摘要
基于N-S方程的三维气动外形的优化设计一直以来都因为存在众多的设计变量和庞大的计算量问题而无法实现,为了解决这些问题,文中发展了一套机翼气动优化设计系统,该系统基于模拟退火粒子群优化算法、松散式代理模型以及动网格和改进的HicksHenne函数变形技术,选取翼根、翼梢各18个设计变量进行参数化。按某型飞机设计要求,在给定设计指标下进行优化设计,研究了基于松散式代理模型管理框架下,训练样本采用拉丁超立方方法对Kriging代理模型的预测能力进行评估,以及应用该代理模型框架进行了亚音速机翼的气动外形优化设计,结果显示,优化后机翼的气动特性有着显著提高。
Aim. The introduction of the full paper discusses relevant matters and then proposes developing what we believe to be a better optimal design, which is explained in sections 1 through 4. The core of section 1 is that, with the particle swarm optimiZation and simulated annealing algorithm, the moving grid method and the improved Hicks- Henne function deformation technology, we parameterize the 18 design variables of the tip of a wing and its root respectively. Section 2 briefs the sampling method with the loose type of surrogate model. To train the samples, sections 3 and 4 use the Latin Hypercube method to evaluate the prediction ability of the Kriging agent model, whose prediction precision is shown in Fig. 4. The core of section 4 is that we use the agent model to perform the optimal design of the subsonic wing of a certain typical aircraft ; the optimal design results, given in Figs. 5 through 8, and their analysis show preliminarily that our design method is effective for improving the aerodynamic characteristics of subsonic wing.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2011年第4期515-519,共5页
Journal of Northwestern Polytechnical University
关键词
参数化方法
动网格
模拟退火粒子群优化算法
代理模型
Navier-Stokes equations
particle swarm optimization
simulated annealing algorithm
agent model
moving grid method
