High performance and high fidelity deterministic neutron transport methods for nuclear reactor lattice physics simulations

用于核反应堆晶格物理模拟的高性能和高保真度确定性中子输运方法

• 批准号: 2764533
• 金额: --
• 依托单位: Imperial College London
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2764533
• 关键词: performance; fidelity; deterministic; neutron; transport

The aim of this PhD project is to develop novel, self-adaptive. numerical algorithms on modern, multi-core and many-core, high performance distributed computing (HPC) hardware architectures that enable high-fidelity spatial and energy resonance self-shielding of large-scale nuclear reactor lattice physics simulations. The numerical algorithms will use the latest shared and distributed memory algorithms on multicore hardware architectures that have been augmented by manycore (GPU) hardware acceleration. The spatial and energy resonance self-shielding algorithms will comprise a suite of different resonance self-shielding algorithms ranging from equivalence-in-dilution methods (EM), sub-group methods (SGM), embedded self-shielding methods (ESSM), finite element discontiguous supportmethods (FEDS) and ultrafine self-shielding methods (USSM). This will enable a range of fidelities of spatial and energy self-shielding to be performed so that the relevant computational efficiencies and accuracies of the methods can be assessed. It will also enable methods that can investigate resonance inference effects (or mutual self-shielding) between different nuclides within host materials within the nuclear reactor core. The eventual aim will be to develop numerical algorithms that can perform spatial and energy self-shielding for multidimensional single and multiple nuclear fuel assemblies (so called colour-set nuclear reactor lattice physics calculations) as well as potentially compact nuclear reactor cores. The spatial discretisation will use exact geometry conforming Non-Uniform Rational B-Spline (NURBS) enhanced methods for resolving the spatial variation in the neutron distribution which will complement the high-fidelity energy resonance self-shielding methods. Space-energy adaptivity as well as hierarchical numerical solution algorithms in energy will also be exploited to improve the computational efficiency of the space and energy resonance self-shielding algorithms.

这个博士项目的目的是开发新颖的，自适应的。在现代多核和众核、高性能分布式计算（HPC）硬件架构上的数值算法，能够实现大规模核反应堆晶格物理模拟的高保真空间和能量共振自屏蔽。数值算法将在多核硬件架构上使用最新的共享和分布式内存算法，这些硬件架构已通过众核（GPU）硬件加速进行了增强。空间和能量共振自屏蔽算法将包括一套不同的共振自屏蔽算法，包括稀释等效法（EM）、子群法（SGM）、嵌入式自屏蔽法（ESSM）、有限元不连续支撑法（FEDS）和超细自屏蔽法（USSM）。这将使一系列的空间和能量自屏蔽的准确性进行，以便有关的计算效率和准确性的方法可以进行评估。它还将使能够研究核反应堆堆芯内主体材料内不同核素之间的共振推断效应（或相互自屏蔽）的方法成为可能。最终的目标是开发数值算法，可以执行空间和能量自屏蔽的多维单一和多个核燃料组件（所谓的颜色设置核反应堆晶格物理计算），以及潜在的紧凑型核反应堆堆芯。空间离散化将使用符合非均匀有理B样条（NURBS）增强方法的精确几何形状来解决中子分布的空间变化，这将补充高保真能量共振自屏蔽方法。还将利用空间能量自适应性以及能量方面的分层数值求解算法来提高空间和能量共振自屏蔽算法的计算效率。