For several years, Grant County Public Utility District (PUD) struggled with high temperatures – and subsequent failures – of the thrust bearings of two units at its 907-MW Priest Rapids hydro project. To solve the problem, the utility installed pressure transducers to measure load on the bearings, then used the measurements to more accurately adjust the bearings’ position.
The Priest Rapids project is on the mid-Columbia River in Washington. The powerhouse is equipped with ten vertical Kaplan turbines. Each umbrella-style generator is rated at 95 MW. Unit commissioning started in 1959 and was completed in 1961.
English Electric designed and manufactured the thrust bearing assembly for each unit. This assembly consists of ten tilting pad shoes supported with an equalizing table. Each shoe is pre-loaded with an adjusting screw. The thrust runner is in two pieces bolted together, and then bolted to a thrust collar. The thrust collar is shrunk fit 0.05-inch onto the generator shaft.
During start up and commissioning, it was necessary to scrape a depression in the babbitt to prevent overheating of the bearings. The scrape patterns evolved through the years with minor bearing wipes, developing their own patterns of areas needing a depression scraped. Often, a bearing set would require a wear-in period and more than one scraping before the temperatures settled. As a result of frequent high temperatures in the thrust bearing assembly, Grant County PUD conducted an inspection and scraping every four years.
Modifying the thrust runner and bearing
In 1995, after several bearing failures, PUD engineers decided to investigate alternatives to hand scraping for solving the overheating problem. On two units they measured bearing movement relative to the support, pressure at the high lift port, and temperature distribution across the leading and trailing edges of the thrust bearing shoe. From an analysis of the test results, the engineers concluded the oil wedge between the thrust runner and thrust bearing (also referred to as an oil film) pressurized the split between the two thrust runner halves, causing it to open slightly. This opening – aided by centrifugal force – allowed oil to flow out the end of the split. The oil leak resulted in unequal load on the thrust bearing, allowing the outer radius of the bearing to move up and briefly make contact with the thrust runner. This contact caused the temperature to rise on the outer radius and across the top of the thrust bearing shoe. As a result of the temperature increase, the bearing deformed into a crowned shape; this caused the top of the bearing to wipe if a depression was not scraped into it.
PUD engineers conducted testing of the thrust bearings and thrust runner to resolve the overheating and failure issues. During data collection, a physical observation of the thrust runner split leak confirmed this unusual phenomena. PUD crews reported that, when the unit was rotated manually (with the high-pressure lift system energized), a stream of oil “squirted” out the split in the thrust runner. Fretting corrosion between the thrust runner and thrust collar near the runner split was attributed to the split becoming pressurized and moving the thrust runner slightly.
As a result of the analyses, the thrust runner on every unit was removed and the thrust runner split machined to achieve a tight fit. The connecting bolts on each thrust runner were shortened to provide more resistance to flexing. All the thrust bearing sets were machined flat without the scraped depression. Another modification involved replacement of a threaded plug for the high-pressure lift system port with a plug that was welded flush. Additionally, all the radial anchor grooves – which were intended to help hold the babbitt in place but can be a source of bonding problems – were machined flat.
From 1995 to 2004, as a result of the modifications to the thrust runner and bearing, PUD was able to discontinue the previously described four-year cycle of scraping, inspection, and maintenance on the thrust bearings. The number of forced outages caused by thrust bearing problems improved from an average of almost one unplanned outage a year to one unplanned outage every three years.
Then, in December 2004, a bearing failure occurred on Unit 1. Three of the ten bearing shoes had sections of babbitt completely removed. This failure was attributed to babbit bond failure on an older rebabbitting process. In the areas of babbitt removal, the bearing shoe had no tin left. The babbitt was previously bonded to the shoe with a trimetal copper process and anchor grooves.
In May 2005, the Unit 1 thrust bearing failed again. A visual inspection showed the babbitt contained fatigue cracking on all the bearings shoes. PUD staff discovered three of the eight bolts between the thrust runner and thrust collar had broken. They also found a crack in the thrust runner split joint, resulting from the additional stress caused by the broken bolts. Powertech Labs performed material failure analysis, consisting of micrographs and scanning electron micrograph pictures. This analysis showed that the three bolts failed in tension due to fatigue.
