CRANFIELD UNIVERSITY

ELEFTHERIOS ANDREADIS

DESIGN OF A LOW SPEED VANEAXIAL FAN

SCHOOL OF ENGINEERING

MPhil THESIS
 CRANFIELD UNIVERSITY

SCHOOL OF ENGINEERING
DEPARTMENT OF POWER ENGINEERING AND PROPULSION

MPhil THESIS

ACADEMIC YEAR 2010-11

ELEFTHERIOS ANDREADIS

DESIGN OF A LOW SPEED VANEAXIAL FAN

SUPERVISOR: DR JOAO AMARAL TEIXEIRA

OCTOBER 2011

This thesis is submitted to fulfill the requirements for the degree of Master
in Philosophy

©Cranfield University 2011. All rights reserved. No part of this publication

may be reproduced without the written permission of the copyright owner.
 ABSTRACT
The ventilation of industrial areas and tunnels is a safety requirement and
characterizes the quality of a working environment. Low speed fans are used to
achieve the required ventilation level. An attempt to design a low speed
vaneaxial fan, meeting the specifications of a given fan that is already in the
market takes place in this Thesis. The project was conducted with the support of
the Fläkt-Woods Company and the main target is to design a ventilation fan,
meeting the requirements for pressure rise, volume flow and size, of an existing
model. The efficiency improvement is driven in part by the new national and
international legislation concerning the operation of electrical equipment.
Companies require higher efficiencies without compromising safety features of
the fan and the fan capability to operate at high temperatures.

A low speed fan design procedure is established based on the available

literature and design tools. The free vortex approach is employed, which provides
acceptable efficiency and relatively simpler design. The design procedure can be
used to design a fan given a set of customer requirements. Many software tools
are used to design the fan. A Matlab code for the blade design is developed and
other codes are used to establish the final fan design. The effectiveness of the
design procedure is verified with CFD simulations carried out as part of this
project. Three new designs that are developed with the established design
procedure are presented in this Thesis. The new designs differ in the hub to tip
ratio, the rotational speed and the number of the blades and the vanes. The
experience acquired from the analysis of the performance of the first new design
is used to improve the performance of the following designs in order to achieve
the best efficiency possible. The effect of tip clearance is investigated thoroughly
in the new designs because the tip clearance has a major impact on the fan
performance and safe operation of the fan at high temperatures. The mechanical
integrity of the fan is examined last to verify that the fan can operate in high
temperature.

The target of improved efficiency (higher than 79%) is achieved in one of

the fan designs attempted and it was calculated 82%. The off design
performance of the new fan is satisfactory as well. This new design can be
further optimized, since the modification of minor design features is in itself a
methodology that can incrementally improve the efficiency of a low speed fan.
The new fan can operate at high temperatures (400°C), however the safety factor
at this temperature is 1.25 for combined steady mechanical and thermal loading
and it can be further improved either through the use of materials with better
resistance in thermal loading or with an increased tip clearance.

i 
 
 ACKNOWLEDGEMENTS

I would like to express my gratitude to Dr Joao Amaral Teixeira for his

supervision, guidance and assistance throughout the preparation of this study.

I am particularly grateful to Professor Pericles Pilidis for his support during

this project.

I would also like to thank Mr Iain Kinghorn, Technical Manager of Fläkt-

Woods Company, Colchester, UK for the initiation of this project and for the
useful information that he provided.

Last but not least, I would like to thank from my heart my beloved wife
Lillian, who inspired and motivated me to successfully complete this MPhil.

ii 
 
 LIST OF CONTENTS
LIST OF FIGURES……………………………………………………………….. (vi)
LIST OF TABLES………………………………………………………………… (x)
SYMBOLS………………………………………………………………………… (xi)

1 INTRODUCTION…………………………………………… 1

2 Literature Review………………………………………….
2.1 Definitions………………..……………………………………………. 4
2.2 Fan Laws...…………………………………………………………….. 13
2.3 Flow Coefficient Φ and Work Coefficient Ψ……………………….. 18
2.4 Axial Flow Fans……………………………………………………….. 21

3 Preliminary Fan Design Method…………………..……

3.1 Introduction…………………………………...………………………… 30
3.2 Method Selection…….………………………………………………… 30
3.3. Radial Equilibrium……...………………………………………………. 32
3.3.1 Free Vortex Design…...……………………………………………….. 33
3.3.2 Non Free Vortex Design………………………………………………. 34
3.4 Hub to tip Ratio Definition……………………………………………… 35
3.5 Airfoil Selection…..…………………………………………………….. 37
3.6 Blade Geometry………..………………………………………………. 38
3.7 Blade Design….…..……………………………………………………. 39
3.8 Number of blades………………………………………………………. 43
3.9 Number of Guide Vanes…..…………………………………………… 43
3.10 Influence of Hub to Tip Ratio.…………………………………………. 44
3.11 Influence of Blade Angle……...……………………………………….. 46
3.12 Influence of Tip Clearance…..………………………………………… 46
3.13 Three Dimensional Effects .…………………………………………… 47
3.14 Losses………………..……...…………………………………………… 48

4 Fan Design Tools…………………………………………….

4.1 Introduction………………………………………………………………. 50
4.2 MATLAB codes………………………………………………………….. 50
4.2.1 MATLAB code for Osborne method…………………………………… 51
4.2.2 MATLAB code for Blade Geometry……...……………………………. 53
4.3 Lewis Software…………………………………………………………... 54

iii 
 
 4.4 ANSYS Turbogrid…..…………………………………………………… 57
4.5 GAMBIT………….……….……………………………………………… 57
4.6 ANSYS CFX-Pre……………………………………………………….. 58
4.7 ANSYS CFX-Solver…...……………………………………………….. 59
4.8 ANSYS CFX-Post…...………………………………………………….. 60
4.9 Other MATLAB codes ..……………………………………………….. 62
4.10 ANSYS Mechanical...………………………………………………….. 65

5 McKenzie’s Method Validation.……………………….

5.1 Introduction………………………………………………………………. 66
5.2 McKenzie's example…...………………………………………………. 66
5.3 Geometry creation……………………………………….……..………. 67
5.4 Meshing…………………..……………………………………………… 70
5.5 Physics and Fluid properties - Boundary Conditions……………….. 71
5.6 Solver……..………………………………………………………………. 73
5.7 Results Report ………………………………………………………….. 73
5.8 Validation-Verification of the results………………………………….. 76
5.9 Results evaluation………………………………………………………. 78
5.10 Conclusion……………………………………………………………….. 79

6 New Designs………..……………………………………….
6.1 Introduction……………………………………………………………… 80
6.2 Fläkt-Woods fan………………………………………………………… 81
6.3 New Design……………….…………………………………………….. 83
6.4 Design1………………………………………………………………….. 84
6.4.1 Blade Geometry…………………………………………………………. 84
6.4.2 Blade Design...………………………………………………………….. 88
6.4.3 Meshing…..……………………………………………………................ 89
6.4.4 CFX-Pre………………………..…………………………………………. 91
6.4.5 Results……………………….…………………………………………... 92
6.4.6 Design1 Optimization………………………..………………………….. 98
6.4.7 Validation-Verification of the results.…………………………………. 100
6.4.8 Off-design performance of Improved Design1 …………………...... 101
6.5 Design2…………………………………………………………………… 103
6.5.1 Design2 Concept….……………………………………………………… 103
6.5.2 Blade Geometry-Design-Meshing- Optimization…………………….. 104
6.5.3 Results..……………………………………………………...................... 105
6.5.4 Turbulence model comparison…………………………………………. 108
6.5.5 Verification………………….…………………………………………....... 108

iv 
 
 6.5.6 Off-design performance……………………..…………………………. 108
6.6 Design3………………………………………………………………….. 110
6.6.1 Design2 Concept….……………………………………………………. 110
6.6.2 Blade Geometry-Design-Meshing- Optimization……………………. 111
6.6.3 Results..…………………………………………………….................... 112
6.6.4 Verification………………….…………………………………………..... 114
6.6.5 Off-design performance……………………..…………………………. 114
6.7 Tip Clearance Effect..……………………..……………………………. 116

7 Mechanical Integrity…………………………………………
7.1 Introduction……………………………………………………………… 121
7.2 Blade Stressing………………………………………………...………… 121
7.3 Centrifugal Stresses...…………………………………………………… 123
7.4 Fluid Forces………………………………………………………………. 124
7.5 Stresses due to Thermal Gradients.…………………………………… 126
7.6 Blade loading simulation software……...……………………………… 127
7.7 Blade loading simulations………………………………………………. 128
7.8 Creep life calculation……………….……………………………………. 135
7.9 Conclusions………………………….…...………………………………. 138

8 Conclusions and Recommendations…………………….

8.1 Project Summary………………………………………………………… 140
8.2 Conclusions………………………………………………………………. 140
8.3 Recommendation for further work……………………………………… 141

REFERENCES……………………………………………………………………........ 142

APPENDIX "A" Design2 Data..……………………………………………………... A-1

APPENDIX "B" Design3 Data………………………...…………………………….. B-1

APPENDIX "C" Tip Clearance 3%…………………………………………….……. C-1

v 
 
 LIST OF FIGURES

Fig 2.1 Vaneaxial Fan with outlet vanes around the motor and mouth bell 4
inlet (Bleier,1998)………………………………………………………
Fig 2.2 Fan stage Velocity triangles (Ramsden, 2008)……………………. 5
Fig 2.3 Relative Velocities – Angles in a blade (McKenzie,1997)………… 6
Fig 2.4 Deflection (Ramsden, 2008)…………………………………………. 6
Fig 2.5 Chord, Pitch, Stagger angle (McKenzie, 1997)……………………. 7
Fig 2.6 Stagger angle (Angoy, 1984)………………………………………… 8
Fig 2.7 Deflection vs Outer angle (Lewis, 1996)……………………………. 8
Fig 2.8 Inlet-Outer angles, Camber, Stagger angle (McKenzie, 1997)…… 9
Fig 2.9 Angle of Incidence i (Ramsden, 2008)……………………………… 10
Fig 2.10 Deviation δ (Lewis, 1996)…………………………………………….. 10
Fig 2.11 Performance characteristics of a fan (Osborne, 1977)……………. 12
Fig 2.12 Same Fans with different rotational speeds (Bleier,1998)…………. 14
Fig 2.13 Similar Fans with different sizes (tip diameters 27’’ and 40.5’’) 15
(Bleier)…………………………………………………………………….
Fig 2.14 Similar Fans with different tip diameters and rotational speed 17
(Bleier,1998)……………………………………………………………..
Fig 2.15 Values for φ and ψ from actual fans (Osborne, 1977)……………… 19
Fig 2.16 Values for φ and ψ from actual fans (Bleier, 1998)…………………. 20
Fig 2.17 Smith Charts for (a) 50% reaction (b) 70% reaction (Lewis, 1996).. 20
Fig 2.18 Smith Charts for (c) 90% reaction (Lewis, 1996)…………………… 21
Fig 2.19 Tunnel Ventilation (www.roadtraffic-technology.com)…………....... 22
Fig 2.20 Electronics Cooling (www.qmed.com) ……………………………….. 23
Fig 2.21 Characteristic line of a low speed fan (Strohmeyer, 2009) ………… 24
Fig 2.22 Propeller fan (www.canadablower.com)……………………………... 24
Fig 2.23 Tubeaxial fan (www.cincinnatifan.com)………………………………. 25
Fig 2.24 Vaneaxial fan (www.flaktwoods.com)………………………………….. 25
Fig 2.25 Vaneaxial fan with inlet vanes (Bleier, 1998)………………………… 26
Fig 2.26 Vaneaxial fan with outlet vanes (Bleier, 1998) ……………………… 27
Fig 2.27 Various type axial fan performance (Osborne, 1977)……………….. 28
Fig 2.28 Performance curve of a vaneaxial fan (Bleier, 1998)………………… 28

Fig 3.1 Lift ratio, k in variation with space to chord ratio (Dixon,1998)…… 32
Fig 3.2 (a) free vortex axial fan (b) non free vortex axial fan (Lewis,1996)... 34
Fig 3.3 Hub Diameter calculation (Bleier, 1998)………………………………. 36
Fig 3.4 Work Coefficient vs flow coefficient diagram (McKenzie,1988)…….. 37

vi 
 
 Fig 3.5 NACA 65010 airfoil coordinates [http://www.mh- 38
aerotools.de/airfoils/javafoil.htm]………………………………………
Fig 3.6 Blade geometry for fixed air angles (McKenzie, 1997) …………… 40
Fig 3.7 Efficiency contours (McKenzie, 1997)……………………………….. 40
Fig 3.8 Lift coefficient variation with angle of attack for a 4-digit NACA airfoil 42
(Bleier, 1998) ………………………………………………………….
Fig 3.9 Volume Comparison of 29in vaneaxial fan (5 blades, 11 vanes) with 45
hub to tip ratios 52% and 68% [Bleier, 1997]……………………….
Fig 3.10 Influence of Blade angle (36in, 1750 rpm, 13-33 angles) (Bleier, 46
1998)…………………………………………………………………….
Fig 3.11 Summary of the influence of the tip clearance, (Bleier, 1998)…… 47
Fig 3.12 Effect of tip clearance on the characteristics of the blade tip section, 48
(Bass, 1987)………………………………………………………….
Fig 3.13 Loss coefficient variation with incidence (McKenzie, 1997)……. 48

Fig 4.1 Comparison of work and flow coefficient of the new design with the 51
existing designs………………………………………………………..
Fig 4.2 Deflection in variation with the Outer flow angle for the hub 52
rotor…………………………………………………………………….
Fig 4.3 3D Plot of blade sections (output from MATLAB code)…………. 53
Fig 4.4 3D plot of blade sections (output from MATLAB code)…………. 54
Fig 4.5 GUI of program STACK…………………………………………….. 55
Fig 4.6 2D display of the 4 blade profiles (STACK program)……………… 55
Fig 4.7 Stator vane in ANSYS Turbogrid………..…………………………… 57
Fig 4.8 Induct domain in GAMBIT…………………………………………….. 58
Fig 4.9 Flow domains in ANSYS CFX-Pre…………………………………… 59
Fig 4.10 GUI of ANSYS CFX-Solver……………………………………………. 59
Fig 4.11 Blade to Blade view of velocity contour……………………………. 60
Fig 4.12 Pressure distribution across the blade (pressure/suction side)….. 61
Fig 4.13 Variation of pitch to chord ratio across the radii……………………. 62
Fig 4.14 Variation of relative velocity angles across the radii……………….. 63
Fig 4.15 Variation of stage reaction across the radii…………………………. 63
Fig 4.16 Variation of work and flow coefficient across the radii………………. 64
Fig 4.17 Blade deformation for normal operating conditions for aluminum 65
6061-T6 …………………………………………………………………

Fig 5.1 2D display of the 4 blade profiles (STACK program)……………… 68

Fig 5.2 3D plot of the blade using MATLAB code………………………….. 69
Fig 5.3 Representation of the blade, the hub and the shroud in Turbogrid.. 70
Fig 5.4 Representation of the stator vane mesh…………………………….. 71
Fig 5.5 Simulation domain (rotor and stator)………………………………… 72

vii 
 
 Fig 5.6 Convergence plot……………………………………………………… 73
Fig 5.7 Velocity contours at 0.1 of the span…………………………………. 74
Fig 5.8 Velocity contours at 0.5 of the span…………………………………. 75
Fig 5.9 Velocity contours at 0.9 of the span………………………………….. 76
Fig 5.10 Velocity Vector Diagram at 0,1 span…………………………………… 77

Fig 6.1 Tubeaxial vs vaneaxial efficiency in variation with volume flow 81

(Flaktwoods)……………………………………………………………
Fig 6.2 Tubeaxial vs vaneaxial total pressure rise in variation with volume 82
flow (Flaktwoods)………………………………………………………..
Fig 6.3 Work Coefficient vs flow coefficient diagram (McKenzie,1997)….. 84
Fig 6.4 Relative Velocities – Angles (McKenzie,1997)…………………….. 85
Fig 6.5 Blade geometry for fixed air angles (McKenzie, 1997) …………. 86
Fig 6.6 Efficiency contours (McKenzie, 1987)……………………………….. 87
Fig 6.7 Blade and Vane of Design1 in ANSYS Turbogrid…………………. 90
Fig 6.8 Induct domain from GAMBIT ………………………………………… 90
Fig 6.9 Outduct domain from ANSYS Meshing …………………………….. 91
Fig 6.10 Simulation Model in CFX-Pre ……………………………………….. 92
Fig 6.11 Rotor-Stator Velocity Contour Plot at 0.1 span……………………… 93
Fig 6.12 Rotor-Stator Velocity Contour Plot at 0.5 span……………………… 94
Fig 6.13 Rotor-Stator Velocity Contour Plot at 0.9 span……………………… 94
Fig 6.14 Rotor-Stator Velocity Vector Plot at 0.9 span……………………….. 95
Fig 6.15 Rotor-Stator Velocity Contour Plot at 0.8 span…………………….. 96
Fig 6.16 Rotor-Stator Velocity Contour Plot at 0.7 span…………………….. 96
Fig 6.17 Rotor-Stator Velocity Contour Plot at 0.6 span……………………. 97
Fig 6.18 Rotor-Stator Velocity Contour Plot at 0.6 and 0.7 span (optimized) 99
Fig 6.19 Rotor-Stator Velocity Contour Plot at 0.8 and 0.9 span (optimized) 99
Fig 6.20 Volume flow in variation with total pressure rise…………………….. 102
Fig 6.21 Volume flow in variation with efficiency………………………………. 102
Fig 6.22 Rotor-Stator Velocity Contour Plot at 0.1 span……………………… 105
Fig 6.23 Rotor-Stator Velocity Contour Plot at 0.5 span……………………… 106
Fig 6.24 Rotor-Stator Velocity Contour Plot at 0.9 span……………………… 106
Fig 6.25 Stator Velocity Vector Plot at 0.1 span………………………………. 107
Fig 6.26 Stator Velocity Vector Plot at 0.5 span………………………………. 107
Fig 6.27 Stator Velocity Vector Plot at 0.9 span……………………………… 107
Fig 6.28 Volume flow in variation with total pressure rise…………………… 109
Fig 6.29 Volume flow in variation with total efficiency………………………… 109
Fig 6.30 Rotor-Stator Velocity Contour Plot at 0.1 span...…………………… 113
Fig 6.31 Rotor-Stator Velocity Contour Plot at 0.5 span……………………… 113

viii 
 
 Fig 6.32 Rotor-Stator Velocity Contour Plot at 0.9 span……………………… 114
Fig 6.33 Volume flow in variation with total pressure rise…………………….. 115
Fig 6.34 Volume flow in variation with total efficiency………………………… 116
Fig 6.35 Volume flow in variation with static pressure rise……………………. 117
Fig 6.36 Effect of tip clearance in static pressure rise and brake horse power 118
(Bleier, 1997)……………………………………………………………...
Fig 6.37 Volume flow in variation with total efficiency 119
Fig 6.38 Effect of tip clearance in efficiency and noise level (Bleier, 1997).. 120

Fig 7.1 Pressure and stress distribution against volume flow (Cory, 2005) 122
Fig 7.2 Centrifugal loading (Haslam, 2007)…………………………………… 123
Fig 7.3 Bending Moments from gas forces (Haslam, 2007)…………………. 125
Fig 7.4 Bending Moments from centrifugal loading (Haslam, 2007)………… 126
Fig 7.5 Strength and Young modulus variation with temperature for typical 127
carbon steel (Cory, 2005)……………………………………………..
Fig 7.6 Blade deformation for normal operating conditions for aluminum 129
6061-T6 ………………………………………………………………….
Fig 7.7 Stress distribution for normal operating conditions for aluminum 129
6061-T6 …………………………………………………………………
Fig 7.8 Stress distribution for normal operating conditions for aluminum 130
6061-T6 ……………………………………………………………….
Fig 7.9 Blade deformation for operation in 50o C for aluminum 6061-T6 131
Fig 7.10 Blade deformation for operation in 100o C for aluminum 6061-T6 132
Fig 7.11 Blade deformation for operation in 200o C for aluminum 6061-T6 132
Fig 7.12 Blade deformation for operation in 300o C for aluminum 6061-T6 133
Fig 7.13 Blade deformation for operation in 400o C for aluminum 6061-T6 133
Fig 7.14 Blade deformation variation with temperature under constant 134
mechanical loading for aluminum 6061-T6 ………………………
Fig 7.15 Three phases of creep (Andreadis, 2009)……………………….. 136
Fig 7.16 Larson Miller parameter for aluminum 6061-T651 for C-=20.3 137
(Kaufman, 2008)………………………………………………………

ix 
 
 LIST OF TABLES

Table 5.1 Mesh Dependency Study……..……………………………………… 77

Table 6.1 Final rotor blade design (Design1)…………………………………… 87

Table 6.2 Final stator vane design (Design1)…………………………………. 88
Table 6.3 Design1 Rotor………………………………………………………….. 89
Table 6.4 Design1 Stator…………………………………………………………. 89
Table 6.5 Optimized Stator……………………………………………………….. 98
Table 6.6 Improved Design Performance……………………………………… 99
Table 6.7 Results Verification…………………………………………………… 100
Table 6.8 Off design cases……………………………………………………….. 101
Table 6.9 Design point performance fοr k-ε and k-ω turbulence models…….. 108
Table6.10 Design point performance with tip clearance 1.5% and 3.0%......... 117

Table 7.1 Deformation variation with temperature……………………………… 134

Table 7.2 Deformation variation with temperature………………………………. 135

Table 8.1 Summary of 3 new designs…………………………………………….. 140

x 
 
 SYMBOLS

Air Angle α
Angle Velocity ω
Average Velocity W
Blade Angle β
Blade Height h
Blade Width L
Brake Horsepower BHP
Camber (angle) θ
Chord C
Cross Sectional Area A
Deflection (angle) ε
Density ρ
Deviation (angle) δ
Efficiency η
Enthalpy H
Flow Coefficient φ
Hub Diameter d
Incidence (angle) i
Lift Coefficient CL
Mass Flow M
Noise N
Number of Blades/Vanes z
Power Consumption PC
Pressure Coefficient Cp
Radius r
Relative Velocity V
Rotational Velocity U
Solidity σ
Space S
Stagger angle ζ
Static Pressure Ps
Static Pressure p
Temperature T
Time t
Tip Diameter D
Total Pressure Pt
Volume Flow Q
Water Column (pressure unit) WC
Whirl Velocity Uw
Work Coefficient ψ
 

xi 
 
 Introduction

1. INTRODUCTION
The role that ventilation fans play in modern societies is well
understood by the public at large from the heat removal in personal
computers to the large ventilation fans visible in road tunnels. Historically
these have been the poor relation of turbomachines, basic in design, cheap to
acquire and often with a poor efficiency. However an added environmental
consciousness and the passing of national and international legislation are
changing this landscape. The use of low efficiency designs is likely to attract
penalties in the future and this will shift the way consumers look at the
efficiency of ventilation fans and hence the premium they are prepared to pay
for efficient machines.

For companies and designer this shift represents commercial

opportunities together with technical challenges. Both can be tackled through
additional research and the use of new technologies many already in use in
the propulsion end of turbomachinery design practice. The drivers for design
updating are not only in direct answer to the concerns about efficiency but
also due to the fact that in recent years the ventilation of industrial areas and
tunnels have become more stringent as safety requirements and the quality of
working environments are improved.

There are various types of fans that can meet the requirements for
every use. The main categories are the axial flow fans, which are commonly
called “fans” and the centrifugal fans which are called “blowers”. The axial
flow fans, which are the focus of this project are categorized as propeller fans
(a propeller with a motor), tubeaxial fans (a propeller fan with a cylindrical
housing) and vaneaxial fans (tubeaxial fans with a vane stage upstream or
downstream).

The target of this project is to review and illustrate the procedures that
can be employed to design axial flow industrial fans. To this end an existing
vaneaxial fan is taken as the datum configuration and a design procedure to
obtain an updated design with better efficiency is employed and described.
The starting point for the new design is a set of specifications that matches
the datum fan features (size, rotational speed) and its performance.

In some respects this work is expected to have a didactic dimension.

The author has experience in the performance of turbofan engines but the
study of low speed ventilation fans required the adoption of new terminology
and the appreciation of factors specific to these machines, from the flow
conditions marked by low Reynolds numbers, and thus inherently lossier than

1 | P a g e  

 
 Introduction

its higher speed counterparts, to the economic realities that determine the
design space.

