AI-driven property prediction and design optimisation of composite materials: A review

Abstract

This review examines the transformative impact of artificial intelligence (AI) and machine learning (ML) on composite materials design, property prediction, and optimisation. AI/ML integration has revolutionised conventional composite development, advancing materials discovery, design optimisation, and manufacturing process management across all phases from initial design to in-service monitoring. The work provides a structured classification of AI/ML techniques in composite materials. Supervised learning algorithms achieve good accuracy in predicting mechanical properties, while unsupervised learning analyses microstructures and identifies defects, and reinforcement learning optimises designs. Convolutional neural networks demonstrate exceptional effectiveness in defect identification and microstructural analysis. Physics-informed neural networks successfully integrate physical principles into data-driven models, achieving high accuracy while maintaining physical consistency. Key applications include predicting mechanical properties (tensile strength, elastic modulus, fatigue behaviour), thermal/physical properties, and multi-property models. Industrial implementations in the aerospace, automotive, and energy sectors have shown significant improvements. Current challenges include limited data availability, model interpretability, and regulatory compliance. Emerging technologies, such as quantum machine learning, federated learning, edge computing, digital twins, autonomous discovery platforms, and blockchain, represent promising future directions that require standardised datasets, transparent AI frameworks, and collaborative validation protocols.