PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION

by
Mark A. Corbo
Project Design Engineer
and
Stanley B. Malanoski
Manager, Engineering Services
Mechanical Technology Incorporated
Latham, New York

encounter major difficulties in the field . It has been the authors'

Mark A. Corbo is a Project Design En- experience that most turbomachinery users encounter little
gineer with Me~'hanical Technology difficulty in determining their machine's natural frequencies due to
Incorporated, a high teclm%gyengineering/ the large number of resources available in that area. However,
({)Ilmlling firm. In this positioll, he is re- problems often arise when they must translate this information into
sponsible for perjonlling analytical studies, an accurate prediction or whether or not their design will
troubleshooting, and design Gudits in tire experience torsional vibration problems. Accordingly, this
areas of r%rdynamics, fluid-film lubrica- presentation concentrates on the steps that should be taken once the
tion, and hydraulics for various customers natural frequencies have been found.
within the tllrbomachinery industry. Prior A cursory review is presented of popular procedures, such as
to joining MT! ill 1995, he spellt 12 years ill Holzer's method, for obtaining the machine's natural frequencies
the aerospllce industry designing and and mode shapes. This area is purposely limited in detail since
analyzillg pumps, valves, controls, and electromechanical compo- there are many excellent resources in the literature that may be
/lollS fiJI' gas IlIriJine engines. Hi.l· fields of expertise include consulted for a more rigorous treatment.
rotordYllamics, jOllrnal bearings, incompressible and compressible The generation of an interference or Campbell diagram is then
flow, cumputational fluid dynamics, stress analysis, finite element treated in far more detail. Of panicular interest is generation of the
analysis, dynamic simulations, and mechanical design. He holds upward sloping lines representing the system's excitation frequen-
B.S. and M.S. degrees (Mechanical Ellgineering) from Rensselaer cies . The various excitation sources commonly found in
Polytechnic Institute. He is a member of ASME. turbomachinery, such as gears, vaned impellers, and electric
motors, are discussed along with the excitation frequencies that
each introduces into the system. The unique problems associated
Stanley B. Malanoski is a graduate with the stanup of systems driven by synchronous motors are also
Mechanical Engineer and Manager of described.
Engineering Services at Mechanical Once the interference points have been generated, the user then
Technology Incorporated's Technology generally has two choices for dealing with them. Either design
Division. He has over 30 years of industri- changes. such as alteration of couplings, are implemented to
al experience in the areas of eliminate the interferences or the interference points are subjected
turbomachinery design, analysis, and trou- to funher analysis. Many users automatically opt for the first alter-
bleshooting. Mr. Malanoski '.I' fields of native. since they believe they must avoid resonance conditions at
special competence are in the management all costs. While this is an admirable and wonhy goal under ideal
of engineering personnel and programs; circumstances, the cost of achieving it is often unwieldy.
dYllamic analysis of practical rotor-bearing Instead. the procedure provided herein advocates analysis of all
systems: squeeze-film damper design, analysis and application; intelference points prior to the implementation of costly design
and j/llidJilm bearing!~eal and system designs, in eluding process changes. The analysis might be as simple as inspection of the
fluid (gas and liquid) lubrication He is author or coauthor of over appropriate mode shape or the unit's torque 1'5 speed curve.
50 technical publications and hundreds of Mechanical Technology Interference points can frequently be eliminated from considera-
IncOIporated tee/mical reports. tion based on these inspections revealing that the induced torques
are negligible. The resonant points that cannot be thereby removed
should be investigated using a damped forced vibration analysis.
ABSTRACT Detailed guidelines for performing the damped analysis are
One of the foremost concerns facing turbomachinery users today presented herein. Methods for detennining the magnitudes and
is that of torsional vibration. In contrast to lateral vibration locations of excitation torques for various machinery classes are
problems. torsional failures are especially heinous since the first given. Procedures are provided for obtaining damping coefficients
symptom of a problem is often a broken shaft, gear tooth, or for typical sources such as impellers, shaft material hysteresis, and
coupling. The difficulty of detecting incipient failures in the field couplings. Finally, users are provided with ground rules for
makes the performance of a thorough torsional vibration analysis utilizing the calculated cyclic torques and stresses to determine
an essential component ot' the turbomachinery design process. their design's adequacy.
Tho: authors' purpose is to provide users with a practical design If the analysis identifies problem areas. practical and relatively
procedure that can be used to cnsure that their systems will not simple rcctification methods arc provided. Lastly. a complete
189
190 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

step-by-step analysis procedure is given that summarizes the entire continue past until the spring was twisted by an equal amount in
preceding discussion. This methodology can be utilized in the the opposite direction. Oscillations would continue indefinitely
design of virtually any turbomachinery system the user may with the system continually exchanging the potential energy of the
encounter. spring for kinetic energy in the inertias. These vibrations havc no
effect on the system's average speed, which remains constant.
INTRODUCTION This phenomenon is known as undamped free torsional
Torsional vibration is a subject that should be of concern to all vibration and is fully analogous to the well-known linear mass-
turbomachinery users. The word "users," utilized throughout this spring system . Regardless of the initial conditions existing prior to
work, refers to all engineers, including designers, analysts., vibration, the system always oscillates at a specific frequency,
managers, and operators, involved in the design, manufacture, known as the undamped natural frequency. The natural frequency
and/or operation of turbomachinery. By some accounts, torsional is a function of the disks' inertias and the shaft's stiffness and is.
vibration is the leading cause of failures in turbomachinery drive thus, a characteristic of the system.
trains. Some typical effects of uncontrolled torsional vibration are Forced vibration can be illustrated with the same system by
failed couplings, broken shafts, worn gears and splines, and superimposing a sinusoidally varying torque on the steady torque
fractured gear teeth. Accordingly, a thorough torsional vibration of either of the disks. The resulting imbalance bctween driver and
analysis should be included as an integral part of the design load torques would cause all elements to vibrate ubout the equilib-
process. A thorough analysis procedure that can be practically rium position. Consequently, all elements would experience
implemented by turbomachinery users is the primary subject here. sinusoidal fluctuations in torque and speed about their average
Although it is felt that all types of engineers who work with tur- values.
bomachinery can gain a flavor for the subject from the information The magnitudes of the induced cyclic torques in the second disk
presented herein, the presentation is primarily directed towards and connecting shaft would be dependent on the dynamic charac-
mechanical engineers. Specifically, it is meant to aid those teristics of the system. The response is controlled by the ratio of the
mechanical engineers responsible for the design and analysis of excitation frequency ,t o the natura'l frequency, as is illustrated in the
turbomachinery drive trains. It is the opinioo of the authors lhal well-known undamped response curve for a one degree of freedom
most torsional vibration problems experienced in the field can be sys·tem shown in Figure 2. The {)rdinate on this figure is the ratio
prevented by taking prudent action during the design process. of the induced shaft torque to the excitation torque and is referred
Although many turbomachinery users are intimately familiar to as the dynamic magnifier.
with the fundamentals of torsional vibration, the authors are
acquainted with some who are 1I0t. For their benefit, a brief review
of the basics is, thus, in order.
For illustrative purposes, any turbomachinery assembly can be
3.0
approximated by two inertias or disks connected by a torsional
spring as is shown in Figure I. One inertia can be taken to represent
the system's driving element, which is usually a turbine or motor li;
while the second corresponds to dle driven load such as a com- :g 2.0
Cl
pressor or pump impeller. The torsional spring between them is a
simplified representation for the interconnecting shafting.
'"
::::E
u
'E
as Resonance
c
>. 1.0
Cl POint

Torsional
Spring
°OL---~--~---L--~--~L:::3I.O::~===4~.=O==~==?5.0
Frequency Ratio (ww n)

Inertia 2 Figure 2. Undamped Response Curve for One D.O.F. System.

It is seen that when the driving frequency is equal to the system's

natural frequency, the dynamic magnifier is theoretically infinite.
Inertia 1
This condition is known as resonance and represents a potential
problem for the system. Although the actual response in practice is
not infinite due to the presence of finite damping in all systems.
Figure 1. Two Inertia Torsional System. huge amplifications can still occur which produce large cyclic
stresses in the shaft.
When the machine is at resl, the two inertias are stationary and These high stresses can often lead to shaft fatigue. Additionally.
the shaft is unstressed in its free position. Whel'! the machine is the large generated peak torques can overload components such as
started and brought up to steady 'speed, the driving and load torques gears, splines, and couplings. Therefore, the essence of torsional
become equal and the two disks rotate at the same velocity. vibration analysis is identification of all resonance points and
However, the shaft is twisted away from ils free position by an determination of the system's ability to withstand them.
angle equal to the transmittedtorquc divided by its spring rate.
This position will be referred to as the cquilibrium position. COlllpari.I'ON with Lateral Vibration
If the driver and load torques were then suddenly removed from Many readers may be more familiar with lateral vibration than
the assembly, the twisted. spring would uncoil, propelling thc two its torsional counteflJarl. Such knowledge is helpful since the two
disks in opposite directions with respect to the shaft. If there was phenomena are similar in many ways. However, there are several
no damping present, the disk:; would rotate {o the free position and important differenccs. including the following:
 PRACTICAL DES IGN AGAINST TORSIONAL VIBRATION 191

• In lateral vibrations, the natural frequencies arc often functions Uncommoll electrical machines. The discussion will be limited
of the operating speed due to their dependence on tluid-film to the motor and generator types most often found in turbomachin-
bearing stiffness. Conversely, torsional natural frequencies are ery. Specifically, DC and three phase AC machines will be the only
independent of operating speed. types addressed. Additionally, all discussion of synchronous
• In lateral vibrations, large vibratory motions can cause severe motors will refer to those having rotors with salient poles. The user
problems. On the other hand. the primary parameters of concern in who encounters a machine not mentioned herein is strongly
torsional vibration are induced torques and stresses. The ac tual dis- advised to consult with the manufacturer for torsional vibration
placements are usually of academic interest only. characteristics.
Measurelllent of torsiollal vibratioll. This paper is primarily
• In the most common lateral modc. synchronous whirling, the
concerned with the design and analysis process so testing proce-
shaft does not undergo stress reversals. The bending stress is essen-
dures are not addressed. Additionally, like many of the other topics
tially constant and shaft fatigue is not a concern . Conversely,
on this list. the subject of measuring devices and techniques is
torsional vibrations always induce cyclic stresses that can lead to
worthy of a paper of its own. Simmons and Smalley [14).
fatigue.
Hershkowitz lIS], and Wachel and Szenasi [16) all provide
• Unlike lateral vibrations. torsional vibration problems usually detailed descriptions in this area .
cannot bc corrccted by balancing the machine more precisely. Electrical circlIit damping. Only mechanical sources of damping
Topics Not Covered arc discussed herein. These include damping that occurs in electric
motors and generators due to their torque vs speed characteristics.
Although this presentation is intended to be comprehensive, However, any damping that occurs due to electrical control circuits
there are some subjects within the field of torsional Vibration that varying (he excitations to electrical machines is well beyond the
are not addressed. The amount of information related to this topic scope of this presentation . The interested reader should see
is far too voluminous to cover in a single tutorial. The authors have, Hammons 117).
therefore, attempted to limit the covered material to areas that a SlIbsynclllvnolls resonallces ill IIIrbogenerators. There have
practical mechanical engineer needs to know in order to perform a been a slew of papers written on this relatively rare phenomenon,
complete torsional analysis on a turbomachinery drive train. which is characterized by resonant interactions between the elec-
Accordingly, the following subjects are either partially or totally trical network and the mechanical drive train. In order to evaluate
neglected in this work: this potentially destructive mechanism, a comprehensive model
Distributed parameter lIlodels. Since most practica l syStems encompassing both the network and drive train must be utilized. A
can be modelled accurately using them, the discussion is limited to description of the steps required to prepare and run such a model
lumped parameter models. The procedures utilized with a would be worthy of a tutorial of its own. A substantial list of refer-
distributed parameter model are similar to those specified herein, ences on this complex subject is provided in a 1992 article [18J.
although they are more complicated. Vance [11 points out that the Malsynclllvnizariol! of synchronous motors and generators. If
main difference is that the ordinary differential equations of the the electrical control circuits are not designed properly, malsyn-
lumped model are replaced by partial differential equations. chronization can occur and generate pulsating torques that are
Distributed parameter models are discussed to various extents by many times the machine's rated torque. The consequences of this
Ker Wilson [2], Eshleman [3], Bisshopp [4), and Triezenbetg [5]. are often catastrophic. However, since this effort is written
Coupled lateral-torsiollal vibrations. There are some instances, primarily for mechanical engineers, this subject is not covered. The
particularly in machines containing gears, where a strong coupling user is advised, however, to maintain effective communication with
between lateral and torsional modes can arise. The procedure the cognizant electrical engineers during the entire design process
required to analyze such a phenomenon would require an entire to avoid potential problems. Further elaboration on this topic is
paper of its own. Additionally, Simmons and Smalley [6] claim ,that provided by Rana and, Schulz [19], Joyce, et al. [20], and Undrill
the presence of lateral vibration has very little impact on the and Hannell [21).
torsional natural frequencies and mode shapes. Thus, with the
exception of the effect it can have on journal bearing damping, this
UNDAMPED ANALYSIS
phenomenon is not addressed. Lund [71 provides a description of
this topic for the interested reader. The first step in any torsional analysis procedure is determina-
Reciprocating machines. In order to maintain this work at a tion of the system's natural frequencies and mode shapes. To
manageable length, its scope is limited to rotary machines. The accomplish this, a lumped parameter model, consisting of disk and
classic texts by Ker Wilson 12) and Nestorides 18) are among the shaft elements, is usually generated. The disks represent the
large number of works that describe the excitations generated in system's significant inertial components while the shaft elements
reciprocating engines and methods for converting their reciprocat- behave as torsional springs. All springs are assumed to behave in a
ing masses into equivalent rotary inertias. linear fashion such that the torque they exert is directly propor-
Torsional vibration dall/pers. Although the damping introduced tional to their twist angle. A schematic representation of a three
by common turbomachinery components is discussed in great disk. two shaft model is presented in Figure 3.
detail. devices that are used for the sole purpose of introducing Since real systems contain energy dissipation elements, known
damping into the assembly are not addressed. These devices, which as dampers. along with inertia and stiffness elements, these techni-
include Lanchester dampers, specialized oil-filled couplings , and cally also should be included in Figure 3. However, the addition of
Holset couplings, are described by O'Connor 19], Den Hartog and damping makes the determination of natural frequencies consider-
Ormondroyd [lO). and Brown [111. ably more difficult. Additionally, the vast majority of the available
Electrical-mechallical analogies. Some authors have solved literature concedes that the error introduced by ignoring damping
both undamped and damped torsional vibration problems by first in the calculation of natural frequencies is practically negligible.
converting the mechanical system into an equivalent electrical Thus. all natural frequencies and mode shapes discussed herein
circuit. The equivalent circuit is then either built and tested or will be those associated with the undamped system.
analyzed to determine its dynamic characteristics. The results, in The number of degrees of freedom possessed by the system is
terms of voltages and currents, are then converted back into their equal to the number of disks in the model. Any system , such as that
mechanical equivalents. McCann and Bennell [121 and Pollard of Figure 3. which docs not have a shaft element connected to
1131 hoth illustrate the usc of this method. ground has a trivial natural frequency of zero. representing the case
192 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

4. Likewise, the second mode contains two nodes. the third has
Shaft Elements three. etc.
Prior to describing the analysis procedure, it should be noted
that natural frequencies are properties of the entire system. If any
component is changed. the torsional characteristics can be drasti-
cally altered. Thus. whenever a system is changed. a new torsional
analysis should be performed. Additionally. since torsional
response is a system property. nothing is gained by analyzing the
individual components by themselves.

Modelling
The first task to be accomplished in the analysis procedure is
generation of the lumped model. Firstly, all significant inertias in
the system should be identified as disks. These include impellers.
Disk Elements propellers. motor and generator rotors, gears, and coupling hubs.
The choice of the nlllnber of disks Lo include is usually a com-
promise. If every single inertia that exists in the assembly were
Figure 3. Three Disk TtJrsiollo/ System.
represented, the modelling and solution time would likely be pro-
hibitive. On the other hand, if complex turbomachinery trains were
where all elements rotate together as a rigid body. The number of modelled as two disk systems, as was suggested in the
nontrivial natural frequencies that can be obtained for such a INTRODUCTION section, the loss of accuracy would probably be
system is. thus, one less than the number of disks in the model. unacceptable.
Each natural frequency has an alisociated mode. shape which Ano.i her consideration is that the number of natural frequencies
describes the shape the shaft twists into duriog free vibration. A that can be calculated is limited by the number of disks in the
representative mode shape is showlil ill Figure 4. The abscissa model. The analyst must ensure that enough disks are included
values represent the axial positions of the disk elements while the: such that all natural frequencies that could reasonably be expected
ordinates correspond to the angles of twist occurring at each disk. to be excited within the machine's operating speed range are
As is the case with lateral vibrations, it is meaningless to refer to determined.
absolute displacements since they are Iheoretically infinite. The All disk elements must be assigned a value for mass polar
only information that can be obtained from a mode shape is Ihe moment of inertia. The inertias for components bought from
relationship between the displacements at the various disks. vendors such as electric machines and couplings can usually be
Accordingly. these curves are arbitrarily normalized such ,t hat the obtained from the manufacturer. The remaining inertia values are
system's maximum displacement is equal to one radian. Mode generated either by test or calculation. The classic works by Ker
shapes are often referred to as nQnnal to reflect the facr that they Wilson [2] and Nestorides [8] present equations for the inertias for
arc all orthogonal to each other. disks of almosl every conceivable configuration. Continuously dis-
tributed inertias may require the use of numerical integration.
Maximum Displacement
The most common inertia element, a hollow disk, has an inertia
/ Set to One Radian
given by the following :

(I)

where:
J Mass polar moment of inel1ia (lbm-in 2 )
P Material density (lbmlin 3 )
Do Outside diameter (in)
Di Inside diameter (in)
L Length (in)

Axial Position (in.) Once the inertias are determined, the torsional spring rates for
the shaft sections which interconnect the disks must be found. The
general equation for the torsional stiffness of a shaft is as follows:

k = G· Ip I L (2)
where :
k Torsional stiffness (in-Ibf/rad)
G Shaft material shear modulus (psi)
-1 lp Area polar moment of inertia (in 4 )
Figure 4. Representative Mode Shape. L Length (in)

Once again, Ker Wilson [2l and Nestorides [8\ provide

The natural frequency associated with a particular mode shape equations for virtually every shaft configuration to be found in
can be easily identified by counling the number of nodes in the practice.
mode shape plot. Nodes are points thaI undergo zero detlection and Since the model is the foundation upon which the entire analysis
are located at all poinls where the mode shape plot crosses the l(- procedure is based, it is imperative that it represent the actual
axis. The lowest natural frequency is known as rhe fundamental machine accuralely. General guidelines for generating good
and has a mode shape with only one node, as is illustratl'd in Figltre m()del~ arc as follows:
 PRACTICAL DES IGN AGAINST TORSIONAL VIBRATION 193

• Disk elements are usually axially located at the center of gravity water. Ker Wilson [2] gives the following equation for the inertia
of the impeller that they represent. of the fluid:
• If a disk element is extremely rigid, the portion of the shaft that
lies within that element is assumed to have zero deflection. Shaft lf = .25 • lp • PR (3)
eL:;ment lengths are, thus, calculated up to the face, not the
centroid. of such an impeller. whe re:
lf Inertia of entrained tluid
• If a disk clement is not extremely rigid. its stiffening effect on the
~~
Dry inertia of propeller
shaft carrying it is modelled by assuming that the shaft ends at Propeller pitch/diameter ratio
what Nestorides [81 calls a point of rigidity within the impe ller. As
is illustrated in Figure 5, the shaft is assumed to defl ect in its Although the same correction should probably also be made for
normal fashion up to this point. Beyond this point, there is no pump impellers, there are no reliable methods, other than the use
deflection. Nestorides [8] provides equations for locating the point of test data, that the authors are aware of. The authors have often
of rigidity for several common configurations. merely used the dry inertias without encountering any problems.

