Performance of Metaheuristic Algorithms for Optimizing Variable-Speed Planning Problem of Geneva Mechanisms with Some Very Large Design Variables

Performance of Metaheuristic Algorithms for Optimizing Variable-Speed Planning Problem of Geneva Mechanisms with Some Very Large Design Variables

"The variable-speed planning problem for the optimum kinematic performance of Geneva mechanisms has been solved using the teaching-learningbased optimization (TLBO) in literature. TLBO requires a tremendous number of evaluations of the objective function to find satisfactory results. The unique feature of the optimization problem lies in the extremely large range of the values for design variables, which can be up to 483,677 times the prescribed range. Thus, the optimization problem demands a very high exploration capacity and a certain exploitation capacity, and serves as a benchmark problem to test the performance of metaheuristic algorithms. In this work, several known metaheuristic optimization algorithms, i.e. the cuckoo search (CS), differential evolution (DE) with a fixed (DE), random (RDE) and chaotic (CDE) scale factor, and combined-mutation DE algorithm (CMDE) are employed. Findings show that the best-to-worst sorting of performance is as follows: CMDE, DE, RDE, TLBO, CDE, CS. The exploration capacity of DE and CMDE using a fixed scale factor is superior to that of RDE, CDE and CS using a random or chaotic scale factor in the mutation operation of difference vectors or biased random walk. The user-supplied parameters for the metaheuristic algorithms on solution quality are investigated."