Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air

Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air
GT2009
GT2009
June 8-12, 2009, Orlando, Florida, USA
June 8 – 12, 2009, Orlando, Florida, USA

GT2009-59392
GT2009-59392

ROTORDYNAMIC COMPARISON OF BUILT-UP VERSUS SOLID ROTOR CONSTRUCTION

J. Jeffrey Moore, Ph.D. Andrew H. Lerche

Southwest Research Institute® Southwest Research Institute®
San Antonio, Texas, USA San Antonio, Texas, USA

ABSTRACT reduces diaphragm deflection, and is less expensive to

manufacture [3].
Most manufacturers of multi-stage centrifugal
compressors for the oil and gas industry utilize a solid shaft
rotor construction. The impellers use a shrink fit onto the
shaft with spacers in between the impellers. With the
introduction of the guidelines in the 7th edition of API 617,
built-up rotors for centrifugal compressors using a tie-bolt are
recognized by API. This study compares the rotordynamic
performance of the identical compressor using both a tie-bolt
design and a more conventional solid rotor for a two-stage
pipeline application. A full API 617 lateral analysis is
performed on the two designs, assuming identical impeller
flow path, stage spacing, and hub diameter. The critical speed
and unbalance response are computed, and a full Level 2
stability analysis is performed for each case. The results show
the tie-bolt construction to be slightly lighter and stiffer, FIGURE 1 – EXAMPLE OF BUILT-UP ROTOR SHOWING
resulting in a higher critical speed and improved rotordynamic TIE-BOLT (COURTESY OF SOLAR TURBINES INC.)
stability.
Several authors have reported rotordynamic issues
INTRODUCTION associated with tie-bolts and built-up rotors, including
inadequate tie-bolt stretch, tie-bolt resonance, interface
Built-up rotors are used in many turbomachines, including
stresses, and unbalance issues [4, 5]. The tie-bolt natural
gas turbines, expanders, and rocket engine turbopumps. For
frequency is a function of the axial tension and internal
example, Lipiak and Schroeder [1] show its use in a 160MW
supports. If the rotor speed encounters the tie-bolt natural
gas turbine. However, most manufacturers of centrifugal gas
frequency, then a destructive resonance can result. If solid
compressors for the oil, gas, and processing industries utilize
diaphragms are used, the rotor must be re-stacked with the
solid rotor construction. The American Petroleum Institute
diaphragms requiring re-assembly of the rotor. The residual
(API) 617 standard [2] for centrifugal compressors now also
unbalance will generally increase often requiring trim
recognizes the built-up rotor construction method. The built-
balancing during factory testing. These drawbacks for built-
up rotor (see Fig. 1) has the potential to be lighter and stiffer
up rotors can be minimized through proper rotordynamic
since a hollow round beam has a higher natural frequency than
design and manufacturing techniques.
a solid one. Furthermore, a built-up rotor maximizes the shaft
stiffness to equal the flow path diameter at the impeller inlet This paper discusses the modeled rotordynamic behavior
while solid rotors require impeller spacers that reduce this of a built-up rotor for a C452 centrifugal compressor built by
stiffness diameter. However, the multiple interfaces between Solar Turbines (Fig. 2). This two-stage compressor is
the built-up rotor components can reduce the shaft stiffness, typically used for natural gas pipeline transmission
depending on the interface diameter and design used. Built-up applications. Comparisons are made with a second
rotors have other advantages over solid rotors by simplifying rotordynamic model developed using identical geometry and
re-staging of the compressor and permitting solid diaphragms operating conditions, except modified to simulate a solid rotor
(not split) to be used, which eliminates split line leakage, construction. Various rotordynamic characteristics are

1 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo

 compared, including critical speed location, amplification that do not use rotor spacers with an associated diameter
factor, and rotordynamic stability. reduction, most compressor designs use rotor spacers.
The rotor models were analyzed using the XLTRC2 [6]
software developed by the Turbomachinery Laboratory at
Texas A&M University. This is a finite element code, which
has the capability to predict undamped critical speeds and
mode shapes, unbalance response, and damped natural
frequencies (which provide stability prediction).

FIGURE 2 – PHOTOGRAPH OF SOLAR C452

CENTRIFUGAL PIPELINE COMPRESSOR
(COURTESY OF SOLAR TURBINES INC.)

