Optimized independent components for parameter regression

In this paper, a modified ICR algorithm is proposed for quality prediction purpose. The disadvantage of original Independent Component Regression (ICR) is that the extracted Independent Components (ICs) are not informative for quality prediction and interpretation. In the proposed method, to enhance the causal relationship between the extracted ICs and quality variables, a dual-objective optimization which combines the cost function wTXTYv in Partial Least Squares (PLS) and the approximations of negentropy in Independent Component Analysis (ICA) is constructed in the first step for feature extraction. It simultaneously considers both the quality-correlation and the independence, and then the ICR-MLR (Multiple Linear Regression) method is used to obtain the regression coefficients. The proposed method is applied to the quality prediction in continuous annealing process and Tennessee Eastman process. Applications indicate that the proposed approach effectively captures the relations in the process variables and use of proposed method instead of original PLS and ICR improves the regression matching and prediction ability.