Dyrobes Rotordynamics Software

http://dyrobes.com

The INFLUENCE of BEARING ASYMMETRY on ROTOR STABILITY

Edgar J. Gunter, Ph.D.

RODYN Vibration Analysis, Inc.
Professor Emeritus Mechanical and Aerospace Engineering
Former Director, Rotor Dynamics Laboratory, Univ. Of Virginia
Fellow ASME
[email protected]

Abstract
This paper deals with the influence of bearing asymmetry on the stability characteristics of a
rotor under the influence of internal friction and aerodynamic cross coupling forces. The Jeffcott
rotor is first examined for stability under each of these excitation mechanisms. It is seen that
bearing asymmetry plays an important role in improving rotor stability. Of great importance,
however is the relationship of the vertical bearing stiffness to the shaft stiffness. When the
vertical bearing stiffness exceeds the shaft stiffness, then the influence of asymmetry is
diminished. Also of importance is the strength of the instability mechanism. As this value
increases, then it becomes increasingly difficult to stabilize the rotor through the effects of
bearing asymmetry alone. The inclusion of bearing damping is of major importance in order to
achieve significant improvements in stability. There is an optimum amount of bearing damping
to be used to apply maximum stability. High bearing damping in combination with high bearing
stiffness values, however, causes a degradation in rotor stability. The stability of multistage
compressors in five pad bearings is examined because of the apparent benefit of these
asymmetrical stiffness bearings in improving stability. It will be demonstrated that many
previous five pad bearing designs for multistage compressors had serious stability problems
because of the excessive vertical stiffness values generated by these bearings with L/D aspect
ratios less than one and preloads exceeding 0.3. It will also be shown that improvement in
stability may be achieved by the use of the 5 pad bearing with reduced preloads or by the use of
the 4 pad load between pad bearing design with extended bearing aspect ratios. This design
approach usually eliminates the need for a squeeze film damper which is required in many of the
previous rotor designs using 5 pad bearings.

1
Background and Introduction
The use of the tilting pad bearing for compressors and industrial turbines has lead to significant
improvements in turbomachinery performance and operating speeds. As compressor designs have
increased to rotors with impeller stages exceeding 8 and speeds in excess of 10,000 RPM, serious
stability problems have been encountered in some high pressure compressors because of the
generation of internal Alford type aerodynamic cross coupling forces developed in these machines.
This condition is not detected under normal factory low pressure performance runs but occurs under
full operating conditions in the field.
The increase in number of stages creates a very flexible rotor design with a reduced critical speed.
The higher the operating speed is above the critical speed, the more sensitive is the unit to self
excited forces. This leads to the occurrence of fractional frequency whirling of the shaft in which
the shaft motion has a large subsynchronous component related to the rotor first critical speed.
Although the whirl motion may often appear to be constrained at the bearings, the center span
motion may exceed the labyrinth seal clearances causing rubs or even shaft failure.
A detailed report by Smith in 1975 on fractional frequency whirling of a high pressure compressor
was reported at the Fourth Annual Turbomachinery Symposium sponsored by Texas A&M
University. The 9 stage rotor was operating at speeds in excess of 10,000 RPM and was supported
by 5 pad bearings. In an attempt to improve the stability, several bearing modifications were
attempted in order to increase the bearing asymmetrical stiffness properties to promote higher
stability.
This paper was of considerable interest to the author because the motivation for the bearing
modifications was based upon my dissertation and ASME paper on the influence of anisotropic
supports on rotor stability. The primary consultant on the project was my thesis advisor from the
University of Pennsylvania. It should be noted that a detailed analysis of the rotor bearing stability
characteristics was not possible at that time. The Lund programs on tilting pad bearings and
complex eigenvalue analysis ( damped natural frequencies) were not available for use at that time.
In the paper presented by Smith are shown a number of bearing modifications attempted in order to
increase the bearing anisotropic properties. These bearing modifications proved to be unsuccessful.
A final design incorporating a squeeze film damper was also of limited success as it was improperly
designed as a long bearing damper which generated excessive stiffness and damping.
The final design, in order to operate at the higher speeds and pressure ratios, involved a new shaft
design with a reduced bearing span and increased shaft diameter. Of interest to note that a later
consulting group (using the Lund bearing and stability programs) concluded that the zero preload
bearing design would be as good as the installed damper. It was reported that the total cost of the
retrofit and production loss exceeded 25 million dollars!
This paper presents how this situation may be avoided in the future by the proper selection of the
type of tilting pad bearing, choice of preload and bearing aspect ratio. It is estimated today that
fractional frequency whirl problems encountered in multistage compressors supported by the 5 pad
bearing configuration has caused hundreds of millions of dollars in costs from redesigns and losses
in production.

2
Jeffcott Rotor on Rigid Supports
It first will be examined the influence of internal friction on the simple Jeffcott rotor as shown in Fig. 1.
The Jeffcott rotor has a single disk of approximately 100 lbs and is mounted on a flexible uniform shaft
with a length of 40 in and a diameter of 4 in. The rotor is constrained at the ends with simple supports that
provide no damping.

Fig 1 Jeffcott Rotor on Simple Supports

The shaft stiffness is given by:

(1)

The natural frequency of the Jeffcott rotor is given as follows:

(2)

It is of interest to note that a typical 8 stage compressor may be reduced to a Jeffcott model for the purposes
of stability analysis by using the modal mass and stiffness generated by a critical speed analysis based on
rigid bearing supports. This helps to quickly determine if a particular bearing design may be suitable.

Fig 2 First Critical Speed Mode Shape of Jeffcott Rotor at 9607 RPM

3
Figure 2 represents the first critical speed mode shape of the Jeffcott rotor. Note that the motion is
constrained at the bearings. It is interesting to note that most compressor manufacturers place monitoring
probes at the bearings to monitor the motion. For the case of very stiff bearings, there will be very little
motion observed at the bearing location while the center span is in large amplitude whirling. This can lead
to rotor failure without the probes shutting down the system.
Models of Internal Rotor Friction
There are several models of internal friction that have been used in analysis. The first type of internal
friction Model is referred to as hysteretic damping and is give as follows:

(3)

In this model the internal hysteretic damping is independent of frequency. The value of alpha is small in
comparison to unity. This effect generates a force perpendicular to the displacement vector.
The model of internal friction that will be employed is the viscous friction model as shown below.