ROTORDYNAMIC MODELING
In order to predict the behavior of the two rotors, FIGURE 3 – SHAFT MODEL GEOMETRY PLOT FOR
rotordynamic models were generated for a built-up (with tie- STACKED TIE-BOLT ROTOR
bolt) rotor as well as a solid rotor. Figure 3 provides a
graphical definition of the mass-elastic model of the stacked
tie-bolt rotor, while Fig. 4 shows the solid rotor. Each rotor
model is composed of 66 elements and is derived from mass
elastic data, Computer Aided Design (CAD) geometry, and
other information provided by the Original Equipment
Manufacturer (OEM). The element numbering starts at the
drive end, shown on left in the model. The drive coupling is
concentrated at Station 6 of the models. The journal bearings
are located at Stations 10 and 58. The vibration probes are
located adjacent to the bearings at Stations 9 and 59.
Labyrinth seals are located at Stations 31, 42, and 44.
Figure 3 shows the tie-bolt rotor, and the red elements
indicate a steel material that contributes to the mass and FIGURE 4. – SHAFT MODEL GEOMETRY PLOT FOR SOLID
stiffness of the rotor. The blue elements represent a steel ROTOR
material and contribute only to the mass of the rotor. The dry
gas seals and the tie-bolt nut are both represented by the blue TILTING-PAD BEARING ANALYSIS
elements. The silicon carbide portion of the dry gas seals Bearing stiffness and damping characteristics were
shown contribute only to the mass of the rotor. The yellow determined using the XLTRC2 software. The software solves
elements represent the impellers and contribute only to the the two-dimensional Reynolds equation. The adiabatic option
stiffness of the rotor since the mass properties of the impellers in XLTRC2 was used for these analyses with an 80% carryover
are modeled separately as added masses. The green elements assumption. This assumption implies that 80% of the hot oil
in Fig. 3 represent the tie-bolt, which contributes to mass and exiting one pad goes into the downstream pad. The remainder
stiffness, and includes the axial tension. is made up by fresh oil supply. This approach provides
Figure 4 shows the solid rotor, and the red elements reasonable correlation to the measured temperature from the
indicate a steel material that contributes to the mass and factory test. ISO 32 oil supplied at 125°F is used.
stiffness of the rotor. The green elements represent the Figure 5 is a bearing schematic, which shows the
impellers and contribute only to the stiffness of the rotor, since definition of Cb (assembled clearance) and Cp (pad clearance)
the mass properties of the impellers are modeled separately as in terms of differences in radii of shaft and pad, together with
added masses. The solid rotor is modeled with the stiffness their relationship to preload (m).
diameter reduced by 6.35 mm along the impeller sections to
account for shrink-fit type impellers with spacers in between.
While some manufacturers use solid rotor construction designs

2 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo

 in Figure 6 and Figure 7. This mode is well above the running
speed and therefore will have little effect on the rotor
response.
Since damping is not included in the undamped critical
speed map, it only provides guidance as to likely critical speed
values. The unbalance response analysis to be presented in the
next section will define more accurately the speed of the
predicted resonant peaks.
forward
backward
f=3995.2 cpm
K=44624769 N/m

FIGURE 5. – SCHEMATIC TO DEFINE CLEARANCES AND

PRELOAD FOR A TILTING PAD BEARING
forward
CRITICAL SPEED ANALYSIS backward
f=7467.2 cpm
Figure 6 shows the undamped critical speed map for the K=47439933 N/m
tie-bolt rotor at nominal bearing clearance and temperature.
The critical speed map plots the rotor natural frequencies as a
function of bearing stiffness. Also plotted are the nominal
bearing stiffness values (Kxx and Kyy) versus speed for both
bearings. The intersection of the average bearing stiffness and
the natural frequency represents the undamped critical speed.
Damping will affect the location of these critical speeds, as
will be described later. The undamped critical speed map
shows the location of the first five critical speeds and shows
forward
the sensitivity of these modes to bearing stiffness. Most rotor-
backward
bearing systems operate between the extremes of low and high
f=14597.7 cpm
bearing stiffness and below the first bending (third) mode. K=55380009 N/m

FIGURE 6. – CRITICAL SPEED MAP FOR TIE-BOLT

ROTOR
Figure 7 presents the first four mode shapes of the tie-bolt
rotor. The first mode is a symmetric mode with relatively high
rotor bending. The second mode predominately involves out-
of-phase motion near the ends of the rotor with a mid-point
node. The third mode involves two nodal points, with in-
phase motion at the shaft ends. The mode shapes of the tie-
bolt rotor and solid rotor are very similar for the first three FIGURE 7. – CALCULATED UNDAMPED MODE SHAPES
modes. The tie-bolt resonant mode is the fourth mode shown FOR THE TIE-BOLT ROTOR

