No. | Title・Author (Affiliation) |
---|---|
025 |
Development of CFD Method to Predict Drag Differences Due to Tire Profiles by Reproducing Rotational Deformation Shusei Tanaka・Sae Takahashi・Jun Ikeda・Kosuke Nakasato (Nissan Motor) Vehicle aerodynamic drag changes by tire profiles. The profile is deformed by the vehicle load and rotational centrifugal force, therefore both effects should be taken into consideration to predict aerodynamic performance with high accuracy. However, there are few studies that consider rotational deformation due to the difficulty of reproducing the shape. In this study, we developed a CFD method to predict aerodynamic drag differences due to tire profiles by reproducing rotational deformation of tire based on measurements of the profiles in rotating with load condition. |
026 |
Mechanism of Low Reynolds Number Oscillatory Flow Past Ahmed body Yusuke Atsumi・Suguru Shiratori・Itsuhei Kohri・Hideaki Nagano・Kenjiro Shimano (Tokyo City University) This study addresses the mechanism of low Reynolds number oscillatory flows past Ahmed body for the case of the slant angles 29 and 31 degrees. We report the structure of energy supply from a time-averaged field to a deviated oscillatory field, and their analogy to the well-known centrifugal instability. |
027 |
Development of Analyzing Method of Condensation Water Splashing on Electric Parts in a Vehicle by using MPS Method Yasuhiro Ohshima・Hisao Nishimori・Yusuke Imai・Hiroshi Kamatani (Toyota Motor) We have confirmed whether there are problems of condensation water splashing on electric parts installed a vehicle through design reviews and by spraying water to concerned area. However, it is difficult to clarify the water splashing route and it is needed to spend a lot of hours to identify the route. We developed the analyzing method of condensation water splashing on electric parts in a vehicle and made the water splashing route easier by considering simulation setting conditions and parameters on MPV method. |
028 |
Feasibility Study of Automated Design Method for Air Conditioning Ducts (First Report) Hiroshi Tanaka・Hiromune Kanamori・Hiroyuki Umetani (Toyota Systems)・Kenichi Ichinose (Toyota Motor) For automatic design, it is necessary to have a technology to derive shapes under the constraints of multiple design variables and multiobjective functions. In this paper, we report on the development of an evolutionary optimization calculation workflow for an air-conditioning duct that automatically performs a loop of shape change, fluid calculation, and property extraction to derive the optimal shape, and on the finding that a Pareto solution can be extracted even under the conditions of a very large number of design variables and multi-objective functions. |
029 |
Feasibility Study of Automated Design Method for Air Conditioning Ducts (Second Report) Hiromune Kanamori・Hiroyuki Umetani・Hiroshi Tanaka (Toyota Systems) In the first report, it was found that the evolutionary optimization calculation method under multi-design variables and multi-objective function constraints requires a large amount of computer resources to reach the optimal shape. In this paper, we report on our investigation how to reduce computational resources by using AI methods, and report that we were able to obtain an optimal shape with approximately the same performance using only one-tenth of the computational resources of evolutionary methods. |
030 |
Implementation of an Aerodynamic Reduced Order Model (ROM) based on Geometric Deep Learning (GDL) for Quick Design Review Bhanu Prakash Samala・Jiri Hajek・Paul Marston・Rahul Varadhan・Enric Aramburu (IDIADA Automotive Technology) The authors will present their experience in implementing different ML technologies, such as CNN or GNN to obtain aerodynamic forces in short timescales. This paper will focus on GNNs, which is currently the most promising approach to learn how to simulate fluid dynamics in geometrically complex domains. Simulation grids are actually graphs, so grid results can be directly translated into GNNs and vice versa, providing high efficiency and versatility compared to other predictive ML methods. The authors will also present an industrial application of GNN for drag and flow field prediction, ultimately allowing interactive analysis of new vehicle designs. |