Novel CFD High Performance Methods for Advanced Simulations in Turbines

用于涡轮机高级仿真的新型 CFD 高性能方法

• 批准号: 2306417
• 金额: --
• 依托单位: University of Oxford
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2306417
• 关键词: Novel; CFD; Performance; Methods; Advanced

Modern gas turbines rely on cooling techniques to operate at gas temperature well in excess of the melting temperature of the metal parts. Cooling is performed by ejecting cold air through small orifices on the blade surfaces. This gives rise to complex flow patterns that are difficult to predict with traditional methods.Recent advances in computer architecture bring scale-resolving simulations of cooling flow closer to common practice. The computer features that are needed are very specialised and difficult to exploit whilst preserving the flexibility and accessibility of the software to general developers.Therefore, there is work to be done on the high-performance computing (HPC) aspects of the simulations in order to analyse physical questions such as the interaction between the passage flow and the cooling flow.Research questionHow and how much by can we improve the speed of CFD codes for gas-turbine applications by exploiting all the features of modern hardware?Can we design the code in a way that is completely opaque to the hardware, so that the same source can make the most of both x8086 and GPU based computers?How much of the error comes from modelling turbulence and flow interactions?ApproachStudy the effect of memory mapping layouts on the execution speed of CFD calculations on x8086 and GPU based computers.Study techniques to enable fine-grained and instruction level parallelism within an industrial software.Devise novel layouts and code generation techniques based on sub-cell resolution.Perform scale-resolving simulations of turbine blades with cooling devices.Novelty/Physical sciencesHPC techniques and sub-cell resolution techniques for scale resolving simulations in gas turbines cooling systemsThis will allow us to answer questions about the interaction between the cooling flow and the passage flow and improve the prediction of thermal paint tests.

现代燃气轮机依靠冷却技术在远超过金属部件熔化温度的气体温度下运行。冷却是通过叶片表面上的小孔喷射冷空气来进行的。这导致了传统方法难以预测的复杂流动模式。计算机体系结构的最新进展使冷却流的尺度分辨模拟更接近于常见实践。所需的计算机功能非常专业，难以开发，同时保留了软件的灵活性和一般开发人员的可访问性。因此，在高性能计算（HPC）方面还有很多工作要做为了分析物理问题，如通道流和冷却流之间的相互作用，研究问题如何以及如何提高速度，利用现代硬件的所有功能，为燃气轮机应用提供CFD代码？我们能否以一种对硬件完全不透明的方式设计代码，以便同一个源代码可以充分利用基于x8086和GPU的计算机？有多少误差来自湍流和流动相互作用的建模？方法研究内存映射布局对基于x8086和GPU的计算机上CFD计算执行速度的影响。研究在工业软件中实现细粒度和指令级并行的技术。设计基于子单元分辨率的新颖布局和代码生成技术。执行带冷却设备的涡轮机叶片的尺度解析模拟。新奇/物理科学HPC技术和子单元燃气轮机冷却系统中尺度解析模拟的单元解析技术这将使我们能够回答有关冷却流和通道流之间相互作用的问题，并改进热漆试验的预测。