A multi-task engineering design intention recognition approach based on Vision Transformer and EEG data

Engineering product design involves a variety of tasks and scenarios, including design modeling, design calculation, process planning, etc. When performing these design tasks, designers generate constantly shifting design intentions. Accurately recognizing these design intentions allows for a more thorough exploration of design processes from the perspective of cognition, facilitating the advancement of intelligent engineering design. Electroencephalogram (EEG) technology has emerged as an effective tool in recent years, which can provide direct insight into designers’ cognitive processes and intentions. However, the current application of EEG technology in engineering design faces difficulties in adapting to multi-task scenarios and rarely targets the design process directly. This study proposed a design intention recognition approach based on Vision Transformer (ViT) and EEG data applicable to multiple engineering design tasks. An image-like representation matrix is introduced to organize designers’ EEG data with the retention of its spatial and frequency features. Then, standard EEG data under different design intentions as well as the EEG data from real design processes is utilized to train and fine-tune a ViT-based design intention recognition model. An experiment workflow for collecting the two types of EEG data is also presented, along with detailed examples of three design tasks. The comparative experiment results and the case study demonstrates the feasibility of the proposed design intention recognition approach.