EXPERIENCES AND APPLICATIONS OF API 617

FULL STABILITY ANALYSIS (LEVEL-II) ON CENTRIFUGAL COMPRESSORS

by
Jiming Li
Product Engineer
Elliott Turbomachinery Co., Inc., Ebara Group
Jeannette, Pennsylvania
Pranabesh De Choudhury
Engineering Manager
Pran RDA Consulting Inc.
Greensburg, Pennsylvania
Mike Sharples
Consulting Production Engineer
Unocal Corporation
Anchorage, Alaska
and
John Wright
General Manager of Engineering
Elliott Turbomachinery Co., Inc., Ebara Group
Jeannette, Pennsylvania

Professional Engineer in the State of Pennsylvania and is a

Jiming Li is a Product Engineer at Elliott member of ASME and STLE.
Company, in Jeannette, Pennsylvania. In
his present assignment, he has had re-
sponsibility for rotordynamics, torsional Mike Sharples is a Consulting Production
dynamics, vibration troubleshooting, de- Engineer with Unocal Corporation, in
velopment and application of advanced Anchorage, Alaska. While with Unocal, he
damper seals and bearings, and engineer- has worked 25 years in the North Sea, SE
ing support to the compressor product line Asia, and the USA, mainly being involved
and marketing. He has written several with the design, installation, and operation
technical papers for ASME journals and is of oil and gas production facilities. In his
a member of ASME. current position with Unocal Alaska, he
Dr. Li received a BSME degree (1982) from Zhengzhou Institute has supported various gas compression
of Technology, and obtained MSME (1995) and Ph.D. (1999) addition and modification projects.
degrees (Mechanical Engineering) from Texas A&M University. Mr. Sharples has a degree (Chemical
Engineering) from Canterbury University, New Zealand, and, prior
to working for Unocal, worked in New Zealand and Australia.
Pranabesh DeChoudhury has worked for
Elliott Company, in Jeannette, Pennsylva-
nia, for more than 32 years. As a Senior ABSTRACT
Consulting Engineer, his responsibilities To cover the remaining operational life of an offshore platform, it
included rotorbearing dynamics, bearing was necessary to go from a single stage compressor to a two-stage
design and analysis, torsional dynamics, compressor. The design of the new two-stage compressor required
blade vibration analysis, and troubleshoot- two different impeller arrangements to cover the range of
ing field vibration problems. He has suction/discharge pressures anticipated. One rotor has nine wheels
recently retired from Elliott Company and and another one has 11 by adding one impeller at the outboard ends
has started Pran RDA Consulting Inc. of both compression sections. With a rotor flexibility ratio around of
Dr. DeChoudhury obtained a BSME 2.5, the new replacement compressor is a borderline machine on the
degree (1963) from Jadavpur University, an MSME degree from API 617 (2002) stability screening map. The stability study indicated
Bucknell University, and a Ph.D. degree (Mechanical Engineering, that the compressor failed to meet the Level-I stability criteria at the
1971) from the University of Virginia. He has authored and coau- design conditions. A detailed stability investigation became
thored over 25 technical papers on various turbomachinery topics necessary though the machine was predicted to be stable based on an
and has been awarded several U.S. Patents. He is a registered original equipment manufacturer’s (OEM) empirical method.
35
36 PROCEEDINGS OF THE THIRTY-FOURTH TURBOMACHINERY SYMPOSIUM • 2005

