profile

∂p
Figure 209 Couette and Poiseuille Flow Due to Pressure Gradient
∂θ

Providing for effects of Couette and Poiseuille flow the general form becomes:

TL (θ ,t ) = f L .TL ( (θ −∆θ ),(t −∆t )) + (1 − f L ) .TL (θ ,(t −∆t )) + T j (θ ,t ) Eq. 8.6.2

Where To is the temperature perturbation due the shaft orbit**

and f L is a factor to express the proportion of advancing to static boundary
flow (i.e. f L = 0.5 for Couette (linear) flow velocity profile)
** Note this can be a negative temperature due the fact it represents a distortion of periodic
variation of the steady oil film temperature.

 1 1  1
Where T j (θ ,t ) = FE& .  − . Eq. 8.6.3
 H j (θ ,t ) H o(θ )  H j (θ ,t ) .cL .ρ L
 

281
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

And FE& is a factor collecting together the bearing specific terms (viscosity,
speed and bearing dimensions) with j = 1, 2 where 1 and 2 indicate forward
and backward whirl orbits respectively.

The boundary conditions applied to the perturbed film temperature model

were that the temperatures of the lubricant at the oil inlet location were equal
to the pad lubricant inlet temperature plus the temperature variation due the
shaft orbit To . All convective flow was assumed to reset due the mixing
occurring at the oil inlet. This condition ensures that the temperature
accumulation does not continue to rise with each orbit iteration evaluated,
allowing a thermal periodic steady state condition to be determined.

The numerical procedure applied to solve the perturbed heat flow conditions is
by means of a time step iterative process. The oil film temperature condition
within the bearing is initiated at time t=0 to the zero obit condition and then a
stepwise time stepping solution is performed for the discretised angular grid.
Each orbit consists of n increments and the time stepping procedure is
continued for an integer number of obits until a converged film temperature
solution is obtained.

It is now important to take the dynamic thermal conditions in the lubricant film
and relate this to the temperature conditions within the journal rotor. The
thermal transport from the lubricant film to the journal can be expressed as a
heat flux from the lubricant to the journal. The heat flux is governed by the
gradient of temperature differential at the journal lubricant boundary and the
rate of heat transfer from the lubricant film into the journal. The rate of heat
transfer across the lubricant film is high when compared to that of the
conductive heat transfer of the journal itself, largely due to the significant
difference in respective conduction time constants. However some time
dependant influence results from the heat transports within the fluid film due to
the small values of ∆t when operating at high rotational speeds. The
instantaneous heat flux from the lubricant film into the rotating journal is given
by Gomiciaga et al [18] as :

kL ∂TL
qJ (θ J ,t ) = . (η = 0, θ J , t ) Eq. 8.6.4
h(θ J ,t ) ∂η

q(θ J ,t ) = Journal heat flux W/m2

282
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

k L = Lubricant thermal conductivity

η = Cross film non-dimensional ordinate, where 0 = journal lubricant
boundary

Equation 8.8.6.4 relates the heat transfer rate variation as a function of the
associated dynamic film thickness, where increased heat transfer occurs for
thinner films over a given time period t. The heat flux relationship described by
equation 8.6.4 can be incorporated into the 1 dimensional film temperature
∂TL
model by substituting with the difference in mean cross film lubricant
∂η
temperature and journal surface temperature for a given location a time t.

The asymmetric heating of the journal is then determined by the time

averaging of the instantaneous heat fluxes, which are based on the combined
steady and perturbed temperatures To and T j , around the bearing for each
time step ∆t for each synchronous orbit. These heat fluxes are then applied
as boundary conditions to the journal thermal finite difference model. This
results in the temperature differential and phase angle of the hot spot. The
converged heat flux conditions are representative of the temperature
distribution and hot spot phase angle that results in the shaft journal. Figure
210 shows the heat flux profile for a converged solution for a forward whirl
orbit, note the hot spot phase lag which now results with respect to the point
of minimum dynamic film thickness.

( )
TL(θ ,t ) = To(θ ) + T j (θ ,t ) .I h (θ ) Eq. 8.6.5

283
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Figure 210 Plot of a converged heat lubricant-journal heat flux

R j = 0.0375 m, Ω = 837 rads/s, cr=70 µ m , ε = 0.25 , σ = 0.2

Key assumptions used in convective modelling :

• Specified factor to govern flow/mixing around bearing circumference

• ISO thermal conditions at lubricant boundaries, i.e. no heat is lost
within lubricant film
• No conduction takes place across or circumferentially around the oil
film.
• Lubricant film considered as two boundary layers – ‘static’ layer
(bearing side) and dynamic layer (journal side)
• Average cross film temperature is the volumetric mean of the dynamic
and static lubricant boundary layer temperatures.
• Lubricant temperature conditions are set to be equal to the pad inlet
temperature at oil inlet grooves
• Axial conditions are averaged with no axial flow effects
• Laminar flow around bearing
• ISO viscous lubricant properties

284
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

8.7 Shaft Bearing Journal Temperature Distribution Model

The analytical and numerical procedures presented in sections 8.2, 8.3 and
8.4 are now applied in a combined procedure to provide a time domain
prediction of the thermal heating model of the bearing and journal. The shaft
temperature differential and thermal bow are predicted for a given set of
operating conditions.

The solution procedure was performed by use of a computer program

developed by the author in the MatLab [49] environment as the matrix
handling techniques available provide a means of efficient programming and
solution times.

8.7.1 Finite Difference Grid

The finite difference grid is constructed in the polar coordinate system and the
associated temperature matrices initiated from the desired initial condition
temperature of the journal. The number of circumferential and radial control
volumes is specified to give suitable resolution and accuracy for the condition
being analysed.

The grid resolution is kept to a minimum to result in as larger ∆t as possible

to give the shortest possible solution times whilst keeping errors and
temperature resolution within acceptable boundaries. Figure 211 shows a
typical finite difference structured polar grid. The maximum time step
increment required is governed by one of two factors. The first factor is that of
maintaining a stable solution when stepping through time domain increments
as defined by equation 8.7.1. The second consideration when specifying the
time step ∆t is matching the shaft rotation speed to give a suitable number of
points N around the shaft during one rotation. Where the ∆t is given by :

2.π
∆t = Eq. 8.7.1
ω .N

285
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Figure 211 Structured 2D Polar Finite Difference Grid

8.7.2 Initialisation and Boundary Conditions

The temperature matrix of the finite difference grid is initialised to a constant

temperature across the journal. The temperature used for the initial condition
is based on mean pad oil inlet temperature resulting from the steady state
temperature solution given by equation 8.4.1 for the zero orbit condition.

The boundary conditions are then applied to the grid for each time step. The
boundary conditions are applied by means of a heat flux to the exterior nodes
of the finite difference grid control volumes. The heat flux distribution is
determined by the methods outlined in section 8.3 for each time increment
and corresponding position around the shaft orbit. After each time step the
boundary conditions are shifted round by ∆θ in the direction opposite to shaft
rotation to simulate the effect of the fixed finite difference grid rotating relative
to the bearing.

8.7.3 Time Domain Solution

The model is then solved for a given time period of ∆t increments or until
convergence is obtained on the temperature within the shaft journal centre
node. The temperature convergence is tracked for each orbit cycle and can be

286
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

extracted against the time t. The complex shaft bend angle development is
also tracked against time and this observed with reference to the residual
mechanical unbalance location. The development of the thermal bend angle is
can then be used to provide an indication or stable and or unstable spiralling
vibration for a given set of running conditions in terms of operating speeds,
response characteristics and shaft orbits. No specific procedure for stability
assessment is presented here, but the procedures presented by Keogh et al
[9,10] and Childs et al [35] may be applied to the complex bend angle data
resulting from the analysis.

A sample 2D temperature plot of a converged case from the finite difference

model is shown in Figure 212. The convergence of the thermal solution is
considered to have occurred when the central node temperature variation with
time reduces to a predefined level. A typical plot of the central node
temperature with iteration cycle (time) is presented in Figure 213. In all
converged cases the cross sectional temperature distribution was seen to be
of a linear profile.

Figure 212 Finite Difference Grid Converged Journal Temperature Distribution

50mm diameter bearing, 5000 rpm, ε =0.3, σ 1 =0.2, forward orbit, 2 oil inlets

287
Chapter 8 – Analytical Evaluation and Development
of Journal Temperature Distribution

Figure 213 Typical Convergence Trend of Central Node Temperature

288
Chapter 9 – Analysis of Results and Discussion

Chapter 9

Analysis of Results and Discussion

9. Analysis of Results and Discussion

9.1 Introduction

This chapter provides a detailed review of the experimental test results

obtained within this research as presented in section 7. Experimental Results.
Review of theoretically produced results, by means of the developed
theoretical model, is also presented. Where possible both qualitative and
quantitative comparisons are made between the various tests conditions
assessed during the experimental work with respect the various key
influencing parameters. Specific attention is given to the correlations between
the rotordynamic behaviour of the rotor and the measured thermal conditions
in the bearing and journal.

Comparison is made between the results of the theoretical models and

experimental test results where possible. Attention is also given to the
discussion of the methods employed in the existing theoretical work.

The analysis of the experimental and theoretical work addresses the key
areas of rotordynamic performance, thermal measurements and theoretical
modelling.

9.2 Rotordynamic Performance

The rotor dynamic performance of the rotors and bearings is in general

agreement with the analysis performed at the design stage. The rotor critical
speed points were located at the expected frequencies for both the fixed and
variable geometry bearing arrangements. Figure 214 and Figure 215 present
the predicted and measured rotor response for the flexible rotor in fixed
geometry and variable geometry bearings respectively. The effects of the
residual mechanical unbalance have been removed from the measured
results to enable comparison to the calculated rotor response. The calculated
and measured results data presented relate to the discrete added mass only.

289
Chapter 9 – Analysis of Results and Discussion

12 180
Measured Y Vibration Measured Y Phase

Pk-Pk Shaft Vibration Filtered 1X (microns)

Calculated Y Vibration 135 Calculated Y Phase

Shaft Vibration Filtered 1X Phase (degrees)

10

90
8
45

6
0

4 -45

-90
2
-135
0
-180
0 2000 4000 6000 8000 10000 12000
0 2000 4000 6000 8000 10000 12000
Shaft Speed (rpm)
Shaft Speed (rpm)

Probe X
25 180
Measured X Vibration Measured Phase Angle
135
Pk-Pk Shaft Vibration Filtered 1X (microns)

Calculated X Vibration Calculated Phase Angle

Shaft Vibration Filtered 1X Phase (microns)

20
90

45
15

10 -45

-90
5
-135

0 -180
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000

Shaft Speed (rpm) Shaft Speed (rpm)

Probe Y
Figure 214 Measured and Calculated Rotordynamic Response for DR0428
(rotor residual mechanical unbalance removed)
25 180
Measured X Vibration
135
Pk-Pk Shaft Vibration Filtered 1X (microns)

Calculated X Vibration
Shaft Vibration Filtered 1X Phase (microns)

20
90

45
15

10 -45

-90
5
-135 Measured Phase Angle
Calculated Phase Angle
0 -180
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000

Shaft Speed (rpm) Shaft Speed (rpm)

Probe X
25 180
Measured Y Vibration Measured Y Phase
Pk-Pk Shaft Vibration Filtered 1X (microns)

Calculated Y Vibration 135 Calculated Y Phase

Shaft Vibration Filtered 1X Phase (degrees)

20
90

15 45

0
10
-45

5 -90

-135

0
-180
0 2000 4000 6000 8000 10000 12000
0 2000 4000 6000 8000 10000 12000
Shaft Speed (rpm)
Shaft Speed (rpm)

Probe Y
Figure 215 Measured and Calculated Rotordynamic Response for DR0673
(rotor residual mechanical unbalance removed)

290
Chapter 9 – Analysis of Results and Discussion

It can be seen that for the fixed geometry bearings generally good agreement
is obtained for the amplitude responses and also good correlation with the
phase characteristics (it should be noted that the phase angles will have a
‘bulk’ shift due to the offset location of the key phasor with reference to the
unbalance locations used on the test rotor and rotordynamic calculations).

For the variable geometry bearings, good agreement is obtained once again
but with this case the predicted amplitudes deviate significantly from the
actual levels when running at the higher speeds > ~7000 rpm. It is most likely
that the cause of this is that a discrepancy exists between the predicted and
actual bearing oil film parameters of stiffness and damping. In practice the
rotor generally exhibits slightly higher damping characteristics than is shown
in the predicted work. However, overall the rotordynamic predictions compare
well to the actual measured values for both amplitude and phase.

9.3 Shaft Journal Temperature Measurements

The experimental work essentially breaks down into two distinct areas of shaft
journal temperature measurement. The first area addresses the measurement
of influence component journal dynamic displacement and operating speeds
on the journal temperature distribution development. The aim was to isolate
the rotordynamic characteristics (other than unbalance response) and any
thermal bend development from the journal temperature differential
development. This allows for observations and an understanding of the
relationships coupling the dynamic bearing conditions and journal temperature
distributions.

