Identifying fibre orientations for fracture process zone characterization in scaled centre-notched quasi-isotropic carbon/epoxy laminates with a convolutional neural network

This paper presents a novel X-ray Computed Tomography (CT) image analysis method to characterize the Fracture Process Zone (FPZ) in scaled centre-notched quasi-isotropic carbon/epoxy laminates. A total of 61 CT images small specimen were used fine-tune pre-trained Convolutional Neural Network (CNN) (i.e., VGG16) classify fibre orientations. The proposed CNN model achieves 100% accuracy when tested on same scale as training set. However, drops maximum 84% unlabelled specimens having larger scales potentially due their lower resolutions. Another code was developed automatically measure size FPZ based identified 0°plies largest which agrees well with manual measurement (on average within 3.3%). whole classification and process can be automated without human intervention.