For this reason the reader will find in Chapter 2 of this thesis a
thorough explanation of fan terminology along with other aspects of the
literature review.

The design of a fan starts from the operational requirements. It is very

important for the designer to know the working environment of a fan and the
specifications that the customer requires. The dimensions, the motor, the
volume flow, the pressure rise are the basic characteristics should be defined
at the preliminary stage of the design. In this project these features are
defined, because the given fan is a machine employed in the field rather than
a theoretical model.

Taking all the above into consideration, the next step is the
presentation of the available design procedures which is covered in Chapter
3. There are plenty of examples in the available literature relative to the
design procedure of a fan (Bleier 1998, Osborne 1977, McKenzie 1997, Lewis
1996) and an effort was made to implement features for every author in order
to achieve the best result possible.

After establishing the design procedure, the design tools are presented
in Chapter 4. The tools are used to create the geometry, generate the
computational mesh and simulate the cases. A Matlab code was created to
contribute to the design procedure. Other codes that were found in the
literature were used for the generation of the blade geometry. CFD software is
used to assemble the meshes, set and solve the cases and post process the
results. Finite element software is used for the investigation of the mechanical
integrity of the fan.

In Chapter 5 the investigation of the design procedure takes place

using CFD analysis. The concept of the design method investigation using
CFD is simple. The blade and geometry features which were the outcome of
the selected design procedure are used to create the geometry of the fan for
the simulation. After creating the geometry, CFD is used to solve the flow field
and acquire the performance of the fan that came from the simulation. The
final step is to investigate if the simulated performance of the fan meets the
requirements that were used for the preliminary design of the fan.

The main task in this project is the simulation of the performance of the
existing fan and for the new design fan, as well. The use of the computational
fluid dynamics software is the key point for this stage. In Chapter 6 the
simulations that were carried out are presented. The results of design point
2 | P a g e  

 
 Introduction

calculation and off design cases are analyzed for 3 different design cases.
The tip clearance effect is simulated and presented as well.

In Chapter 7 the mechanical integrity of the fan is examined. The effect

of the combined thermal and mechanical loads to the impeller is presented.
The mechanical integrity is investigated in order to verify that the fan can
operate in high temperature conditions.

The conclusions of this project and the recommendations for further

work are presented in Chapter 8. Designing a fan only from the customer
requirements is a compromise between many factors. The experience
acquired from the 3 new designs and the interesting opportunities for further
work are then presented.

3 | P a g e  

 
 Lite
erature Rev
view

2. Literature
e revie
ew

2.1 Definitions
The description of the flow in a low speed d fan and d the various
definitions forr the flow characteristics repre esent an essential
e p
part of the fan
dessign underrstanding. The fan design ha as its ow wn terminoology and the
clarrification of the terms is very important for
f the ressearcher. TThis sectio
on is
deddicated to the
t presen ntation and
d definition of the term
ms that ch
haracterize
e the
geoometry of axial
a flow fa
ans. A vaneaxial fan is shown in Figure 22.1.

Figu
ure 2.1 Vaneaxial Fan witth outlet vanes around th
he motor and
d bellmouth in
nlet (Bleier,1998)

The fllow in a fan is always th hree dime ensional, however the

und
derstandingg of the tw
wo dimensioonal flow in
n a fan is the
t first ste
ep towards
s the
fan design. The
T basic terminoloogy applies to the 2D 2 flow, so it is more
m
con
nvenient to present thhe fan term
minology with 2D reprresentationns.

The flo
ow in a faan is visua h 2D veloccity triangles which are
alized with
illusstrated in Figure 2.2
2 and theyy depict th
he velocityy vectors in
n a fan sttage

4|Page
 Literature Review

(rotor-stator) and the relative velocity angles. From the 2D Figure 2.2 many
terms can be defined.

Figure 2.2 Fan stage Velocity triangles (Ramsden, 2008)

The relative velocity V is the velocity of the flow which enters the fan
rotor with an axial velocity Va , relative to the moving blade which has a
rotational velocity U.

The relative velocity at the leading edge of the rotor blade is V1 and at
the trailing edge is V2. The relative velocity at the trailing edge V2 is different
from V1 because of the deflection of the flow through the blade passage. The
absolute velocity at the exit of the rotor blade is V3 which it is the vectorial sum
of the V2 and U vectors. V4 is the velocity downstream of the stator at the exit
of the fan. The relative velocities angles are α1 and α2. The angles α0, α3 and
α4 are absolute velocities angles at inlet, downstream of the rotor and
downstream of the stator respectively.

5|Page
 Lite
erature Rev
view

The whirl velocity Uw or Vw is the tangential velocity

v that is added to
flow
w by the action of th he rotor. To ogether wiith the axiaal velocity it defines the
roto
or exit abssolute velo ocity V3. The
T whirl velocity Vw is a me easure of the
eneergy that thhe rotor bla
ade adds to o the flow. Figure 2.33 presents an alternaative
deppiction of th
he triangless of velocitty in an axial fan stag
ge.

Fig
gure 2.3 Rela
ative Velocities – Angles in a blade (M
McKenzie,19
997)

The de eflection (Figure 2.4) is the ang gular differrence betwween the an ngle
of the relativee velocity at
a the outlet of the bla
ade α2 and the angle of the rela ative
veloocity at the
e inlet of th
he blade α1. There is s a limit to the amoun nt of deflecction
thatt can be achieved byy a blade row, which h is defined d by the V2/V1 ratio also
a
called de Ha aller numb ber. This limit is fix xed by th he inter blade pass sage
diffu
usion. The e point wh here the incoming flow meetts the blade surface e is
dennoted the stagnation
s point. The flow accelerates ove er the sucttion side off the
bladde becausse of the camber of o the pro ofile and consequen
c ntly the sttatic
pressure decreases loccally. In co ontrast the e flow dece elerates in
n the press sure
sidee and the static
s presssure increaases.

F
Figure 2.4 De
eflection (Ramsden, 2008
8)

6|Page
 Lite
erature Rev
view

Space or Pitch is the leng gth of the arc betwe

een two sim
milar pointts in
two
o consecutiive blades (Figure 2.5).

Chord is the disttance betw

ween the le
eading edg
ge with the
e trailing edge
e
of the blade (F
Figure 2.5)).

The Sppace (Pitc ch) to Choord ratio (S/C) has a direct efffect on thee air
ning (or de
turn eflection) and
a therefo
ore on thee pressure rise. Many authors use
the opposite ratio,
r chordd to space
e (pitch) ratio, which is commonly referre ed to
as solidity.
s Cllosely spacced bladess have a laarge degreee of overlap and due to
the influence of the adjacent blad des they ha ave a highher lift and consequeently
highher deflecttions can be
b achieve ed. On the contrary, sparsely sspaced bla ades
will have a low blade e overlap and this lack of channeling
c g implies low
defllection.

e chord to space (pitch) ratio (C

Soliditty (σ) is the C/S).

Figure 2.5 Chord,

C pitch and stagger angle (McKe
enzie, 1997))

Stagge er angle (ζ) is the in

nclination to
t the axia
al direction
n of the ch
hord,
the line joining
g the leading and tra
ailing edge (Figures 2.5,
2 2.6).

7|Page
 Lite
erature Rev
view

F
Figure 2.6 Sta
agger angle (Angoy, 1984)

Figure 2.7 illustrrates the cascade

c performanc
p ce for vario ous deflecction
and d three space to chord ratios,, 0.5, 1.0 and 1.5. The nomin nal deflecttion,
leftmmost figure
e, is the deflection
d at 80% off the stallin
ng deflection (maxim mum
attaainable defflection). For a given outlet ang
gle the defllection deccreases as
s the
pitcch to chord ratio incre
eases (Lew wis, 1996).

Figure 2.7 Deflectio

on vs Outer angle
a (Lewiss, 1996)

8|Page
 Lite
erature Rev
view

To obtain high pressure

p ra
atio, the sta
agger angle should be as high h as
posssible. [MccKenzie 19 997, Osbou urne 1977, Lewis 19 996]. The upper limit on
gger increase is sett by the blade overla
stag ap conside
erations. A
As the staggger
anggle increasses the ove erlap reducces for a given
g space to chordd ratio and this
can
n reduce th he attainab
ble pressurre rise. This can be overcome by increas sing
the chord, bu ut in that ca
ase there is also a limit
l that comes
c from
m the diffus
sion
duee to the inccreased resstrictions in
n the flow passage.
p

High stagger
s angle bladess produce steep fan
n characterristics, tho
ough
thiss is usuallyy undesirab
ble, since small mov
vements in
n the operaating pointt will
produce significant cha anges in teerms of the pressure
e rise and
d volume flow.
f
Fro
om this point a lower rotor blade stagger, leading to acteristics that
o fan chara
are flatter , woould be desirable.

The ou utflow angle dependds to a fairrly small extent

e on tthe cambe
er θ,
whe
ereas it is the
t stagge er angle ζ which
w really
y has a big
g effect.

Inlet angle β1 is the angle of the bla ade at the leading

l ed
dge and ou
utlet
anggle β2 is th
he blade an ngle at the e trailing ed
dge (Figuree 2.8). The
e blade ang
gles
are related too the air angles
a thro
ough the in ncidence at
a the lead ding edge and
with
h deviationn at the traiiling edge.

Figure 2.8 Inlet-outtlet angles, camber

c and stagger
s angle
e (McKenzie
e, 1997)

Cambe er (θ) is th
he differencce betwee en the outeer angle β2 and the inlet
ang
gle β1 and is related to
t the defle
ection of th
he flow (Fig
gure 2.8).

ngle of Inciidence i, Figure

The an F 2.9, is the diffe
erence bettween the inlet
rela
ative velociity angle and the blad
de angle at
a the leading.

i = a1-β1 (2.1)

9|Page
 Literature Review

The importance of the incidence angle is great, because the smooth

distribution of the flow at the pressure and at the suction sides of the blade
depend on the incidence

Figure 2.9 Angle of Incidence i (Ramsden, 2008)

The fluid Deviation δ is defined as the difference between the outlet

flow angle and the angle of the blade metal at outflow, Figure 2.10. Deviation
is related to the outlet of the blade in a manner similar to the incidence in
relation to the inlet of the blade.δ = a2-β2 (2.2)

Figure 2.10 Deviation δ (Lewis, 1996)

10 | P a g e
 Literature Review

Aspect ratio is the ratio of the mean blade span to the mean blade
chord.

Hub to tip ratio is the ratio of the diameter of the hub to that of the tip
of the blade. Because the location of the tip is sometimes ambiguous (the
stator tip may be on the hub) this is often referred to as the hub to casing
ratio.

The tip clearance is the margin between the tip of the blade and the
casing (shroud). It has a crucial effect on the blade performance. In blade
design the tip clearance is scaled with the blade height, but this doesn’t take
into account the blade chord.

The diffusion factor relates the relative velocity and the whirl velocity
with the space to chord ratio and is defined from the following equation

=1− + (2.3)
∗ ∗

where σ is the solidity. The diffusion factor is suggested to be 0.45 to

obtain a reasonable blade design. A value higher than 0.6 yields separation
and is thereby not recommended. Having established all the velocities, the
solidity can be calculated in a straightforward manner (Strohmeyer, 2009).

Efficiency (Osborne, 1977) Fan efficiency is the ratio of the output

power to the mechanical input power. The total efficiency is the ratio of the
total air power to the measured fan input power and the static efficiency is the
static air power to the measured fan input power.

The measured fan input power is defined in every case as the power
which is absorbed by the impeller or by the entire fan (to include the losses in
bearings, transmission, etc)

Reaction is the ratio of the static pressure rise at the rotor divided by
the total stage pressure rise. The reaction distribution usually varies across
the radii.

Fan Total pressure is the difference between the total pressures at

the Fan Outlet and Inlet (Osborne, 1997)

Fan Static Pressure is the Fan Total Pressure minus the Fan Velocity
Pressure. The Fan Static Pressure is not the rise in static pressure across the
fan, but it is defined by the Bernoulli equation. (Osborne, 1997)

11 | P a g e
 Lite
erature Rev
view

Fan Veelocity Prressure is the veloc city pressure corresp

ponding to the
ave
erage veloccity at the fan outlet (found by dividing th
he volume flow of the
e air
t area off the fan diischarge orifice) (Osb
by the borne, 19997)

Figure 2.11 Pe
erformance characteristic
c cs of a fan (O
Osborne, 197
77)

The Tootal Pressuure is the sum of the static prressure an nd the velo
ocity
pressure acco
ording to Bernoulli’s
B e
equation fo
or incomprressible flow:
2
Pt=Ps+1/2*ρ*V
+ (2.4)

At zerro volume flow

f the Faan total pre
essure is equal
e with the Fan sttatic
pressure, beccause the fanf velocitty pressure e is zero. In contrastt when the
e fan
worrks with oppen inlet and
a outlet, the fan to
otal pressuure is equaal with the fan
velo
ocity presssure, becauuse the fan n static pre
essure is zero.
z As thhe volume flow
deccreases the e fan total pressure e increases s and consequently the fan sttatic
pressure incre eases untill a peak.

The highest efficciency is not

n achiev
ved at the highest sstatic press
sure
poinnt, but in a slight higher vollume flow because of the hiighest rate e of
incrrease of th
he fan inputt power.

These relationsh hips are sh hown diaggrammatica ally in Figu

ure 2.11. The
maxximum effiiciency doe es not occcur at the point
p of hig
ghest total pressure, but
in a lower tota
al pressure
e point.

12 | P a g e
 Literature Review

Air power is the result of the pressure rise multiplied by the volume
flow. This occurs considering that work is done when a volume of air is moved
in a distance l, in a pipe of constant area A and with constant pressure p. The
work is then ∆p*A*l and the power is (∆p*A*l)/t. The product (A*l)/t is the
volume flow Q so the air power is ∆p*Q.

The air power can be total or static, depending on the kind of pressure
that is used in the above equation.

2.2. Fan Laws

The main quantities that characterize the performance of a fan are the
volume flow, the pressure rise and the power consumption.

The performance of a fan that operates in a specific condition is

governed by a set of variables. These can be employed to describe the
performance of a related fan design operating in another condition, using the
so called fan laws. The fan laws are derived from the consideration of
dimensional analysis relationships. The variables used in the fan laws can be
the size, the rotational speed and the density of the gas. So, by applying the
fan laws the volume flow Q, the static pressure rise SP and the power
consumption in two different fans can be related with the variables that govern
the two fans.

In case there is a requirement for converting the performance of a fan

to a different rotational speed the following laws are applicable (Figure 2.12).
The dashed lines depict the performance of a 1750rpm fan and the normal
lines the performance of a 2625 rpm fan.

The Volume Flow is proportional to the fan speed (2.5) and the Static
Pressure is proportional to the square of the fan speed (2.6)

Q rpm P rpm
∝ (2.5) ∝ (2.6)
Q rpm P rpm

Total Pressure and velocity pressure also vary as the square of the fan
speed

The Power consumption PC is proportional to the cube of the fan

speed (2.7) and the noise N varies according to formula (2.8)

PC rpm rpm
∝ (2.7) N − N ∝ 50 ∗ log (2.8)
PC rpm rpm

13 | P a g e
 Lite
erature Rev
view

The effficiency re
emains thee same andd has the same tren nd for diffe
erent
volu
ume flow range.
r Thiss trend was expected
d because the power consump ption
and essure are vary both proportion
d Static Pre nally with th
he square of rotation
nal

Figu
ure 2.12 Sam
me Fans with
h different rottational spee
eds (Bleier,19
998)

The se econd case e is the va

ariation of the
t fan size. The size of the faan is
defined by th he tip diameter D. The fan laws are applicable e in this case
c
asssuming thee geometricc proportioonality of the
t fans (ssame number of blad des,
samme other size
s ratios, same blade angle es). The efffect of fan size on the
perf
rformance of the fan is illustratted in Figuure 2.13. The
T dasheed lines de
epict
the performan nce of a 17750rpm fan and the normal lines the perrformance of a
26225 rpm fan. The dash hed lines depict
d the performance of a 27 7in fan and the
norrmal lines the
t perform mance of a 40.5in fan n.

The Vo
olume Flow
w is proporrtional to cu
ube of the size

Q D
∝ (2.9)
Q D

14 | P a g e
 Lite
erature Rev
view

The Sttatic Presssure is prroportional to the sq

quare of the fan sppeed
(2.1
10) and thee Power co
onsumption is proportional to the
t fifth po
ower of the
e fan
speeed (2.11)

P D C
PC D
∝ (2.1
10) ∝ (2.11)
P D PC
C D

The no
oise varies according to the follo
owing form
mula

D
N − N ∝ 50 ∗ log (2.12))
D

The effficiency reemains almmost the same.

s The
ere is a m
minor incre
ease
(<1%) for larg
ge fans (Ble
eier, 1998))

Figure 2.13 Similar Fa

ans with different sizes (tip diameters 27’’ and 40.5
5’’) (Bleier)

The effficiency cu
urve and the
t static pressure curve
c in th
he smaller fan
are more steeep, so the range
r of operation iss smaller th
han in the llarge fan. This
T
is due
d to the fact that in a smalle
er fan the maximum volume flow is about 4
15 | P a g e
 Literature Review

times lower than the maximum volume flow in a larger fan, while the static
pressure rise is about 2.5 times lower.

The power consumption increase in the large fan is much higher than
the static pressure rise increase, because the larger fan has to move
quadruple volume of air and achieve 2.5 times more static pressure. The
power consumption is proportional to the product of the static pressure rise
and the volume.

The third variable that affects the fan performance is the air or gas
density. The fan may operate in various altitudes or the working fluid may
have different density than air.

In this case the Volume Flow will remain the same: = 1 (2.13)

Static pressure will vary directly as the density ρ: ∝ (2.14)

Power consumption will vary directly as the density ρ: ∝ (2.15)

It is interesting to investigate the effect in the performance in a case

where two variables change at the same time as illustrated in Figure 2.14.
The above mentioned set of rules can then be combined to give the predicted
performance. In the case of the simultaneous change of rotational speed and
size the fan laws application has the following results:

Q D rpm
∝ ∗ (2.16)
Q D rpm

P D rpm
∝ ∗ (2.17)
P D rpm

PC D rpm
∝ ∗ (2.18)
PC D rpm

The following table refers to an example where there is variation of size

and rotational speed with reciprocal ratios

D rpm
∝ (2.19)
D rpm

16 | P a g e
 Lite
erature Rev
view

Figure 2.14 Similar

S Fanss with differen
nt tip diamete
ers and rotattional speed (Bleier,1998
8)

The im
mportant observation n in Figuree 2.14 is that
t the sstatic press
sure
remmains the same
s and the
t efficienncy as welll. So by inccreasing th
he size and d by
deccreasing thhe rotationaal speed byy the same
e percentage the volu ume flow rises
r
morre than the e static prressure whhich remains the sam me and this results in a
flattter pressure curve. On the other hand the efficiency and th he noise leevel
remmain almosst the same e.

Other things
t that can be no
oticed are as
a follows

a. The largerr fan has flatter

f static pressuree curve be ecause the the
connversion fa
actor for the
e static pre
essure is smaller
s tha
an the one of the voluume
flow
w
b. The efficie
ency curve in the larg
ger fan is flatter as weell

2.3
3 Flow Coeefficient Φ and Worrk Coefficiient Ψ
There are nume erous fans in the ma arket and the assesssment or the
com
mparison of
o their performance is difficultt when the
e commonn performa
ance
17 | P a g e
 Literature Review

variables (size, static pressure, rotational speed) are used. The initial design
of a fan also requires a data base where the performance of a number of fans
can be meaningfully compared. If, for example, the performance of two fans
regarding the static pressure and the volume flow has to be compared, the
performance has to be converted to take into account the relative sizes and
the rotational speeds of the two fans.

For this reason dimensionless coefficients can be employed. For the

fans design, the main dimensionless coefficients are the flow coefficient φ and
the work coefficient ψ. These coefficients are derived from the air volume and
the pressure rise at the point of maximum efficiency and with standard
working fluid (standard density) at the mid span.

The Flow or Volume Coefficient φ relates the volume flow Q with the
size of the fan D and with the rotational speed U. Through this coefficient the
capacity of the fan to move the air can be categorized. The definition of the
Flow Coefficient φ is (Osborne 1977, Bleier 1998, Eck 1973):

Q
φ= (2.20)
D
π∗ ∗U
4

Where Q is the volume flow, D the outer diameter, U=ω*r

Another definition for φ which is also more convenient is the following

φ∗ = (2.21)

With some manipulation it follows that φ∗ = (2.22)

Where ν = which is the hub to tip ratio

The quantity φ* is denoted as φ by some authors (McKenzie 1997,

Ramsden 2008, Lewis 1996). The coefficient φ gives the order of magnitude
of the size of the fan.

The Work or Pressure Coefficient ψ relates the pressure rise in the fan
with the dynamic pressure of the peripheral velocity. The magnitude of this
coefficient typifies the pressure rise capability of the fan. The definition for ψ
for incompressible flow is

ΔP
ψ= (2.23)
1
∗ρ∗U
2

18 | P a g e
 Lite
erature Rev
view

where ρ is the de
ensity

The ∆ΡΡ may be the total pressure rise or moost commoonly the sttatic
pressure rise
e and this is the usseful ∆P and
a not the
e design ∆
∆P (Osbo orne,
197
77).

Figuress 2.15 and

d 2.16 illusttrate actua
al values of φ and ψ from fans that
are found in the
t markett. These va alues can be the starting point for a new
w fan
dessign.

Fiigure 2.15 Va
alues for φ and
a ψ from actual fans (O
Osborne, 197
77)

The wo
ork coefficcient for hig
gh speed cases
c thuss for comp pressible flows
is defined
d as the Stagn nation Enthhalpy rise over
o the rootational sspeed squa ared
(Lewwis 1996, McKenzie 1977)

ΔΗ
ψ= (2.24)
U

19 | P a g e
 Lite
erature Rev
view

Figure 2.16 Values

V for φ and ψ from actual fans (Bleier,
( 1998
8)

It is in
nteresting to investiggate how the efficie ency is rellated with the
dim
mensionlesss coefficien nts φ, ψ an nd for this particular case there e are availa able
whaat are kno own as Smith chartts (Lewis, 1996). In n these ch harts, ano other
dim
mensionlesss coefficien nt is incorp
porated, the de Halle er number w which as seen
s
is defined
d as the
t ratio off the outlett relative ve
elocity to the inlet rellative veloc
city

dH = V2/V1 (2.25)

The dee Haller number

n is a measuure of the diffusion of the bllade
passsage and it must beb higher than
t 0.7 in
n order to
o avoid larrge separa
ation
lossses (Ramssden, 20088)

Figure 2.17
7 Smith Charrts for (a) 50% reaction (b
b) 70% reacttion stage (L
Lewis, 1996)

20 | P a g e
 Literature Review

Figure 2.18 Smith Charts for a 90% reaction stage (Lewis, 1996)

Smith charts are illustrated in Figure 2.17 and Figure 2.18. Every point
in the charts represents the maximum efficiency that can be achieved for a
pair of φ and ψ, taking also into account the de Haller number.

The object of this study is a vaneaxial fan, which is a fan stage where
the stators are placed downstream of the rotor. After reviewing the literature
the main conclusion regarding the range of the φ and ψ parameters for this
particular vaneaxial machine is the following:

Flow coefficient φ: 0.2-0.3

Work coefficient ψ: 0.2-0.4

Other important conclusions concerning the effect of the dimensionless

coefficients to the fan design, according to (Eck, 1973), are:

a. When the product φ*ψ is maximum the fan has the maximum
capacity with minimum size, which gives the cheapest design
b. The maximum value of ψ gives the minimum noise generation
c. The higher the φ the larger the capacity of the fan - The higher the
ψ the higher the pressure rise

2.4 Axial Flow Fans

An axial fan may be defined as a fan where the flow of the air is
basically parallel to the axis of the rotor throughout its transit through the
machine. The main difference between a fan and a compressor is that the
purpose of a fan is to move air while the purpose of a compressor is the
21 | P a g e
 Literature Review

increase of pressure. The role of low pressure axial fans therefore is the
moving of large volumes of air at a relatively low speed and with a modest
pressure rise.

