[3H5-GS-3] Data mining: Applied data mining (2)

[3H5-GS-3-05] Data-driven design approaces for mechanical design using machine learning

〇Kazuo Yonekura^1,2, Hitoshi Hattori¹, Hiroki Saito¹, Katsuyuki Suzuki² (1. IHI Corporation, 2. The University of Tokyo)

Keywords:Mechanical Design

The aim of this paper is to show effectiveness of utilizing deep learning into mechanical design process. We propose a data-driven design framework for mechanical design process. It consists of three approaches; prediction of performance using deep regression model, shape generation with specified performance using generative model, and shape modification using reinforcement learning. We describe each approach that is separately published, and show numerical experiments that shows better performance.