CFD benchmark of flat plate fuel assembly

. Diversification of the nuclear fuel suppliers is an inseparable component of nuclear safety of power and research nuclear reactors. So far, the Czech research reactor LVR-15 has been operated with the Russian fuel IRT-4M, because there was no alternative to this fuel. Recently, the French company CERCA has developed an alternative to this fuel Flat Plate Fuel Assembly (FPFA). Hence, the influence of the different fuel characteristics (especially neutronic, thermomechanic and hydraulic) on nuclear safety has to be provided. This work involves a comparison of the flow characteristic of these two types of fuel assemblies using the ANSYS Fluent 2021 R1 and STAR CCM+ 2022 CFD programs. Moreover, these computational models of FPFA fuel are validated in an experimental facility.


Introduction
The determination of a neutron-physical, thermohydraulic and thermomechanical properties of any nuclear fuel are crucial for nuclear safety.
To diversify the research nuclear fuel market, Cerca has developed a new Flat Plate Fuel Assembly (FPFA), which should replace the current fuel IRT-4M for the LVR-15 reactor.
It was necessary to simulate the flow characteristics of both fuel types (mainly pressure losses) due to the different design, construction and material. Simulated data of FPFA model was verified on mockup model of FPFA in experimental loop build in CTU laboratories. The simulation was done in CFD codes ANSYS Fluent and STAR-CCM+.

Research reactor LVR-15
The reactor LVR-15 in research centre Řež has been in operation without major design changes since 1989. It is a light-water tank-type reactor with a nominal thermal output of 10 MW. The core is cooled by forced downward circulation of coolant. The reactor vessel is a pool-type with a volume of 22 m 3 . The core is located about 3-4 m below the surface (pressure in core 130 to 140 kPa). The reactor is cooled by a coolant flow up to 2000 m 3 ·h -1 and maximal temperature 56°C. So far, the reactor has been operated with Russian fuel IRT-4M with a UO2 matrix. New fuel FPFA has U3Si2 matrix. [1,2] An interesting feature of the LVR-15 reactor is the downward flow, i.e. against the natural circulation. The reactor was designed with a downward flow because fuel elements are sitting on a grid. In the axial direction they are secured only by their own weight.

Project LEU-FOREvER
LEU-FOREvER stands for Low Enriched Uranium Fuels fOR REsEarch Reactors. The project brings together major European institutions (CV-Řež. ÚJV, CEA, FRAMATOME -CERCA, ILL, NCBJ, SCK-CEN, TechnicAtome and TUM) in the field of nuclear research. The project has two goals. The first goal is the diversification of nuclear fuel suppliers for Soviet-type research reactors, e.g., VR-1 (Vrabec. FNSPE CTU in Prague) or LVR-15 (CV-Řež. ÚJV). Until now, the company ROSATOM had a monopol on this fuel and the reactors were only compatible with IRT-4M fuel. To offer different choices, FRAMATOME -CERCA developed FPFA fuel assemblies that can replace IRT-4M fuel. The second goal is the conversion of HPRRs (High Performance Research Reactors) from highly enriched to low-enriched fuel. [1,2] 2 Comparison of IRT-4M and FPFA fuel The main difference between fuel assemblies is in geometry, material and manufacturing process of uranium plates. In IRT-4M is 8 plates bend in a rectangle shape and set concentrically. Developers of FPFA fuel have problems with bending the plates, in corners appears cracks, for this reason, they decide for 22 straight plates. The matrix UO2 was replaced by U3Si2 for better accident resistance In case of interruption of the coolant circulation through the reactor core, passive cooling of the fuel is necessary. U3Si2 has a higher passive cooling capability, and it achieves up to 10 times higher thermal conductivity compared to UO2 at temperatures above 1400 °C, which reduces the probability and severity of a possible nuclear accident. [4]

Experimental loop
The structure and the main parts of the experimental loop can be seen in figure 4 with detailed the FPFA model in figure 5.  The experimental data in Figure 6 shows pressure drop, mass flow and temperature are directly dependent.

Measurement uncertainty
This experiment was dealing with the uncertainty of the ultrasonic flow meter GE AT600, Rosemount 1154 pressure transmitter, and thermocouples type K.
In the following measurements have been established uncertainties type A (UA) based on statistical means expressed by standard deviation and type B (UB) depends on the error of the measuring device, surrounding environment, used method or approach etc.
Transmitter supplier states accuracy ±0.25 % of range 0.6 bars includes combined effects of linearity, hysteresis, and repeatability. Error is valid only at a temperature 20±2 °C.
Thermocouples have an accuracy of ±0.75 %. [7,8] The ultrasonic flow meter according to the GE measuring and calibration list has accuracy depending on mass flow. For mass flow in the experiment was considered accuracy ±0.46 %.
Combined uncertainty is the square-root of the linear sum of square uncertainty components.