The second area of experimentation introduced the flexible rotor which

introduces more complex rotordynamic conditions that are engineered to
provide thermally sensitive conditions with respect to shaft bend stability. The
same observation principles apply as outlined for the rigid rotor, for the flexible
rotor tests but with the additional of the dynamic amplification effects at the
bearing location with more complex rotor phase characteristics. The
introduction of a second bearing geometry type, variable geometry bearings,
also presents some further change in operating conditions.

Considering in detail the relationship between the journal temperature and

rotor response, and more specifically the response at the bearing journal the
results of the rigid rotor with the fixed geometry offset half bearings will be
evaluated.

291
Chapter 9 – Analysis of Results and Discussion

A principal point of interest when assessing the stability conditions with

respect to journal thermal bend development are the relationships governing
the rate of 1st harmonic temperature distribution and the response
characteristics at the bearing. When considering this at the rotordynamic
design stage the ability to establish steady state unbalance response is well
proven but the relationships and mode of development with basic operating
conditions such as speed, unbalance force, bearing response and orbit shape
is less well understood at a practical level. To date no controlled experimental
data has been published for journal temperature measurement for fixed
geometry bearings and more specifically offset half or two lobe type
configurations. In addition the theoretical modelling work presented to date in
the literature considers only plain cylindrical bearing geometry which is a
profile rarely applied in high speed rotating machinery, due to rotordynamic
stability (half speed whirl). The test data obtained from the rigid rotor test
embodies the more complex interactions the take place with non cylindrical
geometry present with typical high speed plain bearings.

1.4
300 g.mm

1.2 200 g.mm

1st Harmonic Temperature

1
Differencial (deg C)

100 g.mm
0.8

0.6
40 g.mm
0.4
4 g.mm

0.2
0 g.mm

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Shaft Rotation Speed (rpm)

Figure 216 Measured 1st Order Shaft Journal Temperature Differential for
Rigid Rotor

The experimental test data presented in section 7 and Figure 216 show the
fundamental (1X) harmonic component content of the temperature distribution
for the rigid test rotor NDE journal (DR03). The data is provided across a
running speed range of 0-8000 rpm for a range of unbalance magnitudes of
0g to 7.5g applied at a radius of 40mm, resulting 0-300 g.mm of unbalance.
Typical residual mechanical unbalance levels for an industrial rotor of this

292
Chapter 9 – Analysis of Results and Discussion

size/weight running at 8000 rpm would be 25-100g.mm. The unbalance mass

location is consistent for all conditions shown and was located at the NDE
flange.

It can be clearly seen that a steady increase in temperature differential is

developed as the shaft speed is increased. The temperature differential also
increases in magnitude as the amount of mechanical unbalance is increased.
It can be observed that the relationship of temperature differential with speed
shows a near linear condition for the smaller amounts if unbalance mass. As
the amount of unbalance mass is increased (100, 200 and 300 g.mm) the
relationship becomes exponential.

A log-log plot (Figure 217) shows that for the higher unbalance conditions an
exponential relationship fits well with the captured data. The expression
∆T = AT .N BT describes the journal temperature differential, ∆T , where AT is
function relating the rotor vibration response to the magnitude of journal
temperature differential, and BT is the relationship of rotor journal
velocity/lubricant shear stress and temperature differential characteristic. For
the conditions shown here exponents of 1.75, 1.71 and 1.58 were found to
correlate for 100g.mm, 200g.mm and 300g.mm unbalance conditions
respectively, see Figure 217.

10
Differencial (deg C - log10)
1st Harmonic Temperature

300 g.mm
200 g.mm
1

100 g.mm

0.1

0.01
1000 10000
Shaft Rotation Speed (rpm - log10)

Figure 217 Log-Log Plot of DR03 Tests for 100, 200

The maximum temperature differential observed in the rotor was 1.31 oC at

the maximum unbalance condition of 300 g.mm. The temperature differential
in the rotor remained steady when running at constant speed under steady
conditions. As seen in the rotor response vibration results, this is consistent
with the steady vibration amplitude and phase measurements at both the drive
and non drive end bearing.

293
Chapter 9 – Analysis of Results and Discussion

1.4
300 g.mm

1.2
200 g.mm

1st Harmonic Temperature 1

Differencial (deg C)
0.8
100 g.mm

0.6

0.4

0.2

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Shaft Rotation Speed (rpm)

Figure 218 Analytical Curves for Rigid Rotor Unbalance Loads for 100, 200
and 300 g.mm. ∆T = AT .N B T

The relationship between journal temperature differential and the response of

the rotor is shown in Figure 219 where the peak temperature differentials have
been extracted for each unbalance condition when running at 8000 rpm.

1.4 180
-180 Deg Pos 147
1.31 125
1.2 135
1.17 99.9 101
1st Harmonic Temperature

1.14
Differencial Phase Angle (deg)

85.05
1st Harmonic Temperature

90
Differencial (deg C)

1
45
0.8 0.787
0
0.6
-45
0.4 0.403 -86.58
0.311 -90
0.2 -127.9
0.139 -135
-180 Deg Pos
0 -180
0 40 80 120 160 200 240 280 320 0 40 80 120 160 200 240 280 320
Unbalance (g.mm) Unbalance (g.mm)

Figure 219 1st Harmonic journal temperature differential and phase angle with
respect to introduced mechanical unbalance

The data in Figure 219 presents the temperature differential and phase shift
with respect to the unbalance mass introduced. The phase data is taken with
reference to the datum (Green) thermistor sensor and it should be noted that
this not a direct measure of the phase angle with reference to the point of
minimum film thickness on the journal, this is determined and presented later
in this section. An additional data point is included which shows the measured
temperature differential when the unbalance mass was rotated by 180
degrees. The magnitude of ∆T for the 180o location is 1.14 oC compared to
1.31 oC for the datum location, and it is also seen that the ∆T phase rotates

294
Chapter 9 – Analysis of Results and Discussion

by some 188.3o, providing good correlation with the unbalance location and
the temperature differential development. The difference in ∆T and phase for
the two conditions is most likely as a result of the residual mechanical
unbalance in the rotor shaft. However, the thermal conditions within the
bearing are governed not by the specific forces involved but by the geometric
relationships in terms of dynamic whirl orbit. As it is not easy to present the
relationship of the rotor response by a single measurement, due to the
interactions of the orbit shape and bearing bore profile, resulting in more
complex film thickness functions, the modulus of the X and Y probes is taken
as a measure of whirl orbit size. It has been stated in the literature [33,34] that
a near linear relationship between whirl orbit size and journal temperature
differential exists, however this does not give detail of how, if at all, the orbit
shape or bearing profile affects this relationship. The peak temperature
differential measured for the rigid rotor are shown in Figure 220 against the
modulus of rotor dynamic response when running at 8000 rpm. The
temperature differential follows a near linear increase for the low levels of
response up to approximately 18% of clearance, beyond which it appears to
level off above 20% Cd.

1.4
1.31
1.2
0.908
1st Harmonic Temperature

1
Differencial (deg C)

0.8
0.787

0.6

0.4 0.403
0.311

0.2
0.139

0
0 0.1 0.2 0.3 0.4
Pk-Pk Resultant Rotor Response at NDE Brg (%Cd)

Figure 220 1st Harmonic journal temperature differential against orbit size

A key feature in the development of the thermal instability is the phase of the
‘hot’ spot relative to the minimum film thickness location and how this relates
to the location of mechanical unbalance. In order to establish the phase lag it
is necessary to understand the relationship of the shaft orbit with respect to
the bearing profile. This then needs to be corrected with respect to the

295
Chapter 9 – Analysis of Results and Discussion

rotordynamic phase lag when evaluating the thermal bow feedback and its
influence on rotor stability.
180

135
110.55

Phase Lag from h min (deg)

1st Harmonic Temperature
90
86.17

45 50.62 46.08
29.50
0

-45

-90

-135 -142.51 1st Harmonic Delta T Phase Lag

-180
0 0.1 0.2 0.3 0.4
NDE Flange Unbalance (g.mm)

Figure 221 Rigid Rotor 1st Harmonic Temperature Phase Lag with Respect to
Rotor Response at 8000 rpm

For the rigid rotor tests, once the point of minimum film thickness is
determined for the given operating condition, the ‘hot’ spot phase lag angle is
as shown in Figure 221 for six pk-pk response conditions. At the lower
response conditions the phase lag is seen to exceed 90 degrees. At the
higher response levels the phase lag can be seen to steadily decrease with
increased rotor response. The phase lag for the conditions between 0.1-0.35
Cd ranges from 86o to 29.5o.

Considering the ∆T phase lag condition for the lower responses, the orbit
shapes can be seen to differ. Figure 222 shows whirl orbits for the 0.07 and
0.34 Cd response conditions respectively. It can be seen that the whirl orbits
are highly elliptical, with the high response test resulting in a greater ellipse
ratio (ratio of ellipse major axis t minor axis). Significantly the ellipse angle
orientation is rotated round by some 60-70 degrees in the direction of rotation
for the lower response condition from that measured for the max response
condition. The orbit phase angle steadily rotated in a direction opposing shaft
rotation as the response orbit increased in magnitude, whilst accompanied by
an increase in ellipse ratio. The dynamic hmin location point with reference to
the bearing will have an influence on the convective heat transfer and how this
relates to the location of the hot spot on the journal. As previously discussed
the elliptical orbit can be considered to consist of both forward and backward
circular whirl orbits. Where the forward whirl component has been associated
with phase lag conditions and backward whirl with a phase lead. The

296
Chapter 9 – Analysis of Results and Discussion

combination of these conditions with the pre-load bore geometry, gives a

complex set of interactions, all of which will have an influence on the
temperature differential and phase angle.

1.5 1.5
Shaft Orbit Shaft Orbit
Point of hmin Point of hmin
Shaft Rot Angle Shaft Rot Angle

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5

Figure 222 Rigid rotor whirl orbits for 0.07Cd and 0.35Cd conditions

Little information has been presented in the literature with respect to ‘hot’ spot
phase angles, specifically with reference to changes in operating conditions.
Gomiciaga et al [18] however, in his numerical CFD study does indicate a
reduction in phase lag angle with an increase in forward whirl orbit diameter
up to 0.2 Cd for a plain cylindrical bearing with two oil inlets.

8
7.68
Shaft Whirl Orbit Ellipse Ratio

6
6.03

3 2.81
3.06
2.74
2 2.02

0
0 0.1 0.2 0.3 0.4
Pk-Pk Resultant Rotor Response at NDE Brg (%Cd)

Figure 223 Peak-Peak rotor response with respect to whirl orbit ellipse ratio

Also of interest is the relationship observed for the ∆T phase lag

development of the journal ‘hot’ spot with relation to shaft operating speed.
Evaluation of several of the unbalance conditions showed that the ∆T phase
lag was seen to steadily increase with an increase in shaft rotation speed. For

297
Chapter 9 – Analysis of Results and Discussion

the conditions where notable temperature differentials (40, 100, 200 and 300
g.mmm cases) and response were observed, the increase in phase lag angle
followed a linear relationship with speed. Figure 224 shows the phase lag
angle with respect to shaft rotation speed for the 300 g.mm test case with
imbalance positioned at the datum location. It can be seen that the phase lag
increases from ~20 degrees to ~40 degrees between 3000 and 8000 rpm.

90

67.5
Differencial Phase Angle (deg)

45
1st Harmonic Temperature

38
26 30
33
22.5 26
20

0
-11
-12
-22.5

-45

-67.5

-90
0 2000 4000 6000 8000 10000
Shaft Operating Speed (rpm)

Figure 224 1st Harmonic Temperature Differential Phase Angle for 300 g.mm
Unbalance of Rigid Rotor

Considering the flexible rotor arrangements, the rotors were designed to be

sensitive to unbalance excitation and in particular prone to development of
overhung section thermal bow. The rotor was operated in both fixed geometry
and tilting pad bearings, across a speed range up to 11000 rpm. The test
programme investigated a number of unbalance conditions and locations to
assess the impact on rotordyanmic performance and shaft journal thermal
development. Unbalance loads, up to 100 g.mm were introduced at the NDE
flange of the rotor, for both bearing configurations. The rotor was operated at
both shaft multi speed test runs (~10-15 minutes at each speed condition),
and also for extended periods at higher operating speeds between 7000 and
11000 rpm, where zones of instability were expected. The theoretical
calculations for instability of the rotor, in line with the published procedure,
predicted thermal instability at operating speeds between 8000 and 10500
rpm at unbalance levels of nominally 20 g.mm, 20% of the actual maximum
imbalance levels applied.

298
Chapter 9 – Analysis of Results and Discussion

Following extensive periods of running thermal bend instability was not

encounter for any condition. In areas identified as prone to instability the rotor
was run for up to 60 minutes and vibration levels were always observed to
maintain stable amplitude and phase parameters. Small changes in phase
were observed, however these were found to be associated with small
changes in oil inlet temperature. The temperature response of the rotor journal
was observed to be very rapid, with stable conditions being obtained within
30-60 seconds following a speed change. Therefore it was concluded that
further extended running periods were likely to result in unstable
developments. The levels of rotor imbalance was limited to 100 g.mm due to
reaching a predetermined limit of pk-pk vibration measured at the NDE
bearing.

