An efficient data-driven optimization framework for elastically isotropic lattice structures

• A data-driven framework is developed for optimizing elastically isotropic lattice structures. • A sample database and surrogate model are constructed to accelerate the optimization process. • The NSGA-II algorithm enables online optimization and achieves isotropy in just over ten seconds. • Twelve optimized microstructures exhibit high near-isotropic behavior. The elastic properties of an isotropic lattice are direction-independent, ensuring a consistent stress–strain response of the structure to loads in any spatial direction. This paper presents a data-driven approach to optimize the elastically isotropic properties of lattice structures efficiently. Based on the level set method, the implicit description strategy is employed to achieve geometric modeling for various types of basic anisotropic lattice microstructures. The elastic properties of the anisotropic microstructures are characterized using the numerical homogenization method. Based on these modeling and characterization works, a sample database and kriging surrogate model are constructed to establish the mapping relationships among design parameters, effective elastic properties, and relative densities. After completing the offline steps, an intelligent optimization algorithm (NSGA-II) is introduced to optimize the elastically isotropic property online. Twelve typical isotropic microstructures are studied. The optimization results highlight the high efficiency of the proposed method, with the time for achieving highly near-isotropic properties being just over ten seconds. Meanwhile, the prediction accuracy exceeding 99% for each microstructure with a specified volume fraction. Finally, finite element analysis and experiments verify the elastic property and isotropic behavior of the desired isotropic lattices.