BERT and Pareto dominance applied to biological strategy decision for bio-inspired design. (January 2023)

• Record Type: Journal Article
• Authors: Sun, Feng; Xu, He; Meng, Yihan; Lu, Zhimao; Chen, Siqing; Wei, Qiandiao; Bai, Chengying
• Is Part Of: Advanced engineering informatics. Volume 55(2023)
• Publication Date: 2023-01
• Subjects: Bio-inspired design -- Biological strategy -- BERT -- Pareto dominance -- Preference-based; Computer-aided engineering -- Periodicals; Engineering -- Data processing -- Periodicals; 620.00285
• DOI: 10.1016/j.aei.2023.101904
• Languages: English
• ISSNs: 1474-0346

Abstract

Abstract: Bio-inspired design (BID) approaches have provided numerous novel proposals for innovative design. Meanwhile, it resulted in an explosion of alternatives. Choosing a small amount of the personalized recommendations is becoming more and more difficult, and engineers may prefer fewer goals. Furthermore, engineers may choose designs that have variety and feasibility in BID. However, the diversity and feasibility of BID are often contradictory. Engineers need to apply multi-objective optimization methods to find better BID to ensure implementable. Aiming to overcome the above problems, based on the biological strategies from the most popular BID-AskNature, we proposed an interdisciplinary approach with the bidirectional encoder representation from transformers (BERT) and Pareto dominance for biological strategy decision (BPBSD) in the BID. First, we use BERT to find potential biological strategies similar to the keyword of BID. Then, aiming at the functional diversity and the feasibility of BID, we used the Pareto dominance theory to solve the contradiction. Finally, we verify the effectiveness of the proposed approach with three different application cases. Experiments show that the proposed BPBSD can balance the relationship between diversity and feasibility in BID. It is hoped that this work provides practical guidance for BID.