Session: 3.4 - CFD for Nuclear Thermal Hydraulics

Paper Number: 158076

158076 - Large Eddy Simulation of TALL-3D Facility Using Spectral Element Method 

Abstract: 

The pool-type design of advanced metal-cooled reactors offers several significant benefits, including reduced risk of coolant leaks, simplified structural design, and enhanced passive safety. In particular, the primary coolant is contained within a large pool, which minimizes the potential for leaks and provides an extra barrier against sodium leakage. This configuration allows for a more compact primary system, reducing the overall size of the reactor while also enhancing safety through substantial thermal inertia; the large volume of coolant acts as an immense heat sink, enabling passive decay heat removal via natural circulation even in the event of system failures. The TALL-3D liquid-metal facility was introduced to support these advancements as part of the SESAME project (Simulations and Experiments for the Safety Assessment of MEtal-cooled reactors). This experimental facility, utilizing Lead-Bismuth Eutectic (LBE) coolant, features a large, pool-type enclosure that exhibits complex 3D flow effects and serves to advance the European transmutation demonstration for accelerator-driven systems. In this study, Large Eddy Simulation (LES) is performed using the GPU-accelerated spectral element code NekRS to investigate the forced, transient, and natural convection operational conditions of the TALL-3D facility. The Neknek framework in NekRS is employed to couple multiple subdomains with distinct mesh configurations, enabling interaction through boundary conditions. Neknek simulations usually involve multiple MPI ranks, with each session assigned a unique communicator, allowing for independent fluid-thermal simulations. These sessions can vary in polynomial orders and physical models, providing flexibility in simulation configurations. In this work, Neknek solves the high-flow and low-flow sessions. The low-flow region is assigned a mesh with significantly fewer Gauss-Lobatto-Legendre (GLL) points compared to the high-flow region, which improves computational efficiency and optimizes resource distribution. This is very important for high-fidelity simulations like LES. The Neknek simulation results, including mass flow rate and temperature time-series data, were compared with experimental measurements, demonstrating good agreement across all operational conditions of interest. Additionally, Reduced-Order Modeling (ROM) techniques were applied to the simulation data using Proper Orthogonal Decomposition, which extracts the flow field's most important features or patterns from large datasets of time-dependent LES flow solutions. ROM offers significant advantages over Full-Order Modeling (FOM). In FOM, the number of unknowns to solve at each time step grows with the discretization of the domain. Each additional mesh element introduces variables that must be solved simultaneously, leading to a significant increase in computational demands, particularly for high-fidelity simulations. By utilizing NekROM, the dominant flow patterns under the operational conditions of interest are efficiently identified and characterized, providing valuable insights into the underlying flow dynamics while significantly reducing the computational cost. 

Presenting Author: Tri Nguyen Penn State University

Presenting Author Biography: