Sequential Hybrid Approximate Optimization Approach and Its Applications in Structural Design

Abstract

This paper presents a sequential hybrid approximate optimization (SHAO) algorithm suitable for structural design optimizations. A hybrid approximate model is introduced and further employed in predicting structural analyses more accurately while also requiring significantly fewer training samples. Furthermore, an adaptive sampling strategy is utilized to create a balance between its ability to locate the global optimum and computational efficiency within the optimization process. Consequently, the optimal searching efficiency of the SHAO algorithm is substantially enhanced. Efficiency and reliability of the proposed method are demonstrated through several benchmark structural design cases. Numerical results herein obtained reveal the proposed SHAO becomes more efficient when compared to conventional SAO and most existing meta-heuristic methods in terms of quality of solution, computational cost and convergence rate.