Sequential Radial Basis Function Using Support Vector Machine for Expensive Design Optimization
Renhe Shi,Li Liu,Teng Long,Jian Liu
DOI: https://doi.org/10.2514/1.j054832
IF: 2.624
2017-01-01
AIAA Journal
Abstract:To improve the efficiency and quality of solving expensive engineering design optimization problems, recently metamodel-based design and optimization technologies have been widely employed. In this paper, a novel adaptive metamodel-based optimization method called sequential radial basis function using support vector machine is proposed to solve the practical engineering optimization problems involving computationally expensive objective and constraints. The proposed metamodel-based optimization method uses the radial basis function metamodel to approximate the expensive simulations in the interesting sampling region, which is a subregion of the design space. To identify the interesting sampling region, a novel feature space fuzzy c-means clustering method is proposed to obtain the cluster center of superior cheap points classified by support vector machine. The procedure of sequential radial basis function using support vector machine is presented first, and the interesting sampling region and feature space fuzzy c-means clustering methods are discussed in detail. Several numerical benchmark problems are used to compare the proposed method with other well-known metamodel-based design and optimization methods including efficient global optimization, multiple surrogate efficient global optimization methodology, mode-pursuing sampling, constraint importance mode pursuing sampling, and constrained optimization by radial basis function interpolation. The comparison results show that the proposed method generally outperforms the competitive methods in terms of efficiency, convergence, and robustness. Finally, sequential radial basis function using support vector machine is applied in an Earth observation satellite multidisciplinary design optimization problem. The results illustrate that compared with genetic algorithm and sequential quadratic programming, the proposed method can find a better solution with fewer function evaluations, which demonstrates the practicality and effectiveness of the proposed method in solving real-world engineering design optimization problems.