Technical Brief:

Natural Frequency Whirl Instability

in an Industrial Gas Turbine
Chris White, BSc, MEng
Principal Engineer, Rotating Equipment Reliability
Wood
 The machine
• Gas turbine driven compressor set ‘A’ (1 of 3)
• Engine is a two-shaft 70 MW industrial gas turbine, bearing layout
as shown below
Shaft / bearing
layout

Gas gen Power

Compressor turbine turbine

Gas generator
 The issue
• Bearing 1 failed after 16,000 hours operation
• Vendor replaced Bearing 1 in situ
• Vibration returned to normal, operation continued, but:
• At 17,500 hrs, vibration at Bearing 2 stepped up to 50µm / 2 mils pp,
and continued to trend up
• Vendor recommended to remove engine from service for off-site
overhaul, operator requested a second opinion

Shaft / bearing arrg’t.

1 2 3 4 5
 Problem narrative
Shortly after Bearing 1 replacement, Bearing 2 exhibited increasing
vibration amplitude trend (from 16,600 – 17,200 hours), then:
• An unexplained step change back to normal without intervention
• At around 17,500 hrs, high vibration returned at Bearing 2
2.0 Brg 2 increasing Return to
Brg 2 mils p-p

trend elevated
1.5
Human-machine levels
interface display – 1.0
Bearing 2 vibration
amplitude trend 0.5
Step change to normal
0.0

17.1
16.2

16.3

16.4

16.5

16.6

16.7

17.0

17.2
16.9

17.3
16.8

17.4

17.5
Engine kHours
 Data acquisition
• Human-machine interface only records overall vibration levels
• Detailed vibration data from protection panel is not archived
• Vendor attended site and acquired run-up data from gas generator
shaft cranking speed of 3,300 RPM (not 0 RPM), using a
multichannel dynamic signal analyser
• Vendor analysis report recommended withdrawal from service;
data was later passed on to Wood’s vibration, dynamics and noise
team for further analysis
Multichannel
dynamic signal
analyser
 Data review – start-up amplitude trends
• At gas generator cranking speed of 3,300 RPM, large amplitude
fluctuations observed at Bearing 2 to 90 µm / 3.5 mils pp
• Settles down as speed ramps up, however:
• At 13,500 RPM, a
Overall vibration
step increase to amplitude trends
~100 µm / 4 mils Bearing 2
bearings 1, 2 and 3
pp at Brg 2 only
• No issues were Bearing 1
noted on the
power turbine Bearing 3
 Data review – waterfall amplitudes (Brg 2)
• High vibration amplitudes appear at around 13,500 RPM, reaching
their worst at 14,200 RPM
• Main frequency component was at ~0.27 – 0.3x turning speed
(65.0 -67.5 Hz), at over 75 µm / 3 mils p

5.0
1x
Displ (mils pp)

Bearing 2 4.0
start-up 3.0
waterfall at
onset 2.0

1.0

0.0
0 100 200 300 400 Freq (Hz)
 Data review – start-up waterfall plot (Brg 2)
• Waterfall plot (read from rear to front) shows the high amplitude
whilst cranking is at 1x, also suggests critical circa 4000 RPM
• At 13,500 RPM, we 3,330
see onset of a large

GG Shaft Speed (RPM)

3,330
4,870
sub-synchronous
7,226
frequency component 9,131
– initially, around 0.3 11,693
1x 12,679
orders of turning 13,376
speed, but drops ~0.3x
14,352
down a little Bearing 2 14,512

start-up 0 100 200 300

waterfall Freq (Hz)
 Data review – start-up waterfall plot (Brg 2)
• Closer examination of the onset of this sub-synchronous frequency
component indicates it tracks with running speed during the final
speed increase (~0.3x) 13,376

GG shaft speed (RPM)

13,376
• But then ‘locks on’ to 13,396
a frequency of 65Hz, 13,573
13,785
around 0.27 orders of 0.3x
13,998
turning speed 14,190
14,398
14,476
Zoom-in on 0.27x
14,478
bearing 2
start-up
0 100 200
waterfall
Freq (Hz)
 Data review - Brg 2 orbits at normal op speeds
• Two inner loops (forward precession), 3 phase markers per cycle
• Not a closed cycle because not an integer fraction

Bearing 2 Bearing 2
orbit plot at orbit plot,
instability instability
onset fully
developed

Brg 2 @ 13,555 RPM Brg 2 @ 14,190 RPM

(Onset) (Fully developed)
Data review – polar and bode plots (1x amp and phase)
Indicate critical at around 4,100 RPM / = 68 Hz, good match for
the 0.3 orders vibration which appears at 13,500 RPM
Brg 2X Bearing 2
polar plot
1x vibration phase lag

1x vibration amplitude

Bearing 2
bode plots
 Data review – bode plot comparison
Problem machine Bearing 2 bode plots Healthy machine Bearing 2 bode plots

