AUTHOR=Park So-Hyun , Jo Young Beom , Ahn Yelyn , Choi Hae Yoon , Choi Tae Soo , Park Su-San , Yoo Hee Sang , Kim Jin Woo , Kim Eung Soo 

TITLE=Development of Multi-GPU–Based Smoothed Particle Hydrodynamics Code for Nuclear Thermal Hydraulics and Safety: Potential and Challenges

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 8 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2020.00086

DOI=10.3389/fenrg.2020.00086

ISSN=2296-598X

ABSTRACT=Advanced modeling and analysis techniques are always essential for development of safe and reliable nuclear systems. The numerical methods in nuclear thermal-hydraulics and safety analysis are generally based on mesh-based (or grid-based) methods. These mesh-based methods have a long history and thus are very mature both mathematically and numerically. Based on their robustness and efficiency, they have been successfully applied to many applications for decades. However, the mesh-based methods suffer from difficulties in handling complex phenomena accompanied by highly non-linear deformations, which are frequently encountered in the recent nuclear safety issues related to natural disaster and severe accidents. Recent advances in Lagrangian-based CFD techniques have opened up possibilities for addressing this issue effectively. In this study, a Lagrangian-based CFD code (named as SOPHIA) has been developed based on Smoothed Particle Hydrodynamics (SPH), one of the best-known Lagrangian methods which can extensively handle various physics because of its simplicity in expressing and solving mathematical equations. The SOPHIA code incorporates basic conservation equations (mass, momentum, and energy) and various physical models including heat transfer, turbulence, multi-phase, surface tension, diffusion, etc. In addition, to handle multi-phase, multi-component, and multi-resolution flows, the code newly formulates density and continuity equations in terms of a normalized-density. One of the biggest technical challenges in the Largrangian-based CFD is its high computational cost. Therefore, the code is parallelized using multi-GPUs through multi-parallel computing techniques. Through the optimized algorithm based on these techniques, the computational performance has been improved drastically and the code obtains the capability of high resolution and large scale simulation. In order to demonstrate its applicability to nuclear safety-related issues, this study performs the simulations on three benchmark experiments related to nuclear safety: (1) water jet breakup of FCI, (2) LMR core sloshing, and (3) bubble lift force. The simulation results are compared with the experimental data both qualitatively and quantitatively, and they shows a good agreement with highly reliable visualizations.