Sh aft Assumed to - Geared Systems

Tw is! Freely Over
Tn is Length IDiSK When creating models, systems containing gear meshes require
special handling. Since all of the various shafts run at different
speeds, to facilitate solution, it is customary to convert the

~}
elements' parameters to the equivalent values that they would have
V Pointof
if they were all on the lowest speed shaft. The resulting equivalent
SHAFT Rigidity
~ single shaft model has exactly the same dynamic characteristics

-
and natural frequencies as the actual system. The concept of equiv-
alent values is analogous to the common practice of combining
-1 No DeHection
Assumed Over
This Length
electrical resistors or mechanical springs in series or parallel to
obtain an equivalent resistance or spring rate.
The use of an equivalent system assumes that the meshing gears
rotate together without separation throughout the entire vibratory
Figure 5. Point of Rigidity. period. Assuming no separation means that each gear mesh con-
tributes only one degree of freedom to the system, despite the fact
• When a shaft is joined to a nonrigid coupling or impeller by an that there are two disks in each mesh. For a given gear mesh, the
interference fit, the shaft should be assumed to twist freely over a parameter values for elements on the low speed shaft are
length equal to one-third of the overlap. The remainder of Ihe unchanged. However, elements on the high speed shaft must be
overlap s.hould be assumed rigid. transformed via the following equations:
• When a shaft is joined to a nonrigid coupling or impeller by a
keyed joint, the shaft should be assumed to twist freely over a 1eq = 1. N2 (4)
length equal to two-thirds of the overlap. The remainder of Ihe
overlap should be assumed rigid. (5)
• lItihzation of some solution algorithms requires Chat shaft
where:
clements be assumed massless . If this is Ihe case, it is usually suf-
leq Equivalent inertia referenced to low speed shaft
t1cien! to apply one-half of the actual shaft inertia to each of the
1 Actual inertia
disks on either end of the shaft element.. However, if the inertia of
N Gear ratio (N > 1.0)
a shaft element tums out to be of a comparable magnitude to those
keq Equivalent stiffness referenced to low speed shaft
of the major disks in the system, a more accurate procedure is
k Actual stiffness
called for. In this case, it is best to divide the shaft e'l emw! into a
number of disk and shaft elements, with each disk representing a
portion of the shaft's inertia. The above equations are utilized, one gear mesh at a time, to
replace two shafts by an equivalent one. The parameters for the
• Couplings should be modelled as a shaft having the coupling's high speed shaft are all increased by the square of the gear ratio to
spring rate between two disks whose inertias are each equal to one- reflect the larger energy levels they operate at. The equivalent
half of the coupling's total inertia. elements are axially located on the equivalent shaft in the same
• Flanges should be treated as shaft elements having diameters positions they occupy in the actual arrangement. Equations (4) and
equal to their bolt-circle diameters. (5) are used sequentially on each gear mesh in the assembly until
all parameters have been referenced to the lowest speed shaft in the
• When a distributed inertia is divided up into shaft and disk system. Once a single equivalent shaft has been obtained, any of
elements, the accuracy of the model increases with the number of the well known solution procedures, which will be discussed
elements. shortly, can be implemented.
• Although gear teeth have inherent flexibility, for most practical
cases, they can be considered to be torsionally rigid. In general, Nonlinear Couplings
gear tooth flexibility is significant only in the calculation of very
Many couplings that include rubber elements to provide
high natural frequencies or when a system contains multiple gear
damping have decidedly nonlinear stiffness characteristics similar
meshes. If it is desired to account for tooth flexibility, Nestorides
to those shown in Figure 6. The effective spring rate, which is the
[81 should be consulted for the appropriate equations.
slope of the curve, is, therefore. a function of the applied torque
• Disks that represent propellers operating in wate r should have and angle of twist. This nonlinear feature makes the calculation of
their inertias increased to account for the mass of the entrained natural frequencies a somewhat more difficult task.
194 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

Torque This procedure is repeated until all natural frequencies have

been found. Obviously, the higher natural frequencies will not have
a I X running speed that lies within the operating range. These fre-
quencies should have their corresponding running speed calculated
using 2X excitations or the lowest order number excitation that is
appropriate.
It is seen that the presence of the nonlinear coupling causes the
Spring Rate =Slope system's natural frequencies to be dependent on speed, as is often
the case in lateral systems. The above procedure assumes that the
natural frequency is the value that occurs at the speed where
Angle of Twist
resonance with the lowest order excitation possible occurs. These
are the most important natural frequency values since low order
excitations are the ones most likely to cause problems in the field.

Hydralllic Couplings
Another component that requires special attention when
modelling is the hydraulic coupling. Hydraulic couplings consist
of two radially vaned impellers that are mechanically independent
Figure 6. Nonlin.ear Coupling Behavior. of one another. Torque is transferred from the driving to the driven
shaft via kinetic energy of the working nuid. Accordingly, these
To ov(! r.~ ome this difficulty, the authors reco mmend utilization, devices afe also called hydrokinetic couplings.
of a techn ique provided 'b y Andriola [22] . The method consists of For any operating condition, the torques can-ied by the two
converting the co upling's torque vs diSJilacement curve into a curve shafts are identical. Since the coupling cannot transmit power at
of coupling spring rate vs shaft speed, as is illustrated in Figure 7. 100 percent efficiency, .a slighl reduction in speed occurs across the
Since the spring rate is the instantaneous slope ·o f the torque-d is- coupling. This reduction is referred to as slip and typical values are
placement curve, its relationship to transmitted torque is known . A from one to three percent of the driving shaft's speed. The slip per-
curve similar to Figure 7 can then be generated using the load's centage is generally independent of operating speed and varies
torque-speed charact~r,i stic. inversely with the transmitted torque.
By virtue of the speed difference, the nuid in the driving
impeller is subjected to a higher centrifugal force than that in the
follower. A circulating flow pattern is, thereby, established in the
coupling. Flow moves outwards in the radial passages of the driver
and inwards in those of the follower. This circulating now is the
essential mechanism by which torque transmission transpires.
(/) Hydraulic couplings are sometimes used as speed reducers. In
(/)
CD
general, they can be designed to yield virtually any speed ratio
c: between the two shafts, with the restriction that the driving shaft's
:t::
;; speed must be the higher of the two. As noted previously, unlike
(J)
geared reducers, the two shafts experience equal torques. There is,
C>
.!;;; thus, no need to reflect inertias and spring rates across the coupling
a.
::J
by the square of the speed ratio.
o With regards to modelling, a review of the literature uncovered
U a nearly unanimous opinion that hydraulic couplings should be
treated as zero spring rate elements that effectively divide the
assembly into two independent torsional systems. The lone dis-
senting voice, however, was that of Ker Wilson [2], hardly one to
be treated lightly. He claims that the zero spring rate model is only
an approximation since, in reality, restoring forces in the nuid yield
a small but finite spring rate. These restoring forces are generated
Running Speed (rpm) by centrifugal forces in the nuid and are, therefore, proportional to
Figure 7. Nonlinear Coupling Stiffness Vs Speed Relationship. the square of running speed. He provides the following equation
for effective stiffness:
Once the relationship between spring rate ano running speed is
kh = 0 5 • RPM2 / 585 (6)
determined, then an iterative procedure is implemented to find the
natural frequencies. The steps to be taken are as follows:
where:
• Guess a value for the natural frequency. Hydraulic coupling spring rate (in -lbf/rad)
kh
• Assuming a once per revolution excitation, calculate the running o Outside diameter of coupling impellers (ft)
speed corresponding to that natural frequency. RPM= Speed of driving member (rpm)
• Using this speed and the curve of coupling spring ratc liS running
speed, determine the instantaneous coupling stiffness. Thus, when modelling assemblies contammg hydraulic
couplings, the user has two options. Firstly, the machine can be
• Using this coupling stiffness, calculate the natural frequency. treated as a single entity and the above equation can be utilized to
• If the calculated natural frequency matches the guess value, a determine the coupling's effective stiffness. On the other hand, the
solution has been found. If not, go back to the beginning and try coupling can be assumed to have zero stiffness and the two
another guess value. resultant systems can be modelled separately. Even Ker Wilson [21
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 195

acknowledges that the latter treatment is usually satisfact.ory for systems is Holzer's method. The basis of this method is that free
practical systems. vibration can occur with no external torques acting on the system
only if the vibration frequency is a natural frequency of the system.
Analysis Methods In this procedure, a guess value of frequency is selected and a
In this section, methods for calculating a system's torsional Holzer table is generated. The table is statted by assuming that the
natural frequencies are described. This section is intentionally first disk in the system vibrates with the arbitrary amplitude of one
made brief because, in the authors' experience, determination of radian. The torque required to generate this vibration is supplied by
the natural frequencies is usually not a problem for most turboma- the adjacent shaft and is calculated and tabulated. Each disk is then
chinery users. It is in the later utilization of this information where sequentially looked at and the torque ill the shaft behind it required
most users encounter difficulties and the paper's main thrust is to sustain vibration is calculated.
directed accordingly. These calculations are continued until the last disk is reached.
As stated previously, the natural frequencies of a system are the This disk, of course, has no shaft behind it. Thus, the torque
frequencies that it can vibrate at indetinitely, in the absence of required to keep this disk in motion reveals the system's mode of
damping, without any external forcing function up.plied. The vibration. If this torque is positive or negative, it must be supplied
essential method for obtaining natural frequencies is similar to by an external source and the system can only execute forced
those used for any generic spring-mass system. Equations of vibration at the guessed frequency. However, if this torque is zero,
motion are written for every disk in the system, using Newton's the system is in free vibration and the guessed value is a natural
laws. Only spring and inertial torques are involved since there are frequency. -
no external torques and damping is ignored. The motion of each The above procedure is repeated with various guess frequency
disk is then assumed to be perfectly sinusoidal, as follows: values until the desired natural frequencies are found. Each guess
value for the natural frequency requires its own Holzer table. The
e = e" • sin rot (7) curve presented in Figure 8 is utilized to choose successive guess
values and zero in on the natural frequency. This curve is a g;neric
where: plot of the torque required to vibrate the system's last disk, known
e Angular position of a given disk as a function of time as the residual torque, vs the guessed value of natural frequency.
80 Amplitude of angular vibration All frequency values where the curve crosses the x-axis are natural
ro Angular frequency (rad/sec) frequencies. It is seen that odd natural frequencies are approached
t Time (sec) from positive residuals and even ones from negative residuals. The
analyst, thus, knows in which direction to move the guess value
This assumption implies that aU disks vibrate in phase with each based on the sign of the residual torque. For instance, if the second
other which is known to be tfue for undamped free vibration. natural .·requency is being sought and the residual torque resulting
Equation (7) .is then differentiated twice and (he results are substi- ~rom the l-ast guess is positive, the next guessed frequency should

tuted into the original equations of motion. The values of <.t) that be made lower.
solve these equations are the natural frequencies.
Utilization of the above method is straightforward when dealing
with very small numbers of disks. For example, the simplest
system imaginable, consisting of a single disk attached to a 1st Natural
grounded spring, can be shown to have the following natural Frequency
frequency:

ron = (k / J) .5 (8)
I Frequency (0))
where:
ron Natural frequency (rad/sec)
k Torsional spring rate (in -lbf/rad)
J Mass polar moment of inertia (lbf- in-sec 2 )

This is seen to be of identical form to the natural frequency

equation for a simple linear mass-spring system. It is seen that
Figure 8. Residual 7iJrque Behavior in Holzer's Method.
inertia is analogous to mass and torsional stiffness is analogous to
linear spring rate.
The mechanics for generating Holzer tables are explained in
The next basic system contains two disks separated by a single
detail in Den Hartog's classic book [23J and many other vibration
shaft, as was shown in Figure 1. The natural frequency for this
system is: texts and are omitted here . However, it should be noted that once a
natural frequency is found, the cognizant table contains other
useful information including the mode shape, the inertial torque
(9)
acting on each disk, and the angle of twist in each shaft.
Holzer's method lends itself very nicely to programming on a
The above procedure has been used for other combinations of
digital computer. Computers can proceed through the mechanics
shafts and disks and the resulting equations can be found in many
and perform the necessary iterations virtually instantaneously.
introductory vibration texts. However, once the number of degrees
of freedom exceeds three, closed-form solutions, if available, There are many natural frequency programs in existence today that
are based on Holzer's method or variations thereof.
become extremely unwieldy. For such systems, other methods
There arc other popular procedures in existence that bill them-
should be utilized.
selves as transfer matrix methods. Each shaft and disk element in
a system has an associated transfer matrix which describes the rela-
Holzer's Method
tionship between the torques and displacements on either side of
Probably the most well-known procedure for analyzing torsional the element. Transfer matrix elements. therefore. are functions of
196 PROCEEDINGS OF THE TWENTY-FIFTH TURI10MACHINERY SYMPOSIUM

inertia, stiffness, and frequency. In these procedures, all of the indi- occasionally generate el1'Ors. Thus, the authors advocate that an
vidual element transfer matrices are multiplied together to obtain independent hand calculation of the fundamental natural frequency
expressions for the torque and displacement of the last disk in be made to serve as a check for the computer analysis.
terms of those at the tirst. Since this is essentially what Holzer's There are several means by which this hand calculation can be
method does, transfer matrix methods can be considered as performed. The simplest case occllrs when one of the couplings has
variants of Holzer's method. a spring rate that is an order of magnitude lower than that of any of
the other shaft elements. In this situation. which is not at all
Matrix-Eigellvalue Methoils uncolllmon in turbomachinery, virtually all of the deflection in the
Most other procedures used for undamped analysis today fall fundamental mode will be taken in the springy coupling. The
under the heading of matrix-eigenvalue methods. Th ~e methods machine t:an then be approximated as a two disk system by simply
are essentially the same as the basic method described for simple adding together all of the inertias on each side of the coupling.
systems earlier since they also involve solution of the dillerential Equation (9) can then be used to obtain the natural frequency.
equations of motion. The only major difference is that matrices are The same basic procedure should also be utilized for more
utilized to simplify the math. general syStems. The basic objective is to reduce a complex system
Matrix methods are begun by writing the equations of motion in to a reasonable approximation that has three or less degrees of
matrix form. The general undamped equation is: freedom. tr this is done, a closed form equation like Equation (9)
can be straightforwardly implemented.
[kl' fe] + [1]' [al = [0] ( 10) Two basic princi,p'les are utilized in the reduction of systems.
The tirst is that relatively small inertias have very little effect on the
where: fundamental frequency. These disks should, thUS, be ignored and
[k] Stiffness matrix the shaft elemcnts on either side of them can be combined as
[1] Inertia matrix springs in series. The second principle is that shafts having rela-
[e] Angular displacement vector tively large spring rates behave as if they were rigid in the
[a] Angular acceleration vector fundamental modc. Therefore, these elements should be discarded
and the inerli'as on either side of them can be added together.
Both the stiffness and inerlia matrices are square matrices In addition 10 comparison of the two fundamental frequency
having 1he same number of rows as there are disks in the system. values, several other items can be used to validate the analysis. In
Additionally, the inertia matrix is diagonal. The stiffness and systems that have a single weak link coupling, there should be a
inertia matrices are generated from the knowfl disk inertias and considerable gap between the first and second natural frequencies.
shaft spring rates. The two vectors have the same number of rows Additionally, as stated above, the fundamental mode shape should
as the matrices and represent the displacements and acce lerations have ,t he majority of its deflection in the springy coupling and the
of the individual disks . These are the unknowns to be solved for. node sbould be located somewhere in this vicinity. Furthermore,
Once Equation (10) is generated, the next step is to assume the displacements in a two disk system obey the following
simple harmonic motion as follows: equation:

(13)
tel = [eol • sin wt (II )
Thus, whichever side of the machine is represented by the lower
The above is the matrix equivalent of Equation (7). Substituting
inertia disk should exhibit more displacement in the fundamental
the above into Equation (10), the so-called eigenvalue equation is
mode shape.
obtained:
For more general systems, the mode shapes can be checked to
see if they obey the following equation which represents conserva-
1[1]-1 • lk] - [co 2]} • tel = [01 (12)
tion of angular momentum:
where: ( 14)
[w 2 ] = Diagonalized eigenvalue matrix
Thus, with signs accounted for, the sum of the products of the
Matrix methods, which can be found in many mathematics texts, inertias and their displacements should equal zero. This rule can be
are then used to tind the eigenvalues. These are the values of w 2 used to check any mode shape either by hand calculation or by
which satisfy the above equation. The Hatural frequencies are then eyeball.
obtained by merely taking the square roots of the eigenvalues. If the above checks are utilized to validate the analysis, the user
Associated with each eigenvalue is an eigenvector, [8]. These can be reasonably contident that the results are accurate. The
vectors provide the mode shape corresponding to each natural undamped analysis can then be considered complete and the next
frequency. It should be noted that the assumption of no damping in step, generation of the Campbell diagram, can be taken.
the system results in the eigenvalues being purely imaginary and
the eigenvectors being real. GENERATION OF CAMPBELL DIAGRAMS
As is the case with Holzer's method, there are many computer As was stated previously, one of the primary objectives of
programs available which perform the above calculations. torsional vibration analysis is the identification of all potential
Regardless of which method is used, several experts estimate that resonant points. Since most practical systems have numerous
the resulting natural frequencies are accurate to within three to five natural frequencies and multiple sources of excitation, this is, by
percent. This naturally assumes utilization of a reasonable model. no means, a trivial task. A device which greatly aids in the deter-
mination of these points is the Campbell diagram (aka interference
Results Verification diagram). In addition, Campbell diagrams provide an excellent
After the natural frequencies and mode shapes are obtained from overview of the system's torsional vibration situation, analogous to
the computer. many would consider the undamped analysis to be the function provided by critical speed maps in lateral systems.
complete. However, the authors recommend taking one more step A Campbell diagram should always be generated as soon as the
because computer solutions and analysts have been known to undamped analysis is completed. A representative diagram for an
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 197

ullgeared system is depicteJ ill Figure 9. The natural frequencies Hertz by the order number. Connection of this point with the origin
are plotted as horizontal lines and the operating speed range is des- by a straight line generates the desired excitation line.
ignateJ by vertical lines. When plotting the speed range, the upper The most comlllon excitations are at once per revolution (I X)
limit should represent the point where the ovcrspeed trip system and twice per revolution (2X) . These arc generic sources that can
kicks in m.ther than the highest desired running speed. The upward arise via various mechanisms. I X excitations can be generated by
sloping lines are harmonics of speed that represent the system's conditions such as rotating unbalance. eccentricity, and misalign-
potential excitations. Intersections between these lines and the ment. Excitations of 2 X are usually due to misalignment,
natural frequency lines that occur within the operating speed range ellipticity. or certain non-circular shaft cross-sections such as
are referred to as interference points and are indicators of potential keyways. It is the standard practice of the authors to include both
resonances. There are two of these points illustrated in Figure 9. of these excitations in the analysis of all torsional systems.
The speeds cOlTesponding to interference points are known as In addition to these generic excitations, each system contains
critical speeds. individual components that also generate excitations. In general ,
Although determination of the horizontal and vertical lines is these sources include the system driver, such as a motor or turbine.
rather straightforward, generation of the excitation lines requires and load, which is often an impeller. Additionally, all gear meshes
considerable insight. There are many potential excitation sources are excitation sources . The various excitation sources will be
which occur in common turbomachinery drive trains . Essentially, specifically addressed in the text to follow.
any mechanism which is capable of generating a periodic fluctua-
Gear E-.:citativtls
tion in the system's transmitted torque is a potential excitation
source. Gears generate pulsations at several different frequencies. Errors
in the generation of the gear teeth and in the mounting of the gear
Operating hub on the shaft can lead to unbalance, eccentricity, ancVor mis-
Speed
alignment. Such errors lead to fluctuations at a frequency of once
Range
per revolution of the cognizant shaft. Additionally, any errors that
result in gear ellipticity generate torque variations at twice shaft
200 speed. Thus, the generic I X and 2X excitations previously alluded
to must always be considered in gear meshes.
Furthermore. gears can produce disturbances at their meshing
frequency and higher harmonics of it. For each shaft, meshing
N frequency is equal to the number of teeth on that shaft's gear mul-
E. 150 tiplied by shaft rpm . These disturbances can be attributed to a
>.
u phenomenon which Schlegel. et a1. 124]. refer to as engagement
c:
Q)
:J
impulse.
CT Whenever a given tooth meshes, it inherits a portion of the load
~
u.. 100 that had been carried by the previously engaged teeth. The removal
~ of load from these teeth, which had been deflected by the loading,
:J
<a allows them to relax and spring back to their free position. This
z relaxation generates a tangential acceleration in the gear bodies
50 which prevents the newly engaged teeth from meshing smoothly.
The meshing is, therefore, impactive. As a result, an impulsive
force is generated in each gear along their common line of action.
f, = 20 Hz
causing the mesh's transmitted torque to pulsate each time a new
tooth comes into mesh.
O~----~-----L----~------~----~--~
o 1000 2000 3000 4000 5000 6000 Whenever gears are present in a system, therefore, the
interference diagram for each shaft should include three excitation
Speed (rpm)
lines. Their corresponding order numbers should be equal to one,
Figure 9. Represelltative Campbell Diagram. two, and the number of teeth on that shaft's gear. Technically,
excitations corresponding to higher harmonics of gear mesh
frequency should also be included. However, all sources consulted
In general, the excitations occur at integral multiples of the shaft dismiss these hannonics because of their negligible magnitudes.
speed. These multiples. which represent the number of vibrations
which occur during each shaft revolution. are referred to as order Impeller Excitatiolls
numbers. For any given interference point, the order number is Another common excitation source is the vaned impeller which
equal to the natural frequency divided by the critical speed, when may act as either the load or the driver of the system. This category
each arc expresseJ in the same units. Excitations having order encompasses a wide variety of dynamic energy transfer devices
numbers of one. two. and four are depicted in Figure 9. including pump impellers, compressor and turbine rotors, and fans.
In all of these devices, torque variations occur at blade-pass
Excilatio/lS frequency due to pressure disturbances resulting from vanes
It is customary, as is shown in Figure 9, to use mismatched units passing a stationary object such as a volute or diffuser entrance .
in Campbell diagrams. The abscissas are usually expresscd in rpm The order number for an impeller is, thus, equal to its number of
while the ordinates are in Hertz. Thus. generation of the excitation vanes .
lines is not as straightforward as it would be if consistent units Additionally. impellers occasionally operate within casings that
were utilized, since in that ease the slope would be merely equal to contain several equally spaced obstacles that can generate pressure
the order number. fluctuations when passed by an impeller blade. Such a situation
To generate an excitation line of a given order. a speed on the occurs in configurations employing vaned diffusers or volutes with
graph should be arbitrarily selected. The abscissa value for that l1lultiple cutwatcrs. These casings can generate excitations having
speed should thcll he divided by 60 to convert the speed into Hertz . an order number equal to the number of stationary vanes or
The ordinate is then obtained by Illllltiplying the abscissa value in cutwater.-;.
19S PROCEEDINGS Of THE TWENTY-FIFTH TURI:lOMACHINERY SYMPOSIUM