After this second failure, PUD staff disassembled the entire bearing support mechanism in Unit 1, including the equalizing table, to inspect it for wear or failures. On the surfaces of the thrust runner and thrust collar where they face each other, staff found excessive fretting corrosion. The corrosion caused the thrust collar to be out of tolerance.
To fix the problem, PUD staff would need to machine the thrust collar in place. However, staff concluded in-place machining would be too risky. It would require design and fabrication of a custom machine tool. The high original tolerances for flatness on the thrust collar would be difficult for a custom machine tool. If the flatness tolerance was degraded further during an in-place machine operation, complete disassembly would be required to repair the damage.
The next option – to dismantle the unit even further to machine the thrust collar using a standard available machine tool – was too costly. This option would result in lost power generation from the unit for about two months.
The PUD ultimately decided to replace the thrust bearing runner with a spare and put the unit back in service. The PUD decided the damaged and out-of-tolerance thrust collar would not be repaired at this time.
Before placing the unit back on line, the PUD added monitoring instrumentation to provide feedback on the operating characteristics. PUD personnel placed a resistance temperature detector (RTD) inside the six thrust bearings with no monitoring instrumentation. The PUD already used these RTDs on four of the bearings to measure temperature in the thrust bearing shoe.
Figure 1: Grant County Public Utility District connected pressure transducers to each of ten bearings in Unit 1 at its Priest Rapids hydro project to measure oil pressure at the high-pressure lift port.
In addition, they connected a pressure transducer to each bearing to measure the oil pressure at the high-pressure lift port, as shown in Figure 1 on page 72. The end nut of the bearing was drilled and tapped to allow the pressure at the port of the high-pressure lift system to be measured while the unit was on line. This modification does carry the risk of developing leaks between the oil port and the pressure transducer. A significant leak in the tubing could lead to the loss of the oil wedge and result in a bearing wipe. The risk was minimized by careful installation and pressure testing of all the tubing.
Grant County PUD had one month to assemble the unit and put it into operation, to meet future power demand. The short time available prohibited the use of a less risky, more traditional method of measuring the pressure, such as load cells or submersible transducers installed close to (but not on) the bearing.
In September 2005, the PUD began placing Unit 1 back in service. First, the thrust bearings were pre-loaded with the adjustment screw that holds the thrust bearing shoe up against the thrust runner, according to standard torquing procedures. However, two of the bearings did not leak oil out of the edges of the bearing with the high-pressure lift system energized. If a bearing shoe does not leak oil, the load on that particular bearing is too high. In addition, the unit would not rotate manually.
A second attempt was made at torquing the adjusting screws, yet the results remained the same. The PUD staff had no success with additional attempts to flood the bearing by jogging the pump, raising the pressure relief on the pump, and manually rotating the unit with winches. They were able to get the unit to rotate, albeit with difficulty, by lowering two of the thrust bearings by 0.001-inch. Although the problems were not solved, the unit was placed back in service because the outage schedule had been exhausted.
When the unit was in operation, the pressure fluctuation on every pad (measured by the transducers) looked similar to a sinusoidal wave function with a frequency of twice per revolution. During manual rotation while the unit was off line, PUD staff determined the pressure was lowest at the point when the thrust runner split was located directly over the high-pressure lift port (the pressure transducer port). Conversely, the pressure was highest when the split was 90 degrees from the port.
Staff recorded the minimum, maximum, and average value of each pressure transducer. The difference between the minimum and maximum value represented the level of pressure fluctuation on each pad during one revolution. The average value for each pressure transducer was used to qualitatively compare the load on each of the ten thrust bearing pads relative to each other.
When Unit 1 was running, the average difference in bearing pressure was 1,770 pounds per square inch (psi), with the highest average bearing pressure around 3,000 psi and the lowest around 1,230 psi. The average difference between the minimum and maximum values on a single pad was 400 psi.
After 18 days of operation, high thrust bearing temperatures developed, and the unit had to be taken out of service. PUD staff lowered the two highest bearings and raised the two lowest bearings until the bearing pressures (with the high pressure system operating) were closer in magnitude.
After the second adjustment, the average difference between bearing
recordings was 514 psi with the high-pressure lift system running, and 1,111 psi when Unit 1 was on line. The 1,111 psi was an improvement from the previous value of 1,770 psi. Average operating temperature of the bearings was reduced by 2 degrees Celsius (C) under the same approximate operating head and generator load.
