| In the research of multivariate statistical process control(MSPC),more and more scholars have begun to pay attention to the research of matrix data.Compared with vector data,matrix data has a richer form of expression in real life,so it is meaningful to study matrix data.On the other hand,the massive data generated is often a data stream,and this data stream is often generated When the source is mutated,it will also change,so online monitoring of the data flow in the actual production process is necessary.This article mainly studies the online monitoring of matrix value time series.In the first part,the application method of combining traditional principal component analysis(PCA)with multivariate statistical process control theory is studied.Firstly,it discusses the combination of the matrix-valued time series data and the MSPC-PCA method to give the statistical process Control monitoring statistics and their distribution and control lines.Secondly,Monte Carlo data simulation experiments were carried out on this method,which verified the correctness of the theory and the feasibility of practical application,and the results showed that it was excellent.In the second part,the application method of combining two-dimensional principal component analysis(2DPCA)with multivariate statistical process control theory is studied.First,based on the 2DPCA method,the matrix data is orthogonally projected to obtain features,and the monitoring statistics are constructed by fusing these features;Secondly,it is proved that the limit distribution of the monitoring statistics is the chi-square distribution,and the distribution is used for statistical inference.Simulation experiments show that the method is correct in theory;when the sample size is large,the method performs better than similar methods.Finally,the full text is summarized,the main research results of this article are summarized,and the problems that can be studied in the future are further raised. |
PDF Full Text Request