Improving prediction accuracy of an ignition model by weighting using machine learning
機械学習を用いた重み付けによる着火モデルの予測精度向上
- Delivery
- Available on this site
- Format
- Price
- Non-members (tax incl.)：￥1,100 Members (tax incl.)：￥880
- Publication code
- 20214835
- Paper/Info type
- Symposium Text
- Pages
- 1-6(Total 6 p)
- Date of publication
- Dec 2021
- Publisher
- JSAE & JSME
- Language
- Japanese
Detailed Information
| Author(J) | 1) 西井 俊貴, 2) 山﨑 由大 |
|---|---|
| Author(E) | 1) Shunki Nishii, 2) Yudai Yamasaki |
| Affiliation(J) | 1) 東京大学大学院, 2) 東京大学大学院 |
| Affiliation(E) | 1) Graduate School of Tokyo University, 2) Graduate School of Tokyo University |
| Abstract(J) | 予混合気の圧縮自己着火をモデルベースで制御するために，精度が高く計算負荷が軽い予混合気の着火モデルが必要となる．本研究では，Livengood-Wu積分に基づく物理モデルと機械学習を融合させることで，着火遅れ時間の物理ー機械学習融合モデルを構築した．構築したモデルは，ニューラルネットワークによって複数の物理モデルの予測に重み付けを行うといった仕組みであり，実験データに対して良好な予測精度を見せた． Translation |
| Abstract(E) | For model-based control of compression auto-ignition of premixture, an ignition model of premixture with high prediction accuracy and light computational load is required. In this study, a machine learning method was applied to improve the prediction accuracy of an existing ignition model. The method weights the prediction values by multiple ignition models using a neural network considering the driving conditions of the engine. The neural network and model parameters in the ignition models were trained simultaneously using the training data obtained on an engine bench. The prediction accuracy for the test data was better than that of the previous methods. |
