Research Of Fault Diagnosis Method Based On Multivariate Statistical In The Chemical Process

The process monitoring is one of the most important problems in the process industry. The monitoring system can detect the faults and other abnormal events promptly through monitoring the state of the process. The variable which causes the fault also could be detected and pitched. The production process will be safe and the quality of the product will be improved. At present, the batch proeesses play an important role in the production of low-volume, high-value products such as pharmaceuticals, colornats, flavors and biochemical products. Because batch statistical performance monitoring and control only relies on process data and thus general-puprosed, it has become one of the most active research areas in proeess control.This thesis focuses on multivariate statistical process monitoring of the batch process monitoring. On the basis of studying the present situation and developing trends of the statistical process monitoring in the domestic and international field, takes the principal element analysis (Principle Component Analysis, PCA) as the main line, then based on Tennessee Eastman process, a model of fault detection and diagnosis based on statistical analysis is built, and an anticipated result is obtained through analysis and simulation according to the diverse situation.Main research contents and results are as follows:1.Summarizes the research status and development trend in fault diagnosis field. And provides complete introductions based on statistical analysis theory particularly.2. The basic theory of principal component analysis and calculation steps are introducted, and the two statistics of statistical process control SPE (Squares predication Error) and HotellingT2are described in detail.3.Multiway PCA(MPCA)—the method of batch process monitoring and fault diagnosis—is discussed systematieally. The method to deal with three-way batch process data by MPCA is introduced. The MPCA method is applied to the TE process and the results of fault diagnosis is analyzed.4.1n view of the deficiencies of the MPCA, and develops another batch process monitoring and fault diagnosis method-BDPCA which base on the DPCA method. Application the BDPCA into the TE process and batch proeess monitoring and its fault diagnosis results are obtained. Compared with the MPCA, the simulation results of the TE chemical process simulation based on BDPCA monitoring method show that the BDPCA algorithm is feasible and effective.It can accurately monitor the time of the failure when it is happened, effectively reduce the error alarm, and exactly monitor the operation of the system.5.In order to overcome the shortcomings of BDPCA, a forgetting factor is introduced. Therefore, the performance of fault monitoring and diagnosis is improved. This method is successfully applied to the TE process research. The simulation results demonstrate the effectiveness of proposed method.Finally,a summary of the thesis is given and some conclusions are made through the results of the research. Further research interests in the future are prospected.