Application of Artificial Neural Networks to predict fibre orientation in long fibre compression moulded composite materials

Abstract The material properties of compression moulded composites depends on the fibre orientation within a component, which changes as a function of position due to the flow of the material during the manufacturing process. Existing analytical tools are incapable of capturing all of the underlying physics of the problem involving long fibres which deform and interact with one another during the flow process. Artificial neural networks are used to predict the fibre orientation within a compression moulded composite by training the network on fibre orientation data. The fibre orientation data was produced using X-ray micro computed tomography measurements. The proposed artificial neural network framework is able to predict the orientation of the in plane orientation tensor normal components within 0.103 and the through thickness orientation tensor normal component within 0.0095 which is lower than the variability orientation of neighbouring microstructural units. Once trained the artificial neural network framework allows for near instantaneous predictions of fibre orientation presenting an opportunity for increased efficiency of the virtual design process of compression moulded composite materials.