Estimation of the average chain length of polymers with neural classifiers

This paper presents a method for estimating the average chain length of polymers based on neural networks. A simulation of a continuous solution polymerisation reactor, with varying setpoints, is used to train a self organising, winner-take-all, neural net which classifies the different average chain lengths, based on different input patterns. By means of a network parameter, the vigilance parameter, the accuracy-coarse or fine-of the classification intervals can be controlled. This parameter influences also the generalising capacity of the network. Pretrained networks are used to classify sets of untrained input patterns