Target value computing in paper machine grade changes by fuzzy system

Abstract

A fuzzy system is proposed to compute new target values for low level controllers during grade changes in a paper machine. The fuzzy system is designed and tuned in co-operation with human operators of the paper machine. The design method utilizes both heuristic knowledge and estimation techniques based on input-output data. The input fuzzy sets and the rule base are initially designed verbally and thereafter fine-tuned by a least squares algorithm. An industrial example illustrates the iterative design process. It shows that transparency and a priori knowledge included in the model are preserved also during optimization of the fuzzy model.