機械製品の自動設計に向けた基礎検証

（Deep Learningによるトポロジー最適化）

抄録

In recent years, there has been a demand for more efficient design of mechanical products, and research into automatic design systems using AI has been progressing. Ideally, by constructing a surrogate model that learns from actual data that combines required specifications and product structure, it is possible to create a system that presents optimal design proposals when required specifications are input. As a basic verification, a surrogate model was constructed that inputs the analysis conditions for topology optimization and infers the resulting structure. The inference accuracy evaluation showed that the topology optimization results could be faithfully reproduced in approximately 90% of cases, with an interesting trend of improvement in the grayscale region.