The rotor was initially run with no additional overhung mass applied to the
NDE flange when operating in the fixed geometry bearings. As has been
evaluated for the rigid rotor, the journal temperature differential development
is evaluated for several whirl orbit response conditions. For the conditions
assessed the maximum temperature development was at a ∆T of 1.25oC for
a resultant response level of 0.17 Cd. The temperature development was
observed to follow a linear relationship with respect to response magnitude,
see Figure 226. This differs from the rigid rotor results where the ∆T was
seen to reduce development rate at higher response levels. When a least
squares fit curve is applied to the temperature differential against response it
can be seen to cross the axis very close to zero, suggesting a simple linear
response with orbit magnitude. It is however recognised that the response
level experienced on the flexible rotor did not exceed 0.2 Cd which is the point
beyond which the ∆T was seen to drop off on the rigid rotor. A similar
characteristic could be displayed if response magnitudes were increased
further on the flexible rotor configuration. It is also noted that the temperature
differential developed at 8000 rpm on the flexible rotor of ~1oC is higher than
the temperature differential measured at equivalent conditions on the rigid
rotor of 0.75oC for a response of ~0.15 Cd.

299
Chapter 9 – Analysis of Results and Discussion

1.5 1.5
Shaft Orbit
Point of hmin Shaft Orbit

Shaft Rot Angle Point of hmin

1 1 Shaft Rot Angle

0.5 0.5

0
0

-0.5
-0.5

-1
-1

-1.5
-1.5 -1.5 -1 -0.5 0 0.5 1 1.5
-1.5 -1 -0.5 0 0.5 1 1.5

1.5 kg OHM 5.0 kg OHM

Figure 225 Shaft Whirl Orbits for the Flexible Rotor in Fixed Geometry
Bearings with 1.5 kg and 5 kg OHM

A small difference can be seen between the rigid and flexible rotor conditions
in that the shaft orbit is seen to be rotated (~25-30o) around in the direction of
rotation for the flexible rotor case. This rotation tends the orbit major axis
toward the lower clearance quadrant of the bearing due the bearing bore
offset profile. It cannot be concluded that this variation results in the higher
temperature differential measured but it is generally considered that the orbit
orientation will influence the periodic thermal development of the oil film. The
influence of the more complex rotordynamics on the response conditions
within the bearing should also be considered as a potential source for
variations in the characteristics of the journal thermal development between
the rigid and flexible rotor results.

300
Chapter 9 – Analysis of Results and Discussion

1.6 90
84.8 84.1
80

1st Harmonic Temperature Phase ref

1.4
70

to Datum Sensor (degrees)

1st Harmonic Temperature
1.2 1.25
64.2

Differencial (deg C)
60
1
50
0.8
0.64 40
0.6
30
0.4
20
0.23
0.2 1st Harmonic Delta T 10
1st Harmonic Delta T Phase Ref
0 0
0.000 0.050 0.100 0.150 0.200
Pk-Pk Response Modulus (% Cd)

Figure 226 1st Harmonic Journal Temperature Differential for Flexible Rotor
with 1.5 kg OHM with Respect to Orbit Size at 9000 rpm
As was observed with the rigid rotor results the ∆T phase lag was observed
to reduce with increased unbalance response. With the phase lag varying
from ~40o down to ~30o for a dynamic response range of 0.025-0.16 Cd. The
phase lag observed for the flexible rotor was generally lower than that seen
for the rigid rotor at the same resultant bearing response magnitudes. It
should be noted that the orbit shapes and orientations are notably different for
the flexible rotor conditions than those of the rigid rotor and that more complex
phase changes take place across the rotor length for the flexible rotor.

45 120
40.71
40
99.5 35.88 100
Key Phasor to h min (degrees)

35
Phase Lag from h min (deg)
1st Harmonic Temperature

31.97
Rotordyanmic Phase Lag

30 80
74.1
70.8
25
60
20
15 40

10
20
1st Harmonic Delta T Phase Lag
5
Rotordynamic Phase Lag
0 0
0.000 0.050 0.100 0.150 0.200
Pk-Pk Response Modulus (% Cd)

Figure 227 1st Harmonic Journal Temperature Differential Phase Lag for
Flexible Rotor with 1.5 kg OHM with Respect to Orbit Size at 9000 rpm

301
Chapter 9 – Analysis of Results and Discussion

1.4 180 160

1st Harminic 150 149
Temperature 1.25
Diff erencial 135 140 133
1.2
1st Harmonic Phase

Differencial Phase Lag (degrees)

Differencial Phase Ref (degrees)
Ref

1st Harmonic Temperature

1st Harmonic Temperature

1st Harmonic Temperature

90 120 111
1 0.98
Differencial (deg C)

0.89 45 100
0.8
0 80
0.6 63
0.511 -45 60 62
54
0.4 36
-90 40 32
40

0.2 -135 20
0.049 0.055 0.036 0.045 0.908 0.105
0 -180 0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Shaft Rotation Speed (rpm) Shaft Rotation Speed (rpm)

st
Figure 228 1 Harmonic Journal Temperature Differential and Phase Angle
against Speed for a 100 g.mm Unbalance on Flexible Rotor with 1.5 kg OHM

The temperature differential developed in the journal was measured to follow

a flat response up to approximately 6000 rpm with temperature differentials of
less than 0.15oC being developed for an unbalance load of 100 g.mm at the
NDE flange. The temperature differential then increased rapidly to a ∆T of
1.25oC at a running speed of 9000 rpm. A significant phase shift in the ‘hot’
spot location around the journal was observed, with a total phase movement
taking place of ~175o. The temperature distribution and phase shift
corresponds closely to the rotordynamic response as the shaft approaches
the first and second critical speeds, with the first bending critical being located
at ~6800 rpm. The rotor dynamic phase lag was seen be above 90o for the
lower speed/response conditions with this dropping to between 60o and 30o as
the rotor response and ∆T increased at the higher speed conditions. In this
test case there was a general trend for the ∆T phase lag to reduce with
increased running speed, which in some part is a contradiction to the
observations of the rigid rotor tests. However, the response characteristics are
now more complex and whilst speed is increasing a notable increase in
dynamic amplification is also present at the bearing location which would,
from earlier observations, tend to reduce the ∆T phase lag. This is also
coupled with the relationship that the shaft whirl orbit form and orientation now
also changes significantly for each operating condition.

Similar observations were made for the flexible rotor with the NDE additional
rotor overhung mass added for both the fixed and variable geometry bearings.
For the fixed geometry bearings the temperature development follows a more
gradual increase with increased speed as opposed to the more ‘step’ type
change seen for the lower overhung mass condition, see Figure 229. This
correlates well with the changes observed in the rotordynamic response
where a more significant amount of low speed vibration is present but a more

302
Chapter 9 – Analysis of Results and Discussion

damped response is observed. The increase in overhung mass condition was

seen to be more sensitive to unbalance where similar peak vibration levels
were see for the increased mass configuration for approximately 50% of the
unbalance force.

1.2 180
1st Harminic 1.13
Temperature
Differencial 135
1 1st Harmonic Phase
0.95

Differencial Phase Ref (degrees)

Ref 0.97
90
1st Harmonic Temperature

1st Harmonic Temperature

Differencial (deg C)

0.8
45

0.6 0.59 0

-45
0.4

0.20
-90
0.31
0.2 0.17
0.06 0.13 -135
0.05 0.06 0.05 0.03
0.06
0 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Shaft Rotation Speed (rpm)

Figure 229 1st Harmonic Journal Temperature Differential against Speed for a
40 g.mm Unbalance on Flexible Rotor with 5.0 kg OHM

The ‘hot’ spot phase lag for the overhung condition is shown in Figure 229
and it can be seen that the lag angle seems more erratic than observed
previously and also suggests that a phase lead angle exists for high speed
conditions. Whilst it has not been conclusively stated that this case would not
occur, it is in generally contradiction to the observations made here and in
some of the literature.

When the rotordynamic conditions are studied in a little more detail for this
more sensitive rotor configuration it becomes apparent that the relationship of
the journal with respect to amplitude and phase becomes more complex, and
the relationship between the measurement point and the bearing centre must
be considered. Using the rotordynamic model it can be shown that a
significant variation in phase condition is predicted between the bearing centre
line location and the point at which the shaft displacement measurements are
made. Even though this axial distance is separated by 18-20 mm the phase
conditions change notably both in magnitude and characteristic. This variation
is largely due to the fact that at the rotor is operating in a higher order shaft
deflection mode where the NDE bearing location is close to the nodal point
which results in rapid rates of change with respect to phase and amplitude. It
is these conditions which identify the rotor as one that should be sensitive to

303
Chapter 9 – Analysis of Results and Discussion

thermal heating instability. Figure 230 shows the phase angle predictions in
the horizontal and vertical planes for the bearing centre location and the
proximity probe measurement location. It can be seen that the shift in phase
when approaching the first shaft bending critical at ~6000 rpm is significantly
greater than that seen at the bearing centre.

Rotordynam ic Response Plot

Flexible Rotor Model - Offset Half Bearings - Overhung Mass

Sta. No. 8/9: NDE brg
400
H Phase Brg Ctr
350 V Phase Brg Ctr
H Phase Measurement Point
300 V Phase Measurement Point
Response, mm p-p

250

200

150

100

50 Probe Clocking
45 degrees
0
0 3000 6000 9000 12000 15000
Rotor Speed, rpm

Figure 230 Calculated Phase Angles for NDE Bearing Centre and
Measurement Point Locations Flexible Rotor Fixed Geometry Bearings 5kg
OHM

This variance in phase angle between the two locations means that when
determining the phase lag between the minimum film thickness hmin position
on the bearing journal and the ‘hot’ spot, a distorted measurement will be
obtained. To compensate in some way for this variation between the
measurement location and the condition within the bearing, application of the
rotordynamic predictions is used. Good correlation has been observed
between the shaft vibration measurement and the rotordynamic response
prediction, in particular the correlation of the phase angles. The phase angle
prediction and measurement of the proximity probe observed location for the
NDE bearing was compared to the measured rotordynamic response at this
axial point. Figure 231 shows the respective phase angles for the measured
and predicted cases. The predicted rotordynamic phase angles were
corrected for the actual key phasor location used on the physical test rig with
respect the introduced mechanical unbalance.

304
Chapter 9 – Analysis of Results and Discussion

The correlation between the predicted and measured rotordynamic phase

angles is seen to be very good at the speed conditions above ~5000 rpm with
discrepancy in results being of the order of 5-10o.

180 180
Measured Phase Angle Measured Y Phase
135 Calculated Phase Angle 135 Calculated Y Phase

Shaft Vibration Filtered 1X Phase (degrees)

Shaft Vibration Filtered 1X Phase (microns)

90 90

45 45

0 0

-45 -45

-90 -90

-135 -135

-180 -180
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000
Shaft Speed (rpm) Shaft Speed (rpm)

Figure 231 Measured and Calculated Phase Angles for NDE Bearing Flexible
Rotor Fixed Geometry Bearings 5.0 kg OHM – DR0428

Now the ∆T phase lag angles are determined using the rotordynamic
predictions for phase at the bearing centre location. Figure 232 shows the
result of this change, where it can be observed that a more consistent phase
lag angle is observed for the higher running speed conditions. This
characteristic is more in line with expectation with regards to magnitude.
However, at the lower speed conditions <5000 rpm significant errors are likely
to be present due to the poorer correlation between the predicted and
measured vibration phase in this operating range.

180

135
Differencial Phase Lag (degrees)

90
1st Harmonic Temperature

46 42
45 38
28 28 35
22
23
0 -5
-35
-45
-89
-90 -97
-104
-117
-135

-180
0 2000 4000 6000 8000 10000 12000
Shaft Rotation Speed (rpm )

Figure 232 1st Harmonic Journal Temperature Differential Phase Angle for
Flexible Rotor in Fixed Geometry Bearings – Corrected by Calculated
Rotordynamic Phase Shift

305
Chapter 9 – Analysis of Results and Discussion

The application of the predicted rotordynamic phase conditions to the

determination of the ∆T phase lag can provide a more representative
measure of the phase lag angle for the shaft ‘hot’ spot where good correlation
exists between the rotordynamic model and physical measurements. The
method is applied for determination of phase angles for the flexible rotor
conditions where complex rotordynamic responses are observed. Whilst the
procedure should provide a more accurate means of establishing the journal
‘hot’ spot phase lag it will embody some additional error through the numerical
calculation and the precise conditions in the bearing itself remaining unknown.
It is, however, considered a suitable means of establishing the characteristics
and general magnitudes for the data being presented here.

The thermal differential development for the tilting pad bearing was seen to
follow a more gradual increase with speed which was consistent with the
response characteristics for this arrangement. The ∆T was seen to peak at a
running speed of ~8500 rpm which corresponded to a point just below the
second shaft bending critical. Beyond this point, the ∆T was seen to start to
reduce with increasing speed. The ∆T phase angle remained constant up to
the point of the rotor first bending critical, beyond which it can be seen to shift
to a near linear relationship with speed by around -55o, see Figure 233. These
characteristics were consistent for all tests for this bearing configuration where
imbalance was introduced at the NDE flange.

