Cuckoo search and teaching-learning-based optimization algorithms for optimum synthesis of path-generating four-bar mechanisms

Abstract

The optimization of one-phase dimensional synthesis problems for path-generating four-bar mechanisms is a difficult task. Such problems have mainly been solved in the literature using the differential evolution (DE) algorithm and its variants, such as a combined-mutation DE (CMDE). The shortcoming of using DE or its variants lies in the time-consuming parameter tuning that can take up to a dozen iterations. In this study, two well-known population-based metaheuristic optimization methods, i.e. the cuckoo search (CS) and teaching-learning-based optimization (TLBO) algorithms are employed to solve five representative problems. Only one user-supplied parameter is needed for CS and none for TLBO. Findings show that the solution accuracy of CS is superior to that of DE and TLBO. Moreover, CS, with its advantage of parameter tuning only four times, rivals CMDE in solution accuracy. The reason for the superior exploitation capability of CS lies in the special combination of two search modes, i.e. the Levy flight with respect to the best individual and the random walk. To bring the performance of TLBO into full play, the number of function evaluations is two times the original one. However, TLBO still fails to provide much more accurate solutions for the first two problems.