50 ORBI T Vol . 26 No.

3 2006
APPLICATIONS
Dave W. Fahy, P.E.
Senior Field Engineer
Optimization & Control Services
GE Energy
dave.fahy@ge.com
T
his case history reviews many of the problems
encountered during the startup of a wet gas
compressor machine train at a petrochemical
facility. The machine train consisted of a 1500
kW (2000 hp) single-wheel steam turbine, speed
increasing gearbox, and centrifugal compressor
that was skid-mounted and assembled by the
Original Equipment Manufacturer (OEM). A factory
test run verified that the unit conformed to the
end-users specifications, and it was shipped to
the field for installation.
The unit was subsequently subjected to excessive
piping forces causing the turbine casing to distort.
Casing distortion was severe enough to break
the shaft upon initial startup and to permanently
deform a replacement shaft during subsequent
startup attempts. Secondary damage included
repeated failure of seals and bearings.
Resolution of the problems required numerous
piping design changes, extensive foundation
repairs, and replacement of the turbine-to-gearbox
coupling. Additionally, the turbine was returned to
the manufacturer for testing to ensure the casing
had not suffered permanent distortion. Seal and
bearing design changes were also made during
the turbine testing.
The extensive problems proved very costly
several million U.S. dollarsin terms of labor and
material as well as lost production due to a six-
month delay in plant startup.
Pipe Strain
Vol . 26 No. 3 2006 ORBI T 51
APPLICATIONS
Pipe Strain Basics
S
kid-mounted machine trains are becoming more
common as packaged units are considered to be
quick and easy to install (Figure 1). Machine trains can
be test run in the factory, shipped to the customer, and
installed as a single unit.
Care must be taken during shipping and installation
of the assembled unit, particularly to the mounting
and adhesion mechanism between the skid and the
concrete foundation. External forces applied to the
machine train at the suction and discharge nozzles
(pipe strain) must be kept within the OEMs force and
moment specifications, preferably no more than half
the allowable limits. Even though the accompanying
piping system is carefully designed to be within these
limits, problems can occur in at least two ways. One
is that the calculations themselves can be in error.
The other is installation-related. Those responsible for
installing the piping, hangers, and associated apparatus
can introduce strain not accounted for in the design
if specifications and calibration requirements are not
carefully adhered to.
This case history does not attempt to determine the
specific causes of the excessive pipe strain. Instead, it
focuses on the effects such pipe strain introduces in
rotating machinery, and how pipe strain was deduced
as a fundamental root cause of problems on the wet gas
compressor under consideration.
Figure 1 Skid-mounted Wet Gas Compressor Train showing (left to right) compressor, speed-increasing
gearbox, and single-wheel steam turbine.
Initial On-Site Findings (First Startup)
T
he original steam piping was designed to ensure
that less than 80% of the allowable forces and
moments would be present on the turbine inlet and
exhaust flanges. Nevertheless, it was believed that
pipe strain caused the turbine shaft to fail during its
initial solo run. This occurred when the turbine casing
distorted to such an extent that the carbon steam seal
rings made full contact with the shaft. The resulting rub
generated sufficient heat to bow, and then crack, the
turbine shaft within several hours.
Major revisions to the piping system design were then
implemented. Double expansion joints were added to
the inlet and exhaust sides of the steam turbine (Figure
2). The revisions were such that less than 50% of the
allowable forces and moments would be present on
the turbine inlet and exhaust flanges. Although the
revised piping installation did conform to the design
specifications, there were concerns about the materials
of construction, stability of the supports, and calibration
of the pipe hangers. There were also safety concerns
regarding the use of bellows expansion joints on the
inlet steam line which operated at 50 kg/cm
2
@ 400C
(700 psig @ 750F). Exhaust steam conditions were
11 kg/cm
2
@ 170C (150 psig @ 360F).
The end-user was advised to completely remove the
piping system and install a system with much larger
thermal loops. However, the recommendation was
rejected and operation continued with the existing
modifications.
52 ORBI T Vol . 26 No. 3 2006
APPLICATIONS
Second Startup (First Piping Modification)
U
sing the permanently installed Bently Nevada
3300 machinery protection system and a portable
ADRE

data acquisition system, vibration data was

recorded from slow roll to approximately 5000 rpm.
