Artificial intelligence approach for increasing the fidelity of the second order fibre orientation tensor for use in finite element analysis

Abstract

A more efficient integrated computational approach for the design of injection and compression moulded composites which maps the results of manufacturing simulations as input to structural simulations is proposed. This methodology overcomes the limitations caused by insufficient detail provided by the second order fibre orientation tensor (FOT) output from manufacturing simulations which are needed in advanced structural finite element models that incorporate damage and failure. An artificial neural network (ANN) framework is proposed to predict a fibre orientation distribution function (ODF) from the FOT. The ANN contains the equation for calculating the FOT from the ODF such that the ODF is optimized to reproduce the input FOT. It is shown that the proposed framework is nearly as accurate as using experimentally measured ODF for predicting the stress strain response and failure of a compression moulded composite.