DYNAMIC FUZZY-MEMBERSHIP OPTIMIZATION: AN ENHANCED META-HEURISTIC SEARCH

A new hybrid framework is proposed for optimization of orderable search spaces. It is based on fuzzy-membership of design variable alternatives to the optimal solution. The method has dynamic behavior since the membership values are assigned as any solution candidates arises during the search and they are summed based on their overlaps in the alternatives scope. A trail matrix is also utilized to indirectly share values of the fuzzy memberships further ranked regarding closeness of an individual to the optimal solution both in the objective function and in the design space scope. The method takes benefit of different random, vector-sum and probability-based walks to move new solutions toward the global optimum. Utilizing a generalized cooling procedure, the related thresholds are tuned to choose between different walk types in searching the design space and also a search refinement strategy is developed. Two variants of such a framework is then proposed and compared with each other in addition to some well-known procedures including genetic algorithm and particle swarm optimization. Test results with some treated problems, reveals the superior performance of the proposed algorithm and its special feature in adaptive tuning the diversity index during the search.