Data-driven lightweight and robust design of hybrid TPMS structures

Lightweighting and versatility are key research themes in aerospace, transportation, and robotics. Triply periodic minimal surfaces (TPMS), owing to their unique physical advantages, are considered ideal candidates for achieving these goals. However, existing design approaches fail to effectively explore the complex TPMS design space, especially for hybrid configurations, and exhibit strong sensitivity to the initial topology. This study presents a novel optimization framework for TPMS design. Specifically, the hybrid TPMS structures are employed as internal supports and the mass is minimized under stiffness and strength constraints by regulating topology, periodicity, and wall thickness. To systematically explore the intricate relationship between geometric and physical properties, a two-stage optimization scheme, combining data-driven coarse optimization with gradient-based fine optimization, is proposed. In the first stage, Bayesian optimization is employed to perform a coarse global search and identify a near-optimal topological configuration with fewer evaluations. The solution serves as a high-quality initial guess for the second stage, where a gradient-based method refines local geometric features to achieve optimal structural performance. The optimized structures feature lamellar porous morphologies aligned with the loading direction, enhancing deformation stability and enabling multi-stage energy absorption. Both numerical simulations and mechanical experiments confirm that the present method effectively leverages the intrinsic potential of TPMS architectures. Compared with other advanced designs, this study achieves 5–20% reduction in material consumption while maintaining superior mechanical performance and manufacturability, underscoring its strong promise for engineering applications.