Abstract: To expand the application of Triply Periodic Minimal Surfaces （TPMS） in the field of sports protective equipment, and improve the impact protection capability of sports helmets for the head, a bionic-turtle-shell hierarchical sandwich structure （BHSS） was designed in this study. The feasibility of applying BHSS to sports helmet design was verified by referring to the test methods specified in GB 24429-2009. Taking the maximum specific energy absorption （ESEA） of BHSS and the minimum peak acceleration of the punch （-a） as the dual optimization objectives, a high-precision ensemble surrogate model composed of Kriging （KRG）, radial basis function neural network （RBF）, and response surface methodology （RSM） was constructed. The Pareto frontier was obtained using the NSGA-Ⅱ algorithm. After ranking via the entropy-weighted TOPSIS method and subsequent engineering correction, the optimal parameter combination was determined as follows: carbon fiber layup angle of 0°/90°/±45°s, rubber layer thickness of 1 mm, P-type TPMS cell, and cell wall thickness of 0.4 mm. Compared with the initial model, the optimal model showed a 9.6% increase in specific energy absorption and a 16.68% decrease in peak acceleration. This study can provide certain ideas and theoretical support for the innovative design and performance optimization of sports helmet structures.