(4)

In this model, the internal friction is expressed as a function of the relative velocity in a rotating coordinate
system. When expressed in terms of Fx and Fy in a fixed coordinate system, as shown in Eq 5, it is seen
that the relative damping term generates cross coupling terms of opposite sign in the x and y directions.
In general, it is the presence of cross coupling components of opposite sign that generates instability in rotor
bearing systems.

(5)

Planar Motion With Internal Friction

If planar motion is assumed with zero shaft rotation, then the equations of motion appear similar to the
single mass system with gravity excitation as shown in Eq 6.

(6)

A value of internal damping of Ci equal to 5 Lb-sec/in was chosen. This value represents an amplification
factor of Ac=50. For light damping the amplification factor is related to the log decrement as follows:

(7)

The value of the log decrement in this case is 0.0628. A time transient analysis of the nonrotating system
under the influence of gravity is shown in Fig. 3. The log decrement is formally defined as the logarithmic
ratio of successive amplitudes of motion. In Fig. 3 the maximum deflection of 0.774 mils is twice the static
under gravitational loading. The natural frequency may be estimated from the static deflection by the
equation Ncr =60/(2pi)xù = 9.55 sqr(g/ä) = 9539 CPM.

4
 Fig. 3 Transient Motion of Jeffcott Rotor at Zero Speed
With Gravity Excitation

Stability of Jeffcott Rotor With Internal Friction on Rigid Supports

The stability analysis over a speed range will be computed by the finite element program DYROBES which
allows for the computation of the system damped eigenvalues and time transient motion. When the earlier
studies of this stability mechanism were explored, only the threshold of stability was computed ( zero log
decrement) using a state of the art UNIVAC digital. The linear and nonlinear time transient motion studies
were computed at that time using an analog super computer ( which is no longer in existence). Similar and
more advanced models may now be computed with finite element techniques and PC computers which have
replaced the “super computers” of yesteryear.

Table 1 Internal Friction Cross Coupling Stiffness and Damping Coefficients

Acting At Rotor Center Over a Speed Range

5
Table 1 represents the stiffness and damping coefficients for internal friction acting at the rotor center.
Over a speed range from 0 to 20,000 rpm. Note that as speed increases, the cross coupling coefficients
increase as give by Eq. 5.

Fig. 4 Log Decrement of Jeffcott Rotor With Internal Friction

With Speed Range Of 0 To 20,000 RPM

Figure 4 represents. the log decrement vs speed for the Jeffcott rotor. There are two modes shown in the
figure. The upper mode is the backward mode. As speed increases, the backward mode log dec increases.
Thus internal friction can not cause the system to whirl backwards. The log dec for the forward mode
continues to reduce with speed until the critical speed of 9550 RPM is reached. At this point the log dec
is zero.. This speed is referred to as the threshold of stability speed. Thus we see that for the classical
Jeffcott rotor on fixed supports, the system becomes unstable at speeds above the critical speed.

Fig. 5 Whirl Motion At 10,000 RPM Fig. 6 Whirl Motion At 20,000 RPM

6
Stability of Jeffcott Rotor With Internal Friction on Undamped Symmetric Supports
Flexible supports will now be added to the Jeffcott rotor as shown below in Fig. 7.. A bearing stiffness of
Kxx=Kyy = 125,000 RPM will now be added to the model.

Fig. 7 Jeffcott Rotor With Flexible Supports Fig. 8 Critical Speed With Kb=125,000 Lb/in
Nc = 6,737 RPM

The incorporation of finite bearing support stiffness of 125,000 Lb/In causes a reduction in the critical to
Nc = 6,737 RPM. The introduction of a symmetrical support flexibility will not only lower the critical
speed, but also cause a reduction in the onset of stability speed. This is the general characteristic of self
excited whirling whether caused by fluid film bearings or aerodynamic cross coupling and internal friction.
The incorporation of undamped symmetric supports causes a reduction in the critical speeds and lowers the
stability threshold. However, when damping is also included, such as a squeeze film damper, the stability
may be greatly improved.

Fig. 9 Log Decrement of Jeffcott Rotor With Internal Friction

With Kb=125,000 Lb/In For Speed Range of 0 To 20,000 RPM
Figure 9 represents the forward and backward log decrements of the Jeffcott rotor with internal friction with
the addition of the flexible bearing supports. In the absence of bearing damping, the forward mode
becomes unstable at a lower speed due to the reduction of the critical speed.

7
Stability of Jeffcott Rotor With Internal Friction on Damped Symmetric Supports
Figure 10 represents the Jeffcott model with damped symmetric supports. Figure 11 represents the stability
map generated for various values of bearing damping. Of interest is the base curve that shows that if we
simply incorporate simple flexible supports without damping,
the critical speed is reduced and also the threshold of stability.
The analysis shows that once the rotor speed exceeds the
undamped critical speed, the system becomes unstable under the
influence of internal friction.

Fig. 11 Stability Map With Symmetric

Fig. 10 Model of Jeffcott Rotor With Symmetric
Damped Support (Gunter)
Damped Supports ( Gunter)
In Fig. 11 is shown the influence of support damping D ratio and the bearing stiffness ratio R in changing
the stability threshold. The simplified stability criterion as shown in Fig . 11 may be expressed as:

(8)

Figure 11 shows that as the support ratio R becomes larger then unity ( softer support conditions), the
stability increases with an increase of bearing or support damping. It should be noted that Eq. 8 is an
approximate equation based on small damping values. It will be shown that there is a limit to the amount
of effective damping that can be incorporated into the flexible supports to improve stability. As the
damping becomes excessive, a point is approached which is referred to as bearing or damper lockup. This
occurs with squeeze film dampers with tight clearances and high loading conditions. With excessive
support damping, the log decrement no longer continues to increase, but then begins to decrease. When
this begins to occur we have passed the pont of optimum damping.
Figure 12 shows the mode shape and log decrement with zero support damping at a speed of 20,000 RPM.
The system is unstable with a log decrement of -0.1523. The whirl frequency corresponds to the undamped
natural frequency of 6,743 CPM. The motion at the bearings is in phase to the motion of the center disk.

