1. CASE HISTORY
CASE HISTORY
PAY B A C K P R O F I L E :
One-Shot Balancing For A Gas Turbine
How collaboration between OEM, End User, and
the Bently Nevada ® team saved over 1MM USD
Editor’s Note: Mr. Foiles, a rotor dynamics specialist with GE Energy at the time the
events in this story took place, now works for BP. This case history transpired in early
York Lee
Field Services Manager, Bently Nevada® Products 2001, nearly a year prior to the acquisition of Bently Nevada by GE. Although the case
GE Energy history describes the remarkable results achieved by cooperation between the
york.lee@ge.com customer, the turbine manufacturer (in this case, GE), and Bently Nevada, this level of
OEM collaboration was not then – or now – intended to be unique to GE. The Bently
William C. Foiles Nevada team continues to work in the strictest confidence with all OEMs, both in the
Rotating Equipment Specialist supply of condition monitoring products and in the application of machinery
BP
diagnostics expertise. We welcome the opportunity to publish additional articles where
bill.foiles@bp.com
collaboration with any of our OEM customers produced favorable results for all parties.
Introduction
This case history chronicles a machinery balancing job at a large Asian
chemical complex, demonstrating how collaboration between the customer,
the machinery OEM, and Bently Nevada service personnel achieved sub-
stantial improvements in the cost and time required to execute this service.
The Setting
The chemical complex is fully integrated with an adjacent refinery, pro-
ducing a combined total of over one million tons per year of propylene,
ethylene, and other petrochemical derivatives. The world-scale operation
also contains an integrated cogeneration facility for powering both the
refinery and chemical complex.
Unlike many cogeneration installations where electricity is considered the
primary “product” and steam is considered a byproduct, this petrochemi-
cal complex’s use of cogeneration is just the opposite: steam is the essential
product, required by the hydrocarbon cracking process; electricity, in this
instance, is the byproduct.
This explains why the economics and machinery objectives that drive a
cogeneration process will often differ from the power generation sector to
the petrochemical sector. When electric power is the primary product,
machinery efficiency is paramount because competitive pressures focus on
the lowest cost of generation. In contrast, the continuous processing indus-
tries, such as this chemical complex, have processes with enormous down-
time costs – often millions of dollars per day. When a cogeneration process
cannot run, steam cannot be produced, and the multi-million-dollar-per-
day petrochemical process that relies upon the steam must likewise stop.
[Vol.25 No.1 2005] ORBIT 5

2. CASE HISTORY
THIS END USER’S PRACTICE IS TO HAVE
THEIR GAS TURBINES RUN WITH VERY LOW VIBRATION LEVELS.
VIEW AS SEEN FROM DRIVER END 4YD 4XD
(UCZ-710) 4 (UCZ-709)
3YD 3XD
(UCZ-707) 3 (UCZ-708)
LOW-SPEED SHAFT Kφ (UCZ-720)
7YD 7XD 8YD 8XD
(UCZ-715) 7 (UCZ-716) (UCZ-701) 8 (UCZ-702)
GAS TURBINE
GEARBOX
GENERATOR
1 HIGH-SPEED SHAFT Kφ (UCZ-721)
2YD 2XD
1XD 1YD 2 (UCZ-703)
(UCZ-704)
(UCZ-706) (UCZ-705)
5YD 5XD
(UCZ-712) 5 (UCZ-711)
HIGH SPEED GEAR SHAFT 6YD 6XD
(UCZ-714) 6 (UCZ-713)
LOW SPEED GEAR SHAFT
GEAR END VIEW AS SEEN FROM DRIVER END
MACHINE TRAIN DIAGRAM FOR GAS TURBINES SHOWING TRANSDUCER ARRANGEMENT | FIG. 1
The cost of this lost production 3000 rpm generators through tical design. The gearbox input
far eclipses machinery efficiency speed-reducing gearboxes. The shaft utilizes offset-halves radial
concerns. As a result, machinery machines are rated at 65MW and bearings.
reliability – not efficiency – is para- operate at 5,230 rpm. There are All radial bearings are fitted with
mount. eight radial bearings in each orthogonal (X-Y) proximity
machine train. The gas turbines probes. Two Keyphasor® (Kφ)
Machinery each utilize two tilting pad radial phase reference transducers are
At the heart of the cogeneration bearings of four-segment, load- installed on each train; one on
facility are two identical GE between-pad design. Radial bear- the gas turbine drive end and one
MS6001FA (“6FA”) single-shaft ings on the generator and gearbox on the generator drive end. See
industrial gas turbines, driving output shaft are sleeve-type, ellip- Figure 1.
