Neural network Boolean prediction of melt fracture

The extrusion of polymer melts through die openings can result in flow instabilities or melt fracture. Melt fracture is a roughness or distortion encountered at high extrusion rates with all polymer melts. The flow instability affects the ease of processing of the material and the roughness subsequently affects the quality of the final part. Melt fracture is generally considered a Boolean phenomena, in that it is present above a critical stress and is not present below the critical stress. The prediction of melt fracture from knowledge of simple material characteristics and basic operating conditions would be a very useful tool in helping to minimize the melt fracture. In this work, the neural network methodology is employed, as an engineering tool, for the prediction of melt fracture from basic material characteristics and operating conditions. The sigmoid threshold function inherent in the methodology, allows for Boolean ouput. The experimental data are obtained on a commercial extrusion blow moulding machine. The input parameters of the network are the die gap, the melt temperature, the material zero shear viscosity and the material power law index