Fault Diagnosis Of TE Process Based On Information Fusion Theory

With the development of science and technology,as well as the diversified demand for chemical products in our life.The scale of production in modern chemical process is increasing day by day,also the process of production is more complicated and automatic.The automation is the embodiment of artificial intelligence,but it also increases the coupling of the system,so as to increase the risk of failure.In view of the characteristics of modern chemical process,this paper proposes a new method of information fusion fault diagnosis based on support vector machines and neural network.The preliminary diagnosis results by SVM and BP neural network are used as the evidence,and then the D-S evidence theory is used to fuse the two evidence.In this paper,the Tennessee Eastman process is the research platform,the data in the TE process as the experimental simulation data.This paper mainly completed the following work:(1)A fault diagnosis method based on Principal Component Analysis and SVM were studied.Using PCA to extract the fault feature of TE process data,and then use optimized SVM to diagnose the TE process.Simulation results show that this method has a higher failure rate than single SVM method,the diagnosis results of this method as an evidence of D-S evidence theory.(2)A fault diagnosis method based on PCA and BP neural network were studied.Using PC A to extract the fault feature of TE process data,and then use BP neural network to diagnose the TE process.Simulation results show that this method has a higher failure rate than single SVM method,the diagnosis results of this method as an evidence of D-S evidence theory.(3)In view of the requirement of the D-S evidence theory,the probabilistic outputs of SVM and BP network are constructed respectively.Finally,the two groups of evidence are synthesized by using Dempster's rule,and the diagnostic results based on Information fusion are obtained.Simulation results show that the method of information fusion is better than PCA and SVM in fault recognition rate,and it is also better than PCA and BP network.The method of information fusion improves the fault recognition rate of TE process.