Development of machine learning methods for mechanical problems associated with fibre composite materials: A review

Fibre composite materials (FCMs) are widely used in the aerospace, military defence, and engineering manufacturing industries due to their high strength and high modulus. Understanding the constitutive laws, defect detection, impact dynamic response, tribological behaviour and fatigue failure of FCMs is essential in these industries because the mechanical behavior of FCMs is often influenced by various factors, including fiber arrangement and matrix properties. Due to the anisotropic and heterogeneous nature of FCMs, research on their mechanical properties often relies on costly experiments with poor reproducibility and computationally intensive simulations. In contrast, machine learning (ML) methods can rapidly uncover data relationships and are highly reproducible. Moreover, modern FCM manufacturing and testing techniques have generated large amounts of data. This article not only provides a comprehensive analysis of the application of ML methods but also emphasizes the applicability and future trends of different ML approaches in FCMs. In constitutive model building, deep neural network models can consider the subtle connections between multiple parameters, thereby revealing deeper relationships among the data. In defect detection and impact dynamics problems, convolutional neural network models can effectively extract information related to mechanical performance from images. This paper provides inspiration for the application of ML methods to solve mechanical problems and guide the optimal design of FCMs.