Initial low speed data suggested a light rub might have
occurred. Because these vibration amplitudes were
small, the plant elected to continue with the startup
and machine speed was increased. Vibration levels
remained low into the 4000-rpm range, but at approxi-
mately 4380 rpm, the vibration started to increase.
Intermittent rubs occurred, indicated by the flat areas
on the timebase waveform (Figure 3). Vibration levels
were as high as 18 m pp (0.72 mil pp) at the coupling
end. At 4880 rpm and with operating clearances
diminished due to the turbine casing distortion, a hard
rub occurred in the seal area at the coupling end of the
turbine (Figure 4). The startup was aborted as a result,
and turbine speed was reduced to 500 rpm.
Optical measurement points were then installed on
the turbine, baseplate, soleplate, and support I-beams.
Initial measurements were taken with the steam piping
Figure 2 Initial modications to piping arrangement surrounding the steam turbine. View is looking down
from above the steam turbine.
Other concerns that surfaced during the initial on-site
inspection of the machine involved the handling of the
skid and the foundation integrity. It was learned that
a spreader bar used to lift one end of the skid bent
during the original installation and the unit dropped to
the ground. The exact height from which the unit was
dropped was never verified but was believed to be
3060 cm (1224 in). The quality of the anchor bolt
material, the anchor bolt installation, and the grouting
technique used to secure the skid were also questionable.
Vol . 26 No. 3 2006 ORBI T 53
APPLICATIONS
Figure 3 Orbit/timebase plot at 4380 rpm showing evidence of light rub at turbine inboard bearing during
second startup.
Figure 4 Orbit/timebase plot at 4880 rpm showing evidence of hard rub at turbine inboard bearing during
second startup.
54 ORBI T Vol . 26 No. 3 2006
APPLICATIONS
disconnected and the unit cold. The second set of
measurements was taken after the unit had operated
at 500 rpm for approximately four hours. Normally,
optical alignment data is taken at operating speed, but
vibration data indicated that operation above 500 rpm
would be unsafe.
Significant information was obtained regarding the
horizontal movement of the turbine casing relative to
the turbine bearing housings (Figure 5). The inboard
and outboard bearing housings moved 0.225 mm and
0.175 mm (0.009 in and 0.007 in) south, respectively.
Inboard and outboard turbine casing measurement
points moved 0.750 mm and 0.425 mm (0.030 in and
0.017 in) north, respectively. Thus, the bearing housings
moved in an opposite direction from the turbine casing.
Bearing housing movement was also revealed by shaft
alignment readings, indicating shaft movement of 0.300
mm (0.012 in) south during the steam pipe installation
and warm-up. It was apparent the turbine was being
distorted in the horizontal direction.
The only horizontal driving force present, other than
normal casing expansion, was that exerted by the
piping system. Optical measurements indicated the
turbine coupling end of the baseplate moved vertically
0.60 mm (0.024 in) more than the governor end. That
movement was not believed to be thermal in nature. The
baseplate and support structure temperatures were
measured with an infrared pyrometer and found to be
uniform. Vertical movement was thought to be a result
of the gross horizontal distortion of the baseplate. It
was recommended that the piping system design and
installation be reviewed and modified as necessary
to minimize the forces on the turbine. The turbine was
disassembled and inspected. Evidence of heavy rubbing
was present in the carbon ring seal area. A runout check
of the turbine shaft in the seal area indicated there was
permanent damage. Preparations were made to further
modify the piping system and repair the turbine.
Third Startup (Second Piping Modification)
T
he turbine seals and bearings were inspected.
Both bearings exhibited signs of rubbing but were
cleaned up and reinstalled. Broken carbon steam
seal rings were replaced. Inlet and discharge steam
block valves, originally located within 2 m (6 ft) of the
turbine, were moved to the steam header locations.
Relief valves, located on the exhaust side of the turbine,
were also moved further downstream. The idea was to
minimize the amount of weight in the piping to reduce
the forces that could possibly act on the turbine. Pipe
hangers and supports were inspected and re-cali-
Figure 5 Horizontal movement of steam turbine casing relative to bearing housings during steam
piping warm-up.
