Performance Analysis of Nek5000 for Single-Assembly Calculations

Author:

Merzari Elia1,Rahaman Ronald1,Min Misun1,Fischer Paul1

Affiliation:

1. Argonne National Laboratory, Lemont, IL

Abstract

The ExasSMR project focuses on the exascale application of single and coupled Monte Carlo (MC) and computational fluid dynamics (CFD) physics. Work is based on the Shift MC depletion, OpenMC temperature-dependent MC, and Nek5000 CFD codes. The application development objective is to optimize these applications for exascale execution of full-core simulations and to modularize and integrate them into a common framework for coupled and individual execution. Given the sheer scale of nuclear systems, the main algorithmic driver on the CFD side is weak scaling. The focus for the first four years of the project is on demonstrating scaling up to a full reactor core for high-fidelity simulations of turbulence. Full-core fluid calculations aimed at better predicting the steady-state performance will be conducted with a hybrid approach in which large eddy simulation is used to simulate a portion of a core and unsteady Reynolds-averaged Navier-Stokes handles the rest. This zonal hybrid approach provides an additional scaling dimension besides the number of assemblies. The present manuscript focuses on performance assessment using assembly-level simulations with Nek5000. We discuss the development of two benchmark problems: a subchannel (single-rod) problem to assess internode performance and a larger full-assembly problem representative of a small modular reactor (SMR). We note that current SMR assemblies are considerably simpler than pressurized water reactor assemblies since they contain no mixing vanes. This feature allows for considerable reduction in the degrees of freedom required to simulate the full core. We discuss profiling and scaling results with Nek5000, describe current bottlenecks and potential limitations of the approach, and suggest optimizations for future investigation.

Publisher

American Society of Mechanical Engineers

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3