Digital Coding Inverse Design of Hybrid TPMS Lattice Metamaterials With High Fidelity Under Limited Data via DWT‐CNN

Hybrid triply periodic minimal surface (TPMS) lattices provide a versatile platform for tailoring mechanical performance, but efficient optimization remains challenging because of their discrete combinatorial design space. This study considers two complementary TPMS‐derived unit cells: a Gyroid‐Diamond‐derived hybrid configuration (GD) and a Fourier‐expansion‐based improved Primitive configuration (FP). A 16‐bit binary representation is used to encode their spatial arrangement in a 4 × 4 × 4 lattice. To enable rapid response prediction, a lightweight multibranch convolutional model integrating discrete wavelet transform features is developed and coupled with active‐learning‐assisted database construction. Using ≈2000 finite element samples, the model achieves an independent‐test overall stress R 2 above 0.99, while complete‐database R 2 values for yield stress and specific energy absorption (SEA) reach 0.935 and 0.9697, respectively. The trained surrogate is embedded in a genetic algorithm for inverse design under yield‐stress, SEA, and multiobjective constraints. Finite element verification shows prediction errors below 4% for optimized configurations. The resulting designs reveal clear mechanical complementarity: increasing FP content lowers yield stress, whereas increasing GD content enhances SEA. By integrating digital encoding, multiresolution surrogate modeling, active learning, and inverse optimization, the proposed framework provides an efficient route for customized lightweight energy‐absorbing lattice design.