Development of advanced nodal diffusion methods for modern computer architectures
Development of advanced nodal diffusion methods for modern computer architectures
复制标题
为现代计算机体系结构开发先进的节点扩散方法
DOI：
--
复制
发表时间：
1988
期刊：
影响因子：
--
通讯作者：
H. Rajic
中科院分区：
文献类型：
--
作者：
H. Rajic
一系列高效的多维多群高级中子扩散节点法，已在顺序、向量和向量并行计算机上实现。在CRAY X-MP/48上，三维真实感基准问题可以在低于0.73Mflops(33.86Mflops)的情况下以矢量化方式求解。向量并行实现在Alliant FX/8上的加速比高达9.19。这些结果表明，illico方法基本上保持了它与有限差分方法相比的所有速度优势(以前在标量计算机上演示过)。
发展并实现了一种自洽的高阶节点扩散方法。结果表明，该方法的精度(最大装配功率误差为0.02%)与使用装配尺寸粗网格的超细网格有限差分方法的精度相当。将该方法应用于横向泄漏近似的数值研究。
已发展和评估了用于全球核反应堆多群扩散计算的节点法，这些方法明确地考虑了组件核性质的不均一性。对零阶变截面节点法进行了系统的分析。对KWU压水堆耗竭基准问题的分析表明，当燃耗非均质性出现时，没有显式处理非均质性的普通节点法累积了显著的系统误差。已经认识到，使用节点内空间可变的材料属性与当前的均化方法不兼容。因此，设计了一种生成广义空间可变节点横截面的方法。
发展并实现了一种显式处理扩散系数的空间相关性的节点法。在非均质性很强的问题中，它被证明是有效的。
发展了一种用于节点微观耗竭分析的一致性燃耗校正方法。新方法使每个组件一个节点的耗尽计算变得极其准确和计算效率极高。节点耗竭法是以KWU 2周期压水堆耗竭问题为基准的。这一结果与用昂贵的商用耗尽代码如VERVE、SIMULATE-3和NEM-BC获得的结果进行了很好的比较。
A family of highly efficient multidimensional multigroup advanced neutron diffusion nodal methods, ILLICO, have been implemented on sequential, vector, and vector-concurrent computers. Three dimensional realistic benchmark problems can be solved in vectorized mode in less than 0.73 s (33.86 Mflops) on a Cray X-MP/48. Vector-concurrent implementations yield speedups as high as 9.19 on an Alliant FX/8. These results show that the ILLICO method preserves essentially all of its speed advantage (previously demonstrated on scalar computers) over finite difference methods.
A self-consistent higher order nodal diffusion method has been developed and implemented. The method is shown to yield results nearly as accurate (0.02% maximum relative assembly power errors) as those of very fine mesh finite difference methods, using an assembly size coarse mesh. The method was applied to a numerical study of the transverse leakage approximation.
Nodal methods for global nuclear reactor multigroup diffusion calculations which account explicitly for heterogeneities in the assembly nuclear properties have been developed and evaluated. A systematic analysis of the zero order variable cross section nodal method has been conducted. Analyzing the KWU PWR depletion benchmark problem, it is shown that when burnup heterogeneities arise, ordinary nodal methods, which do not explicitly treat the heterogeneities, suffer a significant systematic error that accumulates. It has been recognized that the use of in-node spatially variable material properties is incompatible with the current homogenization methods. A procedure for generating generalized spatially variable nodal cross sections has therefore been devised.
A nodal method which treats explicitly the space dependence of diffusion coefficients has been developed and implemented. It is shown to be effective in problems where very strong heterogeneities occur.
A consistent burnup correction method for nodal microscopic depletion analysis has been developed. The new method makes extremely accurate and computationally highly efficient one node per assembly depletion computations possible. The nodal depletion method is benchmarked on the KWU 2-cycle PWR depletion problem. The results compare very well to the published results obtained with expensive commercially available depletion codes like VENTURE, SIMULATE-3, and NEM-BC.