CGAN-driven intelligent generative design of vehicle exterior shape

In recent years, with the rapid advancement of intelligent generative algorithms and the continuous improvement of computing power, the application of artificial intelligence methods in product design has become increasingly widespread. Data − driven intelligent generative design approaches have offered new options for the development of industrial design. However, current research on intelligent generative design primarily focuses on computer science and lacks the input of complex design parameters with significant engineering implications, such as structural, functional, and performance parameters. Against this background, taking the design of small electric vehicles as an example and adhering to engineering design principles, this paper proposes an intelligent generative design framework and method for industrial product design. In the process, a specialized dataset covering multiple categories of vehicle exterior shape designs was constructed. This dataset consists of thousands of vehicle exterior shapes. Based on domain − related feature engineering, each datum is described by an engineering semantic description specifically tailored to the exterior shapes of small electric vehicles. These descriptions include vehicle functional classification, Kansei descriptions on performance, and structural feature parameters. According to the characteristics of the dataset, a Conditional Generative Adversarial Network (CGAN) algorithm model was constructed and improved to learn the distribution patterns of 3D shapes and engineering semantics. Moreover, a webapp for the intelligent generative design and management of vehicle exterior shapes was developed based on the trained CGAN. Finally, a case study was conducted. The CGAN model was applied to generate new design schemes for small electric vehicle exterior shapes in accordance with new engineering design requirements from the perspectives of structure, functionality, and performance. This research provides theoretical and methodological support for intelligent generative design that takes into account functional, performance, and structural requirements, and enables the rapid generation of near − optimal design solutions for further detailed design. The code and dataset are publicly available at https://github.com/1506438785/IGD.