Artificial neural network technique to predict dynamic fracture of particulate composite

Abstract

In this paper, the artificial neural network technique using a multi-layer perceptron feed forward scheme was used to model and predict the mode-I fracture behaviour of particulate polymer composites when subjected to impact loading. A neural network consisting of three-layers was employed to develop the network. Artificial neural network was constructed using six input parameters such as shear wave speed ( CS), density ( D), elastic modulus ( Ed), longitudinal wave speed ( CL), volume fraction ( Vf) and time ( t). The influence of input parameters on the output stress intensity factor and crack-initiation fracture toughness were found to be in the order of t > CS > D > Ed > CL > Vf. The degree of accuracy of prediction was 92.7% for stress intensity factor. In this regard, artificial neural network can be used in the modelling and prediction of fracture behaviour of particulate polymer composites under impact loading.