Principal components regression control for multivariate autocorrelated cascade processes

Abstract

A comprehensive methodology for multivariate autocorrelated cascade process control using multiple regressions in conjunction with principal component analysis (PCA) is presented in this paper. PCA is used to alleviate the multicollinearity problem among input variables and reduce the dimension of the variables. An integrated principal components selection rule is proposed to reduce the number of variables. A multiple regression model with estimated parameters based on a selected number of principal components is developed. The autoregressive error correction model is used to address the issue of dependent residuals from the multiple regression model. Further, interpretation of out-of-control signals based on the principal components in process monitoring is discussed. The implementation of the proposed methodology is demonstrated considering a real life example of the sugar-beet pulp drying process. A general guideline for controlling multivariate autocorrelated processes also is developed.