UNDERSTANDING THE IMPACT OF ARCHITECTURAL DEFECTS ON COMPOSITE MATERIALS PERFORMANCE USING ARTIFICIAL INTELLIGENCE

Composite materials exhibit complex mechanical behaviours primarily influenced by their internal architectures, whether unidirectional, woven, or non-crimp fabrics. Variabilities in these internal architectures, such as defects, can locally alter the material's mechanical performance, which in turn can change of response on the structural scale under certain loading condition. Developing an understanding of the link between the internal architecture and mechanical response can be challenging in most cases. Experimental and modelling approaches can be cost-prohibitive and resource-intensive. Additionally, its often-complex task to extract meaningful patterns from the large datasets needed to fully characterise a material system. This study explores the use of Artificial Intelligence (AI) techniques to address these challenges and enhance our understanding of defect-induced variabilities in composite material performance.

In this work, the application of several AI techniques to composite materials will be presented. Deep Convolutional Recurrent Neural Networks are employed to model the non-linear, time-dependent mechanical response of composites taking into account the presence of defects. Additionally, pattern recognition methods are used to identify and quantify the influence of repeatable architectural features on material performance. Furthermore, deep learning techniques leveraging are applied to X-ray CT scan data to isolate key architectural elements that dominate the composite’s mechanical response using Gradient Weighted Activation Maps. This work demonstrates that AI-driven approaches offer a powerful and efficient means to providing new insights into the relationship between internal defects and composite materials performance. These methods present significant potential for improving the design, manufacturing, and optimization of high-performance composite structures.