3 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo

 Table 1 summarizes the undamped natural frequencies for amplitudes in Fig. 8 and Fig. 9. The amplification factor for
the two cases. The first and third critical speed is slightly the first critical speed of the solid rotor at the center of the
higher for the tie-bolt case due to the lighter rotor. The second rotor is 2.11, which is slightly higher than the tie-bolt rotor.
critical speed, however, is identical for the two cases. The
Both rotor responses meet API unbalance sensitivity
bearing stiffness is lower for the tie-bolt rotor at lower speed
requirements, which requires that the response due to two
due to the lower weight.
times API unbalance be less than the vibration limit of:
TABLE 1. – SUMMARY OF FIRST THREE UNDAMPED
CRITICAL SPEEDS FOR EACH ROTOR/DRIVER CASE
12000
Vmax = 25.4 = 28.5 μm (1)
N

RESPONSE TO UNBALANCE
The undamped critical speed analysis provides an
approximate indication of where critical speeds will occur for
the rotor-bearing system. At critical speeds, the system’s
sensitivity to unbalance excitation is maximized. However,
the calculation of response to unbalance is required to account
for the full interaction of rotating forces caused by unbalance,
rotor elastic deformation, and bearing forces induced by both
rotor displacement and velocity at the bearings. Including all
these interactions typically shifts the frequency of peak
response away from and higher than the undamped critical
speed. It also enables a measure of likely severity of response FIGURE 8. – PROBE 1 RESPONSE COMPARISON TO
at the critical speeds to be determined. UNBALANCE AT CENTER (1X API)

The calculation of response to unbalance for the tie-bolt

and solid rotors have been performed using the rotor models
presented with an unbalance magnitude of 6350 W/N (or 4
W/N using U.S. Customary units), where W is the rotor
weight and N is the Maximum Continuous Speed (MCS) of
9,500 rpm. Since the model is linear, the vibration amplitude
will scale linearly with unbalance magnitude. Therefore, a
unbalance magnitude of 4 times the API unbalance (following
API 617 guidelines) would result in an amplitude 4 times that
shown in the plots.
Figures 8 and 9 compare the tie-bolt rotor and solid rotor
response to unbalance applied at Station 36 of the model, near
the mid-span location of the rotor. This unbalance scenario is
meant to excite the first critical speed. Figure 8 corresponds FIGURE 9. – MID-SPAN RESPONSE COMPARISON TO
to the response at probe 1, while Fig. 9 shows mid-span UNBALANCE AT CENTER (1X API)
response. The figures include horizontal and vertical rotor
vibration amplitude for the tie-bolt rotor and solid rotor. Figure 10 presents the response to an unbalance located at
The first critical speed excited by unbalance appears to be both shaft ends, with each unbalance 180 degrees out-of-phase
critically damped. The tie-bolt rotor shows a peak response from the other. The unbalance is calculated using the API
for both probes and Station 36 (center of rotor) between 6,000 formulation of 6350W1/N at one end, and 6350W2/N at the
and 6,470 rpm. The amplification factor at the center of the other end, where W1 and W2 are the static loads at bearing 1
rotor, based on the half power points method, is approximately and bearing 2, respectively. This unbalance configuration
1.84. According to API 617 7th Edition, a separation margin is intends to maximize excitation of the second mode, based on
not required for amplification factors less than 2.5. the second undamped mode shape.

The first critical speed of the solid rotor falls slightly The tie-bolt rotor shows the second critical speed to be
below the tie-bolt rotor when excited with the mid-span between 9,660 and 9,670 rpm. The response at probe 2 shows
unbalance. The solid rotor shows a peak response for both the amplification factor to be about 1.83.
probes and Station 36 (center of rotor) between 5,770 and
6,120 rpm as shown by the horizontal and vertical response

