You are on page 1of 8

CASE HISTORY

Diagnosing a Steam Turbine


This case history documents how GE Energys System 1* software was used in identifying the root cause of a unique high vibration problem encountered during the post-maintenance start-up of a critical mechanical-drive steam turbine at the EQUATE Petrochemical complex in Kuwait.

Governor Problem at EQUATE Petrochemical

Joy.P.Francis. Specialist, Condition Monitoring EQUATE Petrochemical Company, Kuwait puthenjf@equate.com


3 6 O R B I T Vol.29 No.1 2009

CASE HISTORY

Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 7

CASE HISTORY

Machine Description
The machine detailed in this case history is a 1547 kW back-pressure steam turbine (Figure 1) driving a high-pressure multi-stage boiler feed pump at 4200 rpm. The turbine is forced lubricated, incorporates tilting-pad thrust bearings at the steam end, and uses a Woodward Peak 150 governor for speed control. The entire train is monitored continuously via a Bently Nevada 3500 Series Machinery Protection System connected to System 1 software.

Events Leading up to the Machinery Problem


During a turnaround in April 2006, the existing labyrinth seal was replaced with a special bearing isolator (BI) as a part of a modification to arrest an oil leak from the turbine bearings. This particular BI is slightly wider than conventional designs; it also incorporates an extra oil deflector ring at the bearing ends. The cross-sectional diagram of Figure 2 shows the BIs and the oil deflectors on the turbine rotor.

Figure 1. Back-pressure steam turbine running at 4200 rpm and driving a multi-stage, highpressure boiler feed pump (not shown in photo to right of coupling guard).

Initial Vibration Problem on Outboard Bearing


On 5 April 2006, the turbine was started for an uncoupled test run, and the shaft vibration amplitude at the exhaust end bearing recorded more than 100 microns pk-pk (3.94 mils pk-pk), reaching the trip setpoint. With System 1 software in place and providing continuous information on machine condition, valuable data were captured and available for diagnosis. However, since the Keyphasor* probe was installed on the pump (not the turbine), only unfiltered vibration signals were available for analysis.
[Editors Note: Good engineering practice is to install a Keyphasor phase reference transducer on the driver rather than on the driven machine, allowing a phase reference for uncoupled test runs, exactly as in the scenario described here. This is why industry standards such as API 670 require the phase reference transducer to be mounted on the driver rather than driven machine only.]

Figure 2. Cross-sectional diagram showing arrangement of bearing isolators and oil deflectors on turbine rotor.

shows the results of this run; analysis of this time signal indicated high runout at the drive end probe landing area. This was initially puzzling because the turbine inspection report showed acceptable shaft runout values, conflicting with the most recently acquired data. The turbine was shut down and runout tests were

The unit was then run at slow-roll speed to determine whether the readings were in fact high vibration or simply excessive mechanical/electrical runout. Figure 3

conducted on both bearings. Runout of 100 microns pk-pk (3.94 mils pk-pk) and 26 microns pk-pk (1.0 mil pk-pk) were detected at the exhaust and inlet ends respectively.

3 8 O R B I T Vol.29 No.1 2009

CASE HISTORY

Figure 3. Timebase signatures gathered during slow-roll conditions from turbine exhaust (top) and inlet (bottom) bearing radial probes respectively. The waveforms are indicative slow-roll runout or so-called glitch.

Figure 5. Following removal of the oil deflector rings and reinstallation of the original labyrinth seals, three separate startups were attempted but aborted due to high vibration amplitudes.

Figure 4. Cross-sectional diagram of turbine rotor showing close proximity of oil deflector rings to radial vibration probes. Because the face of the oil deflector rings was not perfectly perpendicular, the wobble was observed by the probes and accounted for the high slow-roll runout readings.

Figure 6. Timebase signatures gathered during 44 rpm slow-roll conditions from turbine exhaust (top) and inlet (bottom) bearing radial probes respectively. The amplitudes were considerably lower than before removal of the oil deflector rings, but still above API recommended limits.

Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 9

CASE HISTORY

Investigation into the source of the runout focused on the newly installed oil seal deflectors (Figure 4). These were found to have high face runout and were located in very close proximity (~ 2 mm) to the radial probes. The deflector is shrunk fit to the shaft and the face runout (i.e., wobbling action) of this ring was within the observable sideview field of the radial probe; consequently, it introduced erroneous readings into the radial vibration signal. Based on these findings, it was decided to revert back to the previous labyrinth seals due to time constraints and the practical difficulties in straightening and correcting the perpendicularity of the welded oil deflector rings. Figure 7. Bode plots during uncoupled run from turbine exhaust (top) and inlet (bottom) bearing radial probes respectively. The amplitudes were within expected ranges. exhaust end bearings as shown in Figure 7, staying below 40 microns pk-pk (1.6 mils pk-pk). In addition to the Bode plots of Figure 7, spectrum plots were also examined. These showed a high-amplitude component occurring at a very low frequency (approximately 0.625 Hz). This was initially thought to be spurious signal noise; however, analysis of the shaft centerline plots revealed a different scenario. The shaft centerline plot is polar representation of the shafts centerline as a function of time or changing speed. It is often used during startup or coastdown to show how the shaft climbs up the oil wedge to assume a stable position within the bearing clearance, graphically depicting the average radial position of the shaft within its bearings at various speeds. The shaft centerline plot of Figure 8a does not show expected behavior; instead, it depicts an almost instantaneous (note that the time stamps are all within two seconds of one another) change in shaft position from the bottom of the bearing clearance at the 6:00 position to an unstable position in the top half of the bearing clearance circle. This abrupt movement of the shaft towards the X-probe during the coupled startup is highly unusual and the timebase plot of Figure 8c confirms the nearly instant step-change in the average value. Essentially, it shows a DC step change (the gap voltage) upon which is superimposed the AC vibration waveform.
Continued on page 42.
4 0 O R B I T Vol.29 No.1 2009

Continuing Problems
The turbine rotor was dismantled to revert back to the previous labyrinth seals. The outage was also used to perform an overhaul of the governor actuator to correct hunting of the governor valve that had been observed during the previous start-ups. Following this work, the turbine was restarted on 9 April 2006 for a coupled run. Unfortunately, the machine experienced what appeared to be very high vibration amplitudes of 150 microsn pk-pk (6 mils pk-pk) immediately after start-up at the steam end x-probe as shown in Figure 5. As a result, three separate startup attempts had to be aborted. Further start-up attempts were abandoned until detailed vibration analysis could be conducted. The slow-roll runout levels were again checked on the steam and exhaust ends at very low speeds of around 44 rpm (Figure 6). The runout levels were 33 and 13 microns pk-pk at steam and exhaust ends respectively. Though the magnitude of this runout was above the API limits, it was not enough to explain the excessive momentary vibration increase at start-up during the coupled run. Consequently, the root cause of the high vibration remained unanswered. An external speed probe was temporarily installed to allow the capture of filtered vibration data, and the turbine was tested uncoupled by running up to full speed. The vibration levels were acceptable at both steam and

SIDEBAR ARTICLE

CASE HISTORY

Considered one of the worlds leading petrochemical producers, EQUATE Petrochemical Company is a Kuwait-based international joint venture between Petrochemical Industries Company (PIC), The Dow Chemical Company (Dow), Boubyan Petrochemical Co m p any (BP C) an d Q urain Pet ro ch emical Industries Company (QPIC). As a world-class petroc h e m i c a l p r o d u c e r, EQ UAT E currently provides markets in the Middle Eas t, A sia, Africa and Europe with high-quality petrochemical products. E Q U AT E p r o d u c e s a n d i t s commercial face, the EQUATE Marketing Company (EMC), marketsa wide range of p o lyet hy l en e (PE ), f ro m lin e ar l o w d en si t y (LLDPE) to high molecular weight high-density resins (HDPE). Recently EMC has also begun selling Styrene Monomer (SM) produced from Greater EQUATEs expanded facilities. EQUATE also produces Monoethylene Glycol (MEG) and

