能動学習を用いた最悪条件の自動検出手法の開発

抄録

Vehicle drivability is an important factor that is directly linked to comfort and product value. In the development of drivability performance improvement, there are cases that screening evaluation is performed by actual vehicles or simulation with CAE for the target phenomenon. However, the worst condition of the drivability phenomenon depends on the hardware and control specifications of each vehicle and it is difficult to extract by vehicle test. It is necessary to comprehensively test with a simulation or actual vehicles, and it is difficult to consider it in a limited resource. On the other hand, the objective function can be optimized with the minimum number of experiments by using Active Learning. In this study, we propose Bayesian Optimization that can predict the worst condition with minimum CAE calculation as a prediction method. Furthermore, we propose a development process using Bayesian Optimization. By using this method, a reduction in the number of calculations can be expected. Therefore, it can be used as a search method for the worst condition of drivability phenomenon at pre-screening in the initial stage of vehicle adaptation.