Application of artificial neural networks in predicting early failures in industrial pumps

Maintenance optimization is a critical aspect of industrial operations, ensuring reliability, efficiency and cost effectiveness. Traditional maintenance strategies, such as corrective maintenance, often lead to excessive downtime or avoidable resource consumption. This research work focused on the maintenance optimization of centrifugal pumps used in Nigerian refineries. In this study, The Griswold pump (101-PM-2A) was used as a case study and its operational data as well as field performance readings were obtained from Warri Refining and Petrochemical Company (WRPC).The maintenance optimization model employed in this research is Artificial Neural Networks (ANN) to predict failures and suggest maintenance routine for the pump while in operation.  The result shows that the vibration values exceeding 29Hz may damage the pump system. It is highly recommended that the operation of the pump has to be halted should a vibration reading close to 29Hz be registered on the vibrometer. The findings demonstrate that ANN-based maintenance strategy outperforms traditional approaches by drastically minimizing failures, optimizing resource allocation and extending equipment lifespan.