椭圆槽气膜冷却结构的优化研究

Optimization Research of Film Cooling Structureswith Cratered Cylindrical Hole

关键词:  优化  支持向量机  代理模型  遗传算法

Key words:  Optimization  Support vector machine  Surrogate model  Genetic algorithm

摘要:

针对椭圆槽气膜冷却结构展开优化研究，优化参数为椭圆长轴长度[L1]、椭圆短轴长度[L2]和槽深h，优化目标为最大化气膜绝热冷却效率。首先对椭圆槽气膜冷却物理模型进行计算流体力学方法求解，获得一定容量的数据样本；并基于数据样本训练支持向量机参数，建立优化代理模型；最后引入遗传算法在优化区间内进行寻优，获得最佳的椭圆槽结构。在吹风比为0.6的工况下，优化后的[L1]，[L2]和h分别为2.22D，2.65D和0.74D(D为气膜孔径)，冷却效率较参考结构提升了42%；在吹风比为1.2的工况下，优化后的[L1]，[L2]和h分别为3.60D，2.70D和0.67D，冷却效率较参考结构提升了73%。通过优化，气膜孔下游截面肾型涡对得到有效抑制，而反肾型涡对则被强化，气膜展向覆盖能力明显增强。优化结果表明了支持向量机代理模型和遗传算法在气膜冷却结构优化的有效性。

Abstract:

In order to optimize the film cooling structure of the cratered cylindrical hole， an optimization study was conducted. Some influence factors， such as the length of the major axis of the ellipse [L1]， the length of the minor axis [L2] and the depth of the cratered h， are taken into considerations. The optimization objective is to maximize the adiabatic cooling efficiency of the film. Firstly， the physical model of film cooling in cratered cylindrical hole was solved by CFD simulation. A certain capacity data sample was obtained. Then， based on the data samples， support vector machine (SVM) parameters are trained， and the optimal agent model can be established. Finally， genetic algorithm is introduced to optimize the interval and get the best cratered cylindrical hole structure. Under the condition of blowing ratio of 0.6， the optimized [L1]， [L2] and h are 2.22D， 2.65D and 0.74D， respectively (D is the diameter of the film). The cooling efficiency increased by 42% compared with the reference model. Under the condition of blowing ratio of 1.2， the optimized [L1]， [L2] and h are 3.60D， 2.70D and 0.67D， respectively. And the cooling efficiency is improved by 73% compared with the reference model. After optimization， the kidney vortex at the downstream of the film hole was effectively suppressed， while the anti-kidney vortex was strengthened， and the coverage of the film was significantly enhanced. The results show the feasibility of SVM agent model and genetic algorithm in the film cooling structure optimization.