This work presents a data-driven design approach to hierarchical hybrid structures with multiple lattice configurations. Two design variables are considered for each lattice substructure, one discrete variable indicating the configuration type and the other continuous density variable determining the geometrical feature size. For each lattice configuration, a series of similar lattice substructures are sampled by varying the density variable and a corresponding data-driven interpolation model is built for an explicit representation of the constitutive behavior. To reduce the model complexity, substructuring by means of static condensation is performed on the sampled lattice substructures. To achieve hybrid structure with multiple lattice configurations, a multi-material interpolation model is adopted by synthesizing the data-driven interpolation models and the discrete lattice configuration variables. The proposed approach has proved capable of generating hierarchically strongly coupled designs, which therefore allows for direct manufacturing with no post-processing requirement as required for homogenization-based designs due to the assumption on scales separation.
Data-driven, Topology optimization, Hierarchical design, Hybrid structure, Multiple lattice configurations