MFSE-based two-scale concurrent topology optimization with connectable multiple micro materials

• Published in: Computer methods in applied mechanics and engineering 2023-12, Vol.417, p.116409, Article 116409
• Main Authors: Sun, Zhaoyou, Liu, Pai, Luo, Yangjun
• Format: Article
• Language: English
• Subjects: Connectable microstructure design; Homogenization method; Material-field series expansion; Two-scale concurrent topology optimization
• ISSN: 0045-7825; 1879-2138
• DOI: 10.1016/j.cma.2023.116409

The concurrent design of different lattice material microstructures and their corresponding macro-scale distributions has great potential in achieving both lightweight and desired multiphysical performances. In such design problems, the lattice microstructures are usually separately optimized on the basis of the homogenization method, and the possibly poor connectivity between them is a key factor that severely hinders the fabrication and application of optimized two-scale structures. To handle the microstructure connectivity issue, this paper proposes a novel microstructure connectable strategy to bridge the gap between the two-scale and the full-scale model using the material-field series expansion (MFSE) method. Assuming different lattice material types in several pre-defined macro-scale regions, describe all types of microstructural topology with different portions of one material field function and update them simultaneously during the optimization process. Benefits from the material field definition with spatial correlation, the microstructures are well-connected without requiring additional constraints in the topology optimization model. The energy-based homogenization method is utilized for bridging the two-scale with different microstructures, while a decoupled sensitivities analysis for the microscale is employed to enhance the computation efficiency. Additionally, the proposed method significantly reduces the dimension of design variables, resulting in lower optimizer spending. The effectiveness and efficiency of the proposed method are demonstrated by several benchmark two-scale problems. Compared to density-based connectable methods, the proposed framework is easy to implement and reduces computational time by an order of magnitude in the 2D case.