Figure 13 represents the system first mode when the bearing or support damping is increased to Cb = Cxx
= Cyy = 200 Lb-Sec/in. The system is now stable with a log decrement of 1.1868. The support damping
8
has also increased the damped natural frequency from the undamped value of 6,743 CPM to 7,681 CPM.
Note that the mode shape is no longer planar. The bearing motion is now lagging the shaft centerline disk
motion by almost 45 deg!

Figure 14 shows the mode shape and log dec with the damping further increased to 500 lb-sec/in. In this

Fig. 12 Unstable Whirl at 20,000 Rpm at Nf = 6,743 Fig. 13 Stable Whirl at 20,000 Rpm at Nf = 7,681
CPM With Cb = 0, Log Dec = -0.1523 CPM With Cb = 200, Log Dec = 1.186
case, the log dec has been reduced to -.687 and the
whirl frequency has increased to 9,121 CPM. The bearing motion is now lagging the disk motion by about
75 deg. It is obvious that the shaft mode shape is no longer planar. Figure 15 represents the extreme case
in which the damping has been increased to the value of Cb=5,000 lb-sec/in. This case represents “damper
lockup” in which the bearings approach pinned supports and the log dec approaches zero. Any further
increase in damping will cause the rotor to go unstable! Hence we see that Eq 8 is only an approximation
and accurate only for small bearing lag angles.
There is an important conclusion that can be gained from the behavior of this “simple Jeffcott rotor”. Many
large finite element programs and a number of rotor dynamic codes use component modal synthesis for
dynamic analysis of aircraft engines and turbochargers with nonlinear bearings. This method is based on
planar modes and does not have suitable accuracy to correctly compute highly nonlinear bearings!

Fig. 14 Stable Whirl at 20,000 Rpm at Nf = 9121 Fig. 15 Stable Whirl at 20,000 Rpm at Nf = 9521
CPM With Cb = 500, Log Dec = 0.687 CPM With Cb = 5000, Log Dec = 0.0094

9
Stability of Jeffcott Rotor With Internal Friction on Asymmetric Supports
Figure 16 represents the single mass Jeffcott rotor on damped flexible supports (1). The bearing
stiffnesses in the horizontal and vertical Kx and Ky
are identical for both bearings. It is assumed that
only cylindrical motion occurs and the operating
speeds are well below the conical mode. The
equations of motion are represented by 4 degrees of
freedom.
If bearing or support mass is included in the model,
then the system characteristic equation for stability
analysis is 8th order. However, if bearing mass is
neglected, then the characteristic equation is
reduced to a 6th order equation. A further reduction
in the characteristic equation is achieved assuming
low values of internal friction. This reduces the
system stability equation to a 4th order system.
While this is a reasonable assumption for the study
of internal friction on rotor stability, it will be
shown that this is not an accurate assumption for the Fig. 16 Jeffcott Rotor on Asymmetric Supports
case of aerodynamic cross coupling forces acting on
the rotor.
Figures 17 and 18 represent the relative stability of the Jeffcott rotor for various ratios of vertical
flexibility R as a function of the bearing asymmetry ratio á. As the vertical support flexibility is
increased, the stability is reduced with symmetrical bearings.

Fig. 18 Stability For Various Values of R With

Fig. 17 Rotor Stability With Asymmetric
Asymmetric Supports and Zero Brg Damping
Supports and Zero Bearing Damping

10
Figure 19 represents a 3 dimensional model of the stability characteristics of the rotor with increasing
flexibility R in the vertical direction for various values of bearing asymmetry á = Kx/Ky. As shown
before, if we simply make a uniform flexible support with no included damping, the system critical
speed is reduced along with the rotor stability threshold. Therefore simple support flexibility
reduces the stability threshold. Figure 20 shows the important feature that as the value of internal
friction increases ( smaller A values) then the influence of bearing asymmetry is reduced!

Fig. 20 Influence of Internal Friction Value on

Fig. 19 3 Dimensional Stability M odel W ith Stability For Various Values of R W ith
Asymmetric Supports and Zero Brg Damping Asymmetric Supports and Zero Brg Damping

Fig. 21 Stability Plot W ith Bearing Asymmetry and

Zero Bearing Damping

Figure 21 represents a plot of the system log decrement with R = Ky/Kx = 2, Ci = 50 lb-s/in and
zero bearing damping. In this case, the stability threshold has been increased to over 10,800 RPM.
This is a dramatic improvement over the original reduced stability threshold caused by incorporating
the symmetric flexible support.

11
 Discussion of The Influence of Bearing Asymmetry on Rotor Stability Improvement With
Internal Friction
When bearing asymmetry is incorporated into the system, then two planar modes are created that are
different in the horizontal and vertical directions. Figure 22, for example, represents the horizontal
mode corresponding to the horizontal bearing stiffness value of 1.25e6 lb/in with a natural frequency
of 6,753 CPM. Figure 23 represents the vertical natural frequency of 7,715 CPM corresponding to the
higher vertical bearing stiffness of 2.5e5 lb/in.

Fig. 22 Horizontal M ode At Zero Rpm Fig. 23 Vertical M ode At Zero Rpm
Kxx=1.25e5 Lb/In, 6753 CPM Kyy=2.50e5 Lb/In, 7715 CPM

Figure 24 represents the damped mode shape of the rotor at 20,000 RPM with the above specified bearing
stiffness coefficients. The rotor is stable with a log decrement of 0.232. This figure should be contrasted
to Fig 25 which shows a highly unstable mode at 20,000 RPM. In this case the bearing stiffnesses in both
the horizontal and vertical directions are identical.
Dr. Bert Newkirk, in 1924 was not able to explain why this phenomena occurred. In my thesis presentation
in 1965, I was unable to articulate exactly why this was possible. However, it was pointed out at that time
that the influence of bearing asymmetry was the greatest when the value of internal friction was small. For
larger values of internal friction, the influence of bearing asymmetry diminishes rapidly as shown in Fig.
20. The reason for this effect was first explained clearly to me by Dr. Wen Jeng Chen of Eigen Technology.
( It is of interest to note is that all figures presented in this paper use the DyRoBeS rotor dynamics software
developed by Dr. Chen ). The reason that bearing asymmetry improves stability with internal friction is that
it creates elliptical mode shapes as shown in Fig. 24. This minimizes the value of internal friction developed
in the rotor. As an example, if the rotor is vibrated in a plane, the internal friction simply acts as
conventional damping. In Fig. 25, the mode shape is circular, maximizing the effect of internal friction.