Furthermore, it can be easily seen that disturbances can also be In addition to the above, AC synchronous motors also generate
generated each time a rotor blade pascses by a stationary vane or large: pulsating torques at twice slip frequency during starting. The
cutwater. The order number generated by this phenomenon is given discussion of this unique phenomenon will be held off to a later
by the following etlufltion .·mm Ker Wilson 1211: section.
In general. DC machines generate excitations that are relatively
( 15) small during steady operation . Additionally, when exposed to a
short, DC machines merely experience an abrupt change in torque
where:
kve!. Unlike AC machines, there are no torque pulsations. There
n Excitation order number
are many applications, therefore , where DC motors and generators
Nr Number of blades on rotor
are non-factors with respect to torsional vibration.
N, Number of stationary vanes or cutwaters
Ch Highest common factor of Nr and Ns Variahle Frequency Drive Excitatiolls
Thus, a vaned impeller operating within a vaned diffuser can All of the AC machines discussed so far have been assumed to
generate three separate excitations. operate at constant speed. These machines are supplied with a
constant electrical frequency equal to line frequency. In contrast to
Propeller E\cilUtillllS these, mallY turbomachinery users are opting for the tlexibility
Propellers behave in much the same fashion as bladed impellers . provided by variable speed motors. Such motors are controlled by
In marine applications, the reactiol'l torque from the water varies variable frequency drives (VFD) which alter motor speed by
each time a blade passes the ship'.s rudder. Thus, propellers also varying the electrical frequency supplied to the motor's terminals.
generate torque fluctuations at blade-pass frequency, althougJ1 the These drives represent an additional source of torsional excitation
tluctuations are usually much larger than in impellers. Propellers to be oonsidered during the analysis phase.
also create torque variations at Integral multiples of blade-pass Variable frequency drives contain a static frequency converter
frequency but, by all accounts, they are usually negligible. which electronically transforms the constant line frequency (60 Hz
in the United States) to the desired electrical frequency for driving
Electrical Machine Excilllliolls the motor. Since the speeds of synchronous and induction motors
are essentially proportional to electrical frequency, any desired
The pulsating torques that arise in electric motors and generators speed may be obtained by merely selecting the appropriate driving
can be separated into two categories; those that occur when the frequency. Most static converters are capable of producing electri -
machine is running at constant speed and those that transpire when cal frequencies that arc both above and below the supply frequency.
the machine is accelerating or exposed to an electrical fault . Most static converters consist of a rectifier in series with an
Although both categories mllst be investigated during the analysis inverter. The rectifier converts the AC signal at line frequency into
phase, the latter is the more likely to cause serious problems. a DC signal. The DC signal is then subsequently converted back
Therefore, in contrast to all of the excitations addressed heretofore, into an AC signal at the desired frequency by the inverter. The
electrical machine excitations are often of a transient nature. outputs of the inverter are voltage and current waves that are very
In general, the best source of information regarding excitations nearly sinusoidal. However, since they are not perfect sinusoids,
produced by electric machines is the manufacturer. There are periodic variations in the driven motor's output torque are created.
virtually an infinite number of designs for motors and generators There are several different types of variable frequency drive in
and each one has its own unique excitation mechanisms. However, common usage today. The speed of induction motors is usually
to familiarize the user with the more common excitation sources, varied via control of the stator frequency. Pulse width modulators
some basic guidelines will be provided . (PWM), voltage source inverters (VSI), and current source
When running at steady speed, motors and generators can inverters (CSI) are popular devices that accomplish this.
produce torque fluctuations via various mechanisms. Most AC Synchronous motors arc usually driven by load commutated
motors and generators produce fluctuations at line frequency (60 inverters (LCI) which are similar to CSIs.
Hz in the United Stales) and twice line frequency by a number of The excitation characteristics of the various devices that control
di fferent phenomena. Thus, systems containing AC machines have stator frequency (PWM s, VSIs, CSIs, and LCls) arc practically
horizontal excitation lines at line and twice line frequency. identical. In these drives, the excitation frequencies are given by
Additionally, many machines create oscillations having an order the following equation:
number equal to the number of magnetic poles in the machine .
In addition to the oscillations produced dunng steady runnmg, (16)
AC machines generate torque fluctuations when subjected to short
circuits across their terminals. These Iluctuations also occur at line where:
and twice line frequency. Although these excilations lire of a fex Excitation frequency
transient nature, the magniludes of the peak torques are usually fe E1eclrical frequency output by inverter
many times the motor's rated torque. Because of this, most authors n Number of pulses in converter
strongly advise avoiding natural frequcflcies near Wand 120 Hz in k 1,2,:3 ...
assemblies containing AC motors· or generators, if at all possible.
Another situation in which AC motors generate f1uctualing
torques is in the first instants after power is initially applied to In both synchronous and induction motors, the motor speed is
them. Both induction and synchronous motors generate large approximately related to the supplied electrical frequency by the
following:
transient torques at line frequency (hat die out rapidly. However,
since the transients take much longer to decay ill induction motors,
they arc far more likely to generate problems via this mechanism . RPM = 120· fe / Np ( 17)
Thus, although the analysis of systems utili zing AC machines is
anything but ~imple, the generation of the excitation lines on the where:
Campbell diagmm is relatively straightforward. In gencral, the RPM = Motor speed (rpm)
excitations are at 6U Hz, 120 Hz, and sometimes at pole-passing fe Electrical frequency (Hz)
frequency. Np Number of poles in motor
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 199

The excitations are. therefore. upward-sloping lines whose order Operating Speed
number is obtained by combination of Equations (16) and (17), as Range
follows:

nord = n • k • N p I 2 (18) 2000

VFD
.The value of n is either six or 12 since all practical converters
utIlize SIX or 12 pulses. Since the number of poles must be an even
number, the order numbers obtained from Equation (18) are always
N 1500
Integers. E.
In contrast to the preceding devices, wound rotor induction VFD Excitations
{j'
c: Inac1ive In This
motors are sometimes controlled by drives that Mayer 125] refers CD
:::l Region
to as static Kramer drives. These drives are very di fferent from CT

those previously discussed since they do not alter the electrical £ 1000
frequency supplied to the motor's terminals. Instead, they control
motor speed by varying the motor's slip. Unlike the other YFDs,
whose excitation frequencies increase with motor speed, static
Kramer drives act in the opposite direction. These drives produce 500
pulsations at harmonics of the slip frequency. which is defined as
follows :

(19)
5000 6000
Speed (rpm)
Slip frequency
Line frequency Figure 10. Variable Frequency Drive Excitatiolls.
Electrical frequency corresponding to motor speed
state forced vibration analysis. In stark contrast to these are the
torque pulsations that are created by AC synchronous motors
The generated IlaITllonics are either multiples of six or 12,
during the starting process . Interferences resulting from this phe-
dependent on the number of pulses in the drive. The relationship
nomenon can usually only be evaluated via a transient forced
between electrical frequency and motor speed is given by Equation
vibration analysis.
(17).
Unlike induction motors, synchronous motors are not self-
In the speed range between zero and 10 percent of maximum
starting. This is due to the fact that the stator's magnetic field
speed, most variable frequency drives produce t1uctuating torques
begins rotating at synchronous speed virtually instantaneously
that are several times larger than those generated over the
after power is applied. With the stator field rotating and the motor
remainder of the speed range . Because of this, most users choose
at rest, alternating forward and reverse torques are applied to the
to begin the active operating range at a speed above 10 percent.
rotor. This causes the rotor to swing back and forth by minuscule
Thus, when induction motors are started, they behave in the same
amounts and effectively prevents the buildup of any significant
m~nn~r as in a fi~ed speed system since the variable frequency
accelerations in either direction.
drIve IS kept out ot the loop during starting. Once the speed reaches
Since synchronous motors are not self-starting, they are
the ?perating region, the drive is activated and its torque pulsations
normally equipped with squirrel-cage windings (aka amortisseur
are mtroduced to the system . Thus, in the interference diagram for
win.dings) which provide starting torque and also provide damping
such systems, the drive's excitations are shown only within the
dUrIng steady state running. These windings are utilized to acce1-
operating speed range, as is shown in Figure 10.
er.nte the motor as an induction motor from zero speed to a speed
The above discussion generally applies only to induction
slIghtly less than synchronous speed. Starting is usually performed
motors. When synchronous motors are controlled via LCIs. the
WIth no voltage applied to the rotor's field winding. When syn-
LCI is usually active throughout the entire starting process. The
chronous speed is approached, DC field voltage is applied and the
assembly must, therefore, endure the higher pulsating torques
rotor is pulled into synchronism.
produced by the LCI in the low speed range. This is usually a
Pulsating torques are created during this procedure due to the
transient condition. however, since the operating range is normally
fact that synchronous motor rotors contain salient poles that are
set above 10 percent speed.
magnetic protrusions enclosed by field coils. The resulting
In addition to the pulsations generated by the variable frequency
asymmetry causes the motor's output torque to vary as a function
drive, the normal t1uctuations produced by AC machines are also
of rotor position. This effect is in direct contrast to pure induction
present. When YSI class drives are employed, these pulsations
motors which have symmetric rotors and a generated torque that is
occur at one and two times the electrical frequency across the
independellt of rotor location.
motor's terminals. The order numbers for these pulsations can be
Synchronous motors are often modelled as having two axes of
obtained from Equation (18) using values of one and two for nand
symmetry, the direct axis and the quadrature axis. The direct axis
one for k. These excitation lines, thus, have shallower slopes than
refers to a centerline that passes directly through one of the rotor's
those produced by the YFDs . In systems using static Kramer
salient poles. When the direct axis is in perfect alignment with the
drives, the puls~tions merely occur at line and twice line frequency
magnetic field set up by the stator, the magnetic reluctance
SInce the electrIcal supply frequency is not variable.
between rotor and stator reaches a minimum value. Accordingly,
the torque generated in this position, known as the direct axis
Synchronous Motor Startllp Excitations torque, represents an extreme (maximum or minimum) value
The majority of the preceding excitations transpire when the produced during a revolution of the rotor.
machine is running at constant speed. Any potential damaging On the other hand, the quadrature axis is a centerline which is
mterferences gent:rated by them could be evaluated by a steady perpendicular to the direct axis. The condition where this axis is
200 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

aligned with the stator's magnetic field represents the point of United States. As the motor is accelerated, the excitation frequency
maximum magnetic reluctance. The resulting torque, which is decreases linearly until it reaches zero when the rotor achieves syn-
referred to as the quadrature axis torque, represents the opposite chronous speed.
extreme from the direct axis torque. A Campbell diagram is presented in Figure 12 for a hypotheti-
Accordingly, as the rotor rotates, the torque varies in a roughly cal system being driven by a synchronous motor. Synchronous
sinusoidal fashion between the limits imposed by the direct axis speed is assumed to be 1800 rpm and all other excitations arc
torque and the quadrature axis torque. The mean value of these two omitted from this figure for clarity. It is easily seen that the motor
torques is known as the average torque and represents the torque generates interferences with all natural frequencies that are below
available to provide acceleration to the system. Superimposed on twice line frequency. assumed to be 120 Hz.
top of this torque is a pulsating torque whose magnitude is one-half
the difference between the direct and quadrature axis torques. The 140 f4 = 136 Hz
various torque components are illustrated in Figure II as functions ~------------------~------
of speed for a hypothetical synchronous motor. 1 Twice Line Frequency = 120 Hz

Crossover
100 Point
Quadrature Axis Torque
100
Q)
:::l
80
2' f3 = 86 Hz
~ N
~ 60 ;;. 80
iii >. Synchronous Motor
a: 0
c: Excitation at 2x
"0 - - -- Q)
:::J
&i 40 0- Slip Frequency
60
~
rf Direct Axis Torque ~
20
40
Synchronous
20 40 Speed = 1800 rpm
Percent of Rated Speed
20
fl = 13 Hz
Figure II. Synchronous Motor Torque-Speed Characteristics.

oLo-----4~00----~8700~--1~2~070--~~~~
The frequency of the torque pulsations is the rrequency at which
the stator's rotating magnetic field passes a rotOr pole. Since the Speed (rpm)
stator's magnetic field rotates at synchronous speed, the excitation
Figure 12. Synchronous Motor Campbell Diagram.
frequency is a function of the difference between synchronous
speed and rotor speed, which is known as slip spood. Specifically, The results of Figure 12 are, by no means, unique to the example
excitations occur at twice slip frequency where slip frequency is selected . In fact, most systems driven by synchronous motors will
defined by the following equation: experience resonances with all natural frequencies below 120 Hz
during starting. This is a significant source of potential problems
(20) since most practical turbomachinery drive trains have several
natural frequencies within this range.
where: The discussion has so far been based on the assumption that the
fstip Slip frequency (Hz) motor is started with no tield excitation. If the UScI' chooses to
fl Line frequency (Hz) apply field excitation during starting , the pulsating torques at twice
Ns Synchronous speed (rpm) slip frequency are unchanged . However, an additional pulsation at
N Rotor speed (rpm) slip frequency is introduced and the motor's average torque is
reduced. Since both of these consequences are detrimental , almost
The motor's synchronous speed is given by the following all synchronous motors are started without field excitation and that
equation: shall be assumed henceforth.
Since the resonances occur only during starting, they are of a
(21) transient nature and, thus, may not be as troublesome as resonances
occurring within the steady operating speed range. However, all
where : interference points generated by synchronous motors must be
Ns Synchronous speed (rpm) investigated further using techniques to be discussed herein . The
fl Line frequency (Hz) unique problems associated with analysis of synchronous motor-
Np Number of poles driven systems will be discussed in a later section.
It should be noted that the above discussion applies only to
The ramifications of Equation (20) are extremely important. It is motors driven at line frequency. If a variable frequency drive is
seen that the frequency of torque pulsations, twice slip frequency, utilized in the system, the motor is in synchronism at all speeds.
decreases as rotor speed increases . Thus, at zero speed, the excita- The twice slip frequency pulsations are, thus, absent. However, the
tion frequency is equal to two times line frequency or 120 Hz in the pulsations produced by the drive must then be dealt with.
 PRACfICAL DESIGN AGAINST TORSIONAL VIBRATION 20t

Geared Systell/ Exall/ple The first item to address in generation of the Campbell diagram
The excitation frequencies generated by each of the discussed is calculation of the order numbers for the motor and variable
sources are summarized in Table I. It should be noted that when an frequency drive. Since there are two poles in the motor, use of
AC motor is driven by a VFD, the excitations generated by both the Equation (17) reveals that the speed in rpm is equal to 60 times the
motor and the drive need to be accounted for. Once the relevant electrical frequency in Hz. Thus, the speed and electrical frequency
excitations are determined for a system, they should bc plotted on are equal when expressed in consistent units. The induction motor
a Campbell diagram to determine interference points. A hypothet- pUlsations, which occur at one and two times the electrical
ical example will now be provided to illustrate the general frequency, therefore, are represented as I X and 2x excitations on
procedure . the Campbell diagram.
The order numbers for the VFD are obtained by inputting six for
n and two for Np in Equation (18). The resulting excitations are,
Table 1. Swnma/}' of Excitation Sources Gnd Frequencies. thereby, at 6X, 12X, 18X, 24X, etc. For the sake of clarity only
Excilalion Source Excitation Frc--.9.ucncles
the first three excitations will be utilized in the example Campbell
G~ncri~ I,X (unbalance. ccc~n(rkily, 0,," x Speed
diagrams.
misalignment. ('tt:.}
The order numbers for the motor shaft are as follows:
Generic 1 X (misalignment. elliPticity. etc.) Two x S""erl Generic I X excitation: 1
Gear M..: ~ II Con!Oisting of Pinion Pinion Shafl: Generic 2X excitation: 2
with Np Teeth Maling One x Pinion Spt...'eJ
with Gear Having NG Teeth Two x Pinion Speed Induction Motor excitation: 1,2 (duplicates of the above)
N" x Pinion Speed Variable Frequency Drive excitation: 6, 12. 18
Gear SharI:
One ;It Gear Speed
Gear mesh excitation: 25
Two x. Gear S~cd
No x Gear Speed The cOlTesponding order numbers for the pump shaft are:
impeller wilh N. Black.'s Rotating NR x Speed Generic I X excitation: I
1nsidc Casing with Ns Cutw;.ucr.- Ns x Spe...d Generic 2X excitation: 2
n x S[>C<.'d (n i~ivcn b}'Equiltio n (15)
Impeller blade excitation: 8
r\C Momr or Generator ~dlh Np Pok:\ Line 1'''''llK.'flCY (60 II,) Pinion mesh excitation: 10
(Fi«d frequencJ or Twice line Frequency ("l20 HLl
Static Kra~rOrivc) Np I Spt....'<1
AC MOlar with N. Polos 11, • N. 1 Speed There are several options for displaying the above on Campbell
(VanabkFr<qucncy N". Sp«<! diagrams. Perhaps the simplest method is to prepare separate
Drive C.ofl(rolTing Stator)
diagrams for each shaft, as is illustrated in Figures 14 and IS. The
Variable Frequency Drive I/~ .l N ~ N, X Speed
(S"'OI' Frequency Control) with
natural frequencies are the same on both diagrams since the gear
N • N. x Spectl
N Pubrs Dnvin~ AC Motor 1.5 x N x N p ' Speed ratio has no impact on them. The excitation lines are simply drawn
wilh N p' Polt..~ 2 x N x N. x Spt."d at the slopes indicated by the above order numbers. which are all
Sialic Kramer Dri ve with N Puh~cs N x Slip Frequency integers. Both diagrams are then checked for interferences.
2 x N x Slip Fn..quency
It is seen from Figure 14 that there are six interferences, all des-
Synchronous Motor Two x Slip Frequency
ignated by circles, occurring in the motor shaft. Likewise, the
(Fix.d FrL"QuenCy Onv.)
pump shaft generates four interference points. It must be remem-
bered that any interference found on one shaft's diagram can cause
It has been the authors' experience that the generation of vibration problems in the other shaft since the two shafts are
Campbell diagrams for geared systems is a source of confusion for assumed rigidly connected at the gear mesh.
many users . To help to alleviate this, the example system consists Another method of handling the example system is to plot all
of a motor and pump operating on two different shafts, as is shown information on a single interference diagram, as is shown in Figure
in Figure 13. The motor is assumed to be a two-pole induction 16. In this case, both motor and pump speed ranges are indicated
motor driven by a variable frequency six pulse VSI drive. The on the diagram. All excitations are plotted on the figure at their
motor speed is assumed to vary between \000 and 1800 rpm. The actual slopes. A problem that arises from this method is the
gear mesh provides a 2.5 to I ratio, resulting in a pump speed range apparent indication of false interference points. An example of this
of 2500 to 4500 rpm. The pump impeller is assumed to have eight is intersection point B, between the second natural frequency and
blades while the pinion and gear tooth numbers are 10 and 25 the 8X line representing impeller blade excitation. Since this point
respectively. The variable frequency drive is inactive at all speeds is located within the motor speed range, it appears to be a bona fide
below the operating range. interference point. However, when it is remembered that the 8x
excitation is only related to pump speed, it is seen that this point is
PUMP
l:-1P(LLER meaningless. Likewise, points C, D, and E are also invalid.
PINION
Because of the large potential for this type of confusion, the
authors do not recommend the use of this method.
A much better method of plotting al\ of the relevant information
on one diagram is illustrated in Figure 17. In this diagram, every-
thing has been referenced to motor speed. The motor excitations
are drawn in the same manner as in the previous plots. However,
TV(] - Po....£ all excitations acting on the pump shaft must have their order
INDUCTION
H(]TOR numbers multiplied by the gear ratio in order to be referenced to
motor speed . For instance, the generic I X excitation on the pump
Figure 13. Example Motor-Dril"('n Pump. shaft is drawn with an order number of 2.5 . The other pump shaft
excitations are represented as 5X. 20x, and 25X lines.
Comparison of Figure 17 with Figures 14 and IS reveals that the
The undamped analysis is assumed to be complete and there are two methods yield the same interference points. For instance. point
four natural frequencies within the range of interest. These occur at A in Figure 15 is an interference with the fundamental mode at a
50, 220, 340. and 500 Hz. pump shaft spced of :1000 rpm. Point F of Figure 17 occurs at a
202 PROCEEDINGS OF THE TWENTY-FIFTH TUR130MACHINERY SYMPOSIUM

Motor
---. Speed ... - -
Range

600
600

500
500
N
;S
N
~ 6-
c:
(I) 400
:l ~ 400
0- c:
~
Q)
=> f3 = 340 Hz
0-
~
300 u..
300

'2 = 220 Hz
200 200 -

100 100

Motor Speed (rpm) Speed (rpm)

Figure 16. Comhil/ed Motor/Pump ClIInphell Diagrall1.

Figure 14. Motor Campbell Diagram.
25x
lOx
700
700

600

600
f4 = 500 Hz
500

'4 =500 Hz N
500 ;S
400
~
c: 13 = 340 Hz
Q)
:l
0-
N
6- 400 ~ 300
~
c:
(I)
f3 = 340 Hz
=>
0-
<I)
200 -
at 300

100
200
lx
500 1000 1500 2000 2500
100 Motor Speed (rpm)

Figure 17. Com/Jined Motor/Pump Campbell Diagram.

