Professional Documents
Culture Documents
Malfunction Diagnosis:
Fluid-Induced Instability
By
Charles T. Hatch
Table of Contents
Fs = K r (1)
1
f=
where K is the effective spring constant of the bearing at that eccentricity ratio, and r is
the distance from the center of the bearing.
At the same time, a quadrature component that acts in a tangential direction in the
same sense as rotor rotation. It turns out that this tangential force, Ft, is a function of
bearing damping, fluid circulation, rotor speed, and distance from the equilibrium
position:
Ft = jDλΩ r (2)
where D is the bearing (or seal) damping, λ (lambda) is the Fluid Circumferential
Average Velocity Ratio, Ω (capital omega) is the angular velocity of the rotor (the speed
of the rotor in radians/sec), and r is the distance from the center of the bearing.
The j is − 1 . Practically, all that means is that the action occurs at 90°. (See
Reference [1] for much more detail.)
What is λ? Put most simply, λ is a measure of the amount of fluid circulation in the
bearing. It is defined as the ratio of the average angular velocity of the fluid to the angular
velocity of the rotor. For a plain cylindrical, fully lubricated (360° lubricated) bearing, λ
is typically a little under ½, around 0.49 or so. But the value of λ can be influenced by the
geometry of the bearing, the rate of end leakage out of the bearing, the eccentricity ratio
of the rotor in the bearing, and the presence of any pre- or antiswirling that may exist in
the fluid.
Note that the strength of the tangential
force depends not only on the rotation Start
speed, Ω, but also on the strength of fluid
circulation around the rotor (λ). It is much
stronger (λ is much higher) when the rotor
is surrounded with fluid (the fluid-film
bearing is fully, or 360° lubricated).
Properly loaded fluid-film bearings are
normally only partially lubricated, and λ is
usually small. Thus, properly loaded
bearings are unlikely to be a source of very
large tangential forces unless the bearing
becomes flooded with an excess of
lubricant. Note that fluid-film bearings can
become unloaded, for example because of
misalignment, transition to fully lubricated Figure 2. A stable rotor with a fluid-film bearing
operation, and generate high tangential is turning X to Y (CCW) and is moved by a
disturbance to the Start position. Because of the
forces. fluid-related tangential force, the rotor cannot
To see how the tangential force affects move immediately back to the center of the
rotor stability, imagine a rotor rotating in bearing, but follows a decaying spiral path. The
the center of a fluid-film bearing. If the frequency of precession is a natural frequency of
rotor is displaced from the center to some the rotor system.
position and released, the spring force tries to push the rotor back toward the center. But,
at the same time, the tangential force tries to push the rotor in a direction 90° from the
center in the direction of rotation. The rotor begins to move at some angle determined by
2
f=
the ratio of the tangential force and the spring force (Figure 2). Once moving, the
damping of the bearing produces a force that is opposite to the direction of instantaneous
rotor motion and tries to slow down the rotor. And, the inertia of the rotor comes into
play. All of these forces act together in a complicated way. The result is that the stable
rotor cannot return immediately to the center, but instead follows a decaying spiral path
back to the center of the bearing.
The precession frequency of this spiral path is equal to the rotor system damped
natural frequency. This is similar to what happens if you pluck a guitar string. The string
vibrates at its natural frequency and slowly decays back to the original equilibrium
position.
As rotor speed increases, the tangential force becomes stronger while the other forces
do not (the spring force and damping forces are independent of rotor speed). Eventually a
speed will be reached where the tangential force becomes so strong that the rotor is
unable to return to the original position at all. If the rotor speed remains constant, the
rotor will continue to orbit around the original equilibrium point of the rotor or perhaps
even orbit around the bearing in an increasing spiral. The rotor is then said to be in fluid-
induced instability.
It can be shown that, when the rotor system is operating in fluid-induced instability,
the frequency of precession of the rotor is equal to an undamped natural frequency of the
rotor system. For our simple rotor system the undamped natural frequency, ω n (lower
case omega), is given by this expression:
K
ωn = (3)
M
where K is the spring stiffness, and M is the rotor mass. (The damped natural frequency
of our rotor model is a more complicated expression, and is also a function of rotor speed
[1].)
