Conference Agenda

Nuclear thermal hydraulic analyses generally employ system and subchannel codes to examine the safety and performance characteristics of a given reactor system. More recently, 3D CFD modelling has been widely used to produce higher fidelity analysis but such methods can mainly be used for local phenomena. This research seeks to develop a coarse-grid CFD model using a subchannel approach to compute wall effects so as to create a computationally cost-effective model for the two-phase flow dynamics within the nuclear reactor core. The subchannel CFD (SubChCFD) technique previously developed for single-phase flow at the University of Sheffield is first implemented in OpenFOAM^TM CFD solver. The 5X5 bare rod bundle case from the NESTOR experiment is used for validation. Model predictions are compared well against experimental data. This single phase model has then been extended to the homogeneous equilibrium two-phase flow model concept and models for the wall effects (including wall boiling) and phase crossflows are being developed. In the full paper, the OpenFOAM^TM implementation and validation for both single and two-phase boiling flows will be discussed and evaluated.

The presence of spacer grids and mixing vanes in a typical Pressurized Water Reactor Fuel Assembly leads to significant turbulence in the coolant sub-channels, which plays a very important role to enhance the heat transfer performance of the fuel assembly. In this preliminary work, OpenFOAM is used to build a single coolant channel model with fuel rods, spacer grid, and mixing vanes, with periodic boundary conditions, to study the complex fluid flow behaviour induced by the mixing vanes and spacer grid. The flow field results obtained from OpenFOAM with the Reynolds Averaged Navier-Stokes (RANS) turbulence models are found to match well with experimental results from the MATIS-H benchmark. In addition, multiple hyperparameters are varied to study their effects on computational resources required and stability, and accuracy of results.

The CFD task group of the OECD Nuclear Energy Agency (NEA) Working Group on Analysis and Management of Accidents (WGAMA) drives collaborative activities in Computational Fluid Dynamics for Nuclear safety studies since the early 2000s. This paper presents recent achievements and the current major objectives of the group, illustrated by examples of code benchmarks and reviews of reference documents such as the updated Best Practice Guidelines.

A recently published Technical Opinion Paper identifies main challenges towards the perspective use of CFD in safety studies. Several ongoing and future activities identified and initiated will be discussed. In the qualification process of scientific computational tools , the quantification of uncertainties (UQ) plays a major role in the estimation of the trust of a given evaluation. As far as CFD is concerned, several specificities, mostly induced by the computational cost have been recently analyzed within an expert group composed of CFD, UQ and data science specialists. This paper shares the outcomes and potential future activities on the topic.

In the context of the qualification process, identification of suitable application-oriented validation data is an important step. WGAMA is currently considering an important task concerning the update and extension of the CSNI Code Validation Matrix (CCVM) for reactor coolant system and containment thermal-hydraulic phenomena of current and advanced water-cooled reactor including SMR. In the field of CFD, a connected activity is foreseen on the identification and analysis of validation data for specific safety applications. The needs and way forward will be discussed within this paper.

To meet the requirements of three-dimensional numerical simulations for the flow
and heat transfer phenomena in nuclear reactors, this article introduces WINGS-CFD （Workbench of Intelligent Nuclear reactor desiGn and Simulation-Computational Fluid Dynamics）, an Computational Fluid Dynamics (CFD) code for nuclear reactors developed by Nuclear Power Institute of China, which is designed based on the object-oriented and hierarchical architecture principles with a high degree of extension. This article comprehensively outlines the overall design concepts of WINGS-CFD, covering aspects such as theoretical models, numerical discretization methods, and code architecture. Several numerical calculations for typical reactor scenarios involving flow and heat transfer conditions are conducted using WINGS-CFD and the commercial CFD software Fluent, and the results demonstrate that the computational accuracy of both codes is comparable. Furthermore, WINGS-CFD is capable of supporting large-scale numerical simulations of neutron transport and flow heat transfer with billions of cells and it exhibits superior parallel performance. This WINGS-CFD code provides an autonomous numerical technique for the fluid dynamics and heat transfer analysis of reactor systems.

In the present work, a novel methodology for the numerical simulation of single-phase flow in rod bundles, within a coarse-mesh context, is presented. The proposed approach aims to fill the gap between standard CFD and subchannel modeling, targeting a stronger balance between the accuracy of the numerical solutions and a reduced computational effort. For this purpose, different numerical techniques were designed and combined in custom applications within the OpenFOAM framework. In the first place, wall models for the turbulent quantities that rely on the use of empirical correlations were implemented. The use of these models reduces the traditional restrictions on the mesh refinement at the walls and consequently introduces significant gains in computational efficiency for a similar level of accuracy when compared to standard simulations. On the other hand, special numerical schemes were implemented to include the capability of handling piecewise linear pressure distributions in a finite volume context, i.e., discrete changes in pressure across selected locations that mimic the behavior of unresolved, localized geometrical scales. This approximation is used to model spacer grid effects without explicitly including the spacer grid in the geometry. Additionally, new numerical techniques that allow the use of different meshes for each equation are explored. The combination of the proposed coarse-mesh approximations is validated against high-resolution numerical simulations and experimental data.

With the advancement of nuclear technology, the demand for precise simulation of flow fields in reactor rod bundles has significantly increased. Traditional low-order eddy viscosity models often lack the accuracy needed for complex subchannel flow fields, while high-order numerical methods like Large Eddy Simulation (LES) provide high-fidelity flow structures but are computationally expensive, limiting their practical application. To address the challenge, this study introduces a Condition Adaptive Eddy Viscosity Turbulent Model (CAEVTM) tailored for specific subchannel conditions in rod bundle with Mixing Vane Grids(MVGs). The goal is to achieve a balance between high accuracy and low computational cost by integrating high-order numerical simulations with machine learning techniques. The research begins with performing scale-resolving simulation to obtain high-fidelity flow structures. Key physical quantities sensitive to subtle flow changes are then selected to ensure the model's responsiveness. Subsequently, gradient-based techniques are employed to calculate the sensitivity of each grid point, and a optimization method is used to optimize the correction coefficients, enhancing the model's accuracy. A neural network is then trained to map the operational conditions to the correction coefficients efficiently. Finally, the trained neural network is incorporated into the secondary eddy viscosity model to construct the CAEVTM. Results demonstrate that CAEVTM significantly improves the accuracy of flow field predictions in specific subchannels while greatly reducing computational costs. The proposed CAEVTM successfully combines high-order numerical simulations with machine learning techniques to achieve efficient and accurate simulations of flow fields in specific grid subchannels.