Complex process monitoring using modified partial least squares method of independent component regression

In this paper, first, some disadvantages of original partial least squares method of independent component analysis (ICA-PLS) are analyzed. Then ICA-PLS is modified for regression purpose. Disadvantages of the original ICA-PLS algorithm are as follows: 1) the regression coefficient matrix and residual matrix cannot been given so that the computation time may increase with the number of samples: and 2) ICA-PLS lacks the ability to give better monitoring performance when the correlation structure of measured variables is nonlinear, which is often the case for industrial processes. To solve the above problems. we modified the original algorithm in following aspects: 1) the regression coefficient matrix and residual matrix in ICA-PLS are given so that the computation time is decreased: and 2) to solve the nonlinear problem, ICA-PLS and kernel trick is first combined for nonlinear regression purpose, which is called iterative ICA-KPLS in this paper. The iterative calculation of ICA-KPLS will be time consuming when the sample number becomes larger. Hence, the regression coefficient matrix and residual matrix in ICA-KPLS are given to avoid the expensive computation time when the number of samples is huge. The proposed methods are applied to the quality prediction in fermentation process and Tennessee Eastman process. Applications indicate that the proposed approach effectively captures the relations in the process variables and use of ICA-KPLS instead of ICA-PLS improves the predictive ability. The expensive computation time is avoided by using the coefficient matrix and residual matrix. (c) 2009 Elsevier B.V. All rights reserved.