An independent party was contracted to conduct a full stability

audit on both rotors. Compressor cross-coupling was evaluated
uniquely and lumped at each impeller location in the analysis
model. Additionally, the destabilizing forces from all labyrinth seals
in the compressor were calculated and included in the detailed
stability analysis. The study results indicated that both rotors would
most likely be susceptible to severe instability problems even using
a modified center labyrinth seal with a shunt-injection and swirl
brakes. Consequently, the effect of damper seals on the rotor
stability was further investigated independently. It was concluded
that a honeycomb seal could bring the log-decrement to a distinc-
tively positive value for the nine-wheel rotor, but not succeed in
raising the log-decrement to positive for the 11-wheel rotor.
Because of faster than anticipated declines in well deliverability,
it was decided to bypass the nine-wheel configuration and install
Figure 1. Platform C-12 Compression Train.
the 11-wheel configuration. The 11-wheel rotor used a modified
center labyrinth seal with a shunt/brake deswirl system. As a
Table 1. C-12 Replacement Compressor Design Conditions and
backup, the nine-wheel rotor and the inner casing were modified to
Features.
fit with a pocket damper seal. The 11-stage compressor was suc-
cessfully commissioned and has run well in the field for more than
one year. Overall shaft vibration level was low and the predicted 25MB4/5I 25MB5/6I
subsynchronous vibration was not observed. In this paper, a corre- 1st stage 2nd stage 1st stage 2nd stage
lation study was carried out and a comparison was made to discuss
how to calculate the Wachel number, and how to practically Inlet P (psi) 275 587 115 ~ 165 329 ~ 434
evaluate the contribution of the labyrinth seals and the gas deswirl Inlet T (° F) 70 90 70 90
devices in a detailed (API Level-II) rotor stability analysis.
Discharge P (psi) 597 1245 334 ~ 439 1065 ~ 1145
BACKGROUND OF COMPRESSOR Discharge T (° F) 200 227 234 ~ 260 264 ~ 307
REPLACEMENT PROJECT Gas Molecular Wt 16.25 16.25
Steelhead Platform is located in the Cook Inlet region of Alaska Flow (lbs/min) 3460 ~ 3910 1771 ~ 2400
and is operated by Unocal Alaska, part of Union Oil of California.
Primary function of the Steelhead Platform is to recover natural Design Speed (rpm) 9950 ~ 10240 10250 ~ 11620
gas from the Grayling Gas Sands (GGS) Reservoir. The platform Max. Power (Hp) 10700 10700
was installed in the mid 1980s, and started production in 1987, at
Max. Speed (rpm) 12000 12000
which stage there was a single 5000 hp gas compressor train (C-
12). During the early 1990s, a second 6500 hp gas compression Bearing Span (in) 73.34 73.34
train (C-13) was added to the platform, to provide redundancy and Flexibility Ratio 2.48 2.52
the soon to be required parallel gas production train. In 1997, the
original C-12 compressor package was removed and replaced with No. of Impeller 9 11
a 10,400 hp package comprising a gas turbine driving a single
stage 25MB4/4 compressor directly.
In 2002, after more than 15 years of gas production from the The compressor rotors are fitted with dry gas seals and
Steelhead Platform, Unocal Alaska was faced with optimizing supported on two conventional five-shoe, tilt-pad journal bearings.
project end of life gas recovery operations, and the C-12 replace- The diameter of the journal bearing is 5.0 inches and the bearing
ment project was initiated. The project goal was to attempt to cover load is applied in the direction between two bottom pads. The
remaining operations with a single compressor change, to detailed bearing geometry is provided in Table 2. Figure 2 is a
complete the work with minimum downtime, and to minimize the cross-section sketch of the 25MB5/6I inner casing to show the
risk of any post startup issues. Due to the continuous decline of the impeller arrangement and the axial locations of seals. Shaft seals
gas well pressures, it was necessary to optimize the design of the and impeller eye seals are of a typical straight-through labyrinth.
replacement compressor across multiple operating points. Abradable stationary seal materials are used with rotating labyrinth
The replacement compression train consists of a new back-to-back teeth and allow tighter seal clearances to achieve a higher effi-
centrifugal compressor directly driven by the existing gas turbine, as ciency. It is tolerable for the rotating teeth to wear into the
shown in Figure 1. Considering the variation in the production con- sacrificial stator material during transient periods without damage
ditions and the amount of gas to be transmitted, two exchangeable to the rotor. The back-to-back compressor has a long intermediate
inner casings were designed for the same vertical assembled com- center labyrinth seal between the two compression stages.
pressor outer casing. The first inner casing has an expendable Compared to other labyrinth seals in the compressor the center seal
two-stage rotor with four and five impellers in the first and second would sustain the largest differential pressure. The original design
stages, respectively. The second rotor has a total of 11 impellers by of the center seal is a type of teeth-on-rotor labyrinth seal, which
adding one impeller at the outboard ends of the compression stages. rides against an abradable stationary component with a smooth
Compared to the original back-to-back eight-wheel compressor, the bore surface.
maximum continuous operating speed (MCOS) is increased by 19
percent, but the machine power goes up only 3 percent because of the Table 2. Journal Bearing Parameters for C-12 Replacement
new design at a lower suction pressure and lower gas flow rate. On Compressor.
the other hand, the new compressor rotors have more impellers on
the shaft and are more flexible. For convenience the two new com- Diameter Bearing Assembled Diametral Pivot
L/D
pressor rotors are referred to as 25MB4/5I and 25MB5/6I in this (in) Preload Clearance (in) Location
paper, respectively. The main operating conditions for both rotors are
5.00 0.425 0.402 0.007 55%
given in Table 1 along with the compressor design parameters.
 EXPERIENCES AND APPLICATIONS OF API 617 37
FULL STABILITY ANALYSIS (LEVEL-II) ON CENTRIFUGAL COMPRESSORS

issues. The new compressor has smoothly run more than one year
with a low level synchronous vibration.

STABILITY DESIGN AND AUDIT

ON C-12 REPLACEMENT COMPRESSOR
According to the progress arrangement of the project, original
rotordynamics review on both rotors 25MB4/5I and 25MB5/6I was
Figure 2. Cross-Section Sketch of C-12 Replacement Compressor finished prior to the release of the API 617, Seventh Edition
25MB5/6I. (2002). After June of 2002, the customer decided to proceed with
the second rotor with 11-wheels first. The rotor stability was rein-
The 25MB4/5I compressor successfully passed its mechanical vestigated including the effect of labyrinth seal, deswirl remedy,
test at the manufacturer’s facility. However during the performance and damper seals. A stability study per API 617 Level-I require-
test with a partial load, a small subsynchronous vibration (0.04 ments was performed on the rotors as well for internal reference
mils peak-to-peak) showed up and the overall vibration level of the usage. Meanwhile the user of the compressor contracted a third
compressor was below 0.5 mils. As shown in Figure 3, the low party to conduct an independent rotor stability design audit, and
level 70 Hz subsynchronous peak existed for all operating points the manufacturer was responsible for providing rotor models and
away from surge line and its frequency was coincident with the other required information except for the aerodynamic excitation
first critical speed of the rotor. This phenomenon is not abnormal forces to enable the third party to conduct the full stability analysis.
in performance tests and field operations with a loaded centrifugal Original Stability Study on C-12 Compressor Rotors
compressor. Due to possible excitation sources such as normal
flow turbulence, pressure fluctuation, and noise, some low-level For the C-12 compression train, the molecular weight of the
forced vibration likely exists at rotor natural frequencies. Based on process natural gas is 16.25 and the expected average gas density
the aerodynamic excitation force calculated using the manufac- varies from 1.24 to 1.84 lb/ft3. The flexibility ratio CSR (MCOS/first
turer’s proprietary equation, rotordynamics analysis illustrated that rigid support critical speed) of the new rotor is about 2.5. Figure 4
the compressor would be stable at the full rated load. shows the rated design positions of the 25MB4/5I and 25MB5/6I of
the C-12 train in an API 617 Level-I screening map along with
several reference cases. These referenced compressors have been
operating well in the field. Apparently the new replacement rotors
are more flexible than the original eight-wheel compressor rotor by
increasing the CSR value from 2.0 to 2.5. The map also indicates
that the subject two C-12 rotors are still within the envelope of
previous design experience and applications of the OEM.
3.5

25MB5/6I
3.0
25MB4/5I Region B
2.5
CSR

Figure 3. Vibration Spectrum at Partial Load Performance Test.

