Hybrid Differential Evolution and Sequential Quadratic Programming Algorithm

Abstract

Optimization algorithms are very important in product design. They could be divided into two classes: traditional local search methods and heuristic global ones. Sequential quadratic programming (SQP) algorithm has been known as one of the most prominent and fastest methods, but its local exploitation characteristic leads to the fact that it could be easily trapped by local optimum. However, heuristic methods such as differential evolution (DE) possess better convergence quality although their convergence speed is not good enough. This paper proposes a hybrid differential evolution and sequential quadratic programming algorithm, denoted as DE-SQP. At first, SQP adopts active set method and range space method to solve quadratic programming problems. Then, SQP is combined with DE. Experiments using benchmark optimization problems and engineering design problems are presented and DE-SQP is compared with other global optimization algorithms. Results demonstrate that DE-SQP is reliable, effective and efficient.