6 ORBIT [Vol.25 No.1 2005]

3. CASE HISTORY
POINT: UCZ-704 55° Left 1X
From 15FEB2001 20:00:00 To 14MAR2001 19:00 Steady State
21FEB2001 07MAR2001
19:00 19:00
90
PHASE LAG:
30 deg/div
270
90
150
10 µm pp/div
AMPLITUDE:
TREND OF 1X VIBRATION
100
AMPLITUDE AND PHASE AT
50
#2 BEARING Y-PROBE PRIOR
0 TO CORRECTIVE BALANCING
19:00 19:00 | FIG. 2
21FEB2001 07MAR2001
TIME: 1 Day/div
Monitoring Systems A High Vibration Problem the end user’s 55 µm alert levels
The end-user chose Bently Nevada The cogeneration units were in the (see Table 1 on page 10). Figure 2
products and services for protect- process of being commissioned in shows a one-month trend of 1X
ing and managing the critical late 2000 / early 2001. This end amplitude and phase from the #2
machinery in their facility. Data user’s practice is to have their gas bearing Y-probe. Notice that the
Manager® 2000 (DM2000) soft- turbines run with very low vibra- phase is stable and the amplitude is
ware is used for mechanical condi- tion levels – generally well below approximately 110 µm, twice the
tion monitoring and Bently the alarm levels discussed in ISO alert alarm value established by the
PERFORMANCE™ software is Standard 7919-4. As such, radial end user.
used for thermodynamic perform- vibration alarms for their proxim- Data obtained during test stand
ance monitoring. The systems are ity probes were set at 55 µm (2.17 commissioning, as well as transient
tightly integrated with the plant’s mils). While one unit (train B) was and steady-state data obtained at
process control system, allowing running very smoothly, the other site, suggested unbalance as the
correlation of machinery condition unit (train A) was exhibiting vibra- most likely source of the high vibra-
data with process data. tion amplitudes that – while still tion. After conferring with GE’s on-
The end-user also has a comprehen- within standard acceptance criteria site technical advisor, the end user
sive services agreement covering – were higher than the end user’s enlisted the assistance of a Bently
product repair and support, system preference. Specifically, by March Nevada machinery diagnostics
integration support, and machin- 2001, three of the four proximity engineer to confirm unbalance as
ery diagnostics support via 24/7 probes installed on the #1 and #2 the source of the problem and to
call-out to the Bently Nevada serv- bearings were consistently showing assist in balancing the machine. A
ice team at a nearby field office. vibration amplitudes well above decision was made to perform a
[Vol.25 No.1 2005] ORBIT 7

4. CASE HISTORY
2-plane balance on the gas turbine, dynamics engineers were brought regarding their machines. The
and the plant’s machinery engineer into the discussion, and their OEM possesses deep knowledge
informed his management and involvement was pivotal. that may simply be unavailable any-
operations personnel that correct- where else, and should be consulted
ing the unbalance would take 3-4 Consulting the Experts whenever possible. Fortunately, in
days and multiple starts of the tur- While Bently Nevada field engi- this case, the relationship between
bine. A location and weight for the neers are “OEM-neutral” in their the OEM (GE), the end user, and
initial trial balance was proposed approach to diagnosing and cor- Bently Nevada personnel was that
(but not actually installed). recting machinery problems, they of a team tasked with solving a
However, as we will discuss next, it also know that there is no substi- problem, and the parties were all
was at this juncture that GE’s rotor tute for consulting the OEM able to work cooperatively.
Why Both Proximity and Velocity Transducers are Important
It is noteworthy that the gas turbines in this case history had proximity probes and casing-mounted
velocity transducers installed. The velocity transducers were mounted vertically on the #1 and #2 bearings
of each gas turbine, while the proximity probes were arranged in an X-Y configuration as shown in Figure
1. It is customary for GE to install both types of transducers on many of their industrial gas turbines, and
the authors strongly advocate this as a best practice for any gas turbine (such as the 6FA) that exhibits
compliant casing and support structures [1,2].
The shaft-relative measurements provided by proximity probes are typically more sensitive to rotor-
related vibration problems such as imbalance, rubs, misalignment, and bearing instabilities. In contrast,
casing-mounted transducers are typically more sensitive to problems originating in the casing, sup-
ports, and piping. Best practice for many industrial gas turbines – particularly those that have very large
frames – is to use both shaft-relative (i.e., proximity probe) and casing-mounted (i.e., seismic velocity)
transducers. In this particular case history, the data from the proximity probes suggested that the unbalance
was more pronounced than would have been concluded by looking merely at the velocity transducers.
