An industrial process optimization approach based on input and output statistical data analysis

Abstract

A major concern in process industry is to improve the quality of final products. This is highly dependent on the raw materials composition and on the industrial processes settings. The main issue in operating these systems is to identify a correlation between the process settings and the quality of the final product. The aim of this work is to first build a model of the industrial process. Then we search for a set of parameters in order to minimize an objective function based on the quality of the final product. Since the number of parameters of these processes may be important (several hundred in some instances), we perform a Support Vector machines Regression (SVR) method as multiple regression to model the manufacturing process, based on the input (various settings) and output (product quality) data. The settings optimization using the regression function is done by a heuristic. It is based on an iterative descent method applied iteratively on each parameter. The proposed approach is used on a fluidized bed combustion boiler in the context of paper industry. The experiment confirms the efficiency of the approach.