Ed Wilcox is the Machinery Team Lead

with Lyondell/Equistar, in Channelview,

Texas. He is responsible for trobleshooting,
repair, and condition monitoring of rotat-
ing equipment at both the Lyondell and
Equistar facilities. Prior to this, Mr. Wilcox
worked for more than 10 years with Conoco
and Citgo as a Rotating Equipment Engi-
neer. He has authored several papers in the
areas of rotordynamics, vibration analysis,
and performance testing.
Mr. Wilcox has a BSME degree from the University of Missouri-
Rolla, and an MSME degree from Oklahoma State University. He
has also done postgraduate work at the Georgia Institute of
Technology in the areas of lubrication, rotordynamics, and
vibration. He is a Vibration Institute Level III Vibration Specialist
and a registered Professional Engineer in the State of Oklahoma.
ABSTRACT
Accurate and meaningful condition monitoring is necessary to
prevent both severe equipment damage and unnecessary shutdowns.
One of the most important aspects of condition monitoring is the
evaluation of startup and shutdown vibration data. A considerable
amount of upfront time must be committed to gather condition data
when the machinery is known to be in good condition to provide an
adequate picture of the current condition. Examples are given of
problems that were correctly evaluated only through access and
comparison to past startup/shutdown data as well as information
obtained from a rotordynamic model of the system.
INTRODUCTION
This paper describes how transient vibration can be used to
provide much more information about a piece of machinery than
steady-state data alone. Transient vibration data are defined as
vibration data taken when the machinery starts up or shuts down
(i.e., large changes in speed). Most of the machinery described is
of the high-speed nature with hydrodynamic journal bearings and
noncontacting eddy current displacement probes, though many of
the principles apply to low-speed applications with antifriction
bearings. A brief description is given of the type of diagnostic
equipment required to gather transient data. The different type of
transient data plots are described along with the different insights
they each provide. Likewise, several case studies are presented
where transient data were used to evaluate the stability of the
machinery as well as determine the source of the excitation forces.
Another important benefit of transient data is that they provide a
method of verifying machinery natural mode frequencies and
rotordynamic models.
IMPORTANCE OF TRANSIENT DATA
Transient vibration provides a wealth of information that is not
available from steady-state data. Among these are:

The ability to compare the vibration amplitude, phase, shaft

position, etc., at speeds above and below the shaft critical speeds

The best signature of the overall rotor natural frequencies, which

are normally a result of the complex relationship between the rotor
supports (bearings, seals, etc.) and the dynamic nature of the rotor
itself
Just as important as the comparisons listed above is the ability to
compare transient data before and after a problem exists in the
machinery. Assessments of machinery condition are made much
easier when the ability to compare both before and after are
available.
Likewise, it is very important to evaluate many different types of
plots for displaying vibration data. Using only one type of plot
limits the machinery specialists ability to obtain an adequate
picture of the machinery condition.
BASIC ROTORDYNAMICS
A basic understanding of rotordynamics is required to best
interpret transient vibration data. A simplified model of a between-
bearing rotor is provided by the Jeffcott rotor (Figure 1). The
unbalanced disk is located in the center of the massless rotor on
rigid supports. The only damping in the system is due to the
viscous drag of the air. While the Jeffcott rotor is a very simple
model, it can provide a wealth of insight into the effects of
synchronous vibration in turbomachinery. An endview of the
Jeffcott rotor shows the relationship between the center of mass
(M) and the geometric center (C) (Figure 2). The synchronous
phase angle () is the difference between the unbalance vector (U)
and displacement vector (R). The amplitude of synchronous whirl
and phase angle for the Jeffcott rotor as it passes through the rotors
first balance resonance (critical speed) is shown in Figure 3. The
response amplitude equation is shown in Equation (1).
(1)
A physical explanation of these graphs shows that well below
the critical speed, the response vector R and unbalance vector Uare
in phase (i.e., ~ 0). As the rotation speed is increased, the
unbalance vector U begins to lead R. As the critical speed is
transversed, passes through 90 degrees, then asymptotically
approaches 180 degrees at speeds above the critical. Thus at super-
critical speeds, the rotor center of mass (M) moves to the center of
the whirl orbit and stands still, while the shaft center whirls around
the center of mass. Additionally, the effect nonfrequency
dependent forces have on the whirl amplitude is shown in Figure 4.
The actual synchronous response will be some combination of
Figures 3 and 4.
An illustration of the spring, damping, and unbalance forces is
shown in Figure 5 for the Jeffcott rotor at speeds below, at, and
above the first balance resonance. As can be seen, far below the
91
TROUBLESHOOTING TURBOMACHINERY
USING STARTUP AND COASTDOWN VIBRATION DATA
by
Ed Wilcox
Machinery Team Lead
Lyondell/Equistar
Channelview, Texas
r
u
k
m
c
m
=