Also, the value of velocity transducers is in understanding the relationship between casing motion and
shaft-relative motion. Recall that proximity probes measure the relative motion between their mounting
location (often, the bearing housing) and the shaft. If the housing is quite stiff, casing motion will be
negligible and does not generally factor into the diagnostics of the machine for activities such as bal-
ancing and malfunction detection. However, if this motion is appreciable – as is the case on a 6FA – it
cannot be ignored. The authors are familiar with the use of Bently Nevada dual probe monitors on large
steam turbines, which allow the signals from proximity probes and bearing-mounted seismic trans-
ducers to be vectorially combined for shaft relative, casing absolute, and shaft absolute signals. In our
opinion, these monitors – while historically used only on large steam turbines – should also be considered
on a case-by-case basis for use on certain industrial gas turbines as they provide many diagnostic and
machinery protection benefits.
8 ORBIT [Vol.25 No.1 2005]

5. CASE HISTORY
GE’S ROTOR DYNAMICS ENGINEERS WERE
BROUGHT INTO THE DISCUSSION, AND
THEIR INVOLVEMENT WAS PIVOTAL.
As GE’s rotor dynamics experts in ginal unbalance state. Then, the try and balance the machine on the
Atlanta became aware that the end- machine is stopped, a test mass is first run, it is merely to quantify
user was proposing to field balance added at a known location, the how the machine responds to the
the machine, they were confident machine is restarted, and the result- addition of a known balance mass.
this could be accomplished success- ing vibration is measured. This is
fully with their assistance, but there known as the trial run. With this Balancing With Prior Data
were several important issues that information, the machine’s re- Influence Coefficients are a charac-
had to be considered: sponse to the addition of mass (i.e., teristic of the machine. They do not
its Influence Coefficients) can be change unless something other
r 6FA machines are balanced
computed, and the amount and than unbalance is wrong – such as
during factory tests; however,
location for the final corrective bal- a shaft crack – altering the relation-
GE had never performed an
ancing weight is determined. The ship between rotor excitation and
in-situ (i.e., field) balance of the machine is stopped once again, the rotor response (i.e., the Transfer
6FA design. balance mass is added, and the Function). When a rotor’s Influ-
r The unit was still under machine is restarted. This is known ence Coefficients are already
warranty. as the “correction run.” known, balancing can theoretically
Two-plane balancing is a similar be accomplished in a single “shot”
r In order to meet the end user’s
process, but requires two trial runs, since the rotor’s response to the
objectives of minimal starts/stops
since weight must be added inde- addition of weight has already been
and the fewest possible runs to
pendently at each balance plane to established.
balance the machine, prior see the response, allowing compu- Unfortunately, while GE had
knowledge of the Influence tation of two Influence Coefficients access to a database of Influence
Coefficients would be helpful. instead of one. Ideally, a minimum Coefficients for similar 6FA units,
Without this information, data of three runs (two trial and one they did not have the Influence
would have to gathered empir- correction) will be required; how- Coefficients for the particular unit
ically at the site by starting ever, machines do not always in question. Nor did the database
and stopping the machine mul- respond exactly as anticipated to contain field data from a unit that
tiple times and installing trial the addition of balance masses. was coupled to a gear. However, all
balance masses. Consequently, more runs are some- was not lost. It was reasoned that by
times required to obtain the desired consulting this database of rotor
Balancing Without Prior Data results. dynamic response data from a pop-
Single-plane balancing normally Also, without prior knowledge ulation of similar 6FA machines,
takes a minimum of two “runs” – of the machine’s Influence Co- it may be possible to statistically
one “trial run” and one “correction efficients, the initial amount of the determine the approximate Influ-
run.” First, the vibration is meas- required correction is – at best – an ence Coefficients for the rotor
ured with the machine in its ori- educated guess. The goal is not to under consideration [3].
