Session: 10-03-02: CFD Applications - II

Paper Number: 96817

96817 - Aerodynamic Analysis of a Car Based on Computational Fluid Dynamics and Machine Learning 

Vehicles play an essential role in daily life, which could greatly benefit people by making travel more accessible and faster. As one of the most important performance parameters, the aerodynamic performance of a vehicle attracts more and more attention in the industry. The aerodynamic performance could impact a vehicle's power, economy, and handling stability, and the aerodynamic analysis is one of the most important parts of the car designing process.  Intending to design more energy-saving cars, my research focuses on the aerodynamic performance of a car and investigates how the car body shape, or more specifically, the angles of the front and the rear window, can affect the aerodynamic drag and lift on a moving car. In this research, I use computational fluid dynamics as well as a machine-learning algorithm to investigate the problem and determine the optimal designs. First, I use a computational fluid dynamics tool to simulate the flow over the 50 two-dimensional cars with varying angles of the front / rear windows. Then, by using these computational fluid dynamics simulation results as a training database, I develop a machine learning-based algorithm to quickly determine the optimal design of the angle of car winds. Usually, the computational fluid dynamics simulations take a long time to get the results, so here, I especially use machine learning to lever the results from the CFD simulation and further determine the optimal design with a greatly reduced simulation time. Finally, by using the results from the two-dimensional simulation as a reference, I model three realistic three-dimensional cars (including an old-fashioned car with vertical windows, a pickup car, and a step-back car). Also, I simulate the turbulent flows over the designed cars, and the velocity field, pressure field is observed. It shows that the car shape has great effects on the aerodynamic performance of a car, especially velocity and pressure fields near the weak region. Results further show the step-back car has a better aerodynamic performance among the three designs, which could reduce the aerodynamic drag up to 30%. This study of the aerodynamic performance of a car, such as the shape of the front / rear windows could greatly reduce the fuel consumption, which further could help environmental protection by reducing emissions. This research also has a great significance to new product development, optimization of the car because the machine learning-based algorithm could use to quickly determine the optimal designs.  As regards the methodology used in this research, by combining the computational dynamic simulation and machine learning together, my research provides an easy use approach for the design and aerodynamic analysis of next-generation cars.

Presenting Author: Xingchuan Ma Portsmouth Abbey School

Presenting Author Biography: Ryan is a student in Portsmouth Abbey school. He had always been keen on different science topics. He is best at doing topics that focuses on fluid mechanics and computer science. He would like to use his research to solve the problems encountered in real world. He did a project by using aerodynamic analysis on different cars and applying machine learning to the datas from the analysis. He is also a member of the science magazine writer in school.

Authors:

Xingchuan Ma Portsmouth Abbey School