Vol . 26 No. 3 2006 ORBI T 55
APPLICATIONS
brated. Field engineers from the pipe design company
inspected and approved the installation.
During an uncoupled run-up of the turbine, vibration
data indicated a rub was occurring. This was revealed
by the truncations on the timebase waveform and the
accompanying flat areas on the Orbit (Figure 6). The
primary vibration component was 1X (synchronous),
with forward precession as the full spectrum plots
(Figures 7a and 7b) show. The unit was shut down and
the bearings, seals, and shaft were inspected. The shaft
had turned blue just outside the steam seal area on the
coupling end as a result of the friction generated by the
rub. Although the seals and shaft were damaged, the
decision was made to run the unit again to facilitate
further diagnostics.
As during the second startup, optical measurement
points were again installed on the steam inlet and
exhaust flanges as well as various other locations on
the piping system. The unit was operated at 500 rpm for
approximately three hours while vibration and optical
data were acquired. Several interesting observations
were made:
1. Vibration data indicated the occurrence of rubs.
2. Optical data indicated the occurrence of turbine and
baseplate distortion.
3. Visual inspection indicated the flanges on the expan-
sion joint closest to the turbine inlet were distorted at
least 3 mm (0.120 in).
4. The wobble foot on the turbine was noticeably
distorted and did not return to its original position
when the unit cooled down. The turbine supports had
to be loosened before the wobble foot returned to its
original position.
An internal inspection of the turbine, performed
immediately after the shutdown, revealed evidence of
heavy rubbing (Figure 8). One of the carbon seals had
Figure 6 Orbit/timebase plot at 4800 rpm showing evidence of rub at turbine inboard bearing during
third startup.
56 ORBI T Vol . 26 No. 3 2006
APPLICATIONS
spun and was broken in several pieces and the shaft
was permanently damaged. The rotor was removed and
the internal alignment of the turbine case was checked.
Significant runout in the nozzle ring [0.75 mm (0.030 in)]
and seal area [0.325 mm (0.013 in)] warranted the return
of the turbine to the manufacturer for closer inspection.
Shop Test at OEM Factory
C
oncerned that damage may have occurred when the
skid was dropped during the original installation, the
turbine was completely disassembled and inspected. No
damage was found. The nozzle ring and seal areas were
adjusted back into tolerance. The turbine was placed on
the test stand with a new rotor, bearings, and seals, and a
cold optical survey was conducted. The unit was coupled
to a water brake, an 1121 kW (1500 hp) load was applied,
and vibration data was collected. After one hour at load,
hot optical measurements were made.
Figure 8 Turbine cross-section showing rub
location.
Figure 7 Full spectrum showing predominance of forward precession for 1X component during third
startup at a) turbine outboard bearing and b) turbine inboard bearing.
a)
b)
Vol . 26 No. 3 2006 ORBI T 57
APPLICATIONS
Figure 9 Shaft centerline plot showing vertical rise of the shaft within a) turbine outboard and b) turbine
inboard bearings.
a)
b)
58 ORBI T Vol . 26 No. 3 2006
APPLICATIONS
Vibration data indicated that the rotor rose straight
up in each bearing [100 m (4 mils) governor end and
37.5 m (1.5 mils) coupling end]. This resulted in a high
shaft attitude angle (Figures 9a and 9b), which would
contribute to low rotor eccentricity within the bearing
and possibly a fluid instability. Polar and Bode plots
were normal during the transient period of the run, while
Orbit and timebase data indicated several light rubs and
one severe 100 m pp (4.0 mil pp) rub. The severe rub
cleared when the machine speed was reduced. Light
rubs occurred intermittently, and maximum vibration
amplitudes cycled from one end of the machine to the
other as the back-pressure to the machine was changed.
Bearing Modification
O
verall vibration levels were generally between 12
and 25 m pp (0.5 and 1.0 mil pp) when no rubs
were present. High shaft attitude angles, sensitivity of the
machine to load and back-pressure, and the intermittent
rubs suggested a bearing load problem. After inspection
of the bearings, the decision was made to rotate each
pressure dam bearing approximately 70 degrees within
the bearing housing. The modification was accomplished
by simply repositioning the locating pin in the bearing.
