序列近似优化方法及其在火箭外形快速设计中的应用

针对序列近似优化方法在代理模型构造与采样策略等方面的不足,基于采样点局部密度,引入与局部密度成反比的样本点影响体积概念,通过优化总影响体积确定径向基函数最优核宽度,满足序列近似优化过程不同规模、非均匀样本条件下的函数近似需要;建立潜在可行域最大距离加点准则,并与潜在最优加点准则结合平衡算法的探索能力与开发能力;建立三步收敛判定准则;构建改进序列近似优化的算法流程。对于Golinski减速器的优化设计问题,算法在目标函数调用42次后便搜索到全局最优解,体现了其良好的全局寻优能力与搜索效率。以"天航二号"火箭为例,建立其外形优化问题数学模型,所提优化方法在调用原始计算模型165次之后便搜索到全局最优解,大大提高了设计效率,同时飞行试验也表明设计结果满足要求。

Sequential Approximate Optimization Study and Application in Rocket Shape Rapid Design

Sequential approximate optimization method has shortcomings in two respects as surrogate models establishing and infill strategy at present. Based on local density of sampling points, optimal kernel width of radial basis function (RBF) is obtained by means of total influence volume optimization. Potential feasible region infill strategy was proposed, and potential optimal strategy was applied together, both exploration and exploitation capacity of sequential approximate optimization algorithm were satisfied. Convergence criterion was seted up. Algorithm flow process was constructed. For Golinski reducer optimization problem, global optimal solution was solved after calculating original model 42 times. Shape optimization mathematical model was established for TH-II rocket (TianHang), global optimal shape was gained after 165 times original model calling using proposed method. TH-II rocket aerodynamic shape was proved reliable by flight testing.