A Generative Design and Drag Coefficient Prediction System for Sedan Car Side Silhouettes based on Computational Fluid Dynamics. (June 2019)

• Record Type: Journal Article
• Authors: Gunpinar, Erkan; Coskun, Umut Can; Ozsipahi, Mustafa; Gunpinar, Serkan
• Is Part Of: Computer aided design. Volume 111(2019)
• Page Start: 65
• Page End: 79
• Publication Date: 2019-06
• Subjects: Generative design -- Computational fluid dynamics -- Design support system -- Design sampling -- Machine learning; Computer-aided design -- Periodicals; Engineering design -- Data processing -- Periodicals; Computer graphics -- Periodicals; Conception technique -- Informatique -- Périodiques; Infographie -- Périodiques; Computer graphics; Engineering design -- Data processing; Periodicals; Electronic journals; 620.00420285
• DOI: 10.1016/j.cad.2019.02.003
• Languages: English
• ISSNs: 0010-4485

Abstract

Abstract: A design support system is developed in this work that can be integrated into the car side silhouette design tools and can estimate the drag coefficient of a given silhouette. This task is typically performed via two manners: namely wind tunnel testing and computational fluid dynamics (CFD) simulations. Due to the high computational cost for these two approaches, it is impractical to employ them during the silhouette conceptual design stage in a real time. Therefore, a mathematical model is obtained in this study for the drag coefficient estimation of a given silhouette. First, the desired number of silhouettes are generated via a generative design (silhouette sampling) technique so that the silhouettes are evenly distributed in the silhouette design space. Each silhouette is then tested via computational fluid dynamics simulations, and their corresponding drag coefficients ( C D s) are obtained. A training dataset is formed with the silhouette geometries and C D s of the silhouettes, and a mathematical model that can estimate the drag coefficient ( C D ) of a silhouette is finally obtained via principal component analysis (PCA) followed by regression/neural network methods. These three steps are repeated until a desired level of reliable mathematical model is obtained. Finally, three generative design test cases are illustrated based on the mathematical model obtained to predict C D of a given silhouette. Highlights: A generative design technique is employed for the generation of car side silhouettes. Silhouettes are regularly sampled in the design space. Drag coefficients for the silhouettes are computed via CFD simulations. A mathematical model is obtained using machine learning techniques. Silhouette optimization is carried out using the mathematical model.