There is another way to look at the fluid-induced instability problem. For a simple
rotor model, rotor dynamic response is determined by the following expression (adapted
from reference [1]):
F∠δ
A∠α = (4)
K − Mω 2 + jD(ω − λΩ )
where A∠α is the amplitude and phase of the response, F∠δ is the amplitude and
phase of a force that is rotating at some arbitrary frequency ω (lower case omega) that is
different from the rotor speed, Ω. The denominator of Equation (4) is referred to as the
Dynamic Stiffness, KDS.
K DS = K − Mω 2 + jD(ω − λΩ ) (5)
K D = K − Mω 2 (6)
3
f=
K Q = jD(ω − λΩ ) (7)
The Dynamic Stiffness in Equation (4) is what keeps the rotor from flying off into
space when subjected to a force. The higher the stiffness, the smaller the response, and
the relationship between the Direct and Quadrature parts controls the changes in phase
between the input force, F∠δ , and the output response, A∠α .
If the Dynamic Stiffness were to become zero, there would be nothing to restrain the
rotor and the response would be infinite. The rotor would be unstable. The only way this
can happen is if the Direct Dynamic Stiffness and the Quadrature Dynamic Stiffness
become zero at the same time. Thus, fluid-induced instability occurs when both terms of
the Dynamic Stiffness become zero simultaneously:
K − Mω 2 = 0
(8)
jD(ω − λΩ ) = 0
From these two expressions, by using a little algebra we can obtain a very important and
useful expression for the rotor speed at which the rotor system first begins to go unstable,
called the Threshold of Stability:
1 K
Ω TH = (9)
λ M
Note that only three parameters control the Threshold of Stability, λ, K, and M. (In
reality, system damping in rotor systems can also play a role, but it is a relatively small
effect and our simple model neglects it.) Note also the presence of the undamped natural
frequency from Equation (3).
There is an important point regarding this equation: If the rotor speed is less than
ΩTH, then the rotor system will be stable. Or to look at it another way, if ΩTH is above our
operating speed, then the rotor system will be stable. Thus, to ensure rotor stability, all
we have to do is keep the Threshold of Stability above our operating speed.
4
f=
Rotor
Bearing Center
Bearing Wall
Mass
M
Bearing
Stiffness
Shaft Stiffness
Shaft
Spring KS
Bearing KB
Combination Spring
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Foundation
Bearing Eccentricity Ratio
Figure 4. The rotor system behaves as though the
Figure 3. Rotor system spring stiffness is a series shaft spring, KS (blue), is connected in series with
combination of shaft spring stiffness (blue) and the bearing spring, KB (green). In any series
bearing spring stiffness (green). Bearing stiffness combination of springs, the stiffness of the
is a strong function of eccentricity ratio, while combination is lower than the weakest spring in
shaft stiffness is independent of eccentricity ratio. the series. Because the bearing stiffness is a
For any operating position in the bearing, the strong function of eccentricity ratio, the
combination stiffness (red) is less than the combination stiffness becomes controlled by the
weakest spring (see Figure 4). shaft stiffness at high eccentricity ratios (see
Figure 3).
nears the wall, the stiffness starts to increase dramatically and becomes extremely high
very close to the bearing wall. It is just this effect that produces the high load carrying
capacity of a fluid-film, hydrodynamic bearing. The bearing stiffness is a strong function
of the position of the rotor in the bearing and can vary over a very wide range of values.
Up to now, we have treated the spring stiffness, K, as a single spring that acts toward
the center. However, the situation is usually more complicated than that. A flexible rotor
can be thought of as a mass that is supported by a shaft spring, which is in turn supported
by a bearing spring (Figure 4). Thus K actually consists of two springs in series, the shaft
spring, KS , and the bearing spring, KB.
