文献类型：期刊文献

英文题名：Automated emotional design generation for NEV wheel hubs: Integrating StyleGAN2-ADA and WOA-SVR within Kansei engineering

作者：Wang, Yi[1];Zhou, Meiyu[1];Wang, Zhengyu[2];Cai, Weilin[1];Sun, Xin[1,3];Zhu, Huijuan[1]

机构：[1]East China Univ Sci & Technol, Sch Art Design & Media, 130 Meilong Rd, Shanghai 200237, Peoples R China;[2]COMAC Shanghai Aircraft Customer Serv Co Ltd, Ind Design Inst, Shanghai 200241, Peoples R China;[3]Qinghai Univ, Sch Mech Engn, 251 Ningda Rd, Xining 810016, Peoples R China

年份：2026

卷号：298

外文期刊名：EXPERT SYSTEMS WITH APPLICATIONS

收录：;EI(收录号：20254519475173);WOS:【SCI-EXPANDED(收录号:WOS:001593358000001)】；

语种：英文

外文关键词：Generative design; StyleGAN2-ADA; WOA-SVR; Kansei engineering; NEV wheel hub design

摘要：Wheel hub design is significant in the New Energy Vehicle's (NEV) aesthetics and user emotional connection. However, the existing design process faces two main challenges: traditional concept methods suffer from efficiency bottlenecks, and designers' subjective judgments make it difficult to capture users' emotional preferences. Recently, Generative Adversarial Networks (GANs) have been introduced to industrial design. However, conventional GAN-based methods typically require large-scale datasets to achieve high-quality design generation, which limits their applicability in vertical product domains where data are scarce. To overcome these challenges, this paper proposes an emotional design generation approach for NEV wheel hubs by combining StyleGAN2 with Adaptive Discriminator Augmentation (StyleGAN2-ADA) and Whale Optimisation Algorithm-Support Vector Regression (WOA-SVR). First, the StyleGAN2-ADA is trained on a limited NEV wheel hub dataset to achieve automated design generation. Second, Ward hierarchical clustering dentifies representative samples, while morphological analysis deconstructs design features. Subsequently, Factor Analysis (FA) categorises Kansei words to extract principal emotions. After collecting users' emotional ratings of wheel hubs, WOA-SVR constructs a "design features-emotional needs" mapping model. Finally, the generated images are used to validate the implementation of the generation and prediction models. The case study demonstrates that the proposed method not only generates diverse design alternatives aligned with users' emotional preferences but also provides reliable predictions of their emotional responses, thereby systematically improving the NEV wheel hub design process while preserving its emotional impact.