2
2
2 2
Figure 1. The Jeffcott Rotor. (Courtesy Vance, 1988)
Figure 2. End View of the Jeffcott Rotor. (Courtesy Vance, 1988)
Figure 3. Synchronous Response and Phase Angle of the Jeffcott
Rotor. (Courtesy Vance, 1988)
Figure 4. Synchronous Response to Constant Force, I.E.,
Misalignment. (Courtesy Vance, 1988)
first balance resonance, the shaft stiffness is approximately equal
to the unbalance force. However, at resonance, the damping is the
only term that restricts the displacement due to unbalance.
Likewise, as the speed increases past the resonance, the whirl
amplitude, r, approaches the value of the center of mass eccen-
tricity, u (assuming there are no other excitation forces such as
misalignment, internal friction, etc.).
Figure 5. Synchronous Response Below, At, and Above the Critical
Speed.
Since at resonance, the damping is the only force that contains
the Jeffcott rotor displacement, measuring displacement as the
rotor passes through resonance is a good method to determine the
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 92

amount of damping in the rotor system. The synchronous amplifi-

cation factor (Q) is a measure of the severity of a resonance, as well
as an indication of the damping present. While there are several
methods available to measure Q, the half power method is used
here because it is the most widely accepted. The basic definition is
given below in Equation (2).
(2)
The frequency bandwidth is the difference between the
frequencies where the amplitude is 3 dB below the maximum at
peak resonance. This corresponds to 0.707 times the amplitude at
resonance. This method has been adopted by the American
Petroleum Institute (API) for measuring the synchronous amplifi-
cation factor (Figure 6).
Figure 6. API Definition of Synchronous Amplification Factor.
Generally, machines with synchronous amplification factors
below four are considered to be very well damped. Values between
four and eight are considered adequately damped and are normally
stable. Between eight and 15 are considered marginal, though
many older machines may operate in the area. Amplification
factors above 15 are considered to have insufficient damping and
are very likely to experience rubs on startup and stability problems.
The synchronous amplification factor is an important diagnostic
parameter for evaluating the stability of the machinery and
comparing the machinery condition before and after a problem
exists with the equipment. As can be seen in Figure 3, the amplifi-
cation factor increases as the system damping decreases. Since
most of the damping in high-speed machinery comes from the
hydrodynamic bearings, changes in the synchronous amplification
factor can indicate a change in the damping provided by the
bearings. Which in turn can indicate a change in the condition of
the bearings.
Since most turbomachinery is supported on hydrodynamic
bearings, the effects of flexible supports need to be considered as
well. The Jeffcott rotor can be modified to have flexible bearing
supports (Figure 7). Flexible supports are very desirable from a
rotordynamics standpoint because they:

May reduce the dynamic load transmitted to the casing.

Allow the damping in the bearings to be more effective (i.e.,

softer bearings allow more velocity at the bearings,
which is required for damping).
Figure 7. Jeffoctt Rotor with Flexible Supports. (Courtesy Vance,
1988)
The calculated transmissibility ratio (F
B
/F

) for different values

of damping is shown in Figure 8. As can be seen, the damping
lowers the force transmitted to the bearings at resonance, but
actually causes it to be higher for /
n
> 1.14. This is in contrast
to Figure 3, where damping has a desirable effect on whirl
amplitude for the entire speed range. Generally speaking, it is
desirable (from a synchronous response and stability viewpoint)
for the bearing to shaft stiffness ratio (K
B
/K
shaft
) to be as low as
possible, with the obvious restriction that the low bearing stiffness
should not cause the machine to operate closer to a balance
resonance (Barrett, et al., 1978).
Figure 8. Transmissibility Versus Shaft Speed Ratio. (Courtesy
Vance, 1988)
Furthermore, rotor support stiffness is not always in the
principal direction (i.e., principal stiffnesses produce a force
opposite in direction to the rotor displacement). The stiffness and
damping coefficients of a bearing or seal support can be divided
into the following (Figure 9). The K
xy
and K
yx
terms are
commonly called cross-coupled terms (if the shaft is displaced in
the vertical direction, the cross-coupled stiffness produces a
reaction force in the horizontal direction) and reduces the positive
effects of damping if K
xy
(K
yx
) > 0. For this reason, when K
xy
>
0, the amplification factor of the rotor is increased and the stability
is reduced. Equation (3) gives the response of a rigid rotor on
flexible supports, including cross-coupled effects. Note that when
K
xy
> 0, it reduces the damping term and increases the amplifica-
tion factor. For this reason, when the amplification factor is
measured from field data, it is actually a measure of the
combination of direct damping and cross-coupled stiffness present
in the machine.
(3)
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 93
Q
sonance Frequency
Frequency bandwidth dB
N
N N
cr
=

Re
@ 3
2 1
r
u
k
m
c
m
k
xy
=

2
2
2 2
2 2
Figure 9. Support Coefficients. (Courtesy Bielk and Leader, 1994)
DIAGNOSTIC EQUIPMENT
Obtaining startup/shutdown vibration data require somewhat
specialized diagnostic equipment. While many handheld vibration
analyzers today have the ability to take startup/shutdown data, they
are normally limited to one or two channels with limited plot capa-
bilities. For a piece of high-speed turbomachinery, with
noncontacting shaft displacement probes, a system similar to
Figure 10 is required to obtain adequate startup/shutdown vibration
data. While many people today will not include the tape recorder,
it should be understood that the digital systems today do not record
the raw vibration signal, only certain digitized elements. The tape
recorder provides a copy of the vibration data that can be used for
analysis repeatedly. Often times when the digital system is used by
itself, the operator may inadvertently make an error on the
acquisition parameters. For example, maybe the maximum speed
for the keyphasor is set too low or the trigger between shaft speeds
is too low or too high. If the digital system is used by itself,
important data may be lost and may not be obtainable because the
machinery cannot be shut down or started up again or, worse,
because it was damaged. One way to improve the reliability of the
data acquisition process is to simulate the machinery shutdown or
startup prior to the actual event. This can be accomplished with a
rotor kit (Figure 11). The rotor kit can usually be adjusted to have
similar acceleration/deceleration times as the machine in question.
This allows the machinery specialist to review the data provided by
the diagnostic equipment prior to the actual machinery event and
make sure that all the data required are recorded.
The following concerns should be addressed before the data
acquisition process begins:

Number of samplesWhile the analog tape record has

somewhat unlimited recording space, the newer digital systems do
not. Most digital systems have a limited number of vector and
waveform samples that can be recorded before the data have to be
written to the hard-drive of the computer. When this happens, data
acquisition stops, while the data is being written to the hard-drive.
For this reason, the sampling method (i.e., delta speed or delta
time) must be adjusted based on the machinery configuration to
ensure that enough data are taken for analysis, but not so much that
the maximum samples are exceeded.

Max frequency/number of linesDepending upon the

machinery configuration and the purpose of the analysis, the
maximum frequency and number of lines for the fast Fourier
transform (FFT) algorithm become very important. On machinery
that is driven by electric motors, the acceleration time is many
times quite rapid (i.e., less than 10 seconds). For these applications,
to get adequate data the maximum frequency should be as high as
reasonable to reduce the time for data acquisition. Similarly, the
Figure 10. Diagnostic Equipment Required for Startup/Shutdown
Data.
Figure 11. Rotor Kit.
number of lines should be as low as possible while still providing
adequate resolution. If the specialist is interested mostly in
synchronous vibration, the resolution will not be as important as
the number of samples obtained. However, if multiple nonsyn-
chronous frequencies are present in the vibration signal, the
resolution obviously becomes more important.

