Tailoring Energy Absorption of Curved-Beam Lattices Through a Data-Driven Approach.

Abstract

Programmable mechanical metamaterials demonstrate significant potential for realizing high-performance mechanical responses, particularly in the field of energy absorption. In this study, a novel curved-beam thickness gradient lattice structure (CBTGLS) is proposed. Based on an intelligent inverse design framework integrating deep learning and genetic algorithms, the beam thickness and curved-beam control points of the CBTGLS were optimized to maximize its total energy absorption (EA) and specific energy absorption (SEA). Furthermore, this research employed interpretability methods, such as Shapley Additive Explanations (SHAP) and Partial Dependence Plot (PDP), to analyze the influence mechanism of geometric parameters on energy absorption performance, aiming to enhance design efficiency and establish a clear design rationale. The results indicate that the optimized CBTGLS exhibits significant improvements in both EA and SEA. Specifically, compared to a baseline straight-beam lattice structure possessing an identical thickness gradient, SEA of the optimized CBTGLS was enhanced by 49.12%. Among the investigated parameters, beam thickness was identified as having a particularly significant impact on performance. Furthermore, it was observed that a curvature profile bending more towards the outer side of the unit cell is more beneficial for enhancing the energy absorption capabilities of the lattice structure.