O~~~~~~~~~~~==_ motor speed of 1200 rpm which is equivalent to a pump speed of

o 1000 2000 3000 4000 5000
3000 rpm. The two points are, thus, identical.
Pump Speed (rpm)
The reader might notice that there are a total of 10 interference
points in Figures 14 and 15 while Figure 17 conlains on ly nine.
Figure 15. Pump ClIrlljJhell Diagram.
The exrlanalion for Ihis is Ihal point G from Figure 14 and point H
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 20J

in Figure 15 both represent interferem;es where the fourth natural points are subjected to inspection and only those that are not
frequellcy is excited by the gear meshing frequency. These points eliminated in this step undergo further analysis.
are, thus, redundant and are represented by a single point, point 1, The first step of the inspection procedure is the tabulation of all
in the combined diagram. Thus, the procedure of Figure 17 is interference points. For all points, the following three items should
perfectly acceptable as long as the introduction of non-integral be noted:
order numbers doesn't bother the user from an aesthetics stand-
• Excitation source
point.
• Running speed
INSPECTION OF INTERFERENCE POINTS • Normal mode which undergoes excitation
After the Campbell diagrams are completed, the next step in the The basis of the inspection method is that the only undesirable
analysis procedure is the identification of intersections between effects of torsional resonance are large cyclic torques and stresses .
natural frequency and excitation lines that represent true interfer- Unlike lateral vibration, the vibration amplitudes arc normally not
ences. The statement made previously that intersection points that a concern. Thus, the major concerns are of a structural nature. If
occur within the operating speed range are the only ones consid- the induced cyclic torques arc large enough to cause a gear mesh
ered to be interferences is not entirely true. After all, an or coupling to be overloaded, the situation is unacceptable.
intersection point shown by analysis to lie just outside of the speed Additionally, if the cyclic stresses are large enough to cause fatigue
range could, in actuality, be within the range if the analysis failure of a shaft, then there is a problem . Furthermore, if torque
contained inaccuracies. reversals occur at gear or spline interfaces, there could be a
In order to prevent the erroneous elimination of interseclion problem . However, any resonant point that does not lead to one of
points, the margin between the intersection point and the top or those three situations can generally be dismissed from
bottom limit of the speed range should be calculated using the consideration .
following equation : With the above criteria in mind, the first thing the analyst should
do is look at the amplitude of free vibration, as revealed by the
MAR = oN I N I • 100 (22) mode shape, at the location of the excitation torque. Since there are
usually multiple locations where the generic I X and 2x excita-
where: tions can occur, this process applies primarily to fluctuations
MAR = Margin (percent) arising from a specific source such as a motor, gear mesh, or
NI Top or bottom limit of speed range (rpm) impeller. If the mode shape reveals that the excitation source is
oN Difference between speed at intersection point and N I located near a node, the excitation will have virtually no effect on
(rpm) the system . Any interference point having this characteristic can be
immediately dismissed.
Once the margin is determined, the question arises as to how The above rule applies when the excitation is in the form of a
much margin is required to eliminate a potential interference point. pulsating torque, which is the type of excitation generated by all
To a certain extent, this is a matter of judgment. Obviously, one components discussed herein except for gears. As will be described
percent is insufficient since the accuracy of the natural frequency later, gears can generate two very different types of excitations,
calculation is no better than three to five percent. On the {Jther torques and displacements. Gear torque excitations behave in
hand, few analysts would be concerned if a 50 percent margin were exactly the same manner as all other torque excitations and are,
demonstrated. thus, subject to the above rule. However, displacement excitations
The API codes for pumps, compressors, and turbines specify a are radically different in that they are most effective when located
minimum margin of 10 percent between operating speed and inter- at nodes. Therefore, they cannot be eliminated on the basis of the
seclion poinl speed. A survey of the available literature reveals a above argument.
general consensus that the allowable margin should be somewhere The mode shape can also be used to select only one point to be
between 10 and 20 percent which is consistent with the authors ' analyzed for the case where a specific mode is excited by more
experience. In general, the authors require a minimum margin of than one sOurce. In general, the effectiveness of a given excitation
15 percent. Under special circumstances, the authors have is a strong function of its amplitude in the mode shape. Thus, if one
permitted mild deviations to this rule but under no circumstances point has an excitation located at a point of much higher amplitude
would they condone a margin of less than 10 percent. than the other points, that point will result in the worst case cyclic
Once the required margin has been decided upon, all intersection torques and stresses, as long as the excitation torque magnitudes
points unable to meet the requirement should be identified as inter- are about equal. The worst case interference point would then be
ference points. To these points should be added all transient submitted to a forced vibration analysis but the remaining points
intersection points generated by synchronous motor and variable could be ignored. Once again, this mle applies only to torque
frequency drive excitations. excitations.
Once the tnle interference points have been determined, the user A third way in whieh the mode shapes can be of assistance is by
generally has two choices for dealing with thent. Either desiglll their indication of the relative magnitudes of induced torques at
changes, such as alteration of couplings, are implemented to various locations . In general. the largest cyclic torques will occur
eliminate the interferences or the interference points. are subjected in shaft regions exhibiting the most twist, or highest slopes, in the
to further analysis. Since the first alternative is often costly andlor mode shape. These regions are usually at or near nodes. Therefore,
impractical, the authors recommend that the second course be if a system had a region which was an unquestionable weak link,
followed. such as a quill shaft, interference points having a shallow slope in
The further analysis route can sOllnd daunting to the uninitiated this region could probably be safely eliminated.
user since forced, damped analyses are often relatively lime- Another clement that should be accounted for is the torque \'.1"
consuming. If the fact that systems often contain many interference speed characteristic of the driven load. In many turbomachines ,
points, as is the case in Figure 17, is also consider.ed, the analysis such as centrifugal pumps, the transmitted torque is proportional to
alternalive may appear to be totally impractical. Fortunately, many the square of the operating speed. Therefore, if a given mode had
interference poinls can be dismissed from consideration by mere inlerferences at more than one speed. the point at the highest
inspection. In the procedure advocated herein, all interference speed would correspond to the highest transmitted torque. Since
204 PROCEEDINGS OF THE TWENTY-FIFfH TURBOMACHINERY SYMPOSIUM

excitation torques arc often a fixed percentage of transmitted • I X Excitations : one percent of transmitted torque
torgue, this point would most likely represent the worst case from
a stress standpoint. Thus, if all other things were equal, the lower • 2X Excitations: 0.5 percent of transmitted torque
speed points could be dismissed. The authors cannot claim to have veritied the magnitude given
Another factor that should be accounted for is the difference in for 2X excitations. However, the one percent of transmitted torque
energy levels between the various normal modes . In general, the figure is in excellent agreement with the authors' experience with
energy level of a given mode varies inversely with the mode turbomachinery. Furthermore, the general trend exhibited, whereas
number. Thus, the fundamental mode contains the most energy higher order number excitations have lower excitation magnitudes,
followed by the second mode, and so on. As a consequence, the is ill agreement with the literature and the authors' experience.
lower numbered modes tend to be more responsive to excitations Thus, these recommendations appear to be quite reasonable to the
and are more likely to encounter stress problems. According'ly, authors.
with all other variables being equal, if two modes are excited ut the
same torque level and location, the lower mode will experience Gear Excitatiolls
higher stresses. This principle can often be used to eliminate higher
mode interference points from consideration. With regards to gear excitations, there are strong disagreements
Unfortunately, the above principles cannot be written as hard within the literature . Andriola [22], Ehrich [26], and Poritsky [271
and fast rules that can be followed for any system encountered, all claim that modern gears are cut so accurately that their excita-
since there are too many variables that must be taken into account. tions of torsional vibration are essentially negligible. However,
Instead, they are meant to be used as guidelines to be combined there are at least as many references that cite gear excitations as
with the judgment and experience of the analyst. In many cases, being the primary cause of observed vibration problems.
these principles can be used to eliminate most interference points The authors believe that it would be unrealistically optimistic to
from consideration which greatly reduces the amount of rigorous merely disregard gear excitations. Accordingly, the following list
analysis that must be performed. of general trends with which the literature is in agreement may be
The general principles discussed above can be summarized as of some assistance to the user:
follows: • The excitation magnitude decreases as the accuracy of the gear
• Excitation torques located near nodes have very little impact on improves.
the system. • The excitation magnitude usually decreases as the order number
• For a given mode, the WOfst case point is the one whose pulsating increases .
torque location has the highest free vibration amplitude.
• Helical gears provide less excitation than do spur gears.
• For a given mode, the maximum cyclic ,t orques occur at the
locations where the mode shapes have the largest slopes. • Spiral bevel gears provide less excitation than do straight bevel
gears.
• In many machines, the worst case point for a give" mode is the
one that occurs at the highest speed. • Fine pitch gears are usually quieter than coarse pitch gears.
• Interferences with the fundamental and lowti:r numbered modes As was brietly touched upon earlier, there are two different
are usually the most dangerous. forms that gear excitations can take. The first type is a torque exci-
tation which is similar to the excitations produced by other
FORCED VIBRATION ANALYSIS components. The sources that can produce the generic I X and 2X
If performance of the preceding steps leav,e s the analyst with 110 excitations, such as unbalance, misalignment, and ellipticity, are
interference points of concern, the analysis procedure is finished also present in gears. Additionally, the gear contact point can move
and the system can be declared fully satisfacto~y. Although this radially due to tolerances in the tooth profile. All of these sources
optimum situation will occur occasionally, most times there will generate pulsating torques that are applied to the system at the
still be at least one undismissed interference point remaining. In gears.
that case, the next step in the analysis procedure, performance of a Additionally. gears can also produce a very different type of I X
forced vibration analysis, should be taken. excitation known as a displacement excitation. This is also due to
In a forced vibration analysis, a known excitation is applied to errors in the tooth profile, but it is concerned with movement of the
the system and the resulting vibratory torques and stresses are cal- contact point in the tangential direction only. This tangential
culated for each shaft element. Damping due to various sources is deviation from the nominal position is sometimes referred to as
included in the model. The induocd torques and stresses are ,then static transmission error. Although this deviation is very small, it
compared to nllowable values to determIne if the system is struc- call lead to large motions in other portions of the system due to
turally adequate. resOnance effects.
The mechanism by which tangential errors generate excitations
Excitation Magnitudes is variation of the gear velocity ratio. The velocity ratio of a pair of
The first step in performing a forced vibration analysis is deter- mating gears is constant only when motion of the driver produces
mination of the magnitudes of the exc,i tations. The various perfectly smooth rotation of its mate. Such a situation occurs only
turbomachinery components that can excite machines and their if both gears are made perfectly. The imperfections that are present
excitation frequencies were discussed ill a previous section. In in any real pair of gears produce small variations in pitch line
order to determine typical magnitudes for ,t heir fluctu ati ng torques, velocity which imply that the gears undergo accelerations. These
an extensive literature search was underlake!l. Unfortunately, with accelerations generate dynamic forces in the gear mesh which can
the notable exception of the recent work by Wuchel and Szenasi cause the mesh's transmitted load to substantially exceed its steady
[161, the authors found a dearth of information in this area. value .
Therefore, in many cases, the user should obtain test data from the This increase in maximum transmitted load is often referred to
machine of interest or manufacturer's recommendations in order to as dynamic loading and is accounted for in most popular gear
determine accurate values. sizing methods. The variation in force cannot be directly calculat-
Wachel alld Szenasi (161 recommend the following values for ed from the gear errur since it is strongly dependent on the
the generic I X and 2X excitation torques in turbomachincry : dynamics uf the entire system. Therefore, these excitations must he
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 205

input to the model as periodic displacements at the pitch diameter Impeller Excitations
whose amplitudes are equal to the static transmission errors.
For pump impeller excitations, Reference [16] gives the
Although displacement excitations could theoretically occur at
following value:
any multiple of gear speed, the higher order excitations could only
occllr if the static transmission etTOr were distributed symmetrical- 'ex = I / ( I 00 • N) • '( (24)
ly about the gear's circumferenl:e. Since such symmetry is
extremely unlikely, the analysis is run using the maximum error at where:
a frequency of onl:C per revolution for each gear in the mesh. 'ex Excitation torque
As mentioned beforc, displacemcnt excitations behave very dif- Nj Number of blades on impeller
ferently from torque excitations. Specifically, maximum
amplification of the tangential error at the gears occurs when the
'l Transmitted torque
gears are located at a node. In contrast to torque excitations, the Thus, if Equation (24) is utilized, the excitation torque will be
further the gears are located away from a node, the smaller is the significantly less than one percent of transmitted torque.
effect of their displacement excitations. Heretofore. the authors have never been aware of a correlation
Ker Wilson [2] explains this apparently anomalous behavior by between number of impeller blades and excitation torque
noting that a shaft performing undamped torsional vibrations is magnitude, although such a relationship does exist for propeIIers.
usually considered to be fixed at its nodes. He claims that this is Thus, the authors recommend use of the more conservative rule of
equivalent to placing a disk having infinite inertia at all of the ArtiIIes, ct ai. (29], which is to use one percent of transmitted
nodes. Applying a tixed displacement excitation to a node is, thus, torque for all impellers.
equivalent to moving an infinite inertia by that same amount.
Obviously. the effect of this on the system would be considerable. Propeller Excitations
Wachel and Szcnasi (16) provide the following recommended The literature is fiIIed with disagreement on the subject of exci-
magnitudes for gcar torque excitations: tations provided by propellers. In his classic book, Den Hartog [23]
• I X Excitations: one percent of transmitted torque recommends use of a valuc of 7.5 percent of transmitted torque.
Poritsky (27) echoes this value. Yates [30] provides several values
• 2X Excitations: 0.5 percent of transmitted torque of torque, dependent on the number of blades, which span the
The authors' reaction to these values is the same as that range from four to 12 percent of transmitted torque. On the other
expressed above with respect to the generic I X and 2X excitations. hand, Ehrich [26] claims that the excitation torque is only about
Unfortunately, Wachel and Szenasi [16] provide no guidclines for three percent of transmitted torque. Andriola [22] seems to
excitations at tooth meshing frequency. For these excitations, the synopsize the above authors by claiming that the fluctuating torque
authors recommend that a conservative value of one percent of may vary all the way from two to 15 percent of average torque,
transmitted torque be utilized. depending on number of blades and application. As always, the
A 1988 seminar 128J provides an alternate method for determin- best approach is to obtain test data for the unit in question, but, in

- ing pulsating torque magnitude from the gear's error profile. The
first step is to determine. either by test or tolerance study, the radial
variation of the gear contact point during one complete revolution
the absence of such, its probably best to conservatively use the
maximum value of 15 percent.
Electrical Machine Excitations
of the gear. The excitation torque may then be obtained from the
following equation: The excitations produced by electric motors and generators can
be divided into several categories. The first category is normal
running at operating speed. For this condition, Wachel and Szenasi
(23) [16] recommend the foIIowing values for AC motors:

where: Excitations at line frequency: one percent of transmitted

'ex Excitation torque torque
OR Total radial variation in gear contact point Excitations at twice line frequency : 0.5 percent of transmitted
Rp Gear pitch radius torque
'l Torque transmitted through gear mesh
These values are in general agreement with the authors' experi-
ence and are recommended for use with AC generators also.
As stated above, displacement excitations are input to the model As was stated previously, the torque pulsations generated during
by specifying the tangential compone nt of the gear error. steady running of DC motors and generators are often negligible.
Before the discussion of gear excitations is concluded, an inter- Unless the manufacturer says differently, these excitations can
esting footnote should be mentioned. In his book, Ker Wilson [2] normally be ignored.
noted that gears can also generate "phantom" excitations that are The second category is the electrical short circuit. As stated pre-
determined by the characteristics of the gear-cutting machine and viously, when AC motors experience short circuits, they develop
"hereditary" excitations due to the traits of the machine which torque pulsations at line and twice line frequency. When motors
made the gear-cutting machine. The authors are familiar with an experience large pulsating torques, as is the case during short
actual case in which these types of excitations caused severe circuits, the peak torque is often expressed in the peculiar units of
vibration problems in several units of a particular product line. "per unit" (pu). This is a shorthand method of expressing the ratio
Repeated examinations and inspections could not find any defects of peak torque to the motor's rated torque. For instance, if a short
in the gears capable of producing the observed vibration. It wasn't circuit caused a peak torque that was five times the motor's rated
until the gear-cutting machine in the vendor's shop was disassem- torque, its magnitude would bc specified as 5.0 pU. Accordingly.
bled that the culprit was identified as a slightly defective gear in the per PoIIard [31], typical magnitudes of short circuit torques are as
machine. Although there isn't much that can be done in the design follows:
process to prevent this type of occurrence, knowledge of this phe-
nomenon provides the user with an appreciation for the subtlety of Excitations at line frequency: 7.0 pu
torsio nal vibration sources. Excitations at twice line frequency: 3.5 pu
206 PROCEEDINGS OF THE TWENTY-FIFrH TURBOMACHINERY SYMPOSIUM

Wolff and Molnar [32j report that these torques may be even The torque values presented above are provided merely for the
higher, claiming levels of 10.0 pu for line frequency excitations purpose of giving the user a general idea of the magnitudes under
and 5.0 pu for those at twice line frequency. Grgic, et al. [33), claim consideration. As always, the most accurate numbers for any
that the excitatioll torques generated by shorts are signiticanll:y system can be obtained from the motor manufacturer.
reduced if a variable frequellcy drive is in operation when the short
occurs. Unfortullately, they do not specify a numerical value for Dampillg
this attenuation. As is the case with everything else discussed
herein, the best source of accurate information is Ihe electrical Once the excitation torques are determined, the next step is iden-
machine's manufacturer. titication of all sources of damping in the system. There are two
The third category is the large transient torques at line frequency types of damping that can occur in turbomachinery, external and
that occur instantaneously upon powering of AC motors. Wolff and internal. External damping is energy dissipation that is dependent
Molnar [32] claim that the peak torque generated by induction on the absolute velocity of a particular inertia. This kind of
motors can range from 3.0 to 10.0 pU. Pollard [31] estimates that damping is modelled as a dashpot between the cognizant inertia
the corresponding peak torques in synchronous moto rs are about and ground. Examples of external damping include damping at
7.0 to 8.0 pu. impellers and motors.
The fourth category is the starting of AC synchronous motors. On the other hand, internal damping is dependent on the differ-
Since these must be evaluated using a transient analysis, the torque ence in angular velocities between two adjacent inertias. Whereas
vs speed characteristics, for both average and pulsating torques, external damping is associated with a particular disk element,
should be obtained from the manufacturer. Many authors have internal damping is associated with a specitic shaft element.
observed that the maximum pulsating torque during starting can be Internal damping is modelled as a dashpot in parallel with the
as high as the motor's rated torque. torsional spring representing the appropriate shaft element. Shaft
material hysteresis and damping in .c ouplings are modelled as
Variable Frequency Drive Excitatiolls internal dampers.
Both external and internal damping elements are modelled as
The pulsating torques generated by variable frequency AC motor
linear dashpots, as is illustrated in Figure 18.
drives are essentially of constant magnitude in the range between
10 and 100 percenl of rated speed. As stated before, the magnitudes Internal
are much larger between zero and 10 percenl of rated speed.
Approximate magnitudes are provided in Table 2 for the 'various J Damper J

I------~"-----it
harmonics occuning in a six pulse LCI, as taken from Wolff and
Molnar [32].

