Centrifugal Pump Failure Trend on a Cooling-Water Loop
A chemical plant cooling-water loop has two 75 kW centrifugal pumps arranged duty/standby. Over 28 months the duty pump has failed five times: three mechanical-seal leaks, one bearing seizure, and one coupling failure. Seal replacements lasted 4, 9, and 11 months. On the last run, vibration rose from 2.1 to 4.8 mm/s RMS in six weeks and bearing temperature climbed 8 C. Operators run the pump at 60-70% of design flow with the discharge valve throttled. The suction strainer differential pressure log shows 0.6 bar, against 0.2 bar when clean. No root-cause analysis has been performed; the standard response is to replace the failed part and return the pump to service. An unplanned outage costs about $18,000 in lost production, and a maintenance shift costs $1,200. The reliability engineer proposes quarterly preventive maintenance on both pumps, replacing seals and bearings whether or not condition data indicate a problem.
1. Evaluate the quarterly PM proposal. Would it address the failure history? Justify your answer with the evidence in the setup.
Read the worked answer
The proposal is not supported by the evidence. The failures are not a single wear-out mode: seals, bearing, and coupling point to different stresses, and seal life varied from 4 to 11 months, which is not a stable interval. The rising vibration and bearing temperature before failure, together with 0.6 bar strainer differential pressure against 0.2 bar clean and sustained 60-70% throttled operation, suggest a chronic causal condition such as cavitation, suction recirculation, misalignment, or contamination. Replacing seals and bearings quarterly would consume parts and introduce infant-mortality risk without removing the cause. It would also apply PM to the standby pump, which has no operating hours. A better immediate action is to clean or replace the strainer, inspect the suction line, review the operating point, check alignment and lubrication, and collect a vibration spectrum. Use the failure data to identify the bad actor rather than scheduling by calendar.
2. Identify the most likely failure mechanisms and list the measurements you would take before changing the maintenance strategy.
Read the worked answer
Most likely mechanisms are low net positive suction head or cavitation from the fouled strainer and throttled operation; discharge recirculation at 60-70% flow causing radial load, shaft deflection, and seal or bearing distress; and possible misalignment or coupling wear. Collect suction and discharge pressure, flow, pump speed, NPSH margin, and strainer differential pressure. Take a vibration spectrum and phase reading, bearing temperature trend, lubricant analysis, seal-face photographs, shaft runout, and alignment data. Compare the operating point with the pump curve. Confirm whether the coupling failure followed bearing degradation. A root-cause analysis should test these hypotheses, not merely record which parts were replaced.
3. State the decision rule you would use to choose condition-based versus interval-based maintenance for this pump, and what would make you revisit it.
Read the worked answer
Use condition-based maintenance when a measurable precursor precedes failure and the cost of monitoring is less than the expected failure cost. Use interval-based PM when failure is age-related, a clear wear-out pattern exists, and the residual risk is acceptable. Here the vibration and temperature trends are precursors, so condition monitoring plus correction of the operating point is preferred. Revisit the decision if failures become random and undetectable, if monitoring proves unreliable, or if the standby pump develops a start-up reliability problem. Review after six months using failure rate, downtime, and cost per operating hour.