Aimed at solving the problem of optimization convergence and the global optimization, this research proposes a Combined Optimization Methodology, which combines the global and local optimization methods. The former one can obtain global optimum solution but converge slowly, while the latter one can converge fast and obtain a local optimum solution and can be sensitive to initial value. Firstly, the approximation function of the original problem was built. Then the approximate optimum solution was obtained by the global optimization method. The approximate optimum was taken as the initial value, and the real optimum solution is obtained by the local method optimizing the original problem directly. In order to achieve better approximation, the RBF was improved and Shape Parameter Optimization Radius Basis Function was developed by using the surrogate model. Both of the methods were used in the near space aircraft wing optimization. Results show that the modified interpolation method is more accurate than the Kriging model, and the Combined Optimization can obtain the optimum and improve the convergence speed.%针对优化中收敛速度和优化解全局性的问题,提出了一种联合优化方法:构造原问题的近似模型,使用全局优化方法对近似函数进行优化,得到优化点作为初值,再使用局部优化方法对原问题进行优化.为了获得对原问题更好的近似,改进了径向基插值方法,以优化误差的方法来选择参数.利用临近空间机翼模型的优化对算法进行了测试,结果表明,优化参数的径向基插值方法提高了高维问题的近似能力,联合优化能够得到较好的优解,并提高了收敛速度.
展开▼