Table 2. Six Pulse Lei Excitation Torques.

t:(citalion Torque
Excilalio4'l (Per",,", of Average Torque)
Fr~qu~n<.:'Y
(fmI110-IO %S~ )

6X ElectOl:al Frcqucm;y 100 20

K
12X Elcrtrica.l Frequency 40

18X EIC(.;lrical Fn.:qucncy

24X Ekclrit.:al Frequency 10

Murphy [34] claims that for any VFD, the only excitalion that is External
usually significant is that of the lowest order. Accordingly, for a six Dampers
pulse drive, the six times frequency eXcilatioll is, by far, the most
important. The 12, 18, and 24 times frequency excitations are Figure 18. Modelling of Damping.
usually small. He proceeds to provide the following representative
The torques generated by each oppose the angular velocity and
torque magnitudes for the six times frequency excitation:
are given by the following equations:

Pulse Width Modulator: 5-10 percent of full load torque (25)

Voltage Source Inverter: 10-20 percent of full load torque
Current Source Inverter: 20-30 percent of full load torque (26)

In his nomenclature, full load torque is the maximum torque where:

delivered at a given speed. Even if the actual transmitted torque is "exl Torque generated by external damper (in-lbf)
somewhat less than this value, he claims that dIe pulsating torque "jnt Torque generated by internal damper (in-lbf)
magnitudes will still obey the above. relations. C CXl Damping coefficient of external damper
Sheppard [35] gives torque amplitudes for current source (in -lbf-sec/rad)
inverters that are identical (Q those above. On the other hand, Cjl\l Damping coefficient of internal damper
Hudson [36] reports of a CS I that produces excitations of only one (in -lbf- sec/rad)
to two percent of rated torque. Furthermore, the same source that {() Angular velocity (rad/sec)
provided the large values listed above, Murphy [341, agrees that
special design techniques can be utilized to limit the maximum There are two common means for specifying the magnitude of
excitation torques to one or two percent of rated torque. damping provided by a speciticd source. The first and most
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 207

straightforward is through use of the damping coefficient used in damping coefticient is equal to the instantaneous value of the slope
the above equations. When this method is used, the damping is the of the torque vs speed curve at the operating point.
same for all vibration modes. The reason such a characteristic results in damping requires
On the other hand, damping is also commonly specitied via the some elaboration. As stated previously, torsional vibration results
damping ratio which is detined as follows: in instantaneous speed tluctuations about the average machine
speed. If a component has a positive torque vs speed characteristic,
(27) then during the portion of the vibration cycle when the speed is
above the mean value, the resisting torque increases. The addition-
where: al resisting torque attempts to slow the system down and, thus,
r Damping ratio opposes the vibratory motion.
Damping coefficient (in-Ibf-sec/rad) On the other hand, during the time when the speed is below the
Critical damping coeftieient (in -Ibf- sec/rad) steady state value, the resisting torque goes down. This reduction
in opposing torque tends to accelerate the system and, thereby, also
The critical damping coefficient is the lowest value of damping opposes the torsional vibration . Therefore, the torque-speed char-
that will prevent oscillations when the system is initially displaced, acteristic of the load tends to oppose the vibration throughout its
released, and allowed to freely vibrate until it comes to rest. It is, entire cycle and, thus, acts as a source of damping.
thus, analogous to the well-known critical damping coefficient for Of course, this phenomenon is dissipative only if the slope of the
the simple mass-spring system. However, for multiple degree of torque V.f speed characteristic is positive. If the slope were
freedom systems, the critical damping value is not constant, but, negative, the situation would be exactly the opposite of that
rather. is dependent on the vibration mode being considered. The discussed above and the load would act to enhance the vibration.
detining equation is as follows: In that case, the damping would be negative and instability could
result if there weren't other damping sources in the system to
(28) suppress the vibration. Fortunately, most practical impellers and
propellers provide positive damping over their entire operating
where: ranges. The same cannot be said, however, of generators since their
cern = Critical damping coefticient for nth mode torque vs speed curves sometimes exhibit negative slopes at speeds
(in-Ibf-sec/rad) below rated speed .
Oln = Natural frequency for nth mode (rad/sec) The above discussion also applies to drivers such as motors and
ln = Effective inertia for nth mode (in-lbf-sec 2 ) turbines with one major modification. Although the magnitude of
the damping coefficient is still equal to the slope of the torque vs
The effective modal inertia, 1". is a weighted summation of the speed curve, the sign is reversed. That is, in order for the driver to
inertias of the individual disks. The procedure for calculating it is supply positive damping, the slope of its torque-speed curve must
provided by Eshleman [3} . It is, therefore, seen that any damping be negative. This sign change is necessary because the driver is an
element defined by a damping ratio will have a variable damping energy source rather than an energy absorber.
coefficient that is dependent on the vibration mode under consid- When generating the torque vs speed curve for the purpose of
eration. calculating a damping coefficient, it is essential that all other
Before the various sources of torsional vibration damping are relevant parameters be held constant. This means that the tluid
discussed, it should be pointed out that coulomb friction, which is properties and tlowrate must be held constant for an impeller, the
independent of angular velocity, does not provide damping of vessel speed must be constant for a propeller, and the voltage must
torsional vibration. At first, this seems counter to intuition, since be held fixed for an electrical device.
one tends to picture a simple torsional spring-mass systcm where Failure to follow the above rule can lead to extremely mislead-
the system is twisted and then released. Such a system would ing results. For instance, turbines are often used as drivers in
perform free vibrations indefinitely in the absence of friction . If a turbomachinery. In such cases, the steady state torque vs speed
coulomb friction source were added to the system, energy would curve for the turbine usually has a positive slope, retlecting the
be dissipated as heat and the system would eventually come to rest. higher power generated at high speeds. The reason the turbine
In this case, damping has unquestionably taken place. produces more torque at higher speed is that the tlow rate increases
However, the above model is an oversimplification of a with speed. Thus, if one were to merely take the slope of the steady
turbomachinery drive, since the latter is running at some average stale curve, the erroneous conclusion that the turbine is a source of
speed when torsional vibration is initiated. Since the machine is negative damping would be arrived at. Thus, it is imperative that
already rolating prior to excitation, the driver must be providing the flow rate be held constant during generation of the curve so that
sufticient torque to overcome any coulomb friction that exists in the true positive damping characteristic will emerge.
the system. When torsional vibrations commence, resulting in As is the case with generators, some motors are sources of
variations in angular velocity, the coulomb friction's magnitude negative damping. This is especially common with AC induction
remains unchanged since it is independent of velocity. The drive and synchronous motors that often have a positive torque vs speed
torque and coulomb friction torque remain perfectly in balance, characteristic over a large portion of the speed range below syn-
regardless of the amplitude of vibration. Unlike the above chronous speed.
simplified case, the coulomb friction continuously acts in one Although the torque-speed curve method just described is the
mosl accurate method for obtaining damping coefficients for load
direction. It, therefore, theoretically does absolutely nothing to
and driver elements, there are other approximate equations
resist the vibration and is, thus. not a damping element.
provided in the literature. For instance, Nestorides [8) provides a
detailed method for calculating the damping factor for propellers
Load Dampillg
as a function of their dimensions. Additionally, many authors
One of the primary sources of damping in turbomachinery approximate the torque-speed relation for a propeller by the
systems is the torque vs speed characteristic of the load. This following:
applies regardless of whether the load is an impeller, electric
generator. or propeller. 1f the load is such that its resisting torque (29)
increases as speed is increased. with all other parameters held where:
cunstant, then it is a source of damping. Furthermore , the viscous '(prop = Propeller torque (in-lbO
208 PROCEEDINGS OF THE TWENTY-FlFrH TURBOMACHINERY SYMPOSIUM

k = Constant Fluid absolute viscosity (lbf-seclin 2 )

Wprop = Propeller angular velocity (rad/sec) Disk angular velocity (rad/sec)
n = Propeller exponent Disk outside radius (in)
Disk inside radius (in)
Most authors give the value of n as being between two and three . Radial clearance or axial clearance (assumed equal) (in)
By differentiation with respect to w prop ' the damping coefticient is Disk length (in)
then given by the following:
Differentiation of the above with respect to angular velocity
(30) yields the following equation for damping coefficient:

where: C v = 11: • ~l • (RoJ • (Ro + 2L) - Ri4) Ih (33)

c prop = Propeller damping coefficient (in-Ibf-sec/rad)
where:
For approximation purposes, Ker Wilson [2] claims that most c" Damping coefficient (in -Ibf-sec/rad)
propellers yield damping coefficients that are between five and 10
percent of critical damping. loumal Bearing Damping
The damping from turbomachinery impellers is usually much Another damping source is the tluid-film journal bearing. The
smaller than that arising in propellers. Yates [30] claims that viscous friction that occurs in these bearings is speed-dependent
turbines provide damping that is between 0.15 and 0.50 percent of and is, therefore, theoretically a provider of damping. However,
the critical value. Smalley [37) claims a typical value for compres- many authors note that the damping due to viscous friction in
sor rotors is 0.2 percent of critical. In many practical impellers, the bearings is virtually negligible. The interested reader can demon-
torque is approximately proportional to the square of the speed. strate this by merely differentiating the classical journal bearing
Therefore, Equations (29) and (30) can be used with a value of two friction equation with respect to angular speed and calculating the
for n to estimate the damping coefficient if at least one torque resulting damping coefficient.
speed point is known. However, journal bearings can be significant, even dominant,
Nestorides [8) provides the following equation for estimating sources of damping if any lateral motion accompanies the torsional
the damping coefficient of a DC generator: vibration. Lateral motion can arise if there is a coupling between
torsional and lateral modes. Such coupling is very common in
c gen = 600,000· V I (V - Vc ) • hp I RPM2 (31) geared systems due to the radial load at the gear mesh. When the
journal moves laterally, it forces the fluid in the radial clearance to
where: flow circumferentially, generating a squeeze film effect. Simmons
DC generator damping coefficient (in -Ibf-sec/rad) and Smalley [6). Draminsky [38], and Shannon [39] all report of
~en
Total generator voltage (terminal voltage plus internal cases where this squeeze film action was the preeminent source of
drop) (Volts) damping in the system. Furthermore, Smalley [37] claims that this
Counter EMF of driven machinery (motor or storage mechanism can yield damping ratios as high as 0.10.
battery but not an external Although this mechanism can be highly significant, especially in
resistance) (Volts) geared systems, its magnitude is very difficult to calculate. Per
hp Generator power (hp) Smalley [37], in order to determine the damping coefficient, a
RPM= Shaft speed (rpm) combined torsional-lateral model of the gear mesh needs to be
created and executed. Since such an undertaking is often compli-
Ker Wilson [2] states that the above equation is also applicable cated and time consuming, it does not always represent a practical
to DC motors. alternative . Accordingly, the authors recommend that journal
Both Ker Wilson [2] and Nestorides (8]' state that the damp,i ng in bearing damping be completely ignored in ungeared systems. In
standard AC motors and generators is negligtble. h is common, geared systems, the beneficial aspects of squeeze film action can
therefore, to equip these machines with damper w,i ndings. When be accounted for by judiciously increasing the system's modal
this is done, the damping coefficient can be app~oximatedt tlsing damping ratio, to be discussed later on.
Equation (31). Ker Wilson [2] claims thai generators with damper
windings can achieve damping as high as 25 percent of the critical Shaft Hysteretic Damping
value.
All of the damping sources discussed up to now qualify as
external damping phenomena. Accordingly, their damping is asso-
Windage Damping ciated with a specific disk element. The remaining damping
Another source of external damping is the windage frictional sources to be described are internal damping mechanisms which
loss that occurs due to viscous t1uid drag on immersed impellers. are linked to shaft elements.
Windage friction generates damping, unlike coulomb friction, Material hysteresis is the dissipation of energy that occurs when
since the frictional force generally increases with speed. This a metal is subjected to a complete cycle of stress changes within its
damping is usually negligible for impellers surrounded by air such elastic region. Energy is dissipated because all real metals slightly
as compressor and turbine rotors. However, pump impellers deviate from idealized Hooke's law behavior due to plastic defor-
immersed in a highly viscous liquid can generate damping torques mations at the microscopic level. These cause the material's
that are not insignificant. Ker Wilson [2] provides the following behavior during unloading to be slightly different than it was
equation for the torque due to fridion acting on all surfaces of a during loading. An exaggerated depiction is shown in Figure 19 of
hollow disk: the stress-strain behavior of a real material taken through a
complete stress cycle. It is seen from the figure that the strain
slightly lags behind the stress such that there is a finite residual
strain in the material when all stress is removed.
where: The implication of this is that when a shaft is twisted in either
TV f"rictional torque (in -lbO direction and then released . it does not return to its initial position
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 209

Stress A Matcrial constant

() Amplitude of cyclic shear stress (psi)
m Hysteresis exponent

Accordingly, the cyclic energy loss in a hollow circular shaft that

obeys the relation of Equation (34) is as follows:

Initial
Loading

Energy loss per cycle (in -lbf/cycle)

Strain Outside radius (in)
Inside radius (in)
Length (in)

At stresses below the quasicritical limit, most authors agree that

the hysteresis exponent is between two and three. An exponent
espoused by several authors for steel is 2.3. Above the quasicritical
stress, the exponent increases dramatically and can reach values as
high as 10.5, according to Dorey [40]. Dorey [40] also found
experimentally that the quasi-critical stress for steel is between
Figure 19. Material Hysteresis Loop. 12,000 and 30,000 psi. Several sources note that this limit can be
approximated by dividing the material's ultimate tensile strength
of zero detlection. In order to bring the shaft back to the unstrained by five.
position, a finite amount of work must be added to the system. This Unfortunately, Equation (35) does not lend itself easily to forced
work is necded to make up for the hysteresis energy dissipated vibration analysis. To facilitate calculation, shaft hysteresis is
during twisting and releasing of the shaft. This energy loss is due usually modeled as a viscous internal damping element. The
to internal friction within the material and is released in the form assumption of viscous damping implies a value of 2.0 for the hys-
of heat. If a material is taken through a complete stress cycle, the teresis exponent so it is relatively accurate for shafts at low stress
total energy dissipation is equal to the area enclosed in the stress- levels. However, the error introduced is considerable for highly-
strain loop. stressed shafts .
There has been a plethora of technical articles written on the Thc equivalent viscous damping coefficient can be calculated
subject of material hysteresis. Some general characteristics of this for a given mode of vibration by using the following:
phenomenon are as follows:
(36)
• The energy lost due to hysteresis during cycling is a very small
fraction of the elastic energy at the maximum stress point. where:
• Energy loss is essentially independent of frequency of cycling. c Damping coefficient (in-Ibf- sec/rad)
Dcq Specific energy loss (in -lbf/cycle/i n3)
• The energy dissipated per cycle is strongly dependent on the VOL = Volume of material in shaft (in 3)
cyclic stress level in the material. Dissipation and damping Natural frequency (Rad/sec)
())n
increase as stress increases. Angle of twist in shaft (rad)
8,h
• Above a certain stress level. known as the quasicritical stress, the
relationship between applied stress and hysteresis loss chamges Unfortunately, the damping coefficient is seen to be dependent
dramatically. Above this point, relatively small stress level on the amplitude of vibration which is usually the unknown
increases result in large increases in energy dissipation. quantity. An iterative solution is needed, therefore, if Equation (36)
is to be utilized. This is far from a trivial task and is not the method
• Hysteresis, as specified by the energy lost per cycle per unit preferred by the authors.
volume, is a material property. It is totally independent of part size Instead, like the majority of sources consulted, the authors
and form. recommend use of a representative damping ratio which is applied
• The hysteresis exhibited by a given material at a specified stress to all shaft elements in the model. The literature is filled with
level is a function of the material's stress history. Parts that have various recommendations for the magnitude of this ratio. Ehrich
previously been taken above the quasicritical stress level will [26] specifies a value of 0.16 percent of critical for typical rotor
exhibit significantly larger damping at the same stress level than steels. Draminsky [38] essentially agrees, citing a value of 0.2
those that have never crossed this threshold . percent. Ker Wilson [2] claims typical factors of 0.2 to 0.4 percent
of critical. Rieger [41] claims that the damping is closer to 0.5
• In general. configurations where most of the material is at a high
percent. Finally, a 1988 seminar [28] recommended use of 0.5 to
stress level provide higher levels of damping. Thus, for a given
1.0 percent.
torsional stress level. a hollow shaft provides more damping than a
No matter which of the above values is selected, it is readily
solid shaft having the same volume since a greater portion of the
apparent that the hysteretic damping is relatively small. Because of
material is in the vicinity of the highly-stressed outer surface.
this, the authors recommend lumping the hysteresis damping with
In general, the relationship bctween energy dissipation and several other indeterminate damping sources to be discussed
cyclic stress level can be expressed by the following equation: shortly in one overall empirical damping ratio.
Another source of internal damping is couplings containing
--- (34) rubber elements. The mechanism by which these dissipate energy
where: is exactly the same as that just discussed for steel shafts. The major
D Spccific cnergy loss (in - lbf/cycle/in') difference is that rubber gencrates an order of magnitude more hys-
2)0 PROCEEDINGS OF THE TWENTY-FIFrH TURBOMACHINERY SYMPOSIUM

teresis than do metals. Rubber couplings can generate damping as which are brought into relative motion by the vibration. Locations
high as 25 to 30 percent of critical 1281. Of course. the best source where this l'OIlIlllonly occurs include bolted tlanged joints, splines,
for accurate damping data is the coupling manufacturer. keyed connectiolls, and shrink tits. It should be noted that this
mcchanism is very different from the coulomb fril'tion described
Hydraulic Couplillg Dumping previous!y sillee there is no relative motion or friction prior to the
Another type of coupling that can provide substantial damping onset of vibration.
is the hydraulic coupling that was described earlier. As was pointed Unfortunately, there are no practical methods for calculating the
out. this coupling has the benclicial aspect of effectively dividing a magnitude of slip damping. However, it appears that it plays a
torsional system into two independent systems. This permits the major role in many real systems. Many authors have observed that
isolation of vulnerable components from signiticant sources of the damping occurring in practice is almost always signiticantly
vibration such as synchronous motors. greater than the calculated value. Many attribute this discrepancy
In addition to its isolation characteristic. the tluid coupling also to the presellce of slip damping.
provides damping to the machine by virtue of viscous friction Because of the difficulty in cakulating hysteresis and slip
between the impellers and coup!ing tluitl. Per Ker Wilson [2], damping, many experts advocate lumping the two together in an
although this damping l'an be significant, it is usually considerably empirical damping ratio that is applied to every shaft element in the
less than that of a marine propeller at the same power level. He model. This is the method recommended by the authors . All other
proceeds to then provide the following equation for a coupling sources of damping sueh as impellers, motors, and journal bearings
operating with !ube oil at three percent slip: should be simultaneously included in the model.
There is a considerable dift'erenee of opinion in the literature
c = 600,000· (I + 1/ n 2 ). hp / RPM 2 (37) regarding the percentage of critical damping that should be
assumed. A representative sample of recommended values is
where: provided in Table 3. Unfortunately, the tabulated numbers are
c Damping coefticient (in-Ibf-sec/rad) somewhat nebulous, since the majority of the sources neglect to
n Order number of mode being excited specify whether a geared or ungeared configuration is under con-
hp Average power transmitted through coupling (hp) sideration. The authors have found through experience that
RPM= Speed of driving shaft (rpm) assuming damping of one percent of critical is conservative for
ungeared systems.
This equation is also independently given by Nestorides [8J. It
should be noted that the above equation should only be used for
approximate ballpark numbers, since damping is a fu nction of fluid Tahle 3. Recommended Dalneillg Ratios [42,43, 44,45,46].
type, operating temperature. and slip that are not accounted for in
this equation. For accurate damping values, the coupling manufac- SoUTce Danlping
turer should be consulted. R3Iio (r)

Speed Varilllioll Damping Enrich )26) 0.004

A source of damping that is not provided by a specitic Grgic. et 31. [J:t) 0.005
component is apparent damping due to speed variations. This Fro;' ct al. [42) 0.01
source is not an energy dissipation mechanism per se; rather, it rep-
MI.:Cann and Hennelt [121 0.005 - (1.020
resents the fact that speed variation can occur in real machines,
resulting in imperfect resonance. If the nlllning speed is assumed Thames and Heard 143) D.OOS] - 0.025
to vary in a periodil' manner. Ne storides [8] gives the following E van.Ii, ct al. {441 0 .0167
formula for the effective increase of al'l damping coefficients in the
Vance II) 0.01:; - 0.020
system:
\Vadld and SZl:!n:Jsi 1161 0 .0167 - 0.025
c = Co / cos(B • R / n:) (38) Kcr \Vilson 121 0.025

where: Yat<> 130) 0.025

c Apparent damping coefficient (in -Ibf- sec/rad) Cllenl45) 0 .01 - 0.05

Co Damping coefticient cakulated in normal manner Re ferenco 128) 0.01 -0.05
R Number of torsional vibration cycles that occ ur in one
Artilles . et ill. r2~) 0.01 - 0.05
speed variation cyclc
Simrnons and SmallL:y 161 ().(II -0.06
The coeflicient B is obtained from the following: Chen, et al. )46) o.m - 0 .05

B = (RPM""" - RPM min ) / RPM mean (39)

where:
RPM max Maximum machine speed (rpm) There is little doubt that the damping in geared systems is
RPM min Minimum machine speed (rpm) generally higher than for ungeared ones due to the squeeze film
RPM mcan Average machine speed (rpm) effect in fluid-fi!m journal bearings. Ker Wilson [21 goes so far as
to claim that the damping is doubled. Wright 147] is more cautious.
In most practical systems. speed variation damping is small and claiming that the damping ratio increases from 0.0125 to 0.020 .
can often be neglected, UnfOJ1unately, there are few other recommendations regarding the
amount by which damping should be increased when dealing with
Slie Dall//,illi;
a geared system. It seems safe . however, to moderatcly incrcase the
Another important source of damping in torsional systems is slip ernpiril'al factor used for ungeared assemblies if there are tluid-
damping. Slip damping arises from friction between twu surfnces Ii 1m heari ngs present.
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 211

Steudy Stu/l' Analysis (41 )

Once the excitation torques and damping coefficients are
determined, the ste;tdy state forced damped analysis can be
Equivalent damping coefficient referenced to low speed
performed. In general, a separate analysis should he performed for
shaft
each interference point remaining under consideration . As was the
c Actual damping coefticient
case with the undamped analysis, the mechanics or setting up and
N Gear ratio (N > 1.0l
solving the relevant equations are not of prime importance since
there arc many good computer codes available for doing this. Once However, when equivalent systems are used, the displacements
again, the primary emphasis is placed on generation of the correct and torques occurring in the equivalent elements have no physical
input values and proper utilization of the analysis results. meaning. In order to determine the displacements and torques
Forced torsional vibration occurs when a sinusoidal excitation is occurring in the actual physical shafts. the procedure for generat-
applied to a machine operating at a constant speed, N. The system ing the equivalent system must he reversed. That is, the equivalent
response is obtained by solving thc differential equations of displacements and torques at the lowest speed shaft must he
motion . In general, the solution is known to consist of two compo- converted through all of the appropriate gear meshes until the
nents, the complimentary solution and the particular solution. The desired shaft is reached. The conversion equations for any gear
complimentary solution is a transient free vibration at the natural mesh arc as follows:
frequency that decays to zero exponentially with time. Yates [301
asserts that the complimentary response typically takes about 20 8 = 8.:q • N (42)
vibration cycles to die out. Conversely, the particular so lution is a
steady state term that continues for as long as the excitation is (43)
applied. In steady state analysis. the particular solution is the only
term of interest. where:
For any system, the pal1icular solution consists of a continuous 8 Angular displacement at high speed shaft
oscillation at the excitation frequency, Cll. This oscillation is super- 8 eq Equivalent angular displacement at low speed shaft
N Gear ratio (N > 1.0)
imposed on the steady running speed, N. The net result is that the
disks no longer rotate at a constant speed. Instead. their angular , Torque at high speed shaft
velocity varies sinusoidally between the limits of N -wand N +w, 'cq Equivalent torque at low speed shaft
assuming Nand (t.l are in consistent units.
Unlike the undamped case, the disks do not vibrate in phase with NUlllinear Couplings
each other during forced vibration. The system's damping intro- As was the case in the undamped analysis, nonlinear couplings
duces phase shifts to all disks such that their individual motions are demand special attention during the performance of a forced
given by the following equation: vibration analysis. Tuplin 1481 advocates use of an equivalent
linear spring that would contain the same amount of strain energy
8 = 80 • sin (wt - $) (40) as the actual coupling when subjected to a specified angie of twist.
where: Since the strain energy accumulated by a nonlinear coupling is
8 Angular position of a given disk as a function of time equal to the area under the torque vs angle of twist curve (Figure 6)
80 Amplitude of angular vibration up to the specitied twist angle, 81' the equivalent stiffness can be
w Angular frequency (rad/sec) calculated from the following:
t Time (sec)
<!J Phase angle between displacement and excitation (rad) (44)

The lumped model used for the damped analysis is essentially where:
the same as that utilized for the undamped calculation. The only keq Spring rate of equivalent linear element (in -Ibflrad)
difference is that internal and external damping elements must be 8, Angle of twist in coupling (rad)
added in the appropriate locations. In .t orsional systems, as in other A Total area under, \'S 8 curve up to 8,
vibrating systems, damping has n significant impact on the
response only at frequencies near resonance. However, since the It is seen that the equivalent stiffness is a function of the angle
only forced vibration analyses that will be discussed here in are of twist whi ch is an unknown. Accordingly, as was done in the
concerned with the resonant condition, damping shall always be undamped analysis, an iterative procedure must be implemented .
included. The steps to be taken are as follows:
In the analysis to be discussed, torque and/or displacement exci- • Assume an angle of twist, 8" in the coupling.
tations are applied to the system at appropriate locations. Since
they are applied at a resonant frequency, the response would be • Calculate the equivalent spring rate using Equation (44).
infinite if there were no damping in the system . However. since • Using this stiffness, perform the forced vibration analysis and
damping is present, the response builds up until the total energy determine the actual twist in the coupling.
input to the system by the various excitation sources equals the • If the actual twist in the coupling matches the assllmed value, a
total energy dissipated by the dampers. At this point, the system solution has been found. If not, go back to step I and try another
begins steady state vibration . guess value.
Geared Systems Since each interference point will most likely induce a different
Geared systems are handled in the same manner as in the amollnt of twist in the coupling, the above procedure should be
undamped analysis. All inertias and spring rates are converted to repeated for each point analy zed.
equivalent values at the lowest speed shaft via Equations (4) and
EleClriclIl Macizilles
(5). Both internal and external damping clemcnts located on the
high speed shaft are converted to equivalent values in the same The reader may lind thc various comhinations of electrical
manner, as illustrated in the following: machinc types and drives to be sOlllcwhat conrusinf',. To alleviate
212 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

any confusion, the various cases that should be investigated for Excitation torque (in - Ibn
different motor-drive combinations are listed in Table 4. The cases Angular displacement at excitation location (rad)
tabulated as steady state can be analyzed using the procedures to be
discussed next. The transient cases will be addressed in a later Likewise, the energies dissipated per cycle in external and
section . It should be emphasized that only those cases whose exci- internal dampers, respectively, are given by the following:
tation line generates an interference point need to be analyzed.
Thus, for many systems, many of the listed cases will not require
(46)
analysis.

