Armed with powerful predictive maintenance tools, investigators took the right steps to eliminate a problem with a critical cooling-water pump before it could impede the progress and up the cost of a planned shutdown.
This case study details the use of oil analysis and other powerful condition-monitoring techniques in the identification and resolution of problems with critical equipment at a large nuclear power plant. The equipment in question was a centrifugal cooling-water pump used in the shutdown process before an outage. It goes without saying that in a nuclear plant, each piece of critical equipment must have a spare that is capable of functioning at 100%. In this case, two cooling-water pumps were both required to function at 100% during shutdown. These units were rotated during shutdown events, with one being the main pump and the other the spare.
Data results & evaluation
Both centrifugal pumps were normally sampled every six months for routine in-house oil analysis. Pump #2 (P2) was to serve as the main pump for an upcoming outage; Pump #1 (P1) was to serve as the spare. During routine evaluation, the spare pump (P1) showed a high lubricant viscosity of 58.2 cSt. Table I illustrates the results for a more complete evaluation comparing the results to the previous sampling period and the new oil reference.
Analytical ferrography was run on the oil sample and indicated large amounts of brass sliding wear and small ferrous-particle wear.
Evaluation of the main pump (P2) indicated no major wear metals. With an oil viscosity of 33.2 cSt and an ISO Cleanliness Code of 19/16, this unit was deemed ready for use during the outage.
The problem came about in properly preparing P1 as backup for the outage. The investigation found that the wrong oil had been added to the backup pump and a flush and change-out was scheduled. (The results of the oil analysis from the bearing housing are shown in Table II.) At that point, the pump was drained, flushed and run for 30 minutes.
Based on the results in Table II, investigators concluded that the oil had been added without a flush, resulting in a higher viscosity than the added new oil, and that the wear particles in the bottom had been agitated—which, in turn, led to high wear particles in the sample.
After flushing, P1 was filled with new oil. It was then run for 12 hours with vibration and infrared thermographic readings taken to identify any bearing defects. The results of a subsequent oil sample evaluation are shown in Table III.
The flush and new oil brought the viscosity back in range, but wear metals were still high. The question was where were they coming from, and would they fail the pump if it had to be run for 10 days during the outage? Infrared thermography showed no problems in the pump; the vibration readings, though, turned up the following results:
- All readings low
- Bearing defect frequencies very low
- Possible very early inner- and outer-race defects (spike energy)
- Bearing pre-Stage 1 failure
The investigators, therefore, began trying to identify the source of the wear metals and the effect on the pump’s reliability. The first step was to determine the metallurgy of the pump components and to utilize a powerful technique in the determination of the wear particles’ composition by element and amount. These particles were evaluated with use of a scanning electron microscope (SEM), as depicted in Fig. 1 (click for image).
SEM utilizes electron versus light to form an image of the particle, thus allowing for accurate determination of the element and its percentage in a particle. The samples were prepared with the use of a glass ferrogram slide and a patch. Each of the preparation techniques had its advantages and disadvantages, and the results are a composite of both.
Before the results of SEM could be utilized, a comparison needed to be made to the actual composition of the pump components that could cause the wear. The pump and its components are illustrated in Fig. 2.
Individual manufacturers were contacted to obtain the exact metallurgy of the components they supplied for the P1 unit. (Table IV lists the elemental composition of these components.)
After matching the SEM results with the component metallurgy, the investigators concluded that the wear came from the inboard cover (bronze) and the deflector (grey cast)—components that have only a minor impact on pump performance. The next question was whether the high levels of abrasive wear generated by these two components would affect bearing life.
The ultimate feasibility of utilizing P1 as a backup without repair had to be established. An outage was scheduled and not enough time was available to repair the pump. Each day of delay would cost $1 million. That’s why it was so crucial to determine if the pump could last a maximum of 10 days (or 240 hours).
The L10 rating of the bearing based on its dynamic load was 5,000,000 hours under ideal conditions. The penalty for a contaminated lubricant reduced the factor by about 500. The L10 life of the bearing was calculated to be 96,700 hours with the contamination penalty.
Stage 1 failure of a bearing means there is probably 10% L10 life remaining—which meant P1′s bearings had about 10,000 hours of life left in them. Since the maximum life required for these bearings to function in the pump for the outage was 240 hours, there was a safety factor of 40 or more.
Based on the bearing calculations, it was determined that no repair was required until after the outage. The following recommendations were made:
- Vibration readings were to be taken at least once per day during the outage to detect early signs or acceleration of bearing damage.
- P1 was to be scheduled for teardown and inspection of its bearing-housing components soon after the outage.
- All parts were to be in stock for repair.
- After the outage, the P1 unit was repaired. At that time, the major problem was found to be in the inboard defector disks. Improper installation of these disks was resulting in metal-to-metal contact, which, in turn, was generating high levels of wear metals.
All the right steps
Because of the enormous consequences of failure, nuclear power plants require an extremely high level of equipment reliability. This case history is a good example of a highly competent reliability group utilizing the proper predictive maintenance tools to identify a problem early—and make the right decisions to deal with it.
As noted in this example, criticality is not based simply on the size of the equipment. The condition of a centrifugal cooling-water pump with a small lubricant reservoir—a pump used only during outages—was the key factor in whether an outage would be on schedule and not incur a large penalty for any delay.
While oil analysis was the main tool, infrared thermography and vibration analysis were also utilized in arriving at the correct conclusion. Identifying the individual components in the bearing housing and contacting the OEMs for complete metallurgical information was a key step. The use of SEM, a very sophisticated tool not normally used in oil analysis, supplied the comprehensive information necessary to determine exactly which components were causing the wear particles in the system. Case closed. LMT
The author thanks Kevan Slater, one of the most practical and knowledgeable reliability professionals in the industry, for his assistance in the preparation of this article. In addition to Figures 1 and 2, Slater also supplied Tables I, II, III and IV.
Contributing Editor Ray Thibault is based in Cypress (Houston), TX. An STLE-Certified Lubrication Specialist and Oil Monitoring Analyst, he conducts extensive training in a number of industries. Telephone: (281) 257-1526; e-mail: firstname.lastname@example.org