用于气膜冷却的机器学习湍流模型
Machine-Learning-Based Turbulence Model for Film Cooling
张振 1叶林 1苏欣荣 1袁新1
作者信息
- 1. 清华大学能源与动力工程系,北京 100084
- 折叠
摘要
在气膜冷却中,主流和射流的相互作用会产生三维涡系结构和湍流掺混,它们引起了温度的对流和湍流扩散输运.由于上述现象的三维特性,传统的湍流模型无法准确预测气膜冷却效果,限制了燃气轮机冷却结构的分析与设计水平.本文采用湍流模型反演机器学习框架,基于流场关键位置的温度数据,对湍流普朗特数进行了具有空间分布的修正.所发展的机器学习湍流温度扩散模型可以显著降低冷却有效度的预测误差,且具有较好的泛化能力.
Abstract
In film cooling,the interaction of the main flow and the jet produces a three-dimensional vortex structure and turbulent mixing,which cause convective and turbulent diffusion transport of energy.Due to the three-dimensional nature of the above phenomenon,the traditional turbulence model cannot accurately predict the film cooling effect,which limits the analysis and design level of the gas turbine cooling structure.In this paper,the inversion machine learning framework for turbulence models is used to correct the turbulent Prandtl number with a spatial distribution based on the temperature data at key positions in the flow field.The developed machine learning turbulent temperature diffusion model can significantly reduce the prediction error of cooling effectiveness and has good generalization ability.
关键词
气膜冷却/湍流模型/机器学习Key words
film cooling/turbulence modeling,machine learning引用本文复制引用
基金项目
国家科技重大专项(J2019-Ⅲ-0007-0050)
出版年
2024
NETL
NSTL
万方数据