By: Dr. Matthew Smillie, Consultant Engineer, Quest Integrity Group
Download the PDF version here.
A number of technological and scientific advances through the ages have helped us reach the standard of living we enjoy today. One of the most important was the development of anti-friction bearings. These bearings allowed the design, construction and operation of innumerous mechanical devices for countless mechanical purposes. For example, rotating machinery relies on anti-friction bearings to support and align their rotating components during operation. They also help prevent catastrophic clashes between the rotating and stationary parts.
The rotors of heavy-duty gas turbines are exposed to the heat of combustion during their operation. This thermal loading, in addition to the mechanical loading due to self-weight and rotation, provides a challenge to the designers of gas turbines when aiming for the reliability demanded by the commercial requirements of operation.
This article briefly introduces the particular issues of bearings in heavy-duty gas turbines, and then reviews a case study of a gas turbine bearing that could no longer hold its load and the situation surrounding its failure.
Large gas turbines (greater than 20 MW power output) can be generally split between the aero-derivative models and the heavy-duty or frame-type industrial turbines. Aero-derivatives, because of their aviation heritage, size and general multiple-shaft layout, tend to use typical rolling-element bearings to support and locate the shafts.
Because of the lower restrictions on weight and size of components, heavy-duty gas turbines exclusively use large and rigid fluid film plain journal bearings to support and locate the rotor radially within the casings. Some older designs required three bearings to maintain the required clearances along the shaft (e.g. MS7001EA, MS9001E). Smaller, more modern designs with stiffer rotors reduced this to two bearings (e.g. MS6001B, GT13E2, etc.). Plain-journal bearings or tilt-pad bearings are used, depending on the particular manufacturer’s preference and requirement for accommodating misalignment due to rotor flex or casing distortion during operation of the turbine. Although it is not addressed further in this article, a thrust bearing is always utilized to locate the shaft axially.
All modern heavy-duty gas turbines follow the same form factor: ambient air enters an axial-flow compressor and then exits into one or many combustion chambers at high pressure and temperature due to the compression. Then, due to combustion, a further rise in temperature occurs, and the resultant hot-gas stream is expanded through an axial-flow turbine, running on the same shaft as the compressor, to extract power. The hot combustion gases flow through the turbine, between the shaft and casing (both protected from most of the heat by a combination of heat shielding and cooling air) before exiting the rear of the machine. The exhaust temperatures of most heavy-duty gas turbines are generally between 500 °C and 600 °C, depending on the duty, fuel and efficiency of the turbine.
It is in this hot exhaust gas that the hotend bearing must sit, relying on extensive thermal management of the exhaust frame, lubrication oil and cooling air to maintain its position (and that of the spinning rotor) and to maintain its own integrity. Many design solutions from the major manufacturers have been incorporated to minimize the impact of thermal distortion on the hot-end bearing (e.g. MHI’s tangential struts for their M501 and M701 models or Alstom’s adjustable bearing housing supports in their GT13E2 model).
The MS6001B (or Frame 6B) is a widely utilized General Electric heavy-duty design. It is generally used for power generation, at both 50 and 60 Hz, outputting between 38 and 42 MW depending on the model. The design is "E-class”, derived from 1970s technology, but has been continuously upgraded since its introduction in 1978. Although it uses a reasonably flexible built up rotor, it requires only two radial bearings for support (Figure 1). Both are plain elliptical journal bearings. Power is taken off the hot (turbine) end of the machine.
|Figure 1 MS6001B rotor. Arrow indicates the hot-end bearing journal|
The particular unit in question was in a co-generation role, supplying steam and power to neighboring industry. The unit had seen approximately 120,000 fired hours and an average of five starts per year after commissioning. Due to demand fluctuations, the unit was running base load during the day and part load at night.
The unit had a history of vibration and alignment issues during commissioning and while in service. Alarm levels of vibration (approximately 13 mm/s) were often exceeded during start-up. Base load levels of vibration were within the OEM limits but always close to alarm. Two bearing temperatures for the hot-end bearing were measured as part of the control package. These temperatures were always noted to be high, but below alarm. A number of bearings had been replaced during the life of the unit, based on their condition during scheduled or opportunistic maintenance.
The replacement cost of the bearing was low, so there were little to no cost pressures to solve what appeared to be an acceptably managed "quirk” of the unit.
One afternoon the alarm temperature (130 °C) on the hot-end bearing was reached. The load on the unit was reduced immediately. A subsequent reduction in the bearing temperature was observed. Plant operations decided to monitor and shutdown the unit if the temperature exceeded 150 °C. It is important to note that there was no trip limit on bearing temperature in the OEM package.
Later in the evening, the alarm temperature on the bearing was reached again. Thirty seconds later 150 °C was reached and a manual shutdown was initiated. Two minutes after the shutdown was initiated, the bearing temperature was measured at 200 °C. The vibration levels increased, causing an automatic trip of the unit five minutes after the shutdown was initiated. During the rundown, bearing temperatures up to 420 °C were measured. The tin-based white metal used in journal bearings melts between 180 and 240 °C (depending on the lead content).
