MS148 - Advances in machine learning for composite materials
Keywords: Composite, Machine learning
Over the past few decades, sectors like aerospace and wind energy (among others) have experienced significant impacts from the integration of high-performance composites. However, a challenge in modelling and designing composites is the limited computational efficiency of precise high-fidelity models. Conventional optimization methods for design often lead to complex procedures due to the large dimensions of the design space and the computational burden associated with high-fidelity simulations. In recent times, machine learning approaches have emerged as promising techniques to enhance the efficiency and reliability of various approaches for modelling of composites . These techniques offer a powerful approach to capture and understand the intricate behavior of composite materials by leveraging vast datasets and identifying underlying patterns. By training models on different sources of data, machine learning can accurately predict physical and mechanical properties, failure mechanisms, and durability, enabling efficient material design and reducing the need for extensive and costly experimental and computational testing programs . Furthermore, machine learning can facilitate the optimization of composite processing parameters, aiding in the achievement of desired material properties and reducing manufacturing defects. The application of machine learning techniques in composite materials not only enhances our understanding of these advanced materials, but also enables their broader utilization across industries such as aerospace, automotive, and renewable energy, where lightweight and high-strength materials are in high demand. In this mini symposium, we will discuss the state-of-the-art developments in usage of machine learning techniques for modelling, design, and process optimization of composite materials. Contributions on a wide range of topics are encouraged, from traditional and physics-enhanced machine learning for surrogate modelling, to learning and exploiting latent representations of material behavior, generative learning for material optimization, efficient data assimilation and active learning, among others.