Degradation Assessment and Remaining Useful Life Prognostic of Centrifugal Pump Using Multi-Sensor Process Monitoring

Taotao Zhou,Yan-Fu Li

DOI: https://doi.org/10.1109/aparm49247.2020.9209577

2020-01-01

Abstract:Centrifugal pump is the most widely used industrial rotating machinery and typically serve under harsh operational condition. The safe operation of these pumps is important and lots of research efforts have been devoted to avoid the pump failures. Recently, data-driven approaches arise to be a promising option to discover the underlying failure patterns, which would complement the limitations of conventional physical-based approaches susceptible to uncertain filed conditions. However, most recent studies are conducted based on vibration data. This in turn assumes the pump is equipped with prerequired vibration monitoring capability that would be usually expensive and even not feasible. This paper proposes a degradation assessment approach that exploits the multi-sensor process monitoring already exists in the pump system, for instance, flow rate, pressure, electric current, electric voltage. Specifically, the pump conditions are identified via a Bayesian nonparametric clustering algorithm. Then a distance-based degradation index is designed to characterize the degradation evolution. Finally, a state-space based predictive model is developed to achieve the remaining useful life prognostic. The proposed approach is validated by the process monitoring data collected from the centrifugal pumps tested in the laboratory. The results demonstrate the effectiveness of the proposed approach to characterize the pump degradation and predict its remaining useful life.