Material similarity algorithm for process cases retrieval based on granular computing

Aiming at the problem of larger error in the similarity calculation of material type, material brand and heat treatment method in process cases retrieval, and according to the analysis that the essence of three kinds of similarity calculation is similar degree comparison of different materials on the physical and mechanical properties, the idea of unifying three kinds of similarity as material similarity and its algorithm based on granular computing is put forward. On the basis of five kinds of performance index values of material samples in the process case library, including hardness, tensile strength, elongation, impact toughness and thermal conductivity, the corresponding machinability grade value of every performance index is determined through comparing with the general material machinability classification table. By the integrated application of fuzzy tolerance quotient space theory and rough set theory, the objective weight of every performance index is calculated through analyzing its significance from different granularity of quotient spaces, and the multiplication combination weighting method based on subjective and objective weights is used to determine their comprehensive weights. The material similarity is calculated according to the weighted Euclidean distance. An application example verifies the feasibility, rationality and validity of the proposed algorithm. © 2014 Journal of Mechanical Engineering.