Extraction of evolutionary factors in smart manufacturing systems with heterogeneous product preferences and trust levels

Abstract

Smart products, as an increasingly significant part of smart manufacturing systems, have brought users extreme experiences. To provide a significantly better user experience and succeed amidst intense market competition, smart products undergo continuous and iterative evolution guided by diverse user requirements. However, users' requirements are heterogeneous, ever-changing, and misleading owing to the heterogeneous familiarity and deceptive interactions, leading to an erroneous identification of product evolutionary factors and inferior subsequent evolution. In existing works, the heterogeneous preferences and trust levels of users are considered superficially, and the extracted product evolutionary factors fail to adequately fulfill users' requirements. To address this issue, a reliable extraction method for evolutionary factors of smart products with heterogeneous product preferences and trust levels is proposed. First, the fuzzy similarity measure is introduced and a group classification method based on this measure is designed to filter out users without reference value. Second, the trust score induced order weighted averaging (TS-IOWA) operator is employed to obtain the comprehensive trust level of an individual user. Third, the dual trust aggregating operator is improved by two corrective factors being incorporated with the Einstein product and the Einstein sum to overcome the deficiency: an over-reduction of trust level and an over-expansion of distrust level, and then employed to aggregate multiple trust/distrust values and sort the evolutionary factors. Finally, two case studies on the App rating system and the evolution of smart wristbands are provided to verify the effectiveness of this proposed method. Throughout the analysis, the ranking accuracy of users' trust levels has increased by 25%, and the matching ratio between the extraction results and user requirements has increased from 57.14% to 85.7% with an improvement of 50%. The results demonstrate that the proposed method is capable of obtaining more credible evolutionary factors of smart products.