节点文献

基于代理模型的异型气膜孔设计优化方法及实验验证

Surrogate Model Based Optimization of Shaped Film Cooling Hole and Experimental Validation

【作者】 张浩;

【导师】 袁新;

【作者基本信息】 清华大学 , 动力工程及工程热物理, 2019, 硕士

【摘要】 气膜冷却是燃气轮机重要的外部冷却方式,气膜孔孔型设计对气膜冷却有重要的影响。传统设计优化工作往往存在计算量过高、优化效率较低的问题,同时在设计优化过程中还需要考虑到气膜孔当地流动特征,使设计优化得到的气膜孔更适用于真实环境。本文构建了一套基于代理模型的高效气膜孔设计优化系统,分别考虑在平板无压力梯度环境下和弯通道内横向压力梯度作用下进行气膜孔设计优化。为了提升气膜孔设计优化效率,本文开发了一套基于多精度代理模型的设计优化系统。与普通代理模型相比,本文开发的多精度代理模型可以将经验关联式与传统数值计算方法进行数据融合,进而降低高准确度代理模型的构建成本,有效提升气膜孔设计优化效率。同时针对多精度代理模型,对普通拉丁超立方采样方法进行改进,使拉丁超立方采样样本空间均匀性更好,进一步提升多精度代理模型的预测准确度。研究了平板无压力梯度环境下成型孔的后倾角、侧扩角、孔长径比对气膜冷却的影响,并将这三个孔型参数作为设计变量,采用基于多精度代理模型的设计优化方法,在平板无压力梯度环境下对成型孔进行设计优化。采用PSP测量方法对优化孔型进行实验验证,证明了多精度代理模型的优化效果。本文针对Kriging代理模型,开发了基于期望提升准则的自适应采样方法,经过验证发现该采样方法可以提高代理模型在全局最优点附近的预测准确度,并在一定程度上降低代理模型的构建成本。本文将透平端壁环境抽取出弯通道的几何特征,研究了弯通道内横向压力梯度作用下的流动特征,并在弯通道内研究了成型孔复合角对气膜冷却的影响。研究发现,在壁面横流影响下大复合角反而会削弱气膜冷却效果,并基于该现象提出了横流作用下大复合角气膜孔涡结构模型。采用基于自适应采样的代理模型设计优化系统,在弯通道横向压力梯度作用下进行成型孔孔型设计优化,选取后倾角、侧扩角、复合角作为设计变量,得到最优孔型并进行分析。本文构建的基于代理模型的设计优化系统有助于提升气膜孔孔型设计优化效率,提出在设计过程中需要进一步考虑真实工况,对气膜孔设计优化具有一定指导意义。

【Abstract】 Film cooling is an important external cooling method for modern gas turbine,and the design of film cooling hole is critical to film cooling performance.Numerical opti-mization plays an important role in film cooling hole design and optimization.Not only is it important to improve the efficiency and accuracy of numerical optimization meth-ods,but also it is necessary to take the real flow environment into consideration.Hence,in this current work we have developed a high-efficient film cooling hole optimization framework,which is separately applied in shaped film cooling hole design under flat plate environment and turbine endwall environment with transverse pressure gradient.In order to improve the efficiency of film cooling hole design optimization,firstly we developed a multi-fidelity surrogate model based optimization framework.Com-pared with traditional optimization works,this current work considers data fusion method,which allows to take other preliminary design methods such as empirical correlations into consideration apart from the common-used numerical simulation.The development of multi-fidelity model helps improve the construction efficiency of high-accurate surrogate model while maintains the prediction accuracy at the similar level.In the same time,we also developed an improved latin hypercubic sampling method based oncriteria.This sampling method provides better space-filling performance than normal latin hypercu-bic sampling,and hence helps improve the prediction accuracy of multi-fidelity model to some extent.We also investigated the influences of shaped film cooling hole parame-ters,i.e.,laidback angle,lateral angle and hole length-to-diameter ratio,on film cooling performance under flat plate environment.Furthermore,the shaped film cooling hole is optimized under flat plate environment using the multi-fidelity model based optimization framework.Generic algorithm and sequential quadratic programming is used as the op-timization method to make sure the global optimum point is determined.In the end,PSP measurement and 3D printing technology are used to verify the optimized hole design and the accuracy of multi-fidelity model.In order to improve the efficiency of Kriging model based optimization,an adaptive sampling method is developed in this work.It is found that the sampling method based on expected improvement(EI)can help improve the prediction accuracy in the region near the global optimum point,and in the same time reduce the whole sampling point number.In this current work,the geometric feature in turbine endwall is extracted as a curve channel,which aims to simulate the transverse pressure gradient in turbine endwall environment.The flow features in the curve channel are carefully investigated,and it is pointed out that the most obvious flow characteristic in flow curve under transverse pressure gradient is the cross flow inside the boundary layer.An optimization campaign is further carried out using the Kriging model based optimization framework,and the adaptive sampling method based on EI criteria is used to improve the optimization accuracy and efficiency.Three key hole design parameters are chosen for the optimization,which are laidback an-gle,lateral angle and compound angle.The optimization result indicated that the optimum hole design is not obtained with largest compound angle,which is against the traditional conclusion for compound angle in flat plate environment.In order to better explain this interesting phenomenon,a physical model is further proposed to illustrate the influences of endwall cross flow to kidney vortices with large compound angle in curve channel.In conclusion,the surrogate model based optimization framework helps improve the optimization efficiency and accuracy of film cooling hole design.It is also pointed out that we should consider the realistic operation condition during film cooling hole design,which might help the optimum hole design more feasible.

  • 【网络出版投稿人】 清华大学
  • 【网络出版年期】2021年 02期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络