摘要
涡轮端壁受复杂流场的影响热负荷不均匀,通常需要非规则的气膜孔布局进行应对.由于多个气膜孔喷射出流的叠加效应具有极强的非线性特性,传统的气膜冷却或发散冷却经验关联式在此场景容易失效,设计中常需要针对布局专门开展实验或者数值计算.本文建立基于条件式生成对抗网络的图-图翻译模型,以表面气膜孔分布为输入特征,以全温全压工况下的冷却效率为输出特征,捕捉到气膜叠加的非线性规律.研究结果表明,模型能实现在输入任意孔的分布下快速预测.此项研究结果可大幅缩短端壁气膜冷却的设计时间,便于后续对孔的分布进行优化.
The heat load on the turbine endwall is uneven because of the complex flow field.It usually requires irregular film cooling hole layouts.Because the superposition effect of multiple film cooling jets has extremely strong nonlinear characteristics,the empirical correlations of traditional film cooling or transpiration cooling are ineffective.The design of layouts often requires experiments or numerical calculations.In this paper,the image-to-image translation based on conditional generative adversarial networks is used,with the surface film cooling hole distribution as the input,and the cooling efficiency under full temperature and pressure conditions as the output,to capture the nonlinear law of film superposition.The results show that the model can predict rapidly given the layouts of arbitrarily distributed holes.The method of this paper can greatly reduce the time of design of film cooling on the endwall and facilitate the subsequent optimization of the film cooling hole layouts.
作者
戴维
杨力
饶宇
DAI Wei;YANG Li;RAO Yu(Institute of Turbomachinery,School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《工程热物理学报》
EI
CAS
CSCD
北大核心
2020年第10期2420-2424,共5页
Journal of Engineering Thermophysics
基金
国家自然科学基金资助项目(No.51906139)
上海市青年科技英才扬帆计划项目(No.19YF1423200)。
关键词
气膜冷却
涡轮端壁
生成对抗网络
深度学习
film cooling
turbine endwall
generative adversarial networks
deep learning