(47)
Table 4. Relevallt Cases for Variolls Electrical Machines.
where :
Muchinc Drive Case Excilalion Steady Stale
Type Type Freqllcncil.!~ ()( Ed Energy dissipated per cycle by damper (in -Ibn
Tr.l(l~icnl
(On Natural frequency (rad/sec)
AC I. Steady Running Line FlttJUCncy Sleady Slate c Damping coefticient (in -Ibf -sec/rad)
Generator 2. Steady Running 2 • 'Line F""Iucncy Sleady Stale
with Np 3. Sicady Running Nr~ RI'M Sleady Slale 8 Angular displacement at damper location (rad)
Poks 4. Shon C;",uil Line frt"qt~llcy Tmnsicn(
5. Soon Circuit 2 , line Fr<qucncy Trolflsicnl
In the above equations, the unknown parameters are the angular
AC MUlor Fixed I. Steady Runn ins lLine Frt-quency Steady Slate
with Np Frequency 2. Steady Running 2 x lLinc Frcquenl:Y Steady SIal<'
displacements. However, the ratios of the various displacements
Poles 3. Steady Running N p" RPM Steady Stale are tixed by the mode shape. Combining these ratios with the
4. Shon Circuit Une Frequency Tr,m~icnt
5. Shon Circuit 2 x line Frequency Trun!,;(CIII
above equations, the steady state amplitudes at each disk can be
6. Inilial Stan Line Frequency TrclflSien( solved for.
7. Runup 10 Spe ...xt 2 x Slip Frequency Transienl
(synchronous only)
"12 Moto< Vllliabic I. Steady Running IJ ~ x Np x RPM Steady Stuh!
Complex Holzer Method
with Np FR,'qlu~ncy 2. Steady Running N. x RPM St""dy Stal"
Poles' wilhN 3 Sr~ad'y Running 1/2:\ N x Np :\ RPM Sleady Siale
The preceding method is reasonably accurate for most torsional
PUhiCl" 4. Sleady Running N. Np x RPM Sleady Stale systems, since they are usually lightly damped. However, it is not
(e.",,1'1 5. Sleady Running I.S x N x Np x RPM Steady Slale
Slalic 6. Steady Running Steady Stale appropriate for systems that contain a large damping element such
2 x N x No x RPM
Kmtncr) 7. Short Circuil
1 /1 :~ Np x RPM Tr<1nsienl as a hydraulic coupling. To permit analysis of such systems, Den
R Short Circuit Tf'J.n.sicnl
9. Inilial Stan Np x RPM
Trans.ienl
Hartog and Li 149] propose a method that utilizes a variation of the
Line ,Frequency
(induclion onlyl Holzer table. Although their method is more complicated than the
AC Molor Sialic I. Sleady Running Line hCqucfiC)' Steady State energy balance method, it does not suffer from any of its limiting
withNp Il (r.amcf 2 Steady Running 2 x Li"" I'rcqucncy Steady Stale
Poles Drive with J. Steady Runnin g N•• RPM Steady Slale
assumptions.
N Pulses 4. Steady Runn ing N • Slip Frequency I Steady Stale In the Den HartogiLi procedure [49], a Holzer table containing
IS Sready Runnin g 2N • Slip Frequen"y Sleady Stale
6. Short Circuit Line FreqUL"OC)' Transient
complex quantities is generated. The real part of each quantity rep-
7. Short Circ.:uit 2 x Lioo frequency Tran~icnl resents the portion that is in phase with the motion of the shaft's
8. Initial Start Line Frequency T m.n sienl
first disk while the imaginary part is 90 degrees out of phase with
that motion. As with Holzer's method of finding undamped natural
frequencies, this method requires generation of a table for each
guessed value of the vibration frequency. The remainder torque is
Analysis Methods determined for the last disk for various frequencies until the
There are many different methods that can be employed to solve frequency which yields the minimum remainder torque is found.
steady state forced torsional vibration problems. A few of them This frequency is the damped natural frequency which for most
will be briefly discussed to give the user a flavor for the type of torsional systems is not much different from the undamped value.
techniques commonly utilized. However, as is the case with The damped natural frequency is found by setting the amplitude
undamped analysis, the authors do not believe that the mechanics at the first disk arbitrarily to one radian. The amplitudes at the
of the solution procedures are of primary importance, since there various disks in the table for the damped natural frequency
are many computer codes available to the user. Accordingly, represent the mode shape of damped vibration. After determination
descriptions of the various methods are kept brief. of the natural frequency, this Holzer table is manipulated further
until the appropriate disk torques are equal to the excitation
Energy Balance Method torques. The resulting displacements are then the actual forced
vibration amplitudes of the system.
Perhaps the simplest method for steady state resonant forced
vibration analysis is that which Ehrich 126] refers to as the energy
Modal Analysis
balance method. The basis for this procedure is that, at steady state,
the amount of energy input to the system by the ex.·c itation tmques Modal analysis is a relatively simple, approximate method for
during a vibration cycle must be equal to that diss,ipated by the detemlining the induced torques in a system undergoing forced
damping elements. The key assumption utilized in this method is vibration. Anilles, et al. [29], provide an equation for the angular
that the damping level is so light that it has no impact on the displacement at a given disk as a function of the excitation torques
system's natural frequency and mode shape. The assembly is, applied at other disks. The equation consisls of a weighted
therefore, assumed to execute forced vibration at its undamped summation of all of the mode shapes of the system, which is tech-
natural frequency with its undamped mode shape. Implementing nically an infinite number. However, Artilles, et al. [29], claim that
this assumption, the energy input by each excitation is: relatively accurate results can be obtained using only the first four
or five mode shapes.
(45) The modal nnalysis method is not exnct because it assumes that
thc damping in the system is small. However, this is a good
whcre: assumption for the majority of torsional systems encountered in
Ei Energy input per cycle by excitation source (in -Ibn practice. Additionally, this procedure suffers from the limitation
 PRACTICAL DES IGN AGAINST TORSIONAL VIBRATION 213

that the only damping input is a modal damping ratio which is R Shaft radius (in)
applied to all shaft elements. Thus. individual damping sources Ip Area polar moment of inertia (in 4 )
located at specific positions cannot be incorporated in the model.
However. since many torsional analyses are conducted with an Analysis Methodologv

- cmpirical damping ratio as the only damping input, this is not as

much of a restriction as it may first appear to be. Vance [I J.
Eshleman [31. and Poritsky 1271 all provide more detailed descrip-
Once the analysis procedure is selected, it should be used to
evaluate each remaining interference point. one at a time. Since
each interference point is driven by a specific excitation
tions of this analysis procedure. source, there should only be one excitation torque applied to
the model. The location of this torque is tixed if it is due to a
Matrix-EigclIl'allle Mcthods
specific component in the system, such as an impeller or motor.
Matrix-eigenvalue methods . which were previously described However, if the excitation is a generic one such as the I X and 2 X
for the natural frequency analysis, can also be used for forced drivers resulting from unbalance or misalignment, it could be
vibration calculations. The matrix equation for forced vibration is located at literally any disk in the system. In this case, the authors
an extension of Equation (10) and is as follows: advocate application of the excitation at the disk which exhibits the
maximum displacement in the appropriate mode shape. This
Ik]· \81 + [cl • [wi + [11 ·Ia) = It] (48) represents the worst case from an induced torque and stress
standpoint and is. therefore, conservative.
where: Additionally, if any of the interference points arise from dis-
[k] Stiffness matrix placement excitation at a gear mesh, that case should be analyzed
[cl Damping matrix separately. In order to accomplish this. the solution algorithm
[1) Inertia matrix being used must be tlexible enough to accept a displacement input.
[t] Excitation torque vector Badgley and Laskin [50J discuss how to achieve this and provide a
[8J Angular displacement vector sample computer program.
IwI Angular velocity vector The authors have saved time on many occasions by executing an
[al Angular acceleration vector initial run with the damping due to specific sources ignored. The
only damping included is the generic damping ratio of 0.0 I applied
The stiffness and inertia matrices are the same as those used in to each shaft element in the model. As discussed previously, this
the undamped analysi s. The damping matrix is a combination of accounts for hysteretic and slip damping effects and represents the
the damping coefficients of all of the external and internal dashpots minimum amount of damping that can reasonably be expected. The
in thc system. Vance [I J provides specifics for generating each of results obtained from this run are, thus, conservative.
these matrices . If the results from this initial run are deemed acceptable
As in the undamped analysis, matrix methods are used to according to the criteria to be discussed later, then the analysis for
determine the eigenvalues and eigenvectors. Due to the inclusion that particular interference point is complete. The analyst has saved
of damping , these arc all complex . The damped natural frequencies the often considerable time and effort required to determine the
are equal to the square roots of the imaginary portions of the eigen- damping coefficients due to the various individual sources. If the
values. The displacements of the individual disks are obtained results are not acceptable, the analyst then must calculate the
from the eigenvectors. appropriate coefficients and rerun with them applied.
Even if the initial run is not successful in validating the design,
Shaji Torqlles alld Stresses the analyst need not calculate every single damping coeft1cient for
the more rigorous run. As is the case with excitation torques,
All of the forced vibration analysis procedures are designed to
damping elements only have a significant impact on the system if
obtain the amplitude and phase angle of each disk in the system.
they are located in regions of significant vibratory activity, as
Once these are determined, the cyclic torque in each shaft is easily
de tined by the cognizant mode shape. This refers to the amplitude
obtained from the following :
at the appropriate disk for an external damper and the angle of twist
in the accompanying shaft for an internal dashpot. Thus, when
running the analysis, the user can ignore damping elements located
in dormant areas of the system.
where:
tl2 Cyclic torque occurring in shaft between disks I and 2 Analysis Results
(in - Ibf)
Spring rate of shaft between disks I and 2 (in - Ibflrad) The results of a steady state analysis consist of the angle of twist ,
Angular displacement of disk I (rad) cyclic torque, and cyclic stress occurring in every shaft element in
Angular displacement of disk 2 (rad) the model. Additionally, the torques and detlections for each disk
Damping coefticient for shaft between disks I and 2 clement arc also often presented. These results should be utilized
(in - Ibf-sec/rad) to determine the adequacy of the system, based on three general
Angular velocity of disk I (rad/sec) criteria.
Angular velocity of disk 2 (rad/sec) Shaft Fatigll e Stress
Once the cyclic torque acting on a particular shaft is known, the The first criterion is the structural integrity of each shaft
resulting maximum shear stress can be calculated using the element. Each shaft is subjected to a combination of mean stress
following equation from strength of materials: due to the average transmitted torque and cyclic stress due to the
torsional vibration. The cyclic stress obtained from the vibration
(50) analysis is a nominal value, calculated using Equation (50). based
on the macroscopic properties of the shaft cross-section.
where : The nominal cyclic stress should be increased by appropriate
S, ivlaximum shear stress in shaft (psi) stress concentration factors to account for the presence of keyways,
t Torque actin g Oil shaft (in - Ibn splines . fillet rad ii, etc. The corrected cycli c stress should then be
214 PROCEEDINGS Of THE TWENTY-FIFrH TURBOMACHINERY SYMPOSIUM

plolleu along with the mcan stress on a Goodman uiagram for the recommendation. which has an inbuilt safety factor of 2.0. should
shaft material. Thc equivalent fully-reversing stress can then be only be used for ballpark estimates.
obtained hy drawing a straight line from the pI oiled point to the
material's ultimate strength in shear. As is illustrated in Figure 20. Peak Tort/lit'
extension of this line to the y-axis yields the equivalent stress The sccond criterion to be considered is the peak torque
value. occlilring in components such as gears and couplings. The peak
torque is simply equal to the sum of the mean transmitted torque
anu the induced cyclic torque. Most gears and couplings are rated
for the maximum continuous torque that they may carry. Although
Equivalent the peak torque in vibration occurs only once in a cycle, it is often
Fully-Reversing taken by the same tooth on at least one of the gears. Such a tooth
Stress is. thereby. continuously exposed to this torque. Thus, the authors
/ recommend that the peak torque be kept below the continuous
rating for these components.

Torque Rel't'r.l'lIls
The third anu final criterion concerns components that operate
with backlash such as gears, splines, and geared couplings. Under
normal circllmstances, such components are continuously loaded
in one direction and the backlash is a tOtal nonfactor. However, if
the magnitude of the induced vibratory torque at one of these
locations exceeds the average transmitted torque, the applieu
torquc becomes instantaneously negative. This is an indication of a
Ultimate rcversal in direction of the applicd torque <Inti is known as a torque
Shear reversal.
Stress When a torque reversal occnrs, the drive surfaces in the gear or
Mean Stress spline separate. The negative applied torque causes the teeth to
move through their backlash until they make contact on their
Figure 20. Typical GOOdlllllll Diagram.
reverse sides. Once the torque changes sign again. the teeth are
driven back through the backlash until they resume contact on their
Alternately, the equivalent fully-reversing stress can also be cal- original surfaces.
culated from the following formula: This phenomenon causes the teeth to continuously pound
against each other. This causes shock loading on the teeth which
(51 ) results in loads much greater than the normal design loau. This
often eventually leads to pitting and wear of the tooth surfaces
where: which is frequently followed by tooth fatigue fracture. Torque
Seq Equivalent fully-reversing s[ress reversals, therefore, can have catastrophic consequences.
SUit Material ultimate strellgth in shear There arc many references that advocate the avoidance of torque
Slllcan Mean stress reversals at backlash interfaces under all conditions. Although this
Scyclie Cyclic stress is a worthy and desirable goal. the authors feci it is overconserva-
tive. There is no question that torque reversals cannot be tolerated
This equivalent stress should then be compared with the when the system is operating at anywhere close to its maximum
material's endurance limit. corrected t'or all of the appropriate torque. However, at very low transmitted torque levels, thc motion
factors such as size. tempcrature. and surface finish. The endurance through the backlash is less likely to cause problems.
limit must be used sinee resonance at a normal operating speed can Torque reversals generate problems because the impact loading
induce a virtually intinite number of stress cycles . The endurance they produce yields gear loadings that are higher than those
limit and equivalent stress shoulu then be combined to determine designed for. Determination of the amplification of the transmitted
the safety factor as follows: load due to impact requires a sophisticated dynamic analysis that is
outside the scope of this work. However. Shigley [52J confirms that
(52) the universal rule of thumb of doubling the transmitted load to
account for impact effects is conservative. Therefore, if a torque
where: reversal occurs when the transmitted torque is below onc-half of its
SF Safety factor maximum value, the resulting gear load would still be below the
Endurance limit (psi) design load.
Sc
Equivalent fully-reversing stress (psi) Accordingly, torque reversals occurring at these transmitted
Seq
torque levels may be acceptable. However, it still must be remem-
Most of the references consulted recommcnd a minimum safety bered that torque revcrsals introduce a pounding effect that is
factor of 2.0. The authors are in complete agreement with this rec - absent during normal power transmission. This hammering, even at
ommendation. Thus. if the calculateu safety factor is greater than low force levels, could eventunlly lead to wear. Thus, in order for
2.0. the shaft can be considered structurally adequate. If not, design a torque reversal situation to be accepted, it must be validated from
changes need to be instituted to correct the situation. a tribological standpoint.
If the material fatigue properties are not immediately available.
Transient Allalysis
Mil. Std. 167 [511 provides a rough rule of thumb for estimating
the maximum allowable value of equivalent fully-reversing stress. There are some situations. such as the starting of synchronous
They recommend that the equivale nt stress in steel shafts be kept motors, where a steady state solution is totally inappropriate. In
below 112) of the material's ultimatc tensile strength . This these cases. a transient analysis must be pcrforrncu. These analyses
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 215

are very different from their steady state counterparts, both in the is continued in a time-marching manner until all transients decay
procedures used and the criteria for acceptability. Some of the and the system reaches the steady state condition . This procedure,
major differences between transient and steady state situations are therefore. estimates the time history of all relevant parameters in
as follows: the system. Evans, et. al. 1441. Chen 1451, anu Szenasi and von
• The m,lximum vibratory torques and stresses in transient Nimitz [541 all discuss numerical integration methous in much
problems are often less than the steady state values. This is because greater detail.
they often do not have sufticient time to build up to the steady state It should be noted that several authors comment on the large
values. Chapman r531 provides the following equation for the amount of computer time requireu to perform transient analysis
buildup of torque at resonance with no damping: using numerical integration mdhous. The user should. therefore.
be advised that reduction of the number of elements in the model
Tpeak = (N + .5) • n: • Texc (53) shoulu be investigated prior to running this type of analysis.