For these two springs connected in series, the stiffness of the combination is given by
these equivalent expressions:
1 KB KS
K= = = (10)
1 1 K K
+ 1 + B 1 + S
K
S K B KS KB
For any series combination of springs, the stiffness of the combination is always less
than the stiffness of the weakest spring. The weak spring controls the combination
stiffness.
Typically, the bearing stiffness at the center of the bearing (eccentricity ratio = 0) is
much lower than the shaft stiffness. In that case, the ratio KB/KS is small. In that case, the
middle of Equation (10) tells us that the combination stiffness is a little less than KB. In
other words, at low eccentricity ratios, the bearing stiffness is the weak stiffness and it
controls the combination stiffness.
5
f=
On the other hand, the bearing stiffness close to the bearing wall (eccentricity ratio
near 1) is typically much higher than the rotor shaft stiffness (Figure 3). Because of this,
the ratio KS/KB is small. Then, the third of Equation (10) tells us that the combination
stiffness is a little less than KS. Thus, at high eccentricity ratios, the shaft stiffness is the
weak stiffness and it controls the combination stiffness.
To summarize, at low eccentricity ratios, the bearing stiffness controls the rotor
system stiffness. Therefore, any changes in bearing stiffness will show up immediately as
changes in the overall, rotor system spring stiffness, K. On the other hand, at very high
eccentricity ratios, the constant shaft stiffness is in control, and the overall rotor system
spring stiffness will be approximately independent of changes in bearing stiffness.
Remember that the natural frequency of the rotor system is given by K / M , and it
is function of the rotor system spring combination stiffness, K. Thus, the natural
frequency of the rotor system is a function of bearing eccentricity ratio and depends on
which stiffness element, bearing or shaft, is the weakest at any particular time.
Because fluid-induced instability causes the rotor to precess at its natural frequency,
the frequency of precession in fluid-induced instability will depend on the eccentricity
ratio. This is the key to understanding the difference between whirl and whip.
Whirl
Figure 5 shows a full spectrum cascade plot of a startup of a rotor system where the
rotor is initially centered in a fluid-film bearing. When the rotor speed reaches the
Threshold of Stability, the rotor enters fluid-induced instability and develops large
amplitude, forward, subsynchronous vibration. The subsynchronous vibration tracks rotor
speed for a while and then transitions to a constant frequency. We define whirl as the
rotor behavior that tracks rotor speed and whip as the behavior that locks to a particular
frequency.
Remember that fluid-induced instability causes the rotor to precess at the natural
frequency of the rotor system. Thus, at the beginning of fluid-induced instability, the
rotor starts precessing at the natural frequency that corresponds to the stiffness at the
center of the bearing. For this reason, this is called the low eccentricity natural frequency
of the rotor system. Note that the bearing stiffness at the center of the bearing is less than
the rotor shaft stiffness, so the bearing stiffness controls the overall rotor system stiffness
and therefore the rotor system natural frequency.
When the rotor starts into fluid-induced instability, the rotor starts to spiral away from
the stable operating point, the diameter of the orbit increases, and the rotor begins
orbiting about the center of the bearing in an approximately circular orbit (inset in Figure
5). While the average eccentricity ratio is close to zero, the dynamic eccentricity ratio
increases as the orbit diameter increases. If the fluid-film bearing had a constant stiffness
profile, this orbit diameter increase would continue forever until the rotor system
destroyed itself. However, this is not the case, and the rotor moves into a region of the
bearing where the stiffness is higher.
This higher stiffness increases the natural frequency of the rotor system ( K / M )
and increases the Threshold of Stability (see Equation (9)), and the rotor settles into a
“stable” limit cycle of subsynchronous vibration at the new natural frequency of the rotor
system. It would stay this way if the rotor speed remained constant.
6
f=
However, because the system is starting up, the rotor speed increases further and
again exceeds the new Threshold of Stability. The orbit diameter increases further until
the bearing stiffness drives the Threshold of Stability high enough to “restabilize” the
rotor. This cycle continues, and the subsynchronous vibration frequency (which is the
natural frequency of the rotor) tracks running speed as the rotor accelerates. The rotor
operates on the continually changing Threshold of Stability.
