Research On Fault Detection And Diagnosis Of Complex Chemical Process

Data driven model is an effective means to solve the problem of fault detection and diagnosis,especially for complex industrial chemical process control and detection problems;among them,the method of multivariate statistical analysis has become a data driven model in the field of complex fault detection is widely applied to a class.The main contents of this research is to study the complex multi fault detection and diagnosis method of variable chemical field,and by Tennessee Eastman(TE)simulation platform of the proposed method is validated by simulation analysis.Therefore,the data used in this paper are generated and collected by the TE platform,and through the following two methods to verify the effectiveness of the exposition.In view of the complex industrial chemical processes usually have time dynamic characteristics and nonlinear process,and the data are shown as the self correlation and crosscorrelation analysis,put forward a kind of dynamic independent component analysis and support vector data description(DICA-SVDD)process fault detection model.First of all,the dynamic augmented matrix structure containing time delay variable l,using independent component analysis(ICA)algorithm is applied to dynamic augmented matrix and dynamic independent component process data;then,the dynamic independent component is applied to the support vector data description(SVDD)algorithm,we construct the statistical limit and statistics.The realization of online fault detection;finally,Tennessee Eastman(TE)numerical simulation analysis of experimental platform,verify the validity of the proposed model.And with the original based on principal component analysis(PCA)comparative analysis of the test results by the model of fault detection,the comparative results show that the proposed model is based on the detection performance of the PCA algorithm model of fault detection is better,validate the proposed detection model.The process of industrial system often has the characteristics of many process variables.When the fault occurs,it is the first problem that how to locate the main abnormal variable in the first time.In order to solve this problem,an integrated model of process fault detection and diagnosis based on independent component analysis and principal component analysis(ICA-PCA)as well as least absolute shrinkage and selection operator(Lasso)regression algorithm is proposed.First of all,the process of data extraction of Gauss signal and non Gauss signal by the method of ICA-PCA,I~2,T~2 and SPE to construct statistics and determine the statistics of its corresponding time limit,detect the fault detection process,the implementation process;then,the fault information obtained by fault detection,and establish fault diagnosis model of Lasso regression algorithm based on the main abnormal variable positioning and the choice of the fault,fault diagnosis.Finally,the simulation analysis of TE process data is carried out,and the results are compared with those of the distributed PCA contribution graph method.The results show that the proposed method is effective and feasible.