Failure Analysis, Energy Loss Quantification, and Time-Series Forecasting for Submersible Pump Maintenance at an Iron Ore Mine in Nigeria

This study investigates submersible pump failure patterns at an iron ore mine in Nigeria, quantifies associated energy losses through a theoretically derived and empirically validated formula, and develops a time-series forecasting model to support proactive maintenance scheduling. Purpose: to provide an integrated failure-energy-forecast framework applicable to diesel-powered mining dewatering systems where pump failures translate directly into generator fuel waste. Methodology: operational logs (January 2024 – May 2025; n = 68 failure events across 17 months) were analysed for failure mode, horsepower rating, and location. An energy waste formula derived from three-phase AC power principles — E_waste = √3 × ΔI × V × t × PF / 1000 — was validated against site meter readings from five independent events (mean deviation: 2.0%, max: 2.5%). An ARIMA(2,1,2) model, selected by AIC minimisation and validated by Ljung-Box diagnostic, was used to forecast 2026 failures. Findings: motor burnout dominated at 70% of failures; 7.5 HP pumps recorded 47.1% of events (χ²(3) = 18.12, p < 0.001); Collection Tank 2 generated 32.4% of failures and an estimated 34.4% of cumulative energy waste (178.6 kWh total). No statistically significant seasonal effect was identified (Mann-Whitney U, p = 0.127). The ARIMA model projects 32 failures in 2026 (95% CI: 18–52). Implications: a tiered preventive maintenance framework is recommended, with Priority 1–2 interventions estimated to reduce energy waste by 35–45% and generate monthly savings of NGN 8,000–11,000.