Rotor Speed (rpm)
Whip
irl
Wh
of Stability
Threshold
Figure 5. Full spectrum cascade plot of a rotor system startup. The rotor system starts into fluid-induced
instability (in whirl) at about 2400 rpm, the Threshold of Stability. At this time, subsynchronous, forward
precession begins at a frequency near 0.475X. The initial whirl frequency is about 1300 cpm, which is the
low eccentricity natural frequency of the rotor system. As speed increases, the whirl orbit becomes larger,
the bearing becomes stiffer, and the rotor system natural frequency shifts to a higher frequency. Thus, the
whirl tracks at a sub multiple of running speed. At about 2900 rpm, the high 1X rotor vibration associated
with a balance resonance causes the rotor to operate at a high dynamic eccentricity ratio. The resulting
higher bearing stiffness pushes the Threshold of Stability temporarily above running speed, and the fluid-
induced instability disappears. After the resonance, 1X vibration declines, the orbit diameter decreases, the
bearings stiffness decreases, and the Threshold of Stability once again falls below running speed; thus the
fluid-induced instability reappears. When the rotor dynamic motion reaches high eccentricity, the rotor shaft
becomes the weakest spring in the system, and the instability frequency locks in to the high eccentricity
natural frequency in whip. The orbit inset shows the orbit of the rotor inside the bearing in whirl, and the
magenta circle shows the approximate bearing boundary. At this dynamic eccentricity ratio (about 0.6), the
bearing controls the spring stiffness of the rotor system (see Figure 3). The pair of Keyphasor dots are
shifting slowly in a direction opposite to rotation. This indicates that the frequency of vibration is a little
less than 1/2X.
7
f=
For the simple rotor we have been discussing, the frequency of precession in fluid-
induced instability can be found from Equation (9). Because the rotor is precessing at the
natural frequency, and because the rotor is operating on the Threshold of Stability, then
the precession frequency must be λΩ. If λ is about 0.48, then the subsynchronous
precession in whirl will take place at a frequency of about 0.48X. This relationship is not
so simple for more complex rotor systems, but it does provide a useful ballpark estimate.
Whip
At some point in the startup, the dynamic eccentricity ratio of the rotor begins to
approach the bearing wall, and the bearing stiffness becomes so high that it is no longer
the weakest spring in the rotor system. The shaft spring becomes the weakest spring, and
it controls the stiffness of the rotor system. And, the shaft spring cannot be changed by
changes in vibration amplitude or eccentricity ratio.
Because of this, the subsynchronous vibration (and natural frequency) of the rotor
system asymptotically approaches a constant value (see Figure 5). We call this value the
high eccentricity natural frequency. In this region, the frequency of precession of the
rotor system remains constant, and we define this instability region as whip. The high
eccentricity natural frequency often correlates well with the “nameplate critical” in
rotating machinery
because normally-
Whip
loaded fluid-film
bearings operate at high
eccentricity ratios.
Whip vibration is
Rotor Speed (rpm)
8
f=
bearings. If a bearing becomes flooded with an excess of lubricant, it is possible for fluid-
induced instability to appear immediately as whip and display no whirl region at all
(Figure 6).
For whirl and its characteristic speed tracking to occur, a mechanism must be
available that can modify the rotor natural frequency. This condition is most easily met
when the rotor operates near the center of a fluid-film bearing (perhaps due to
misalignment). If such a natural frequency modifying mechanism is lacking, then fluid-
induced instability will start as whip.
Subsynchronous Vibration
The primary symptom of fluid-induced instability is forward, subsynchronous
vibration. The frequency of the subsynchronous vibration is usually less than 0.5X
(Figures 5 and 6). The frequency of the subsynchronous vibration in whirl is related to
the fluid swirling rate (the Fluid Circumferential Average Velocity Ratio, λ) in the region
of the rotor system that is the source of the problem. In whip, however, the frequency of
vibration will lock to a rotor system bending mode (Figure 6). The subsynchronous
frequency can range from 0.3X to 0.8X or higher if fluid has been preswirled before entry
into the region causing the problem.
