Neural network based approach for optimisation applied to an industrial nylon-6,6 polymerisation process

The basic idea of the proposed optimisation method is to replace the model equations by an equivalent neural network (NN) that mimics the phenomenological model, and use-this NN to carry out a grid search, mapping all the region of interest. The proposed optimisation approach was applied to the industrial process of nylon-6,6 polymerisation in a twin-screw extruder reactor. This corresponds to the finishing stage of an industrial polymerisation plant. A qualitative optimisation procedure is used taking in account safe operation conditions, wear and tear of the equipment, product quality and energy consumption. The chosen operational variables are then checked with the phenomenological model. This approach provides more comprehensive information for the engineer's analysis than the conventional non-linear programming procedure. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.