Fig. 24 Stable M ode At 20,000 RPM W ith

Fig. 25 Unstable M ode At 20,000 RPM W ith
Bearing Asymmetry , Ci=5 Lb-S/In
Symmetrical Bearings , Ci=5 Lb-S/In

12
Jeffcott Rotor With Aerodynamic Cross Coupling on Rigid Supports
Another mechanism that can cause self excited whirl motion is referred to as the Alford effect(4). This
effect can be generated by aerodynamic forces in both turbines and compressors. This mechanism
generates cross coupling terms similar to the cross coupling terms in a fluid film bearing. As Alford
first described the effect for turbines, the radial displacement of the turbine wheel causes a local change
in the wheel efficiency. This local change in efficiency generates forces to act perpendicular to the
deflected centerline of the shaft. For small displacements, this effect may be represented as two similar
out of phase bearing forces acting at the disk center. Therefore for the Jeffcott rotor on rigid or stiff
supports, the two governing equations of motion would be as follows:

(9)

In the analysis of stability of the Jeffcott rotor under the influence of aerodynamic cross coupling, it will
be assumed that there is some positive damping acting at the rotor center. This damping term is shown
as C2. This small value of damping is assumed to be 5 lb-s/in and will cause an amplification factor of
25 at the critical speed.

Fig. 26 W hirl M otion W ith Aerodynamic Q Fig. 27 Bearing Forces W ith Aero Q at 20,000 RPM
at 20,000 RPM - Stiff Brg Supports Stiff Brg Supports - Kbrg = 1.0e7 Lb/In

Figure 26 represents the whirl orbits of the Jeffcott rotor at 20,000 RPM for 20 cycle of shaft motion
with an assume aerodynamic cross coupling value of 10,000 Lb/In and center plane damping of 5 Lb-
s/In. In addition, a rotating unbalance of 1.6 oz-in has been added. This value was chosen to displace
the rotor center approximately 1 mil. The amplitudes shown if Fig. 26 represent mils of motion. If the
rotor were stable, then the orbit radius at this speed above the critical speed would be 1 mil. Since there
are no nonlinear shaft forces acting on the rotor, there will be no limit cycle achieved. Hence the
motion will continue to increase.
Figure 27 represents the bearing forces transmitted at 20,000 RPM with the assigned aerodynamic cross
coupling of 10,000 Lb/In and the stiff bearing support values of 1.0e7 Lb/in. If perfectly rigid bearings
are assumed, then the bearing forces can not be computed! It is seen that the bearing forces continue
to increase in time and will become unlimited for the assumed linear system. The frequency of the
linear growth takes place at the system natural frequency.

13
Jeffcott Rotor With Aerodynamic Cross Coupling on Flexible Supports
Figure 28 represents the Jeffcott rotor whirl orbit in which the bearing or support system is reduced in
stiffness to a value of Kbrg =2.5e5 Lb/In. In this model, the bearing stiffness values are identical in both
the horizontal and vertical directions. The reduction in bearing stiffness has caused the rotor obit to
increase greatly over the orbit as shown in Fig. 26. In this model, a small amount of nonlinear shaft
stiffness was added to develop a limit cycle of motion.
Figure 29 represents the bearing forces transmitted with the reduced bearing support system. In a linear
system, the forces would continue to increase in time as shown in Fig. 27. A small nonlinear shaft
stiffness component was added to create limit cycle motion and constrained bearing forces as shown in
Fig. 29. Figure 29 shows the bearing forces transmitted vs time. Without the nonlinear shaft stiffness
addition, the bearing forces would become unlimited with time. The nonlinear response is apparent after
0.05 seconds of motion. The beating motion occurs due to the interaction of the unbalance with the
nonlinear shaft response.

Fig. 28 W hirl W ith Aero Q at 20,000 Fig. 29 Bearing Forces W ith Aero Q at 20,000 RPM W ith
RPM ,-Kxx=Kyy=2.5e5 Lb/In Symmetric Bearing Supports - Kbrg = 2.5e5 Lb/In

Fig. 30 W hirl W ith Asymmetric Bearings Fig. 31 Bearing Forces W ith Aero Q at 20,000 RPM With
Kxx = 1.25e5, Kyy = 2.5e5 Lb/In Asymmetric Bearing Supports - Kxx = 1.25e5 Lb/In

Figure 30 shows the whirl orbit with the horizontal stiffness reduced in half to 1.25e5 Lb/In. The rotor
system has been stabilized by the asymmetric support. In Fig. 26, with the symmetric support, the
maximum Q value is 8,500 Lb/In. With the asymmetric support, the maximum Q value has increased
by a factor of 4 to 35,000 Lb/In! Figure 31 shows the bearing forces with the asymmetric support.

14
Stability of Multistage Compressors in Tilting Pad Bearings With Aerodynamic Cross Coupling
In the early 1950's, compressor design was limited in speed and number of stages, usually 8 or less. The
reason for this was the occurrence of self excited whirling with fixed fluid film bearings. At that stage of
development, there did not exist the computer tools for computation of fluid film bearing characteristics,
rotor damped eigenvalues (stability) or instrumentation to measure vibrations and determine frequency
content. The concept of tilting pad thrust and radial bearings, however, had been in existence for many
years. The 5 tilting pad bearing seemed a natural advancement for centrifugal compressor applications
because of its superior stability characteristics over fixed plain or elliptic bearings currently in use.
Lund, in 1964 (who many of us consider to be the father of modern rotor dynamics), presented his classical
paper on “Stiffness and Damping Coefficients for the Tilting Pad Journal Bearing”. At that time, however,
there did not exist the software to compute rotor stability. It was 10 years later that Lund published his
article on rotor stability based on the transfer matrix approach. At this same time, another significant
development was the introduction of the Bently noncontact induction probes to monitor vibrations. The
ability to perform an FFT analysis of the motion was a later development. Thanks to the cold war, the
NAVY spent significant funds on FFT analysis and instrument development. While working on the Space
Shuttle engines, I received a major grant from HP for their latest analyzer. It came in three components.
The reason was that it have to be able to fit through the front hatch of a nuclear attack submarine!
Compressor design in the 1970's progressed rapidly with designs over 10 stages and speeds in exceeding
10,000 RPM. The instability mechanism that Alford reported in 1965 now began to effect these new high
speed, high pressure compressor designs with self excited subsynchronous whirling. A major problem was
that this behavior was not detected on the test stands as few full high pressure tests were conduced then.
This was a particularly difficult problem when the subject compressor was located on an off shore platform
in the North Sea. Many of these compressor problems were corrected by incorporating squeeze film
dampers with the tilting pad bearings. We will show that, in many cases, resorting to a squeeze dampers
is not necessary. An alternate design using a 4 pad bearing configuration may be sufficient.

