Composite structure optimization for satellite using discrete dynamic radial basis function metamodel

To improve the efficiency of satellite composite structure optimization (SCSO), this paper proposes a novel trust region based dynamic radial basis function method for discrete variable optimization problems, denoted as DTR-DRBF. In DTR-DRBF, the samples are mapped to unique integer samples for constructing RBF metamodel, and then integer genetic algorithm (GA) is employed to optimize the RBF metamodel to obtain the potential optimum of the real optimization problem. Finally, the obtained optimum is mapped back to the true discrete potential optimum. Moreover, a sequential maximin LHD sampling scheme is utilized to enhance the global exploration capability of DTR-DRBF in a promising region identified by trust region method. The proposed DTR-DRBF is tested with a benchmark problem compared with several well-known discrete variable optimization methods to demonstrate the merits of DTR-DRBF in terms of convergence, efficiency and robustness performances. In the end, DTR-DRBF is applied in a real-world satellite composite structure optimization problem. In the optimization problem, a finite element model of the satellite composite structure is constructed first. Then the thicknesses of the composite material are optimized by DTR-DRBF to minimize the satellite mass considering the modal frequency constraints. The results show that the mass of the optimized satellite decreases significantly with the modal frequency constraints being satisfied, which illustrates the effectiveness and practicality of DTR-DRBF in solving engineering satellite composite structure optimization problems.