Abstract:Aiming at the problems of low diagnostic accuracy and high misjudgment of the traditional single gas regulator fault diagnosis model,a fault diagnosis method based on empirical wavelet transform and improved D-S evidence theory was proposed to diagnose the fault state of gas pressure regulator.EWT was used to preprocess the data collected by the sensor and the energy entropy of each component was calculated,which was used as the input vector of the mixed diagnostic model based on three models,namely generalized regression neural network,elman neural network and grey relational entropy analysis.Then,the basic probability assignment of three models was established according to D-S evidence theory to realize the decision fusion of fault information,and the evidence correlation coefficient method was introduced to correct the weight of the decision importance and the conflict problems of evidence bodies.Experimental results show that the diagnostic accuracy of EWT and the improved D-S evidence theory model reaches 95.0%,and it is better than the single model GRNN,Elman and GREA in terms of mean error,mean square error and maximum error percentage.