4 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo

 The solid rotor shows the second critical speed to be for the first mode of vibration. The total cross coupling is the
between 9,580 and 9,650 rpm, which are slightly lower than weighted sum from each impeller.
the second critical speed of the tie-bolt rotor. This response
Table 3 shows a total impeller cross-coupling for the tie-
amplitude is slightly higher than the response in the tie-bolt
bolt rotor of 3.61 x 106 N/m using the API method, for design
rotor. The response shows the amplification factor for the
conditions. The inputs to calculate the impeller cross-coupling
solid rotor to be about 1.86.
for the solid rotor are the same as the tie-bolt rotor with
The second critical speed for both the tie-bolt rotor and exception to the modal amplitudes of the impellers. The solid
solid rotor is very near the MCS of 9,500 rpm when driven by rotor impeller cross-coupling value is 3.59 x 106 N/m.
the turbine. However, the amplification factor remains below
TABLE 3. – IMPELLER AERO EXCITATION
2.5 for both rotors and no separation margin is required by API CALCULATIONS FOR TIE-BOLT ROTOR AT MCS
guidelines.
Aero Cross-coupling Analysis Revised 03/18/08
Mars 14,000HP Checkpoint Inputs

Tie Bolt - Design Conditions

Units
Wheel(I):- 1 2
Speed rpm 9500 9500
Power KW 4764.928 5439.316
Power % Total 46.70% 53.30%
Power
Power hp 6387.3 7291.3
Mole Weight 17.38 17.38
Density Out kg/m^3 58.03 69.56
Density In kg/m^3 49.18 58.03
Stage Density Ratio 1.18 1.20
Section Density Ratio 1.41 1.41
Diameter mm 519.684 538.48
Diameter in 20.46 21.20
B2
FIGURE 10. – PROBE 1 RESPONSE TO ENDS OUT OF B3 mm 19.22 17.39
PHASE B3 in 0.757 0.684
Diffuser Pinch
Station No. 32 43
TABLE 2. – UNBALANCE RESPONSE SUMMARY Modal Amplitude 0.998 0.958
Estimated
Unbalance Shaft Critical
Amplifi-
Imp Kxy (API) lb/in 9643.232574 10996.94993
Rotor MCS Configu- Location of Speed
cation
ration Response [rpm]
Factor
Center 6000 1.84
Total
Tie- Imp Kxy (API) 3.61E+06 N/m (20,640.18 lb/in)
9500 Mid Span Probe 1 6430
bolt
Probe 2 6470
Center 5770 2.11
Solid 9500 Mid Span Probe 1 6100
Probe 2 6120 The calculation of damped natural frequencies and
Tie-
9500 Shaft End
Probe 1 9660 associated log decrement provides a measure of rotor stability.
bolt Probe 2 9670 1.83 This analysis uses the rotor model, the bearing analysis
Probe 1 9650
Solid 9500 Shaft End
Probe 2 9580 1.86 (synchronously reduced coefficients), and the aerodynamic
cross-coupling values. The labyrinth seal coefficients were
calculated for both impeller eye seals and the balance piston
seal and consist of tooth on rotor designs with abradable
STABILITY ANALYSIS stators. The balance piston seal contains a shunt injection to
For this study, the authors utilized an adapted version of minimize pre-swirl entering the seal. The XLLaby [6] code is
the Wachel method [7] as found in API 617 [2]. The resulting used for the seal predictions. The calculations made are
formula for each impeller is: consistent with API requirements for a Level II analysis.

⎛ 2.865 × 10 7 ⎞ ⎛ Hp ⎞⎛ ρ d ⎞ 2 Figures 11 and 12 provide damped eigenvalue mode

K xy = ⎜⎜ ⎟⎟∑ ⎜⎜ ⎟⎟⎜⎜ ( )
⎟⎟ η (1) shape plots for the rotors operating at MCS with all labyrinth
⎝ N ⎠ ⎝ D ⋅ B3 ⎠⎝ ρ s ⎠ seals, and API level aero cross-coupling. The figures indicate
where Hp is the stage power (Watts), N is the rotor speed the whirl frequency for each rotor-driver combination. The
(rpm), D is the impeller diameter (mm), B3 is the diffuser tie-bolt rotor has a frequency of 4,327 rpm and a log
passage width (mm), ρd is the stage discharge density, ρs is the decrement value of 1.14, while the solid rotor has a frequency
stage suction density, and η is the impeller modal amplitude of 4,229 rpm and log decrement of 0.997. All of these
calculations are based on nominal bearing clearance and

