Cognitive shape similarity assessment for 3D part search

Abstract

Mass customization aims to satisfy diverse customer requirements with high product variety while maintaining reasonable manufacturing cost and lead time. Allowing customers to perceive product differentiation is a critical factor for most design methods developed for mass customization. This study examines 3D part search from the human cognitive perspective. We designed and conducted a quasi-factorial experiment to understand how structured variations of four factors—the shape, type, dimension, and location of the feature volume of a part model—affect human judgment of part similarity. The corresponding factorial similarity values were computed with different shape signatures in the form of the feature adjacency graph. The human responses were obtained by paired comparisons of test parts, and quantified as the cognitive similarity. Statistical analysis of the experimental results showed that the type and shape factors played an important role in the subjects’ judgments. Back-propagation neural networks were trained to model the correlations between the cognitive and the factorial similarity values. The performance of the networks validates our idea of incorporating human cognition into assessment of 3D part similarity. This study presents a systematic approach for personalized part search that reflects individual perception of shape similarity.