These machines find great applicability over a wide range of

engineering applications. Mechanical ventilation uses fans so that control over
air movement is obtained and the environmental conditions in a given space
are not subjected to the uncertainties of the natural ventilation systems. In
many applications it is essential to control the contamination such as in
industrial environments or tunnels as illustrated in Figure 2.19.

Figure 2.19 Tunnel Ventilation (www.roadtraffic-technology.com)

Another application of axial flow fans is thermal management and an

example for this is the cooling of electronic components, Figure 2.20. This is
one of the main concerns for the achievement of the higher packaging
densities which are required by the general trend of microminiaturization.

22 | P a g e
 Literature Review

Figure 2.20 Electronics cooling fans (www.qmed.com)

A fan is a machine used to produce a continuous flow of air by

aerodynamic action. Fan engineering requires a good knowledge of fan types
(centrifugal, axial, and mixed) together with the theory of their operation and
design, and of the fan laws, fan testing, fan systems, fan mechanics
(materials strength) and fan selection.

The selection of a fan for air-conditioning or ventilation applications

involves choosing the most inexpensive combination of size, arrangement and
type while providing stable operation. The main advantages of axial flow fans
are their high efficiency, compactness and simplicity of installation. However,
the pressure development along an axial fan is considerably lower than a
centrifugal fan for a given impeller diameter.

The fan selection for a particular application is usually concerned with

its characteristic curve which defines a relation between pressure and volume
or mass flow. For incompressible flow due to the low developed pressure rise,
the characteristic curve of a small axial fan, for standard atmospheric
conditions, can be expressed as pressure difference against mass flow rate at
a given rotational speed as illustrated in Figure 2.21.

23 | P a g e
 Literature Review

Figure 2.21 Characteristic line of a low speed fan (Strohmeyer, 2009)

In the simplest form of an axial fan, the air approaches the rotating
component of the fan (impeller) with a given axial velocity. The work done on
the fluid by the impeller torque adds a rotational component to the velocity. As
a result the absolute velocity at the exit is higher than the axial velocity, thus
some of the total pressure that was developed by the fan does not appear as
useful total pressure.

There are 3 main types of axial fans: the propeller fan of Figure 2.22,
the tubeaxial fan of Figure 2.23 and the vaneaxial fan of Figure 2.24.

Figure 2.22 Propeller fan (www.canadablower.com)

24 | P a g e
 Lite
erature Rev
view

Figure 2.23 Tubeaxxial fan (www

w.cincinnatifa
an.com)

Figure
e 2.24 Vanea
axial fan (ww
ww.flaktwoods.com)

The prropeller fann is used when the lower cosst is the m

more imporrtant
facttor of thee design. Often the ey do not have ae erofoil blad
des and their
t
efficciencies arre quite low
w.

The tub beaxial fan

ns is a prop
peller fan with
w a cylin ndrical hou
using (shro
oud).
It has more so ophisticate
ed design, increased static presssure and efficiency and
a huub to tip ra
atio range from
f 0.3 to
o 0.5 (Bleie
er, 1998)

When there is a requiremeent for high

her pressu
ure and higgher efficie
ency
guid
de vanes are
a added to the fann and in thaat case the
e fan is ca
alled vaneaaxial
fan. In fans where
w the loss of usseful press
sure is hig
gh (outlet velocity an
ngle

25 | P a g e
 Lite
erature Rev
view

greater than 15 degree es, McKen nzie 1997, Osborne 1977) guiide vanes are
ded to the
add e design in order to o remove the rotational comp ponent of the
velo
ocity and convert
c som
me of the dynamic
d prressure intto useful sttatic press
sure.
Thee vaneaxia al fans havve higher hub
h to tip ratio than the tubea axial fans. The
hub
b to tip ratio
o for vanea
axial fans ranges
r from
m 0.5 to 0..8 (Bleier, 1998)

The guuide vaness can be lo

ocated on the outlet side or on n the inlet side
s
of the
t fan bla
ades. In th
he case of inlet vanees the stattic pressurre is produ
uced
onlyy from thee fan blad des; the contribution
c n of the innlet guide vanes is the
preparation off the flow for
f the fan blades as shown in Figure
F 2.255.

Figure 2.2
25 Vaneaxial fan with inle
et vanes (Ble
eier, 1998)

In the case of downstrea

d m guide vanes
v a percentage
p e of the sttatic
pressure is produced
p a the fan
at n blades and
a anothe
er percenttage from the
nes as show
van wn in Figure 2.26.

26 | P a g e
 Lite
erature Rev
view

Figure 2.2
26 Vaneaxial fan with outllet vanes (Bleier, 1998)

The guuide vaness provide an increas se in the static presssure and the
efficciency of a vaneaxial fan compparing with an impelleer fan or a tubeaxial fan.
In most
m applications the
e outlet va
anes are preferable
p than the inlet vaness. In
speecific casess where thhe motor is
i installedd at the fro
ont of the fan for beetter
acccessibility, the
t inlet gu
uide vaness are the be
etter solution.

Howevver, inlet guide vanes v havve importtant disad dvantages in

perf
rformance comparing with th he outlet vanes, esspecially iin high sttatic
pressures (Ble
eier, 1998)) as followss:

a. The shap pe of the e inlet va anes is more

m criticcal, than the
dowwnstream guide
g vane es
b. The opposite spin thatt the in
nlet vanes give to the flow re esult
highher noise
c. The inlet vanes
v are staggered d in a perm
manent po osition so they
t
cann provide the nece essary inittial inlet angle only in the design point p
connditions. Th
heir efficie
ency is deccreasing inn of design
n cases wwhile the no oise
incrreases

27 | P a g e
 Lite
erature Rev
view

F
Figure 2.27 Various
V type axial
a fan perfformance (O
Osborne, 197
77)

In Figu
ure 2.27 th he perform mance of a fan is illu ustrated in
n terms of the
dim
mensionlesss pressure e flow characteristic ψ and the e dimensioonless volu
ume
flow
w characteristic φ forr all types of
o axial fan
ns. In Figure 2.28 thee performa
ance
of a vaneaxial fan is illusstrated in terms
t of sttatic pressu
ure and vo
olume flow.

F
Figure 2.28 Performance
P curve of a va
aneaxial fan (Bleier, 1998)
28 | P a g e
 Literature Review

The performance curve starts from the point of free delivery (static
pressure=0). If the flow restrictions increase in the environment or ducting
system where the fan operates, the static pressure increases and the volume
flow decreases. This is due to the fact that with increasing flow restrictions,
the volume flow will decrease and the axial velocity will decrease. The inlet
angle of the relative velocity will increase (same rotational speed and lower
axial velocity) and the angle of attack will increase and as a result the static
pressure will increase (operating range).

However this will happen until the maximum useful angle of attack
when the peak of static pressure is reached. After that point, the angle of
attack will increase above the stall point of the blade, the flow will not be able
to follow the contour of the blade, separation will occur and the static pressure
will finally decrease simultaneously with the volume flow (stall dip region).

After the stalling dip region it would be expected that the lift coefficient
will decrease until the point of zero flow and zero static pressure is reached
(dashed line). However, the fan continues its operation albeit with low
efficiency and high noise. When the fan operates in a stalled condition the
flow is to some extent thrown outwards by centrifugal force and in this way
static pressure is produced until the point of zero flow.

29 | P a g e
 Preliminary Fan Design Method

3. Preliminary Fan Design Method

3.1 Introduction
Various aerodynamic techniques have been employed for the design of
low speed fans. The most powerful design methods are the three dimensional
methods which are available due to the advance in computational
aerodynamics. However, the application of these advanced methods to
industrial fans is only justified for special cases (Smith, 1989)

The use of the two dimensional airfoil characteristics is generally

adequate for the design of an industrial fan. The design of an industrial fan
based on the 2-D airfoil characteristics is very well documented in literature.
For a preliminary design of an industrial fan the 2-D approach will result in a
fan design with acceptable performance.

The purpose of a fan is to move a required volume of air and at the

same time to overcome the resistance of the area by providing the necessary
pressure rise. The volume flow Q and the pressure rise ∆P are the main
design parameters for a fan and it is common to categorize fans according to
the volume flow (size) and pressure rise.

3.2 Method Selection

The preliminary design of an axial fan can be approached in two main
ways:

a. Treating the fan as a series of isolated airfoils

b. Regarding it as a rotating cascade

The choice of the method depends on the space to chord ratio of the
fan; the ratio of annulus area to the blade area. The space to chord ratio
depends on the loading, so relatively early in the design procedure the
appropriate method becomes apparent.

In the isolated aerofoil approach each blade is considered as a series

of two-dimensional airfoil sections. The lift and drag forces that act upon them
are resolved into the torque and thrust planes respectively and the integration
of them from root to tip gives the shaft torque and static pressure rise across
the rotor disc. Knowledge of the local lift and drag forces, which are derived
30 | P a g e
 Preliminary Fan Design Method

from the local air velocity relative to the blade, is required for the calculation of
the static pressure rise. The designer controls the velocity and pressure
distribution within the annulus to maintain radial equilibrium.

The main limitation to the isolated aerofoil approach is the distance

between the blades. When the blades are sufficiently close to each other,
there is a divergence between the theoretical and actual values of the lift
forces, because their pressure fields interact and seriously restrict the
development of the full lift forces.

This issues can be corrected using empirical rules that derive from
experiments, however the effect of these empirical corrections deteriorate as
the distance between the blades decreases. According to Bass 1987, for
solidity values greater than 0.7 the isolated airfoil method is not
recommended. For closer spacing an alternative approach is needed and
unless facilities are available for a more elaborate computation of the flow
field and the aerofoil surface pressures recourse must be had to the cascade
method.

According to the cascade method the row of blades is regarded as a

row of rotating diffusing passages and the deflection imparted to the flow is
considered rather than the lift forces on the individual blades, although the
latter can be expressed in terms of inlet and outlet flow angles. Total pressure
loss corresponds to blade drag and the deviation of the flow at the outlet from
the condition of tangency to the camber line is analogous to the lift slope of an
airfoil (Bass, 1987).

Experiments showed that the lift that an airfoil produces as an isolated

airfoil is not the same with the lift that is produced from the same airfoil as a
part of a cascade. Dixon, 1998 defines as k as the ratio of the lift of the blade
is cascade to the lift of the blade isolated. The variation of k with the space to
chord ratio for different stagger angles is depicted in Figure 3.1. For lower
space to chord ratios (less space between blades) and especially for high
stagger angles k is high and the designer should be careful with the design
assumptions. The effect of closely spaced blades is difficult to predict in an
analytical solution, however the result will be obvious in a CFD simulation.

31 | P a g e
 Preliminary Fan Design Method

Figure 3.1 Lift ratio, k in variation with space to chord ratio (Dixon,1998)

In general and according to Turner, 1966 conventional compressor

stages are designed by the cascade method, while high stagger low solidity
ducted fans are designed on modified isolated aerofoil theory.

3.3 Radial equilibrium

In an axial fan the flow must be maintained in equilibrium between the
radial static pressure gradient and centrifugal field. As a result, the choice of
the work distribution along the blade is influenced.

A vortex is a mass of fluid that rotates. This phenomenon occurs in

axial fans in order to balance the forces, because of the pressure difference
and the rotating speed.

Usually the preliminary design starts by calculating the behavior of a

blade at mid radius. The behavior at the mid radius of the fan represents the
total behavior of the blade, but still the conditions across the radii have to be
established. The aerodynamic loading at each radius must be carefully

32 | P a g e
 Preliminary Fan Design Method

designed, because the variation of performance at each radius can cause

difficulties in the choice of the blade geometry and the blade profile.

The first step is to define the aerodynamic performance as one moves

along the radii. There are several assumptions that can be made and from
these assumptions different design philosophies arise.

3.3.1 Free Vortex Design

The free vortex is the simplest design and offers the best efficiency
possible. In a free vortex condition the axial velocity is constant across the
radii while the swirl velocity is inversely proportional to the radii. That means
that as one moves from the hub to the tip the swirl velocity is lower in order to
keep the contribution of pressure rise constant from tip to root. This is
associated with constant axial velocity across the annulus and the result is
that no downstream energy losses from the redistribution of kinetic energy
occur.

The free vortex pattern is like that of the water draining in a bath tub,
where the rotational velocity near the plug hole is high but reduces in
magnitude as the radius away from the drain increases.

The first assumption that can be made is that the total energy remains
the same the same across the radii. The total energy is expressed by the
Bernoulli’s equation for incompressible flow along a streamline:
1 2
Pt= Ps + ∗ ∗ =constant (3.1)
2

This equation can be manipulated to the more convenient form:

Uw * r = constant, (3.2) where Uw is the whirl velocity

The above equation confirms the previous assertions regarding the fact
that the axial velocity remains constant across the radii while the swirl velocity
is inversely proportional to the radii. This characterizes the “free vortex”
design. The difference in the flow between a free vortex design and a non free
vortex design is illustrated in Figure 3.2.

33 | P a g e
 Preliminary Fan Design Method

Figure 3.2 (a) free vortex axial fan (b) non free vortex axial fan (Lewis,1996)

An important and practical advantage of the free vortex design is that it

is stable and simplifies the fan design whenever it can be adopted.

A major disadvantage however is that a free vortex design implies high

root loading. Since the hub velocity is much less than the tip velocity, this
difference must be compensated for through a high twisting of the blade at
root in order to maintain the same pressure rise contribution across the whole
span. As a very approximate indication a free vortex approach is unlikely to be
satisfactory for hub/tip ratios less than about 0.3 although for very lightly
loaded fans a smaller value may be possible. Its use in extreme cases results
in very large root chords and highly twisted blades in this region according to
Bass, 1987. Examples of the preliminary design of a fan blade employing a
free vortex methodology can be found in Lewis,1996.

Where conditions require it, very considerable work grading from root
to tip can be employed giving much reduced root loading and a much more
manageable design but care is required to maintain the radial equilibrium
mentioned above. Values of root/tip work loading of 2.0 or more have been
successfully achieved using the flow deflections of the cascade approach
combined with an assumption of a free vortex swirl distribution

Bass, 1987 states that in many cases the design can be based in the
free vortex swirl distribution while at the same time considerable work grading
from root to tip can be employed in order to improve the design.

3.3.2 Non Free Vortex Design

Besides the free vortex design, radial equilibrium can also be achieved
with either constant swirl velocity, forced vortex or the mixed vortex designs.

34 | P a g e
 Preliminary Fan Design Method

In the constant swirl design approach the swirl velocity is kept constant
across the span:

Uw = K, (3.3) where K is a constant

In the forced vortex design the swirl velocity is proportional to the

radius:

Uw =K*r (3.4)

The mixed vortex design in turn is a combination of free vortex and

forced vortex:

Uw = K*r + K/r (3.5)

Examples for the above mentioned radial equilibrium approaches may

be found in Lewis, 1996.

3.4 Hub to tip ratio definition

According to the procedure that is described in Bleier’s handbook,
(Bleier, 1998) the input data for the design of a fan are the Volume Flow Q,
the Static Pressure Ps, the Fan Diameter D and the rotational speed rpm.
These requirements are specified by the customer in any case.

The hub diameter can be calculated from the following equation

=( )∗√ (3.6)

or from Figure 3.3

In Figure 3.3 the hub diameter can be found if the Static Pressure and
the Rotational speed are set.

35 | P a g e
 P
Preliminary
y Fan Design Method
d

Figure 3.3 Hub Dia

ameter calculation (Bleierr, 1998)

The hub diametter is a critical

c parrameter, because
b a smaller hub
diam
meter coulld invite th
he occurren
nce of turb
bulence an
nd reverse flow nearr the
hubb, which ca
an lead to stall.
s

Furthermore as McKenzie,
M 1997 statees the size
e of the fan
n annulus area
a
is im
mportant because
b th
he lower th
he fan annulus veloccity, the lowwer the los
sses
duee to the difffusion of th
he annuluss velocity in
n the duct.

McKen nzie definees the hub to tip ratioo by plottin

ng the valu ues of ∆Η//Va2
2
andd Va/U at the diagram m in Figure
e 3.4. The values
v of ∆Η/Va
∆ and
d Va/U cann be
calcculated after assumin ng a hub to
o tip ratio. This is an iterative pprocedure until
a saatisfactory value of Pressure
P C
Coefficient Cp
C is achie eved.

Cp=1-((V2/V1)2 where V2/V V1 is the de

d Haller number. T The de Ha aller
nummber cann not be mo ore than 0.7,
0 otherw
wise separration will occur at the
trailling edge area.
a As a result the maximum Cp that is acceptable is 0.5

36 | P a g e
 Preliminary Fan Design Method

Figure 3.4 Work coefficient vs flow coefficient diagram (McKenzie,1997)

Osbourne, 1979 on the other hand uses established values of φ and ψ

that were mentioned in Figure 2.15 to obtain the hub to tip ratio.

According to Bleier, 1997 the hub to tip ratio for Vaneaxial fans varies
from 0.5 to 0.8.

3.5 Airfoil selection

The choice of aerofoil for low pressure fans is not critical and in
practice often depends largely on the availability of enough aerodynamic data
to cover the range of working conditions encountered by the different radial
stations along the blade (Bass, 1987). Provided that the operating lift/drag
ratios are at least 50, a readily achievable target, the section efficiency is not
very sensitive to small changes in drag.

The C4 airfoil is chosen from McKenzie,1997 for the preliminary design

of a fan. Bleier, 1997 used a NACA 65 profile for his preliminary design of an
axial fan.

37 | P a g e
 Preliminary Fan Design Method

Figure 3.5 NACA 65010 airfoil [http://www.mh-aerotools.de/airfoils/javafoil.htm]

A NACA 65 airfoil profile is illustrated in Figure 3.5. The knowledge of

an airfoil profile coordinates and the aerodynamic characteristics of an airfoil
are necessary for the selection of the airfoil. A very useful application for the
creation of various airfoils and the airfoil analysis may be found in the
following website http://www.mh-aerotools.de/airfoils/javafoil.htm.

3.6 Blade Geometry

Best efficiencies for vaneaxial fans are obtained with airfoils as
cross sections of blades. These airfoils have high lift to drag ratios.
Additionally for a good efficiency the airflow of an axial fan should be evenly
distributed over the working face of the fan wheel which means that a free
vortex design is preferred.

The velocity of the rotating blade increases with the radius. This can be
compensated for by twisting the blade resulting in larger blade angles near
the hub and smaller blade angles near the tip.

38 | P a g e
 Preliminary Fan Design Method

The airfoil blades are manufactured from aluminum or steel castings

and they are usually twisted and tapered. There are several advantages and
disadvantages in a tapered airfoil according to Bleier, 1998:

1. Wide chord at the tip: high pressure, high efficiency and quiet
operation, but cause of considerable motor overload at the point of no delivery
but fans with such blades have a deeper stalling drop

2. Medium chord at the tip: this reduces the maximum static pressure
and the no delivery overload

3. Narrow chord at the tip: non overloading brake horsepower

characteristic, flatter pressure curve and efficiency slightly lower, but
acceptable.

McKenzie, 1997 considers that the tip should not exceed the hub chord
by more than 20% for centrifugal loading reasons. In his preliminary design
example of an industrial fan, the tip chord is 2/3 of the hub chord to ensure a
reasonable size of the airfoil and a modest thickness to chord ratio.
Osborne,1977 suggests a tip chord that is 75% of the hub chord in order to
achieve a solidity of unity for a specific number of blades.

3.7 Blade design

The air angles can be calculated from the velocity triangles of Figure
2.2 and the equations resultant from the free vortex design method. The blade
angles derive after the consideration of the incidence and the deviation is
taken into account. According to McKenzie, 1988 the peak efficiency can be
achieved when the angle of incidence lies between +5ο to -10ο as it is
illustrated in Figure 3.6.

39 | P a g e
 P
Preliminary
y Fan Design Method
d

F
Figure 3.6 Bla
ade geometrry for fixed air angles (MccKenzie, 198
88)

A trial and
a error procedure
p is used by
y McKenzie, 1997 to o find the most
m
apppropriate in ncidence. The
T variattion of inciidence affe
ects the space to ch hord
ratioo which ha as to be between a range of values
v in order
o to givve the des
sired
efficciency as it is shown in Figure 3.7.

Figure
e 3.7 Efficien
ncy contours
s (McKenzie, 1997)

40 | P a g e
 Preliminary Fan Design Method

The following equations are used for the calculation of the blade
angles. An analytical example may be found on McKenzie, 1997

a1=atan(U/Va) (3.7)

tan a2=tan(a1)-∆Η/U*Va (3.8)

tan a3=U/Va-tan a2 (3.9)

Cpi (rotor)=1-(cos a3)2 (3.10)

S/C (rotor) = nblades*(0,567-Cp) (3.11)

tan am= (tan a1+tan a2)/2 (3.12)

tan ζ = tan am -0.15 for low staggers or tan ζ = tan am -0.213 for high
staggers (3.13)

θ= (0,5-0,31*(S/C)(1/3) (3.14)

β2=β1-θ (3.15)

deviation δ=α2-β2 (3.16)

deflection e =a1-a2 (3.17)

i = α1- β1 (3.18)

θ=2(β1-ζ) (3.19)

S/C=δ/(1.1+0.31θ)3 (3.20)

The equations for incidence and space to chord ratio are underlined to
show the relation between these two design features. McKenzie, 1997 found
that the maximum efficiency occurs when the mean angle and the stagger
angle are related according to the equation tan ζ = tan am -0.15 or tan ζ = tan
am -0.213. This is the result of the experiments that were made on cascades
with C4 airfoil blades. For a different basic airfoil, the stagger angle is
correlated with the mean air angle with a modified equation.

From the above mentioned procedure it is clear that the airfoil profile in
every section from hub to tip will be different. In McKenzie’s example C4 is the
base profile, but in every section the camber and the stagger can be different
in order to maintain the principle of the free vortex design which is ensures the
same amount of pressure rise for every section.

Bleier, 1997 follows a similar procedure. The velocities and the air
angles are calculated for every section and then the base airfoil is modified
41 | P a g e
 P
Preliminary
y Fan Design Method
d

acccordingly inn order to give the desirable

d static
s presssure rise. The equa ation
thatt is used to
o calculate the lift coe
efficient of every section is the following:

= 6,6456 ∗ 10 ∗ ∗ ∗ ∗ (B
Bleier,1998
8) (3.21)

Ps : Sta
atic Pressu
ure rise in Pascals

zB : Number of blades
b

CL : Lift coefficie
ent of the airfoil at the
e specific ra
adius

L : Blade width

W : Average
A relative veloccity of the flow
f (m/s)

From the above equation the CL can be calcula ated and frrom the CL the
neccessary mo odification can be ma ade to the base proffile. It is ob
bvious thatt the
dessigner sho ould have available the nece essary datta for the aerodyna amic
chaaracteristicss of an airf
rfoil. From the CL, the
e angle of attack
a can be found and
fina
ally the blad
de angle can
c be calcculated. Fig gure 3.8 illu
ustrates a CL vs anglle of
atta
ack diagram m which ca an be used d to obtain
n the anglee of attack for a givenn lift
coeefficient.

Figu
ure 3.8 Lift coefficient variation with angle
a of attac
ck for a 4-dig
git NACA airffoil (Bleier, 19
998)

A simmilar appro oach is suggested

s by Osborne, 197 77. After the
calcculation off the veloccities and the air anngles, the solidity is selected and
from
m solidity the
t CL can n be foundd using thee following equation which derrives
from
m the radiaal equilibriu
um and thee free vorte
ex design.
42 | P a g e
 Preliminary Fan Design Method

∗ ∗C ∗u =u (3.22)

where is the solidity and u is the mean velocity (V1 + V2)/2

From the C , the angle of attack can be found and finally the blade
angle can be calculated.

The procedure to calculate the vane angles is similar.

3.8 Number of blades

The number of blades is a compromise between many design features;
the blade chord, the space to chord ratio, aspect ratio, noise level and
turbulence. A certain design can be modified be changing the number of
blades and their width (chord) to achieve the same pressure rise.

High number of blades with small chord result in a loss in efficiency

and higher noise. Low number of blades with large chord result in a better
efficiency, but the hub will be too wide and bulky. A good practical number of
blades for vaneaxial fans is 5-12 according to Bleier, 1998 and the designer
can select the number of blades from this range.

The number of blades according to McKenzie, 1997 is the result of the

combination of the calculated space to chord ratio and the value of the aspect
ratio that is selected by the designer. An aspect ratio of 1.5 is recommended
by McKenzie, 1997 for the preliminary design of an industrial fan. Osbourne,
1977 selects 9 blades for his preliminary design, within the Bleier's range 5-
12.

