An Enhanced Mode-Pursuing Sampling Method with Contracting-Expamding Design Space Mechanism

Abstract

As a promising metamodel based design optimization (MBDO) method, the mode-pursuing sampling (MPS) method is developed to solve modern engineering design optimization problems with expensive simulations. However, the efficiency and convergence performance of the original MPS significantly decays as dimensionality of the underlying optimization problem grows. In this article, the limitation of MPS in efficiency and global exploration capability are firstly discussed. Then, in order to improve the overall performance of MPS, an enhanced mode-pursuing sampling method using contracting-expanding design space based MPS, as named CEDS-MPS, is proposed. In the proposed CEDS-MPS method, a contracting-expanding mechanism is used to intelligently identify the promising region for accelerating local exploitation and improving global exploration. Several well-known low-dimensional and high dimensional benchmark problems are applied to test CEDS-MPS through comparison with the original MPS and some other existing MBDO methods. The optimal solution, number of function evaluation and CPU time consumption are recorded for comparison. The comparison results indicate that CEDS-MPS outperforms the original MPS and other competitors in terms of local exploitation ability, global exploration capability, and efficiency performance.