Efficient Design of Automotive Structural Components via De-Homogenization

In the past decades, automotive structure design has sought to minimize its mass while maintaining or improving structural performance. As such, topology optimization (TO) has become an increasingly popular tool during the conceptual design stage. While the designs produced by TO methods provide significant performance-to-mass ratio improvements, they require considerable computational resources when solving large-scale problems. An alternative for large-scale problems is to decompose the design domain into multiple scales that are coupled with homogenization. The problem can then be solved with hierarchical multiscale topology optimization (MSTO). The resulting optimal, homogenized macroscales are de-homogenized to obtain a high-fidelity, physically-realizable design. Even so MSTO methods are still computationally expensive due to the combined costs of solving nested optimization problems and performing de-homogenization. To address these issues, this paper presents an efficient de-homogenization method that can be applied to any macroscale topology in order to obtain a high-fidelity multiscale structure. In contrast to prior de-homogenization methods, an alternative representation of the rectangular hole microstructure is proposed so that it is only dependent on the local density and stress distributions. Consequently, MSTO methods are not needed for the de-homogenization method to be applied. This makes the method applicable to any conceptual design, including those from simple single-scale TO codes. Additionally, the proposed de-homogenization method avoids the expensive mapping optimization problem associated with most projection de-homogenization methods by clustering sub-domains of the structure into discrete orientation angles. Pre-computed microstructures for each of the discrete angles are then assembled into each sub-domain. The proposed method is showcased for the design of a simple bumper and hood structure. The material distribution of each component is optimized in a two-dimensional TO problem for maximum stiffness.