The number of blades for a preliminary design of a fan depends on the

experience of the designer. The decision procedure starts from a set of
assumptions (aspect ratio, number of blades, blade chord, and solidity) while
the other design features can be calculated based on that initial set of
assumptions. If the result is not satisfactory the procedure can be revisited.

3.9 Number of Guide vanes

According to Bleier, 1998 there are two practical rules for the stator
guide vanes:

43 | P a g e
 Preliminary Fan Design Method

a. The number of guide vanes should be greater than the blades,

because they should be closer to each other in order to reduce the danger of
blocking the flow because of the overlapping.

b. The number of rotor blades and the number of stator blades should
have no common divisor, otherwise two blades can pass simultaneously two
vanes which result an increase in the noise level. When two blades can pass
simultaneously two vanes wakes are created and 2 disturbances are
produced per one revolution (2nd engine order excitation) and these
disturbances create noise. The failure is not very possible, because the speed
is relatively low.

McKenzie calculates the number of vanes after assuming a vane

chord. The value of the vane chord should be inside the range of the
maximum blade chord and the minimum blade chord. The chord is then
trimmed in order to give an integer and prime (preferably) number of vanes.

The shape of outlet vanes has not great importance. They can have an
airfoil shape or a uniform thickness with the proper curvature (cheaper). The
vanes have in most cases constant thickness (Bleier, 1998)

3.10 Influence of hub to tip ratio

Vaneaxial fans have converging tail pipes and expanding diffusers for
static regain. The hub diameter is higher in vaneaxial fans comparing with
tubeaxial fans, in order to maintain good performance. A smaller hub diameter
might result in the occurrence of turbulence and possible reversed air flow
near the hub (stalling).

The Static Pressure increases with the square of the hub diameter and
for a constant tip diameter with the square of the hub to tip ratio.

In Figure 3.9 the effect of hub to tip ratio to the static pressure rise is
illustrated which is a result of Bleier’s, 1997 experiments in a 29 inch fan with
same number of blades (5), same rotational speed (1750rpm), same angle
(16) but different hub to tip ratios (52% and 68%).

44 | P a g e
 P
Preliminary
y Fan Design Method
d

F
Figure 3.9 Vo
olume Compa
arison of 29in
n vaneaxial fan
f (5 bladess, 11 vanes) with hub to tip
t
ratios 52% and 68% (BBleier, 1998)

The efffect of hub

b to tip ratio
o can be su
ummarized
d according
g to Bleier,,
199
98 as follow
ws

a. The fan with

w higherr hub to tip
p ratio pro
oduces higher maxim
mum
stattic pressurre rise.

b. According to the fan laws thee static pre

essure rise
e ratio shoould
havve been proportiona al to the hub
h to tip ratio. The
e static pre
essure rise is
slightly lowerr because in the seecond fan the blade tip is slig ghtly narro
ower
2
((0.68/0.52) X 3.05=5.222 and the maximum m is 4.75)

c. The staticc pressurre value where we e have tthe maxim mum

efficciency is not
n the pre edicted on
ne, once again
a beca e narrower tip
ause of the
2
bladde ((0.68/00.52) X 2.1=3.6 and the maxim ency occurss at is 3.4)
mum efficie

d. an deliverss higher vo
The first fa olume but in lower sttatic presssure.
Thee Volume Flow
F Q varriation withh Static Pre
essure Ps line is morre flat than
n the
line
e of the seccond fan

e. The decreease in thee static pre

essure (stalling dip) iss much hig gher
in th
he second fan. Fans with highe
er hub to tip ratio havve higher sstalling dips
s

45 | P a g e
 Preliminary Fan Design Method

3.11 Influence of blade angle

The higher the blade angle the higher the volume flow (larger blade
passage), the higher the pressure, the higher the energy consumption for the
same rotational speed. In Figure 3.10 illustrates the different performance
curves of volume flow variation with static pressure for same size and same
speed vaneaxial fans with different blade angles. The blades are single
thickness blades (not airfoils) and there is a 12o twisting from hub to tip.

Figure 3.10 Influence of Blade angle (36in, 1750 rpm, 13-33 angles) (Bleier, 1998)

3.12 Influence of tip clearance

The tip clearance is an important feature of an axial fan. The main
advantage of an increased tip clearance is the safety margin that it offers in a
hostile working environment (high temperature). The main disadvantage is the
deterioration of the fan performance. According to Bleier, 1998, an increase
in the tip clearance has the following effects:

a. Volume flow decreases slightly

b. Maximum Static Pressure decreases considerably

c. Power consumption decreases, but not as much as the pressure

d. Mechanical efficiency decreases considerably

e. Small increase in the noise level at the free delivery

46 | P a g e
 P
Preliminary
y Fan Design Method
d

A smalller tip clea

arance resu
ults in optim
mum perfo
ormance in
n all respec
cts.
The
e effect of tip
t clearance is summ marized in Figure 3.1
11.

Figure
e 3.11 Summ
mary of the in
nfluence of th
he tip clearance, (Bleier, 1998)

3.13 Three dimensiional effe

ects
The previous
p annalysis asssumes tha at the flow
w over the blades is two
dim
mensional and
a that th he loading distribution
n can be obtained
o from root to
o tip.
Howwever, in ana actual fan
f there are
a factors s that weree not taken n into accoount
durring the 2-D analyssis. At th he blade root, an aerodynam
a mically diffficult
situ
uation arise
es in whichh the skewwed bound dary layer of the hub b is swept into
the corner forrmed by thhe convex surface off the blade and the h hub wall giv ving
rise
e to an areea of stalle
ed flow in this corne
er. A similaar situation
n exists at the
47 | P a g e
 P
Preliminary
y Fan Design Method
d

blad
de tip. Th he situation
n is furthe
er complic
cated by the t presennce of a high
h
cenntrifugal field to which
h the bound
dary layer at rest or moving
m slo
owly relativ
ve to
the blade is subjected.

The influence off rotation on o a rotatting aerofooil with twoo dimensio onal
chaaracteristicss is great if the extre
emely high
h radial accceleration is include ed in
the calculations. Furthe ermore acccording to Bass, 198 87 the stall initiates at
a an
incidence that is lower than what is expecte ed notionaally while th
he 3-D effects
of the tip clea
arance reduce the exxpected pe erformance e notably aas is illustra
ated
in Figure
F 3.12
2.

Figu
ure 3.12 Effe
ect of tip clea
arance on the
e characteris
stics of the blade tip section, (Bass, 1987)

Bass, 1987 state es that in a situation

n where a considerab ble numbeer of
larg
ge continuo ously operrating fans is required, small ch
hanges in efficiency can
represent sig gnificant co osts. Evenn in this case, prooviding tha at the loadding
connditions aree not too severe,
s hig
gh efficien
ncies are attainable
a by the carreful
appplication of relatively simple dessign measu ures.

3.14 Losses
The fan
n operation requires the supplyy of energ gy for the aacceleratio
on of
the fluid and energy
e to overcome
o f
forces thatt resist the fluid flow.

48 | P a g e
 P
Preliminary
y Fan Design Method
d

The tottal energy that is sup

pplied to th
he fluid can
n be expre
essed in te
erms
of volume
v floww Q and pressure
p lo
oss ∆p, because pressure is energy per unit
volu
ume of fluid essure losss is evidentt at each change of d
d. The pre direction off the
flow
w, change in section area, obstruction.

The fan
n losses arre due acccording to Cumpsty,
C 2
2003 to skiin friction, flow
sepparation, se
econdary flows
f and energy
e diss
sipation at the system m’s dischaarge.
Duee to the flluid friction
n within thhe real staage, some e of the o ordered kin netic
eneergy entering the blade rows is dissipa ated as he eat during g the diffus sion
process, whicch is practiccally loss of
o useful ennergy. The e skin frictio
on arises from
f
the flow “scrubbing” ag gainst flow
w surfaces s. Separattion and the second dary
flow
ws can occcur through bends or discon ntinuities (blade edg ges, tip, duct
d
missalignmentss, sudden enlargeme ents of the duct etc).

Lewis, 1996 defined a losss coefficient for the rotor as w well as for the
stattor, in orde
er to standdardize the
e calculatio
on of losse
es. The losss coefficieents
are functionss of flow coefficientt φ, the workw coeffficient ψ, the Reyno olds
nummber Re, the stage reaction and a the in
nlet Mach number ((Lewis, 19 996).
McK Kenzie, 19 997 definess a loss co
oefficient as
a well andd he assocciates the loss
coeefficient witth the incid
dence, Fig
gure 3.13. At high incidence th he flow can nnot
follo
ow the geo ometry of the
t convexx side of thhe blade annd separattion occurss. At
larg
ge negativve incidence the flo ow reache es very high
h velociities and can
beccome chokked.

Figure 3.13 Lo
oss coefficient variation with
w incidencce (McKenzie
e, 1997)
49 | P a g e
 Fan Design Tools

4. Fan Design Tools

4.1 Introduction
The theoretical calculation of the fan is the first approach towards the
preliminary fan design. The theory, the equations and the method that were
mentioned in the literature review (Chapter 2) and in the design procedures
(Chapter 3) are employed to generate a fan geometry and hence from the fan
geometry to a simulation model for the CFD calculations. Furthermore, the
preliminary design of a fan requires the performing of a number of tasks. In
this project these steps are the following:

1. Definition of the geometry; calculation of the design parameters

according to the specifications for the fan
2. Creation of the geometry; processing of the design data in order to
create the rotor and the stator geometry
3. Computational domain(s) creation
4. Meshing of the computational domain(s)
5. Assembly of the subdomains, physics and boundary condition set
up
6. Setting up and simulate the case
7. Results processing

Many design tools are required for the completion of the above
mentioned tasks. In this chapter the design tools that were used for this
project will be presented.

4.2 MATLAB codes

For the calculation of the design parameters the main software that is
used is MATLAB.

The McKenzie procedure that was described in a previous chapter was

implemented in an EXCEL spreadsheet. All the equations and the solution
flow were written in the EXCEL spreadsheet, so that the effect of varying
several parameters in the design could be investigated. An attempt was made
to convert this excel spreadsheet into a more sophisticated and automated
MATLAB code, but there was no time available within the schedule of the

50 | P a g e
 Fan Design Tools

present work; however it can be a future task. In this excel spreadsheet the
user can change the design parameters of the case in every step of the
procedure according to his experience and the level of familiarization that he
has with the fan design in order to achieve the desirable result.

4.2.1 MATLAB code for Osborne method

The Osborne, 1977 method for a preliminary fan design was encoded
with MATLAB and the result is quite interesting. This code can be used for the
study of Osborne procedure for a preliminary fan design. The initial design
factors of this code are the dimensions of the fan, the rotational speed, the
duty and the efficiency. One of the limitations of this code is that there are no
calculations for the stator, because there is no such analysis in the Osborne
procedure. One step further could be the implementation of the stator’s design
in this code. The results are the pressure rise (total and static), velocity
vectors, blade angles at the hub and at the tip and the blade geometry in a
text file that was named “profile.curve” file.

Nevertheless there are four output diagrams as well. The first of them
compare the flow and work coefficient of the current design with the flow and
work coefficients of existing designs (Figure 4.1).

Figure 4.1 Comparison of work and flow coefficient of the new design with the existing
designs

51 | P a g e
 Fan Design Tools

The second diagram (Figure 4.2) illustrates the cascade performance

for a typical deflection at the rotor’s hub. This diagram can be utilized in order
to find the upper and lower limit for the space to chord ratio for a certain set of
deflection and outer angle.

Figure 4.2 Deflection in variation with the Outer flow angle for the hub rotor

The last output diagram (Figure 4.3) depicts all the sections of the
blades stacked, in order to help the designer decide if the computed geometry
corresponds to an acceptable design.

52 | P a g e
 Fan Design Tools

Figure 4.3 3-D plot of blade sections (output from MATLAB code)

4.2.2 MATLAB code for blade geometry

The definition of the blade and vane geometry is the most important
task in fan design. Another MATLAB code was created to process the design
data and provide the necessary geometry for the rotor and the stator.
Although there is available software that can create blade geometry (Lewis for
example), this code was created to suit the specific design outputs of
McKenzie’s preliminary design method. Generally, this code can create the
blade geometry when specific data is imported.

The inputs for this code are the stagger angle, the camber angle and
the chord of every section of the blade and the basic airfoil profile which is
used for the particular design (uncambered). The code implements these
features in every section of the blade and stacks all the sections into the
center of gravity. The output is a text file (profile.curve) that can be used for
the later stages. Another important output, as it is shown in Figure 4.4, which
is a 3D plot with the final sections of the blade, which again helps the designer
to decide if the geometry produced is an acceptable design.

53 | P a g e
 Fan Design Tools

Figure 4.4 3D plot of blade sections (output from MATLAB code)

4.3 Lewis software

Another tool that was used for the creation of the blade geometry,
which is an alternative to the above mentioned MATLAB code, is a code that
was developed by Lewis, 1996. The program for the section profile generation
is called STACK and it can only be used in a DOS environment. For this
purpose it is required to use a DOS emulator program (like DOSBOX or
equivalent).

The inputs for this software is the camber angle, the position of
maximum camber, the stagger angle, the chord, the profile thickness and the
maximum thickness position. There is a limited selection of airfoil profiles that
can be used, but there is an option to use a user defined profile. For this
project the inputs were the results from the McKenzie’s low fidelity 1-D
simulation procedure that was presented in the previous Chapter 3. Figure 4.5
depicts the Graphical user interface of STACK code.

54 | P a g e
 Fan Design Tools

Figure 4.5 GUI of program STACK

The output of STACK is a file with the 2-D (x,y) coordinates of the
blade or the vane for each section. In these results is very easy to add the
third coordinate z, which is the radii of every profile. Figure 4.6 illustrates the
two dimensional graphical representation of 4 blade sections stacked at the
center of gravity of each section.

Figure 4.6 2D display of the 4 blade profiles (STACK program)

55 | P a g e
 Fan Design Tools

The STACK code has a few disadvantages that make its use limited:

a. The maximum number of sections is 11, although this number is

enough for preliminary design

b. There are limited available airfoil profiles.

c. The results are 2-D, however, as it was explained above, the

third coordinate can easily be added using EXCEL.

d. The base coordinates of the available profiles are not many and
the profiles especially at the leading and the trailing edges were not perfect.
The final points that define the profile of every section are 41 which is a
relatively small number of points.

The above mentioned limitations didn’t affect the coordinates

generation and the blades that were designed using STACK code had
acceptable shapes for a preliminary design. STACK was used successfully for
the profile creation in many design cases of this project.

4.4 ANSYS Turbogrid

ANSYS TurboGrid is a powerful tool that is focused on rotating
machinery analysis. Its main purpose is the mesh creation, while preserving
the underlying geometry. These meshes, which are separate files, are used in
the ANSYS workflow to solve blade passage flow cases. The main input for
this software is the blade geometry which is made up of three files: the
“profile.curve”, the “hub.curve” and the “shroud.curve”. These three files
provide the coordinates that define the geometry of the blade, the hub and the
shroud respectively.

The user can select the basic features of the mesh and then trim it in
order to achieve the most appropriate mesh for a particular case. Figure 4.7
illustrates a stator vane that was created from the coordinates files that were
loaded

56 | P a g e
 Fan Design Tools

Figure 4.7 Stator vane in ANSYS Turbogrid

4.5 GAMBIT
GAMBIT is also a meshing tool; however the geometry must be
created in GAMBIT before the meshing. Its purpose is the geometry (not CAD
geometry) and the mesh creation of surfaces and bodies. Domains without
rotating parts can be easily designed and meshed with GAMBIT. GAMBIT is
used to create geometry and meshes for CFD calculations with FLUENT,
which is not developed particularly for rotating components, although rotating
machinery cases can be solved with FLUENT. The non rotating domains do
not require the complexity of Turbogrid, so they can be created using
GAMBIT. GAMBIT needs practice and guidance in order to acquire quick
results, but generally it is friendly user software. GAMBIT was used for the
geometry and mesh creation of the domain that is illustrated in Figure 4.8.

57 | P a g e
 Fan Design Tools

Figure 4.8 Induct domain in GAMBIT

GAMBIT development stopped after ANSYS bought FLUENT. There

are no newer versions and updates for this software. Alternative software for
meshing is ANSYS Meshing and ICEM.

4.6 ANSYS CFX-Pre

ANSYS CFX-Pre is the physics and solution parameters definition pre-
processor for ANSYS CFX-Solver. The meshes are imported into CFX-Pre
after being produced in various mesh generators and the physical models are
selected for the CFD simulation. The file that is produced is the input file for
the CFX-Solver. ANSYS CFX-Pre is used to assembly the domains, set the
physics and the boundary conditions and finally set the case before the
solver. CFX Pre is a powerful tool which can help the user to create a
simulation model relatively quickly by using a specific option for
turbomachinery cases. The practice and the guidance are mandatory for the
user in order to acquire the necessary experience. Figure 4.9 illustrates a
four domain assembly where the boundary conditions are set.

58 | P a g e
 Fan Design Tools

Figure 4.9 Flow domains in ANSYS CFX-Pre

4.7 ANSYS CFX-Solver

CFX-Solver is the CFD solver and allows the user to control the CFX-
Solver interactively and view information about the emerging solution. CFX-
Solver can be operated from a command line as an alternative and this can
be particularly useful for batch mode operations CFX-Solver uses the output
of the CFX-Pre.

Figure 4.10 GUI of ANSYS CFX-Solver

59 | P a g e
 Fan Design Tools

4.8 ANSYS CFX-Post

CFD-Post is designed to allow easy visualization and quantitative

analysis of the results of the CFD simulations. It has a graphical user interface
that includes a viewer panel in which all graphical output from CFD-Post is
plotted. In CFX-Post there is a “Turbo” option in the same manner as it is
found in CFX-Pre. The turbo workspace improves and speeds up post-
processing for turbomachinery simulations. The analysis of the results
requires extensive experience by the user and of course guidance in order to
spot the good and the weak point of the design and find the ways to improve
or to implement changes to the fan.

The following Figure 4.11 illustrates the blade to blade view of the
relative velocity flow field of a fan stage. The visualization of the flow helps the
designer to identify any problems in the design or to confirm a successful
design.

Figure 4.11 Blade to Blade view of the relative velocity field

The chart of Figure 4.12 depicts the pressure distribution over the
pressure and suction sides of the blade. The nearly automated production of

60 | P a g e
 Fan Design Tools

such plots simplifies considerable the analysis task.and contributes to the

speeding up of the design cycle.

Figure 4.12 Pressure distribution across the blade (pressure/suction side)

4.9 Other MATLAB codes

The tools that were used for this project are presented in this chapter.
During the course of this project a few more MATLAB codes relative to fan
design were developed.

The first code can calculate the air angles in a vaneaxial fan and
produce useful diagrams in order to understand the performance of a
cascade. The following Figures are the results of the first program run for a
vaneaxial fan with hub to tip ratio = 0.68, φ=0.25 and ψ=0.31.

61 | P a g e
 Fan Design Tools

Figure 4.13 shows the pitch to chord ratio across the radii for the rotor
and the stator. The variation of the pitch to chord ratio is high in the rotor and
as a result the blade has to be tapered. A tapered blade has a smaller chord
as one moves from hub to tip. In contrast there is a small variation of the pitch
to chord ratio at the stator.

Figure 4.13 Variation of pitch to chord ratio across the radii

In Figure 4.14 the relative velocities angles are illustrated and the swirl
velocity angle at the exit of the blade as well. The variation of the air angles
(beta1, beta2) across the radii is in a good range according to Figure 2.4. The
swirl angle at the exit of the blade is quite high and this proves that the use of
downstream vanes improve the efficiency, because they convert the swirl
velocity to static pressure rise.

62 | P a g e
 Fan Design Tools

Figure 4.14 Variation of relative velocity angles across the radii

Figure 4.15 depicts the variation of the reaction across the radii is
illustrated. The range of reaction is from 0.76 at the hub to 0.89 at the tip. The
reaction range are inside the literature limits 0.7-0.9 (McKenzie, 1997).

Figure 4.15 Variation of stage reaction across the radii

63 | P a g e
 Fan Design Tools

Figure 4.16 shows the distribution of the flow and work coefficients
across the radii. Both values decrease as we move from hub to tip.

Figure 4.16 Variation of work and flow coefficient across the radii

One more output of this MATLAB code is also a table with values for
the design variables. The input data appear first and then the variation of the
coefficients and reaction across the radii.

Another MATLAB code was created in order to help in the creation of

the profile coordinates for the NACA65 airfoil series. The inputs of this code
are the camber and the maximum position of the camber and the output is a
file with the coordinates of the profile.

A similar MATLAB code was also developed for the extraction of the lift
and drag coefficient of every airfoil for different Reynolds numbers. The inputs
are the camber, the position of the maximum camber and the Reynolds
number and the output is the Cl and the Cd and the angle of attack
respectively.

4.10 ANSYS-Mechanical
The mechanical integrity of the new designs is examined using the
Static-Structural module of ANSYS. The geometry can be created or imported

64 | P a g e
 Fan Design Tools

and then meshed. The boundary condition and the loading conditions which
can be thermal or mechanical gradients are set. The outputs of the software
are the deformation and the stress field which are very useful in order to
estimate the mechanical integrity of the fan. The visualization of the results
that is illustrated in Figure 4.17 is another important feature of the software,
because the designer can review the results more efficiently.

Figure 4.17 Blade deformation for normal operating conditions for aluminum 6061-T6

65 | P a g e
 McKenzie’s Method Validation

5. McKenzie’s Method Validation

5.1 Introduction
In this Chapter an attempt is made to simulate the performance of a fan
that was designed according to McKenzie’s method. The target is then to
compare McKenzie’s method results with the output of a CFD analysis.

As it was mentioned in the Chapter 3 the McKenzie method is the most

appropriate method for the design of the new fan. The preliminary design of
an industrial fan is presented by McKenzie, 1997.

The preliminary design of an industrial vaneaxial fan is analyzed step

by step and the target of this analysis is to evaluate the results using CFD.
According to McKenzie’s example the design procedure starts with the
specifications of volume flow, pressure drop, size and rotational speed and
the outcome of the process is the blade and geometry features of the
vaneaxial fan.

The concept of the method comparison using CFD is simple. The blade
and geometry features which were the outcome of McKenzie’s preliminary
design will be used to create the geometry of the fan for the high fidelity 3-D
simulation. After creating the geometry, CFD can be used to solve the flow
field and acquire the performance of the fan that came from the high fidelity 3-
D simulation. The final step is to investigate if the high fidelity 3-D simulated
performance of the fan meets the requirements that were used for the
preliminary design of the fan.

The results of this investigation will show how reliable is McKenzie

method and if a fan can be designed efficiently according to this method.
When it comes to the final conclusion it is going to be taken into account that
McKenzie’s procedure refers to a preliminary fan design and it is not a final
design. Many simplifications and assumptions were made during the
procedure which affected the performance of the fan.

5.2 McKenzie’s example

The specifications are usually requirements of the client. The inputs for
this example are the following:

1. Volume flow : 5 m3/s

66 | P a g e
 McKenzie’s Method Validation

2. System Duct diameter : 0.75m

3. Duct System pressure drop : 500Pa
4. Fan rotational speed : 1500rpm

Many assumptions need to be made in order to obtain a proper

solution for the preliminary fan design:

1. Diffusion efficiency : 80%

2. Static pressure recovery : 80%
3. Total to total pressure rise efficiency : 85%
4. Tip clearance : 2% of blade span

A few more considerations and assumptions were made during the

procedure in order to improve the design like the approximation of the number
of the blades and the vanes, the taper of the blade etc.

The final result was the design features of the fan which are the rotor
blade geometry and the stator vane geometry. In particular these features are
the following:

1. Hub to Tip ratio

2. Number of Blades/Vanes
3. Airfoil specifications
4. Blade metal angles
5. Camber angle for every profile
6. Stagger angle for every profile
7. Chord length for every profile

All the above geometry features are necessary in order to build the
model of the vaneaxial fan.

5.3 Geometry creation

The definition of the geometry was the first step for the problem set up.
The next step is the creation of the geometry of the flow region and the whole
the computational domain for the CFD calculations.