2.0

However, the potential instability problem was still a customer Original Region A
concern. An independent rotor stability audit was performed in 1.5 25MB4/4I
accordance with the customer’s request. Aerodynamic destabiliz-
ing forces including Wachel numbers and all labyrinth seals were
considered in the full stability investigation. A unique procedure 1.0
0.0 1.0 2.0 3.0 4.0 5.0
was adopted to estimate the Wachel numbers. The results of the
Average Gas Density (lbm/ft^3)
study increased the concern related to an instability problem and
indicated that a special damper device such as a honeycomb seal Figure 4. Platform C-12 Compressor Positions on API Screening
had to be installed to improve the rotor stability. Available analyt- Map.
ical tools, application experience, and stability criteria were
thoroughly discussed by the manufacturer and the customer. It was In the original rotordynamics examination, the rotor unbalance
mutually agreed to install a hybrid shunt-brake system as a direct response, critical speeds, and separation margins related to the
counteraction by modifying the discharge volute. The objective of operating speeds were calculated per API 617, Sixth Edition, and
the change was to reduce gas swirl velocity in the leading portion the results were acceptable. The rotor stability was reviewed based
of the center labyrinth seal and consequently to reduce the destabi- on the manufacturer’s empirical method and criteria. To estimate
lizing cross-coupling force. the level of the load-dependent destabilizing force applied on the
However, due to well-pressure declines greater than forecasted, compressor rotors, aerodynamic cross-coupling stiffness was
it was decided to first install the 11-wheel bundle 25MB5/6I with predicted using the manufacturer’s proprietary equation in the
the added shunt-brake device in 2003. Meanwhile as a stability following form:
contingency plan per the customer’s request, the spare bundle
25MB4/5I was changed to fit with a center damper seal. Although Q = F ( Pi , Ti , Pd , Td , M w , N max , N rl ) (1)
two kinds of gas damper seals, a honeycomb seal and a pocket seal,
were available and both were analyzed as well, the pocket-damper The total cross-coupling stiffness was estimated based on com-
seal was used finally in the spare bundle. In September 2003, the pressor design and operating conditions including the maximum
replacement compressor 25MB5/6I was run at various points to operating speed (Nmax), the gas molecular weight (Mw), the inlet
check rotor stability and confirmed no subsynchronous vibration and discharge gas pressures (Pi and Pd), the inlet and discharge gas
38 PROCEEDINGS OF THE THIRTY-FOURTH TURBOMACHINERY SYMPOSIUM • 2005

temperatures (Ti and Td) of the compressor, and the rotor rigid minimum safety factor should not be less than 2.0. If the compres-
support critical speed (Nr1). In this paper, a cross-coupling stiffness sor falls into Region “B,” a more conservative safety factor of 10.0
of 18,200 lb/in and 15,300 lb/in was predicted for the subject is needed to meet the Level-I stability criteria.
25MB4/5I and 25MB5/6I rotors, respectively, using the above As shown in Figure 4, the two rotors of the C-12 compressor,
empirical equation. In general, this is an end-to-end approach to 25MB4/5I and 25MB5/6I, belong to borderline machines in the API
estimate the aerodynamic cross-coupling generated by all identi- screening map. A safety factor of 10.0 would be required for the
fied and unidentified destabilizing sources in the compressor. subject rotors to satisfy API Level-I requirements. To the authors’
Therefore, the effect of labyrinth seals on rotor stability was not knowledge, it is analytically almost impossible for such flexible
separately investigated originally. compressor rotors to sustain so large a stability safety factor without
The predicted total cross-coupling stiffness was evenly distrib- using an exceptional damping device. Following the procedure in
uted at impeller locations in the log-decrement calculation to the API Level-I stability study, the rotor stability (log-decrement)
evaluate the compressor stability. Table 3 gives the rotor basic and was still calculated at the cross-coupling excitation levels QA and
aero log-decrements predicted at the maximum operating speed of 2QA, and the predicted results are shown in Table 4. Clearly, both
12,000 rpm and the average bearing clearance. Although the second rotors did not have a satisfactory safety factor and even their log-
rotor 25MB5/6I had two more wheels than the first rotor 25MB4/5I, decrement could not reach the minimum acceptance value of 0.1 in
the expected cross-coupling was lower because of a lower pressure the minimum and nominal bearing clearance cases at the excitation
range. Consequently the predicted aero log-decrement was slightly level of QA. It could be concluded that both 25MB4/5I and
higher for 25MB5/6I. The predicted log-decrement values were 25MB5/6I rotors did not pass the API 617 Level-I stability criteria.
within the experience envelope and accepted by following empirical Note that the predicted aero log-decrements in Tables 3 and 4 are not
criteria. Note the nonsynchronous coefficients of the journal bearing comparable because these aero log-decrements were calculated with
were adopted with the OEM’s empirical method. People should be different bearing coefficients and aerodynamic excitation forces.
aware that the log-decrements are not directly comparable if
different types of bearing coefficient are used in rotor stability Table 4. Log-Decrements per API 617 Level-I Stability Study
analyses. Normally using synchronous bearing coefficient yields a Approach.
larger log-decrement at the same aerodynamic excitation level.
Minimum Average Maximum
Table 3. Rotor Stability/Log-Decrements Based on the OEM’s Total Bearing Bearing Bearing
Method. Case QA Clearance Clearance Clearance
(lb/in)
QA 2QA QA 2QA QA 2QA
No. of Total Aero Basic Aero
Rotor 25MB4/5 28,600 -.057 -.310 .012 -.253 .154 -.137
Wheels Q (lb/in) Log-dec Log-dec

