Based On The Research Of The Fault Diagnosis Of Multivariate Statistical Analysis

With the development of science and technology, there have been great changes in modern industry, including increasing production scale, increasingly complicated equipment structure, more stringent requirements for automatic control, and increasing cost of investment. There would be unimaginable loss unless the system and equipment fault could be diagnosed in time. Therefore, research on process monitoring system with functions of monitoring, control and fault diagnosis has been an urgent need in the development of modern industry, with both great theoretical significance and application value.In the field of fault diagnosis, the method based on the analysis of multivariate statistics is of great value. Taking it as a theoretical basis, the thesis researches into methods of fault diagnosis based on principal component analysis. By applying these methods into the fault diagnosis of Tennessee-Eastman Process, the thesis has achieved good fault detection results. The thesis mainly consists of the following contents.The thesis elaborates on the research methods in the field of fault diagnosis, and especially gives a summary of the development and current situation of the methods of fault diagnosis based on the analysis of multivariate statistics.The thesis introduces the statistics and statistical control chart frequently-used in the methods of fault diagnosis based on multivariate statistics, analyzes the fundamental ideology of fault detection and diagnosis of multivariable system based on principal component analysis, and points out the shortcomings of principal component analysis in the fault diagnosis of nonlinear systems.In order to diagnose the fault in nonlinear systems, the thesis introduces the concept of Kernel function and the method of Kernel principal component analysis. Considering the shortcomings of the method of Kernel principal component analysis, the thesis improves the method and research the method of Kernel principal component analysis based on feature vector selection.The thesis explores a process monitoring system based on Tennessee-Eastman process and has won good effects of monitoring. The research results of algorithm have been successfully applied in online fault detection module of monitoring system, which is a proof of the effectiveness and practicality of algorithm.