5 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo

 temperature and meet the API requirement for a minimum log higher response amplitudes than the tie-bolt rotor. The shaft
decrement value of 0.1. ends out-of-phase unbalance response for the solid rotor
shows a similar second critical speed but with a 7% higher
Damped Eigenvalue 3-D Mode Shape Plot response amplitude than the tie-bolt rotor. The stability
As-Is Rotor (W/Tiebolt) analysis shows that the solid rotor has a 12% lower log
MCS 9500rpm
forward
decrement value and a 10% lower stability threshold than the
backward tie-bolt rotor. Both rotor designs meet API 617, 7th edition [2]
f=4327. cpm requirements.
d=1.1351 logd
L og Dec V ersu s Aero Kxy
(9500rp m M C S)
1 .3 50 0
T iebolt Rotor 9500rpm M CS
S olid Rotor 9500rpm M CS
1 .1 50 0

Q0 = 2.62e7 N/m

Q0 = 2.94e7lb/in
0 .9 50 0 Tieb o lt R oto r
API Aero: Q 0/Q a=8.14

0 .7 50 0

Log Dec
0 .5 50 0

Qa = 3.61e4 lb/in
S olid Ro tor
AP I Aero : Q 0/Q a=7.31
0 .3 50 0

FIGURE 11. – FIRST DAMPED MODE WITH LABYRINTH

0 .1 50 0
SEALS AND API AERO KXY, TIE-BOLT ROTOR
-0 .0 50 0

0.00E+00

5.00E+06

1.00E+07

1.50E+07

2.00E+07

2.50E+07

3.00E+07

3.50E+07
Damped Eigenvalue 3-D Mode Shape Plot -0 .2 50 0

Solid Rotor Aero K xy [ N/m ]

MCS 9500 RPM

forward FIGURE 13. – LOG DEC STABILITY THRESHOLD
backward SUMMARY
f=4229.2 cpm
d=.9968 logd
ACKNOWLEDGEMENT
The authors would like to thank Solar Turbines for
providing the compressor geometry and operating conditions
used in this study. We would also like to thank SwRI for
permission to publish this work.

REFERENCES
[1] Lipiak, G. and Schroeder, A., 2007, “Lifetime Extension
for Siemens Gas Turbines”, VGB Powertech, Vol. 87/2007.
FIGURE 12. – FIRST DAMPED MODE SHAPE WITH
LABYRINTH SEALS AND API AERO KXY , SOLID ROTOR [2] API 617, 2002, “Axial and Centrifugal Compressors and
Expander-Compressors for Petroleum,” Chemical and Gas
Figure 13 compares the log decrement sensitivity to aero Industry Services, Seventh Edition, American Petroleum
cross-coupling Kxy of the tie-bolt rotor and solid rotor. The Institute, Washington, D.C.
figures also show the Q0 and Qa values, which are defined as
the cross-coupling value necessary to produce zero log [3] Kurz, R., 2004, “Industry Benefits from Efficiencies of
decrement and the API specified cross-coupling amount under Modular Gas Compressor Design,” Pipeline & Gas Journal.
nominal operating conditions, respectively. This plot was also
[4] Kirk, R.G., et al., 2007, ”Diesel Engine Turbocharger
used to determine the stability threshold, based on API aero
Rebuild and Experimental Testing,” Proceedings of the
cross-coupling, which is defined as Q0 / Qa. The tie-bolt rotor
ASME/STLE International Joint Tribology Conference
has a stability margin of 8.14, while the solid rotor has a
(IGTC), San Diego, CA.
stability margin of 7.31, representing a 10% decrease in the
stability threshold using solid rotor. [5] Carek, G.A., 1988, “Improved Method for Stress and
Compatibility Analysis of Multi-component Rotating
CONCLUSION Systems,” NASA Technical Memorandum 100884, NASA
The analytical results presented in this paper indicate that Glenn Research Center.
the tie-bolt rotor and solid rotor meet the required API [6] XLTRC2, 2006, User’s Manual, Turbomachinery
separation margin criteria [2] when operating at nominal Laboratory, Mechanical Engineering Dept., Texas A&M
clearances and temperatures. The solid rotor is approximately University, College Station, Texas.
21 kg heavier (5% greater) than the tie-bolt rotor and has
slightly lower undamped critical speeds. The mid-span [7] Wachel, J. C. and von Nimitz, W., 1980, “Assuring the
unbalance response for the solid rotor has about a 5% lower Reliability of Offshore Gas Compression Systems,” European
critical speed, a 15% higher amplification factor, and slightly Offshore Petroleum Conference & Exhibition, London,
England; EUR 205.

6 Copyright © 2009 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/gt2009/70651/ on 02/03/2017 Terms of Use: http://www.asme.org/abo