Fig. 32 Ten Stage Centrifugal Compressor With Tilting Pad Bearings

Figure 32 represents the 10 stage compressor mounted on 5 pad tilting pad bearings. The 5 pad bearing
configuration is very typical for multistage compressors operating above 7,500 RPM. The design is based
on the assumption that this bearing will be a superior bearing for resisting instabilities because of its
asymmetric bearing coefficients and lack of bearing cross coupling terms (neglecting pad inertia effects).
Also shown in the figure is the assumption of the aerodynamic cross coupling as acting at the rotor center.
The actual aerodynamic cross coupling coefficients act at all compressor stations. In the following analysis,
the total effective Q will be assumed acting at the rotor center. This has shown to be a reasonable
assumption. Figure 32 represents a system with 19 nodes and 76 degrees of freedom. To determine the
time transient motion of this system, it will be necessary to integrate the 76 degrees of freedom several
hundred times for each cycle of motion. The characteristic equation is of order 152 for damped roots.

15
 Fig. 33 Critical Speed Map of 10 Stage Compressor

Figure 33 represents the undamped critical speed map of the compressor for various values of bearing
stiffness. Reviewing only the undamped critical speed map, for example, can lead to some dangerous
conclusions and disastrous designs. For example, viewing the critical speed map, we see that for a nominal
bearing stiffness value of 1.0e6 Lb/In, the compressor will be operating on top of the critical speed. This
is not a problem! A far bigger problem would be to increase the bearing stiffness above 1.0e6 Lb/In in
order to elevate the second critical speed above running speed. This, in effect pinched the bearings, causing
small deflections and reducing effective modal damping for the first mode. The rotor system now is highly
sensitive to self excited whirl forces. This is a good example of where a little bit of knowledge can be a
dangerous thing!

Fig. 34 5 Pad Bearing, Load on Pad Fig. 35 Characteristics of 5 Pad

Cb = 5 Mils, Plf=0.5 Bearing, Cb = 6 Mils

Figure 34 represents the bearing pressure profile for the 5 pad bearing with load on pad. Figure 35
represents the bearing parameters and shows the stiffness and damping coefficients for the bearing.

16
Stability of 10 Stage Compressor in 5 Pad Bearings With Aerodynamic Cross Coupling - Cb = 6 Mils
Figure 36 represents the first forward mode shape (eigenvector) at 11,000 RPM with the specified
aerodynamic cross coupling. The whirl frequency is low and is less then 1/3 of running speed. The log
decrement for this mode is -0.99 which make this system highly unstable.

Fig. 36 1st Forward M ode W ith 5 Pad Brg & 6 M il Cb

Unstable M ode - LD = -0.99

Fig. 37 2 nd Forward M ode at 7214 CPM W ith 5 Pad Brg

& 6 M il Cb Stable M ode LD = 1.34

Figure 37 represents the 2nd critical speed. This mode is not of concern and will not be excited by
unbalance since the log decrement is 1.34. The large amplitude at the bearings creates significant damping.
Figure 38 shows the 3rd mode near running speed at 11,078 RPM. This is not a problem for unbalance
response.

Fig. 38 3rd Forward M ode at 11078 CPM W ith 5 Pad Brg

& 6 M il Cb Stable M ode Log Dec = 1.26

17
Stability of 10 Stage Compressor in 5 Pad Bearings With Aerodynamic Cross Coupling - Cb = 4 Mils
In an attempt to improve stability, the bearing clearance has been reduced from 6 mils to 4 mils. Figure
39 shows the resulting bearing pressure as a result of the reduced bearing clearance. Figure 40
represents the bearing parameters and bearing coefficients generated by the reduction in bearing
clearance. The stiffness in the x direction has almost tripled while the vertical stiffness has increased
by 40%. Similar increases in damping are also observed.

Fig. 39 5 Pad Brg W ith Load on Pad , Cb Fig. 40 Characteristics of 5 Pad Brg W ith
= 4 M ils L/D =0.666, Preload = 0.5 Cb = 4 M ils L/D =0.666 Plf = 0.5

Fig. 41 Compressor 1st Forward M ode W ith Reduced Clearance 5 Pad Bearing
L/D=0.666, PLF = 0.5, m = 0.5, Q = 50,000 Lb/In, Log Dec = -0.303

Figure 41 represents the first damped forward mode with the tighter clearance 5 pad bearing. With an
applied aerodynamic cross coupling of Q = 50,000 Lb/ In, the mode is highly unstable with a log
decrement of -0.30. Note that the motion with the stiffer bearing is now more symmetrical. The
increased stiffness has also increased the whirl to 3,230 CPM.

18
Stabilization of a 10 Stage Compressor With a 4 Pad Bearing Design
Once the aerodynamic cross coupling exceeds approximately a Q value of 20,000 Lb/in, it is difficult
to stabilize the system with a 5 pad bearing regardless of clearance values and preloads. The bearing
stiffness and damping values are not independent. Thus to stabilize this class of compressor, it has been
necessary to resort to mounting the 5 pad bearings in squeeze film dampers. This becomes a
complicated design and also may not be adequate for compressors exceeding 10 stages.
For example, one design the author worked on at Fort McMurry (tar sand region of Canada) was an
unstable compressor with 5 pad bearings and squeeze film dampers! The system was still unstable.
The final solution required redesigning 2 internal seals as squeeze film dampers (subject for a future
paper). A later review of this compressor using modern analysis techniques showed that this 8 stage
rotor could easily have been stabilized by the proper design of 4 pad bearings. However, at that time we
did not know what the proper tilting pad bearing coefficients that should be designed for. At that time
and also currently today, the 5 pad bearing configuration appears to be the preferred approach by the
various manufactures. This design approach has been a bonanza for the consultants over the last several
decades for correcting stability problems.
Shaft Modal Stiffness, Kshaft
Before we can properly design a tilting pad bearing for the multistage compressor, it is first necessary
to determine a fundamental property of the system. Without the knowledge of this one parameter, we
could endlessly be computing various 5 or 4 pad bearings without arriving at a satisfactory bearing
design. The basic parameter that is required in order to properly design an effective 4 pad bearing is the
knowledge of the shaft modal stiffness. This value can be computed from the critical speed analysis.
The modal mass corresponding to the rigid support critical speed is computed. If this value is not
available, then a value of 55% of the total rotor weight is of sufficient accuracy. Note that the modal
weight of a uniform shaft is exactly ½ the shaft weight.

