Implicitly inspired prediction approach for design thinking with multi-domain analogical knowledge driven by electroencephalogram data

Abstract

Explaining the potential relationship between analogical knowledge and target design problems is vital in analogical design. Existing studies neglect the role of multi-domain analogical knowledge in stimulating innovative design thinking. Furthermore, when data mining methods are used to evaluate the inspirational effect of analogical knowledge, the cognitive psychological state of designers is not fully considered. To address these issues, an implicitly inspired prediction approach for design thinking with multi-domain analogical knowledge driven by electroencephalogram (EEG) data is proposed. First, the fuzzy best–worst-method (BWM) model is used to screen analogical knowledge across three domains, namely, biology, abstract principles, and engineering case knowledge, which are retrieved from the AskNature platform, TRIZ effect webpage, and patent database, respectively, and then the transfer characteristics and semantic similarity of analogical knowledge are defined to support encoding. Second, an EEG experiment is designed. In the experiment, analogical knowledge from different domains serves as target stimuli, and the subjects are required to conduct knowledge transfer reasoning and scheme evaluation on the analogical knowledge presented in sequence. By collecting EEG data and mining the power density indicators of the frequency-domain features, the cognitive preferences of the subjects toward analogical knowledge are analyzed. Third, a support vector machine (SVR) model is constructed to predict the inspirational effect of analogical knowledge, after which the most suitable analogical knowledge is screened. A practical case study of a metal ore crushing and separation device is employed to validate the proposed approach. The validation results confirm that mining EEG data can explore the inspirational effect of analogical knowledge and parse designers’ psychological states during the design process.