Data-driven topology design using a deep generative model

Abstract In this paper, we propose a sensitivity-free and multi-objective structural design methodology called data-driven topology . It is schemed to obtain high-performance material distributions from initially given in domain. Its basic idea iterate the following processes: (i) selecting dataset of according eliteness, (ii) generating new using deep generative model trained with selected elite distributions, (iii) merging generated dataset. Because nature model, are diverse inherit features training data, that is, distributions. Therefore, it expected some superior current by dataset, performances newly improved. The further improved iterating above processes. usefulness demonstrated through numerical examples.