A NEURAL PREDICTIVE MODEL FOR CHARACTERIZING IMPACT CUSHIONING CURVES

This paper describes a predictive characterisation model for impact cu shioning curves. The model involves establishing apr appropriate set o f random discrete experimental points, network training and curve char acterisation. Preliminary tests have found that the approach is able t o predict quite accurately the impact cushioning curve characteristics to within +/-2% error. The results achieved indicate that this approa ch can substantially reduce the number of experimental points required when characterising new impact cushioning materials. The algorithms u sed to obtain a set of randomly distributed training data and generate the requisite points for curve characterisation are also discussed an d found to be suitable for this purpose. (C) 1998 Published by Elsevie r Science Ltd. All rights reserved.