Data-driven multi-objective optimization and integrated realization of high-performance lattice structures: Design, manufacturing, and experimental validation

Lattice structures are widely valued in aerospace, automotive, and biomedical fields for their low density, high specific strength, and superior energy absorption. This study presents a data-driven framework for the multi-objective optimization and integrated realization of high-performance rod-based lattice structures. A Multi-Mutation Genetic Algorithm assisted by Neural Networks (MMGA-NN) is proposed to efficiently optimize key mechanical metrics, including specific energy absorption, efficiency, load capacity, and relative density. Deep neural networks serve as surrogate models to replace finite element simulations, drastically reducing computational cost. To enhance convergence and solution diversity, a dual-fitness scheme combining weighted and Pareto-based evaluations is adopted. A strict selection mechanism ensures evolutionary stability, while a hybrid mutation strategy—comprising proportional, adaptive, and disaster mutations—enhances global exploration. Optimized lattice designs were fabricated using PEEK and FDM technology, and their mechanical performance was experimentally validated, confirming the effectiveness of the proposed approach. This unified framework for design, manufacturing, and validation demonstrates strong scalability and holds promise for future applications in multi-physics and multifunctional metamaterial optimization.