Generative product design Kansei mapping by an integrated temporal convolutional network-transformer-gated recurrent unit and stable diffusion model

Ceramic incense burners, as important carriers of traditional Chinese culture, are closely associated with religious practices, literati lifestyles, and ritual systems. However, the form design of existing ceramic incense burners remains largely conservative, making it difficult to meet the diversified aesthetic preferences and emotional needs of younger consumers. Therefore, this study proposes a generative form design method integrating Kansei Engineering (KE) with a Temporal Convolutional Network-Transformer-Gated Recurrent Unit (TCN-Transformer-GRU) model. First, a morphological analysis system is established based on representative market samples. Subsequently, seven Kansei descriptors influencing purchase intentions are extracted through online text mining, and their relative importance is determined using a Cloud Model-Entropy Weight Method. A TCN-Transformer-GRU model is then constructed to map users’ Kansei requirements to product form features, enabling the prediction of optimal form combinations with high emotional value. Based on these results, a Stable Diffusion Model (SDM) generates corresponding visual design schemes. Finally, Grey Relational Analysis (GRA) combined with eye-tracking experiments comprehensively evaluates user preferences. The proposed Generative Product Kansei Mapping Framework (GPKMF) improves emotional mapping accuracy and design efficiency, supporting the digital design and innovative inheritance of traditional ceramic artifacts.