1.6 180
1st Harminic
Temperature
1.38 1.36
1.4 Differencial 135
1st Harmonic Phase 1.32
Differencial Phase Ref (degrees)

Ref
1.2 1.26 90
1st Harmonic Temperature

1st Harmonic Temperature

1.20
Differencial (deg C)

1.00
1 0.99 45

0.8 0
0.75

0.6 0.50
-45

0.34
0.4 -90
0.20
0.2 0.14 -135
0.03
0 -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
Shaft Rotation Speed (rpm)

Figure 233 1st Harmonic Journal Temperature Differential for Flexible

Rotor in Variable Geometry Bearings with 5.0 kg OHM

The rotordynamically corrected ∆T phase lag angle was seen to follow a fairly
flat response, steadily decreasing in magnitude from 72o to 31o degrees

306
Chapter 9 – Analysis of Results and Discussion

between 2000 to 9000 rpm. At 10000 rpm this had reduced to just 11o (Figure
234). A small oscillation in phase lag is observed as the rotor traverses the
first shaft bending critical speed.

180

Differencial Phase Angle (degrees)

135

90 72 67
66
1st Harmonic Temperature

63 60 62
49
45 56 58
31
22 11
0

-45
-71
-90

-135

-180
0 2000 4000 6000 8000 10000
Shaft Rotation Speed (rpm )

Figure 234 1st Harmonic Journal Temperature Differential Phase Angle for
Flexible Rotor in Variable Geometry Bearings with 5.0 kg OHM

A comparison of the various temperature differentials for several of the test

conditions for the flexible rotor is presented in Figure 236. It can be seen that
for all cases that a linear trend would tend towards the 0 loci on the plot.
Clearly too few data points are available to provide a numerically conclusive
observation but the characteristics are consistent for all conditions of both
fixed and variable geometry design. It can be seen that in general for the
same nominal response magnitude, the tilting pad bearing results in higher
temperature differentials. For the 5kg overhung mass condition the tilting pad
bearing resulted in approximately twice the ∆T than the fixed geometry
bearings for the same response level. The orbit characteristic of the tilting pad
bearing remained more consistent across the running conditions of both
speed and imbalance load with respect to orientation and ellipse ratio (i.e.
relevant content of forward and backward whirl). The whirl orbit was much
more circular than observed with the fixed geometry offset half bearing, with
typical ellipse ratios of approximately 1.5. Figure 235 shows a typical shaft
whirl orbit for an imbalance load applied to the NDE flange on the flexible rotor
running in variable geometry bearings.

307
Chapter 9 – Analysis of Results and Discussion

1.5
Shaft Orbit
Point of hmin
Shaft Rot Angle
1

0.5

-0.5

-1

-1.5
-1.5 -1 -0.5 0 0.5 1 1.5

Figure 235 Shaft Whirl Orbit for the Flexible Rotor in Variable Geometry
Bearings with 5 kg OHM

2.5
Fixed Geometry Brg 1.5kg OHM @ 9000 rpm
Variable Geometry Brg 5kg OHM @ 9000 rpm

2 Variable Geometry Brg 5kg OHM @ 6000 rpm

1st Harmonic Temperature

Variable Geometry Brg 5kg OHM @8000 rpm

Differencial (deg C)

Fixed Geometry Brg 5kg OHM @ 9000 rpm

1.5

0.5

0
0 0.05 0.1 0.15 0.2
Pk-Pk Response Modulus (% Cd)

Figure 236 1st Harmonic Journal Differential Temperature with Respect to

Response Modulus for Different Arrangements and Conditions

The introduction of the overhang mass on the flexible rotor when running in
the fixed geometry bearings resulted in a reduction in effective ∆T for a given
magnitude of response orbit. Exploring this further it is noted that whilst the
whirl orbits are of a similar magnitude the orientation of the major ellipse axis
was rotated by some 25-30 degrees in the direction of rotation for the lower
overhung mass condition (1.5 kg). An orbit tending to this orientation would
generally result in slightly lower film thickness conditions and also impact on
the convective oil film heat transports. It can not be stated that this is the
cause for the higher temperature differential of the lower OHM tests, but it is a
systemic observation. It should also be noted that, as highlighted previously,
due to rotordynamic influences on phase at the bearing centre this will also

308
Chapter 9 – Analysis of Results and Discussion

change the conditions seen in the bearing from the orbits measured. This can
influence the orbit geometry and hence temperature differential development.
Figure 237 shows the orbits for the 1.5kg and 5kg overhung mass conditions
for the fixed geometry bearings running at 9000 rpm with 50 g.mm of
imbalance.

1.5 1.5
Shaft Orbit Shaft Orbit
Point of hmin Point of hmin
Shaft Rot Angle Shaft Rot Angle

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5

Figure 237 Flexible Rotor Shaft Whirl Orbits for 50 g.mm Unbalance at the
NDE Flange Running at 9000 rpm in Fixed Geometry Bearings

Imbalance mass was also added to the shaft centre location (no imbalance
was included at the NDE flange). Several imbalance magnitudes were
introduced and a near linear relationship to ∆T was observed for response
conditions up to 0.2 Cd. The significant difference was the notable reduction in
∆T when running above the first bending shaft critical. In Figure 238 the peak
∆T can be seen to occur at ~6500 rpm (approximately 300 rpm above the
first shaft bending critical speed) with a magnitude of ~0.7oC, where it then
rapidly drops down to ~0.25oC when running up to 9000 rpm. The ∆T phase
lag can be seen to remain at a steady level when running in the higher ∆T
region at ~40o up to a speed of ~8000 rpm where it then shifts to a phase lead
of ~90o.

309
Chapter 9 – Analysis of Results and Discussion

0.8 180

0.7 0.708 135

Differencial Phase Ref (degrees)

1st Harmonic Temperature

1st Harmonic Temperature

0.6 90

Differencial (deg C)
0.5 45
0.476

0.4 0
0.363

0.3 -45
0.27 0.281
0.2 -90
0.09 1st Harminic
0.1 Temperature -135
0.032 0.045 Differencial
0.027 0.026
1st Harmonic Phase
0 Ref -180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Shaft Rotation Speed (rpm)

Figure 238 1st Harmonic Journal Temperature Differential Running in Fixed

Geometry Bearings with 150 g.mm of Imbalance Located at the Rotor Mid-
span

180
179

135
Differencial Phase Lag (degrees)
1st Harmonic Temperature

90
44 43 39
45
37 36
11
0

-45
-80
-90 -93

-135 -133

-180
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Shaft Rotation Speed (rpm)

Figure 239 1st Harmonic Journal Temperature Differential Phase Running in

Fixed Geometry Bearings with 150 g.mm of Imbalance Located at the
Rotor Mid-span

The focus within this work has principally been on the 1st order harmonic of
temperature disruption as it is this component of thermal differential that will
result in a mechanical bow of the rotor. All other higher order components
maintain a state of equilibrium with respect to thermally induced stress in the
journal (i.e. no resultant displacements). However it was observed that for
many of the tests, particularly for the conditions when running at low levels of
response some form of higher order components were present in the
circumferential temperature distribution. For the purposes of discussion here
these shall be referred to as ‘non harmonic’ components.

310
Chapter 9 – Analysis of Results and Discussion

Figure 240 shows the temperature distributions for the flexible rotor running at
two speeds for the fixed and variable geometry bearing arrangements with an
NDE flange imbalance force of 40 g.mm. Figure 240 (a) and (b) are for the
fixed geometry bearing and as (c) and (d) for the variable geometry. The
green line represents the residual temperature distribution once the 1st
harmonic component is removed. It can be seen that the non-harmonic
component contributes more to the overall measurement as the rotor
response/speed is reduced. Clearly some of this non harmonic content will
result from measurement error, however it can be seen that the magnitude of
this component is notably in excess of the stated measurement accuracy,
(which has been demonstrated by several means). Even if the error range
were tripled the magnitude of the non harmonic component would remain
significant.

Further insight into the presence of this non harmonic component can be
derived from the measurements taken using the rigid rotor. It can be seen
from Figure 241 (a) and (b) that when the unbalance mass is shifted by 180
degrees the residual component has now inverted. If this residual were as
result of asymmetric measurement error then it might not be expected that the
measured residual would shift phase in this manner.

Shaft Sensor Temperature Shaft Sensor Temperature

57.2
66.3
Measured
Measured
66.2 57.15 Calc 1st Harm
Calc 1st Harm
Residual
66.1 Residual
Full 1st Calc
Temperature (deg C)

Full 1st Calc 57.1

Temperature (deg C)

66

57.05
65.9

65.8 57

65.7
56.95
65.6

56.9
65.5
0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position
Sensor Position

(a) DR0428 – 9000 rpm (b) DR0428 – 5000 rpm

Shaft Sensor Temperature Shaft Sensor Temperature
58 50
Measured
57.8
49.9 Calc 1st Harm
57.6 Residual

57.4 49.8 Full 1st Calc

Temperature (deg C)

Temperature (deg C)

57.2
49.7
57

56.8 49.6

56.6 Measured 49.5

56.4 Calc 1st Harm
Residual 49.4
56.2
Full 1st Calc
56 49.3
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Sensor Position Sensor Position

(c) DR0673 – 9000 rpm (d) DR0673 – 5000 rpm

Figure 240 Measured Journal Temperature Distributions for the

Flexible Rotor

311
Chapter 9 – Analysis of Results and Discussion

It is not possible here to attempt to determine and harmonic content of this

residual as too few measurement points are used to extract anything greater
than the 1st harmonic. The residual component can only be used as a guide to
the amount of measured non fundamental temperature distribution present
within the journal.
Shaft Sensor Temperature Shaft Sensor Temperature

52 53

51.95
52.95
51.9

51.85 52.9

Temperature (deg C)
Temperature (deg C)

51.8
52.85
51.75

51.7 52.8

51.65
52.75
Measured
51.6
Actual
Calc 1st Harm
51.55 Calc 1st Harm 52.7
Residual
Residual Full 1st Calc
51.5
Full 1st Calc 52.65
51.45 0 1 2 3 4 5 6
0 1 2 3 4 5 6
Sensor Position
Sensor Position

(a) DR0321 – 4000 rpm (b) DR039 – 4000 rpm

Figure 241 Measured Journal Temperature Distributions for the

Rigid Rotor with 180o Phase Shift in Mechanical Unbalance

From the discussion presented here it is considered that the measurement of

the presence of some non harmonic content is credible. It is not unexpected
that this maybe the case as the shaft orbits have been shown to be of an
elliptical nature for both the fixed and variable geometry bearings, which
means some backward circular whirl is present. A backward circular whirl has
been shown in the theory presented in section 8 to result in some significant
non harmonic content. This coupled with the more complex bearing profile
geometry results in complex periodic time dependent temperature
distributions around the bearing which could quite conceivably result on non
sinusoidal temperature distributions around the rotating shaft journal.

Comparison of Results to Thermal Instability Theory (Balabahdur [19])

The theoretical analysis method presented by Balbahadur [19] predicted

unstable regions for the fixed geometry and variable geometry bearings of
8800-10800 rpm and 7000-9000 rpm respectively for 50 g.mm and 25 g.mm
unbalance loads respectively. As previously stated the rotor remained stable
across all operating conditions, and both rotors where run with unbalance
loads up to 100 g.mm.

312
Chapter 9 – Analysis of Results and Discussion

The predicted temperature differentials for the fixed and variable geometry
bearings were of the order of 2.8oC and 10oC respectively for the 50 g.mm
and 25 g.mm loads. As the method for establishing temperature differential is
not principally driven by the unbalance force but the specified response
magnitudes then it is important that these correlate to the measured
conditions. As stated at the being of this chapter good correlation was
obtained for both magnitude and phase conditions between the measured and
predicted responses for a given level of applied unbalance. This provides
some confidence in the data that was used in the analytical model of
Balbahadur [19] is valid and the temperature differential predications can be
safely compared to the measured levels.

The predicted and measured temperature profiles for two unbalance

conditions, one with fixed geometry bearings and the other with variable
geometry bearings is presented in Figure 242 and Figure 243 respectively.

3 1.2
Calculated Temperature
Differencial (Balbahadur [19]
Method) 50 g.mm
2.5 1st Harmonic Measured 1

Differencial as Measured (deg C)

Balbahdur [19] Methods (deg C)
Differencial in Accordance with

Temperature Differencial 50 g.mm

1st Harmonic Temperature

Calculated Temperature

2 0.8

1.5 0.6

1 0.4

0.5 0.2

0 0
0 1500 3000 4500 6000 7500 9000 10500
Shaft Rotation Speed (rpm)

Figure 242 Measured and Calculated [19] Journal Temperature Differential

Flexible Rotor in Fixed Geometry Bearings with 40 g.mm Unbalance
and 5 kg OHM - DR0428

313
Chapter 9 – Analysis of Results and Discussion

12 1.6
Calculated Temperature
Differencial (Balbahadur [19]
Method) 25 g.mm 1.4
10 1st Harmonic Measured

Differencial as Measured (deg C)

Balbahdur [19] Methods (deg C)
Differencial in Accordance with
Temperature Differencial 40 g.mm
1.2

1st Harmonic Temperature

Calculated Temperature
8
1

6 0.8

0.6
4
0.4
2
0.2

0 0
0 1500 3000 4500 6000 7500 9000 10500
Shaft Rotation Speed (rpm)

Figure 243 Measured and Calculated [19] Journal Temperature Differential

Flexible Rotor in Variable Geometry Bearings with 40 g.mm Unbalance
and 5 kg OHM - DR0673

It can be seen that there are significant differences in the magnitude of

predicted temperature differentials. For the fixed geometry bearing the
predicted temperature differential is 2.8oC which is 2.4 times that measured
on the shaft journal. With the variable geometry bearing the peak predicted
temperature differential is 10.1oC compared to a measured differential of
1.4oC, which is approximately a factor of 7 higher. It should also be noted that
in the variable geometry case the measured temperature was for an
unbalance force of 40 g.mm as opposed to a 25 g.mm unbalance in the
model, which would result in a further in variation.