Fig. 42 Compressor Critical Speed on Rigid Supports

Ncr = 3,271 RPM

Figure 42 represents the compressor critical speed on rigid supports. This frequency is computed at
3,271 RPM. Note that this value could also have been obtained from the critical speed map as shown
in Fig.33. The modal weight was computed as 710 Lb which is 54% of the total rotor weight. The
fundamental shaft stiffness may now be computed as follows:

(10)

19
Design Parameter for 4 Pad Bearing
We have essentially reduced the 10 stage compressor to a Jeffcott rotor! The optimum bearing to shaft
stiffness ratio has been shown to Roptimum = 2Kbrg/Kshaft =1. This design parameter is next to impossible
to achieve with a 5 pad bearing with load on pad. For example, with Fig. 39 with the 5 pad bearing and
load on pad, the R ratio is 2.7. Thus when the values of R exceed unity, there is no amount of bearing
damping that will make a highly stable bearing configuration.
With this bearing to shaft stiffness parameter in mind, we now have a design goal to attempt to achieve.
As previously mentioned, it will not be possible to design any 5 pad bearing to meet these guidelines.
With this in mind, we now examine the 4 pad bearing. What we find out that the 4 pad bearing initially
does little to improve the stability. The stiffness is still too high with the l/d ratio 0f 0.666. We now have
the paradox that to improve stability, we actually have to use a larger bearing with less preload!
Figure 43 represents a 4 pad load between pad bearing design with an enlarged L/D ratio to 1 and a
preload reduced to m=0.2. These changes are, at first, counter intuitive. The reason for the larger
clearance is to “float the shaft” in order to obtain a lower bearing stiffness.

Fig. 43 4 Pad Bearing With L/D=1.0, Preload = 0.2, Cb = 6 Mils

The R ratio for this design is 1.35 which is close to optimum. If the shorter L/D ratio is used, than the
R ratio will be over 2. Thus we are using a larger bearing, not for increased load capacity, but to obtain
a softer bearing support. Fig. 44 shows a stable whirl mode with the enlarged 4 pad bearing.

Fig. 44 Stable M ode at Threshold of Stability With 4 Pad Brg

L/D =1, Cb=6 M ils, Q=60,000 Lb/In, Log Dec =0.096

20
Time Transient Behavior With 5 and 4 Pad Bearing Configurations
Time transient analysis was performed on both the 5 and 4 pad bearing designs to show the difference
in amplitudes of motion and bearing forces transmitted. In both cases, aerodynamic cross coupling of
Q=50,000 Lb/In is included at the mass center as well as an unbalance of 2 oz-in. The rotor system has
a total of 76 degrees of freedom. The motion ids evaluated over 10 cycles of motion. A time step of
1.0e-6 sec was used which requires over 5,000 time steps for each cycle of motion for each degree of
freedom.
Figure 45 shows the unstable transient motion of the rotor with the 5 pad bearings. Figure 46 shows
the motion at the rotor center for 20 cycles of motion. The rotor develops large amplitudes of whirl.

Fig. 45 Transient M otion W ith 5 Pad Brg & Unbalance Fig 46 Center Span M otion W ith
5 Pad Brg For 20 Cycles , Mils
Fig. 47 shows the whirl motion at the 5 pad bearing. Although the initial amplitude is small, it
continues to grow. Shown on both Figures 46 and 47 are timing marks to represent each shaft rotation.
If the motion was totally synchronous, then the timing marks would align.
Of great importance are the large forces over 400 Lb developed after 20 cycles of motion as shown in
Figure 48. The influence of unbalance represents a synchronous response, but is overwhelmed by the
large forces generated due to the occurrence of self excited whirling.

Fig. 47 Bearing M otion W ith Fig. 48 Bearing Forces W ith 5 Pad Brg For 20 Cycles
5 Pad Brg For 20 Cycles, M ils

21
Figure 49 shows the time transient motion with the 4 pad bearing configuration for 20 cycles of shaft
motion. The initial transient motion is caused by the excitation of the first forward mode caused by
unbalance. After about 5 cycles of motion, the initial transient begins to die out, since this is a stable
system. Figure 50 shows the initial motion at the shaft center. The initial maximum excursion is only
0.7 mils caused by the application of the 2 oz-in of unbalance placed at the rotor center.

Fig. 50 Center Span M otion W ith

Fig. 49 Transient M otion W ith 4 Pad Brg & Unbalance
4 Pad Brg For 20 Cycles, M ils

Fig. 51 Bearing M otion W ith

4 Pad Brg For 20 Cycles, M ils Fig. 52 Bearing Forces W ith 5 Pad Brg For 37 Cycles, Lbs

Figure 51 shows the orbital motion at the bearing location for the case of the 4 pad bearing. We can
clearly see the transient motion diminishing after only 5 cycles of motion. If we were to continue the
transient calculations for another 20 cycles of motion, we would have only synchronous circular motion
caused by the applied unbalance.
Figure 52 shows the bearing forces transmitted over 37 cycles of motion for a time period of 0.2 seconds.
The initial oscillations in the bearing forces transmitted is caused by the excitation of the first critical
speed. After about 20 cycles of motion, the forces transmitted are predominately synchronous forces
caused by the rotating unbalance of 2 oz-in. At this speed, the 2 oz-in of unbalance generates a rotating
force of over 470 Lbs. Note that the steady state forces reduce to +1 60 Lbs at each bearing. This case
shows the dramatic improvement of the rotor response with 4 pads as compared to the 5 pad design.