Number of vectors per waveform sampleSome digital systems

do not record a waveform sample for every vector sample (to save
memory and increase sample speed). For this reason, the sample
frequency should be closely examined to give the correct number
of spectra (i.e., more vector samples may be required than are
necessary to obtain adequate waveform samples).

Keyphasor voltageThe keyphasor signal is extremely

important, since most of the digital analysis is dependent upon this
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 94











&
X
X
X
X
Y
Y
Y
Y
value. For this reason, the voltage change produced by the
keyphasor should be set up correctly in the software and monitored
during the data acquisition using the oscilloscope shown in Figure
10.
As can be seen, the digital systems require a large amount of
input from the specialist to obtain meaningful data. For this reason,
the tape backup becomes even more important on critical unspared
machinery.
TYPES OF TRANSIENT DATA PLOTS
Transient vibration data can be displayed in many different
formats. Each different type can reveal information that is not
readily available from the other. For this reason, all the types
should be used when evaluating the condition of a machine. The
different types of plot are:

Bod/polar

Cascade/waterfall

Shaft centerline

Orbit
This paper gives a description of each type of plot along with what
can be determined from it.
Bod and Polar Plots
Bod plots are the most common method of displaying transient
data. They normally display overall or synchronous vibration
(however, nX data can be displayed as well) and corresponding
phase, versus speed (Figure 12). The rotor natural frequencies can
be determined by locating the peak in the vibration amplitude that
corresponds to a phase shift of approximately 180 degrees.
Figure 12. Bod Plot.
Polar plots show the same data as Bod plots, just in a different
format (Figure 13). Sometimes, polar plots make it easier to
determine when the rotor has transversed its critical speed, because
the phase change is obvious (i.e., every 180 degrees on the plot).
For example, from the Bod plot shown in Figure 12 it is not
immediately obvious that this rotor is approaching its second
critical. However, one look at the polar plot in Figure 13, and it is
very clear.
For the most accurate Bod or polar plot, the vector data must be
runout compensated. All probe areas have a certain amount of
physical and/or electrical runout. The vector data are corrected for
this runout by vectorally subtracting the vibration vector at a very
low rpm from all the vector data obtained. Figures 14 and 15 show
both the compensated (dashed) and uncompensated (solid) vector
data. As can be seen, the compensated data show the vibration goes
to zero.
Figure 13. Polar Plot of Centrifugal Compressor Drive-End
Bearing.
Figure 14. Bod Plot Showing Compensated and Uncompensated
Data.
Figure 15. Polar Plot Showing Compensated and Uncompensated
Data.
Case Study 1Centrifugal Compressor with Increasing
Radial Synchronous Vibration after Some Shutdowns
A particular high-speed centrifugal compressor was experi-
encing large increases in synchronous radial vibration after some
(but not all) shutdowns (Figure 16). Likewise, it was noticed that
the increase in vibration only occurred when the compressor
tripped, not when it was intentionally shut down by operations.
Normally, increases in synchronous radial vibration are caused by
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 95
either increased unbalance or bearing clearance. A natural
assumption was that the bearings were being damaged somehow
during coastdown. However, the bearing temperatures were normal
(i.e., < 180F). Likewise, both lube oil pumps, as well as the lube
oil rundown tank, were functioning properly. Additionally, the
probe gap voltages had not increased, indicating that the rotor had
not dropped down in the bearings.
Figure 16. Increase in Radial Vibration after Compressor Trips
over Time.
Records of the transient vibration data from the mechanical test
from the original equipment manufacturer (OEM) were found
showing the shaft radial vibration during startup (Figure 17).
Additionally, transient data were recorded during several of the
compressor shutdowns (Figures 18 and 19). Although the
synchronous radial vibration had increased from 0.4 mils to 4.0 mils
(i.e., Figure 19), the measured amplification factor had remained
approximately the same (~ 4 to 5). This indicated that the damping
produced by the bearings was the same, which meant that unbalance
was probably the source of the problem. At the next available
opportunity, the compressor bearings were inspected and found to be
in excellent condition. The rotor was pulled and found to be heavily
fouled with an oil catalyst residue. The bearings were reused and the
rotor cleaned. The vibration dropped to approximately 0.5 mils with
the clean rotor. Additionally, it was determined that if the compressor
tripped, process gas that contained oil would leak back through the
discharge check valve into the compressor, which was causing the
fouling. If the compressor was shut down normally, a control valve
was closed that prevented gas from backing into the compressor.
Knowing that the source of vibration was probably deposits on the
rotor allowed maintenance to be prepared for a rotor changeout (i.e.,
spare rotor prepared, rigging available, additional crews), which
reduced both maintenance costs and operational downtime.
Figure 17. Transient Data from Mechanical Test at OEM Facility.
Case Study 2Centrifugal Compressor with High
Radial Synchronous Vibration after Previous Shutdown
A high-speed barrel compressor tripped due to an instrumenta-
tion failure. After the compressor was restarted, the nondrive-end
(NDE) radial vibration increased from approximately 0.5 to 2.0
Figure 18. Transient Data after Two or Three Trips.
Figure 19. Transient Data from Last Startup, Shortly before
Overhaul.
mils. All the vibration was synchronous. As in the previous case
study, it was assumed that the bearings had been damaged during
coastdown. The bearing damage was verified when transient data
recorded during a planned shutdown (approximately one year after
the instrumentation failure) were compared to transient data from
a shutdown prior to the instrumentation failure (Figures 20 and 21).
Figure 20. NDE Vertical Vibration Prior to Instrumentation Trip.
As can be seen, the amplification factor on the NDE bearing had
increased dramatically. Inspection of the NDE bearing found that
the lower pads were indeed wiped (Figure 22). After the NDE
bearing was replaced, the compressor vibration returned to normal
(Figure 23).
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 96
O
v
e
r
a
l
l

