Fault diagnosis of engineering systems using neural networks: a practical approach
William Crowther · 1996
A methodology is proposed for diagnosing faults in engineering systems using neural networks (NN). Using the methodology, faults were diagnosed in a simulated potential divider circuit and a simulated spring-mass-damper system. NN training for diagnosis of single faults is relatively fast, and accurate results are achieved. For two simultaneous faults, NN training takes approximately 10 times as long as training for the single fault case. For full-scale problems, fast computing hardware is required to bring network training times within useful time scales.