This paper presents the methods that could achieve the fault diagnosis and predetermination of crucial circuit in the nuclear measurement system.The analog electric current amplifying circuit was selected as the object of our study and it was simulated by PSPICE.The fault characteristics of the analog circuit were extracted from the output shock response through wavelet packet transform method.These characteristics were used as the input information of BP neural network for fault type identification.At the same time,fault prediction research is carried out based on the relevance vector machine(RVM).The calculation results show that,the efficiency of fault diagnosis was 99％for different fault types,and the RVM model optimized with quantum-behaved particle swarm optimization(QPSO)can accurately predict the development trend of circuit faults.This study provides more substantial theoretical support for the maintenance and repair of crucial circuit in nuclear measurement system.
维普
万方数据