Similarity assessment is a cognitive activity that pervades engineering design practice, research, and education. There has been a significant effort in understanding similarity in cognitive science, and some recent efforts on quantifying the similarity of design problems in the engineering design community. However, there is a lack of approaches for measuring similarity in engineering design that embody the characteristics identified in cognitive science, and accounts for the nature of design activities, particularly in the embodiment design phase where scientific knowledge plays a significant role. To address this gap, we present an approach for measuring the similarity among design problems. The approach consists of (i) modeling knowledge using probabilistic graphical models, (ii) modeling the functional mapping between design characteristics and the performance measures relevant in a particular context, and (iii) modeling the dissimilarity using KL-divergence in the performance space. We illustrate the approach using an example of a parametric shaft design for fatigue, which is typically a part of mechanical engineering design curricula, and test the validity of the approach using an experiment study involving 167 student subjects. The results indicate that the proposed approach can capture the well-documented characteristics of similarity, including directionality, context dependence, individual-specificity, and its dynamic nature. The approach is general enough that it can be extended further for assessing the similarity of design problems for analogical design, for assessing the similarity of experimental design tasks to real design settings, and for evaluating the similarity between design problems in educational settings.

This content is only available via PDF.
You do not currently have access to this content.