Diagnosing Faults at Different Process Operating Points Using Neural Networks

Abstract

Abstract Industrial processes often operate at different operating points, mainly to manufacture at different customer specifications. The demonstrated applications of neural networks for fault diagnosis usually consider only a primary operating point. This paper investigates their application for the diagnosis of non-catastrophic faults at different operating points on an industrial nuclear processing plant. Data conditioning methods, including statistical scaling and principal component analysis, are investigated to facilitate fault classification and to reduce the complexity of neural networks. Results are presented to illustrate the performance of trained neural networks for classifying process faults using simulated and real industrial data.