While the magnitudes of ∆T show significant discrepancy there is some

correlation between the characteristics of the temperature development of the
predicated and measured data. For the fixed geometry bearing it can be seen
in Figure 242 that the trend is very well matched to that of the measured ∆T
data. With the variable geometry data in Figure 243 there is some reasonable
correlation between predicted and measured data, but the most notable
correlation is in the drop off in ∆T at higher speeds above ~8000-9000 rpm.
However it would be reasonable to assume good correlation of the
temperature development trend would be attained, as the method is heavily
dependent on the supplied rotordynamic amplitude and phase data, which has
been shown to be of reasonable accuracy.

314
Chapter 9 – Analysis of Results and Discussion

9.4 Comments on Existing Theoretical Models

At current with regard to the thermal modelling there are two well documented
procedures. In addition to this both methods have been used to couple in
some way with the shaft dynamic characteristics. In the case of Keogh [9] et al
an actual rotordynamic formulation was incorporated, albeit a fairly straight
forward case. Balbahadur [19] does not use analysis with respect to
rotordyanmic influences but is based solely on a steady threshold for the
amount of thermally induced shaft bend for the spot analysis performed.
Whilst this does provide a simple method of analysis it unfortunately does not
give any real account to the rotordynamic sensitivity of the rotor with respect
to scale. The use of the stability threshold factor based on rotor mass to
determine the unbalance force raises questions when light weight rotors are
being used, and unbalance forces can quite soon reach levels which account
for a significant percentage of the rotor mass. The bearing stiffness with
respect to balance force also do not necessarily scale with rotor size. Which
could introduce issues with the use of a constant stability threshold factor for
all rotor sizes and configurations.

Whilst the method of Keogh et al [9] provides a much more thorough account
of both the thermal conditions and rotordynamic interactions the procedure is
analytically intensive and as such does not lend itself to easy introduction with
existing rotordynamic codes/procedures. The significant advantage to the
more complex method with respect to reliability, is the ability for the model to
model the convective and conductive heat transports within the oil film and
around the bearing. It is these elements which have been identified as being
key to the prediction of the hot spot location on around the journal, which
dictates the phase angle of the thermal bend. This phase angle is key to
understanding how sensitive a rotor likely to be to thermally induced
instability. This is a significant restriction in the current more applicable tools,
where the methods do not provide any provision for determining the thermal
phase angle. There is a tendency for such methods to predict higher
temperature differentials and in general provide a more conservative
prediction of instability. Whilst conservative design is generally a positive
approach care must be taken that systematic overdesign does not result.

315
Chapter 9 – Analysis of Results and Discussion

Some key observations from analytical procedures :

Balbahadur [19] procedure

• No account for hot spot phase angle prediction

• Uses steady state assumption for oil film temperatures at dynamic
locations and as such would have a tendency to over predict the
bearing temperatures and journal temperature differentials
• Method can be easily applied and built into existing rotordynamic
codes/methods
• Dynamic tilting pad bearing solution provides good account for oil film
functions under dynamic conditions
• No account provided for non circular geometry on fixed profile bearings
• Method suggests more centred orbits are more prone to instability
which is in some contraction to other analytical studies

Keogh et al [9] procedure

• Procedure couples the thermal, hydrodynamic and rotordynamic

problems giving a holistic evaluation
• Method provides for oil film conditions which allow for determination of
not only temperature differentials but hot spot phase lag angle
• Procedure for coupling shaft bend angle as a feedback gain provides
more definitive measure of system stability
• Method is computationally complex and does not readily apply to more
complex bearing configurations without further development
• Computation time would result in difficulties to incorporate with
common rotordynamic analysis methods

In summary the work presented by Keogh et al [9] provides a highly detailed

account of the conditions taking place in a whirling shaft and the effects on
thermal stability, although some assumptions are still introduced. However
due to its highly involved nature it is generally too complex for practical
implementation at an industrial level. The work and method can, however,
provide a good benchmark by which to evaluate other less involved analysis
procedures. The work of Balbaadur [19] introduced a method which can be
built into the practical design procedures and follows the type of approach that
is needed for industrial use. However, some of the assumptions and
idealisations that have had to be made mean that some critical elements of
the data, such as phase angle and rotordynamic interaction are omitted. To

316
Chapter 9 – Analysis of Results and Discussion

increase the reliability of the procedure some method of account for these
elements would be necessary. The works presented in the literature have
clearly advanced the understanding and modelling capability of the
phenomenon but further development of the analytical procedures is required
along with practical test data and results with which to validate the
approaches.

317
Chapter 9 – Analysis of Results and Discussion

9.5 Shaft Journal Temperature Theoretical Model Results

The theoretical model presented in this research consist of three principal

elements :

• The oil film energy model

• The orbit perturbed thermal model
• The time domain finite difference journal model

The principal focus of all of the aspects involved in this procedure is to provide
as simple and as efficient solution method as possible whilst still providing
results of acceptable accuracy. As it has not been part of this research to
perform an extensive parametric study into the application of the presented
journal hearting model, some simple case studies have been considered to
provide and insight as to how well the method compares to existing data or
other published work.

As the principal element in determining the temperature distributions within

the shaft journal are dependant on the lubricant film thermal conditions, it is
key that the energy model used to represent the heat generation within the
bearing provides suitably accurate representation.

9.5.1 Oil Film Energy Model and Temperature Equation

To investigate how the presented oil fil temperature energy equation performs,
a comparison to the case presented by Keogh et al [9] is made for the zero
orbit condition.

Keogh et al [9] uses a complex thermohydrodynamic model incorporating

CFD techniques to determine the steady and perturbed bearing lubricant film
temperature around the bearing. The model used by Keogh et al [9] considers
cross film temperatures in addition to circumferential temperature variation. As
such, for the purposes of comparison to the current research, then the cross
film mean temperature from the results presented by Keogh is used. The
bearing considered is a single inlet plain cylindrical bearing, with the oil inlet
positioned at 180o to the point of minimum film thickness as outlined in Figure
244. The following material properties and parameters are taken for the
considered load case :

318
Chapter 9 – Analysis of Results and Discussion

ε = 0.648 Ω = 1000 rad / s

R j = 0.05 m Lb = 0.035 m cr = 100 µm
o
υ40 = 46 cst υ100 = 5.5 cst TLin = 45 C
ρ L = 850 kg / m3 cL = 2000 J / kg / K

Figure 244 Keogh et al [9] Bearing Case

Using the equations set out in chapter 8, a bearing temperature rise ∆T of

45.6oC was calculated with an effective pad inlet temperature To1 of 50.3oC.
The mean effective viscosity used for the analysis was calculated as 12.9 Cst,
where this initial condition was estimated from the mean temperature rise for a
fully centred journal in the bearing. Figure 246 shows the predicted
circumferential temperature distribution for both that presented by Keogh et al
[9] and that of the current research.

319
Chapter 9 – Analysis of Results and Discussion

100

Mean (Cross Film) Lubricant Temperature

90

80
(degrees C)
70

60

50

40
Keogh [9] Example
Current Research
30
0 45 90 135 180 225 270 315 360
Angle Around Bearing (degrees)

Figure 245 Circumferential Temperature Distribution for Steady Eccentricity

Comparison Between Literature and Research

The results of the example case show good correlation between the mean film
temperature presented by Keogh et al [9] as calculated by a comprehensive
coupled fluid and thermo dynamic approach. The current research
formulation appears to predict a higher pad inlet temperature by ~4.5oC than
that of Keogh et al [9]. Also the overall temperature rise is slightly less than
that of Keogh et al [9] with a ∆T of 48oC and 45.6oC for [9] and the current
research respectively. It should be noted that the temperature distribution in
Figure 245 is the actual lubricant temperature and no effect of axially
averaged heat (flux due to the fractional film function in the cavitation region)
is present.

For the above result, the application of the closed form energy solution of the
bearing temperature distribution shows very promising results when
compared to a well considered and detailed CFD type analysis. This provides
a good indication that the method can provide acceptably accurate results for
the prediction of the oil film temperature where the basic input parameters are
defined. However, it is recognised that a much wider range of application is
required to validate the model across a greater range of parameters and
geometric arrangements before a definitive understanding of the scope and
accuracy of the solution can be made.

320
Chapter 9 – Analysis of Results and Discussion

9.5.2 Oil Film Temperature Perturbations

To evaluate the journal differential heating aspects in the orbit perturbed

thermal model, the same case as presented by Keogh et al [9] was used.
Keogh et al [9] goes on to predict the journal temperature differential for both
forward and backward whirl orbits. The basic operating parameters are used
as for the steady zero orbit case, but now the synchronous orbits are
introduced. The following additional parameters for the dynamic perturbed
case are :

σ 1,2 = 0.1 k L = 0.15 W .m −1.K −1 H J = 50 W .m −2 .K −1

The forward orbit was solved assuming Couette flow across the oil film. The
resulting temperature distribution is shown in Figure 246. As can be seen the
results show an almost sinusoidal distribution with the journal surface hot spot
lagging the minimum film thickness.

Minimum Film
Thickness Point

Direction of Rotation

Figure 246 Calculated Journal Temperature Distribution with a Forward Whirl

Orbit
A harmonic analysis of the temperature distribution is presented in Figure 247,
where the 1st harmonic ∆T and associated phase angle are 7.8oC and -66.9o
respectively. It can be seen that the fundamental harmonic dominates, as
would be expected, but there is a small amount of 2nd harmonic component -
approximately 10% of the magnitude of the 1st harmonic.

321
Chapter 9 – Analysis of Results and Discussion

9 0
Delta T
8 Phase -10

1st Harmonic Delta T Phase Angle

7
-20

Shaft 1st Harmonic Delta T

6
-30

(degrees C)
5
-40
4
-50
3

2
-60

1 -70

0 -80
1 2 3 4 5 6
Harmonic Order

Figure 247 Journal Temperature Distribution Harmonic Analysis

with a Forward Whirl Orbit

The model was also solved for the backward whirl condition. The resulting
journal temperature distribution is presented in Figure 248 where the
temperature distribution can again be seen to be predominantly sinusoidal.
However a more significant point is that the journal hot spot now leads the
point of minimum film thickness on the journal. The temperature distribution
harmonic analysis is given in Figure 249. The phase angle can be seen to
lead by some 55 degrees with the 1st harmonic ∆T being 3.6 oC.

Direction of Rotation

Minimum Film
Thickness Point

Figure 248 Calculated Journal Temperature Distribution with a

Backward Whirl Orbit

The predicted ∆T and phase angles presented by Keogh et al [9] were 10 oC

and 4 oC for forward and backward whirl orbits respectively. The phase lag for
the forward whirl orbit was given as 55o. No specific phase angle was given

322
Chapter 9 – Analysis of Results and Discussion

for the backward whirl orbit, but it was specified that it was leading the point of
minimum film thickness on the journal.

4 120

3.5
100

1st Harmonic Delta T Phase Angle

Shaft 1st Harmonic Delta T 3

80
2.5
(degrees C)

2 60

1.5
40
1

Delta T 20
0.5
Phase
0 0
1 2 3 4 5 6
Harmonic Order

Figure 249 Journal Temperature Distribution Harmonic Analysis

with a Backward Whirl Orbit

The results from the method presented within this research compare
reasonably to the work presented by Keogh et al [9] for the temperature
distributions. Forward whirl orbit is some 2.5 oC (25%) less than the literature
and the backward whirl ~0.5 oC less (12%). The phase lag was however a
little further away from that of Keogh et al [9] by some 11o.

In summary, whilst some variation does exist between the method presented
within the research and the literature case, this still gives a good correlation in
light of the objectives. Some discrepancy would be expected when a
simplified method with greater assumptions is adopted over the more complex
and costly analysis, however the results are still well within acceptable
accuracy levels for such an analysis. It was noted during the evaluation of the
procedure that the predicted phase angle is quite sensitive the convective flow
fraction. With phase angle shifts of 10-20o observed for 10% change in fL,
highlighting the need for further investigation to this feature. However, if such
levels of accuracy as have been seen in this case study can be demonstrated
by this method across a wider number of cases, then use in practical
rotordynamic analysis would be well justified. To account for the variations
seen in phase angle and ∆T , a tolerance could be applied to the calculated
phase angles and temperature differences and a sensitivity analysis
performed to ensure a stable operating range is selected for a rotor. This type
of sensitivity analysis is common place in rotordynamic procedures where
levels of uncertainty for some specific parameters is relatively high.