The geometry of the blade and the vane was created using the Lewis
program STACK that was described in Chapter Design tools. The inputs for
this software is the camber angle, the position of maximum camber, the
stagger angle, the chord, the profile thickness and the maximum thickness
67 | P a g e
 McKenzie’s Method Validation

position of 4 profiles from hub to tip. All these data were the results of
preliminary design in McKenzie’s example. The airfoil that was used was the
C4. Figure 5.1 illustrates the 4 section of McKenzie's blade that are stacked in
their center of gravity

Figure 5.1 2D display of the 4 blade profiles (STACK program)

The output of STACK is a text file with the 2D coordinates (X, Y) of the
4 profiles from hub to tip. The third dimension Z, which is the span, was added
to create the 3D design of the blade and the vane. Two more text files were
written one for the hub and the other for the shroud.

The final output of this procedure was the definition of the coordinates
of the blade, the vane and their hubs and shrouds with 6 text files (3 for the
blade and 3 for the vane). A small modification of these text files was made in
order to have the appropriate format for the next step.

The same coordinates came up using the MATLAB code that was
described in Chapter 4. The output of the MATLAB code is visualized in figure
5.2 For this case it was the data from Lewis software that was used.
68 | P a g e
 McKenzie’s Method Validation

Figure 5.2 3D plot of the blade using MATLAB code

It is important to mention that the final design has one domain for the
rotor and one domain for the stator, which means that no domains for the inlet
and the outlet were defined separately. In McKenzie’s example there was no
reference for the existence of an inlet domain or an outlet domain, so the
design for the CFD calculations has two main domains. Although no separate
domains were defined for inlet and outlet, Turbogrid creates a small domain
for the inlet and the outlet in order to be easier to set the boundaries and the
properties at inlet and outlet.

The next step is the creation of the blade and vane and their domain.
ANSYS Turbogrid is the software that was used to accomplish this task. For
the creation of the blade and the blade domain the 3 text file were imported in
Turbogrid. The bspline option was selected for the curve type and the surface
type. The number of the blades was entered manually in Turbogrid and a few
more parameters were adjusted properly. The tip clearance which consists a
very important parameter was set in Turbogrid as well. The blade is illustrated
in figure 5.3 along with the hub and the shroud.

69 | P a g e
 McKenzie’s Method Validation

Figure 5.3 Representation of the blade, the hub and the shroud in Turbogrid

5.4 Meshing
ANSYS Turbogrid was used for the meshing of the domains as well.
During the pre-process stage the grid generation constitutes one of the most
important steps after creation of the geometry. It is required to subdivide the
flow domain into smaller subdomains which are not overlapping each other.
The flow physics are solved within the domain geometry that has been
created. The result is the generation of the mesh which is made from small
cells.

The method for the topology definition of the grid was the H/J/C/L-Grid
and an O-Grid was included with a width factor of 0.5. For the tip topology the
H-Grid Not matching was selected. For the first approach, a fine mesh size
was selected with a target of 200.000 nodes and the option normalized was
selected for the near wall size specification. The impact of the size of the
mesh was later investigated to prove the independence of the result from the
size of the mesh. The stator domain mesh in Turbogrid is depicted in figure
5.4

70 | P a g e
 McKenzie’s Method Validation

Figure 5.4 Representation of the stator vane mesh

5.5 Physics and Fluid properties - Boundary Conditions

The two domains (rotor domain and stator domain) were imported to
ANSYS CFX-Pre in order to continue the solution of the fluid flow by defining
the physics and the boundary conditions. The option “turbo mode” was
selected for the set up, which is a special mode, allowing users to set up
turbomachinery simulations. One of the main advantages of turbo mode is the
automatic setup of the boundary regions and conditions.

The physics selection is the first step to set up the case. The simulation
is a viscous steady state simulation. There is no heat transfer and the viscous
model that was selected is the k-epsilon model. The k-e turbulence model
offers a good compromise between numerical effort and computational
accuracy and uses always standard wall functions for the near wall treatment.

It is very important for the setup the definition of the inputs at the inlet
and the outlet. CFX-Pre offers many choices for the boundary template. For
this particular simulation the Total Pressure was selected at the inlet and the
Mass Flow at the outlet. The static pressure at the outlet and the velocity at
the inlet are part of the solution, whereas total pressure at inlet and mass flow
71 | P a g e
 McKenzie’s Method Validation

at outlet are inputs. This selection was made , because according to the CFX
manual it gives is a robust simulation model. The inlet that draw flow in from
the atmosphere uses a Total Pressure = 0 boundary condition (e.g. as an
open window) while the reference pressure at the domain in 101325Pa. This
choice can produce more accurate solution, because with this choice a 200Pa
increase in the pressure is for example from 300Pa to 500Pa (more than 50%
increase), whereas in the case where the reference pressure is 0 and total
pressure is 101325, a 200Pa increase would be for example from 101325 to
101525 (less than 1% increase).

The interface of the two domains was defined as “stage”. The stage
model performs a circumferential averaging of the fluxes through bands on
the interface. Stage model allows steady state predictions to be obtained for
multi-stage machines. This model allows steady state predictions to be
obtained for multi-stage machines. The Stage model usually requires more
computational effort than the Frozen Rotor model to converge.

The boundary conditions (inlet, outlet, periodic, wall, symmetry) were

added automatically and after checking that this automatic procedure is
correct, the total pressure, the reference pressure, the relative pressure, the
mass flow and the rotational speed were set respectively to the domains and
the subdomains. The simulation domain in CFX-Pre and the boundary
condition are shown in figure 5.5

Figure 5.5 Simulation domain (rotor and stator)

72 | P a g e
 McKenzie’s Method Validation

5.6 Solver
Another advantage of the “turbo mode” is the auto initialization. The
basic settings for the solver control are the convergence control and the
convergence criteria. The choice for convergence control was the maximum
iterations for this simulation (1000 iterations) and the choice for the
convergence criteria was the RMS (residual target 10-6). Other choices for the
solver were the high resolution for the advection scheme and the first order for
the turbulence numerics.

5.7 Results report

The main result from the high fidelity 3-D simulation is the total
pressure rise at the fan. As it was mentioned earlier the mass flow, the
rotational speed and the inlet pressure were input data according to the
boundary template that was selected, whereas velocity at inlet and the
pressure at outlet are results of the simulation.

The simulation converged in less than 300 iterations as it is illustrated

in figure 5.6, which means that the model was very well defined and the mesh
was appropriate and successful.

Figure 5.6 Convergence plot

73 | P a g e
 McKenzie’s Method Validation

The total pressure at the outlet of the stator domain was found to be
368Pa. The total pressure at the inlet of the inlet of the rotor was set at 0 so
the total pressure rise at the fan was 368Pa.

An interesting part of the results except the pressure magnitude is the

fluid flow around the blade and the vane. It is important to check the flow in
areas where separation and other disturbance is expected. ANSYS CFX-Post
was used to acquire all the above data.

CFX-Post has the turbo option which is very useful for the quicker
manipulation of the results. Very representative 3D plots of the flow can be
easily generated with the “blade to blade” option. The position that were
selected to represent the flow field are the hub (0.1 of the span) the middle
(0.5 of the span) and the tip (0.9 of the span)

Figure 5.7 Velocity contours at 0.1 of the span

74 | P a g e
 McKenzie’s Method Validation

Figure 5.8 Velocity contours at 0.5 of the span

Figure 5.9 Velocity contours at 0.9 of the span

75 | P a g e
 McKenzie’s Method Validation

From the above figures it is clear that the design fulfills the
requirements of the specification. The flow is smooth around the blade and
the vane and there is no sizeable separation. The downstream guide vanes
are beneficial in terms of fluid flow at the outlet. They were used in order to
remove the high rotational component of the velocity at the exit of the rotor.
There is no whirl at the outlet flow so all the rotational component of the
velocity was converted to useful pressure rise.

This is even more obvious at the next figure where the velocity vectors
are plotted at the 0.1 of the span. The high angle of the velocity at the inlet of
the vane can be observed, but after the flow passes the vane, the flow is
horizontal.

Figure 5.10, Velocity Vector Diagram at 0,1 span

5.8 Validation-Verification of the results

For a CFD solution to be credible, it is important to check the results in

many ways. This step can be accomplished through the verification and
validation procedure. Verification and validation procedures are the means by
which a CFD solution can be properly assessed through distinct procedures.
76 | P a g e
 McKenzie’s Method Validation

Verification can be defined as a process for assessing the numerical

simulation uncertainty and when conditions permit, estimating the sign and
magnitude of the numerical simulation error and the uncertainty in that
estimated error [Tu et al, 2008]. A major part of the verification process is to
carry out a mesh dependency study. By this means it is investigated the
impact of the mesh size (coarse or fine) to the results.

For this particular case the performance of two more models was
investigated. In the first model the mesh of its domain had the half size (100k)
of the original mess (200k). In the second case the mesh of its domain had
the double size (400k) of the original one (200k). The total pressure rise in all
three cases is illustrated on the next table

Table 5.1 Mesh Dependency Study

All Domain mesh size (elements) Total Pressure Rise (Pa)

262,954 366
504,364 368
995,320 372

The above values of total pressure rise for different mesh sizes prove
that the results are not mesh dependent.

The validation can be defined as a process for assessing simulation

model uncertainty by using benchmark experimental data and when
conditions permit, estimating the sign and magnitude of the simulation
modeling error itself [Tu et al, 2008].

In this particular case there are no experimental data, so the results

cannot be validated this way. The only results that can be compared with the
high fidelity 3-D simulation results are the results from the low fidelity 1-D
simulation. As it was mentioned earlier in this Chapter the geometry of the fan
that was used at the high fidelity 3-D simulation is based on McKenzie fan
preliminary design method and the initial requirements of volume flow, total
pressure rise, fan size and rotational speed.

The performance of the fan is expected to be similar with the

performance that McKenzie’s fan has. Since the mass flow, the rotational
speed and the size are inputs at the high fidelity 3-D simulation the only
variable that can be checked is the total pressure rise.

McKenzie’s fan was designed in order to provide a total pressure rise

of 500Pa. The total pressure rise from the CFD simulation is 372Pa which is
25% less than the expected performance. This divergence is mainly due to
77 | P a g e
 McKenzie’s Method Validation

the fact that the high fidelity 3-D simulation is a more reliable method because
it calculates the flow field more precisely than the low fidelity 1-D simulation
and it is more thoroughly examined in the next paragraph.

5.9 Results evaluation

The divergence of the high fidelity 3-D simulation results with the
results of the he low fidelity 1-D simulation is significant and a few comments
can be made about this:

a. The total pressure rise from the high fidelity simulation is less
that the low fidelity one and this can be due in the following reasons:

1. The high fidelity 3-D simulation model includes more loss

factors than the low fidelity 1-D model.

2. The low fidelity 1-D model takes into consideration less

loss factors.

b. Since the flow is smooth in the high fidelity 3-D simulation model
that means that there are no extra losses that are due to the flow (separation)
and the model fulfills the flow requirements. Therefore the blade angles are
proper and manipulate the flow effectively.

c. The high fidelity 3-D simulation model takes into consideration

the turbulence of the flow by using the k-e turbulence model. Turbulence
causes energy losses especially close to the boundary layer and these losses
are not taken into account at the low fidelity process.

d. The 3D effects at hub and tip which have serious impact at the
performance of an axial fan, because they increase the losses (Bass, 1987).

e. The interaction of two consecutive blades, because of the small

distance between them, which restrict the development of the full lift forces
and thus decrease the performance (Bass, 1987).

f. The tip clearance effect cannot be simulated in the low fidelity 1-

D simulation. The tip clearance as it was discussed earlier has a major
negative impact in the fan performance

78 | P a g e
 McKenzie’s Method Validation

5.10 Conclusion
In this chapter an attempt was made to create a high fidelity 3-D
simulation model based on a known preliminary fan design, which was
calculated according to McKenzie’s blade design method, in order to
investigate if it is viable to design vaneaxial fans following this method.

There is a significant divergence in the performance between the low

fidelity 1-D simulation model and the high fidelity 3-D simulation model. This
divergence is due to many reasons that were explained in paragraph 5.9, but
the fact is that a vaneaxial fan can be designed according to this method even
if the low fidelity 1-D simulation performance does not entirely match the high
fidelity 3-D simulation model which is closer to the real performance. The low
fidelity 1-D simulation is used to acquire an initial solution that can be later
improved using the CFD analysis tools.

The target of this Thesis is to design of a vaneaxial fan with a specific

performance and a better efficiency than the performance of an existing
model. The above procedure proved that it can be designed, but it should be
taken into account that the performance of the high fidelity 3-D simulation
model which is going to be closer to the performance of the real model, will be
worse than the performance obtained from the low fidelity 1-D solution.

79 | P a g e
 New Designs

6. New Designs
6.1 Introduction
In this Chapter the process leading to the design of a fan with an improved
efficiency compared to the datum vaneaxial fan is described. The dimensions of the fan
and the fan characteristics can be obtained from the specifications. The requirements
are the volume flow, the size and the pressure rise. The dimensions and the
characteristics of an axial fan consist of the following:

1. Outside diameter
2. Hub to tip ratio
3. Number and width of blades/vanes
4. Blades/vanes angles
5. Blade/vanes curvature
6. Fan speed

The design tools that were presented in Chapter 4 and the design procedure that
was analysed in Chapter 3 and validated in Chapter 5 were used to accomplish this
task. The CFD simulations are the main part of this project. The fan design theory is
applied in order to build the simulation model, but only after the post processing of the
results from the CFD simulations, can be it concluded if the design theory was correctly
applied to design a fan with a fan with a respectable performance.

Three new designs were developed and each design derived from the
experience of the previous one.

The analysis procedure for each new design takes place as follows:

1. Geometry definition

2. Domain creation

3. Mesh Generation

4. Domain Assembly

5. Physics set up, boundary definition

6. Case set up

7. Results for the design point

80 | P a g e
 New Designs

8. Design Optimization

9. Results for the optimized design at design point

10. Off design performance

11. Tip clearance effect

6.2 Fläkt-Woods fan

The datum fan is a vaneaxial machine that derives from a tubeaxial fan. The fan
is driven by a motor at 1500rpm and it has a 9 blades impeller. The original tubeaxial
fan had a total efficiency of 69%. The efficiency of the tubeaxial fan was relatively low
and the requirement for improved efficiency arouse from this. The use of downstream
guide vane improves the efficiency, because part of the swirl velocity at the impeller’s
exit is converted into useful static pressure.

Fläkt-Woods added a vane stage with 11 vanes to this particular tubeaxial fan in
order to improve the efficiency of the fan. The improvement in the efficiency is
illustrated in Figure 6.1

tubeaxial
80JMR Performance
vaneaxial
90

80

70

60
efficiency (%)

50

40

30

20

10

0
0 2 4 6 8 10
Volume Flow Rate (m3/s)

Figure 6.1 Tubeaxial vs vaneaxial efficiency in variation with volume flow (Fläkt-Woods)

81 | P a g e
 New Designs

The guide vanes that were added in the tubeaxial fan remove the rotational
component of the velocity and convert some of the dynamic pressure into useful static
pressure. This explains the fact that in the vaneaxial fan the pressure rise is higher for
the same volume flow than in a tubeaxial fan. The increased pressure rise result in the
higher efficiency as well for the vaneaxial fan.

The maximum efficiency was increased by 16.6% and it reached 79.5% at a

volume flow 7.51 m3/s. The point with the highest efficiency was picked as design point
for the new designs. An improvement in efficiency has the following practical meaning:

1. The fan has an improved performance (increased ventilation) in the same

environment (increased volume flow under the same pressure rise) without the
increased energy consumption penalty.

2. The same fan can operate with higher load a different environment (increased
pressure rise) without reducing the volume flow (ventilation level) and without the
increased energy consumption penalty.

The increased pressure rise of the vaneaxial fan is illustrated in Figure 6.2

80JMR Performance tubeaxial

vaneaxial
700

600

500
Fan Total Pressure (Pa)

400

300

200

100

0
0 2 4 6 8 10
Volume Flow Rate (m3/s)

Figure 6.2 Tubeaxial vs vaneaxial total pressure rise in variation with volume flow (Fläkt-Woods)

82 | P a g e
 New Designs

6.3 New Design

The point with the highest efficiency of the Fläkt-Woods vaneaxial fan was
selected as design point. The performance of the given vaneaxial fan at the selected
point provides the requirements for the new design as follows:

1. Volume flow : 7.51 m3/sec

2. Total Pressure rise : 506 Pa

3. Rotational speed : 1500rpm

4. Efficiency > 79,5%

5. Tip Diameter : 0.8m

6. C4 airfoil for the rotor and the stator

Fläkt-Woods stated that the design of the fan can start from a blank paper,
however the above inputs are enough in order to design a fan as it was discussed in the
Design Procedures Chapter and even more design data can be extracted from the
given model in order to start the new design. The given design has 9 blades and 11
vanes, so this configuration can be a starting point for the new design.

The following assumptions were made as well, in order to make the initial
calculations for the new design. These assumptions are based on the assumptions that
were made from McKenzie and Osbourne in their preliminary designs. Real data for
diffusion efficiency and pressure recovery factors cannot be retrieved because they are
proprietary information of the fan manufacturers.

1. Diffusion efficiency : 80%

2. Pressure recovery factor : 85%

There are more requirements for the new fan relating to the extreme operating
conditions under which the fan should operate. According to
http://www.pveuk.com/axial_fans.htm a ventilation fan should be capable to operate
under 400oC for 2 hours. This requirement concerns the mechanical integrity of the fan,
however a relatively large tip clearance of at least 3mm is necessary, which is about
1.5% of the blade span.

83 | P a g e
 New Designs

6.4 Design1
6.4.1 Blade Geometry

Design1 is the first attempt to design a brand new fan from customer’s
requirements. The basic requirements were apposed in the previous paragraph, so the
target for Design1 is to have 9 blades and 11 vanes as the given fan has.

The fan geometry has to be determined first. The McKenzie preliminary design
method is a trial an error method as was discussed in Chapter 5. The first assumption
concerns the hub to tip ratio. The hub to tip ratio consist an essential assumption
because it has a direct effect on the axial velocity. McKenzie, 1997 and Bleier, 1998
have similar procedures to check the choice of the hub to tip ratio. McKenzie calculates
the ratios of ∆Η/Va^2 and Va/U at the hub and plot the values at the diagram of Figure
6.3

Figure 6.3 Work Coefficient vs flow coefficient diagram (McKenzie,1997)

The point should be as close as possible to the line of Cp=0.5 in order to have a
de Haller number of 0.7. A 0.52 hub to tip ratio provides a point close to the line of
Cp=0.5.

The annulus is divided to 10 sections and after fixing the hub to tip ratio, the air
angles of the relative velocities can be calculated for every section.

84 | P a g e
 New Designs

Figure 6.4 Relative Velo

ocities – Angle
es (McKenzie
e,1997)

The rotation
T nal speed was
w increased to 1650 rpm in ord der to mainttain the num
mber
of blade
es to 9 witho
out increasing the blad
de chord more
m than 15cm. The blade
b chord
d is a
comprom mise betwe een the rotational spe
eed and thee number of
o blades. If the rotational
speed was
w 1500rp pm the bladde chord shhould have been incre
eased to 211cm at the hub,
which is not accepttable.

T next ste
The ep is to calcculate the stagger
s ang
gle of everyy section.

ta
an am= (tan a1+tan a2)//2 (6.1)

ta
an ζ = tan am -0.15 (6.2)

The blade angle

T a variess with the incidence and
a the cam mber is a fuunction of blade
b
angle an
nd stagger angle. The following chainc of equuations con
nsist an iterative proce
edure
to deterrmine the proper incidence value for the hub. A value v for th
he incidencce is
assumed and then n the S/C ratio is ca alculated. The
T value of incidencce that gives a
reasonaable S/C rattio is finally selected fo
or the hub section.
s

β1 = α1-i (6.3)

θ=2(β1-ζ) (6.4)
(

β2=β1-θ (6.5)

δ=α2-β2 (6.6)
(

S/C=δ/(1.1+0.31θ)3 (6
S 6.7)

The peak effficiency ca

T an be achie
eved when the angle of incidencce lies betwween
ο ο
+5 to -1
10 (McKenzie,1997) anda the S/C C ratio for good
g efficiency can be
e checked using
u
the diagrams of Fig
gures 6.5 and 6.6.

85 | P a g e
 New Designs

Fiigure 6.5 Blad

de geometry for
f fixed air angles (McKen
nzie, 1988)

Figure
e 6.6 Efficienccy contours (M
McKenzie, 19
988)

The same procedure

T p c
can be followed for the tip secttion in orde
er to checkk if a
reasona
able inciden
nce and S/C
C ratio can be
b obtainedd.
86 | P a g e
 New Designs

Having the S/C ratio fixed at the hub, the S/C ratio at the tip can be calculated
assuming that is proportional to the radius and assuming that the blade chord at the tip
is ¾ of the blade chord at the hub. As it was discussed in Chapter 3 narrower tip will
unfavourably affect the efficiency of a fan, however the centrifugal loading of the blade
is less and the mechanical strength will be increased. The blade height can also be
calculated and after assuming a reasonable aspect ratio of 1.5, the blade chord can be
calculated as well.

All the calculations and the final design of the blade features are summarized in
Table 6.1

Table 6.1 Final rotor blade design (Design1)

HUB TIP
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.8
S/C 0.935 1.098 1.26 1.423 1.585 1.748 1.91 2.073 2.235 2.398
a1 60.36 62.7 64.72 66.48 68.03 69.39 70.6 71.67 72.64 73.52
a2 47.85 53.39 57.61 60.93 63.61 65.82 67.67 69.25 70.61 71.79
tanam= 1.431 1.642 1.847 2.049 2.247 2.443 2.637 2.829 3.02 3.21
(tana1+tana2)/2
tanζ=tanam -0.15 1.281 1.492 1.697 1.899 2.097 2.293 2.487 2.679 2.87 3.06
ζ (rad) 0.908 0.98 1.038 1.086 1.126 1.16 1.188 1.214 1.236 1.255
STAGGER 52.03 56.16 59.49 62.22 64.51 66.44 68.09 69.53 70.79 71.9
CAMBER 26.67 21.71 18.59 16.7 15.69 15.33 15.49 16.1 17.12 18.59
(0.5-0.31*(S/C)^(1/3))
β1=2θ+ζ 65.36 67.02 68.79 70.58 72.35 74.1 75.84 77.58 79.35 81.2
incidence i= (a1-β1) -5 -4.32 -4.06 -4.09 -4.33 -4.72 -5.25 -5.91 -6.71 -7.68
β2=β1-θ 38.69 45.31 50.2 53.87 56.66 58.77 60.35 61.48 62.23 62.61
deviation δ=α2-β2 9.16 8.077 7.411 7.061 6.956 7.051 7.324 7.765 8.38 9.187
deflection e =a1-a2 12.51 9.316 7.113 5.551 4.413 3.566 2.922 2.425 2.035 1.724
am (rad) = 0.961 1.024 1.075 1.117 1.152 1.182 1.208 1.231 1.251 1.269
am (deg) = 55.06 58.65 61.57 63.98 66.01 67.74 69.23 70.53 71.68 72.7
Cl (from Osbourne) =
2*S/C*(tana1- 0.699 0.676 0.65 0.623 0.596 0.571 0.547 0.525 0.504 0.484
tana2)*cosam
CHORD 0.155 0.146 0.139 0.133 0.129 0.126 0.123 0.12 0.118 0.116

The same procedure can be followed to acquire the geometry of the vane. A few
comments for the vane design as follows:

87 | P a g e
 New Designs

1. The a4 angle at the first 3 sections was selected 3o 2o 1o degrees respectively.

The a3 angle is higher at the first sections and the higher the angle the harder is to
straighten the flow. So, the penalty of a small swirl is permitted at the 3 first sections of
the vane (30% of the total span) in order to avoid the higher penalty of separation is
more difficult to straighten the flow.

