idw – Informationsdienst Wissenschaft
Nachrichten, Termine, Experten
Nachrichten, Termine, Experten
Designing new cars is expensive and time consuming. As a result, manufacturers tend to make only minor changes from one model generation to the next. With DriverAerNet++, researchers at the Technical University of Munich (TUM) and the Massachusetts Institute of Technology (MIT) have now developed the largest open-source database for aerodynamic car design. More than 8000 models representing the most common vehicle types will make it possible to create more efficient designs with the aid of artificial intelligence. The goal is to make development processes more cost-effective while improving fuel efficiency and making advances in electric vehicles.
Carmakers often invest many years in the design of a vehicle. This starts with simulations using 3D models before the most promising designs are tested in the wind tunnel. The details and specifications of these tests, including the aerodynamics of a certain design, are generally not made public. Consequently, progress in such areas as fuel efficiency or the range of electric vehicles tends to be slow and limited to the individual companies.
Car designs using generative AI
As a starting point for DrivAerNet++, the researchers used 3D models from 2014 provided by Audi and BMW and representing various car body shapes. In addition, 26 parameters such as length, underbody characteristics and windshield angle were systematically adjusted. The team also conducted complex airflow simulations to calculate wind resistance for the various generated car designs.
The goal was to use DriverAerNet++ as training data for generative AI. This allows enormous quantities of data to be analyzed in seconds to generate innovative designs. Although tools of this kind already exist, the necessary data were not freely accessible in the past.
Improved fuel efficiency and progress with electric cars
Angela Dai, a professor of 3D Artificial Intelligence at TUM, says: “Our dataset can be used as an extensive library in order to generate new designs quickly with the aid of AI models with the goal of designing more fuel efficient cars in the future or improving the range of electric vehicles.”
Mohamed Elrefaie, first author of the study, says: “This dataset lays the foundation for the next generation of AI applications in automotive design, promotes efficient design processes, reduces R&D costs and will promote progress toward a more sustainable automotive future.”
Prof. Dr. Angela Dai
Technische Universität München
Professorship of 3D Artifical Intelligence
angela.dai@tum.de
M. Elrefaie, F. Morar, A. Dai, and F. Ahmed. DrivAerNet++: A large-scale multimodal car dataset with computational fluid dynamics simulations and deep learning benchmarks. In Thirty-eigth Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2024.
Angela Dai, Professor of 3D Artificial Intelligence
Astrid Eckert / TUM
Astrid Eckert, TU ; Free for use in reporting on TUM, with the copyright noted.
Merkmale dieser Pressemitteilung:
Journalisten
Informationstechnik, Maschinenbau
überregional
Forschungsergebnisse, Wissenschaftliche Publikationen
Englisch
Angela Dai, Professor of 3D Artificial Intelligence
Astrid Eckert / TUM
Astrid Eckert, TU ; Free for use in reporting on TUM, with the copyright noted.
Sie können Suchbegriffe mit und, oder und / oder nicht verknüpfen, z. B. Philo nicht logie.
Verknüpfungen können Sie mit Klammern voneinander trennen, z. B. (Philo nicht logie) oder (Psycho und logie).
Zusammenhängende Worte werden als Wortgruppe gesucht, wenn Sie sie in Anführungsstriche setzen, z. B. „Bundesrepublik Deutschland“.
Die Erweiterte Suche können Sie auch nutzen, ohne Suchbegriffe einzugeben. Sie orientiert sich dann an den Kriterien, die Sie ausgewählt haben (z. B. nach dem Land oder dem Sachgebiet).
Haben Sie in einer Kategorie kein Kriterium ausgewählt, wird die gesamte Kategorie durchsucht (z.B. alle Sachgebiete oder alle Länder).
