The influence of fabric architecture on impregnation behavior and void formation: Artificial neural network and statistical‐based analysis

Abstract

This work proposes an approach combining artificial neural networks (ANN) with statistical models to predict injection processing conditions for four reinforcement architectures: plain weave, bidirectional noncrimp fabrics, unidirectional fabrics (Uni) and random fiber mats (Random). Key results allow evaluating the velocity of the flow front by combining processing parameters and creating a three-dimensional response surface based on a properly trained ANN. This investigation is based on a large number of experimental results. The key role played by some physical parameters was associated with predicting the impregnation behavior (velocity of the flow front) during resin injection. The main outcome aims to provide a better control of void content in terms of size and position to the four fibrous reinforcements considered.