L('Ivis Procedure
where:
Tpeak Peak torque In addition to the rigorous proceuures uiscussed above, Lewis
N Number of resonant cydes l55) provides a simplitied procedure for estimation of the dynamic
TCH ' Excitation torque magnifier for a system accelerating or decelerating at a uniform
rate. Although he was working with a single degree of freedom
• In general, transient resonances are not as dangerous as those at system, his results also provide good estimates for torsional
steady state . This is due to two factors. The first is the lower systems with multiple disks. The dynamic magnifier, which Lewis
illlluced torque amplitude mentioned above. Addilionally, transient [55) refers to as the resonance factor, for a transient system is
resonances exist for only a finite number of cycles. Thus, compo- defined as follows :
nents need not endure their indu<;ed stresses for infinite life.
• In transient cases, there is a lag in the appearance of the peak (54 )
response. When a system ,is accelerating, the peak torques will
occur at a speed above the resonant speed and during decel'e ration, where:
the peaks occur at a speed below the ~ritical value. Thus, in both DM Dynamic magnitier
cases, the maximum response docs not oc€:ur until after the steady Tpeak Peak vibratory torque
state resonant point has beel) passed through. t cxc Excitation torque
• The system's dynamics have a large effect on transient perfor-
mance. Acceleration and deceleration rates are critical since they A plot of steady state dynamic magnifier vs excitation frequency
determine the dwell time at resonance. yields the well-known vibration response plot, similar to Figure 2,
• The uncertainty in magnitude of the various dampin.g factors is which has sharp peaks at all of the natural frequencies. When there
less of a concern in transient systems. Steady slate probl'e ms are is no damping in the system, the dynamic magnifier at resonance
extremely sensitive to the assumed values of damping since their is theoretically equal to inlinity.
resonant dynamic magnifiers are nonnal\y extremely high, given As previously pointed out, the dynamic magnifier at resonance
the low levels of damping typical for torsional systems. In transient for a transient situation is usually somewhat less than the steady
situations, the dynamic magnifiers are usually much lower and the state value. Lewis l55) noted that the reduction is strongly
level of sensitivity is cOITespondingly lower. dependent on the acceleration or deceleration rate. He presents this
• Unlike steady state si·t uations, ,i ncreased damping is not always relationship as a function of a dimensionless parameter he calls q.
beneficial. This is because higher damping values tend to increase which is defined as follows:
the speed range over which the dynamic magnifier is ncar its
maximum value. Thus, a situation can be envisioned in which (55)
increasing the system's (jamping level would increase the dwell
time at resonance and the. number of fatigue cycles. where:
q Acceleration (or deceleration) parameter
Analysis Methods fn Natural frequency under consideration (Hz)
Since there are many computer programs available for solving a Acceleration (or ueceleration) rate (Hz/sec)
transient vibration problems, the discussion of solution procedures
will be kept brief. Transient prohlems are sometimes solved using The parameter q represents the number of free vibrations that a
transform methods. These involve the transfer of information system would execute between the time when the system is at zero
between the time and frequency domains via Laplace or Fourier speed and the time when resonance is achieved. In his detinition,
transforms. These procedures are adequate for completely linear resonance occurs when the driving frequency is equal to the natural
systems but encounter problems when confronted with frequency, not at the point of peak response. Thus, q is basically a
nonlinearities. measure of the relative speed of acceleration or deceleration. Small
Modal methods, similar to those discussed for steady state values of q represent rapid transients while a value of infinity cor-
problems, are also used on transient problems. These methods responds to the steady state condition. The dynamic magnifier at
suffer from the same problems as the transform methods in that resonance, thus, increases as q is increaseu. The lowest value likely
they do not readily handle nonlinearities. to be observed in a practical system is approximately 10.
The most robust solution procedures for transient problems are Lewis [55] presents curves of dynamic magnifier vs excitation
numerical methods. In these, the differential equations of motion frequency ratio for various values of q and assorted damping ratios.
for the disks are integrated numerically using methods such as The ma)(imum transient uynamic magnitier can be obtained from
fourth-order Runge-Kutta and the Newmark-B method . The the peak of the appropriate curve. Additionally, Wachel and
analysis begins at time zero with all parameters set to their initial Szenasi 1161 provide a plot of peak dynamic magnitier as a
values. The numerical approximations, which udine a parameter's function of q and the steady state resonant uynamic magnifier.
new value in terms of its previous value, arc then used to determine From both of these sources, it can be seen that the difference
all parameler values aner the lirst limc increment. This procedure between the peak transient anu steady state dynamic magnifiers is
216 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

not very much unless the damping level is very small or the accel- During starting, the net torque available to accelerate the system
eration is extremely rapid. As the system damping is increased, the is equal to the motor's average torque minus the load torque. It is,
transient response curves begin to approach the steady state curves. therefore, helpful to also have the load's torque vs speed curve,
Additionally, it is seen that the dynamic magnifiers for the decel- although many common load torques can be approximated as
eration cases are slightly higher than the corresponding being proportional to the square of the speed. The system's accel-
acceleration values. eration rate in the vicinity of the critical speed is crucial because it
Lewis [55] also provides the following equation for the peak determines the length of time and number of cycles that the
dynamic magnifier in a transient, undamped system: machine spends at resonance.
To estimate the time spent at resonance, Brown [561 recom-
DMpcak = 3.67· q.5 ± .354 (56) mends the definition of a resonant speed band centered about the
critical speed. Chen, McLaughlin, and Malanoski [46] point out
The plus sign should be used for decelerations while the minus
that this band should be relatively small since torsional systems are
sign corresponds to the acceleration case.
usually lightly damped and, thus, have narrow resonant peaks . The
Analysis Methodology time spent within this band is critical for two reasons. The first is
that long resonance times give the vibration amplitude a chance to
During the course of a torsional vibration analysis, there are build up to the steady state value. Additionally, the number of
three common situations that require transient analysis, all damaging fatigue cycles experienced is directly proportional to the
involving electrical machines. The first case occurs when an AC resonance dwell time.
motor is initially powered. As was mentioned previously, this Synchronous motor startups are usually analyzed using a
condition is marked by large motor torque pulsations at line numerical integration time-marching scheme. While these provide
frequency that die out rapidly. If the system has a natural frequency an accurate time history for every element in the system, they are
in the vicinity of line frequency, an analysis of this condition must usually costly and time-consuming. As an alternative, Chen,
be performed. Pollard [31] provides beneficial information for the McLaughlin, and Malanoski [46J offer a simple method for esti-
analysis of this case. It should be noted that this case need not be mating the maximum cyclic torques and stresses that occur at
analyzed when the motor is started using an active variable resonance with the fundamental frequency. The method is modal in
frequency drive since the starting pulsations do not occur in this nature and only requires the motor's torque-speed curve and the
situation. However, as previously noted, this is an anomalous fundamental mode shape for implementation. With this informa-
situation since most YFDs are inactive during starting. tion, the calculation can be peliormed in a matter of minutes.
The second condition is the shorting of the terminals of an AC The potential shortcomings of this method are that it only deals
motor or generator. For this case, torque pulsations occur at one with resonances with the fundamental mode and does not attempt
and two times the electrical supply frequency. Once again, Pollard to determine the number of cycles spent at the maximum stress
[31] provides guidelines for the analysis of this situation. level. The first item may 1\0t be as large of a problem as it may
For machines driven by conventional controllers, the short appear since Pollard [31], Wright [47], and Sohre [57] all claim
circuit analysis needs to be performed only if the syst.em has that synchronous motor resonance problems usually occllr at the
natural frequencies near line or twice line frequency, However, the fundamental frequency . There are several reasons for this. First of
situation for motors driven by variable frequency drives is more all, the fundamental frequency is often characterized by large dis-
complicated. Since the supply frequency to these motors v.aries, the placement at the motor so it is especially vulnerable to motor
likelihood for resonance is much greater. Thus, these devices torque variations. Secondly, the fundamental mode is always the
should be analyzed for shorts at all natural frequencies that could mode having the highest energy level. Finally, the resonance with
potentially be excited. the fundamental occurs at the highest speed due to the inverse
Synchronous Motor Startups slope of the motor's excitation line so the pulsating torque there is
likely to be higher than for other modes.
The above two cases only need to be analyzed in the unfortunate The second criticism can be dealt with by making a conservative
situation where the system has a natural frequency in the vicinity estimate of the number of cycles at maximum stress. As Chen, et
of the excitation frequencies. On the other hand, the thLrd si.tuation, al. [46], point out, the resonance band is usually narrow enough
that of the starting of a synchronous motm, must be analyzed for that the acceleration rate can be taken as constant. This rate can be
all systems containing these motors and hav·i ng a natural frequency estimated by the following equation :
less than 120 Hz. Again, the exception for when the synchronous
motor is started with an LCI should be noted. (57)
The startup of synchronous motors is probably the most signifi-
cant transient phenomenon that must be dealt with when analyzing where:
torsional systems. As was previously discussed, these devices a. Acceleration rate through resonance (rad/sec 2)
generate pulsating torques at twice slip frequency during slarting. 'avc Average motor torque at resonant speed (in -Ibt)
As was shown in the Campbell diagram of Figure 12, these pu,l sa- 'load Load torque at resonant speed (in -Ibf)
tions excite all natural frequencies between zerO and 120 Hertz. JlOt Total system inertia (in - lbf-sec 2)
The tirst thing that must be done when analyzing this phenome-
non is the obtainment of an accurate torque vs speed curve ~romt/iie The time spent at resonance can then be determined by dividing
motor manufacturer. This curve should include both the average the width of the resonance band by the acceleration rate. The
and pulsating torque characteristics over the entire speed range number of cycles at maximum stress can then be obtained by mul-
from standstill to synchronous speed. An illustration of a typical tiplying the vibratory frequency by the time spent at resonance to
torque-speed curve is presented in Figure I I. obtain the number of resonant cycles and assuming that these
Both the average and pulsating torque characteristics play a cycles are all at maximum stress. To provide a point of reference,
large role in determining the severity of the vibration during Sohre [57J claims that a typical startup contributes between 10 and
starting. The pulsating torque's role is obvious since it rcpresents 20 high stress cycles.
the vibratory excitation torque. However, the average 't orque is also Of course, use of this method must be tempered by judgment.
important since it determines the acceleration rate of the drive There are some situations in which the fundamental mode may 110t
train. be the most dangerous . Pm instance, this mcthod would !lot he
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 217

appropriate for a system having a fundamental mode shape with the generic solutions to interference problems to be discussed in a
lillIe activity at the motor. However, in situations where it can be later section can be implemented. Additionally. synchronous
used, this procedure represents a large time and labor saver. motors have some unique characteristics that can also provide
solutions. Some of these traits are as follows :
Allalysis Results
• There is always a point in the speed range where the motor's
Regardless of the condition analyzed and the analysis method direct and quadrature axis torques cross. as is seen from Figure II .
utilized, the primary objective of a transient analysis is to Since the pulsating torque is proportional to the difference between
determine the maximum cyclic torques and stresses induced in the these two numbers. there is no pulsating torque at the crossover
various model elements. Although the goal is the same as for speed. Thus, if the system can be altered to place the resonant point
steady state analysis, the acceptance criteria are significantly close to the crossover point , substantial benefit would result. This
different. can be accomplished either by changing the system to alter the
First of all, the stress levels are not compared to the endurance natural frequency or by moving the crossover point by changing
limit since the number of cycles at maximum stress is much lower the motor's design .
than for a steady state resonance. Furthermore, there is often more
than one stress condition that must be considered for each shaft • Since both the motor and load average torque-speed curves often
since the induced stress verses time profile often looks like that of contain peculiarities, the net accelerating torque is far from
Figure 21. constant over the speed range. The accelerating torque should be
plotted as a function of speed and changes should be enacted to
locate the problem resonant point at a speed where the accelerating
torque is high to minimize resonance dwell time.
• In general, reduction of the pulsating torque is beneficial. The
pulsating torque at a given speed can be easily reduced by altering
the motor's design or by adding a resistance in series with the
motor. The resistance acts to reduce the voltage across the motor's
terminals which reduces the pulsating torque since it is approxi-
mately proportional to the square of this voltage. However, this
change is a double.edged sword since the ratio of pulsating torque
to average torque is approximately constant. The average torque
Time (sec) will, therefore. be reduced by about the same amount. However,
there are some cases, such as when the yield strength or ultimate
Figure 21. Typical Transiellt Respollse. strength is exceeded, w'here the net effect is positive.
• In contrast [0 the above item, there are situations where an
Determination of stntctural adequacy is easiest for the short increase in average torque is beneficial, in spite of the correspond-
circuit condition. Since a given machine can be expected to expe- ing rise in pulsating torque. An example is a system with a
rience, at most, a few short circuits during its lifetime, fatigue is relatively large load torque. In such a system. a small increase in
not a concern. Accordingly, the maximum induced stresses should average ,torque could result in a much faster acceleration rate and
be compared directly to the material's yield and ultimate strengths improved performance.
in shear.
For the other two cases. adequacy should be determined by first • In general, solid pole motors generate more problems during
calculating the equivalent fully-reversing stress for each induced starting than do those having laminated rotors. Thus, the simple
cyclic stress using the same procedure used for steady state change from a device with solid poles to one with laminated poles
stresses. Secondly. the number of cycles at each stress condition could e liminate an overloading problem. However, a switch to the
should be determined by multiplying the number of stress cycles more COstly laminated design must be justified from an economic
per start by the expected number of starts in the lifetime of the standpoint,
drive train. The allowable number of cycles for each stress state • Synchronous motors often have a dip in their average torque
should then be found from an S-N curve for the material. The curve at about one-half of synchronous speed. Since the pulsating
actual number and allowable number of cycles should then be used torque characteristic usually does not contain this dip. resonances
in a cumulative damage algorithm, such as a Miner's summation, close 10. this speed are usually troublesome and should be avoided.
to determine stnlctural adequacy.
In addition to the above check for fatigue, the maximum stress
ELIMINATION OF INTERFERENCE PROBLEMS
also should be compared to the shaft's shear yield strength (or
ultimate strength if yielding is permissible) to assure that a one- If the previous analysis reveals that there are any interference
shot failure does not occur. In the same vein, the maximum torques points that constitute a problem from a stress or gear unloading
should be compared to the maximum torque ratings for gears. standpoint, the user generally has four options for rectification.
splines, and couplings. First of all, the interference point can be eliminated by altering the
In contrast to the steady state situation. torque reversals at com- cognizant natural frequency or, less likely, changing the excitation
ponents with backlash are not cause for panic in the transient frequency. Secondly, the user can make design changes to reduce
situation. The greatly reduced number of reversal cycles in the the excitation applied to the system . Thirdly. the user can add
transient case makes tooth pitting and wear a much smaller damping to the system to attenuate the resonant response. Lastly,
concern. Szenasi and von Nimitz [54J note that transient reversals the user can designate a barred speed range centered about the
are very common and are often allowed for in the design of gears interference point in which the machine is forbidden from
and geared couplings. operating continuously.

Synchronous Motor Characteristics Alteration of Natural Frequencies

If the analysis reveals that the system has a transient stress The option that should be investigated first is removal of the
problem. the designer has several uptions for rectification. Any of problem interference point via change of the natural frequency.
218 PROCEEDINGS OF T HE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

Natural frequencies may be altered by either changing disk inertias The third option is addition of damping to the system. There are
or shaft stiffnesses. Increasing spring rates tends to raise natural many types of couplings available that produce significant
frequencies while increasing inertias has the opposite effect. damping via elastomers or oils contained within. Additionally.
In order to identify candidate elements for modification, the even if the excitation source and critically stressed shaft are located
appropriate mode shape should be reviewed. In general, shaft such that introduction of a hydraulic coupling would not isolate
elements Ihat experience significant twisting in that mode are can- them from each other, such a coupling can introduce substantial
didates for change . This is because the sensitivity of a natural damping to the assembly. Finally, there are many devices such as
frequency to a change in a particular shaft element is proportional Lanchester dampers, whose sole purpose is to provide torsional
to the total strain energy stored in that shaft. Since strain energy is damping, that can be utilized.
proportional to the square of the angle of twist, altering an element Regardless of the damping element chosen. it should be strate-
experiencing considerable twisting will have a large impact on the gically located in order to be effective. As was discussed
natural frequency. On the other hand, shafts having very little previously, dampers are only effective if they are located in regions
twisting can be altered by an order of magnitude and still have a of significant vibration activity for the mode under consideration.
negligible effect on the natural frequency. The appropriate mode shape, therefore, should be utilized to
In turbomachinery, couplings are usually the easiest components determine optimum locations for dampers. Ideally. dampers should
to change. Therefore, the twist occurring in them should be be positioned in places where the mode shape reveals the absolute
examined closely to determine if there is any potential benefit to or relative displacement to be near maximum.
their alteration. Fortunately, the fundamental mode in many turbo- The final option is designation of a certain speed range to be off-
machinery drives is characterized by twisting exthlsively in one limits for steady operation. The desirability of this option is
coupling. This occurs because systems often -contain II coupling dependent on the level of hardship it imposes on the user. In
whose stiffness is an order of magnitude lower than any other shaft general, this option is used only in cases where none of the other
in the system. Since most vibration problems occur with the fun- alternatives can be practically implemented.
damental mode. a simple coupling change is often sufficient to
solve the problem. OVERALL ANALYSIS PROCEDURE
Although impellers are usually not as easily altered as At this point all of the component steps of a torsional vibration
couplings, their moditication can also eliminate a problem inter- analysis procedure have been described in detail. To summarize, a
ference. As is the case wi.lh shaft elements, the mode shape chronological listing of these steps is as follows:
provides a good indication of which impellers should be altered.
The sensitivity of a natural frequency to change of a specific inertia I. Generate lumped parameter model.
is proportional to the percentage of the system's kinetic energy 2. Determine undamped natural frequencies and mode shapes.
stored in that disk. Since the energy is proportional to the square of 3. Verify undamped analysis results using hand calculations.
the maximum amplitude, only disks having significan.t vibratory 4. Generate Campbell diagram.
amplitudes should be considered for modification . • Plot natural frequencies.
If a shaft element or inertia change is found to be sufficient to • Plot operating speed range.
move the problem interference point away from the operating • Plot excitation lines (See Table I for guidelines).
range with sufficient margin. that particular point can be dismissed. 5. Determine all interference points from Campbell diagram.
However, any change to the system can alter all natural frequen- • All intersection points within operating speed range
cies, not just the one being considered. The new natural • All intersection points outside of speed range that don't have
frequencies must, therefore, be calculated and checked for any new adequate margin (see Equation (22».
interferences that may have been generated . • For systems containing AC machines, natural frequencies at
Another way to eliminate a problem interference point is by line and/or twice line frequency
placement of a hydraulic coupling between the excitation source • All intersection points with synchronous motor twice slip
and region where shaft overstressing or gear unloading is taking frequency excitation line
place. As stated previously, hydraulic couplings serve to divide the • All intersection points with active VFD excitation lines
assembly into two independent systems, thus, isolating the two 6. Use inspection techniques to eliminate interference points.
systems from each other. For instance. if torque pulsations from a • Points excited by torques located near nodes
synchronous motor were producing high stresses in a shaft remote • Points excited by gear displacements located away from
from the motor, a fluid coupling placed at the motor's output shaft nodes
would remove the problem. Of course, if this tactic is used, the two • Points that are obviously less severe than other interference
new resulting systems would need to be freshly ana1yzed. points involving same mode
A less common method of eliminating interferences is via • Points where mode shapes show little twist in vulnerable
changing the order number of the exciting component. This option regions
is only viable for order numbers above two since tile I X and 2X 7. Determine excitation torque magnitudes.
excitations are always assumed to be present. Howev:er, if the exci- • Generic I X excitations
tation is caused by the blade pass frequency of an impeller or a gear • Generic 2X excitations
meshing frequency, it can be altered by merely changing the • Gear excitations
number of blades or gear teeth, as long as the change does not • Impeller excitations
adversely impact performance. • Propeller excitations
The second alternative for escaping trouble is reduction of the • AC machine excitations
excitation driving the system. The applicability of this method is • Steady state torques
dependent on the nature of the exciting mechanism. If the excita- • Peak torques during shol1 circuits
tion arises from an electric motor or generator, design changes can • Peak torques during initial powering
often be made which reduce the torque fluctuations. Additionally. • Variable frequency drive excitations
excitations generated by gears, particularly disp'l accment excita- 8. Perform preliminary steady state forced vibration analysis.
tions, can be reduced by applying tighter tolerances to the gear • Identify interference points that must be analyzed (all except
tooth profile. ror those listed under transient analysis).
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 219

• For each cognizant interferencc point, exccute the following then utilized to dismiss points, one by one, until there are none
steps: remaining. When this point is rcached, the system is viablc.
• Apply excitation at appropriate location.
CONCLUSIONS
• Apply damping of one percent of critical to all shaft
elements. A comprehensive procedure for the analysis of torsional
• Determine induced torques and stresses throughout system. vibration has been presented . The methodology is general enough
• Evaluate results : that it should be applicable to any turbomachinery configuration
• All shafts must have infinite lives. that may be encountered. The key points that should be empha-
• Peak torques in gears, splines, and couplings must not sized are as follows:
exceed their continuous ratings.
• A thorough torsional vibration analysis should always be
• Effects of torque reversals in connections having included as an integral pari of the turbomachinery design process.
backlash must be evaluated.
• If results indicate a potential problem, proceed to stcp 9. • The essence of torsional vibration analysis is identification of all
9. Pelfonn rigorous steady state forced vibration analysis. resonancc points and determination of the system's ability to
• Determine damping coefficients for system components. withstand them .
• Loads (impellers, propcllers, generators, etc.) • A Campbell diagram should be generated as soon as possible to
• Drivers (motors, turbines, etc.) provide visibility into the overall situation.
• Windage
• Fluid-tilm journal bearings • Systcms e mploying synchronous motors and/or variable
frequency drives should be treatcd with extra caution .
• Couplings (elastomeric, hydraulic, etc.)
• Cyclic speed variation • Manufacturers of equipment such as motors and couplings
• For each cognizant interference point, execute the following should bc consulted with regards to the torsional vibration charac-
stcps: tcristics of their products.
• Apply excitations and damping ;It appropriate locations.
• The presenoe of resonance points in a system is not an automatic
• Apply damping of one percent of critical to aU shaft
harbinger of disaster. MallY resonant points can be shown to be
elements.
harmless v,ia various methods .
• Determine induced torques and sttesses throughout system.
• Evaluate results (see step 8) . • Torsional vibration analysis is not an exact science. Although
• If results indicate a potential problcm, proceed to step 12. many of the calculation procedures arc relatively sophisticated, the
10. Perform preliminary transient forced vibration anal ysis. role of the skill and judgment of the analyst should not be
• Identifilnterference points that must be analyzed (see Table undcrestimated.
4).
REFERENCES
• For each cognizant interference point, cxecute the following
steps: I. Vance, 1. M., Rotordynamics of Turbo/1/acltinery, New York.
• Apply excitation at appropriate location. New York: 10hn Wiley and Sons (\988).
• Apply damping of one percent of critical to aU shaft 2. Ker Wilson, W., Practical Solution of Torsional Vibratioll
elements. Problems, Volumes I, 2, and 3, New York, New York: 10hn
• Determine induced torques and strcsses in systcm as Wiley and Sons, 3rd Ed. (1956).
function of time.
3. Eshleman , R. L., "Torsional Response of Internal Combustion
• Evaluate rcsults .
Engines," ASME 10urnal of Engineering for Industry, pp. 441-
• Fatigue lives of all shafts must be satisfactory.
449 (May 1974) .
• Peak torques in gears, splincs,. and couplings must not
exceed their maximum rlliings. 4. Bisshopp, K. E., "Forced Torsional Vibration of Systems with
• If results indicate a potenlial problem, proceed to step I I. Distributed Mass and Internal and External Damping," ASME
II . Perform rigorous transient forced vibration analysis. Journal of Applied Mechanics, pp. 8-12 (March 1959).
• Determine damping coefficients for system components (see 5. Triezenberg, D. M., "Characteristic Frequencies and Mode
step 9). Shapes for Turbogenerator Shaft Torsional Vibrations," IEEE
• For each cognizant interference point, execute the following Transactions on Power Apparatus and Systems , pp. 352-357
steps: (January/February 1980).
• Apply cxcitations and dampin g at appropriate locations.
• Apply damping of one percent of critical to all shaft 6. Simmons, H. R. and Smalley, A. 1., "Lateral Gear Shaft
elements. Dynamics Control Torsional Stresses in Turbine-Driven
• Determine induced torqucs and stresses in system as Compressor Train," ASME 10urnal of Engineering for Gas
function of time. Turbines and Power, pp . 946-951 (October 1984).
• Evaluate results (see step 10). 7 . Lund, 1. W., "Critical Speeds, Stability and Response of a
• If results indicate a potential problem, proceed to step 12. Geared Train of Rotors," ASME Journal of Mechanical Design.
12. Implement modifications to eliminate interference problems. pp. 535-539 (July 1978).
• Investigate means of changing natural frcquencies.
8. Nestorides, E., A., Handbook on Torsional Vibration , London:
• Investigate methods of changing excitation frequencies .
Cambridge University Press (1958).
• Investigate means of reducing excitation magnitudes.
• Investigate feasibility of adding damping to system . 9. O'Connor, B. E., "The Viscous Torsional Vibration Damper:·
• If necessary, designate speed range to bc off-limits (usually SAE Quarterly Transactions, pp. 87-97 (1anuary 1947).
last resort). 10. Den Hartog, 1. P. and Ormondroyd, 1., "Torsional-Vibration
13 . Analyze new system resulting from step 12 changes. Dampers," ASME Transactions, papcr APM-52-13 (1930).
It is seen that stcps one through five involvc the identitication of II. Brown, R. N., "Torsional Damping- Transient and Steady
all of the machine's intcrkrence points. The remaining steps are State ," Proceedings of" th e Thirteellth TlIrblJllla chillen·
220 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHINERY SYMPOSIUM