2. The vane was initially assumed to have the average chord of the blade in
order to have a reference point for the design

3. A constant chord was selected for simplicity of the vane design

4. The airfoil of the vane is C4, as the blade airfoil

5. The geometry of the vane was adjusted taking into account that the stator has
11 vanes.

The final design of the stator vane is illustrated in Table 6.2

Table 6.2 Final stator vane design (Design1)

HUB TIP
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.8
a3 33.13 30.62 28.43 26.52 24.83 23.33 22 20.8 19.72 18.74
a3 (rad) 0.578 0.534 0.496 0.463 0.433 0.407 0.384 0.363 0.344 0.327
a4 3 2 1 0 0 0 0 0 0 0
a4 (rad) 0.052 0.035 0.017 0 0 0 0 0 0 0
tanam=
0.352 0.313 0.279 0.249 0.231 0.216 0.202 0.19 0.179 0.17
(tana3+tana4)/2

tanζ=tanam -0.15 0.202 0.163 0.129 0.099 0.081 0.066 0.052 0.04 0.029 0.02
ζ (rad) 0.2 0.162 0.129 0.099 0.081 0.066 0.052 0.04 0.029 0.02
STAGGER 11.45 9.278 7.376 5.682 4.651 3.757 2.975 2.286 1.673 1.126
CHORD 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127
S= 0.119 0.131 0.143 0.155 0.168 0.18 0.192 0.204 0.216 0.228
S/C= 0.936 1.032 1.127 1.223 1.319 1.415 1.511 1.607 1.703 1.799
CAMBER 48.38 42.45 37.86 34.34 29.1 24.55 20.58 17.08 13.97 11.19
β1=ζ+θ/2 35.63 30.5 26.31 22.85 19.2 16.03 13.27 10.83 8.657 6.72
i=a3-β1 -2.51 0.115 2.125 3.669 5.631 7.297 8.73 9.973 11.06 12.02

6.4.2 Blade Design

The definition of the blade and vane geometry is the first step to design a
vaneaxial fan. The next step is to examine if a fan with this geometry can perform the
88 | P a g e
 New Designs

way it was designed for. This can only achieved by simulating the performance of this
fan.

First of all the blade and the vane geometry of each section should be converted
to 3D coordinates. For this purpose the data of Table 6.3 and Table were imported to
the STACK program in order to acquire the 2D coordinates of the 10 sections of the
blade and the vane. Similar result is obtained by using the MATLAB code that was
developed for this cause.

Table 6.3 Design1 Rotor

SECTION 1 2 3 4 5 6 7 8 9 10
CAMBER 26.7 21.7 18.6 16.7 15.7 15.3 18.6 16.1 17.1 18.6
CHORD 0.155 0.146 0.139 0.133 0.129 0.126 0.123 0.120 0.118 0.116
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
STAGGER 52.0 56.2 59.5 62.2 64.5 66.4 68.1 69.5 70.8 71.9

Table 6.4 Design1 Stator

SECTION 1 2 3 4 5 6 7 8 9 10
CAMBER -48.4 -42.4 -37.9 -34.3 -29.1 -24.6 -20.6 -17.1 -14.0 -11.2
CHORD 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127 0.127
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
STAGGER -11.4 -9.3 -7.4 -5.7 -4.7 -3.8 -3.0 -2.3 -1.7 -1.1

The results were manipulated in an excel file and the final output was 2
profile.curve file, which contain the 3D coordinates of the blade and the vane.

The coordinates of the hub and the shroud for the rotor and the stator were
defined in separate files hub.curve and shroud.curve. The definition of the hub and the
shroud derived from the hub to tip ratio, the tip diameter, the tip clearance and the
domain length. The length of the rotor and the stator domain was defined from
experience and from the given fan.

6.4.3 Meshing

These 3 .curve files were imported to ANSYS Turbogrid to create the domain
geometry and visualize the result. A fine size mesh selected with 200000 elements for
each domain. Figure 6.7 illustrates the blade and the vane of Design1

89 | P a g e
 New Designs

Figure 6.7 Blade and Vane of Design1 in ANSYS Turbogrid

Two more domains were designed to simulate the inlet and the outlet of the fan.
The induct domain prepares the free flow to enter the annulus of the rotor domain and
the outduct discharges the flow after the stator domain. These two domains were
designed in GAMBIT and in ANSYS Design Modeler as well. The length of these
domains and their shape were chosen from previous models and experience.

The geometry in GAMBIT has several steps points, edges, faces, volume and
each step takes significant amount of time. The meshing has the same steps as well. In
Figure 6.8 the mesh of the induct domain is illustrated.

Figure 6.8 Induct domain from GAMBIT

90 | P a g e
 New Designs

The ANSYS platform is easier to use; the geometry is designed in ANSYS

Design Modeler which is a CAD like software and the mesh is generated automatically
using the ANSYS Meshing software. The mesh of the ouduct domain is illustrated in
Figure 6.9.

Figure 6.9 Outduct domain from ANSYS Meshing

6.4.4 CFX-Pre

The next step is to assembly the domains in ANSYS CFX Pre. ANSYS CFX-Pre
is the physics-definition pre-processor for ANSYS CFX-Solver. The 4 separate domains
(induct, rotor, stator, outduct) are imported into CFX-Pre and the physical models are
selected for the CFD simulation. ANSYS CFX-Pre is used to assembly the domains, set
the physics and the boundary conditions and finally set the case before the solver. The
simulation model with all the domains assembled is illustrated in Figure 6.10.

The Turbo mode was selected for the assembly of the meshes and the boundary
settings. The boundary template that was selected for this simulation was Total
Pressure at the inlet and Mass Flow at the outlet. The Total Pressure was set to 0 Pa,

91 | P a g e
 New Designs

because the reference pressure of the flow field was already selected to 101325Pa and
the volume flow of the design point was converted to Mass Flow.

Other settings for the CFX-Pre are as follows:

1. Turbulence model: k-ε model

2. Interface model: General connection, stage

3. Pitch change: Automatic

4. Mesh Interface: GGI

5. Convergence criterion: rms residual 10-4

The case was written to a ".def" file to be ready for the solver.

Figure 6.10 Simulation Model in CFX-Pre

6.4.5 Results

The Total Pressure rise was found 487Pa and the Static Pressure rise 479Pa.

92 | P a g e
 New Designs

The efficiency was calculated 78.5% using the following formula:

∗
= (6.8)
( ∗ ∗ )
Total Pressure rise (∆Ρ), Static Pressure rise (∆p) were the result from the
subtract of the pressure in Outduct's outlet minus the pressure at the Induct's inlet. The
torque was the sum of the torque at the rotor blade and the rotor's hub. All the values
were calculated using the "function calculator" of CFX-Post.

The next step is to check the flow at the rotor and the stator. The "blade to blade"
view was selected at CFX-Post, in order to visualize the flow. The flow is examined at
the hub (0.1 of the span), the middle (0.5 of the span) and the tip (0.9 of the span). The
best way to visualize the flow is the velocity variation. The velocity range is set to a
range from 0 m/s to 80 m/s. The results are plotted in Figures 6.11, 6.12 and 6.13
respectively.

Figure 6.11 Rotor-Stator Velocity Contour Plot at 0.1 span

93 | P a g e
 New Designs

Figure 6.12 Rotor-Stator Velocity Contour Plot at 0.5 span

Figure 6.13 Rotor-Stator Velocity Contour Plot at 0.9 span

94 | P a g e
 New Designs

Figure 6.14 Rotor-Stator Velocity Vector Plot at 0.9 span

High separation is noticed at the tip area which has to be examined more
thoroughly. The velocity vector diagram of Figure 6.14 depicts the separation better. It is
obvious that the vane geometry cannot handle the flow which is not able to follow the
curvature of the suction area and separates very quickly causing losses.

The next step is to examine the total area that the separation problem exists. In
Figures 6.15, 6.16, 6.17 the velocity contour is plotted at 0.8, 0.7, 0.6 of the span
respectively. The range of the velocity is fixed to a range 0-80 m/s, therefore the velocity
contours for each span can be immediately compared.

95 | P a g e
 New Designs

Figure 6.15 Rotor-Stator Velocity Contour Plot at 0.8 span

Figure 6.16 Rotor-Stator Velocity Contour Plot at 0.7 span

96 | P a g e
 New Designs

Figure 6.17 Rotor-Stator Velocity Contour Plot at 0.6 span

The separation starts from 0.6 of the span and worsens as the span is getting
greater. This situation should be investigated and the best way to start is the vane
geometry that is depicted in Table 6.2 (incidence line)

i=a3-β1 -2.51 0.115 2.125 3.669 5.631 7.297 8.73 9.973 11.06 12.02

The examination of Table 6.2 shows that the incidence after the 0.6 of span is
greater than 7. Figure 6.5 shows that there is a vertical decrease in the efficiency for
incidence angles greater than 5, so the high separation after 0.6 of the span is not a
surprise.

97 | P a g e
 New Designs

6.4.6 Design1 optimisation

The high separation at the vane and the resulting losses affect the efficiency of
the fan and an attempt to improve the design of the vane is necessary in order to
achieve better performance.

The root cause of the high separation is the high angle of incidence at the
sections above the 0.6 of span. The angle of incidence depends on the stagger and the
camber angle.

i = a3-β1 (6.9)

β1 = ζ + θ/2 (6.10)

The fact that the camber and the stagger can improve the flow over the vane can
be extracted from Figure 6.14. It is obvious that an increase in the stagger and in the
camber of the vane airfoil will provide a less aggressive environment to the flow and the
separation can be reduced.

After many trial and error attempts the stagger angle was increased properly
from section 3 and above (in order to have smoother transition to higher stagger angles)
and the camber angle was adjusted automatically in order to provide the zero swirl at
the fan outlet (a4=0). The new vane geometry is illustrated in Table 6.6

Table 6.5 Optimized Stator

SECTION 1 2 3 4 5 6 7 8 9 10
CAMBER -50.7 -44.8 -41.0 -39.2 -36.4 -33.7 -31.7 -30.1 -28.4 -26.6
CHORD 0.164 0.164 0.164 0.164 0.164 0.164 0.164 0.164 0.164 0.164
DIAMETER 0.416 0.4587 0.5013 0.544 0.5867 0.6293 0.672 0.7147 0.7573 0.8
STAGGER -13.2 -10.9 -9.1 -7.7 -7.1 -6.5 -6.0 -5.7 -5.3 -4.9
i=a3-β1 -2.9 -0.2 1.2 1.5 1.7 2.0 2.1 2.0 2.0 2.3

The new geometry of the vane requires the redesign of the vane from the
beginning. The vane was redesigned using the procedure that was presented in
chapters 6.4.2, 6.4.3, 6.4.4 and 6.4.5. The improvement in design is depicted in Figures
6.18, 6.19

98 | P a g e
 New Designs

Figure 6.18 Rotor-Stator Velocity Contour Plot at 0.6 and 0.7 span (optimized)

Figure 6.19 Rotor-Stator Velocity Contour Plot at 0.8 and 0.9 span (optimized)

The separation decreased significantly in the improved design and the

performance was affected positively. The improvement of stagger angle and camber
increased the performance; the Total Pressure rise was found 529Pa, the Static
Pressure rise 529Pa and the efficiency was calculated 79.3%

Table 6.6 Improved Design Performance

Design1 Optimized Design1 % Improvement

Total Pressure Rise (Pa) 487 529 8.6%
Static Pressure Rise (Pa) 479 525 10.4%
Efficiency 78.1% 79.3% 1.5%

99 | P a g e
 New Designs

6.4.7 Validation-Verification of the results

For a CFD solution to be credible, it is important to assess the results through

verification and validation procedures.

A major part of the verification process is the mesh dependency study. By this
means it is investigated the impact of the mesh size (coarse or fine) to the results.

For this particular case the performance of three more models was investigated.
The original mesh of the rotor and the stator has 200k elements. The cases that were
selected for the mesh dependency study have 50k, 100k and 400k elements. The induct
and the outduct meshes remained the same. The total pressure rise in all four cases is
illustrated in table 6.7

Table 6.7 Results Verification

Rotor mesh (thousands Stator mesh size (thousands Total Pressure Rise
elements) elements) (Pa)
55.6 59.7 530
114.9 119.5 524
235.7 237.8 529
462.8 462.8 534

The above mentioned mesh dependency study that was carried out verify that
the results of the CFD simulation are not grid dependant, as the divergence in the
values of Total Pressure Rise is about 1%.

6.4.8 Off-design performance of Improved Design1

The off designed performance off the fan should be examined as well. The fan is
supposed to be installed in a certain environment (e.g. tunnel) where the operating
conditions are well established. However the different operating altitude or the specific
weather conditions affect the ambient conditions and a flat performance curve is
desirable.

100 | P a g e
 New Designs

The off design performance of the fan was examined at the same range of
volume flow that was used for the off design performance off the given fan. The volume
flow of the off design cases is illustrated in Table 6.8.

Table 6.8 Off design cases

1 2 3 4 5 6 7 8 9 10 11 12
volume
flow 9.17 8.77 8.52 8.32 7.89 7.51 7.03 6.61 6.12 5.66 5.23 4.37
3
(m /s)

The off design performance of Design1 in variation with the off design
performance of the given fan is depicted in Figures 6.20 and 6.21. Figure 6.20 depicts
the variation of volume flow with total pressure rise and Figure 6.21 depicts the variation
of volume flow with the efficiency.

It is interesting to notice that there is no characteristic stall area at Design1. As

the volume flow decreases Design1 there is no stall at the fan, but an area where the
total pressure does not increase but remains the same. After that area and as the
volume flow decreases, the total pressure increases again as it was described in
Chapter 2.

101 | P a g e
 New Designs

Volume Flow vs Total Pressure rise

800
700
Total Pressure Rise (Pa)

600
500
400 given fan

300 design1

200
100
0
3,00 5,00 7,00 9,00
Volume flow (m3/s)

Figure 6.20 Volume flow in variation with total pressure rise

The efficiency curve has similar shape with the efficiency curve of the given fan,
however the efficiency given fan is higher at lower volume flow.

The maximum efficiency of the final model was about the same with the
efficiency of the given fan and a new fan with an increased efficiency should be
designed.

Volume Flow vs Efficiency

100%
90%
80%
70%
efficiency

60%
50%
given fan
40%
design1
30%
20%
10%
0%
3,00 5,00 7,00 9,00
Volume Flow (m3/s)

102 | P a g e
 New Designs

Figure 6.21 Volume flow in variation with efficiency

6.5 Design2
6.5.1 Design2 concept

Design2 is the second attempt to achieve a fan design with better efficiency and
performance. The basis for design2 was design1 and some new ideas incorporated in
the new design.

The velocity contour plots showed that the flow in the rotor was smooth and
uniform across the pressure surface and the suction surface. This fact means that
maybe the blade loading is not as high as the geometry of the blade can stand and an
increase in the blade loading would increase the efficiency of the fan.

The blade loading parameter is the ∆Η/U2 ratio which depends on the fan work
per unit of mass flow ∆H=∆P/eff*ρ. If the efficiency of the fan increases, more energy
will be converted to useful fan work and the blade loading will increase. Therefore the
idea is to keep the main design features the same and increase the design efficiency in
order to increase the blade loading. The CFD simulation will prove if the increase in
efficiency of the analytical solution can be incorporated in the simulation model

The main design features remain the same as follows

1. Hub to tip ratio : 0.52

2. Rotational speed : 1650rpm

The design efficiency was increased to 90% and the blade loading was
significantly increased. The problem that arises from the increased blade loading is the
required increase in the blade chord in order to increase the fan work. The chord length
that was calculated is not acceptable and the only solution is to increase the number of
blades to 11.

The main new features of design2 are the following:

1. Increased design efficiency

2. Increased number of blades and vanes (11 blades and 13 vanes)

103 | P a g e
 New Designs

6.5.2 Blade Geometry-Design-Meshing- Optimisation

The same procedure that was used in Design1 simulation was followed for
Design2 as well.

The McKenzie method that was described in the Chapter 6.4.1 was followed to
obtain the blade geometry. The geometry of the blade and the vane are illustrated in
Table 1 and Table 2 of Appendix “A” respectively.

The new design data, that are illustrated in Table 3 and Table 4 of Appendix “A”,
were imported to STACK software in order to generate the 2D coordinates for the blade
and the vane and the third dimension was added in an excel file. Three new ".curve" file
were generated for the rotor and for the stator as well (profile.curve, hub.curve,
shroud.curve)

Design2 has 11 blades and 13 vanes, so the domains have different pitch
angles. The different pitch angles for the blade and the vane domain are incorporated
automatically by choosing the new number of blades and vanes at the machine data.
The domains of the rotor and the stator were created in Turbogrid with the same
procedure as Design1. The induct and the outduct domains have different pitch angles
and they were designed from the start.

The assembly of the domains in CFX-Pre has the same settings for the physics,
the turbulence model, the boundary conditions and the interfaces.

The efficiency of the new design was 80.5% which is higher than the maximum
efficiency of the given fan. The experience acquired from Design1 was used for the
examination of Design2. The vane performance was checked first and it was shown that
the shape of the vane produced the same separation problems as the vane of design1.

The same optimisation method was followed to decrease the separation and
increase the efficiency of design2. The stagger angle of the vane increased as it was
increased during the optimisation of Design1. The optimized vane geometry is
illustrated in Table 5 of Appendix “A”.

The vane was redesigned and re-meshed. The new stator domain was
assembled with the induct domain, the outduct domain and the rotor domain and a new
simulation took place.

104 | P a g e
 New Designs

6.5.3 Results

The Total Pressure rise for the design point was found 550Pa and the Static
Pressure rise 544Pa.

The efficiency was calculated 82.1% using the following formula (6.8):

∗
=
( ∗ ∗ )
The 82% efficiency is a satisfactory result, because it is 3% higher than the
efficiency of the given fan.

The visualization of the flow is depicted in the following velocity contours of

Figures 6.22, 6.23 and 6.24 for the hub the middle span and the tip. The flow follows the
curvature of the blade and the vane across the span without serious separation at any
point. The modification of the stagger angle worked well and the result is depicted from
the flow of the following Figures. The range of the velocity is fixed to a range 0-80 m/s,
therefore the velocity contours for each span can be immediately compared.

105 | P a g e
 New Designs

Figure 6.22 Rotor-Stator Velocity Contour Plot at 0.1 span

Figure 6.23 Rotor-Stator Velocity Contour Plot at 0.5 span

106 | P a g e
 New Designs

Figure 6.24 Rotor-Stator Velocity Contour Plot at 0.9 span

The velocity vector plot shows that the stator performs as designed. According to
the design data the velocity exit angle is very small at the hub as it is shown in Figure
6.25 and zero at the middle of the vane as shown in Figure 6.26. At the tip area the
velocity exit angle is zero according to the design data, however a small angle is noticed
in Figure 6.27 due to the 3D effects that the tip clearance generates.

Figure 6.25 Stator Velocity Vector Plot at 0.1 span

Figure 6.26 Stator Velocity Vector Plot at 0.5 span

107 | P a g e
 New Designs

Figure 6.27 Stator Velocity Vector Plot at 0.9 span

6.5.4 Turbulence model comparison

Design2 has better performance comparing with the performance of the given fan
and Design1 fan. The simulation was carried out using the k-ε turbulence model. The
same simulation carried out with different turbulence model. The turbulent model that
was used was the k-ω model which does not overestimates the turbulence as the k-ε
model does (Tu et al, 2007). The losses due to turbulence are less and the performance
of the fan is better as shown in the computational result of Table 6.7. The same
conclusion for the turbulence models k-ε and k-ω was drawn from Strohmeyer, 2009 as
well.

Table 6.7 Design point performance fοr k-ε and k-ω turbulence models

Turbulence model Ps rise (Pa) Pt rise (Pa) Total Efficiency

k-ε 526 532 82%
k-ω 557 560 83.2%

6.5.5 Verification

For this particular case the performance of three more models was investigated.
The original mesh of the rotor and the stator has 200k elements. The cases that were
selected for the mesh dependency study have 50k, 100k and 400k elements. The induct
and the outduct meshes remained the same. The turbulence model was the k-ε.

The differences in the computational results for different size meshes are below
1%. The results can be found in Table 6 of Appendix “A”.

6.5.6 Off-design performance

The off design performance of Design2 is examined for the different volume flows
of Table 6.8 as well. The off design performance of Design2 in variation with the off
design performance of the given fan is depicted in Figures 6.28 and 6.29. Figure 6.28
depicts the variation of volume flow with total pressure rise and Figure 6.29 depicts the
variation of volume flow with the efficiency.

The total pressure curve has the same shape with the total pressure curve of
Design1. There is no significant stall region, which means less noisy operation in lower
volume flows.
108 | P a g e
 New Designs

Volume Flow vs Total Pressure rise

800
700
Total Pressure Rise (Pa)

600
500
400 given fan
300 design2
200
100
0
3,00 5,00 7,00 9,00
Volume flow (m3/s)

Figure 6.28 Volume flow in variation with total pressure rise

The efficiency of Design2 is better than the efficiency of the given fan in the
design point area, however it has the same efficiency for high volume flows and lower
efficiency for low volume flows. Although the fan has quieter operation in lower volume
flows, the efficiency of the fan low volume flow is lower than the corresponding
efficiency of the given fan.

Volume Flow vs Efficiency

100%
90%
80%
70%
Efficiency

60%
50% given fan
40% design2
30%
20%
10%
0%
3,00 5,00 7,00 9,00
Volume flow (m3/s)
109 | P a g e
 New Designs

Figure 6.29 Volume flow in variation with total efficiency

6.6 Design3
6.6.1 Design3 concept

Design3 is the final attempt to achieve a fan design that has better efficiency and
performance than the given fan has. The lessons learned from design1 and design2
were a good guide for the new design. A better performance has already been achieved
with Design2.

The following conclusions were drawn from the design analysis of a free vortex
vaneaxial fan:

1. Increasing the design efficiency has the following effects:

a. Lower angle a3 of the relative velocity V3 at the rotor exit

b. Higher angle of incidence

c. Smaller blade chord

2. Decreasing the rotational velocity has the following effects:

a. Higher angle a3 of the relative velocity V3 at the rotor exit

b. Smaller angle of incidence

c. Higher blade chord

3. Decreasing the hub to tip ratio has the following effects:

a. Lower angle a3 of the relative velocity V3 at the rotor exit

b. Higher angle of incidence

c. Higher blade chord

The hub to tip ratio of Design1 and Design2 has the same value 0.52. In Chapter
3.10 the influence of hub to tip ratio was discussed. The rotational speed of Design1
and Design2 was the same as well 1650rpm.

Design3 was the field to experiment with the rotational speed and the hub to tip
ratio. The number of blades and vanes is the same with Design2 11 and 13

110 | P a g e
 New Designs

respectively. The final design is a compromise between the rotational speed, the design
efficiency, the hub to tip ratio and the blade chord.

The above design features (angle a3, angle of incidence i, blade chord) were
selected because they have immediate impact to the final design.

A relative high blade chord is not desirable because it will add material to the
blade, increase the rotor domain length, increase the blade mass

The higher the angle a3 is, the higher is the difficulty to design a stator that can
straighten the flow (a4=0)

The high angle of incidence leads to low efficiency; therefore it must be kept into
a certain range of values.

Design3 target is to investigate the performance of a fan with lower hub to tip
ratio and lower rotational speed. The rotational speed was reduced to 1600rpm and the
hub to tip ratio reduced as well to 0.48, which is closer to the hub to tip ratio of the given
fan. The design efficiency was increased in order to reduce the length of the blade
chord

6.6.2 Blade Geometry-Design-Meshing- Optimisation

The same procedure that was used in Design1 and Design2 simulation was
followed for Design3 as well.

The McKenzie method that was described in the Chapter 6.4.1 was followed to
obtain the blade geometry. The geometry of the blade is illustrated in Table 1 of
Appendix “B”.

Design3 has 11 blades and 13 vanes as Design2 has, however the hub to tip
ratio is smaller, so the domains have different size. The domains of the rotor and the
stator were created in Turbogrid with the same procedure as Design1. The induct and
the outduct domains have different size and they were designed from the start as well.

The assembly of the domains in CFX-Pre had the same settings for the physics,
the turbulence model, the boundary conditions and the interfaces.

The original Design3 was optimized according to the experience from the
optimization of Design1 and Design2. The stagger angle of the vane was increased as it

111 | P a g e
 New Designs

was increased during the optimisation of the previous designs. The optimized vane
geometry is illustrated in Table 2 of Appendix “B”.

The new design data, that are illustrated in Table 3 and Table 4 of Appendix ”B”,
were imported to STACK software in order to generate the 2D coordinates for the blade
and the vane and the third dimension was added in an excel file. Three new ".curve" file
were generated for the rotor and the stator as well (profile.curve, hub.curve,
shroud.curve). The hub.curve file has different data comparing with the hub.curve file of
Design1 and Design2 because of the different hub to tip ratio.