Disassembly of the bearing housing revealed an extensively scored shaft journal, also deposited with remains of the bearing white metal. The top half of the bearing was relatively unaffected. The bottom half had almost all the white metal removed from the steel shell of the bearing, which was itself severely rubbed, overheated and distorted.
The material condition of the rotor journal was of great concern. The journal was part of a stub shaft bolted to the turbine section of the rotor. Replacement of the stub shaft would require shipping to a facility capable of disassembly, replacement of the shaft, and balancing the reassembled rotor before shipping back to site. Metallurgical inspection of the journal surface revealed localized regions of high hardness (exceeding 600 HV) and transformation microstructures that showed that parts of the journal surface had exceeded 720 °C during the incident. The hardening extended to a depth beyond the re-machining limit of the journal. The rotor stub shaft was in a non-serviceable condition.
The consideration of shipping and repair lead times led to the decision to replace the entire rotor, as this would return the machine to service much quicker than any other option. As a replacement rotor was sourced, an investigation into the physical cause of the bearing failure was undertaken. Since most of the physical evidence on the failed bearing itself was missing, examination of bearings previously removed from the unit was undertaken (e.g. Figure 2). The following was noted: - Features typical of long-term fatigue failure of the bearing white metal were observed in a number of the replaced bearings.
- There were indications of edge loading due to slight angular misalignment with respect to the shaft.
- Extrusion of white metal over the edge of the bearing indicated creep of the white metal during service due to high operating temperatures of the bearing.
|Figure 2 Lower half of the hot-end bearing previously removed from machine. The damage seen was typical of edge-loading causing fatigue in the white metal.|
The physical features indicating edge-loading were unexpected, as numerous alignment checks performed at outages showed no indications of serious misalignment. Further inspection of the bearing housing and exhaust frame revealed the following:
- The bearing housing was nose-up relative to the axis of the machine.
- Fiber insulation in the outer skin of the exhaust frame was almost completely missing.
- Internal insulating pads (Figure 3) inside the exhaust frame had dropped in the lower half of the exhaust frame, partially blocking the entry of cooling air on one side on the machine (Figure 4).
|Figure 3 Insulation pads inside exhaust frame protect cooling air for struts (flow path arrowed). Lower half pads had dropped, blocking the flow of cooling air to (bearing supporting) lower struts.|
|Figure 4 Note dropped pad blocking cooling holes on one side of the lower exhaust frame (left). Compare with pad held in place and unblocked holes on other side of lower exhaust frame (right).|
The latter findings were quite significant, as the manual for the turbine [General Electric, MS6001B Gas Turbine Manual, Volume II, 1995] states (emphasis added):
Exhaust frame radial struts cross the exhaust gas stream. The struts must be maintained at a uniform temperature in order to control the central position of the rotor in relation to the stator. This temperature stabilization is accomplished by protecting the struts from exhaust gases with a metal wrapper fabricated into the diffuser. This wrapper also provides a circuit for cooling air. Turbine shell cooling air flows through the space between the struts and the wrapper to maintain uniform temperature of the struts.
This describes the method by which thermal expansion of the hot exhaust frame (and location of the hot-end bearing) is managed. The loss of insulation and partial loss of cooling air resulted in an in-service (hot) condition that could not maintain the required (cold) alignment. The increased loading on the bearing was exacerbated by the inherent vibration of the rotor. The combination of all factors resulted in an increased loading on ”" and higher temperature of ”" the hot-end bearing, directly resulting in a decreased service life and ultimately failure during operation.
During the investigations, a number of other examples of MS6001B hot-end bearings were found with similar indications of distress, revealing that the problems with this unit were not unique. The later GE "F-class” designs have a completely different bearing arrangement using two tilt-pad bearings, and driving from the cold (compressor) end of the machine, which "…eliminates the alignment issue at turbine end bearing” according to the Electric Power Research Institute [Design Evolution, Durability and Reliability of General Electric Heavy-Duty Combustion Turbines: Pedigree Matrices, Volume 3, Palo Alto, CA, 2007].
During reassembly, the insulation and cooling flow was restored to the exhaust frame. The housing alignment was corrected using angled shims. Laser alignment of the bearings was undertaken to ensure the required tolerances were reached in the fully assembled state. A replacement rotor was installed in the machine. Vibration levels and bearing temperatures were considerably lower with the new rotor. The plant installed new instrumentation, alarm and trip limits related to bearing performance monitoring. The unit has been running with little issue for more than a year since the incident.
The damaged rotor was recently shipped to a repair facility for refurbishment. During de-stacking of the built-up rotor, it was apparent that there had been loss of interference fit on a number of compressor wheels that made up the compressor rotor. It is speculative at this stage, but such a loss of interference fit may have influenced both the rigidity and heat transfer through the rotor during thermal transient conditions (such as start-up), leading to distortion (e.g. bowing) during the thermal transients and resulting in increased vibration. It also may have contributed to the noted thermal sensitivity of the rotor and high vibration levels during base load operation.
The author would like to acknowledge the following:
- The affected facility, which graciously gave permission for this case study to be discussed and the findings shared publicly, on the condition of anonymity.
- Dr. Maxine Watson, Director ”" Power Resources, Quest Integrity Group, for her contributions in relation to the tribological aspects of the investigation.