Product similarity assessment for conceptual one-of-a-kind product design: A weight distribution approach

• We proposed a weight distribution approach to assign customer requirements (CRs).
• A weighted distance-based algorithm is developed for case retrieval.
• Similarity assessment is allowed at both product level and part level according to CRs.
• The proposed approach can benefit rapid conceptual design for customized product.
• Case study shows the efficiency and industrial applicability.

Case-based reasoning (CBR) is a promising methodology for assisting conceptual product design. The efficiency of case retrieval determines the quality of design. In one-of-a-kind product design, customer activities are increasingly involved. Under the circumstances customized product design results in the increase of product variety, which produces a large case library. This brings difficulties in managing customized product families, and also results in issues in retrieving similar cases. In the scope of our knowledge, Only limited studies attempt to research case retrieval by addressing these OKP practice for its conceptual product design. Moreover, customer requirements (CRs) are not fully interpreted to guide this process at the preliminary design stage. To cope with these issues, approaches for elicitation of CRs and retrieving appropriate cases play essential roles in successful case retrieval for OKP. This paper proposes a modularized generic product model (MGPM) for managing OKP product families. This structured generic product model can represent a significant number of product/module/component variants in a single product model. Also, the weight distribution model (WDM) is developed for assign weighting factors of CRs to multi levels of product physical configuration. For similarity assessment, a weighted distance-based algorithm is developed to calculate the degree of similarity between the target case and the reference cases. A case study on the fruit chute system is carried out to prove the efficiency and industrial applicability of the proposed approach. The result shows that the propose approach is capable of effectively retrieving similar product/part variants according to specified CRs.