6.6.3 Results

The Total Pressure rise for the design point was found 483Pa and the Static
Pressure rise 478Pa.

The efficiency was calculated 79.5% using the formula (6.8):

∗
=
( ∗ ∗ )
The 79,5% efficiency is not a satisfactory result, because it is the same efficiency
with the given fan.

The visualization of the flow is depicted in the following velocity contours of

Figures 6.30, 6.31 and 6.32 for the hub the middle span and the tip. A severe
separation can be noticed at 0.9 of the span. The modification of the stagger was
proven insufficient and further optimization of the design can take place.

112 | P a g e
 New Designs

Figure 6.30 Rotor-Stator Velocity Contour Plot at 0.1 span

Figure 6.31 Rotor-Stator Velocity Contour Plot at 0.5 span

113 | P a g e
 New Designs

Figure 6.32 Rotor-Stator Velocity Contour Plot at 0.9 span

6.6.4 Verification

Design1 and Design2 results were verified with the mesh dependant study that
was carried out. Design3 was the product of the same procedure; therefore it is believed
that no extra verification is needed for Design3.

6.6.5 Off-design performance

The off design performance of Design3 is examined for the different volume flows
of Table 6.8 as well. The off design performance of design3 in variation with the off
design performance of the given fan is depicted in Figures 6.33 and 6.34. As it was
noticed in Design1 and Design2 there is no visible stall area according to the plot of
Figure 6.33. There is only an area where the total pressure remains constant or
increases in a lower rate

114 | P a g e
 New Designs

Volume Flow vs Total Pressure rise

800

700

600
Total Pressure Rise (Pa)

500

400 given fan

300 design3

200

100

0
3,00 4,00 5,00 6,00 7,00 8,00 9,00 10,00
Volume flow (m3/s)

Figure 6.33 Volume flow in variation with total pressure rise

The efficiency curve has the same shape with the efficiency curve of the given
design. Generally is slightly lower comparing with the efficiency of the given fan, except
the design point area where the efficiency of Design3 is the same with the efficiency of
the given model.

115 | P a g e
 New Designs

Volume Flow vs Efficiency

100%
90%
80%
70%
60%
Efficiency

50% given fan

40% design3

30%
20%
10%
0%
3,00 4,00 5,00 6,00 7,00 8,00 9,00 10,00
Volume flow (m3/s)

Figure 6.34 Volume flow in variation with total efficiency

6.7 Tip clearance effect

The tip clearance was set to 1.5% which is about 3mm for a 200mm blade. It is
interesting to investigate the performance of the fan with higher tip clearance. Bass
1987 states that the behavior of a fan with higher tip clearance can be completely
different.

The main advantage of an increased tip clearance is the safety margin that it
offers in a hostile working environment (high temperature). The main disadvantage is
the deterioration of the fan performance. The performance of Design1, Design2 and
Design3 with higher tip clearance was simulated to prove the deterioration of the
performance. The new tip clearance was set to 3% which is about 6mm for a 200mm
blade.

The new tip clearance was incorporated to the rotor domains of each new
design, while the rest of the domains remained the same. The new cases have the
same settings and the new simulation carried out for the design point and the off design
points. The results for the design point are illustrated in Table 6.8.

116 | P a g e
 New Designs

Table 6.8 Design point performance with tip clearance 1.5% and 3.0%

Design TP (1.5%) TP (3%) % η (1.5%) η (3%) %

Design1 529 422 -20.2% 79.1% 72.6% -8.2%
Design2 550 438 -20.3% 82.1% 74.3% -9.5%
Design3 493 408 -17% 78.0% 73% -6.4%

The comparison of the off design performance data of new Designs with nominal
tip clearance (1.5%) and high tip clearance (3%) are illustrated in Appendix “C”.

The off design total pressure rise comparison for the two different tip clearances
is illustrated in Figure 6.35 for the 3 new designs.

Volume Flow vs Static Pressure rise

800

700
Static Pressure Rise (Pa)

600 design1
design1 3%
500 design2
design2 3%
400
design3

300 design3 3%

200
4,00 5,00 6,00 7,00 8,00 9,00
Volume flow (m3/s)

Figure 6.35 Volume flow in variation with static pressure rise

The comparison of the behavior of the case with the nominal and the high tip
clearance shows that the difference in behavior is similar with the difference in behavior
that Bleier, 1997 noticed in his experiments and are depicted in Figure 6.36.

From Figure 6.36 it can be noticed that for low and high volume flows the effect
of tip clearance to the pressure rise is lower comparing with the effect of tip clearance at

117 | P a g e
 New Designs

volume flows close e to the deesign point area. The same patttern can be
e noticed in
n the
on results in Figure 6.3
simulatio 35

Figu
ure 6.36 Effec
ct of tip cleara
ance at static pressure rise
e and brake horse
h power (Bleier, 1997)

118 | P a g e
 New Designs

The off design efficiency comparison for the two different tip clearances is
illustrated in Figure 6.37 for the 3 new designs.

The efficiency with higher tip clearance is lower in all the range of volume flow.
However in low volume flows the comparison is more difficult, due to the fact that the
effect of tip clearance to efficiency is lower. The lower effect of tip clearance to
efficiency is depicted in Figure 6.38. The value of efficiency at for both high tip
clearance and nominal tip clearance converge to zero for zero volume flow. Therefore
as the volume flow decreases the effect of tip clearance to efficiency is less that in
higher volume flow efficiency

Volume Flow vs Efficiency

90%

85% design1

80% design1 3%
75%
Efficiency

design2
70%

65% design2 3%

60% design3
55%
design3 3%
50%
4,00 5,00 6,00 7,00 8,00 9,00
Volume flow (m3/s)

Figure 6.37 Volume flow in variation with total efficiency

The tip clearance effect can be summarized as follows:

a. Volume flow decreases slightly, because a backflow is expected which is

proportional to the tip clearance

b. Maximum Static Pressure decreases considerably, because the flow

resists to the pressure increase and escape backwards.

119 | P a g e
 New Designs

c. Powe er consump ption decrea

ases, but no ot as much
h as the pre
essure, beca
ause
less volu
ume flow is
s pressurize
ed in a lowe
er static pre
essure.

d. Mech hanical efficciency decrreases conssiderably, because

b the
e pressure loss
is consid
derably higher than the e power co
onsumption decrease.

e. Small increase in the noise

e level at th
he free delivery, beca
ause of the flow
that movves through
h the tip.

Figu
ure 6.38 Effecct of tip cleara
ance in efficie
ency and noisse level (Bleie
er, 1997)
120 | P a g e
 Mechanical Integrity

7. Mechanical Integrity

7.1 Introduction
The mechanical integrity of axial fans depends on the loading of the fan
and the available manufacturing materials. The loading of industrial and
ventilation axial fans is relatively low, because the rotational velocity that
generates the centrifugal loading is relatively low. The variety of the available
materials for fan manufacturing is large and the designer has many options for
the selection of the material.

However, in a few cases the loading of the fan is quite high because there
are requirements for operation in hostile environment. A hostile environment can
be a tunnel where a ventilation fan is required to operate continuously in case of
a fire or a corrosive environment where the strength and the durability of the
material of the fan deteriorate. In cases like this a more careful study should be
take place in order to validate the safe operation of the fan under the adverse
conditions.

In this chapter an attempt will be made to investigate the mechanical

integrity of Design2, which was selected as the design with the better efficiency
and performance. The part of the fan that will be examined is the blade, which
has the higher loading. The blade is the only part of the fan that rotates and the
centrifugal force has significant magnitude. The stress distribution at the blade
and the blade deformation will be estimated for normal operation and for
operation in high temperature condition.

The tip clearance was set at 2.9mm which is 1.5% of the 192mm blade
height. The target is to estimate the deformation of the blade in order to examine
if the tip clearance is large enough to ensure the safe operation.

7.2 Blade stressing

The steady stresses in a rotating blade are negligible comparing to the no
steady stresses. The sources of no steady stresses in fan blades arise from the
fan operation and the working environment. The major sources are as follows
according to Haslam, (2007):

121 | P a g e
 Mecha
anical Integrity

1. Centrrifugal loading that accts at all se

ections of th
he blade, produced
p b
by
the ac
ction of inertia.

2. Gas bending
b moment thatt is produce
ed when th
he fluid passses acrosss
the bllade and ch
hange in mo
omentum and
a pressurre take placce.

3. Bending momen nt: when thhe centrifug gal loading acts at a point whicch
does not lie abovve the centtre of the ro
oot datum section.
s

4. Shear load whicch arises fro

om the cen
ntrifugal unttwisting of the
t blade or
o
as pressure
the ga e.

5. Therm
mal stress that
t arise in
n a hostile environmen
e nt.

The blade stressing g in a fan deepends on the volume

e flow rate. Cory (2005
5)
carrie
ed out tessts on a fa an with Gottingen airfoil bladees. The ressults of the
expeeriments are
e illustrated
d in figure 7.1.
7

Figure 7.1
7 Pressure and
a stress disstribution aga
ainst volume flow
f (Cory, 20
005)

122 | P a g e
 Mecha
anical Integrity

The regu ular line illustrates th he reverse e rotation and

a the bo old line the
forwaard rotationn line. As thhe volume flow decrea ases, the pressure
p inccreases and
the stress
s incre
eases as well.
w Howeve er, an interresting poin
nt is that th
he maximum m
stresss is not coiincident witth the stall point.
p

7.3 Centrifugal stre

esses
The centtrifugal loadd at the blad
des is due tot the rotattion of the cascade
c and
the mass
m that th
he blades have.
h The in
nertial field in aero engines, whe ere the blade
operates, comm mits more than 50% and often 80% of th he material strength to t
overccome the stress
s (Hasslam, 2007)). However in low spe eed fans the e centrifuga
al
loadiing has low wer order ofo magnitud de comparring with ae ero engine es. A simple
mode el of centriffugal loadin
ng is shownn in figure 7.2.

Figure
e 7.2 Centrifu
ugal loading (H
Haslam, 2007
7)

CF = mass * rcg * ω2 (7.1)

The mas
ss of this bla
ade can be written as:

Mass = Density
D * Crross-sectional area * height
h

Thus,

CF= p * A * h * rcg * ω2 (7.2)

mponent with cross-se

For a com ectional are
ea of A the stress
s is:

123 | P a g e
 Mechanical Integrity

σcf= ρ * h * rcg * ω2 (7.3)

It is obvious that the stress is proportional to the square of the rotational

speed, which is proven to be a crucial factor for the mechanical integrity of the
blade. In low speed fans the importance of the rotational speed is downgraded.

The tapering of the blade is an important factor for the centrifugal loading.
A tapered blade has less mass at the tip, which decreases the total inertia of the
blade. Haslam (2007) mentioned that for a 25cm blade the centrifugal loading
can be reduced about 40% if the area of the tip is half the area of the hub,
comparing with a blade with constant blade area for a given rotational speed.

The centrifugal loading can be calculated by dividing the blade into

sections, and calculate the centrifugal force for each section where the rotational
speed is assumed constant. The individual loading of each section is divided with
the cross sectional area and the result is the individual centrifugal stress. The
sum of the individual forces gives the centrifugal loading and the sum of the
individual stresses gives the centrifugal stress.

7.4 Fluid Forces

The change in velocity from the blade leading edge to the blade trailing
edge and the pressure difference across the blade generate stresses and
moments. The fluid forces act in both axial and tangential direction. The gas
loads are considered individually as follows:

1. Axial Bending moment due to Pressure change.

The pressure distribution around a blade can be found by dividing the total
pressure change by the number of the blades. The bending moment can be
found by multiplying the total pressure force around a blade by the distance from
the section under consideration to the centre of gravity of the pressure element.

2∗ ∗ ∗ ∗
= (7.4)

The bending moment can be found by multiplying the total pressure force
around a blade by the distance from the section under consideration to the centre
of gravity of the pressure element.

124 | P a g e
 Mecha
anical Integrity

2∗ ∗ ∗( )∗
= (7.5)

2. Ax
xial Bendin
ng and Tangential Ben
nding mom
ment due to Momentum
m
chan
nge.

e that is prroduced by the change of mome

The force entum is the
e product of
o
the mass
m multip
plied by thee rate of ve
elocity chan nge. The be
ending moment due tot
mommentum chaange is usua ally so smaall that it is ignored.

Specific mass flow = mass flow/unit ann

nular area = m,

= (7.6)
∗( )

∗ ∗2∗ ∗ ∗
( )= (
(7.7)

∗ ∗2∗ ∗ ∗
( )= (7.8)

Figure 7.3 Bending Moments fro

om gas forcess (Haslam, 2007)

125 | P a g e
 Mechanical Integrity

The bending moments in an airfoil are illustrated in figure 7.3

3. Bending Stress due to Centrifugal loading.

The bending moment is the product of inertia load of the section and the
displacement from the section centre of gravity to the line through the centre of
the root

Figure 7.4 Bending Moments from centrifugal loading (Haslam, 2007)

7.5 Stresses due to Thermal Gradients

The stresses produced by the increase ambient temperature within a
blade can be very destructive and are often much greater than any of the other
forms of stress. An industrial fan is required to operate in case of fire for a certain
period for safety reasons. In cases like this the loading is very complex because
there is a combination of centrifugal loading, gas loading and thermal loading.

The maximum operation temperature for a ventilation fan is about 400oC

(Cory, 2005) and it is required to operate in this temperature for 2 hours as it is
stated in the specifications of a emergency smoke extract fan that is already in
the market (http://www.pveuk.com/axial_fans.htm).

126 | P a g e
 Mechanical Integrity

The materials that are used for industrial fans have often poor properties
comparing with the advances alloys that are used in aero engines in order to
reduce the manufacturing cost. Therefore the industrial fan material properties
are affected significantly in high temperatures as it is illustrated in figure 7.5.

Figure 7.5 Strength and Young modulus variation with temperature for typical carbon steel (Cory, 2005)

The important material properties for ventilation fan blades are the
ultimate tensile strength (100-500MPa), the elongation at break (15-20%), the
ductility (4%), creep resistance (400oC), the corrosion resistance, the impact
strength, the fatigue strength.

7.6 Blade loading simulation software

An attempt is made to simulate the blade loading for various combinations
of mechanical loading and thermal loading. The blade that was used for the
simulation is the blade from Design2.

The software that is used for the simulation is the Static Structural
component of ANSYS Workbench. The procedure to carry out the simulation was
as follows:

1. Geometry: The geometry of the blade that was used for the simulation
was exported from Turbogrid.

2. Meshing: the meshing of the blade is a straight forward procedure

especially for a simple geometry like a single blade. For mechanical applications

127 | P a g e
 Mechanical Integrity

and there are three options for the mesh quality: coarse, medium, fine. A fine
mesh was selected for the simulation with 6226 elements.

3. Inertia loading: These loads act on the entire system. The setting of the
rotational velocity and the axis of rotation are set. The rotational speed was set to
172.8 rad/sec which is 1650rpm. The centre of rotation was set to 208mm which
is the radius of the hub of Design2.

4. Structural loads: The aerodynamic forces and moments are imported

from the CFX simulation of the fluid flow.

5. Structural supports: The constraints that prevent movement on certain

regions. In this case the constraint is the blade hub. The blade hub constraint
represents the connection of the blade with the hub.

6. Thermal loads: The thermal loads which result in a temperature field

causing thermal expansion/contraction in the model. The Thermal loading that
were simulate were 0o C, 50o C, 100o C, 300o C and 400o C.

7.7 Blade loading simulations

Simulations were carried out for steady mechanical loading under different
thermal loading. As it was mentioned before the mechanical loading depends on
the rotational speed and the volume flow. The rotational speed is steady in an
industrial fan and the volume flow that was selected is the volume flow of the
design point. The temperature ranges from 50o C to 400o C.

The stress depends only on the loading from the centrifugal force and gas
forces. The temperature contributes only to the total deformation of the blade and
not to the stress. The target is to calculate the deformation of the blade in the
above mentioned conditions and estimate if the tip clearance is high enough to
ensure the safe operation under high temperature.

The material choice is a challenge for the designer. For this particular
study Aluminum 6061-T6 was used. The first simulation carried out for without
thermal loading in order to calculate the stress distribution and the deformation
for normal operation. Figure 7.6 illustrates the deformation distribution of a fan
blade from aluminum alloy 6061-T6. The maximum deformation is 0.077mm at
the thinnest area of the tip.

128 | P a g e
 Mechanical Integrity

Figure 7.6 Blade deformation for normal operating conditions for aluminum 6061-T6

The deformation is higher at the tip because the centrifugal force highest
value is at the tip according to Haslam (2007) and the thinnest area is more
susceptible to deformation. The deformation pattern is the expected one
according to Naeem et al.

The stress distribution at the blade is shown in figure 7.7 and figure 7.8.

129 | P a g e
 Mechanical Integrity

Figure 7.7 Stress distribution for normal operating conditions for aluminum 6061-T6

Figure 7.8 Stress distribution for normal operating conditions for aluminum 6061-T6

The maximum equivalent stress (von Mises stress) at the blade is

approximately 8 MPa or 1.16 ksi which was expected for an industrial fan with
1650rpm. Osbourne (1979) estimated the equivalent stress for a smaller fan
(60% outside diameter) and with lower rotational speed (960rpm) at 0.7 MPa,
therefore 8MPa for a 0.8m fan with 1650rpm is a logical value.

The next step is to simulate the blade operation in an environment with

higher temperature. The first simulation was carried out for a temperature of 50o
C and the deformation is depicted at figure 7.9.

130 | P a g e
 Mechanical Integrity

Figure 7.9 Blade deformation for operation in 50o C for aluminum 6061-T6

The deformation is 0,18mm and the distribution is different comparing with

the deformation distribution at normal conditions of figure 7.6. This is due to the
fact that the thermal gradient generates uniform deformation at every section of
the blade. The combination of the uniform deformation from the thermal gradient
and the deformation from mechanical loading creates this deformation pattern.
The tip clearance is 2,9 mm so the operation in 50o Celsius is safe as it was
expected.

Figures 7.10, 7.11, 7.12, 7.13 illustrate the deformation of the blade for
temperatures of 100oC, 200oC, 300oC and 400oC respectively. The deformation
pattern in figures 7.10, 7.11, 7.12, 7.13 shows that the deformation due to the
thermal gradient dominates over the deformation due to mechanical loading,
from the temperature of 100oC.

131 | P a g e
 Mechanical Integrity

Figure 7.10 Blade deformation for operation in 100o C for aluminum 6061-T6

Figure 7.11 Blade deformation for operation in 200o C for aluminum 6061-T6

132 | P a g e
 Mechanical Integrity

Figure 7.12 Blade deformation for operation in 300o C for aluminum 6061-T6

Figure 7.13 Blade deformation for operation in 400o C for aluminum 6061-T6

In figure 7.11 the deformation in actual scale can be observed. The black
line shows the blade before the deformation, therefore the value actual
deformation of the blade can be estimated. Figure 7.12 depicts the deformation

133 | P a g e
 Mechanical Integrity

pattern at 300oC. It is obvious that the deformation due to thermal gradient

dominates over the deformation due to mechanical loading. In figure 7.13 the
actual deformation for 400oC can be better observed because it is two times
higher than the deformation in 200oC. The blade deformation for various
temperatures is shown in table 7.1

Table 7.1 Deformation variation with temperature

rotational speed Total Deformation

Temperature( C )
(rad/sec) max (mm)

1 172 22 0.077
2 172 50 0.188
3 172 100 0.463
4 172 200 1.042
5 172 300 1.625
6 172 400 2.208

The diagram in figure 7.14 depicts the deformation variation with

temperature and gives a better reflection of the variation.

Deformation vs Temperature
450
400
350
Temperature C

300
250
200
150
100
50
0
0 0,5 1 1,5 2 2,5
Deformation (mm)

Figure 7.14 Blade deformation variation with temperature under constant mechanical loading for
aluminum 6061-T6

134 | P a g e
 Mechanical Integrity

Another way to calculate the deformation of the blade in various

temperatures is to use the material properties. The material properties of
aluminum 6061 T6 were found on the internet at
http://asm.matweb.com/search/SpecificMaterial.asp?bassnum=MA6061t6.

The deformation of a material can be calculated using the coefficient of

thermal coefficient. The thermal expansion coefficient describes by how much a
material will expand for each degree of temperature increase. It was assumed
that the total deformation is the sum of the deformation due to thermal gradient
dominates plus the deformation due to mechanical loading. The results are
summarized in Table 7.2

Table 7.2 Deformation variation with temperature

Deformation Total Deformation

Coefficient Deformation Differe
Temperat due to deformation max (mm)
of thermal due to thermal nce
ure (C) mechanical (mm) from
expansion gradient (mm) (%)
loading (mm) calculated simulation
22 0 0.000 0.115 0.115 0.078 48.77
50 0.0000236 0.236 0.115 0.351 0.188 86.44
100 0.0000252 0.504 0.115 0.619 0.464 33.56
200 0.0000252 1.008 0.115 1.123 1.042 7.78
300 0.0000252 1.512 0.115 1.627 1.625 0.15
400 0.0000252 2.016 0.115 2.131 2.209 -3.49

The maximum deformation for 400oC operating temperature was

calculated 2.2mm, while the tip clearance is 2.9mm. This means there is a 31%
safety margin which is not necessary enough to ensure the safe operation for
long period. The maximum deformation was 1.62mm at 300oC and this means
there is a 50% safety margin which gives a safety factor of 1.5.

7.8 Creep life calculation

The next step is the attempt to examine the influence of the time to
deformation. The influence of time can be calculated using the Larson Miller
parameter for the estimation of the creep life.

Creep is the progressive deformation of a material under a constant load

at elevated temperature. The strength of materials reduces with increasing

135 | P a g e
 Mecha
anical Integrity

temp peratures. This

T happen ns, mainly, because high
h temperratures resu ult in greate
er
mobiility of disloocations byy the mech hanism of climb and in an incrrease in the
equillibrium concentration n of vaca ancies. Pro olonged exposure
e a elevate
at ed
temp peratures can also produce deformations at grain g bounndaries (rre
crysttallization and
a grain grrowth). The e same com mbination off load and temperaturre
can lead
l to pos
ssible destrructive oxidation with possible inttergranularr penetratio
on
of oxxide.

The phen nomenon ofo creep is dependant to load, te emperature and time of o
expo
osure at the e high tem
mperature. Materials
M h
have resista
ance to cre eep loading
g,
which is described by hom mologous temperatur
t re. Homolog gous temperature of a
erial is the
mate e absolute temperatu ure to absolute melting temperrature ratio o.
Usuaally aero-enngines alloys will cree ep at 0.5*T Tmelting to 0.7*T
0 melting temperature
e.
Therre are three
e stages of creep,
c ed in figure 7.15
as itt is illustrate

Figure 7.15
5 Three phasses of creep (A
Andreadis, 20
009)

The mos st popular of

o the time-temperaturre parameteers is the Larson-Mille
L er
parameter. If th
he Larson-MMiller param
meter is kn
nown as we oad and the
ell as the lo
workking tempe erature of a material, a reliab ble predicte
ed creep life can be b
calcu
ulated.

The equa
ation for La
arson Miller parameterr is:

= ∗( ) (7.9
9)
1000

136 | P a g e
 Mechanical Integrity

Where:

T : operating temperature,

t : the time to failure

C : constant, usually about 20 for most of the applications, but it can vary
according to the conditions.

The Larson Miller parameter can be found at the material properties. From
Kaufman’s (2008) database for high temperature aluminum alloys it was
observed that the magnitude of stress, which in the case that is examined is 8
MPa or 1.16 ksi, is relatively very low for an aluminum 6061-T6. The minimum
value for the stress that was used to create the database was 2 ksi and the
maximum temperature 750o F which is approximately 400o C.

The Larson Miller Parameter that was used to calculate the creep life of
the fan derived from the plot of figure 7.16.

Figure 7.16 Larson Miller parameter for aluminum 6061-T651 for C-=20.3 (Kaufman, 2008)

137 | P a g e
 Mechanical Integrity

The lower the value of the Larson Miller Parameter is, the higher the
expected creep life. A conservative choice for the Larson Miller Parameter is 26.
The choice is conservative because for 1.16 ksi of stress there is no LMP at the
diagram which means that the amount of stress can not cause creep whatever
the temperature.

