부산대학교 도서관

Data-driven topology design is a method that enables gradient-free topology optimization like evolutionary algorithms. However, the dimension of design variables that can be handled by this method is limited to about 104 due to the fact that the deep generative model is used as the driving force for the solution search. Therefore, for high-resolution problems with large numbers of design variables, it is necessary to use some method to reduce the dimensionality of the design variables handled by the deep generative model. In this study, we propose a new framework that incorporates an interpolation function-based surrogate representation into the design process and discuss its applicability to high-resolution problems. We demonstrate that the performance of solutions obtained using the proposed method can almost achieve that of the existing method under a lower computational cost by demonstrating the effectiveness of the proposed method.