Clearance of the carbon ring steam seals was increased
by approximately 0.125 mm (0.005 in).
Alignment data indicated typical thermal movements
for that type of turbine. No casing distortion was pres-
ent and no apparent outside forces were acting on the
turbine.
Final Test Run at OEM with
Modified Bearings
T
esting was conducted with a maximum load of 1100
kW (1500 hp). Direct vibration was generally under
12.5 m pp (0.5 mil pp) on all channels (Figures 10a and
10b). No rubs were detected during the run and the
balance state of the rotor was excellent. Synchronous
(1X) frequency was the only component of the vibration
spectrum present at operating speed. Shaft centerline
data indicated the rotor was still riding high in the bear-
ings, resulting in a high shaft attitude angle. However,
the machine was much less sensitive to load and back-
pressure changes due to the bearing modification. The
OEM believed that the high shaft attitude angle was due
to partial arc steam forces lifting the rotor and that full
load steam flow through the nozzle would likely improve
the situation.
Field Re-installation
(Third Piping Modification)
T
he turbine was re-installed and a third piping
modification to strengthen the supports was
completed. It was observed that many of the skid hold-
down bolts were loose. The skid was sounded with a
hammer and was determined to have only about 30%
grout integrity. At that point, the skid should have been
removed, cleaned, and then regrouted. Nevertheless,
the customer instead opted for a pressure grout injec-
tion solution. Anchor bolts were torqued and checked
at 8-hour intervals. Many of the bolts were re-torqued
several times before they held the specified 110 kN-m
torque specification. Before the initial anchor bolt
tightening, the turbine-gearbox and gearbox-compres-
sor alignment was in tolerance. Afterwards, both units
required re-alignment.
Steam piping was connected and the lines were heated.
A dial indicator on the turbine-to-gearbox hub indicated
1 mm (0.040 in) horizontal movement as the steam lines
were heated. Movement over 0.05 mm (0.002 in) was
considered unacceptable. Both steam lines were cooled
down and only the exhaust line was heated by opening
a drain valve. Movement in excess of 0.5 mm (0.020 in)
was measured at the coupling, indicating that the major
problem was with the exhaust line. This convinced the
end-user that the turbine was not capable of withstand-
ing the OEM-specified flange loading. To confirm this,
weights were hung from the exhaust flange and at the
specified load of 287.6 kg (634 lb) a dial indicator on the
hub showed only 0.025 mm (0.001 in) movement. The
end-user was then convinced that the piping system
was flawed and plans were made for a fourth, and
final piping modification: complete redesign of the
steam inlet and exhaust pipe system to allow for larger
thermal loops.
Vol . 26 No. 3 2006 ORBI T 59
APPLICATIONS
Figure 10 Overall vibration amplitude versus shaft rotative speed during shop testing at a) turbine
outboard and b) turbine inboard bearings.
a)
b)
60 ORBI T Vol . 26 No. 3 2006
APPLICATIONS
Success (Fourth Piping Modification)
F
igure 11 depicts the final pipe design as an overlay
of the expansion joint modification. Notice particu-
larly that the thermal loops were expanded down to
grade level. Foundation repairs improved the stability
of the skid, and, approximately six months after the first
startup attempt, the unit finally went into production.
Conclusion
A
s this case history demonstrates, pipe strain can
be a significant source of problems that manifest
themselves in machinery behavior. In this case, the
root problems stemmed from a piping system, and
subsequent revisions did not actually meet engineering
specifications despite analysis with computer programs
and third-party review. A final design employing larger
thermal loops minimized the forces exerted on the
turbine, allowing it to finally be brought into service.
While the initial problems were unfortunate, they were
compounded unnecessarily by attempts to patch a
poor piping design rather than redesign it entirely. Lost
production alone totaled several million dollars, not
to mention labor and material costs. Had the original
recommendations to completely re-do the piping design
been followed, the unit could have been online within
one month.
Figure 11 Final piping arrangement.