NEURAL NETWORKS FOR STEADY-STATE MODELING OF AN EXTRUDER

The inherent complexities of the extrusion process have made the devel opment of both mechanistic and parametric models problematic. This con tribution addresses the issues involved in developing a realistic mode l of an industrial reactive plasticating extruder to enable prediction of extrudate viscosity, which provides a good measure of product qual ity for the process. The complex nonlinearities associated with the pr ocess input-output mapping suggest that neural networks could be an ap propriate modelling paradigm. However, the large number of parameters that had to be used caused problems during model identification, since only a limited data set was available. Resampling techniques were the refore used for model identification and validation, due to their effi cient use of data and their ability to provide realistic inference of the true error rate associated with the identified models. The statist ics obtained are utilised for network structure selection, outlier det ection and the derivation of a distribution for model prediction error s. A final network model is presented with fixed confidence bounds, th e weights of this network are analysed and an input-output mapping of the process is generated. (C) 1997 Elsevier Science Limited.