R
a
d
i
a
l

V
i
b
r
a
t
i
o
n

(
m
i
l
s
)


















Figure 21. NDE Vertical Vibration after Bearing Damage.
Figure 22. Lower Half of NDE Tilt-Pad Bearing.
Figure 23. NDE Vertical Vibration after Bearing Replacement.
Case Study 3Centrifugal Compressor
with High Synchronous Vibration
A centrifugal compressor experienced a large increase in
synchronous radial vibration, mostly on the inboard end (3 to 4
mils) and was shut down to inspect the bearings. The excessively
high vibration had severely damaged the radial bearings (Figure
24) as well as the bearing housing itself. Due to business
constraints, the bearings/housings were replaced/repaired and the
compressor was restarted, although the cause of the high vibration
was unknown. On startup, it was observed that the measured first
critical speed had increased from approximately 4500 cpm to 6800
cpm (Figure 25). Since the thrust load had increased, this seemed
to indicate that the rotating labyrinth on the balance piston might
be rubbing the abraidable stationary, which would raise the first
critical by effectively decreasing the bearing span. The radial
vibration stayed at high, but acceptable, levels for three months
until the compressor was shut down for a scheduled unit outage. A
rub was found in the machine, but not at the balance piston seal.
One of the diaphragms had broken and was being forced forward
against the rotor (Figures 26 and 27). The diaphragm was repaired
and the measured first critical returned to normal (Figure 28).
Figure 24. Damaged Radial Bearing.
Figure 25. Startup Showing Abnormally High First Critical.
Figure 26. Damaged Inlet Guide Vanes/Diaphragms.
Cascade/Waterfall Plots
Cascade/waterfall plots are simply fast Fourier transforms at
different speeds on the same plot. They are used to diagnose
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 97
Figure 27. Cracked Diaphragm.
Figure 28. Bod Plot after Overhaul.
nonsynchronous vibration problems that may appear (and
disappear) at different rotational speeds. One of the most common
problems diagnosed with these plots is oil whirl that turns into
oil whip. An illustration and example of this are given in Figures
29 and 30. As can be seen, the subsynchronous vibration tracks
with the shaft speed until it coincides with the first natural mode of
the machine. At this point, it locks onto this frequency regardless
of the change in rotational speed. This is the point when oil whirl
turns into oil whip. In contrast, Figure 31 shows a centrifugal
compressor that is experiencing rotating stall due to a large drop in
the molecular weight of the process gas. As can be seen, at design
speed (8000 rpm), the rotating stall is exciting the first natural
mode of the compressor at 2700 cpm. As the compressor shuts
down, the subsynchronous vibration drops away. A different
compressor had been experiencing erratic subsynchronous
vibration. A rotordynamic analysis of the compressor reveals that
if the oil seal rings locked up, the first natural mode became
unstable. This was verified, during a startup of the compressor,
when the first mode was excited (Figure 32). As can be seen, the
subsynchronous vibration appears, then drops away as the
compressor starts up.
SHAFT CENTERLINE PLOTS
Shaft centerline plots show the location of the journal center,
relative to the proximity probes. This type of data is only available
when using proximity probes because the actual shaft location is
determined from the DC component of the vibration signal, which
gives the actual distance from the probe tip to the shaft (i.e., not the
vibration). This information is most useful to the machinery
specialist if it is recorded over time, especially as the speed
Figure 29. Illustration Showing Transition from Oil Whirl to Oil
Whip. (Courtesy Vance, 1988)
Figure 30. Cascade Plot Showing Transition from Oil Whirl to
Oil Whip. (Courtesy Eisenmann and Eisenmann, 1998)
Figure 31. Centrifugal Compressor with Rotating Stall/Surge that
Excites the First Mode.
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 98

