25MB4/5I 9 18,200 0.191 0.061 25MB5/6 26,000 -.068 -.293 -.010 -.245 .110 -.145

25MB5/6I 11 15,300 0.191 0.094

It should be pointed out that the stability analysis approach of
API Level-I is conservative. Level-I study results do not directly
API 617 Stability Screening Study reflect the actual operating conditions and the real dynamic
Although manufacturers and users of turbomachinery have been behavior of the compressor investigated. Failing to meet the Level-
familiar with load-dependent instability problems in centrifugal I requirements does not mean an unacceptable stability design and
compressors for quite a long time, there was no uniform procedure just indicates a detailed stability investigation (Level-II) should be
for engineers to follow in compressor stability analyses prior to the performed on these rotors. Obviously, the rotordynamic force coef-
release of API 617, Seventh Edition (2002). The Level-I stability ficients of labyrinth seals in the compressor are essential and
study of API 617 provides a standardized method and the same critical for conducting a full rotor stability analysis. Like the man-
criteria for both manufacturer and users to perform a preliminary ufacturer’s method discussed above, the API Level-I stability study
evaluation on compressor stability. does not explicitly investigate the contribution from any labyrinth
As a practical procedure for industrial applications, the API 617 seal and deswirl stabilizing device either.
Level-I stability analysis sets up a guideline for what is the
An Independent Compressor Stability Design Audit
minimum analytical work to be done to ensure a rotordynamically
stable compressor. Based on a simplified conservative procedure After reviewing the available results of the rotor stability study
defined in API 617 Level-I, those compressors requiring a detailed on the C-12 replacement compressor, the compressor user decided
stability study will be identified. A modified Wachel equation is to contract a third party to perform an independent rotordynamics
provided in Level-I for the evaluation of the anticipated aerody- analysis. The investigation was focused on the stability design of
namic excitation QA. The Level-I stability study also requires using both nine-wheel and 11-wheel rotors. Based on the basic mass-
synchronous bearing coefficients if the rotor is fitted with tilt-pad elastic data transferred from the OEM, the lateral model was
journal bearings. If squeeze film dampers are used in the compres- rebuilt by including an additional stiffening effect to the shaft due
sor, their effect should be included in the Level-I preliminary to the impellers shrunk on the shaft. Bearing coefficients of the
stability analysis. The destabilizing contributions of labyrinth seals five-shoe, tilt-pad journal bearings were regenerated using a
will be further considered separately in the API 617 Level-II study different bearing code developed by the research facility of a major
when a detailed stability analysis is finally necessary. university. The calculated rotor unbalance response results and
According to the Level-I stability requirement, a minimum log- plots, including critical speed map, damped mode shapes, critical
decrement of 0.1 must be achieved at the extremes of journal speeds, and amplification factors, are close to the analysis results
bearing clearance with the anticipated cross-coupling excitation, previously provided by the manufacturer. Therefore the correct-
and certain stability margins are needed to account for any uncer- ness of the rotor models and analytical tools used for the
tainties in the modeling and analysis. The amount of the rotordynamics study was confirmed. However, there was an
aerodynamic cross-coupling to generate a zero log-decrement is apparent discrepancy in the evaluation results of the rotor stability
defined as Q0. In the API 617 Level-I stability study, the ratio of between the manufacturer and the consultant.
(Q0/QA) is called the safety factor of the compressor rotor stability. The independent stability study was conducted at the rated com-
For a compressor within Region “A” in the API screening map, the pressor operating conditions and the blueprint assembled clearances
 EXPERIENCES AND APPLICATIONS OF API 617 39
FULL STABILITY ANALYSIS (LEVEL-II) ON CENTRIFUGAL COMPRESSORS

of the journal bearings and the labyrinth seals. Like the manufac- stability, the log-decrement was calculated at 65 percent of the QT
turer’s approach, nonsynchronous bearing coefficients were adopted as well. It seems that the influence of the labyrinth seals on the
in the stability calculation. Given below is the well-known Wachel’s stability was limited because the predicted log-decrement was still
equation, which was used to estimate the overall aerodynamic cross- negative by removing 35 percent aerodynamic cross-coupling
coupling applied on the compressor rotor in the stability analysis. from the rotor models.

