Neural modelling of polypropylene fibre processing: Predicting the structure and properties and identifying the control parameters for specified fibres

This paper describes the application of artificial intelligence to data der ived from polypropylene drawing carried out at Galashiels using designed ex periments. The topology of the data is visualised in two dimensions with re spect to specific properties to be modelled, as a quality check on the proc ess data. A series of neural network models are used successfully to predic t the tenacity, elongation, modulus and heat shrinkage and also the crystal lographic order and polymer chains orientation of the output fibres from th e draw parameters values. A software harness is constructed for using the n eural predictors to find the draw parameters which come closest to achievin g any specified combination of fibre properties. (C) 2001 Kluwer Academic P ublishers.