Unlike rub, fluid-induced instability almost never produces a pure integer ratio
vibration frequency such as 1/2X, 2/3X, 1/4X, 1/3X, etc. Instead, fluid-induced instability
produces irrational fraction frequencies. However, if the lubricating film breaks down
between rotor and stator, or if the large amplitude instability vibration causes a rub
elsewhere, then fluid-induced instability can lock to an integer ratio.
The subsynchronous vibration caused by fluid-induced instability is almost purely
forward (Figures 5 and 6). This is a very useful way to discriminate between rub and
fluid-induced instability as a root cause. Rub tends to produce significant reverse
components at the subsynchronous frequency.
During a startup or shutdown, whirl due to fluid-induced instability will track running
speed at some sub multiple (Figure 5), while whip tends to lock to a constant frequency
(Figure 6). As can be seen in the figure, it is possible for whip to suddenly appear without
any whirl.
Fluid-induced instability is always associated with a natural frequency of the rotor
system (usually the lowest mode). Often the balance resonance associated with that mode
will appear during startup as 1X vibration (Figure 6). However, if the lowest mode of the
rotor is supercritically (over) damped (as can happen with rigid body modes), then the
rotor will not have a resonance on that mode, and the 1X vibration associated with the
mode will not be visible during startup. This is the case for the machine in Figure 5.
Orbits
If the vibration at the measurement plane is dominated by fluid-induced instability,
then the direct, unfiltered orbit will be predominately forward and circular (Figure 5).
Orbits that are filtered to the instability frequency will always be approximately circular
and forward.
9
f=
The behavior of the Keyphasor dots will depend on the relationship of the
subsynchronous frequency to running speed (the Keyphasor trigger frequency). In
general, the number of Keyphasor dots visible is related to the denominator of the nearest
subsynchronous integer ratio. For subsynchronous frequencies near 1/2X, two Keyphasor
dots will be visible. If the subsynchronous frequency is slightly below 1/2X, then the
Keyphasor dots will slowly drift in a direction opposite to rotation. If the subsynchronous
frequency is slightly above 1/2X, then the Keyphasor dots will slowly drift in the same
direction as rotation. Vibration near 1/3X will produce a set of three Keyphasor dots in
the orbit that behave in a similar way. Vibration near 2/5X (0.4X) will produce an orbit
with 5 Keyphasor dots.
When the subsynchronous vibration is not near an integer ratio, the Keyphasor dots
will tend to form a chaotic pattern consisting of a great many dots (Figure 6).
Note that, under the right circumstances, rub will produce subsynchronous vibration
at a pure integer ratio with locked Keyphasor dots. These dots will not drift around the
orbit with time and will tend to stay in the same location. Because rub produces integer
ratio subsynchronous vibration frequencies (such as 1/2X), Keyphasor dots from a
subsynchronous rub orbit will form a locked integer set. This is a very powerful tool for
discriminating between fluid-induced instability and rub. Locked Keyphasor dots imply
rub, while moving Keyphasor dots imply fluid-induced instability.
Whip orbits, because of the lower subsynchronous frequencies at which it usually
occurs, are more likely to show chaotic Keyphasor dot behavior than whirl orbits.
If the vibration at the measurement plane contains a mixture of 1X and
subsynchronous vibration, then the orbit will be more complex in shape. The
subsynchronous vibration will cause the orbit to continually change shape, but the motion
of the Keyphasor dots (for
frequencies close to an integer Radial
multiple) will still tend to migrate in Load
a small circle (Figure 9 below). Direction
High
Speed
Average Shaft Centerline Position
In classic fluid-induced
instability, the journal will move
about the center of the bearing at a
subsynchronous frequency in a Whip
forward, circular orbit. As the rotor
orbit grows larger in whirl or whip
Low
and begins to move around the Speed
bearing clearance, the average
eccentricity ratio will begin to
approach zero. That is, the average
shaft centerline position will
Figure 7. Average shaft centerline plot showing the
approach the bearing center (Figure transition from stable behavior (black) to fluid-induced
7). Thus, it can be very useful to instability whip (blue) inside the bearing for the data
correlate the onset of shown in Figure 6. The dashed circle shows the bearing
subsynchronous vibration with boundary. As the instability develops, the average
movement of the shaft centerline eccentricity ratio in the bearing approaches zero.
