Authors: Ngangha Victor Oyama, E. G. Saturday, Ph. D, Iluno Nzubechukwu Ugochukwu, Vincent B. Ikuenayo
Abstract
Centrifugal pumps in refinery power plants are critical, continuously operated assets, yet maintenance decisions are often made without quantitative reliability evidence. This study applies Reliability-Centred Maintenance (RCM) to five pumps (70CP-1101 to 70CP-1105) in the power plant of the Port Harcourt Refining Company using 2024 failure and repair records (44 failures, 8,256 h available time per pump). Reliability indices were estimated with an exponential model at stated mission times, with 90% confidence intervals. Pump-level MTBF ranged from 714 h (70CP-1103) to 1,352 h (70CP-1104); availability ranged from 95.19% to 98.25%; one-week reliability ranged from 79.0% to 88.3% and one-month reliability from 36.5% to 58.7%. The between-pump differences in failure rate were not statistically significant (χ² = 1.87, p = 0.76), so pump ranking is treated as prioritisation rather than proof. FMEA, logic tree analysis and a weighted criticality index identified seven of eight failure causes as high criticality (Group A) for 70CP-1103, leading to condition-directed tasks at weekly and monthly intervals. A workload-based integer programming model was then formulated to size the maintenance workforce. The baseline solution requires three engineers and four technicians (₦46.2 million per year) against the current roster costing ₦189.0 million per year; scenario analysis shows that savings remain at 45–76% for workloads up to six times the baseline. The results indicate that the current roster is not driven by pump maintenance workload alone and should be reviewed with other duties considered.
Keywords:Reliability-Centred Maintenance; Centrifugal Pumps; Logic Tree Analysis; FMECA; Availability; Integer Programming; Refinery