HP × M wt ⎛ ρd ⎞ Table 5. Predicted Cross-Coupling and Log-Decrements at the

Q = 6300 × ⎜ ⎟ (2) Rated Speeds.
D × b × N ⎝ ρs ⎠
Cross-Coupling Log-dec
Wachel’s cross-coupling equation resulted from correlation Log-dec Log-dec
Rotor Excitation QT w/o Seals
(100% QT) (65% QT)
studies of rotor instability experienced by centrifugal compressors (lb/in) (90% QT)
earlier and has taken into account all the destabilizing effects from 25MB4/5I 48,600 -0.150 -0.111 -0.013
impeller, labyrinth seals, oil seals, etc. Gas properties, horsepower
25MB5/6I 74,500 -0.384 -0.340* -0.200*
absorbed, and impeller/diffuser geometries should be known to use
this formula. To date, Wachel’s equation is one of most common * Actual calculation point was not available and the log-dec value was curve
fitted data.
formulae used by compressor users and consultants when the rotor
stability needs to be examined. However, there are various
Using shunt injection/swirl brake to retard the preswirl gas at the
approaches of estimating the aerodynamic cross-coupling (Wachel
center labyrinth seal entrance was considered and its influence was
number) in using this equation. The Wachel number can be calcu-
evaluated in the independent stability audit. The study results
lated wheel by wheel with local gas density ratio and local
indicated realistically it was not possible to bring the aerodynamic
horsepower transmitted at each impeller. On the other hand, a total
log-decrement to the positive for both rotors without a damper. A
cross-coupling can be first calculated and then distribute the total
further modification was proposed to replace the existing center
cross-coupling at individual impellers. The most conservative
labyrinth seal with a honeycomb damper seal. Table 6 theoretically
method in estimating the Wachel number is to use the last
shows the effect of the proposed honeycomb seal on the rotor
impeller/diffuser geometries with the overall compressor gas
stability. Unfortunately, the calculated rotordynamic coefficients of
density ratio and the entire power absorbed.
the damper seal and the labyrinth seals were not included in the
In the independent stability investigation, a unique approach
independent audit report. Based on the auditor’s judgment, a well-
was followed to calculate the aerodynamic cross-coupling. Based
designed honeycomb seal should be able to achieve a sufficient
on their experience, the consultant estimated a local Wachel
stability margin for the nine-wheel 25MB4/5I rotor. However, for
number wheel by wheel using a combination of the local wheel
the 11-wheel rotor, even using a honeycomb seal with shunt
horsepower and the overall compressor gas density ratio.
injection could not raise the stability to an acceptable level.
Following this approach, a total aerodynamic cross-coupling QT of
Exceptional countermeasures were predicted to be required to
48,600 lb/in was anticipated for the nine-wheel 25MB4/5I rotor stabilize the 25MB5/6I rotor.
and a larger total cross-coupling QT up to 74,500 lb/in was
estimated for the 11-wheel 25MB5/6I. The estimated Wachel Table 6. Stability Enhancement Study Results.
numbers at the impellers were used as a baseline to further perform
a detailed rotor stability analysis, which would include all Log-dec Log-dec
labyrinth seals in the compressor. The rotordyanmics force coeffi- Rotor using shunt/brake using Honeycomb seal
cients of the labyrinth seals were to be solved at the designed (90% QT) (90% QT)
conditions. To meet this end, the compressor manufacturer further
25MB4/5I -0.052 0.138
provided the third party the detailed gas properties at each
labyrinth seal along with the seal geometry data. Considerable 25MB5/6I -0.333* -0.006
effort was made to calculate the cross-coupling stiffness of the
labyrinth seals and complete the full stability study.
The independent full stability analysis was conducted at the A meeting was held in the manufacturer’s facility to fully review
nominal seal clearance only. Much more effort and engineering the rotordynamic audit results and discuss the potential counterac-
time would be required considering the variation of the assembled tions. Because the well declined very quickly, the user made a
seal clearance. Including impeller eye seals, shaft seals, buffer decision to proceed with the 11-wheel bundle 25MB5/6I first with
seals, and the center seal, there are total of 23 labyrinth seals in the the original modification proposed by the manufacturer. The
25MB5/6I casing and 19 labyrinth seals in the 25MB4/5I casing. A proposal of the OEM was to add a hybrid shunt/brake deswirl
thorough labyrinth seal analysis confirmed that the major seal system at the center labyrinth seal between the two compression
destabilizing cross-coupling was generated by the center labyrinth stages. The second rotor 25MB4/5I (nine-wheel bundle) was
seal, which has the longest axial length and sustains the maximum further modified to fit a center pocket damper seal for further
differential pressure of about 650 psig in the C-12 replacement stability enhancement and used as a stability contingency. The
compressor. Using a particular procedure allowed extraction of the labyrinth teeth on the rotating sleeves were machined off and the
cross-coupling of the labyrinth seals from the baseline Wachel original abradable seal ring was replaced to accommodate the
numbers to avoid duplication of the seal effect in the full stability pocket damper seal at the location of the original center labyrinth
analysis. For the C-12 replacement compressor, it was found that seal. This was a permanent change on the 25MB4/5I rotor, and the
the labyrinth seals contributed about 10 percent to the total aerody- rotor would not be exchangeable with another rotor without
namic cross-coupling. A 10 percent contribution from the labyrinth restaging and using new teethed center sleeves.
seals in a back-to-back compressor was considerably low and no
sound explanation was given by the auditor. It was claimed that the FIELD OPERATION AND RECONSIDERATION
labyrinth seal forces could provide around 35 percent of the overall OF THE STABILITY ANALYSIS
aerodynamic cross-coupling in other centrifugal compressors.
As shown in Table 5, rotor stability (log-decrement) was calcu- Operation Results on Site
lated at the excitation levels 100 percent and 90 percent of the QT In 2003, both the 25MB5/6I contract and 25MB4/5I spare inner
to examine the effect of the labyrinth seals. In order to anticipate bundles were transported to the offshore platform. Shown in Figure
the potential maximum effect of the labyrinth seals on the rotor 5 is the 11-wheel 25MB5/6I casing first installed and the original
40 PROCEEDINGS OF THE THIRTY-FOURTH TURBOMACHINERY SYMPOSIUM • 2005

eight-wheel inner casing/rotor was changed out from the C-12 stall in the compressor. Although the compressor was upset by the
compression train. In the startup check period, the new replace- rotating stall, the vibration peak at the rotor natural frequency was
ment compressor was run at various operation points to verify its still stable, which illustrated that the first forward mode of the
rotordynamic behavior. During the commissioning, the compressor compressor rotor was well damped.
reached a maximum operating speed of 11,090 rpm using the
maximum turbine power output. The rated pressure ratios of the
two stages and the design maximum discharge pressure were
achieved.

Figure 7. Compressor 25MB5/6I Vibration Spectrum near the

Surge Line.