When compared with a healthy example of the same model,

1st critical (as determined by phase change) similar but
amplification factor appears much higher
 Plain bearing
Discussion
• Symptoms bear similarities to oil whirl / oil whip, as DOR
presented in many vibration texts
• Oil whirl arises in plain journal bearings1 because
fluid pressures are asymmetric either side of the
minimum gap – higher on the upstream side
• Resultant bearing force has tangential, destabilising Min gap
component, normal to the bearing support force
Bearing
• Usually, shaft just runs a little to one side of the

component
pressure

Support
shaft centreline in the DOR, but … force

• It is this cross-coupling force which, if it overcomes Destabilising

loading and damping forces, results in oil whirl component

1and other fixed geometry bearing designs

 Discussion
• But the machine in question uses tilt pad journal
bearings (TPJBs), with 5 pads, in a load-on-pad (LOP)
Bearing
configuration pressure
• As each pad can pivot, it constantly adjusts angle to forces

maintain equilibrium pressure across its face;

resultant force always points at shaft centre
• Thus each pad tilts such that the leading edge is a
Tilt pad journal bearing
little further from the journal than the trailing edge
– as we typically see in the wear pattern
Equal pressure
• This bearing type cannot produce
cross coupling forces – but there
are other contenders…
 Diagnosis
• Forward acting cross coupled forces are also produced by aerodynamic
effects, and by fluid rotation at labyrinth seals. Normally, stiffness, loading
and damping is sufficient to suppress these and maintain stability.
• Whilst it cannot be discounted that one or more such forces had increased
here, experience and recent maintenance history suggest a simpler
explanation
• Vendor have advised that such symptoms are commonly seen when
bearing clearances have increased significantly due to wiping or other
damage (e.g., fulcrum wear)
• This results in increased lube end leakage from the bearing, significantly
reducing damping
 Discussion
• Unfortunately, start-up data provided was not captured from 0 rpm,
so we were unable to obtain meaningful Shaft Centre Line plots.
• Such plots enable us to determine shaft running position within the
bearing, and compare with clearance to determine eccentricity ε:
100
14,938 ε=d/c Bearing
Example
80
d = shaft centre to bearing centre Centre d
centre line 60 c = bearing radial clearance
X
X

plot (not from

40 Shaft
this case) Centre
20
ε →0 ε →1
0
Shaft
0 eccentricity
-40 -20 0 20 40
 Hypothesis
• Was Bearing 2 overloaded / damaged whilst running with damage
to Bearing 1? Brg 1
Bearing 2
Brg 2 Brg 3 eccentricity

Good ε ~0.5

ε →1
Brg 1
fail
Brg 2 overload / wear

Brg 1 ε →0
replaced

Brg 2 increased
clearances / unloaded
Data review – cranking speed (3300 rpm) waterfall
At cranking speed of 3,300 RPM, turning speed vibration exhibits
step changes between two clearly distinct ‘states’ at random
intervals (bistable system)
Step changes to
higher 1x
vibration

Bearing 2Y 5.0
Brg 2 mils p-p

cranking 4.0
speed 3.0
waterfall 2.0
1.0
0.0
 Data review – cranking speed (3300 rpm) orbits
• Left orbit shows several rotations at lower stable state,
followed by step to higher state
• Right orbit shows fully in higher vibration state, shape (abrupt
trajectory changes) suggests low friction rub events

Brg 2 lower state Brg 2 fully in higher

Bearing 2
stepping to Higher state Bearing 2
orbit plot
during state orbit plot in
change higher
vibration
state
 Bistable vibration at cranking speed
• We have not found references to such behaviour in the literature
• We do not have an explanation, other than to note that:
– Fluid film bearings are a ‘hardening non-linear system,’ i.e., due to
clearance, stiffness is relatively soft up to a limit of deflection, then
increases sharply at the ‘soft deflection limit’3.
– If clearances increase, this soft deflection limit will increase
significantly, and,
– Ehrich notes that non–linear systems may have multiple solutions.
More pragmatically, is further damage occurring at Bearing 2 during
cranking?
3 Ehrich Handbook of rotor dynamics
 Current status
• Case is not complete – unit not currently required to maintain
the current level of production, still in service as a poor spare
• Most probable diagnosis - likely bearing damage / increased
clearances at Bearing 2, possibly a consequence of earlier
Bearing 1 failure
• If required to be run prior to overhaul, recommended to
minimise cranking time and to reduce lube oil temperature to
lower end of tolerance – increasing viscosity in this manner
increases both effective bearing stiffness and damping
 Lessons learned
• It is essential when acquiring run-up / coast-down data to
ensure data is acquired from standstill, i.e., 0 RPM. This
allows determination of rotor running position from
centreline plots.
• When an initial bearing failure occurs, this may result in
overloading of other bearings on that shaft
• If continued operation with bearing damage is essential,
temporary improvement of vibration amplitudes may be
achieved by lowering lube oil temperatures
Thank you
Questions?