A new scheme of fault detection based on neural nets

This paper presents a new scheme for fault detection based on neural nets. By using a Hopfield neural network for online parameter estimation, process faults caused by parameter changes can be detected. Because of the use of a neural computing technique, the scheme improves the performance of fault diagnosis approaches based on traditional parameter estimation methods with respect to detection speed, accuracy and the ability of realtime processing. Simulation results show that the scheme has good efficacy in fault detection and is especially suitable for detecting faults caused by parameter changes