Set-Based Design Approach for Vehicle Reliability Development Using Bayesian Active Learning
Bayesian Active Learningを用いた車体信頼性開発のためのセットベース設計手法
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.)：￥1,100 Members (tax incl.)：￥880
- Publication code
- 20265361
- Paper/Info type
- Proceedings (Spring)
No.83-26
- Pages
- 1-5(Total 5 p)
- Date of publication
- May 2026
- Publisher
- JSAE
- Language
- Japanese
- Event
- 2026 JSAE Annual Congress (Spring)
Detailed Information
| Author(J) | 1) 宮木 耕太, 2) 菅井 友駿, 3) 西川 幸治 |
|---|---|
| Author(E) | 1) Kohta Miyaki, 2) Tomotaka Sugai, 3) Koji Nishikawa |
| Affiliation(J) | 1) トヨタ自動車, 2) トヨタ自動車, 3) トヨタ自動車 |
| Affiliation(E) | 1) Toyota Motor, 2) Toyota Motor, 3) Toyota Motor |
| Abstract(J) | 車体が耐えるべき最大入力負荷の設定において，悪路走行での車体入力モデルに機械学習によるサロゲートモデルを活用している．しかし，車体入力は非線形性が強い応答場である上に，本目的でのサロゲートモデルの活用では，任意の設計変数範囲においてその最大値を精度よく予測できる必要がある．本稿では非線形かつ複数の極大値を持つ応答に対して，極大値付近に着目し精度よく学習する手法を提案する． Translation |
| Abstract(E) | In defining the target strength of the vehicle body, the maximum load that the vehicle body can endure is estimated by utilizing a surrogate model based on machine learning, which is created from the vehicle body input from rough-road simulations. However, the response surface of the vehicle body input has a strong nonlinearity, and in utilizing the surrogate model for this purpose,It is essential to accurately predict the maximum values within various ranges of design variables. This paper proposes a new method to learn by focusing on the vicinity of the peaks for responses that are nonlinear and possess multiple local maxima. |