323
Chapter 10 - Conclusions

Chapter 10

Conclusions

10. Conclusions

This thesis has considered the phenomenon of synchronous rotor dynamic

thermal bowing initiated and propagated by thermal differential heating of the
bearing journal due to unbalance induced shaft whirl, which under certain
conditions has been observed to induce unstable vibration response
behaviour (the ‘Morton’ effect). A number of industrial cases have been
reported in the literature where synchronous related shaft vibrations have
been observed to increases with time whilst at steady running conditions,
where the phase is seen to alter with changes in vibration amplitude. This
thesis has looked in detail at the theories and has analysed several of the
referenced industrial cases to understand the rotordynamic and bearing
performance characteristics that relate to the promotion and initiation of
thermal bow induced instability. The work presents critique of existing
evaluation methodology and uses a specially designed experimental rig to
obtain controlled experimental data with relation to the bearing induced
thermal differential heating of the shaft journal. This thesis presents an
alternative procedure of theoretical modelling of the thermal development of a
whirling shaft journal to determine shaft temperature differential and phase
data.

It has been shown that it has been possible, using appropriate rotordynamic
models, to reproduce the rotor dynamic response characteristics which
correlate accurately to the measured test results for two cases published in
the literature and an as yet unpublished industrial example. It has been
possible to identify some of the fundamental characteristics that have been
observed to be present in the rotors where thermal rotor bow and instability
has been reported. The study has shown that in cases where the thermal
instability exists, very specific relationships between the rotor operating
deflection shapes are required in combination with the vibration phase
conditions. Whilst a rotor can exhibit high rotor vibrations associated with a
shaft overhung section, a more complex set of conditions relating the
rotordynamic and 1st harmonic temperature distribution phase at the bearing
journal would be required to result in a condition of instability. The

324
Chapter 10 - Conclusions

rotordynamic study also shows that in certain rotor configurations, imbalance

effects at overhung sections in shafts (where the relative distances of bearing
span and overhung section are comparable) could result in unstable
behaviour being observed in opposing bearing locations to those where the
thermal bend development is taking place. In all cases evaluated, the rotor
sensitivity to unbalance was seen in some way to correlate to the level and
severity of instability observed. Where rotors were predicted as being highly
sensitive to imbalance, these were shown to become fully unstable. Those
where transient or cyclical unstable characteristics were displayed were
predicted to be less sensitive to unbalance forces. A common characteristic in
all cases was that, due to the natural vibration response of the rotors, they
operated with mode shapes where dynamic shaft bending is present at or
near to a bearing location. Critical speeds provided some dynamic
amplification in the regions of unstable behaviour.

An experimental test rig design has been presented which has been
rotordynamically engineered to replicate the dynamic conditions observed in
rotors with known thermal rotor bow and unstable ‘Morton’ type
characteristics. The test rig introduces the use of non circular fixed geometry
hydrodynamic bearings, as commonly applied in high speed rotating
machines. To date no controlled operational data for these bearing types has
been presented in the relevant literature. The use of variable geometry
bearings is also included in the test rig design analysis and specification. The
rotor design is predicted to undergo unstable thermal bend conditions within
the speed range for both fixed and variable geometry bearing configurations
under moderate unbalance conditions when calculated in accordance with a
key method presented within the literature [19]. The rig design includes a
method by which journal temperature measurements can be accurately
acquired from the journal for operating speeds up to 14000 rpm. This allows a
range of journal temperature data to be obtained for different operating
conditions for both fixed and variable geometry arrangements. This new area
of experimental data is not covered in the availble literature and has been
generated and presented in this thesis.

The experimental test rig and associated instrumentation has been

manufactured, constructed and commissioned as per the design outlined in
the above paragraph. The test rig has been operated across a speed range of
13,000 rpm in both fixed geometry offset half and variable geometry tilting pad
bearings, running above the 1st bending natural frequency and within 10% of
the 2nd natural bending modes of the rotor. Shaft dynamic displacement,

325
Chapter 10 - Conclusions

phase and journal temperature distribution data has been obtained for a
number of operating conditions and configurations for several states of initial
mechanical unbalance.

Experimental results data is presented which demonstrates good correlation

with predicted rotor unbalance synchronous vibration amplitudes and phase.
Specific results data is presented where the 1st harmonic rotor journal
temperature distributions are extracted and the phase conditions presented
with respect to the initial mechanical unbalance location and the point of
journal minimum film thickness.

The work shows that it was not possible to initiate instability due to the thermal
rotor bow when running in either the fixed geometry or variable geometry
bearings. Magnitudes of initial unbalance mass were introduced that were
consistent with those used in the theoretical model for which instability was
predicted, and also higher unbalance magnitudes of up to 6-8 times greater
were also applied with stable vibrations observed throughout. Running periods
of over 30 minutes under steady speed and inlet conditions were performed.

Shaft journal temperature differentials were measured and a systematic

increase in shaft temperature differential was observed as unbalance and
shaft orbits were increased. The development of the shaft journal temperature
differentials was shown to follow an almost linear relationship with shaft whirl
orbit size for vibrations of <20% of the bearing diametrial clearance for the
fixed profile arrangement. This is consistent with the assumptions presented
in several literature sources. However, the experimental data shows that for
the fixed geometry offset half bearings with orbits >20% of diametrial
clearance, the rate of temperature differential development reduces to a lower
rate.

Actual shaft journal temperature differentials are shown to be significantly

lower than those predicted using the current theoretical model presented in
the literature [19] for the same dynamic displacement conditions. Fixed
geometry bearings were only 35% of predicted levels and variable geometry
lower at ~15%. Maximum temperature differentials were measured of 1.5-
1.7oC.

Shaft ‘hot’ spot phase angle data is measured and systematic relationships
are seen to exist between unbalance conditions, shaft speeds and
rotordynamic behaviour. For cases where little rotordynamic behaviour is

326
Chapter 10 - Conclusions

active at the bearing, it is shown that ‘hot’ spot phase lags reduce with
increased orbit size and tend to increase with elevated speed. Whereas with
flexible rotors undergoing significant rotordynamic response behaviour, it is
shown that these influences create a more complex bearing-journal interaction
and resulting temperature profile and ‘hot’ spot phase lag. In particular it is
shown that the orbit shape and orientation relates to significant variation in
temperature differential development and phase lag angles. This research
shows that the tilting pad bearings produced higher temperature differentials
than the fixed geometry offset halve bearing arrangement for the same orbit
size ratios (%Cd). The work shows that, in all cases, the circumferential
temperature profile consists not only of a fundamental harmonic component,
but that higher order distributions exist within the generated temperature
profile.

Calculated and measured rotordynamic results show that for flexible rotors
undergoing significant shaft deflections and vibration phase shifts, the phase
conditions within the bearing can be significantly different from those at the
measurement point of the rotor, even where the sensors are positioned at the
outer edge of the bearing and the axial distances involved are small (10-15
mm). The impact of these variations result in significant errors being
introduced when determining the phase angle relationships between the
journal hot spot and the initial mechanical unbalance location, unless suitable
compensations are made for these variations.

A novel computational modelling technique for evaluation of the thermal

development within the bearing journal is presented. The technique allows for
the journal temperature development to be modelled in the time domain
without the requirement for complex CFD analysis techniques, enabling
practical application to be made when performing more routine rotordynamic
stability analysis procedures. A simplified energy model is introduced which
allows the steady state bearing temperature distribution to be established by
means of a closed form equation developed within this work. A numerical
procedure is applied to model the dynamic temperature perturbations and
convective heat transfer within the oil film. The resulting steady state and
dynamic oil film temperatures and phase lag angles provide the necessary
boundary conditions for solution of a transient finite difference grid model of
the shaft journal. The resulting journal temperature differential and phase lag
angles are then determined for specific operating conditions. This allows the
necessary stability analysis to be performed. The procedure allows for correct
modelling of the fixed geometry bearing bore profile - a feature which is

327
Chapter 10 - Conclusions

overlooked by means of simplifications in the non CFD techniques presented

in the literature. This thesis is the first theoretical model presented which
predicts the ‘hot’ spot phase lag angle without the need for complex and
highly computationally intensive CFD procedures to be applied.

The oil film energy model in this research is compared against the theoretical
results data in the literature where complex CFD techniques have been
applied. The results correlate to within approximately 5 oC for the conditions
considered. Similar comparisons are made for the dynamic journal
temperature differential and ‘hot’ spot phase lag where variations of
~2.5oC(~25%) oC and -11.9o result.

The key contributions of this work are :

1. Presentation of a detailed characterisation of the rotordynamic and

bearing performance features present within rotors where thermally
induced rotor bow instabilities have occurred in industrial machines.

2. Design, development and commissioning of a dedicated high speed

rotor test rig for the accurate measurement of rotating journal shaft
temperatures in a whirling shaft for both rigid and flexible rotor systems.
The rotordynamic conditions within the test rotor replicate those
conditions identified as significant with respect to the elements
necessary for the development of thermally induced rotor bow and
associated instabilities.

3. The experimentally controlled generation of shaft journal temperature

differential data in whirling shafts for both fixed and variable geometry
bearings. The temperature differentials are presented for a range of
operating parameters, with detailed information relating to the
associated phase lag angles under these conditions. This data can be
used to further develop and validate the theoretical models in place
today and those presented within this work. To date no controlled
experimental data on fixed geometry bearing journal temperature
measurement has been presented in the literature. Moreover this is the
first presentation of detailed phase lag angle data for a range of
operating arrangements and conditions for both rigid and flexible rotor
systems. Whilst it was not feasible to send the rotor unstable within this
work, the rotor was seen to exhibit the desired rotordynamic
characteristics identified as necessary for promotion of the Morton

328
Chapter 10 - Conclusions

Effect instability. This would enable the use of a test rotor of this design
to be used for future studies where alternative bearing configurations
are investigated.

4. Comparison of existing theoretical model predictions to experimental

data for both fixed and variable geometry bearings. A number of
significant differences have been identified between actual and
calculated parameters. More fundamentally, a lack of correlation
between the predicted thermal bow conditions and associated stability
with that observed experimentally has been shown. This highlights the
need for development of theoretical modelling tools currently available
for thermal differential development and stability prediction.

5. Derivation of a closed form energy based equation to describe the

cross film mean lubricant temperature for cylindrical journal bearings.

6. Following from 4 above, a new modelling procedure for the prediction

of shaft journal temperature differential development is presented. The
restrictions present within other procedures from the literature are
addressed with respect to ‘hot’ spot phase lag angle and prediction of
the journal temperature differential that result from the ‘quasi’ steady
state assumption for shaft whirl orbit perturbations. The presented
method uses a significantly lower complexity solution than that of the
CFD approaches, which enables practically implementable evaluations
to be made, whilst retaining sufficient accuracy for appropriate stability
risk assessments to be performed.

329
Chapter 11 – Further Work

Chapter 11

Further Work

11. Further Work

In addition to the work presented here, an extensive programme of work to

focus on the temperature developments within the bearing should be
considered. Through the use of the rigid rotor arrangement where little
rotordynamic influence is present within the bearing, other than simple
unbalance response, it would be possible to generate both journal
temperature differential data and ‘hot’ spot phase lag angle data. From this a
wider range of operating parameters can be considered, such as oil
viscosities, oil supply temperatures, oil supply flows and more significantly
bearing clearance. The work would need to cover a range of bearing types.
The larger collection of acquired data would be used to provide guidance on
selection of bearing configurations which would be suitable for development of
instability in the flexible rotor configuration, validation and development of
theoretical models and also for use as guidance in the design of industrial
machine arrangements.

The introduction of several improvements or design modifications to the test

arrangement and instrumentation setup should be investigated. A significant
aspect which should be considered is the elimination of the mechanical slip
ring configuration for the measurements of the shaft journal temperatures.
Whilst it was possible to obtain acceptable temperature measurements in this
research, the problems and reliability issues associated with the slip ring
method of obtaining signals resulted in significant down time and extended
experimental running. Investigations and development of a telemetric solution
should be performed which would eliminate the mechanical elements from the
data acquisition and, if appropriately designed, reduce the low level noise
present within mechanical slip rings. The inclusion of additional journal
temperature sensors should also be considered so that the harmonic range to
which the measured temperature distribution can be described is increased
beyond the fundamental.

330
Chapter 11 – Further Work

The possibility of introducing proximity probes at the bearing centre location to

enable direct, accurate measurement of the shaft vibration displacements and
phase angles should be explored. It was shown in the current research that
variations can exist between the measurement point and the bearing centre
position in flexible rotors. These variations are sufficient to significantly affect
the analysis of the phase angle measurements of the journal ‘hot’ spot
location, even when axial distances between the bearing centre and
measurement point are small (~10-15 mm). The introduction of this feature
can be applied without too much difficulty for the variable geometry
arrangements but some special consideration and testing will be required
where this is to be applied to typical fixed geometry arrangements. The
inclusion of direct measurement of pad angle for the variable geometry
arrangement would also provide a validation of the conditions present within
the oil film function which has a direct influence on the predicted temperature
differential generation. Some current models in the literature assume this
aspect to follow a ‘quasi’ steady state position for the orbit perturbations,
where in practice inertial effects in the pads will influence the dynamic tilt
angles presented within the bearing.

