Prediction of a Failure Envelope of Short Fiber Reinforced Polymer Composites (SFRPs) Using Artificial Neural Network

Abstract

Short fiber reinforced polymer composites (SFRPs) have become a material of interest due to their high stiffness-to-density ratio, durability, and cost-effective manufacturing of intricate parts. The modeling of an SFRP at the microscale involves complex homogenization techniques and sophisticated algorithmic implementations. The failure envelope of an SFRP gives a decision boundary for safety and failure under multiaxial loading. Additionally, the expansion and translation of the failure envelope describe the plastic behavior under monotonic and cyclic loading. Unlike isotropic materials, the stress tensor cannot be split into hydrostatic and deviatoric parts. It limits a formal mathematical theory for failure criteria for SFRPs. In the present work, an artificial neural network (ANN) is used to predict the failure envelope for a polyamide-glass SFRP specimen under biaxial loading. The data sets for training and validation are taken from the First Pseudo grain Failure (FPGF) model implemented using commercial finite element software. A noise in terms of randomly generated points near the failure curve is used as the test data. The stress or the strain values corresponding to the ANN predictions falling within a defined threshold range are plotted to obtain the failure envelope.