Fault Diagnosis Of The Nuclear Power Plant Valve Based On Neural Network Technology

The electro-hydraulic servo jvalve and safety valve are very important in nuclear powerplants,their works directly affect the operation condition of the entire system,and the causeof the fault reasons also appears in nonlinear,uncertainty. etc. It is extremely necessary todiagnose or predict the faults,and study fault diagnosis methods for both of the valves.This paper first analyzes applications of the electro-hydraulic servo valve in the nuclearpower system,basic structure and fault mechanism,the static characteristic curve,and a briefintroduction to several existing artificial intelligence fault diagnosis methods,such as expertsystems,gray theory,fuzzy theory,artificial neural networks. By the comparison on theelectro-hydraulic servo valve and the valve itself features,i decided to use neural network forfault diagnosis of these two kinds of valves after analyzing their basic principles.characteristic curve of the pressure data is collected by using bench. We take theelectro-hydraulic servo valve outlet pressure and return oil port as the network input data, andthe artificial state of the five failures as the network’s output,then normalize the data for thefault diagnosis.Using the Matlab neural network toolbox,then compare with BP neural network,RBFneural network and Elman neural network by means of electro-hydraulic servo valve’sdiagnosis results,and finally I decide to use RBF neural network as the nuclear power plantvalve fault diagnosis method.This paper also analysis the structure and fault mechanism of safety valve,analysis inputlayer and output layer by using the relevant data of measuring points to prepare for thefollowing troubleshooting...