22
 Industrial Application - Kabob of Alberta
The Kabob case history is a perfect example of fractional frequency whirl occurring in a high
pressure compressor. The Kabob whirl problem was first encountered in 1971and the history and
design changes were presented in some details at the Fourth Annual Texas Turbomachinery
Conference in 1976. This case was of particular interest as two of the references for the motivation
for the bearing redesigns were based on several of my publications on the influence of bearing
asymmetry on stability with internal friction.
The original Kabob rotor is shown in Fig. 53. The bearing span for this system is shown in the figure
as 59.68 inches. Compare this design with the 10 stage compressor as shown in Fig. 32 which has
a bearing span of 64.3 inches. It was reported that in order to have a rotor with controlled whirl
motion, that a final rotor design was required with a reduced length of 53.43 inches and increased
shaft diameter. Based on modern rotor-bearing dynamical analysis, this compressor could have been
stabilized by either a properly designed 4 pad bearing or with a squeeze film damper and the 5 pad
bearing configuration.
However, attempting to stabilize this rotor using the asymmetrical properties of a 5 pad bearing will
not work with the original bearing aspect ratio. Proper bearing design could have saved a reported
loss of $25 million ( in 1971 dollars!). Although Lund had earlier presented a major paper on the
5 pad bearing, his paper on rotor -bearing stability was not presented until 1974. Therefore, the
major compressor manufactures did not have the computer tools for extensive rotor-bearing stability
studies. Kabob has demonstrated that trial and error bearing replacements can be very expensive.

Fig. 53 Original Kabob 9 Stage Compressor Design

Redesigned 5 Pad Bearing For Exaggerated Bearing Asymmetry

Fig. 54 Redesigned Kabob 5 Pad Brg

to Increase Bearing Asymmetry Fig. 55 Computed Brg Coefficients For
Redesigned 5 Pad Bearing

Figure 54 shows the redesigned 5 pad bearing to increase the apparent amount of bearing asymmetry.
Note that the side pads , labeled B, have a reduced bearing width from 1.625 to 1.00 inches. Also the
bearing pad clearance has been increased from 7 mils to 10 mils. Figure 55 shows a computer
simulation of the bearing redesign with variable pad length and pad clearance values.

23
Note that the stiffness and damping values are almost twice as high in the vertical direction as the
horizontal direction. The upper 3 pads contribute very little to the overall stiffness and damping values
for this bearing.
Compressor Stability With Redesigned 5 Pad Bearings
The compressor critical speed on rigid supports for the original rotor design was given as 3,915 RPM.
Although the modal mass of the unit was not specified, we can estimate the effective modal shaft
stiffness to be of the order of 250,000 Lb/In. From Fig. 55, we see that the vertical bearing stiffness
Kyy is 438,000 Lb/In. This places the critical bearing to shaft R ratio (2Kyy/Ks) to be over 3!
Therefore, we can not expect this bearing design to be effective in rotor stabilization. Figure 56 shows
the computed stability corresponding to the bearing coefficients as give in Fig. 55 and a representative
rotor model. The log decrement of -0.405 indicates that this bearing design will be highly unstable.

Fig. 56 Computed 1st Forward Whirl Mode For 9 Stage Kabob Rotor in
5 Pad Bearings - Unstable Mode
Six Pad Narrow Bearing Design With Included Squeeze Film Damper

Fig. 57 Narrow 6 Pad Bearing in 6 M il

Fig. 58 Pressure Profile & Brg Coefficients For
Squeeze Film Damper
Narrow 6 Pad Bearing

Figure 57 shows a 6 pad bearing design with narrow pads. This bearing has an R ratio of 2.6 which is
not a major improvement over the original design. The bearing was mounted in a squeeze film damper.
This damper is flawed as it does not have a central groove! This caused the damper to act as a long
bearing damper. The central groove makes a short bearing damper with adequate damping values. A
much larger clearance must be used with the long bearing damper. The final design error was to
decrease the clearance for the long bearing damper. Thus this damper was ineffective. All of these
problems required a redesigned shorter and stiffer shaft. Proper bearing design could have avoided this.

24
Industrial Application - 11 Stage Compressor in 5 Pad Bearings
Figure 59 shows an 11 stage compressor that initially encountered a sever whirl problem at full load
conditions. At the time this rotor was analyzed, both finite element bearing and stability programs (
based on the Lund transfer matrix method) were available for detailed stability analysis. The system
was analyzed for various conditions of bearing preload and conditions of load on pad and between pad
loading conditions for a 5 pad bearing.

Fig. 59 11 Stage Compressor in 5 Pad Bearings Bearing Span = 70 In, Shaft D = 6 In

( Nicholas, Gunter, Barrett, 1978 )

The 11 stage compressor has a bearing span of 70 inches. The nominal shaft diameter is 6 inches. The
computed effective shaft stiffness is Kshaft = 280,000 Lb/In. Figure 60 shows the relative stability of the
11 stage compressor for a variety of loading conditions with the 5 pad bearings. The highest stability
is achieved by bearing types 1 and 6. Bearing no. 1 is a load on pad design while bearing no. 6 is a load
between pads configuration. In both of these cases, the bearings have zero preload. The zero preload
condition leads to softer vertical stiffness values. The worst two bearing designs are similar to bearings
1 and 6 except that the bearing preload has been increased to 0.5! This causes a large increase in the
vertical bearing stiffnesses causing a substantial reduction in stability.

Fig. 60 Stability M ap for 11 Stage Compressor in 5 Pad Bearings

For Various Values of Preload and Load On and Between Pads

Bearing number 1, which is a load between pad design, has a zero preload. This bearing has a vertical
stiffness ratio of 2Kyy/Kshaft of 2.79. The optimum value is 1. The bearing number 6 is also a zero

25
preload design with load on pad design. The R ratio for this design is slightly higher at 3.50. These two
designs represent the best of the ones analyzed. The worst designs are bearings 5 and 8 which are also
load between and load on pad designs, but with the bearing preload increased to 0.5 in both cases. This
greatly increases the bearing vertical stiffness values.
These two designs have R ratios of 13.4 and 21. These designs both will allow violent self excited whirl
instability to occur. Bearing no. 4 is also very poor in spite of the lower preload value of 0.3. The reason
for this is that the design uses an offset pivot of 55% of arc length. This generates greater hydrodynamic
bearing pressures and hence excessive bearing stiffnesses. Increasing bearing preload or offset ratio in
an attempt to raise stability by increasing bearing stiffness and damping can cause catastrophic failure
to occur when the value of twice the vertical bearing stiffness exceeds an order of magnitude above the
shaft stiffness. At high R values, bearing asymmetry is of no benefit in stability improvement.