For LMP=26 and for 400oC which are converted to 673oK, the time to
rupture t is 2.15 * 1018 hours which is practically infinite.

The time that was calculated was time to fracture. From the material
properties it is known that the elongation at break is approximately 17% of the
length. For a 192 mm blade, the elongation at break will be 32,6 mm, while the
tip clearance is 2.9 mm. This means that the deformation will be critical at the
first phase of the creep period. However, and taking into account the infinite time
to failure, creep does not consist a critical factor for the fan safety.

7.9 Conclusions
The preliminary examination of the mechanical integrity of the new
designed fan can be summarized as follows:

1. There is no concern about mechanical integrity when the fan operates

at normal conditions.

2. The safety factor for 300oC was calculated 1.5 and for 400oC it was
calculated 1.25. The 25% safety margin at 400oC is not acceptable and
alternative solution should be introduced.

a. Selection of a material with better resistance in thermal loading.

This will increase the manufacturing cost of the fan

b. Revaluation of the worst case scenario (400oC). If the upper limit

is 300oC there is no need to improve the thermal strength of the fan.

3. Under the worst loading (400oC), there is no risk of blade rupture

(break) due to creep. The total deformation (elongation) when the material fails
(break) was calculated 32.6mm, while the tip clearance is 2.9mm. This means
that a lot before the break, the deformed blade will reach the shroud and the
operation of fan will stop. The time needed for the blade operating at 400oC to
reach the shroud cannot be calculated.

138 | P a g e
 Mechanical Integrity

The mechanical design of arduous duty fans can be extremely complex

because the thermal gradient effect can only be estimated. Modern materials are
not always fully documented and their limitations may be found only through
experience. Nevertheless, the application of principles from strength of materials
and theory of machines can produce acceptable designs.

139 | P a g e
 Conclusions and Recommendations

8. Conclusions and recommendations

8.1 Project Summary
The main objective of this project was the review of low speed design
methodologies and the application of one such approach to the redesign of a datum
geometry. A suitable fan design procedure has been established based on the existing
literature employing the free vortex design concept. The applicability of the design
procedure, and to a certain extent its degree of accuracy, have been established by
using appropriate CFD software for the simulation of the fan performance. Three new
fans have been designed by using the validated design procedure, the existing design
tools and MATLAB codes that have been developed as part of this study. The efficiency
of the second design was higher than the efficiency of the datum fan and the off design
performance has been demonstrated to be satisfactory. The mechanical integrity of the
new fan is acceptable, but the thermal strength has to be improved in order to give an
higher safety factor when operating in a high temperature scenario such as a in the
vicinity of a fire.

8.2 Conclusions
The fan design process is a compromise between the customer requirements
and the aerodynamic and mechanical constraints. As Bass, 1987 states, it is difficult to
design a very bad fan and it is difficult to design a very good one. Between a very bad
fan and a very good one lie a large number of fans with acceptable performance and
satisfactory efficiency. In this Thesis a design procedure has been established which
was used to design vaneaxial fans with higher efficiency than similar size existing
designs have. Three new fans were designed and optimized and the experience from
the older design was used for the newer design. Finally the best efficiency achieved
was 82.1% which is about 3% higher than the efficiency of the given fan.
The maximum efficiency and the total pressure rise for the design point of the 3
new designs and the given fan are summarized in Table 8.1
Table 8.1 Summary of the 3 new designs
Total Pressure Rise (Pa) Efficiency
Given Fan 506 79.5%
Design1 529 79.3%
Design2 550 82.1%
Design3 483 79.5%

140 | P a g e
 Conclusions and Recommendations

The power consumption is not an important factor for small fans, but it becomes
a significant factor for large fans that operate continuously. In cases like this, small
efficiency improvement can represent substantial cost reduction. In this work has been
proven that the efficiency of a vaneaxial fan can be improved with the careful design of
the downstream guide vanes in order to reduce the losses from separation and convert
the energy of the swirl velocity to useful static pressure rise.

The tip clearance is a critical factor for the performance and the mechanical
integrity of a vaneaxial fan. The increase of tip clearance from 1.5% of the blade span to
3% of the blade span reduces the efficiency by 6.5-9.5%. The mechanical integrity of
the fan is better with higher tip clearance because the fan can operate in higher
temperature for longer period of time.

All the above mentioned results were outputs from CFD simulations. The
performance of the new designs has been examined only with CFD and there are no
experimental data. A divergence was noticed between the performance of the model
that was the output of the design procedure and the performance according to the CFD
results. The main reason for this divergence is the three dimensional effects that are not
taken into account in the design procedure. Another important observation is that the
selection of the turbulence model during the set up of the physics in the CFD has an
immediate impact in the performance of the fan. The k-ω turbulence model does not
overestimate the turbulence as the k-ε model does and the performance acquired using
the k-ω model is better than with the k-ε model. The disadvantage of the k-ω model is
the difficulty in convergence.

8.3 Recommendations for further work

During the design process many compromises and arbitrary choices have been
made. A small change in these choices can influence the performance of the fan. All
these factors can be separately examined in order to estimate their impact to the
performance of the fan.
1. Wide tip blade
Wide tip blades generally give better efficiency than the narrow tip blades. All
the new designs have narrow tip in order to have better mechanical integrity. The
examination of the mechanical integrity proved that the centrifugal loading is relatively
low, so there is enough safety margin to design a wide tip blade for further efficiency
improvement. A wide tip blade can be easily incorporated to the established design
procedure.
2. Free Vortex Design
141 | P a g e
 Conclusions and Recommendations

The free vortex design concept was selected because it is more convenient to
standardize a design procedure based on this design concept. The three new designs
are then free vortex designs Sorensen, 2000 developed an aerodynamic model for
arbitrary vortex flow and Lewis, 1996 as well. These design concepts can be used
alternatively to give a fan design.
3. Rotational Speed
The rotational speed of the new designs is from 1600-1650 rpm. It’s difficult to
achieve this rotational speed with the existing motors. The motors that are in the market
for this range of rotational speeds usually give 1500rpm or 1750rpm and the given fan
has a motor that gives 1500rpm. New models can be further developed with these
rotational speeds.
4. Alternative airfoil
The base airfoil that has been used for the new designs is the C4 airfoil. The
MATLAB code that has been developed can use any airfoil. The NACA 65 airfoil is an
established alternative choice for the airfoil. New design can be developed using the
NACA65 airfoil and incorporated it to the design procedure.
5. Correlation of blade angles with air angles
The design procedure that has been established correlates the blade angles
with the air angle with a formula which was the result of experiments in a C4 cascade.
The use of different airfoil requires the re-establishment of the correlation between the
blade angle and the air angle. Since it is quite difficult and expensive to set a series of
new experiments, a good alternative could be the trial and error method using CFD
simulations.
6. Vane design
The vanes of the new designs have airfoil shape. Airfoil shaped vanes
increase the total cost of the fan. The performance of the new designs can be examined
with single thickness vanes.
7. Coding of the design procedure
The design procedure has been developed in an excel file. The use of data
from Tables and Figures has made the coding of the procedure quite difficult. An
attempt to code the design procedure can make the procedure more convenient for the
user.
8. Material properties
The material that was selected for the fan blades is the Aluminum 6061-T6.
The safety factor for the extreme use under 400o C is 1.25 and it can be further
improved with the use of a material with better thermal strength. The mechanical
integrity of a model with wide tip blades and better material can be additionally
examined.

142 | P a g e
 List of References

1 Strohmeyer, H. (2009), Improving the Efficiency of an Industrial Low

Speed Fan, MSc Thesis, School of Engineering, Cranfield University

2 Bleier, Frank P. (1998), Fan Handbook: Selection, Application, and

Design, McGraw-Hill Companies. Inc., New York

3 McKenzie A.B. (1997), Axial Flow Fans and Compressors,

Aerodynamic design and Performance, Ashgate

4 Osborne William C. (1977), Fans 2nd Edition (in SI/Metric Units),

Pergamon Press

5 Eck, Bruno (1973), FANS Design and Operation of Centrifugal, Axial

Flow and Cross Flow Fans, Pergamon Press

6 Tu J., Yeoh G.H., Liu C. (2008), Computational Fluid Dynamics a

Practical Approach, Butterworth-Heinemann

7 Ramsden, K. W. (2008), Course Notes for Axial Compressor Design

and Performance, Lecture Notes, Cranfield University

8 Lewis, R. I. (1996), Turbomachinery Performance Analysis,

Butterworth-Heinemann

9 Cumpsty, N. A. (1989), Compressor Aerodynamics, Longman

Singapore Publishers Ltd

10 Terzis A., Kalfas A. I., Zachos P., Ramsden K. W. (2009),

Experimental, Performance Evaluation of Axial Ventilators, The 9th
International Conference on Industrial Ventilation

11 Bass, R.M. (1987), Factors influencing the aerodynamic design of low

pressure axial fans. In: Industrial Fans-aerodynamic design, London,
April 09, 1987

12 Palmer J.R., Pilidis P. (2008), Gas Turbine Theory and Performance,

Lecture Notes, Cranfield University

13 Rubini P.A, Turbine Blade Cooling, Lecture Notes, Cranfield University,

2008

14 Walsh Philip P., Fletcher Paul (2004), Gas Turbine Performance

Second Edition, Blackwell

15 Cookson R.A., Haslam A.S. (2008), Mechanical Design of

Turbomachinery, Lecture Notes, Cranfield University

16 Angoy C.H. (1984), Fan Design : a novel flat blade design, Thesis,
School of Engineering, Cranfield University

17 Turner R. C. (1966), Notes on Ducted Fan Design, Ministry of Aviation

Aeronautical Research Council Current Papers

143 | P a g e
 List of References

18 Terzis A, Stylianou I, Kalfas A, Ott P. (2011), Effect of Stators on the

Performance and the Thermal Capabilities of Small Axial Cooling Fans,
The 10th International Symposium on Experimental Computational
Aerothermodynamics of Internal Flows

19 Hay N, Mather J. S. B., Metcalfe R (1989), Fan blade selection for low
noise, Proceedings of the Institution of Mechanical Engineers, Part A:
Journal of Power and Energy August 1989 vol. 203 no. 3 149-154

20 Smith T. W., A practical approach to the design of axial and mixed flow
fans

21 McKenzie A. B. (1988), The selection of fan blade geometry for

optimum efficiency, Proceedings of the Institution of Mechanical
Engineers, Part A: Journal of Power and Energy February 1988 vol.
202 no. 1 39-44

22 Cory W. T. W. (2005), Fan and Ventilation, Elsevier, ISBN 0-080-

44626-4

23 Kaufman G. J. (2008), Parametric analyses of high temperature data

for aluminum alloys, ASM International

24 Naeem M. T., Jazayeri S. A., Rezamahdi N., Failure Analysis of Gas

Turbine Blades, Paper 120, ENG 108, Proceedings of The 2008 IAJC-
IJME International Conference ISBN 978-1-60643-379-9

25 Wallis, A. R. (1983). Axial Flow Fans and Ducts. John Willey & Sons.

26 Sorensen D.N. (2000), Towards Improved Rotor-Only Axial Fans-Part I:

A Numerically Efficient Aerodynamic Model for Arbitrary Vortex Flow,
Transactions of the ASME Vol. 122, June 2000

27 Dixon S. L. (1998), Fluid Mechanics and Thermodynamics of

Turbomachinery (4th Edition),Elsevier

144 | P a g e
 APPENDIX “A”. Design2 Data
Table 1 Final Rotor Blade Design (Design2)
HUB TIP
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
S/C 0.77 0.90 1.04 1.17 1.30 1.44 1.57 1.70 1.84 1.97
a1 60.36 62.70 64.72 66.48 68.03 69.39 70.60 71.67 72.64 73.52
a2 46.43 52.37 56.86 60.37 63.18 65.48 67.39 69.02 70.42 71.64
tanam=
1.40 1.62 1.82 2.03 2.23 2.43 2.62 2.81 3.01 3.20
(tana1+tana2)/2

tanζ=tanam -0.15 1.25 1.47 1.67 1.88 2.08 2.28 2.47 2.66 2.86 3.05

ζ (rad) 0.90 0.97 1.03 1.08 1.12 1.16 1.19 1.21 1.23 1.25
STAGGER 51.44 55.73 59.16 61.97 64.30 66.27 67.96 69.42 70.70 71.82
CAMBER(0.5-
27.85 22.05 18.30 15.91 14.42 13.58 13.22 13.22 13.52 14.10
0.31*(S/C)^(1/3))
β1=2θ+ζ 65.36 66.75 68.31 69.92 71.51 73.06 74.57 76.03 77.46 78.87

incidence i= (a1-β1) -5.00 -4.05 -3.59 -3.44 -3.49 -3.68 -3.97 -4.36 -4.82 -5.36

β2=β1-θ 37.51 44.70 50.01 54.01 57.09 59.48 61.35 62.81 63.94 64.78

deviation δ=α2-β2 8.92 7.67 6.85 6.35 6.09 5.99 6.04 6.21 6.48 6.86

deflection e =a1-a2 13.93 10.33 7.86 6.12 4.85 3.91 3.20 2.65 2.22 1.88

am (rad) = 0.95 1.02 1.07 1.11 1.15 1.18 1.21 1.23 1.25 1.27
am (deg) = 54.54 58.27 61.28 63.75 65.83 67.59 69.11 70.43 71.60 72.63

Cl (from Osbourne) =
2*S/C*(tana1- 0.63 0.61 0.58 0.56 0.53 0.51 0.49 0.47 0.45 0.43
tana2)*cosam

CHORD (mm) 0.155 0.145 0.138 0.133 0.129 0.125 0.122 0.120 0.118 0.116

A-1
 Table 2 Stator Vane Design (Design2)
HUB TIP
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
a3 35.23 32.64 30.37 28.37 26.60 25.03 23.62 22.35 21.20 20.17
a3 (rad) 0.61 0.57 0.53 0.50 0.46 0.44 0.41 0.39 0.37 0.35
a4 3.00 2.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
a4 (rad) 0.05 0.03 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00
tanam=
(tana3+tana4)/2 0.38 0.34 0.30 0.27 0.25 0.23 0.22 0.21 0.19 0.18
tanζ=tanam -0.15 0.23 0.19 0.15 0.12 0.10 0.08 0.07 0.06 0.04 0.03
ζ (rad) 0.23 0.19 0.15 0.12 0.10 0.08 0.07 0.06 0.04 0.03
STAGGER 12.92 10.63 8.63 6.84 5.73 4.77 3.92 3.18 2.52 1.93
CHORD 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13
S= 0.10 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19
S/C= 0.76 0.84 0.92 0.99 1.07 1.15 1.23 1.31 1.38 1.46
CAMBER 50.29 44.38 39.75 36.14 31.02 26.58 22.70 19.26 16.22 13.49
β1=ζ+θ/2 38.06 32.82 28.51 24.92 21.25 18.06 15.27 12.81 10.63 8.67
i=a3-β1 -2.83 -0.18 1.87 3.46 5.36 6.97 8.34 9.54 10.58 11.49

Table 3 Rotor data for Design2

SECTION 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
CAMBER 27.85 22.05 18.30 15.91 14.42 13.58 13.22 13.22 13.52 14.10
CHORD 0.155 0.145 0.138 0.133 0.129 0.125 0.122 0.120 0.118 0.116
DIAMETER 0.42 0.46 0.50 0.54 0.59 0.63 0.67 0.71 0.76 0.80
STAGGER 51.44 55.73 59.16 61.97 64.30 66.27 67.96 69.42 70.70 71.82

Table 4 Stator data for Design2

SECTION 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
CAMBER -50.29 -44.38 -39.75 -36.14 -31.02 -26.58 -22.70 -19.26 -16.22 -13.49
CHORD 0.132 0.132 0.132 0.132 0.132 0.132 0.132 0.132 0.132 0.132
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
STAGGER -12.92 -10.63 -8.63 -6.84 -5.73 -4.77 -3.92 -3.18 -2.52 -1.93

A-2
 Table 5 Final Stator Vane Design (Design2) (optimized)
HUB TIP
DIAMETER 0.416 0.459 0.501 0.544 0.587 0.629 0.672 0.715 0.757 0.800
a3 35.23 32.64 30.37 28.37 26.60 25.03 23.62 22.35 21.20 20.17
a3 (rad) 0.61 0.57 0.53 0.50 0.46 0.44 0.41 0.39 0.37 0.35
a4 3.00 2.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
a4 (rad) 0.05 0.03 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00
tanam=
(tana3+tana4)/2 0.38 0.34 0.30 0.27 0.25 0.23 0.22 0.21 0.19 0.18
tanζ=tanam -0.15 0.23 0.19 0.15 0.12 0.10 0.08 0.07 0.06 0.04 0.03
ζ (rad) 0.23 0.19 0.15 0.12 0.10 0.08 0.07 0.06 0.04 0.03
STAGGER 12.92 10.63 8.83 7.44 6.83 6.27 5.82 5.48 5.12 4.73
CHORD 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13 0.13
S= 0.10 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19
S/C= 0.76 0.84 0.92 0.99 1.07 1.15 1.23 1.31 1.38 1.46
CAMBER 50.29 44.38 40.67 38.91 36.09 33.49 31.44 29.86 28.19 26.39
β1=ζ+θ/2 38.06 32.82 29.17 26.90 24.88 23.01 21.55 20.41 19.21 17.92
i=a3-β1 -2.83 -0.18 1.21 1.47 1.72 2.01 2.07 1.94 1.99 2.25

Table 6 Verification of Design2

Rotor mesh (thousands Stator mesh size (thousands Total Pressure Rise
elements) elements) (Pa)
56.5 58.7 526
113.8 117.3 530
236.9 238.2 532
461.6 463.5 534

A-3
 APPENDIX “B”. Design3 Data

Table 1 Final Rotor Blade Design (Design3)

HUB TIP
DIAMETER 0.384 0.430 0.476 0.523 0.569 0.615 0.661 0.708 0.754 0.800
S/C 0.65 0.77 0.90 1.03 1.16 1.28 1.41 1.54 1.67 1.79
a1 58.92 61.72 64.09 66.12 67.86 69.38 70.71 71.89 72.93 73.87
a2 42.67 50.21 55.67 59.78 62.97 65.53 67.63 69.38 70.86 72.14
tanam=
(tana1+tana2)/2 1.29 1.53 1.76 1.99 2.21 2.43 2.64 2.86 3.07 3.28
tanζ=tanam -0.15 1.14 1.38 1.61 1.84 2.06 2.28 2.49 2.71 2.92 3.13
ζ (rad) 0.85 0.94 1.02 1.07 1.12 1.16 1.19 1.22 1.24 1.26
STAGGER 48.76 54.07 58.18 61.44 64.10 66.30 68.15 69.73 71.09 72.28
CAMBER 30.34 22.59 17.81 14.85 13.06 12.03 11.52 11.41 11.59 12.01
(0.5-
0.31*(S/C)^(1/3))
β1=2θ+ζ 63.92 65.37 67.08 68.87 70.63 72.31 73.91 75.43 76.89 78.29
incidence i= (a1-
β1) -5.00 -3.64 -2.99 -2.75 -2.76 -2.93 -3.20 -3.54 -3.95 -4.42
β2=β1-θ 33.59 42.77 49.28 54.02 57.57 60.28 62.39 64.02 65.30 66.27
deviation δ=α2-β2 9.08 7.44 6.39 5.76 5.40 5.25 5.24 5.35 5.56 5.86
deflection e =a1-a2 16.25 11.51 8.42 6.34 4.89 3.85 3.09 2.51 2.07 1.73
am (rad) = 0.91 0.99 1.05 1.10 1.15 1.18 1.21 1.23 1.26 1.27
am (deg) = 52.23 56.83 60.42 63.29 65.65 67.61 69.28 70.71 71.95 73.04
Cl (from Osbourne)
= 2*S/C*(tana1- 0.58 0.56 0.53 0.50 0.47 0.45 0.43 0.41 0.39 0.37
tana2)*cosam
CHORD 0.170 0.159 0.151 0.145 0.140 0.137 0.134 0.131 0.129 0.127

Table 2 Final Stator Vane Design (Design3)

HUB TIP
DIAMETER 0.384 0.430 0.476 0.523 0.569 0.615 0.661 0.708 0.754 0.800
a3 36.41 33.36 30.73 28.45 26.47 24.72 23.18 21.81 20.59 19.49
a3 (rad) 0.64 0.58 0.54 0.50 0.46 0.43 0.40 0.38 0.36 0.34
a4 3.00 2.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
a4 (rad) 0.05 0.03 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00
tanam=
0.39 0.35 0.31 0.27 0.25 0.23 0.21 0.20 0.19 0.18
(tana3+tana4)/2

tanζ=tanam -0.15
0.24 0.20 0.16 0.12 0.10 0.08 0.06 0.05 0.04 0.03
ζ (rad) 0.24 0.19 0.15 0.12 0.10 0.08 0.06 0.05 0.04 0.03
STAGGER 13.76 11.12 9.26 7.70 7.05 6.49 5.97 5.57 5.17 4.75
CHORD 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139
S= 0.13 0.15 0.16 0.18 0.19 0.21 0.22 0.24 0.25 0.27
S/C= 0.94 1.05 1.16 1.27 1.39 1.50 1.61 1.72 1.84 1.95
CAMBER 51.25 44.08 40.32 37.84 35.02 32.55 30.29 28.55 26.79 24.95
β1=ζ+θ/2 39.39 33.16 29.42 26.61 24.56 22.76 21.12 19.85 18.57 17.22
i=a3-β1 -2.98 0.19 1.30 1.84 1.91 1.96 2.07 1.97 2.03 2.27

B-1
 Table 3 Rotor data for Design3
SECTION 1 2 3 4 5 6 7 8 9 10
CAMBER 30.34 22.59 17.81 14.85 13.06 12.03 11.52 11.41 11.59 12.01
CHORD 0.170 0.159 0.151 0.145 0.140 0.137 0.134 0.131 0.129 0.127
DIAMETER 0.384 0.430 0.476 0.523 0.569 0.615 0.661 0.708 0.754 0.800
STAGGER 48.76 54.07 58.18 61.44 64.10 66.30 68.15 69.73 71.09 72.28

Table 4 Stator data for Design3

SECTION 1 2 3 4 5 6 7 8 9 10
CAMBER -51.25 -44.08 -40.32 -37.84 -35.02 -32.55 -30.29 -28.55 -26.79 -24.95
CHORD 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139 0.139
DIAMETER 0.384 0.430 0.476 0.523 0.569 0.615 0.661 0.708 0.754 0.800
STAGGER -13.76 -11.12 -9.26 -7.70 -7.05 -6.49 -5.97 -5.57 -5.17 -4.75

B-2
 APPENDIX “C”, Tip Clearance Effect
Performance comparison of new Designs with nominal tip clearance (1.5%) and
high tip clearance (3%)

Tip Clearance effect for Design1

Volume Flow vs Efficiency

1
0,9
0,8
0,7
0,6
Efficiency

0,5 given fan

0,4 design1
0,3 tip cl 3%
0,2
0,1
0
3 4 5 6 7 8 9 10
Volume Flow (m3/s)

Volume Flow vs Total Pressure rise

800

700
Total Pressure Rise (Pa)

600

500
given fan
400
design1
300 tip cl 3%
200

100

0
4 5 6 7 8 9 10
Volume Flow (m3/s)

C-1
Tip Clearance effect for Design 2

Volume Flow vs Efficiency

1
0,9
0,8
0,7
0,6
Efficiency

given fan
0,5
design2
0,4
tip cl 3%
0,3
0,2
0,1
0
3 4 5 6 7 8 9 10
Volume Flow (m3/s)

Volume Flow vs Total Pressure rise

800

700

600
Total Pressure Rise (Pa)

500
given fan
400
design2
300 tip cl 3%

200

100

0
3 4 5 6 7 8 9 10
Volume Flow (m3/s)

C-2
Tip Clearance effect for Design3

Volume Flow vs Efficiency

1
0,9
0,8
0,7
0,6
Efficiency

given fan
0,5
design3
0,4
tip cl 3%
0,3
0,2
0,1
0
3 4 5 6 7 8 9 10
Volume Flow (m3/s)

Volume Flow vs Total Pressure rise

800

700

600
Total Pressure Rise (Pa)

500
given fan
400
design3
300 tip cl 3%

200

100

0
3 4 5 6 7 8 9 10
Volume Flow (m3/s)

C-3