= √∑ 2 =1
(2) Expanded uncertainty is the last calculation when estimating uncertainty in measurement based on 95% Gauss law. Uncertainties of flow and pressure measurement depend on mass flow, temperature did not.

CFD simulation
The CFD simulation was using the FPFA CAD model described in Figure 8. The fluid volume model is in Figure 9.

Fluent simulation and mesh independent study
The calculations were performed in the ANSYS Fluent 2021 R1 and Star CCM+ softwares. ANSYS independent mesh study compared 8 IRT-4M and 7 FPFA meshes. The geometry of the calculation model is an approximation of the inverse volume of the fuel assembly. For flow stabilization is empty space before and behind solid parts 300mm. The "Inlet" and "Outlet" parameters were measured 10mm before and behind solid parts. The "hot channel" parameters were measured in the middle (in the middle of length uranium plate). Initial parameters of coolant flow are: temperature 50 °C, pressure 132 kPa and speed 1 m·s -1

Number of cells [-]
Pressure drop Velocity Figure 11. Dependency of the pressure drop and velocity on mesh size FPFA

Star-CCM+ simulation
For the following CFD simulation was used code STAR-CCM+ in version 2022.1. The mesh was chosen polyhedral in combination with Thin mesher and Prism layer mesher settings in the fuel grid area. The number of cells is 10.2·10 6 , the basic cell size is 5 mm. Thin mesher was set to 3 layers, Prism layer mesher also 3 layers. All mesh important parameters are summarized in Table 3 below. Concerning y+ assuming values under 12, the K-epsilon low Re turbulence model and all y+ Wall treatment model was chosen. From Table 4 can be seen slight deviations from the experiment. Deviation reduction can be achieved by further refining of numerical mesh and refining the physical or numerical model. The deviations against ANSYS Fluent may be due to different numerical mesh or different physical models. 3.3 -6.0 -8.7 -11.5 -3.0 The future steps of the benchmark are tuning of a physical and numerical model. Then a more detailed calculation with heat transfer will be made and the results will be extended by other

Number of cells [-]
Pressure Drop Velocity operating parameters (flows, temperatures). Based on verified data of a single element will be made reactor core model. Deviations from CFD codes are relatively small and their use for further calculations appears to be valid and reasonable. Future work will be focused on fuel element roughness approximation, test channel optimization and benchmark of experimental loop with the orifice plate.

Fluent fuel assemblies pressure drop comparison
The dependency below is created from Mesh number 2 for FPFA and Mesh number 7 for IRT-4M. The best possible mesh was chosen based on an independent mesh study. The choice is based on the number of cells and the number of iterations performed in 8 hours.
Every mesh was calculated with 10 different coolant inlet velocities at the inlet to the model from 0.4. to 2.2 m·s -1 with the step 0.2 m·s -1 . Regard to the inlet velocity, the new FPFA fuel has from 50 % to 90 % higher pressure drop than the old IRT-4M fuel. A higher pressure drop is associated with higher heat transfer.

Fluent heat transfer simulation
Heat transfer was simulated only in the case of FPFA. The temperature at the inlet is 45 °C and at the outlet 56 °C. Heat is generated in a matrix (U3Si2 45 % and Al 1050 55 %) with a thickness of 0.51 mm and passes through a wall (Al 5754) with a thickness of 0.38 mm.
The following simplifying hypotheses are assumed in the CFD simulation: • Constant thermal conductivity specific heat and density • Steady state • Constant heat output of fuel 337 MW·m -3 : this corresponds to the heat output of 328 kW of one fuel assembly • In real core configuration the fuel element is surrounded by other elements in experiment and simulation the element is placed in a channel  The Nusselt number was then used for the calculation of the heat transfer coefficient from the wall to the liquid in a turbulent flow. According to the simulation. at an input speed of 0.6 m·s -1 and a mass flow of 3.1 kg·s -1 , the coolant heats up from 45.0 °C to 55.5 °C at a fuel assembly heat power 328 kW. The fuel matrix locally reaches a maximum temperature of 77.0 °C (in the middle uranium plate). The maximum surface temperature of the fuel plates is 72.0 °C.

Conclusion
This work contained experimental and simulation data of flow characteristics for FPFA fuel and only simulation data for IRT-4M fuel. Pressure drop is significantly higher in the case of FPFA fuel. Higher pressure drop is related to higher heat transfer, FPFA should be safer than the precedent IRT-4M fuel. The result of the heat transfer simulation says that the coolant flow velocity in hot channel 3,11 m/s is needed to cool the heat output 328 kW of one single fuel element. Coolant temperature increases from 45 to 56°C. Data obtained by CFD simulation correspond with data obtained by experiment. The deviations between the data are within the measurement uncertainty tolerance range. Simulated data slightly underestimate pressure drop against experimental data, it can be caused by the rougher surface of the mockup model, than was expected. In CFD simulations was assumed default surface roughness. Based on experimentally obtained data will be built future reactor core model.