After the commissioning was successfully performed, the C-12

compression train was brought inline to transmit 84 mmscfd
natural gas from the offshore platform at the design discharge
pressure. Both the user and the OEM were satisfied with the per-
formance and the vibration level of the new replacement
Figure 5. Platform C-12 Compressor 25MB5/6I.
compressor. Considering the exchangeability between the two
rotors in the future, the nine-wheel 25MB4/5I rotor/casing would
Figure 6 shows a typical peak-hold vibration spectrum of the
be sent back for restaging and change out the damper seal with the
compressor, which was measured in a field test period of 49
deswirl-remedied center labyrinth seal. Although the independent
minutes. Operations at various possible conditions indicated that
stability study seemed conservative and did not accurately predict
there was no concern about the rotor instability problem with this the stability in this application, their quantitative results did
11-wheel rotor. Note that this rotor was only fitted with a manu- encourage both the customer and the manufacturer to install the
facturer’s hybrid shunt/brake system and without an exceptional deswirl system in the compressor and helped in preventing the
damper. The rotor vibration was dominated by a small synchronous compressor from any potential rotor instability.
vibration from 0.4 ~ 0.5 mils. Like the phenomenon observed in The full rotordynamics audit could not accurately predict the
the partial load runs performed on the manufacturer’s test floor, a rotor stability by including the contribution of the labyrinth seals in
small nonsynchronous vibration still showed up around the rotor this case. However, the empirical method of the OEM can be
natural frequency of 74 Hz, and its magnitude was stable and less further improved to enable direct consideration of the influence of
than 0.02 mils. labyrinth seals and deswirl devices in the stability calculation. API
617 Level-I stability study has standardized the methodology and
the acceptance criteria of a preliminary stability analysis.
However, the primary goal of the Level-I study is to screen those
compressors that need a detailed analysis (Level-II). There still are
some pending issues in the Level-II stability procedure recom-
mended due to the existing uncertainties in the analytical tools and
the conditions assumed in the stability calculation.
With the implementation of API Level-I stability study require-
ments, more and more compressors need a Level-II study
including the labyrinth seals in the rotordynamics model. But
sometimes it is not possible and not necessary to examine all of the
labyrinth seals in a centrifugal compressor. For the compressor
manufacturers and users, the analysis workload would become
overwhelming and the engineering time would not be affordable to
investigate every seal, especially in the quote stage of a compres-
sor unit. From the industrial practice point of view, a quick and
reliable detailed study is expected.
Figure 6. Compressor 25MB5/6I Vibration Spectrum in Commis- Evaluation of Aerodynamic Excitation
sioning.
Using Wachel Number Formula
The replacement 25MB5/6I compressor was also pushed near Although Wachel’s cross-coupling formula has been widely
the surge line several times to verify the design limit. The spectrum used more than 20 years, the interpretation of the variables in the
plot in Figure 7 clearly shows that a low frequency at 23 Hz equation is quite liberal. Based on the experience from users with
emerged when the compressor was forced to run near its surge line. different backgrounds, some engineers use the entire horsepower
This low frequency vibration was expectantly excited by a rotating and gas density ratio while others would like to use the local power
 EXPERIENCES AND APPLICATIONS OF API 617 41
FULL STABILITY ANALYSIS (LEVEL-II) ON CENTRIFUGAL COMPRESSORS

