Artificial Neural Network-Based Fault Detection System with Residual Analysis Approach on Centrifugal Pump: A Case Study

Centrifugal pump is an instrument that widely used in industry and has become the main driving component. A detection system often needed to prevent damage these pumps because they can interfere with overall performance. Therefore, this study discussed development of a fault for two centrifugal units, namely Medium Pressure Oil Pump (MPOP) Water Injection (WIP). In detecting operating conditions pump, it was residual feature extraction technique time domain statistical approach. Residual generated by using three sub-systems pumping system. Each sub-system modeled artificial neural network feedforward-back propagation architecture. Based on values, classifier designed classify conditions. Then proposed applied condition monitoring scheme. The test results (using data from field) show accuracy 91.67% MPOP 94.8% WIP cases. Meanwhile, above 99% during online simulations.