Neural networks for fault diagnosis of a nuclear fuel processing plant at different operating points

Abstract

Industrial processes often produce at various operating points; however, demonstrated applications of neural networks for fault diagnosis usually consider only a single (primary) operating point. Developing a standard neural-network scheme for fault diagnosis at all operating points may be impractical due to the unavailability of suitable training data for less frequently used (secondary) operating points. This paper investigates the application of a single neural-network for the diagnosis of non-catastrophic faults in an industrial nuclear processing plant operating at different points. Data-conditioning methods are investigated to facilitate fault classification, and to reduce the complexity of the neural networks. Results illustrate the performance of trained neural networks for classifying process faults using simulated and real industrial data.