and density ratio at individual wheels. The minimum flow path cross-coupling. Using shunt-injection and installing swirl-brakes
could be defined at the impeller tip or the diffuser channel. The are two popular options for compressor engineers to break up the
application in the new C-12 replacement compressor indicates that circumferential swirl flow in the labyrinth seal and to solve the
using the overall gas density ratio would result in a conservative load-dependent subsynchronous vibration problems. Apparently, a
estimation of the destabilizing aerodynamic excitation in a com- shunt-injection would slightly degrade the compressor efficiency
pressor. The conclusion is consistent with the authors’ experience due to an increase in the flow leakage through the labyrinth seal.
that the Wachel formula would give a more reasonable amount of Normally the impact of the shunt-injection on the efficiency is
aerodynamic excitation using the local absorbed power and gas minimal if the shunt-injection is used with an abradable labyrinth
density ratio. seal. Sometimes using shunt-injection or swirl brakes cannot com-
The procedure using the entire gas density ratio to estimate pletely eliminate the subsynchronous vibration. Field observations
cross-coupling at each impeller station is one of the considerable and analytical predictions are not consistent. An obvious reason is
conservative approaches. In the nine-wheel 25MB4/5 case, the the improper design and installation of the deswirl system. Another
estimated aerodynamic cross-coupling was 48,600 lb/in, which potential problem could be that it is not sufficient to only use one
was about 2.7 times the excitation predicted by the OEM and 70 deswirl device, shunt or brake. A hybrid shunt/brake design is
percent more than the Wachel number yielded by API’s formula. developed and applied in centrifugal compressors to deal with
The total Wachel number even increased up to 74,500 for the 11- complicated rotor instability problems in the authors’ company. A
wheel 25MB5/6I due to a decline in the suction pressure. It would detailed description of this kind of stability-enhancing mechanism
be of interest to note that both OEM and API’s equations showed is beyond the discussion scope of this paper.
the aerodynamic excitation force to be lower for the 11-wheel No doubt the effect of the major labyrinth seals and deswirl
rotor. Due to an overestimation of aerodynamic excitation baseline devices should be included in a detailed compressor rotordynamic
in the detailed stability analysis discussed above, the destabilizing stability analysis. Besides the diameter, length, and number of
contribution of the labyrinth seals appeared minor. Only the effect teeth of the seal, the seal clearance is another important parameter
of the center labyrinth seal was noticeable, but secondary. to be considered in the cross-coupling evaluation of the labyrinth
Apparently the exaggerating effect in the compressor cross- seal. With a positive gas swirl velocity at the seal entrance,
coupling would be more severe for a low suction gas density typically the larger the seal clearance, the larger the seal cross-
application such as the case studied in this paper. coupling stiffness. When performing a detailed stability
investigation considering the contribution from the labyrinth seal,
The Effect of Labyrinth Seals on Stability the seal cross-coupling and the log-decrement should be examined
Labyrinth seals have been demonstrated to be one of the major at the extremes of the seal clearance.
sources of destabilizing forces resulting in rotor instability
vibration in centrifugal compressors. The forces of the labyrinth A Practical Procedure to Evaluate Compressor Rotor Stability
seal are compressor load-dependent and dependent on the seal Based on the authors’ experience, the cross-coupling of the
geometric configuration as well. A well-known fact is that the major labyrinth seal such as the center seal or the balance piston
rotordynamic force of the labyrinth seal is dominated by its cross- seal should be calculated directly in a full rotor stability analysis.
coupled stiffness coefficient Kxy because other force coefficients The cross-coupling of other labyrinth seals may not be solved
(Kxx, Cxx, Cxy) are relative small. The seal cross-coupling stiffness explicitly due to its time-consuming nature and the uncertainties in
is strongly determined by the preswirl velocity of the process gas the actual conditions. Their contribution can be included in the
at the labyrinth seal entrance in the direction of rotation. The analytical model by using partial Wachel numbers. The objective is
effective damping of a labyrinth seal can be defined in the equation to avoid the possible duplication of the effect of the labyrinth seal
below, in the detailed stability analysis. The simplified full stability study
should be able to save engineering time and not lose the reliability
⎛ Kxy ⎞ of the high-level stability investigation.
Cef = Cxx ⎜ 1 − ⎟ (3)
⎝ Ω Cxx ⎠ A rotor instability vibration is a kind of self-excited subsynchro-
nous vibration. Normally the instability vibration most likely
where Ω is the precession frequency of the first forward mode of a occurs when the rotor running frequency is higher than the first
flexible rotor in a stability study scope. From the above equation, forward natural frequency of the rotor. The damping forces of the
it can be seen that any contribution to promote the cross-coupled journal bearing and the damper seals (if any) applied on the rotor
stiffness would reduce the effective damping of the labyrinth seal. are proportional to the frequency of the potential instability
It should be pointed out that the positive direct damping Cxx of the vibration at a nonsynchronous frequency. Therefore, the authors
labyrinth seal is small, but it is critical to keep the rotor in the used the nonsynchronous coefficients of the journal bearings in a
stable region when the compressor is a borderline machine. proposed full stability study to predict the real dynamic behavior
It has been recognized that most of the seal cross-coupling is of the compressor rotor. Since the effect of the center seal or the
generated by the center labyrinth seal in a back-to-back compres- piston labyrinth seal is split with other aerodynamic excitations in
sor or the balance piston seal in a straight-through compressor. The the proposed procedure, the contribution of the major labyrinth
center seal and the balance piston seal normally take more axial seals in the compressor can be investigated in depth. The following
space and have more labyrinth teeth than other impeller eye seals steps would give a simplified procedure to conduct a detailed
and shaft seals. In the centrifugal compressor, the center labyrinth stability study on the compressor rotors.
seal or the balance piston seal sustains the largest differential
pressure inherently. For a typical subsynchronous vibration associ- • Use the Wachel number formula in Equation (2) to calculate the
aerodynamic excitation (Wachel number) at each impeller based
ated with the first forward mode of the flexible rotor, the effect of
on the local gas density ratio, local wheel horsepower absorbed,
the labyrinth seals on the stability is substantial because the
and gas molecular weight at the rated operating conditions. The
magnitude of the destabilizing force of the labyrinth seal is pro-
flow path width is the minimum from the diffuser channel and
portional to the shaft precession amplitude at the seal location.
impeller tip.
Prior to adopting gas damper seals such as honeycomb seals and
pocket-damper seals, it is easier to enhance rotor stability by • Calculate the rotordynamic force coefficients of the major
reducing the cross-coupling of the labyrinth seal than increasing its labyrinth seals at the extremes of the seal clearance with the rated
direct damping. Reducing the circumferential flow in the labyrinth gas properties around the seals and the given gas preswirl velocity
seal is a direct and effective approach to reduce the seal ratios.
42 PROCEEDINGS OF THE THIRTY-FOURTH TURBOMACHINERY SYMPOSIUM • 2005

Table 9. Predicted Average and Minimum Aerodynamic Log-

• Adjust the magnitudes of the predicted Wachel numbers down Decrements.
with a multiple factor around 0.65 ~ 0.75. The objective is to
separate the excitation of the major labyrinth seal from other aero-
dynamic excitation for conducting a detailed stability analysis. Aero-Log 9-Wheel Rotor 11-Wheel Rotor
Dec (25MB4/5I) (25MB5/6I)
• Incorporate the adjusted local Wachel numbers at the correspon-
ding impellers and the predicted cross-coupling of the major Average Minimum Average Minimum
labyrinth seals into the model of the compressor rotor. If a damper log-dec log-dec log-dec log-dec
seal is used at the center seal location or the balance piston Seal w/o
0.02 -0.06 0.03 -0.05
location, the seal direct stiffness and damping items should be Shunt/Brake
included in the analytical model as well. Seal with
0.11 0.03 0.12 0.04
Shunt/Brake
• First calculate the average aerodynamic log-decrement at the
nominal bearing clearance and the maximum continuous
operating speed, then the minimum aerodynamic log-decrement at CLOSURE
the extremes of the journal bearing clearance and the supply oil The above stability study procedure is practical in engineering
temperature. In the proposed study procedure here, nonsynchro- time-consumption and straight forward for the compressor manufac-
nous bearing coefficients should be used in the log-decrement turers and the users. It permits a thorough investigation of the rotor
calculation if the compressor is fitted with tilt-pad journal stability quickly without missing any main points associated with the
bearings. For typical compressor applications with an average gas load-dependent rotor subsynchronous vibration. This procedure may
density less than 4.0 lb/ft3, the stability design of a centrifugal be considered as an equivalent full stability study per API require-
compressor would be considered as acceptable if the average log- ments before sound validated analytical tools are well developed and
decrement is not less than 0.05 and the minimum log-decrement is become available to accurately estimate the aerodynamic excitation
positive. of the compressor impeller. A minimum log-decrement of 0.05 is
Following the stability analysis procedure above, both rotors always preferred in the stability analysis. Based on the authors’ expe-
25MB4/5I and 25MB5/6I of the C-12 replacement compressor rience, a rotor having a minimum log-decrement of 0.05 with
were studied and the results are presented below. Table 7 shows nonsynchronous bearing coefficients normally can reach the log-
the adjusted local Wachel numbers for the C-12 compressor at decrement value 0.10 required by API Level-II using synchronous
the rated operating conditions. The cross-coupling of the center bearing data. If the predicted minimum log-decrement becomes
labyrinth seal was calculated separately considering with or negative, an exceptional damper such as a honeycomb seal may be
without a deswirl remedy and predicted data were provided in further considered and the seal rotordynamic force coefficients
Table 8. Including the aerodynamic excitations in Tables 7 and should be solved as a function of the rotor whirling frequency at the
8, the log-decrements were calculated using the nonsynchro- given operating conditions. It is noted that the study procedure is
nous bearing coefficients at the MCOS of 12,000 rpm, and the developed for current compressors fitted with dry gas face seals. If
final analysis results are shown in Table 9. The stability the centrifugal compressor has oil seals or fitted with squeeze film
analysis results illustrated that the stability design of the C-12 damper devices, more special considerations should be devoted in
replacement compressor would be accepted with a shunt/brake the stability analysis and the acceptance value could be different.
deswirl system, and the field operation also confirmed that the
compressor vibration was well damped without the instability
REFERENCES
issue. API Standard 617, 2002, “Axial and Centrifugal Compressors and
Expander-Compressors for Petroleum, Chemical and Gas
Table 7. Adjusted Wachel Numbers (65 Percent) in C-12 Industry Services,” Seventh Edition, American Petroleum
Compressor. Institute, Washington, D.C.