Using the experimental results presented within this research, the results of
any further testing in the rigid rotor design and the developed theoretical
models should focus on investigating different bearing configurations,
clearances and overhung sections. The aim of this should be to define an
arrangement where the conditions required for instability of the test rotor can
be replicated. This information should be used to provide data enabling
validation of the relationships between actual instability thresholds and those
predicted in theoretical models.

Further benchmarking and development is required of the newly presented

theoretical model to address a more widely spread number of validation
cases. A more extensive study using some CFD case studies could be used
for comparison when looking to further develop the convective element in the
analytical procedure, specifically with respect to different bearing profile
geometries and where high levels of backward whirl are present. Also more
extensive comparison of the model is needed to experimentally determine
data and greater understand the accuracy and scope of the new procedure
presented here.

Further development of the specific procedures and computer codes

necessary to integrate the techniques for the evaluation of tilting pad bearings

331
Chapter 11 – Further Work

is required and also the inclusion of an integrated Reynolds solution into the
calculation procedures would improve the functionality of the routine. This
could also be extended to incorporate a more comprehensive stability
assessment and method for closer integration to rotordynamic modelling tools,
which would provide a further significant step forward in the practical analysis
of rotors for Morton stability assessment.

332
References and Bibliography

References and Bibliography

References

1. Newkirk, B.L., 'Shaft Rubbing', Mechanical Engineering, Vol. 48, 1926,

p.830.

2. Kroon, R.P. and W.A. Williams, 'Spiral Vibration of Rotating Machinary'

Proceedings of 5th International Congress of Applied Mechanics, Wiley,
New York, 1939, p. 712 – 718.

3. Dimarogonas, A.D., 'An Analytical Study into the Packing Rub Effect in
Rotating Machinery', Dissertation, Rensselaer Polytechnic Institute,
Troy, N.Y., 1970.

4. Dimarogonas, A.D., 'Newkirk Effect : Thermally Induced Dynamic

Instability of High Speed Rotors', ASME, 1973.

5. Dimarogonas, A.D., 'Heat Distribution and Flash Temperatures in

Radial Seals', Wear, Vol. 23, No. 1, 1972.

6. Kellenberger W., 'Spiral Vibrations Due to Seal Rings in

Turbogenerators Thermally Induced Interaction Between Rotor and
Stator', Journal of Mechanical Design, Vol. 102, p.177-184, January
1980.

7. Ericsson U., 'Temperature Distribution in the oil Film of a Vibrating

Tilting Pad Bearing', Thesis Report, Chalmers University of Technology
Goteborg, Sweden, 1980.

8. Schimied, J., 'Spiral Vibration of Rotors', Rotating Machinery Dynamics,

ASME, Vol. 2, p.449-456, September 1987.

9. Keogh P.S. and Morton P.G., 'Journal Bearing Differential Heating

Evaluation with Influence on Rotor Dynamic Behaviour', Proc. R. Soc.
Lond. A 441, p. 527-548, 1993.

10. Keogh P.S. and Morton P.G., 'The Dynamic Nature of Rotor Thermal
Bending due to Unsteady Lubricant Shearing within a Bearing', Proc.
R. Soc. Lond. A 445, p. 273-290, 1994.

333
References and Bibliography

11. de Jongh F.M. and Morton P.G., 'The Synchronous Instability of a

Compressor Rotor Due to Bearing Journal Differential Heating', ASME
Int. Gas Turbine and Aeroengine Cong. and Exp.', The Hague,
Netherlands, Paper 94 GT-35, 1994.

12. Faulkner H.B., Strong W.F. and Kirk R.G., 'Thermally Introduced
Synchronous Instability of a Radial Inflow Overhung Turbine, Part I',
Proceedings of DETC'97, 1997.

13. Faulkner H.B., Strong W.F. and Kirk R.G., 'Thermally Introduced
Synchronous Instability of a Radial Inflow Overhung Turbine, Part I',
Proceedings of DETC'97, 1997.

14. de Jongh F.M. and van de Hoeven P., 'Application of a Heat Barrier
Sleeve to Prevent Synchronous Rotor Instability', Proc. of the 27th
Turbomachinary Symposium, 1998.

15. Larsson B., 'Journal Asymmetric Heating – Part I : Nonstationary Bow',

Journal of Tribology,Vol. 121, p. 157-163, January 1998.

16. Larsson B., 'Journal Asymmetric Heating – Part II : Alteration of Rotor

Dynamic Properties', Journal of Tribology,Vol. 121, p. 164-168,
January 1998.

17. Tucker P.G. and Keogh P.S., 'On the Dynamic Thermal Sate in a
Hydrodynamic Bearing With a Whirling Journal Using CFD
Techniques', ASME Tribology Conf., 1995.

18. Gomiciaga R. and Keogh P.S., 'Orbit Induced Journal Temperature

Variation in Hydrodynamic Bearings', Journal of Tribology, Vol. 121, p.
77-84, 1999.

19. Balbahadur A.C. , 'A Thermoelastohydrodynamic Model of the Morton

Effect Operating in Overhung Rotors Supported by Plain or Tilting Pad
Journal Bearings', PhD Thesis, Virginia Polytechnic Institute and
University, 2001.

20. Balbahadur A.C. and Kirk R.G. , 'Part I : Theoretical Model for a
Synchronous Thermal Instability Operating in Overhung Rotors', 2002.

334
References and Bibliography

21. Balbahadur A.C. and Kirk R.G., 'Part II : Case Studies for a
Synchronous Thermal Instability Operating in Overhung Rotors', 2002.

22. Taylor H.D., 'Rubbing Shafts Above and Below Resonant Speed', G.E.
Technical Information Series, 16709, 1924.

23. Balbahadur A.C., Z Guo and Kirk R.G., 'Synchronous Thermal

Instability Prediction for Overhung Rotors', Proc. of the 32nd
Turbomachinary Symposium, 2003

24. Kirk R.G., Balbahadur A.C., ‘Thermal distorition syncrounous rotor

instability’, 7th International Conference on Vibrations in Rotating
Machinary, 2000

25. Jeffcott H.H., ‘The lateral vibration of loaded shafts in the

neighbourhood of a whirling speed – The effect of want of balance’,
Philosophical Magazine Ser. 6-37 p. 304-314, 1919

26. Reynolds O., ‘On the theory of lubrication and its application to Mr
Beauchamp Tower’s experiments including and experimental
determination of the viscosity of olive oil, Phil. Trans. 177 (i) p. 157-
234, 1886.

27. J. W. Lund and B. Sternlicht, ‘Rotor-Bearing Dynamics with Emphasis

on Attenuation,’ Journal of Basic Engineering, Trans. ASME, Vol. 84,
Series D, 1962, pp. 491–502.

28. J. W. Lund, ’Stability and Damped Critical Speeds of a Flexible Rotor in

Fluid-Film Bearings,’ Journal of Engineering for Industry, Trans. ASME,
Vol. 96, Series B, No. 2, May 1974, pp. 509517

29. Thoma D., Zeitschrift des Vereines deutscher Ingenieure, p. 985, July
25 1925.

30. Kirk G.R., ‘Morton Effect Analysis – Theory, Program and Case Study’,
ISCORMA-3, Cleveland, Ohio, 19-23 September 2005.

31. Marscher W., Illis B., ‘Journal Bearing “Morton Effect” Cause of Cyclic
Vibration in Compressors’, Tribology Transactions, 50 : p. 104-113,
2007.

32. Schmied J., Pozivil J., Walch J., ‘Hot spots in turboexpander bearings:
case history, stability analysis, measurements and operational
experience’, ASME Turbo Expo, Berlion Germany, 2008

335
References and Bibliography

33. de Jongh F., ‘The synchronous rotor instability phenomenon – Morton

Effect’ Proceedings of the 37th Turbomachinery Symposium 2008

34. Morton P.G., ‘Recent advances in the study of oil lubricated journal
bearings’, Proceedings of fourth international conference on
rotordynamics, IFFToMM, Chicago, Illinois, p. 299-305.

35. Childs W.C., Saha R., ‘A new, iterative, synchronous-response

algoritham for analyzing the Morton Effect’, Proceedings of ASME
Turbo Expo, Vancouver, British Columbia, Canada, June 2011

36. Murphy B.T., Lorenz J.A., ‘Simplified Morton Effect analysis for
synchronous spiral instability’, ASME Power Division Special Section,
December 2009

37. Murphy B.T., Lorenz J.A., ‘Case study effect shaft differential heating in
a variable-speed rotating electric machine’, Proceedings of ASME
Turbo Expo, Vancouver, British Columbia, Canada, June 2011.

38. Lorenz, J.A., ‘Implementation of fluid-film bearing shaft differential

heating calculations using commercial CFD software’, MSc Thesis,
B.M.E., University of Minnesota, 2009

39. Dowson D., Hudson J.D., Hunter B., March C.N., ‘An experimental
investigation of the thermal equilibrium of steadily loaded journal
bearings’, Proceedings of the Institute of Mechanical Engineers, Vol
181, p. 70-80, 1966.

40. Sommerfeld A., ‘Zur hydrodynamishen theorie der

Schiermittelreibung’, Z.Math. u. Physik, Vol. 50, p. 97, 1904

41. Ocvirk F.W., Dubois G.B., ‘Analytical derivation and experimental

evaluation of short bearing approximation for full journal bearings’,
NACA Report 1157, 1953

42. Garner D.R., ‘The use of design procedures for plain bearings’,
Industrial Unit of Tribology, University of Leeds, September 1908

43. Cameron A., ‘The Principles of Lubrication’ Longmans, 1996

44. Stachowiak G.W., Batchelor A.W., ‘Engineering Tribology’ Third Edition

, Elsevier Butterworth-Heinemann, 2005

45. Cole J.A., Hughes C.J., ‘Oil Flow and Film Extent in Complete Journal
Bearings’, Proceedings of the Institute of Mechanical Engineers, Vol
170, p. 499-510, 1956.

336
References and Bibliography

46. Wright K., ‘Chebyshev collocation methods for ordinary differential

equations’, Computer Journal, Vol. 6, Issue 4, p. 358-365, 1964

47. San Andres L., ‘An Overview of Tilting Pad Journal Bearings’, Dept. of
Mechanical Engineering, Texas A&M University, Febuary 2000

48. Vance J.M., ‘Rotordynamics of Turbomachinery’, John Wiley & Sons,

1988

49. MATLAB, ‘Programming Environment Software’, The MathWorks Inc

50. Ettles C.M., ‘The analysis of pivoted pad journal bearing assemblies
considering thermoelastic deformation and heat transfer effects’,
Tribology Transactions, Vol. 35, No. 1, p. 156-162

51. XLRotor, ‘Rotordynamic Software Package’, Rotating Machinery Inc

(RMA).

52. JPU, ‘Vickers – JPU Tilting Pad Journal Bearing Performance Program’
V3.21’’ – Custom AGSL code

53. API Standard 617, ‘Axial and centrifugal compressors and expander-
compressors for petroleum chemical and gas industry services’, 7th
Edition, American Petroleum Institute, Washington D.C.

54. HYBJRN, ‘Rotordynamic and Hydrodynamic Bearing Calculation

Program’, Franklin Institute AGSL custom code

55. ISO 1940-1:2003 ‘Mechanical vibration -- Balance quality

requirements for rotors in a constant (rigid) state - Part 1:
Specification and verification of balance tolerances’, International
Organisation for Standardisation
56. Steinhart I.S., Hart S.R. ‘Calibration curves for thermistors’, Deep Sea
Research Vol.15(3), p. 497-503, 1968

57. ‘Industrial Temperature Sensors Datasheet’, Industrial Sensors Ltd,

www.itsirl.com

58. Biggs D.H., ‘Electrical Runout and Eddy Current Displacement

Proximity Transducers (Bently Nevada Literature No. L0360)’, ASME
Paper, September 1975

59. Shannon C.E., ‘Communication in the Presence of Noise’, Proceedings

of the IEEE, Vol. 86, No. 2, February 1998 (reprinted from earlier
paper)

337
References and Bibliography

60. www.cdeep.iitb.ac.in

61. Muneer T., Kubie J., Grassie T., ‘Heat Transfer : A problem solving
approach, Volume 1’, Taylor & Francis, 2003

62. Li Z., ‘Finite Difference Methods Basics’, Centre for Research in

Scientific Computation & Department of Mathematics North Carolina
State University Raleigh, NC 2769

63. Martin F.A., ‘Developments in Engine Bearing Design’, Tribology

International, Vol. 16. no. 3, p. 147-164, 1983

64. Martin F.A., Lee C.S., ‘Feed-Pressure Flow in Plain Journal Bearings’
Tribology International, Vol. 26, No. 3, p. 381-392, 1983

65. Marscher D., McGinley J.J., ‘Vibration Troubleshooting of an Air

Compressor’, Sound and Vibration Journal, May 2006

66. Sundqvist H., Veronis G., ‘A simple finite-difference grid with non-
constant intervals’, Tellus XXII (1970), 1

67. Blake R. E., ‘Basic Vibration Theory’, Shock and Vibration Handbook,
Fourth Edition, Chapter 2 – p. 2.18, 1995

338
References and Bibliography

Bibliography

1. Stroud Engineering Mathematics, K.A. Stroud, Fourth Edition, 1995

2. Stroud Further Engineering Mathematics, K.A. Stroud, Third Edition,

1996

3. Mechanics of Materials Volume 1: An Introduction to the Mechanics of

Elastic and Plastic Deformation of Solids and Structural Materials: v. 1
by E.J. Hearn (9 Jul 1997)
4.
Rotordynamics of Turbomachinery by John M. Vance (1988)

5. Machine Dynamics MSc Course Notes Part 1&2, Cranfield University

6. Signal Processing MSc Course Notes, Cranfield University

7. Shock and Vibration Handbook by Cyril M. Harris (Hardcover - Dec

1995)

8. Applied Thermodynamics for Engineering Technologists by T.D.

Eastop and A. Mcconkey (1993)

9. Handbook of Numerical Heat Transfer by W. J. Minkowycz, E. M.

Sparrow, G. E. Schneider and R. H. Pletcher (1988)

10. Basic Lubrication Theory (Ellis Horwood series in engineering science)

by A CAMERON (1977)

339
Appendices – Appendix A

Appendix A

Numerical Procedure for Measurement of Bearing Pre-load

This appendix details the numerical procedure used to measure the bearing
bore geometry of the fixed pad offset half bearing. The procedure is based
around using traditional clock or CMM readings in the bearing radial plane to
then use a curve fitting approach to determine the shaft preload and
clearance conditions.

