Machine learning guided resolution of mechanical trade-off in polymer composites via stress adaptive interface

Abstract

Developing polymer composites that simultaneously achieve high strength, toughness, and impact resistance remains a fundamental challenge due to inherent trade-offs and brittle interfacial failure. Here, we propose a universal toughening strategy that integrates a bone-inspired trabecular interlock architecture with a thermodynamically driven, stress-adaptive interface to enable efficient energy dissipation under mechanical loading. To address multi-objective optimization in composites design, we further develop a data-driven framework combining Pareto Set Learning and Active Learning, which systematically explores the composition–performance landscape to identify balanced, high-performance formulations. The optimized composites exhibit synergistic mechanical properties: strength up to 250 MPa, fracture toughness exceeding 14 MPa·m1/2, and impact resistance of nearly 4.8 J, surpassing most bioinspired and engineered polymer counterparts. The strategy is scalable, chemically versatile, and broadly applicable, offering a programmable route to next-generation lightweight composites for aerospace, transportation, and protection.