10
f=
11
f=
machine operating condition to a point farther away from the surge point. If the problem
is rotating stall, it should disappear. (An additional complication can result if a broken
part is obstructing the flow path in a compressor. Then it is possible to trigger flow
instabilities when the compressor is operating far from the surge point.)
Rub
Rub can also produce subsynchronous vibration. However, there are significant
differences between the subsynchronous vibration due to rub and that due to fluid-
induced instability.
Fluid-induced instability tends to produce a predominately forward subsynchronous
precession of the rotor. However, the subsynchronous vibration due to rub usually has
significant reverse components. Also,
subsynchronous rub will lock to a
frequency that is a pure integer ratio Fluid Instability Rub
such as 1/2X, 1/3X, 2/5X, or, more
rarely, 2/3X, 3/4X etc. It is improbable
that fluid instability will occur at exactly
an integer ratio like these. If the only
tool available for diagnosis is a simple
spectrum (half spectrum), then, because
of the limited resolution and lack of
precession information of the spectrum,
it can be very difficult to tell the
difference between fluid instability and
rub.
Fortunately, there are better tools Figure 9. Direct orbits showing a mixture of 1X and
subsynchronous vibration for eight shaft revolutions.
available. A direct orbit with a The fluid-induced instability frequency is slightly less
Keyphasor display is the best tool for than 1/2X in whirl (the orbit is from a location some
determining whether or not the distance from the source), while the rub frequency is
frequency is a pure integer ratio (Figure exactly 1/2X. In the instability orbit the Keyphasor
9). If the Keyphasor dots steadily dots slowly migrate against rotation (black arrows) in
a circular path (red), while the rub orbit dots are
change position in the orbit from one locked in place. See the text for more information.
vibration cycle to the next, then the
12
f=
13
f=
1 K
Ω TH = (9)
λ M
From this expression, it can be seen that decreasing λ, the fluid circulation term, will
increase the Threshold of Stability. Increasing the rotor system spring stiffness, K, will
also increase stability. Reducing the rotor mass, M, while theoretically helpful, is not
practical. The following discussions will address the practical aspects of changing λ and
K in real rotor systems.
14
f=
Pressurized
Fluid
Pressurized
Fluid
15
f=
external and internal alignment of the machine should be checked. Correct alignment
should result in properly loaded bearings.
At the design level, fluid-film bearings in a machine should be designed with an
adequate load. Over designed bearings could result in an fluid-induced instability
problem.
16
f=
References
1. Muszynska, A., "One Lateral Mode Isotropic Rotor Response to Nonsynchronous
Excitation," BRDRC Report No. 4, 1991, pp. 1-31; also Proceedings of the Course on
Rotor Dynamics and Vibration in Turbomachinery, von Karman Institute for Fluid
Dynamics, Belgium, 21-25 September 1992. MachineLibrary.
2. Bently, D. E., Hatch, C. T., "Root Locus and the Analysis of Rotor Stability
Problems," Orbit, BNC, v. 14, No. 4, December 1993. MachineLibrary.
3. Hatch, C. T., Bently, D. E., “Moment Equation Representation and Stability Analysis
of a 1-CDOF Overhung Rotor Model With Fluid Bearing and Gyroscopic Effects,”
BRDRC Report 8, 1995. MachineLibrary.
4. Bently, D., Hatch, C., Jesse, R., Whiteley, J., "Dynamic stiffness in whirl and whip,"
Orbit, BNC, v. 19, No. 1, March 1998.
17