Comparative Study Of Online Fault Diagnosis Methods For Multimodal Processes

With the rapid development of science and technology,industrial production has become more and more complicated.The requirements for the reliability,safety and accuracy of complex industrial processes are also increasing.Therefore,Scholars have focused on the research of industrial process fault diagnosis technology in recent years.At present,the production process of complex industry is mostly the Multimode industrial production process.This kind of process has many stable working states because of its own characteristics,environmental factor,changeable production plan and so on.Compared with the single modal process,its craft is complicated and its data usually shows the characteristic of multivariable,nonlinear,dynamic,time varying and so on.If this kind of industrial process goes wrong,not only will it lead to a decline in product quality,but it will also cause personal safety and property loss.Therefore,the Multimodal process fault diagnosis technology is very important for industrial safety.The research content of this thesis is mainly based on the global modeling method.It is a study of ensemble online fault diagnosis method.This thesis divides the method into three parts.They are the online data preprocessing part,the online fault detection part and the online fault classification part.This thesis puts forward an online fault diagnosis method that based on VMD-LLE-ICAMM-NBC for the multimodel process.In the online data preprocessing part,it uses the method of variational mode decomposition to remove the noise and make a detailed comparison with the empirical mode decomposition algorithm.In the online fault detection,this thesis proposes an improved algorithm based on the independent component analysis mixtue model for online fault detection.Using the locally linear embedding reduces the dimensionality of data and combines with the independent component analysis to extract the data's features.Next,utilizing the Bayesian inference structures global model.The online fault classification part,using the method of the improved Naive Bayes classifier classifies the fault.Finally,using the experiment verifies the effectiveness of the method.