Sequential radial basis function optimization strategy using support vector machine for flight vehicle multidisciplinary design optimization

To improve the efficiency and quality of flight vehicle design, multidisciplinary design optimization (MDO) has been widely employed. For the purpose of reducing computational burden in MDO problems, the paper proposes a novel adaptive metamodel-based optimization strategy using support vector machine, noted as SRBF-SVM. In SRBF-SVM, large quantities of cheap points are classified by using binary SVM classifier to elect superior sample points with higher probability of locating close to the actual global optimum. Then RBF metamodel is constructed and updated based on the successively increasing sample dataset, which leads the search process to efficiently converge within limited number of evaluations of high-fidelity analysis models. Several numerical benchmark problems are used to test the efficiency and convergence of the proposed method, and the results show that SRBF-SVM is able to achieve efficient convergence to the global optimum compared to other two well-known metamodel-based optimization methods, including efficient global optimization method (EGO) and mode-pursuing sampling method (MPS). In the end, the application of SRBF-SVM in marine satellite MDO problem is investigated. The total mass of the satellite decreases by 5.43% after optimization and optimization efficiency is much higher compared with multidisciplinary feasible method (MDF), which demonstrates the practicality and effectiveness of the proposed method.