Fig. 61 Critical Speed M ap for 11 Stage Compressor on 5 Pad Bearings

For Various Values of Preload and Load On and Between Pads

Figure 61 shows the critical speed map for the compressor for the various bearing configurations.
Bearing types 5 and 8 are often chosen based on the consideration that the compressor will be operating
below the second critical speed. This type of design consideration , based on API recommendations, is
a recipe for a disaster. It is of interest to note that bearings no. 1 and 6 show that the vertical bearing
stiffness values actually decrease with speed. This is due to the zero preload case in which the top pad
is not adding additional loading.
As the shaft raises in the bearing with speed, the bearing stiffness becomes softer. Compare this
behavior with the preloaded bearings. At speeds above 4,000 RPM, the preloaded top pad can cause a
dramatic increase in the net vertical stiffness values. Hence as speed increases, these bearing designs
become more sensitive to self excited whirl forces. This dangerous condition can be observed from the
critical speed map. If the bearing stiffness at running speed projects down into the indicated danger
region, then this bearing type must be redesigned with either a lower preload or a wider 4 pad design.

26
Discussion and Conclusions
The use of the 5 pad tilting pad bearing configuration with centrifugal compressors has greatly
improved the operating speed range of these machines. This has led marketing departments to
propose compressor designs with speeds in excess of 10,000 RPM and compressors with as many
as 11 stages. The use of the 5 pad bearing has been particularly popular with manufacturers since
this bearing design is not susceptible to self excited hydrodynamic bearing effects.
However, with the increase in speed, number of stages and higher discharge pressures, these
machines have become highly sensitive to Alford type of self excited forces generated internally.
Under normal low pressure factory testing, this sensitivity condition usually goes undetected and
does not appear until the unit is installed and operated under full load conditions.
Vibration monitoring probes are usually placed vertically at the bearing locations. Probes are never
placed at the compressor center span for motion detection because of the difficulties involved. The
safety of the unit is usually based on the monitored motion at the bearings. In some of the earlier
cases of self excited compressor whirling, the bearing preload was increased to improve the bearing
stiffness to raise the critical speed. This bearing retrofit would result in lower whirl motion observed
at the bearings but catastrophic shaft failure on startup.
A number of attempts were made to improve compressor stability by increasing the amount of
bearing asymmetry. Although bearing asymmetry does improve rotor stability both in the cases of
internal shaft friction and aerodynamic cross coupling, the vertical bearing stiffness can not exceed
values in excess of the shaft modal stiffness. The design of multistage compressors in excess of 10
stages leads to shaft stiffness values of the order of 250,000 Lb/in.
With most 5 pad bearings with L/D ratios of 0.5, we obtain vertical bearing Kyy values well in
excess of the ideal values. Under these conditions, bearing asymmetry is of no value in improving
rotor stability. This has required a number of compressors to have squeeze film dampers installed.
The squeeze film damper provides a soft support and an optimized damping value which may be
considerably below the values obtained from the tilting pad bearing.
The paradox of the 5 pad bearing design is that many units may operate with constrained whirl
motion by the simple procedure of reducing the bearing preload. This leads to larger motion
observed at the bearings ( and possibly exceeding API codes). Hence preloads are increased,
reducing bearing motion but leading to shaft rubs and failures. The further paradox of design is that
high stability may be achieved by the initial design of a wider 4 pad bearing with load between pads.
This design “floats” the shaft with a lower bearing stiffness.
Bearing asymmetry is not required for stabilization. This initial design approach is often rejected
by the end user because it appears that the compressor may be operating near a critical speed in
violation of API. This is a non issue since these higher critical speeds are so well damped that they
are not excited or detected during operation. Thus the use of the critical speed map alone can lead
to dangerous designs in an attempt to remove all apparent critical speeds from being near the
operating speed.

27
References
Alford, J., 1965, “Protecting Turbomachinery From Self-Excited Rotor Whirl”, Journal of
Engineering For Power, pp. 333-334
Barrett, L. E., and Gunter, E. J. 1975, “Steady State and Transient Analysis of a Squeeze film
Damper”, NASA CR-2548
Chen, W. J., and Gunter, E. J. 2007, “Introduction to Dynamics of Rotor-Bearing Systems”, Trafford
Publishing, ISBN: 978-1-4120-5190-3
Gunter, E. J., 1966, “Dynamic Stability of Rotor Bearing Systems”, NASA SP-113
Gunter, E. J. And Trumpler, P. R., 1969, “The Influence of Internal Friction on the Stability of High
Speed Rotors With Anisotropic Supports”, Journal of Engineering for Industry, Nov., pp. 1105-1113
Kimball, A. L. 1925, “Internal Friction of Shaft Whirling”, General Electric Review, Vol. 28 p. 554
Kirk, R. G. And Donald, G. H., 1983, “Design Criteria for Improved Stability of Centrifugal
Compressors”; ASME Rotor Dynamical Instability, AMD-Vol. 55
Lund, J. W., 1964, “Spring and Damping Coefficients for the Tilting Pad Journal Bearing”, ASLE
Trans., Vol. 7, No. 4
Lund, J. W., 1974, “Stability and Damped Critical Speeds of a Flexible Rotor in Fluid Film
Bearings”, ASME Journal of Engineering for Industry
Newkirk, B. L. 1924, “Shaft Whipping”, General Electric Review, Vol. 27, p. 169
Nicholas, J. C., Gunter, E. J., Barrett, L. E., 1978, “The Influence of Tilting Pad Bearing
Characteristics on the Stability of High Speed Rotor Bearing Systems”, ASME Topics in Fluid Film
Bearings and Rotor Bearing System Design and Optimization, pp. 55-78
Nicholas, J. C., Gunter, E. J., Allaire, P. A. 1979, “Stiffness and Damping Coefficients for the Five-
Pad Tilting-Pad Bearing”, ASLE Trans., Vol.22, No. 2
Smith, K. J. 1975, “An Operational History of Fractional Frequency Whirl”, Proceedings of The
Fourth Annual Turbomachinery Symposium, Texas A&M Univ. , pp. 115-125

28