Application of Deep Learning to Optimal Design

Hajime Igarashi
Graduate School of Information Science and Technology, Hokkaido University


Topology optimization which searches for optimal shapes by freely deforming materials can lead to optimization results with novel shapes and high performance. For this outstanding property, topology optimization has attained great attentions from industries. Topology optimization needs, however, large computing cost because it need a number of finite element computations. To overcome this difficulty, we have developed a new approach that accelerates topology optimization using a deep neural network. In this method, the deep neural network is trained by the device images and corresponding performances obtained by the preliminary optimization. Then, the trained deep neural network fast computes the performance of devices under test in the main optimization process to suppress the number of finite element computations. In this talk, some engineering examples for reduction of computing time of topology optimization by use of deep learning will be presented.

To read Proceedings, please sign-in.

Protected content here, for members only.
You need to sign in as a Regular JMAG Software User (paid user) or JMAG WEB MEMBER (free membership).

By registering as a JMAG WEB MEMBER, you can browse technical materials and other member-only contents for free.
If you are not registered, click the “Create an Account” button.

Create an Account Sign in 

Sign In
Create an account (Free) About authentication ID for JMAG website

Search Filter

  • All Categories

Proceedings Archives