Figure 32. Waterfall Plot Showing Subsynchronous Vibration
Appear, Then Disappear.
changes (i.e., absolute values by themselves are meaningless).
Shaft centerline plots over long periods of time (i.e., months/years)
can also provide insight into the degradation of a bearing. A journal
develops an oil wedge between it and the bearing at a very low
rotational speed (dependent upon the journal load and bearing
configuration) (Figure 33). This causes the journal to rise in the
bearing when the machinery begins to rotate. As the speed
continues to increase, the journal typically continues to rise in the
bearing, as the pressure developed by the hydrodynamic action of
the wedge increases.
Figure 33. Oil Wedge Developed by Hydrodynamic Bearings.
(Courtesy Zeidan, 1991, Turbomachinery Laboratory)
Case Study 4Centrifugal Compressor
with Excessive Radial Bearing Temperature
The drive-end radial bearing temperature of a high-speed
centrifugal compressor was excessive after overhaul. The
temperature would spike up rapidly to 300+F (150C) if it rained
on the compressor (Figure 34). The bearing clearance was quite
low (0.0045 inch on a 3.5 inch journal at 11,500 rpm), and it was
decided that the bearing housing was contracting on the bearing
during the rainstorm and reducing the clearance further. The
compressor was shut down to install another bearing with higher
clearance. The shaft centerline plots of the drive-end (DE) and
NDE bearings are shown in Figures 35 and 36 during the
coastdown. As can be seen, the drive-end journal did not drop at all
as the machine coasted down from approximately 11,000 rpm.
Examination of the bearings on the DE of the compressor revealed
that the clearance was much too tight (Figure 37). Obviously, the
top pads on this type of machine should not show evidence of wear.
Bod plots for this machine did not indicate a problem (Figure 38).
This type of problem was only evident from the shaft centerline
plots. After a bearing with more clearance was installed, the shaft
centerline plot appears normal for the drive-end bearing on startup
(Figure 39).
Figure 34. Radial Bearing Temperature Spike During Rainstorm
on Centrifugal Compressor.
Figure 35. Shaft Centerline Plot of Compressor DE Bearing with
Inadequate Clearance.
Case Study 5Centrifugal Compressor
with Excessive Radial Bearing Temperature
In a similar problem to the previous case study, a centrifugal
compressor was overhauled and the radial bearing temperature on
the nondrive-end was high (~ 240F) after startup. At the time, it
appeared to be a bad resistance temperature detector (RTD) because
the wires had been cut and spliced during installation. Over the next
year, the temperature slowly crept up until it reached 275F. The
temperature was controlled at or below 275F by adding additional
cooling to the lube system. The vibration technician reported that
the gap voltage from the shaft displacement probes was increasing
(Figure 40). This seemed to indicate that the journal was dropping
down in the bearing. Closer examination of the gap voltage graph
reveals that the voltage is becoming less negative (i.e. the journal
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 99




















