Research On Statistical Process Monitoring Method Based On Distributed Error Generation Strategy

With the continuous deepening of the industrialization process,stricter requirements have been imposed on the safe production and product quality of enterprises.Therefore,the process monitoring method has very important research value in actual industrial production and academic research.At the same time,computer technology and sensor equipment are widely used in industrial systems,and various types of process data collected are completely saved.This has enabled rapid development of multivariate statistical process monitoring methods.However,the traditional multivariate statistical process monitoring(MSPM)is no longer sufficient for modern industrial processes with complex structures and numerous operating units,such as dynamic principal component analysis(DPCA),canonical correlation analysis(CCA),and so on.Therefore,the research work of this paper is based on the existing algorithms.In view of the above problems,several improved and optimized process monitoring methods are proposed.The specific research work is as follows:(1)It is unreasonable for DPCA to directly monitor principal components and residuals with autocorrelation characteristics.Based on the DPCA model,a distributed dynamic process monitoring method based on estimation error(DPCA-IM)is proposed in this paper.Based on the idea of error generation of mechanism model,this method use the iterative method(IM)to calculate the estimated value of each process variable to fit the measured value,so as to eliminate the autocorrelation between process variables.Then the original process data is converted into estimation error.Finally,fault detection is implemented by using estimated error.Through numerical and TE simulation experiments,the superiority of DPCA-IM method in fault detection is verified.(2)The traditional process monitoring methods are to extract the potential characteristic information,so it is impossible to reasonably explain the operating mechanism of the system,thus affecting the problem of fault detection and diagnosis.Based on CCA,this paper proposes a distributed process monitoring method based on CCA explicit relationship discovery.Firstly,CCA algorithm is used to describe the correlation between each process variable and other variables.Then,the explicit relationship between process variables is explained by Regression Model.Then,the error generated by the regression model is used for fault detection,and fault diagnosis is carried out after fault detection.Finally,through numerical and TE simulation experiments,the superiority of CCA-ERD method over other process monitoring methods based on feature information in fault detection and diagnosis is verified.Finally,based on the summary of this paper,the future work in the field of multivariate statistical process monitoring is prospected.