Study On Fault Diagnosis Method Of Chemical Process Based On Adaptive PPA And Bayesian Network

With the development of information intelligence technology,the automation of chemical process becomes higher.In order to ensure the safe and stable operation of the system,it is very important to diagnose the process fault timely,accurately and efficiently.In the process fault diagnosis,the system process mechanism model and expert knowledge are difficult to obtain,while the process data is easy to obtain.Therefore,research on the data-driven fault diagnosis method is of great significance.As chemical process data is dynamic and time-varying,chemical process variables are complex and it is difficult to analyze the root cause of the fault,the adaptive principal polynomial analysis(PPA)method is proposed in fault detection and Bayesian network method is introduced in the fault root cause analysis to study the fault diagnosis of dynamic chemical process.The main work and innovations of this paper are as follows:According to the dynamic time-varying of chemical process,a new adaptive PPA fault diagnosis method is proposed in this thesis,by introducing the sliding window mechanism based on the main polynomial chemical fault diagnosis method.In this proposed method,the detection model is used to detect the dynamic process in real time firstly.If the process is normal,the window data is updated according to the latest process data to update the model.Then the process is detected in real time using the latest model.Finally,the method is applied to the chemical simulation TE process for fault detection,and the results show that the proposed method gives better results.Aiming at the problem that it is difficult to analyze the cause of a fault when a fault is detected in the chemical process,Bayesian network is used in fault root cause analysis for chemical process in this thesis as the Bayesian network is a combination of graph theory and probability theory and it has a strong reasoning ability in dealing with complexity,fuzziness,and uncertainty.As the basis of Bayesian network,the complexity of Bayesian structure modeling will increase exponentially with the increase of network nodes,which will affect the accuracy of the final result.Therefore,an improved artificial bee colony Bayesian network method(D-ABC)is proposed in this thesis,in which artificial bee colony algorithm and differential evolution algorithm are combined and the optimal structure is searched by scoring.Its effectiveness is verified through the standard network.Finally,based on the TE process,through the Bayesian network structure modeling and parameter learning,the Bayesian network connection tree algorithm is used to analyze and reason the fault root cause,which proves the effectiveness of this method.