Figure 36. Shaft Centerline Plot of NDE Bearing.
Figure 37. Damaged Bearing Showing Affects of Inadequate
Clearance.
Figure 38. Bod Plot.
Figure 39. Shaft Centerline Plot of DE Journal with Proper
Bearing Clearance.
is rising in the bearing, getting closer to the probes). This is the
opposite direction that would be expected if the pads were wiping.
Approximately one month later, the radial vibration on the NDE
temporarily spiked up from 0.5 to 2.0 mils. Since a spare rotor was
not available for this compressor, it was decided to shut down the
compressor immediately and inspect the bearings (Figure 41).
Note that the amplification factor is relatively low. Inspection of
the bearing after the compressor was shut down revealed a heavy
varnish on the pads (Figure 42). Furthermore, the babbitt had
fatigued above the RTD on one of the lower pads as well (Figure
43). Close inspection of the bearing and bearing housing revealed
that the porting from the bearing housing to the bearing did not line
up, which was restricting the lube oil flow to the bearing. It was
remarkable that the bearing received any oil at all. The porting was
corrected and the compressor was started with no problems. The
measured Bod plot on startup was unchanged from what was seen
on shutdown (Figure 44). This was expected since the bearing had
not been damaged yet.
Figure 40. Gap Voltage of NDE Probes.
ORBIT PLOTS
An orbit shows the displacement of the journal inside the
bearing clearance. Figure 45 shows a 1 filtered orbit (that is why
it is so smooth). To view the nonsynchronous shaft vibration
requires examining the unfiltered orbit (Figure 46). Orbits are
normally steady-state plots (not actually steady-state, just at one
speed). However, orbits at different speeds can be compared to
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 100
Figure ? Bode plot of compressor with insufficient bearing clearance

















