Abstract: The explosion load poses a serious threat to the safety of buildings and personnel, so the research and development of explosion-proof technology is particularly important. In this paper, the antiknock performance of Foam-filled Auxetic Honeycomb Sandwich Structure (FAHSS) was studied by finite element calculation, the accuracy of the simulation results was validated through comparison with experimental data from the existing literature. The anti-knock properties of the unfilled Auxetic Honeycomb Sandwich Structure (AHSS) are compared and analyzed. On this basis, in order to further improve the energy absorption characteristics of FAHSS and carry out efficient structural design, an approximate model of the antiexplosion performance of FAHSS under explosion load was constructed based on the finite element model combined with the uniform Latin Hypercube experimental design method and radial basis function neural network (RBFNN). The multi-objective optimization of the trained RBFNN model is carried out by genetic algorithm. The results show that the maximum displacement (MaxD) and energy absorption (EA) of FAHSS under explosion load decrease by 9.2% and increase by 3.9% compared with AHSS of the same mass. Compared with AHSS with the same wall thickness, the MaxD of FAHSS was reduced by 20.6% and the EA was improved by 2.1%. The prediction accuracy of the approximate model based on RBFNN for MaxD and -X_ASEA of FAHSS is 0.991 and 0.998, respectively. This approximate model significantly enhances computational efficiency while maintaining high precision. The structure optimization of FAHSS found that compared with the benchmark model, the MaxD of Pareto fronts set decreased by 11.6%, and the -X_ASEA decreased by 9.6%. While MaxD decreased by 5.7% due to the compromise design point, - X_ASEA decreased by 6.8%. The multi-objective optimization using genetic algorithms notably enhances the anti-knock performance of FAHSS. The findings offer valuable insights for the efficient design and optimization of FAHSS.