9-Wheel Rotor 11-Wheel Rotor BIBLIOGRAPHY

Impeller
(25MB4/5I) (25MB5/6I)
Benckert, H. and Wachter, J., 1980, “Flow Induced Spring
1st 602 341
Coefficients of Labyrinth Seals for Application in Rotor-
Stage I

2nd 702 416

3rd 814 487 dynamics,” NASA CP 2133, Proceedings of a Workshop on
4th 1333 563 Rotordynamic Instability Problems in High-Performance
5th 900 Turbomachinery, Turbomachinery Laboratory, Texas A&M
University, College Station, Texas, pp. 189-212.
1st 883 690
2nd 987 787 Childs, D. W., 1993, Turbomachinery Rotordynamics, New York,
Stage II

3rd 1058 892 New York: John Wiley & Sons, pp. 290-354.
4th 1223 968 Doyle, H., 1980, “Field Experience with Rotordynamic Instability
5th 2474 1099 in High-Performance Turbomachinery,” NASA CP 2133,
6th 2021 Proceedings of a Workshop on Rotordynamic Instability Prob-
Total 10074 9165 lems in High-Performance Turbomachinery, Turbomachinery
Laboratory, Texas A&M University, College Station, Texas,
Table 8. Predicted Cross-Coupling of the Center Labyrinth Seal. pp. 3-14.
Seal Fulton, J. W., 1984, “Full Load Testing in The Platform Module
9-Wheel Rotor 11-Wheel Rotor
Cross-coupling Prior to Tow-Out: A Case History of Subsynchronous
(25MB4/5I) (25MB5/6I)
(lb/in) Instability,” NASA CP-2338, Proceedings of a Workshop on
Min. Max. Min. Max. Rotordynamic Instability Problems in High-Performance
Clearance Clearance Clearance Clearance
Turbomachinery, Turbomachinery Laboratory, Texas A&M
Seal w/o
Shunt/Brake
8980 10722 9194 10999 University, College Station, Texas, pp. 1-16.
Seal with Li, J., De Choudhury, P., and Kushner, F., 2003, “Evaluation of
777 -1468 884 -1462
Shunt/Brake Centrifugal Compressor Stability Margin and Investigation of
 EXPERIENCES AND APPLICATIONS OF API 617 43
FULL STABILITY ANALYSIS (LEVEL-II) ON CENTRIFUGAL COMPRESSORS

Anti-Swirl Mechanism,” Proceedings of Thirty-Second Turbo- Compressors,” Proceedings of Twenty-Second Turboma-

machinery Symposium, Turbomachinery Laboratory, Texas chinery Symposium, Turbomachinery Laboratory, Texas A&M
A&M University, College Station, Texas, pp. 49-57. University, College Station, Texas, pp. 3-15.
Lund, J. W., 1974, “Stability and Damped Critical Speeds of a Zhou, R. M., 1986, “Instability of Multistage Compressor K1501,”
Flexible Rotor in Fluid Film Bearings,” ASME Journal of NASA CP-2443, Proceedings of a Workshop on Rotordynamic
Engineering for Industry, 96, pp. 509-517. Instability Problems in High-Performance Turbomachinery,
Nielsen, K. K., Childs, D. W., and Myllerup, C. M., 2001, “Experi- Turbomachinery Laboratory, Texas A&M University, College
mental and Theoretical Comparison of Two Swirl Brake Station, Texas, pp. 63-75.
Designs,” ASME Journal of Turbomachinery, 123, pp. 353- ACKNOWLEDGEMENT
358.
The authors are indebted to the Elliott Company and Unocal
Wachel J. C., 1982, “Rotordynamic Instability Field Problems,” Corporation for permission to publish this paper, and to the many
NASA CP-2250, Proceedings of a Workshop on Rotordynamic individuals within the companies whose support and assistance
Instability Problems in High-Performance Turbomachinery, made this paper possible, with special gratitude to Dr. John Vance
Turbomachinery Laboratory, Texas A&M University, College of the Turbomachinery Laboratory of Texas A&M University for
Station, Texas, pp. 1-19. his support during the manuscript preparation.
Zeidan, F. Y., Perez, R. X., and Stephenson, E. M., 1993, “The Use
of Honeycomb Seals in Stabilizing Two Centrifugal
44 PROCEEDINGS OF THE THIRTY-FOURTH TURBOMACHINERY SYMPOSIUM • 2005