IMPROVED BATCH PROCESS PERFORMANCE BY EVOLUTIONARY MODELING

A key issue for improving the industrial efficiency in energy and mate rial resources usage is to integrate the dynamics of the process and i ts scenario in the actual plant operation decision making. In this wor k, a combination of statistical techniques has been used to build an a utomatic modelling tool based on Neural Networks, that overruns the li mitations of modelling techniques based on the theoretical knowledge o f the process principles. As the overall system can be run simultaneou sly with the process, the tool can be used to continuously readjust it s parameters and to follow evolutionary processes. The resulting model can be applied for process forecasting and control, resulting in impr ovements in the process performance. In the Neural Network field, a ne w method is introduced to test the results and a heuristic is proposed to stop the learning process when the best model has been found. (C) 1998 Elsevier Science B.V. All rights reserved.