[Vol.25 No.1 2005] ORBIT 9

6. CASE HISTORY
TABLE 1 | SUMMARY OF VIBRATION AMPLITUDES
AT GAS TURBINE BEARINGS BEFORE AND AFTER 2-PLANE BALANCING
OVERALL UNFILTERED 1X FILTERED
AMPLITUDE AMPLITUDE AND PHASE
MEASUREMENT LOCATION (µm, pk-pk) (µm, pk-pk – degrees)
Alert Before After Before After
Level Balancing Balancing Balancing Balancing
Bearing #1 Y-axis
(Gas Turbine Non-Drive End) 55 55.5 14.7 49.6 ∠ 63° 6.71 ∠ n/a
Bearing #1 X-axis
(Gas Turbine Non-Drive End) 55 41.1 9.7 36.0 ∠ 153° 2.12 ∠ n/a
Bearing #2 Y-axis
(Gas Turbine Drive End) 55 116 29.9 111 ∠ 181° 21.9 ∠ 268°
Bearing #2 X-axis
(Gas Turbine Drive End) 55 97.7 27.1 92.4 ∠ 266° 19.6 ∠ 11°
A linear regression of this data was other methods of selecting initial machine had been allowed to run
performed, and the expected Influ- trial weights. In contrast, the best- under steady-state conditions at a
ence Coefficients were statistically case scenario would be to balance 50MW load for a minimum of nine
determined. While it was antici- the rotor in a single run by placing hours before data was collected,
pated that these Influence exactly the right weights in exactly allowing any thermal transients to
Coefficients would be close, they the right locations on the first settle out and to help ensure that
were not expected to be exact, since attempt. Realistically, it was data was collected under similar
they were based on a population of expected that the actual result operating conditions.
similar rotors – not the particular would fall somewhere between The results exceeded everyone’s
rotor in question. these two extremes, particularly expectations, substantially reducing
Armed with these “expected” since the population of rotors in the the vibration amplitudes measured
Influence Coefficients, calculation GE database for which Influence at each proximity probe on the #1
of the corresponding “expected” Coefficient data was available was and #2 bearings as summarized
balance correction weights for each very limited. in Table 1.
of the two planes was straight for-
ward. The calculated weights would Results Figure 3 provides another view of
be used for the initial run. The The balance masses were placed as the before/after results, using orbit
worst-case scenario was simply that predicted by the model and the plots obtained from DM2000.
the theoretically determined machine was restarted. Data was Another important result of the bal-
weights – if not exactly correct – obtained immediately prior to and ancing job is conveyed in Figure 4,
would at least be less arbitrary than after balancing, and in each case the a continuation of the trend plot of
IT MAY BE POSSIBLE TO STATISTICALLY DETERMINE THE APPROXIMATE
INFLUENCE COEFFICIENTS FOR THE ROTOR.
10 ORBIT [Vol.25 No.1 2005]

7. CASE HISTORY
BEARING #1 BEARING #2
Y: UCZ-705 135° Right VECTOR: 49.6 µm pp 63° Y: UCZ-704 55° Left VECTOR: 111 µm pp 181°
X: UCZ-706 135° Left VECTOR: 36.0 µm pp 153° X: UCZ-703 35° Right VECTOR: 92.4 µm pp 266°
02MAR2001 20:37:23 Delta Time 1X COMP 02MAR2001 20:37:23 Delta Time 1X COMP
UP UP
BEFORE BALANCING
10 µm/div ROTATION: Y TO X (CW) 5233 rpm 10 µm/div ROTATION: Y TO X (CW) 5233 rpm
Y: UCZ-705 135° Right VECTOR: 6.71 µm pp NA Y: UCZ-704 55° Left VECTOR: 21.9 µm pp 268°
X: UCZ-706 135° Left VECTOR: 2.12 µm pp NA X: UCZ-703 35° Right VECTOR: 19.6 µm pp 11°
16MAY2001 05:20:00 Delta Time 1X COMP 16MAY2001 05:20:00 Delta Time 1X COMP
UP UP
BELOW MIN AMPLITUDE
AFTER BALANCING
10 µm/div ROTATION: Y TO X (CW) 5237 rpm 10 µm/div ROTATION: Y TO X (CW) 5237 rpm
1X FILTERED, COMPENSATED ORBIT PLOTS CONTRASTING VIBRATION AMPLITUDES
BEFORE AND AFTER 2-PLANE BALANCING | FIG. 3
[Vol.25 No.1 2005] ORBIT 11

8. CASE HISTORY
THE SAVINGS ACHIEVED BY BALANCING IN A SINGLE-SHOT
VERSUS MULTIPLE STARTS AND STOPS WERE WORTH “…WELL OVER A MILLION DOLLARS.”