Diethylene Glycol (DEG), which are marketed by MEGlobal, a separate PIC and Dow joint venture. To meet ever-increasing worldwide demand for excellent petrochemical products, EQUATEs shareholders have completed a multi-billion expansion project which greatly increases the existing production of PE and EG. Emerging as Greater EQUATE, this $3 billion global-scale venture added SM to EQUATEs product por tfolio. Paraxylene and Benzene are also produced at the same location from a new wo r l d - cl a s s A ro m a t ic s p l ant which is managed by EQUATE. Hand-in-hand with being a leading market player, EQUATE is an empowered, competency-based organization applying best practices and leading technology i n m a r ke t i n g , s a l e s , re cr u i t m e nt & ca re e r development, environment, health & safety, industrial operations, and administrativeORB I T 4 1 affairs. Vo l . 2 9 N o. 1 2 0 0 9

CASE HISTORY

Figure 8. The shaft centerline plot (a) at left depicts an almost instantaneous change in the shafts average position as can be noted by the timestamps which are all within two seconds of one another. The amplitude trend (b) at upper right provides another view of this sudden change, as does the timebase waveform (c) at lower left which assumes the shape of a step change upon which is superimposed the AC vibration signal.

This sudden change in gap voltage as the shaft abruptly shifted towards the X-probe was detected and interpreted as vibration by the X-probe, resulting in the spiking vibration of Figure 5. Since the phenomenon was observed only during the coupled startup, it was apparent that the shaft movement was steam-induced (the steam throughput is much higher in a coupled run than in an uncoupled run due to the additional load imposed by the pump). As such, attention was focused on the turbine governor control as the potential source of problems (Figure 9). Figure 9. Governor actuator assembly.

Root cause Investigation


Investigation revealed that when the governor actuator had been overhauled during reinstallation of the original labyrinth seals, the actuator stroke was changed, altering the start-up sequence. The governor, initially configured in auto-start mode, had been modified to manual mode; thus, the turbine ramped directly to minimum governor speed without dwelling at the slowroll speed. As a result, the sudden inrush of the excessive steam flow momentarily pushed the rotor upward. This upward force, along with the bearing fluid forces, caused a resultant force in the direction of the X-probe. This phenomenon did not occur during uncoupled runs, as the magnitude of steam throughput is much less than in coupled load runs.

4 2 O R B I T Vol.29 No.1 2009

CASE HISTORY

Figure 10. Trend of vibration data during final startup (bottom) shows vibration levels below 60 microns pk-pk (2.5 mils pk-pk), consistent with normal levels. The shaft centerline plot at upper left also shows normal response, with the centerline climbing up the oil wedge as the shaft speed increases from 0 to 4200 rpm. For a shaft rotating in the counter-clockwise direction, it will slowly climb up the right wall of the bearing with increasing speed, exactly as shown, before assuming a stable position at running speed.

To test this hypothesis, the turbine was started slowly by manually controlling the steam admission. As expected, the turbine ramped slowly to full speed with very low vibration levels and a normal shaft centerline response as shown in Figure 10.

units can only run when this unit runs. Without dynamic data capture and a full complement of plot types such as shaft centerline, slow-roll timebase, and spectrum, the problem would have been extremely difficult to diagnose and would have taken considerably longer. As such, this case history emphasizes the importance of different formats of vibration data provided by GE Energys System 1 software. The appropriate use of this data helps the Equate condition monitoring team to fully comprehend the problem by effective and accurate vibration analysis. In this case, the shaft centerline plot, when examined with respect to time, was particularly instrumental in the accurate and timely diagnosis of both the problem and its root cause.
* denotes a trademark of Bently Nevada, LLC, a wholly owned subsidiary of General Electric Company.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 4 3

Conclusion
Left unchecked, the high initial steam flow caused by the improper governor startup control has the potential to deform the shaft. The ability to diagnose this problem allowed us to not only prevent possible permanent damage to the shaft, it saved us approximately one weeks production losses due to a cascade effect whereby other

You might also like