Predictive Fuzzy Control of Paper Quality

Pulp and paper quality depends on the quality of wood chips which depends on their physical and optical properties. Presently, there is no formally established knowledge concerning the co-influences of the several parameters governing the thermo-mechanical pulp and paper process (TMP). The main goal of this paper is to automatically generate fuzzy knowledge bases (FKBs) to characterize wood chip properties online and apply this information to optimize the TMP process so that pulp quality can be predicted and controlled using wood chip properties (defined by numerical data). The production settings used in this article take into account the hydrosulfite bleaching agent. Learning of the FKBs (using a genetic algorithm) uses measurements obtained from the chip management system (CMSreg). Changes in chip quality are measured by physical information (color analysis and humidity) using CMSreg. The information provided by CMSreg enabled us to predict the ISO brightness of the produced pulp according to a certain charge of hydrosulfite. The developed FKBs are used afterwards to control the optimal hydrosulfite charges using a Monte-Carlo based search algorithm