Inverse design of impact-resistant metamaterials using iterative neural-accelerated evolution algorithms
Author(s)
Sun, Yiyuan
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Compressible materials are essential for protecting people and structures from impulsive loads in applications such as footwear, automotive bumpers, and aviation landing gear. While traditional foams dominate current industry use, impact-resistant metamaterials are pivotal in shaping futuristic solutions. The highly tunable mechanical performance of metamaterials, driven by variations in their spatial arrangement of voids and solids, offers immense potential for optimizing and targeting impact resistance. However, identifying optimal metamaterial configurations can be computationally expensive and difficult. In this study, we present a robust framework that integrates finite element simulations, neural networks, and heuristic algorithms to efficiently discover metamaterial designs with superior impact resistance. We introduce energy absorption efficiency as the performance metric, defined as the ratio of shock energy absorbed by the metamaterial to the theoretical maximum, based on its stress-strain response. Starting with a minimal set of randomized metamaterial designs, our algorithm iteratively refines the neural-accelerated evolution strategy (INAES) to propose metamaterial geometries that maximize energy absorption efficiency. Our findings demonstrate that this approach effectively identifies designs with superior energy absorption efficiency, unlocking new possibilities for futuristic impact resistant systems.
Sponsor
Date
2025-05-01
Extent
Resource Type
Text
Resource Subtype
Thesis