Upper Half
Lower Half
Figure 41. Shutdown Bod Plot of Centrifugal Compressor with
High NDE Radial Bearing Temperature.
Figure 42. One of Lower NDE Bearings with Varnish Buildup
(Babbitt is Not Wiped).
Figure 43. One of the Lower NDE Bearing Pads Showing Dimple
Where Babbitt Failed above RTD.
determine the source of the vibration. Figure 47 shows the radial
vibration on a centrifugal compressor while the speed was
increased from 11,000 to 12,200 rpm. As can be seen, the vibration
increased from approximately 1.4 mils peak-to-peak to over 3 mils
Figure 44. Startup Bod Plot of Centrifugal Compressor after Oil
Obstruction Was Removed.
peak-to-peak. Examination of the orbits before and during
the speed change showed two large internal loops appear at the
higher speed (Figures 46 and 48). This is indicative of a subsyn-
chronous frequency at approximately one-third of running speed
(Equation (4).
(4)
A rotordynamic analysis of the compressor revealed that the first
mode (at one-third of running speed) was only marginally stable.
Additionally, changes in alignment or preload can cause strange
looking orbits. Figure 49 shows how preload in one direction
affects the orbit shape. Figure 50 shows the orbit from an induction
motor that has severe temperature related vibration problems. The
overall vibration amplitude would reach 8 mils on this particular
machine during the heat of the summer and drop to 2 mils during
the winter.
Figure 45. Shaft Orbit.
USING STARTUP/SHUTDOWN DATA IN
CONJUNCTION WITH ROTORDYNAMIC MODELS
Startup/shutdown data can be used as a very important tool to
validate and/or tune rotordynamic models of turbomachinery.
While the accuracy of rotordynamic models has increased, actual
measured critical speeds and amplitudes are still very important in
evaluating a machines condition. Additionally, rotordynamic
models can be used to gain insight into some startup and shutdown
data.
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 101
Vib Frequency
shaft speed
of ernal loops
=
+ # int 1
Figure 46. Large Increase in Radial Vibration during Speed
Change.
Figure 47. Orbit at 11,000 RPM.
Figure 48. Orbit Showing Subsynchronous Component at 12,000
RPM.
Figure 49. Change in Shaft Orbits with Increasing Preload.
Figure 50. Induction Motor with Severe Internal Misalignment.
Case Study 5High Synchronous Vibration
on Centrifugal Compressor after Rerate
A small high-speed, multistage barrel compressor was rerated
because the molecular weight of the gas was lower than the
original design. This required that the speed be increased from
17,500 to 19,500 rpm. Because the compressor would be operating
closer to its second critical speed, the spare rotor was at-speed
balanced. The rerated rotor was installed and the compressor was
restarted. The synchronous radial vibration on the nondrive-end of
the compressor was approximately 1.5 mils. While 1.5 mils is not
normally considered excessive for lower speed machines, at 19,500
cpm this is much too high. The startup Bod plots for the drive-end
and nondrive-end are shown in Figure 51. Note the differences in
shape between the drive- and nondrive-end plots. Comparison to
the predicted response for the minimum and maximum bearing and
seal clearance in Figures 52 and 53 indicates that the high
synchronous vibration is probably caused by higher clearances
instead of imbalance. Additionally, examination of the orbits from
both bearings shows a lubricated rub on the drive-end (Figures 54
and 55). Note that the orbit with the inner loops is on the drive-end
(which has the lower vibration). Examination of the repair records
showed that the bearing and seal clearances on the nondrive-end
were at the maximum of the allowable clearance.
Figure 51. Bod Plots of High-Speed Centrifugal Compressors
with Excessive Synchronous Vibration after At-Speed Balance.
SUMMARY
Transient vibration data provide a wealth of information about
the condition of the machinery that is not available from steady-
state data. Likewise, all the different transient plot types must be
used to get the most from the data. This information can be used to
evaluate the condition of the bearings as well as help determine the
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 102
Comp
Speed
CIY
Radial Vib
Figure 52. Predicted Unbalance Response with Nominal Bearing
and Seal Clearances.
Figure 53. Predicted Unbalance Response with Maximum Bearing
and Seal Clearances.
Figure 54. Orbits of NDE and DE of High-Speed Centrifugal
Compressor.
Figure 55. Closeup of DE Orbit on High-Speed Centrifugal
Compressor Showing Lubricated Rub.
cause of excitation forces such as unbalance and rubs when it is not
immediately obvious. Transient data also provide a method of
verifying machinery natural mode frequencies and rotordynamic
models.
ACKNOWLEDGEMENTS
This paper would not have been possible without the gifts given
the author by his Lord and Savior Jesus Christ. Additionally, the
author will always be in debt to his wife, Angie, and his four
daughters, Lauren, Bailey, Allie, and Peyton, for their patience in
letting him take the time to write this paper. The author would also
like to thank Charlie Jackson and Charlie Rutan for providing
some of the figures in this paper.
REFERENCES
Barrett, L. E., Gunter, E. J., and Allaire, P. E., 1978, Optimum
Bearing Support Damping for Unbalance Response and
Stability of Rotating Machinery, Journal of Engineering for
Power, 100, pp. 89-94.
Bielk, J. R. and Leader, M. E., 1994, Rotordynamics, Short
Course 2, Twenty-Third Turbomachinery Symposium,
Turbomachinery Laboratory, Texas A&M University, College
Station, Texas.
Eisenmann, R. C. and Eisenmann Jr., R. C., 1998, Machinery
Malfunction Diagnosis and Correction, New Jersey: Prentice
Hall.
Vance, J., 1988, Rotordynamics of Turbomachinery, New York,
New York: John Wiley.
Zeidan, F. Y. and Herbage, B. S., 1991, Fluid Film Bearing
Fundamentals and Failure Analysis, Proceedings of the
Twentieth Turbomachinery Symposium, Turbomachinery
Laboratory, Texas A&M University, College Station, Texas,
pp. 161-186.
BIBLIOGRAPHY
Childs, D. W., 1993, Turbomachinery Rotordynamics, New York,
New York: John Wiley.
TROUBLESHOOTING TURBOMACHINERY USING STARTUP AND COASTDOWN VIBRATION DATA 103
PROCEEDINGS OF THE THIRTY-FIRST TURBOMACHINERY SYMPOSIUM 2002 104