Development of 3-dimensional Shape Generative Method using VAE and Its Application to CAE
VAEを用いた3次元形状生成技術の開発およびCAE解析への適用
- Delivery
- Available on the other site
- Click here to order.
- Publication code
- 20234504
- Paper/Info type
- JSAE Transaction
Vol.54 No.5
- Pages
- 1074-1079(Total 6 p)
- Date of publication
- Sep 2023
- Publisher
- JSAE
- Language
- Japanese
Detailed Information
Category(J) | 論文 Translation |
---|---|
Category(E) | Paper |
Author(J) | 1) 谷口 真潮, 2) 新谷 浩平, 3) 小野寺 啓祥, 4) 大塚 紀子, 5) 勝原 忠典 |
Author(E) | 1) Mashio Taniguchi, 2) Kohei Shintani, 3) Hiroaki Onodera, 4) Noriko Ohtsuka, 5) Tadasuke Katsuhara |
Affiliation(J) | 1) トヨタ自動車, 2) トヨタ自動車, 3) トヨタ自動車, 4) トヨタ自動車, 5) トヨタ自動車 |
Abstract(J) | VAE(Variational Autoencoder)を用い複数車種の3次元形状を教師データとし,任意の潜在変数を選択することで多様な車両の3次元形状を生成する技術を構築した.構築技術を用いて生成した3次元形状がCAE解析へ適用可能であることを確認した. Translation |
Abstract(E) | Predictions using machine learning typically require a large number of design variables that contain performance results (aerodynamics, crash, etc.). Proposed technique based on VAE (Variational Autoencoder) can generate new styling by sampling from latent space values. This method was applied to an engine hood outer. The FE mesh was created by VAE and its eigenvalues were analyzed. The results matched well with the results of created from CAD. It was suggested that machine learning improved prediction accuracy. |