Benchmark Case Study
RVACS Benchmark: DNS and RANS Analysis
A benchmark study on buoyancy-driven natural circulation in a representative Reactor Vessel Auxiliary Cooling System (RVACS) geometry, comparing high-fidelity DNS reference data with a wide range of practical RANS simulations.
Overview
The Reactor Vessel Auxiliary Cooling System (RVACS) is a passive decay heat removal concept relevant to several advanced reactor systems. This benchmark considers a representative RVACS closure operating under strong buoyancy, where natural circulation is driven by a large temperature difference between the inner and outer walls.
The case focuses on a transitional natural convection flow at Ra = 7.14 × 107, providing a challenging test for both high-fidelity and engineering-scale CFD methods.
Objectives
- Understand the fundamental physics of buoyancy-driven natural recirculation in a representative RVACS geometry.
- Provide high-quality benchmark reference data from DNS.
- Assess the performance of practical RANS modelling approaches against this challenging flow problem.
- Support both scientific understanding and industrial modelling needs in advanced reactor thermal hydraulics.
Benchmark Configuration
The benchmark represents natural convection within an RVACS-like enclosure separated by a baffle plate, with the hot and cold sides connected only in the lower region. This produces a more complex flow than a simple cavity and introduces recirculation interaction, separation near the baffle tip, and transition to turbulence.
- Flow regime: buoyancy-driven natural convection, transition regime
- Rayleigh number: 7.14 × 107
- Thermal boundary conditions: convective conditions on the hot and cold external walls
- Hot-side external condition: 700 K, 200 W/(m²·K)
- Cold-side external condition: 300 K, 200 W/(m²·K)
- Baffle material: SS310 stainless steel with conjugate heat transfer included
- DNS geometry: three-dimensional with periodic spanwise boundary conditions
- RANS geometry: two-dimensional transient simulations
Simulation Methods
The benchmark combines a high-fidelity DNS reference with a broad comparison of engineering RANS approaches contributed by multiple participants.
DNS Reference
DNS was carried out using the low-Mach-number Nek5000 solver based on the spectral element method, with temperature-dependent variable thermophysical properties.
RANS Comparison Set
Transient two-dimensional RANS simulations were contributed using different solvers, meshes, wall treatments, and turbulence models.
- Solvers: Code_Saturne, Fluent, FEAT
- Model families: standard k–ε, Launder-Sharma k–ε, k–ω SST, Reynolds stress transport models, Billard-Laurence k–ε–v²/k, and quadratic Baglietto-Ninokata k–ε
- Wall treatment strategies: wall functions, all-y+ approaches, and wall-resolved treatments
Participants
This benchmark was jointly initiated by CCP-NTH, Frazer-Nash Consultancy Ltd., and Argonne National Laboratory, with contributions from academic and industrial partners.
- University of Sheffield
- University of Manchester
- Science and Technology Facilities Council (STFC)
- Frazer-Nash Consultancy Ltd.
- Argonne National Laboratory
- Pennsylvania State University
- EDF Energy Research & Development UK
- EDF Energy Nuclear Services
Key Findings
The DNS revealed two interacting natural recirculation loops: one in the hot-side channel and a second spanning the lower cavity and cold-side channel. The onset of transition and strong turbulence production were associated with the lower baffle-tip region, where loop interaction and flow separation occur.
Across the benchmark comparisons, most RANS simulations reproduced the broad mean-velocity structure reasonably well at the probe locations examined, except cases using the scalable wall function, which showed clear deviations.
In contrast, turbulent kinetic energy was much harder to predict. The DNS showed strong peaks and more complex turbulence structure near the hot wall and around the baffle-tip interaction region, while the RANS simulations generally underpredicted these features, especially at the most challenging location where transition and separation occur.
Why this Benchmark Matters
- Provides a high-value reference case for buoyancy-driven natural circulation in advanced reactor safety analysis.
- Captures multiple challenging phenomena in one geometry: buoyancy, transition, recirculation, separation, and conjugate heat transfer.
- Offers benchmark-quality DNS data for validating lower-cost industrial modelling approaches.
- Helps identify where common RANS approaches are reliable and where they remain challenged.
- Creates a foundation for future model development, comparison studies, and AI-assisted thermal hydraulics workflows.
Reference
He, J., He, S., Macpherson, G., Shaver, D., Merzari, E., Wang, W., Liu, B., Cartland-Glover, G., Lim, O., Houghton, T., Katsamis, C., Sigournay, C., and Kyritsopoulos, I. On the CFD Modelling of Natural Convection in RVACS: Analysis of DNS and RANS Simulations, Turbulence, Heat and Mass Transfer 11, 2025.