Db

Rj Rb

χr
Figure A1. Offset Half Bearing Bore Profile System of Coordinates

Where

Cθ = Rb − R j

Cr = Db − 2.R j

340
Appendices – Appendix A

Assuming that the radial clearance around the bearing has the following form :

Cθ = Cr + χ r .cos (θ − φm ) Eq. A1
Where

Cθ = theoretical radial clearance from bearing centre

Cr = machined radial clearance
χ r = bore offset value
θ = angle around bearing
φm = bore offset cant angle

Applying trig identities :

Cθ = Cr + χ r . ( cos (θ ) .cos (φm ) − sin (θ ) .sin (φm ) )

Eq. A2
= Cr + χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm )

The error between the theoretical clearance and the measured clearance
distribution is assumed to be given by :

∈m = CH = −Cr − χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm ) Eq A3

Where

CH = the measured radial clearance at angle θ

Applying the method of least squares fit :

∈r = ∑ ( CH = −Cr − χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm ) )

2
Eq A4

∂ ∈r ∂ ∈r ∂ ∈r
Where = = = 0 to give the least error
∂Cr ∂ ( χ r .cos (φm ) ) ∂ ( χ r .sin (φm ) )

Therefore integrating with respect to Cr , χ r .cos (φm ) , χ r .sin (φm ) gives :

∂ ∈r
= ∑ 2. ( CH = −Cr − χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm ) ) . ( −1) = 0 Eq. A5
∂Cr

341
Appendices – Appendix A

∂ ∈r
= ∑ 2. ( CH = −Cr − χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm ) ) . ( − cos (θ ) ) = 0
∂ ( χ r .cos (φm ) )
Eq. A6

∂ ∈r
= ∑ 2. ( CH = −Cr − χ r .cos (θ ) .cos (φm ) − χ r .sin (θ ) .sin (φm ) ) . ( − sin (θ ) ) = 0
∂ ( χ r .sin (φm ) )

Eq. A7

Rearranging from 1

∑ ( −C ) + ∑ C + ∑ χ .cos (φ ) .cos (θ ) + ∑ χ .sin (φ ) .sin (θ ) = 0

H r r m r m
Eq. A8
∑ ( −C ) + n.C + χ .cos (φ ) .∑ cos (θ ) + χ .sin (φ ) .∑ sin (θ ) = 0
H r r m r m

Rearranging from 2

−∑ CH .cos (θ ) + ∑ Cr .cos (θ ) + ∑ χ r .cos (φm ) .cos (θ ) + ∑ χ r .sin (φm ) .sin (θ ).cos (θ ) = 0

2

−∑ CH .cos (θ ) + Cr .∑ cos (θ ) + χ r .cos (φm ) .∑ cos (θ ) + χ r .sin (φm ) .∑ sin (θ ).cos (θ ) = 0

2

Eq A9

Rearranging from 3

−∑ CH .cos (θ ) + ∑ Cr .sin (θ ) + ∑ χ r .cos (φm ) .sin (θ ) .cos (θ ) + ∑ χ r .sin (φm ) .sin 2 (θ ) = 0

−∑ CH .sin (θ ) + Cr .∑ sin (θ ) + χ r .cos (φm ) .∑ sin (θ ) .cos (θ ) + χ r .sin (φm ) .∑ sin 2 (θ ) = 0

Eq. A10

Putting the above equations into matrix from



n ∑ cos (θ ) ∑ sin (θ )  
 
Cr  
 
∑ CH 
 ∑ cos (θ ) ∑ cos (θ ) ∑ sin (θ ) .cos (θ ) .  χ r .cos (θ ) = ∑ CH .cos (θ ) Eq. A11
2

 ∑ sin (θ ) ∑ sin (θ ) cos (θ ) ∑ sin 2 (θ )   χ r .sin (θ )   ∑ CH .sin (θ ) 



Where n = number of data points

342
Appendices – Appendix A

From xx the equations can now be solved for Cr , χ r .cos (φm ) , χ r .sin (φm ) and
χ r .sin (φm ) from which χ r and φm can be given by A12 andA13.

( 2
χ r = ( χ r .cos (φm ) ) + ( χ r .sin (φm ) ) )
2 2
Eq. A12

 χ r .sin (φm ) 
φm = tan −1   Eq. A13
 χ r .cos ( φ m ) 

χr
Where pad pre-load = m =
Cr

343
Appendices – Appendix B

Appendix B
Test Rig Production Drawings

344
Appendices – Appendix B

345
Appendices – Appendix B

346
Appendices – Appendix B

347
Appendices – Appendix B

348
Appendices – Appendix C

Appendix C

1
Derivation of ∫ (1 + ε .cos (φ ) ) 2
.dφ

The information in this appendix follows through the integration of the film function
term from the oil film energy equation derived in chapter 8 of this thesis.

The problem to be solved is the integration of the term given by C1. The
arrangement does not allow for the application of a standard integral so a more
involved integration process is required. The approach taken here is to manipulate
the fundamental term by means of trigonomic substitutions using partial fractions to
provide a series of steps which allow the terms to be solved using standard integrals.

1
∫ (1 + ε .cos (φ ) ) 2
.dφ Eq C1

Let us first consider the trigonomic substitution :

φ 
s = tan   Eq C2
2

ds =
sec 2 φ( 2 ) .dφ Eq C3
2
Then we have :
2
dφ = 2
.ds Eq C4
s +1
1 − s2
cos (φ ) = Eq C5
s2 + 1
2s
sin (φ ) = 2 Eq C6
s +1

Substituting into C1 :
1 1
∫ (1 + ε .cos (φ ) ) 2
.dφ = 2.∫
 ε . (1 − s 2 ) 
2
.ds

( s + 1) . 1 + s 2 + 1 
2

 
s2 +1
= 2.∫ .ds Eq. C7
( −s 2 − 1 + ε .s 2 − ε )

349
Appendices – Appendix C

Breaking out C7 into partial fractions :

 
 1 2.ε 
= 2.∫ − .ds
 (1 − ε ) . ( s 2 + 1 − ε .s 2 + ε ) (1 − ε ) . ( s 2 + 1 − ε .s 2 + ε )2 
 

Taking constants outside integral and integrating term by term :

4.ε 1 2 1
= .∫ 2
.ds − .∫ 2 .ds
ε − 1 ( s 2 − ε .s 2 + ε + 1) ε − 1 s − ε .s 2 + ε + 1

Grouping squared terms of s :

4.ε 1 2 1
= .∫ .ds − .∫ .ds
ε − 1 ( (1 − ε ) .s 2 + ε + 1) 2
ε − 1 (1 − ε ) .s 2 + ε + 1

1
First considering the LHS integral term 2
:
( (1 − ε ) .s 2
+ ε + 1)
Substituting with t :
Where
1 − ε .s
t= .ds Eq. C8
1+ ε
And
1− ε
dt = .ds Eq. C9
1+ ε

Substituting, with some work, gives :

4.ε . 1 + ε 1 2 1
= .∫ .dt − .∫ .ds Eq. C10
1 − ε . ( ε − 1) . 1 + ε
2
(t 2
+ 1)
2
ε − 1 (1 − ε ) .s 2 + ε + 1

1
Now considering the integral term 2
:
(t 2
+ 1)
Substituting with w :
w = tan −1 ( t ) Eq. C11
and
1
dw = 2
.dt Eq. C12
t +1

350
Appendices – Appendix C

Where
1
cos 2 ( tan −1 ( t ) ) = 2
t +1
and
1
2
= cos 2 ( w ) .dw
(t 2
+ 1)

Substituting into C10 :

4.ε . 1 + ε 2 1
= .∫ cos 2 ( w ) .dw − .∫ .ds Eq. C13
1 − ε . ( ε − 1) . 1 + ε
2
ε − 1 (1 − ε ) .s 2 + ε + 1

w sin ( 2.w )
Recognising standard integral of cos 2 ( w ) = + yields :
2 4

2.ε . 1 + ε .w ε . 1 + ε .sin ( 2.w ) 2 1

= + − .∫ .ds Eq. C14
1 − ε . ( ε − 1) . 1 + ε
2
1 − ε . ( ε − 1) . 1 + ε
2
ε − 1 (1 − ε ) .s 2 + ε + 1

1
Now considering the RHS integral term :
(1 − ε ) .s 2 + ε + 1

Using trigonomic substitution

1 − e .s
t= Eq. C15
1+ e
and
1− e
dt = .ds Eq. C16
1+ e
Substituting into C14
2.ε . 1 + ε .w ε . 1 + ε .sin ( 2.w ) 2. 1 + ε 1
= 2
+ 2
− .∫ 2 .dt Eq. C16
1 − ε . ( ε − 1) . 1 + ε 1 − ε . ( ε − 1) . 1 + ε ( ε − 1) . 1 − ε . 1 + ε t + 1

1
Where using the standard integral of ∫t 2
+1
.dt = tan −1 ( t ) + C

351
Appendices – Appendix C

We complete the integration to give :

2.ε . 1 + ε .w ε . 1 + ε .sin ( 2.w ) 2. 1 + ε .tan −1 ( t )

= 2
+ 2
− +C Eq. C17
1 − ε . ( ε − 1) . 1 + ε 1 − ε . ( ε − 1) . 1 + ε ( ε − 1) . 1− ε . 1+ ε

Now substituting back for w = tan −1 ( t ) :

2.ε . 1 + ε . tan −1 ( t ) ε . 1 + ε .2.t 2. 1 + ε .tan ( t ) −1

= + − +C
1 − ε . ( ε − 1) . 1 + ε . ( t 2 + 1) ( ε − 1) . 1 − ε . 1 + ε
2 2
1 − ε . ( ε − 1) . 1 + ε

2.t
Where sin ( 2. tan −1 ( t ) ) = 2
t +1

1 − ε .s
Substitute back for t = :
1+ ε

 1 − ε .s   1 − ε .s 
2.ε . 1 + ε .tan −1   2. 1 + ε .tan −1  
 1+ ε  ε .2.s  1+ ε 
= + − +C
2  (1 − ε ) .s 
2
1 − ε . ( ε − 1) . 1 + ε
2
( ε − 1) . 1 − ε . 1 + ε
( ε − 1) . 1 + ε .  + 1
 (1 + ε ) 

φ 
And finally substituting back in for s = tan  
2
 φ    φ  
 1 − ε .tan  2    1 − ε . tan  2  
2.ε . 1 + ε .tan  −1   2. 1 + ε .tan  −1  
 1+ ε  φ   1+ ε 
  ε .2.tan    
=  +  
2
−   +C
1 − ε . ( ε − 1) . 1 + ε
2
 2 φ   ( ε − 1) . 1 − ε . 1 + ε
 (1 − ε ) .tan  2  
2   + 1
( ε − 1) . 1 + ε . 
 (1 + ε ) 
 
 
Eq. C18

352
Appendices – Appendix C

Grouping terms and simplifying, following some work, the general solution can be
given as :

 φ  φ 
 tan    2.ε . tan  
dφ 2. η  2   .  ε − 1 + 2
∫ (1 + ε .cos (φ ) ) 2
=− 2
(ε − 1)
. tan −1 
 
η  1 + ε   1 2 φ  
+C




(ε − 1) . (1 + ε ) .  η .tan  2  + 1
2

 
Eq. C19
1+ ε
Where η =
1− ε

The above equation C19 can now be substituted back into the energy equation to
complete the lubricant film temperature expression.

353