Multimaterial Metamaterial Inverse Design via Machine Learning for Tailorable and Reusable Energy Absorption

The demand for precisely tailorable mechanical parameters of energy-absorbing structures is emerging. This paper proposes a machine learning-driven inverse design framework that resolves this multiobjective challenge through 181-dimensional parameter optimization. Our method integrates multimaterial compatibility (TPU/resin/NiTi/Al alloy) with topology-morphing body-centered cubic (BCC) lattices, where nodal coordinates, beam diameters, and material parameters are co-optimized. We delve into studying the effects of material parameters, nodal coordinates, and beam diameter variations on the structural compressive performances by conducting over 20,000 simulation experiments on randomly generated BCC lattice structures using a finite element analysis. Subsequently, the metamaterials with the specific platform stress values (from 0.015 to 4.05 MPa) and specific energy absorptions (from 0.049 to 23.377 J/g) can be inversely designed with the aid of the artificial neural networks and genetic algorithms to pinpoint optimized parameters from a 181-dimensional space. Noteworthily, the metamaterials in NiTi alloy presented a high-level reusability even after five compression cycles (over 50% recovery), demonstrating its advantage in realizing the reusable and desired energy-absorbing performances. This method has been rigorously validated through additive manufacturing and experimental characterization. This work bridges the critical gap between customizable energy absorption and structural reusability.