Figure 2. Here, we see not only the state value of approximately 20 µm. turbine, the effect of corrective
dramatic reduction in 1X ampli- The balancing calculations consid- weights be observed only after the
tude achieved after balancing, but ered this thermal transient effect, unit has been given enough time to
we can also observe the effect that and the corrective weights were cho- stabilize thermally and only when
the corrective balancing had regard- sen to optimize the results expected all other pre- and post-balancing
ing the machine’s thermal transient. during both steady state and ther- operating conditions have been
Notice that prior to balancing, the mal transient operating regimes. made as consistent as possible. This
thermal transient (see reference 2 A final note regarding Figure 4: The helps ensure that changes in vibra-
for additional information) caused large (approximately one month) tion are truly due to balancing –
the 1X vibration amplitude to gap occurring between March 13 not other factors. For a large gas
climb to approximately 140 µm, an and April 4 is due to other work turbine, this can take many hours
undesirably high level for the end being done in the plant that and explains why multiple balance
user. During the re-start immedi- required the unit to be off-line. It runs are undesirable. Not only do
ately after balancing, the thermal was not possible to run the unit at multiple starts detract from the life
transient can again be observed, but speed and load until this other of the hot gas path components,
now the vibration never exceeds 50 work had been completed. It is they can also add several days to the
µm and settles rapidly to a steady- essential that when balancing a gas balancing job.
POINT: UCZ-704 55° Left 1X
From 15FEB2001 20:00:00 To 07APR2001 04:40:00 Steady State
21FEB2001 07MAR2001 21MAR2001 04APR2001
19:00 19:00 19:00 19:00
90
PHASE LAG:
30 deg/div
270
Thermal transient during
90 startup before balancing.
Note 1X amplitude of
approximately 140 µm.
150
10 µm pp/div
AMPLITUDE:
Thermal transient during
100 startup after balancing.
Note 1X amplitude does
50
not exceed 50 µm.
0
19:00 19:00 19:00 19:00
21FEB2001 07MAR2001 21MAR2001 04APR2001
TIME: 1 Day/div
TREND PLOT OF 1X AMPLITUDE AND PHASE SHOWING DRAMATIC REDUCTION IN BOTH STEADY-STATE AND
THERMAL TRANSIENT VIBRATION AT #2 BEARING FOLLOWING CORRECTIVE BALANCING | FIG. 4
12 ORBIT [Vol.25 No.1 2005]

9. CASE HISTORY
THE RESULTS EXCEEDED EVERYONE’S EXPECTATIONS.
A Satisfied Customer alert operators to machinery abnor-
Commenting on these results, the malities that, if left unchecked, can
customer’s on-site machinery engi- lead to expensive failures of equip-
neer was extremely pleased with the ment and process interruptions. It
outcome, noting that the savings also underscores the use of these
achieved by balancing in a single- systems in correcting problems and
shot versus multiple starts and stops in verifying that corrective actions
were worth “…well over a million produced the intended results.
dollars.” OEMs possess deep knowledge of
their machines and including them
He was quick to reiterate the
in discussions – as shown by this
importance of the cogeneration
case history – made the difference
process for reliably producing
between the ability to balance in a
steam 24/7, 365 days a year. “The
single attempt and a more tradi-
number of starts on our gas tur-
tional balancing exercise requiring
bines are very, very low because
several more days, several more
they are base loaded, and our
machine starts, and many, many
emphasis on reliability. Certainly
dollars.
less starts equals better hot compo-
nent life, but it also equals fewer References:
interruptions to our petrochemical
[1] M. DIMOND, Vibration Charac-
process and that translates to mil- teristics of Industrial Gas
lions of dollars for us.” He concluded Turbines, ORBIT magazine, Vol.
21 No. 3, September 1998, pp.
by noting another important out- 18-21.
come. “It was also a great boost to
[2] A.W. VON RAPPARD and A.T.
our credibility with management. HECKMAN, Best Vibration
We did what we said we could do, Monitoring Practice for Large
in much less time than we prom- ABB Gas Turbine Protection and
Machinery Management, ORBIT
ised, not by being lucky, but by magazine, Vol. 19 No. 3, Third
being smart and using all of the best Quarter 2000, pp. 10-13.
resources at our disposal.” [3] L.-O. LARSSON, On the Deter-
mination of the Influence
Coefficients in Rotor Balancing
Summary Using Linear Regression
This case history demonstrates the Analysis, in Proceedings of
Conference on Vibrations in
value achieved when all parties – Rotating Machinery, Cambridge,
customer, machinery OEM, and England, 1976, Institute of
Mechanical Engineers, pp. 93-97.
Bently Nevada – work coopera-
tively to solve problems. It under-
scores the value of continuous
condition monitoring systems to
[Vol.25 No.1 2005] ORBIT 13