Improvements to Mutation Donor Formulation of Differential Evolution

As one of the most promising novel evolutionary algorithms, differential evolution has been demonstrated to be an efficient, effective and robust optimization method for nonlinear optimization. Nevertheless, the convergence rate of differential evolution is still far from ideal when it is applied to optimizing a computationally expensive objective function that is frequently encountered in engineering optimization problems. This paper proposes three new schemes to determine the donor for mutation operation in differential evolution. These modifications to the mutation operator are analyzed and compared empirically by using a suite of artificial test functions. They are further examined with two practical neural-network-based aerodynamic data approximation cases. The simulation results demonstrate that the proposed strategies are capable of accelerating the convergence rate of the differential evolution algorithms .