Development of Pipeline Design Method using 3D Shape Generation via Integration of Variational Autoencoder and CAD
変分オートエンコーダーとCADを連携した三次元形状生成技術による管路設計手法の構築
- Delivery
- Available on the other site
- Click here to order.
- Publication code
- 20264392
- Paper/Info type
- JSAE Transaction
Vol.57 No.3
- Pages
- 530-535(Total 6 p)
- Date of publication
- May 2026
- Publisher
- JSAE
- Language
- Japanese
Detailed Information
| Category(J) | 論文 Translation |
|---|---|
| Category(E) | Paper |
| Author(J) | 1) 木佐貫 祥一郎, 2) 松村 直也, 3) 杉浦 卓也 |
| Author(E) | 1) Shoichiro Kisanuki, 2) Naoya Matsumura, 3) Takuya Sugiura |
| Affiliation(J) | 1) 株式会社アイシン, 2) 株式会社アイシン, 3) 株式会社アイシン |
| Abstract(J) | ヒートシンクなどの管路を有する熱流体製品の性能評価効率化に向けて，サロゲートモデルが活用されている．本技術で性能評価は高速となったが，学習用のCAD形状作成には依然時間を要する．本問題解決に向けた，形状生成AI技術であるVAEと，CADの連携による，3次元形状生成技術の構築と設計活用について述べる． Translation |
| Abstract(E) | Surrogate models are gaining attention as alternatives to CFD simulations due to their computational efficiency. However, constructing such models requires extensive CFD results, and preparing the corresponding CAD geometries is particularly time-consuming. This study aimed to establish a rapid generation technique for CAD models of cooling pipes suitable for CFD evaluation. A variational autoencoder (VAE) was developed to generate pipeline layout. A 3D modeling system using CAD software was developed to convert these layouts into three-dimensional cooling pipe geometries. The resulting models were confirmed to be compatible with CFD. Future work includes enhancing shape diversity and implementing constraints in the VAE-based generation process, as well as applying the developed method to efficiently create training data for surrogate models and establish an optimization workflow using these models. |
