文献类型：期刊文献

英文题名：An intelligent prompt system for product generative design based on style perception preference values

作者：Chen, Yumiao[1];Xu, Yue[1]

机构：[1]East China Univ Sci & Technol, Sch Art Design & Media, Room 108,Xuhui Campus, Shanghai, Peoples R China

年份：2025

外文期刊名：JOURNAL OF ENGINEERING DESIGN

收录：;EI(收录号：20253419019082);WOS:【SCI-EXPANDED(收录号:WOS:001552187300001)】；

基金：This study was partly supported by the Research Project of Humanities and Social Sciences of the Ministry of Education (No. 24YJA760013), the National Natural Science Foundation of China (No. 51905175), Shanghai Soft Science Key Project (No. 24692109400), Xie Youbai Design Science Research Foundation (No. XYB-DS-202301).

语种：英文

外文关键词：Generative design; prompt system; style perception; affective design

摘要：Generative design enhances design efficiency and fosters creative divergence, allowing designers to express their intentions through a prompt system to guide models in the generative process. Currently, mainstream prompt systems use descriptive words to convey expectations for generative design. However, these descriptions often lack accuracy, and prompt systems have certain thresholds. This study focuses on constructing a prompt system based on style perception preference values to optimise the generative design process, ensuring that the generated products align more closely with designers' style perception expectations. Initially, we collected prompt words used by designers in generative design through web scraping and conduct cluster analysis to extract design elements and style perception dimensions. Subsequently, we designed a survey to investigate the influence of different design elements on user style perception. The collected data was employed to train a reinforcement learning framework, establishing a mapping model between design elements and style perception. Finally, we integrated the mapping model with a diffusion model to construct an optimised prompt system, enabling designers to directly generate designs by adjusting style perception preference values. Experimental results demonstrate that this system effectively enhances design efficiency and produces products that better meet designers' expectations.