Symposium, Turbomachinery Laboratory, Texas A&M 29. Artilles, A., Smalley, A. J. , and Lewis, P, "Mill Drive System
University, College Station, Texas ( 1984). Vibrations:' Proceedings or the National Conference on Power
Transmission , Chicago, Illinois (October 1975).
12. McCann, G. D. and Bennett. R. R., "Vibration of
Multifrequency Systems During Acceleration Through Critical 30. Yates , H. G., "Prediction and Measurement of Vibration in
Speeds." ASME Journal of Applied Mechanics, pp. 375-382 Marine Geared-Shaft Systems," Proceedings of the Institution
(Decembcr 1949). of Mechanical Engineers, pp. 61 1-642 ( 1955).
13. Pollard, E. L, "Transient Torsional Vibration Due to Suddenly 31. Pollard, E. I., "Synchronous Motors ... Avoid Torsional
Applied Torque," ASME Journal of Engineering for Industry, Vibration Problems," Hydrocarbon Processing (February
pp. 595-602 (May 1972). 1980).
14. Simmons. H. R. and Smalley, A. J., "Effective Tools for 32. Wolff, F H. and Molnar, A. 1., "Variable-Frequency Drives
Diagnosing Elusive Turbomachinery Dynamics Problems in Multiply Torsional Vibration Problems," Power, pp. 83-85
the Field," ASME Journal of Engineering for Gas Turbines and (June 1985).
Power. pp. 470-477 (October (990).
33. Grgic, A., Heil. W., and Prenner. H., "Large Converter-Fed
15. Hershkowitz, H., "Torsional Vibration Measurement in Adjustable Speed AC Drives for Turbomachines," Proceedillgs
Preventive Maintenance of Rotating Machinery," Scientific- of the Tw enty-First Turbomachinery Symposium,
Atlanta, Inc., Randolph Township. New Jersey. Turbomachinery Laboratory, Texas A&M University. College
16. Wachel, J. C. and Szenasi. FR., "Analysis of Torsional Station, Texas (1992).
Vibrations in Rotating Machinery," Proceedings of the Twenty- 34. Murphy, S. P., "Application of Variable Speed Electric Motors
Second Turbomachinery Symposium, Turbomachinery for Pumps," Proceedings of the Tenth international Pump
Laboratory, Texas A&M University, College Station, Texas Users Symposillm, Turbomachinery Laboratory, Texas A&M
(1993). University, College Station, Texas (1993).
17 . Hammons, T. J ., "Electrical Damping and its Effect on 35. Sheppard, D . J., "Torsional Vibration Res ulting from
Accumulative Fatigue Life Expenditure of Turbine-Generator Adjustable Frequency AC Drives," IEEE Transactions on
Shafts Following Worst-Case Supply System Disturbances," Industry Applications, pp. 812-817 (September/October 1988).
IEEE Transactions on Power Apparatus and Systems. pp. 1552-
36. Hudson, J. H., "Lateral Vibration Created by Torsional
1565 (June 1983).
Coupling of a Centrifugal Compressor System Driven by a
18. "Reader's Guide to Subsynchronous Resonance," IEEE Current Source Drive for a Variable Speed Induction Motor:'
Transactions on Power Systems, pp. 150-156 (February 1992). Proceedings of the Twellly-First TuriJornachinl'I)' Symposillm,
Turbomachinery Laboratory, Texas A&M University, College
19. Rana, R. D. and Schulz, R. P., "Generator Loss of Field:
Station, Texas (1992).
Experience and Studies for AEP's Rockport Plant."
Proceedings of the American Power Conference. 37. Smalley, A. J., "Torsional System Damping." Proceedings of
the 1983 Vibration In stitute Seventh Annual Meeting, Houston,
20. Joyce, J. S., Kulig, T., and Lambrecht, D., "Torsional f atigue
Texas (1983).
of Turbine-Generator Shafts Caused by Different Elecrrical
System Faults and Switching Operations," IEEE PES Winter 38. Draminsky, P., "Crankshaft Damping," Proceedings of the
Meeting, New York, New York (1978). Institution of Mechanical Engineers, pp. 416-432 (1948).
21. Undrill, J . M. and Hannett, L. N., "Turbine-Generntor Impact 39. Shannon, J. E, "Damping Intluences in Torsional Oscillation,"
Torques in Routine and Fault Operations," IEEE PES Winter Proceedings of the Institution of Mechanical Engineers, pp.
Meeting, New York, New York (1978). 387-435 (1935).
22. Andriola, A. D., "Torsional Vibration in Geared-Turbine 40. Dorey, S. E, "Elastic Hysteresis in Crankshaft Steels,"
Marine Propulsion Plants-An Introduction to the Subject," Proceedings of the Institution of Mechanical Engineers, pp.
Transactions of the Society of Naval Architects and Marine 479-5 10 ( 1932).
Engineers, pp. 733-752 (1950) .
41. Rieger, N. E, "The Role Gears Play in Drive-Train Vibrations."
23. Den Hartog, 1. P., Mechanical Vibrations, Fourth Edition, New Machine Design. pp. 115- 119 (July 1969).
York. New York: McGraw-Hill Book Company (1956).
42. Frei, A, Grgic, A., Heil, W, and Luzi, A. , "Design of Pump
24. Schlegel, R. G., King, R. J., and Mull , H. R., "How to Reduce Shaft Trains Having Variable-Speed Electric Motors,"
Gear Noise." Machine Design, pp. 134-142 (February 1964). Proceedillgs of th e Third illtemationa/ Pump SVlI1po.l'illlll.
25. Mayer, C. B .. "Torsional Vibration Problems and Analysis of Turbomachinery Laboratory, Texas A&M University, College
Cement Industry Drives ," IEEE Transactions on Industry Station, Texas (1986).
Applications, pp. 81-89 (JanuarylFebruary 1981). 43 . Thames, P. B. and Heard, T C .. "Torsional Vibrations in
26. Ehrich. F E. Handbook of ROlOrdynamics, New York, New Synchronous Motor-Geared-Compressor Drives," AlEE
York: McGraw-Hili Inc. (1992). Transactions, pp. 1053-1056 (December 1959).
27. Poritsky, H.. "Torsional Vibration in Geared Turbine 44. Evans, B. E, Smalley, A. J., and Simmons, H. R., "Startup of
Propulsion Equipment," ASME Journal of Applied Mechanics, Synchronous Motor Drive Trains: The Application of Transient
pp. A-117 - A-124 (September 1940). Torsional Analysis to Cumulative Fatigue Assessment ," ASME
paper 85-DET-I22 ( 1985).
28 . "Torsional Vibrations," 1988 Seminar 011 Vibration Analysis
and Diagnostic Techniques for Rotating Machinery, 45 . Chen. W. J., "Torsional Vibrations of Sync hronous Motor
Mechanical Technology Incorporated, Latham , New York Driven Trains Using p-Methud," ASME Journal or Vibration
(September 1988). and Acoustics. pp . 15 2- 160 (January 19(5)
 PRACTICAL DESIGN AGAINST TORSIONAL VIBRATION 221

46. Chen. H. M .• McLaughlin, D. W., and Malanoski. S. B .• "A Cudworth, C. J., Smith, J. R .. and Mykura, 1. E, "Mechanical
Generalized and Simplified Transient Torque Analysis for Damping of Torsional Vibrations in Turbogenerators Due to
Synchronous Motor Drive Trains." Proceedings of the Twelfth Network Disturbances," Proceedings of the Third' International
TlIrbol11achinery SymposiulII, Turbomachinery Laboratory. Conference on Vibrations in Rotating Machinery, Institution of
Texas A&M University, College Station. Texas (1983). Mechanical Engineers, pp. 139-144 (September 1984).
47. Wright, J., "Large Synchronous Motor Drives: A Review of the Cudworth. C. J. and Smith. J. R .• "Turbine-generator Shaft
Torsional Vibration Problem and its Solution." Koppers Torsional Oscillations Produced by Imbalanced Transmission
Company, Inc. Network Disturbances," Proceedings of First Parsons
Inlernational Turbine Conference, Institution of Mechanical
48. Tuplin. W. A .. Torsional Vihration, New York, New York: John
Engineers. pp. 241-246 (June 1984).
Wiley and Sons (1934).
Doughty, S .• "A Rayleigh-Type Inclusion of Shaft Inertia in
49. Den Hartog, J. P. and Li, J. P., "Forced Torsional Vibrations
Torsional Vibration Analysis," ASME Journal of Engineering
with Damping: An Extension of Holzer's Method," ASME
for Gas Turbines and Power, pp. 831-837 (October 1994).
Journal of Applied Mechanics, pp. A-276-A-280 (December
1946). Eshleman. R. L.. "Torsional Vibration of Machine Systems."
Proceedings of the Sixth Turhomaclzinery Symposium,
50. Badgley, R. H. and Laskin, I.. "Program for Helicopter
Turbomachinery Laboratory, Texas A&M University, College
Gearbox Noise Prediction and Reduction," MTI USAAVLABS
Station, Texas (1977).
Technical Report 70-12, FI. Eustis. Virginia (March 1970).
Godwin, G. L. and Merrill. E. E, "Oscillatory Torques During
51. MIL STD 167. Mechanical Vibrations of Shipboard Equipment
Synchronous Motor Starting." IEEE Transactions on Industry
(Reciprocating Machinery and Propulsion System and
and General Applications. pp. 258-265 (May/June 1970).
Shafting) (1974).
Graybeal. T. D .• "The Nature of Vibration ill Electric Machinery,"
52. Shigley, J. E., Mechanical Engine('l'illg Desigll, Third Edition,
Electrical Engineering. pp. 712-718 (October 1944).
New York, New York: McGraw-Hill Book Company (1977).
Harris, S. L., "Dynamic Loads on the Teeth of Spur Gears,"
53. Chapman. C. W., "Zero (or Low) Torsional Stiffness
Proceedings of the Institution of Mechanical Engineers, pp. 87-
Couplings," Journal of Mechanical Engineering Science, 11, 100 (1958).
(I), pp. 76-87 (1969).
Hyde, R. L. and Brinner. T. R.. "Starting Characteristics of Electric
54. Szenasi, E R. and von Nimitz, W. W., "Transient Analysis of Submergible Oil Well Pumps," IEEE Transactions on Industry
Synchronous Motor Trains," Proceedings of the Seventh Applications. pp. 133-144 (January/February 1986).
TlIrbomachinery SymposiulIl. Turbomachinery Laboratory,
Texas A&M University, College Station, Texas (1978). Johnson, D. c., "The Excitation of Resonant Vibration by Gear
Tooth Meshing Effects," Proceedings of the International
55. Lewis EM.. "Vibration During Acceleration Through a Conference on Gearing. Institution of Mechanical Engineers.
Critical Speed," ASME Transactions-Applied Mechanics, pp London. pp. 18-23 (September 1958).
253-261 (1932).
Kashay. A. M .• Voelker. E c., and Smalley. A. J., "Dynamic Shock
56. Brown. R. N., "A Torsional Vibration Problem as Associated Phenomena in Rolling Mills," ASME Journal of Engineering
with Synchronous Motor Driven Machines," ASME Journal of for Industry, pp. 647-659 (May 1972).
Engineering for Power, pp. 215-220 (July 1960).
Kenney. C. S. and Shih, S .• "Prediction and Control of Heavy Duty
57. Sohre, J. S., "Transient Torsional Criticals of Synchronous Powertrain Torsional Vibration." SAE Transactions, pp . 805-
Motor Driven, High-Speed Compressor Units," ASME Paper 814 (1992).
65-F£-22 (1965).
Kimball, A. L. and Lovell, D. E .• "Internal Friction in Solids."
ASME Transactions. pp. 479-500 (1926).
BIBUOGRAPHY
Lack. A. and Jahnke. C. B .• "Torsional Vibrations and Critical
Alison, N. L., "Hydraulic Couplings for Internal Combustion
Speeds of Shafts," ASME Transactions. pp. 493-523 (1925).
Engine Applications," ASME Transactions. pp. 81-90
(February 1941). Lambrecht, D. and Kulig. T., "Torsional Performance of Turbine
Generator Shafts Especially Under Resonant Excitation," IEEE
Badgley, R. H., "Mechanical Aspects of Gear-Induced Noise in Transactions on Power Apparatus and Systems, pp. 3689-3702
Complete Power Train Systems," AS ME Paper 70-WNDGP-I (October 1982).
( 1970).
Lazan; B. J .• "Effect of Damping COllstanls and Stress Distribution
Bigret, R., Coetzee, C. J" Levy, D. c., and Harley, R. G., on the Resonance Respollse of Members... ASME Journal of
"Measuring the Torsional Modal Frequencies of a 900 MW Appli.e d Mechanics, pp. 201-209 (June 1953)
Turbogenerator," IEEE Transaclions on Encrgy Conversion,
pp. 99-107 (December 1986). Lazall. B. J .• "Energy Dissipation MechaAisms in Structures, with
Particular Reference to Material Damping," Proceedings of
Braund, D. E, "Torsional Vibration," Proceedings of the Institution Colloquium on Structural Damping, ASME, Atlantic City. New
of Mechanical Engineers, pp. 63-72 (1958). Jersey, pp. 1-34 (December 1959).
Costello, M. J., "Understanding the Vibration Forces in Induction Lees, A. W .• "Dynamic Loads in Gear Teeth." Proceedings of the
Motors," Proceedings ()f the Nilleteenth Turholllacllinery Third International Conference on Vibrations in Rotating
SymposiulIl. Turbomachinery Laboratory, Texas A&M Machinery, Institution of Mechanical Engineers, pp. 73-79
University, College Station, Texas (1':)90). (September 1984).
Cox. C. R., "Design Considerations for Acceptable Cabin Noise Lees, A. W. and Haines. K. A., "Torsional Vibrations of a Boiler
Levels in Light Helicopters," American Helicopter Society Feed Pump," ASME Journal of Mechanical Design, pp. 637-
Prcprint No. SW-70-23 (1970). M3 (October 1978).
222 PROCEEDINGS OF THE TWENTY-FIFTH TURBOMACHlNERY SYMPOSIUM

Lewis, F. M., "The Critical Speeds of Torsional Vibration," SAE Smith, J. R., Mykura, J. F., and Cudworth. C. J., "The Effect of
Journal. pp. 418-431 (November 1920). Hysteretic Damping on Turbogenerator Shaft Torsional
Lewis, F. M" "Torsional Vibration in the Diesel Engine," Oscillations," IEEE Transactions on Energy Conversion, pp.
Transactions of the Society of Naval Architects and Marine 152-[57 (March 1986).
Engineers. pp. 109-140 (1925). Spaetgens, 1'. W., "Ho[zer Method for Forced-Damped Torsional
Mahalingham. S. and Bishop. R. E., "Dynamic Loading of Gear Vibrations," ASME Journal of Applied Mechanics, pp. 59 -63
Teeth," Journal of Sound and Vibration, pp. 179-189 (1974). (March 1950).

Mark, W. 0 .. "Analysis of the Vibratory Excitation of Gear Thomsen, W. 1'., Th eo ry of Vibration with Applications,
Systems: Basic Theory," Journal of the Accollstic Society of Englewood Cliffs, New Jersey : Prentice Hall, 3rd Ed. (1988).
America. pp. 1409-1430 (May 1978). Tuplin, W. A., "Dynamic Loads on Gear Teeth." Proceedings of the
Mayer, C. B" "Torsional Dynamics-An Important Consideration International Conference on Gearing, Institution of Mechanical
in Large Refiner Drive Systems," 1985 TAPPI Engineering Engineers, London, pp. 3 [-42 (September 1958).
Conference, pp. 105-115 (1985). Wachel, J. c., "Design Audits." Proceedings of the Fifteenth
Mitchell, R. W., "The Design Office Problem in the Estimation of Turbolllllchillery Symposium, Turbomachinery Laboratory,
the Torsional Resonance Characteristics of Small Marine Texas A&M University, College Station, Texas (1986).
Diesel-Propulsion Units," Proceedings of the Institution of Wachel, J. c., Atkins, K. E" and Tison , J. D., "Improved
Mechanical Engineers, pp. 133-142 (1943). Reliability through the Use of Design Audits," Proceedings of
Mondy, R. E. and Mirro, J., "The Calculation and Verification of the Twenty-Fol/rth Tlirbolllachinery Symposiulll,
Torsional Natural Frequencies for Tnrbomachinery Equipment Turbomachinery Laboratory, Texas A&M University, College
Strings," Proceedings of the Elevelllh Tlirbo/llachinery Station, Texas (1995).
Symposilllll, Turbomachinery Laboratory, Texas A&M Wahl. A. M. and Kilgore. L. A" "Transiel1l Starting Torques in
University. College Station, Texas (1982). Induction Motors," Electrical Engineering-Transactions , pp .
Mruk , G. K., "Compressor Response to Synchronous Motor 603-607 (November [940) .
Startup," Proceedings of the Seventh Tlirbomachinery Wahl. A. M., "Transient Torques in Induction Motor Drives,"
Symposilll/l. Turbomachinery Laboratory, Texas A&M ASME Journal of Applied Mechanics, pp. A-17 - A-21 (March
University. College Station. Texas (1978). 1941).
Pollard, E. I.. "Torsional Response of Systems," ASME Journal of Walker, D. N., Bowler, C. E., Jackson, R. L. , and Hodges, D. A.,
Engineering for Power, pp. 316-324 (July 1967). "Results of Subsynchronous Resonance Test at Mohave," IEEE
Porter, F. p" "Practical Determination of Torsional Vibration in an Transactions on Power Apparatlls and Sy stems, pp. 1878-1889
Engine Installation which may be Simplified to a Two-Mass (September/October [975).
System," ASME Transactions-Applied Mechanics, pp. 25-61. Walker. D. N., Adams, S. L. , and Placek, R. J., "Torsional
Porter, F. p" "The Range and Severity of Torsional Vibration in Vibration and Fatigue of Turbine-Generator Shahs," IEEE
Diesel Engines," ASME Transactions-Applied Mechanics, pp. Transactions on Power Apparatus and Systems, pp. 4373-4380
25-61 (1927). (November 1981).

Reswick. J. B., "Dynamic Loads on Spur and Helical Gear Teeth," Wang , C. c., "Rotational Vibration with Backlash: Part I," ASME
ASME Transactions, pp. 635-644 (July 1955). Paper 77-DET-I 05 (1977).

Rowett, F. E .. "Elastic Hysteresis in Steel," Proceedings of the Wang, S. M. and Morse, I. E., "Torsional Response of a Gear Train
Royal Society, pp. 528-543 (1914). System," ASME Journal of Engineering for Industry, pp. 583-
594 (May 1(72).
Schlegel, R. G. and Mard, K. c., "Transmission Noi se
Control-Approaches in Helicopter Design," ASME Paper 67- Williams , P. N., McQuin, N. P., and Buckland, J. E., "The
DE-58 (1967). Importance of Complete Drive Train Ana[ysis for Brushless
Salient Pole Motor Drives," Fourth Il1Iernational Conference
Seireg, A. and Hou ser. D. R., "Evaluation of Dynamic Factors for on Electrical Machines and Drives, lEE-England, pp. 210-214
Spur and Helical Gears," ASME Journal of Engineering for ( 1989).
Industry. pp. 504-515 (May 1970).
Shadley, J. R" Wilson , B. L" and Dorney, M. S., "Unstable Sclf- ACKNOWLEDGMENTS
Excitation of Torsional Vibration in AC Induction Motor
The authors would like to recognize and thank their colleagues
Driven Rotational Systems," ASME Journal of Vibration and
at MTI who assisted in the generation of this presentation .
Acollstics. pp. 226-231 (April 1992).
Specifically. publication of this tutorial would not have been
Smalley, A. J., "Transient Torsional Vibration," Mechanical possible without the contributions and talents of Mike Cronin .
Technology Incorporated. Latham, New York (September Sandy MacCue, Donna Rivers, and Shirl Smith.
1974).