Development of a pumping system decision support tool based on artificial intelligence

Abstract

A framework for the development of a pumping system decision support tool based on artificial intelligence techniques has been investigated. Pump fault detection and diagnosis are key requirements of the decision support tool. Artificial Neural Networks (ANNs) were proposed for condition monitoring data interpretation utilising quantitative performance data. In the analysis, the Cumulative Sum (Cusum) charting procedure was successful in incipient fault identification. Various preprocessing techniques were investigated to obtain maximum diagnostic information despite the inherent problems of real industrial data. The orthonormal technique highlighted good generalisation ability in fast machine learning time. ANNs were successful for accurate, incipient diagnosis of pumping machinery fault conditions based on real industrial data corresponding to historical pump faults.