Research On Multiple Faults Condition Monitoring And Diagnosis Method Based On Principal Components Analysis

With the advance of the level of industrial automation, mechanical equipment is having the tendency of large scale, high-speed, serialization, automation and intelligent. When these technological developments reduce the production cost and increase the efficiency, it also make a higher requirement on operation and maintenance of the system. So we need to monitor the condition and make fault diagnosis of industrial process to ensure the operation status satisfy the performance index. Therefore, it has very important significance to modern industry with continuous study on fault diagnosis and making the theory of fault diagnosis a wide application and spread in complex industry.Although process monitoring and fault diagnosis has been studied a lot in some references, most of them studied on the situation of single fault. In fact, complex industrial process is often composed of several subsystems. It becomes more and more common when multiple faults happen simultaneously. In the case of multiple faults happen at the same time, there will be some problems if we still use the traditional method of single fault diagnosis. They will lead to the situation of low accuracy?wrong result and even no result when make fault diagnosis on system. So it has a very important academic value and practical significance to study on the method of multiple faults monitoring and fault diagnosis for complex industrial process. Principal component analysis has always been the research focus because it can only use the historical data to finish the task of condition monitoring. But it has some limitation, such as it only performs well in liner system and it can't make fault identification and separation.In response to these problems, this paper makes improvement to principal component analysis and achieves the multiple faults monitoring and diagnosis with the method improved. The main work in this paper includes:(1) Elaborate the source of fault diagnosis and make a classification on common diagnosis methods; Introduce the background, significance and status of the research;(2) Make a further study on the method of single fault condition monitoring and diagnosis based on principal component analysis. Analyze the method principle and introduce the steps of how to achieve condition monitoring and fault diagnosis based on principal component analysis.(3) According to the problem of nonlinear relationship existing between the variable in actual industry systems, use KPCA to achieve multiple faults monitoring and make simulation verification with tennessee-eastman process.(4) After combining the method of contribution plots and logic analysis, this paper puts forward a method based on KPCA contribution plots for multiple faults diagnosis and then verify this method through simulation experiments.