In September 2006, Unit 7’s thrust bearings wiped. During the bearing replacement, pressure transducers were installed on each thrust bearing so that pressure measurements could be compared to Unit 1. The thrust bearings in Unit 7 were pre-loaded with the standard torquing procedure and then adjusted up and down using the following process, developed by PUD personnel:
First, the high-pressure lift system was energized and the unit was rotated until all the bearing pressures stabilized. Then, the average pressure for each bearing during one revolution was calculated. Third, the unit was raised off the thrust bearings, and several designated bearings were lowered or raised based on the average values. The highest and lowest pressure thrust bearings were adjusted first. Fourth, the high-pressure lift system was de-energized, thus lowering the unit back down on the bearings. Then, the pressure in each thrust bearing was recorded again. This process was repeated until the difference in average pressure could no longer be improved.
This process does take longer (a full day vs. the previous half day of preloading only). However, it provides the ability to ensure the load is equalized across all ten thrust bearings. The more equally distributed the total load is on the bearing, the better the operating characteristics. However, this also highlights a drawback to the pressure transducers vs. load cells. With a load cell, Grant County PUD most likely could adjust the bearings in one step. The pressure transducer value cannot be recorded while making the adjustment, so the entire process was necessary in order to adjust the bearings.
The pressure data collected from Unit 7 was compared to that of Unit 1. The average difference in bearing pressure on Unit 7 was 300 psi while running compared to the 1,111 psi on Unit 1. The average bearing pressures (with the high lift system operating) for Unit 7 differed by 400 psi between the high and low bearing when they first pre-loaded and were improved to an average bearing pressure of 80 psi between the high and low bearing. Operating under the same approximate head and load, the highest bearing pressure on Unit 7 was 1,600 psi while on Unit 1 it was 2,500 psi. The data recorded from Unit 7 confirmed that the thrust bearing adjustment could still be improved on Unit 1.
In December 2006, the PUD conducted a planned outage of Unit 1 to make a third adjustment on the thrust bearings, using the same process as it used on Unit 7. The average difference in bearing pressures was decreased to 182 psi compared to 514 psi with the high-pressure lift system operating only, and 742 psi compared to 1,111 psi with the unit in operation. The adjustment resulted in an average operating temperatures reduction of 0.5 degrees C by evening out the load on the bearings.
Although the evenness of the load distribution in Unit 1 improved significantly from the first adjustment (from 1,770 psi to 742 psi), the improvement did not allow the unit to match the average bearing pressures in Unit 7. Further observation of the data revealed that the average bearing pressures from Unit 1 have a wider spread when the unit is loaded with hydraulic thrust than those on Unit 7. Some of the Unit 1 bearings do not pick up an equal share of the load. The reason for this discrepancy has not been determined.
Units 1 and 7 are now running satisfactorily. During an upcoming planned outage, PUD staff will close a valve immediately downstream of the drilled port in all ten bearing assemblies in Unit 1 to further minimize the risk of leaking oil. The units are scheduled for a major rehabilitation in five to ten years.
Good written record keeping is important for finding solutions to recurring problems. For the Priest Rapids thrust bearing problems, finding historical information took significant effort, and some information had been lost. During a failure involving an unplanned unit outage (especially a reoccurring failure), the focus is often on returning the unit to service and not determing the cause of failure. Yet, observations noted during a failure often help with future problems. Grant County PUD has improved its record-keeping process by archiving the information found so far using Maximo from MRO Software and producing internal reports on recent failures.
The pressure transducers have turned out to be a cost-effective alternative to load cells for measuring the bearing load. They cost less, are quicker to obtain, and are easier to install. Originally, Grant County PUD elected to use the transducers rather than load cells owing to a limited time for instrumentation. To date, the transducers have provided accurate data, useful in making bearing adjustments.
The addition of more temperature data collection and oil pressure data on Unit 1 led to a new adjustment procedure for the troubled unit, which has allowed it to operate without continued thrust bearing failures. The PUD believes the new procedure for thrust bearing adjustment is superior to the old process of torquing that will result in more equal bearing loading and longer bearing life.
– By Dale Campbell and Pat Oldham, mechanical engineers, Grant County Public Utility District, 15655 Wanapum Village Lane S.W., Beverly, WA 99321; (1) 509-754-5088, extension 3157 (Campbell) or (1) 509-754-3541 (Oldham); E-mail: [email protected] or [email protected]