Machine learning-enabled inverse design of bioinspired layered composite structures with maximum auxetic performance.

Abstract

Layered composite structures inspired by biological tissues can exhibit out-of-plane negative Poisson's ratio, but identifying layups that maximize auxetic performance is challenging in high-dimensional designs. Here, we introduce an inverse design framework that searches for laminate layups with minimum Poisson's ratio. The approach combines multi-start resampling with machine learning-guided clustering to map layup families across layer numbers. Analytical relations from laminate mechanics link ply angles to effective properties, and computer simulations with laboratory measurements validate the predicted minima. The analysis resolves three layup categories, explains how shear-strain mismatch across bonded plies drives through-thickness auxetic expansion, and shows that simple symmetry rules reduce the search space. The framework reproduces previously reported minima and uncovers layups that approach lower Poisson's ratios under practical constraints. These results provide a physics-grounded, data-efficient route to engineer layered composite structures with strong auxetic responses and offer concise design rules for impact mitigation, vibration control, and flexible structures.