^{1}

^{1}

^{2}

^{1}

^{2}

The acceptable accuracy for simulation of severe accident scenarios in containments of nuclear power plants is required to investigate the consequences of severe accidents and effectiveness of potential counter measures. For this purpose, the actual capability of CFX tool and COCOSYS code is assessed in prototypical geometries for simplified physical process-plume (due to a heat source) under adiabatic and convection boundary condition, respectively. Results of the comparison under adiabatic boundary condition show that good agreement is obtained among the analytical solution, COCOSYS prediction, and CFX prediction for zone temperature. The general trend of the temperature distribution along the vertical direction predicted by COCOSYS agrees with the CFX prediction except in dome, and this phenomenon is predicted well by CFX and failed to be reproduced by COCOSYS. Both COCOSYS and CFX indicate that there is no temperature stratification inside dome. CFX prediction shows that temperature stratification area occurs beneath the dome and away from the heat source. Temperature stratification area under adiabatic boundary condition is bigger than that under convection boundary condition. The results indicate that the average temperature inside containment predicted with COCOSYS model is overestimated under adiabatic boundary condition, while it is underestimated under convection boundary condition compared to CFX prediction.

The containment phenomenological aspects during an accident have been studied extensively during the last 40 years for light water reactors [

Considerable international efforts were dedicated to better understand related phenomena by performing experiments and analytical assessments of their results. Since it is not possible to perform containment thermal-hydraulics experiments in the existing nuclear power plants due to safety concerns, experiments are performed in special facilities, which imitate containment or their parts [

Two main kinds of codes/approaches are used for simulation of containment thermal-hydraulics, that is, lumped-parameter approach with highly simplified 0D models and 3D CFD (Computational Fluid Dynamics) approach.

The program COCOSYS, a lumped-parameter code, is being developed by the Gesellschaft für Anlagen-und Reaktorsicherheit (GRS) gGmbH, Germany, for the simulation of all relevant processes and plant states during severe accidents in containment of light water reactors. And this code is widely used in nuclear engineering [

The CFX code is a general purpose CFD tool developed by ANSYS Inc. The code solves the conservation equations for mass, momentum, and energy together with their initial and boundary conditions. The discretization of the equations in the CFX code is based on a conservative finite-volume method.

Considerable research has been devoted to the study of the associated phenomena predicted by lumped-parameter code and field code; the development of various computer codes to analyze these severe accidents phenomena is summarized in the review [

Nevertheless containment thermal-hydraulics prediction remains an open question. One outcome of the ISP-47 (TOSQAN, MISTRA, and THAI) [

In present work, a simplified enclosure based on generic containment is adapted in prototypical geometries for comparing different simulation results with separate effects scenario “thermal plume” to illustrate the prediction capacity of COCOSYS and ANSYS CFX.

Irrespective of the nature of the accident, heat and mass transfer play a major role in these accidents. Quite often it is a complex phenomenon involving forced and natural convection heat transfer, metal-water reaction, nuclear heat generation, melting, condensation, diffusive and convective mass transfer, nucleate and film boiling, porous medium, combustion, and detonation.

The analyses presented here aimed at investigating the accuracy of COCOSYS code compared to the CFD codes to provide an evaluation of the applicability to the large-scale, transient problems. To this aim, the assessment must use separate-effect simulation, so we focus on a plume (due to a heat source) process in the present work; on the other hand, because of the thin shell and cylinder structure in CFX model, measures of smaller structure thickness with higher structure conductivity, lower density, and heat capacity are taken in COCOSYS model to eliminate the transient process impact of the structure, so that COCOSYS model and CFX model are comparable.

Figure ^{3}, including a cylinder with radius of 27 m and a height of 27 m, and a hemisphere with radius of 27 m. R-CAVITY (radius

Geometry (a) and mesh (b) for CFD simulation.

After grid independent test (coarse, intermediate, and fine mesh), mesh containing tetra element (element size 1.0 m) is adopted; it is simulated in CFX with a 3D Cartesian geometry model using 90,142 computational elements with 1234,062 nodes. Figure

The input data and nodalisation of the generic containment have been created on the basis of benchmark run-2 COCOSYS code for German PWR simulation [

COCOSYS model.

An overview of the general initial condition and boundary condition is listed as follows.

Adiabatic case:

Initial condition: air at 1 bar pressure and 20°C temperature.

R-CAVITY volumetric heat generation: 3000 kW.

Boundary condition: adiabatic enclosure.

Convection case:

Initial condition: air at 1 bar pressure and 20°C temperature.

R-CAVITY volumetric heat generation: 3000 kW.

USUMP boundary condition: adiabatic condition.

More specific information on the convection condition which follows in Section

Direct comparison of the zone temperature in COCOSYS and discrete point temperature is difficult due to some differences between COCOSYS and CFX; an alternative approach is to take a weighted average of CFX discrete points corresponding to COCOSYS area temperature.

The results are compared in curve charts for transient evolution process in Figure

Comparison of zone temperature at transient state.

The

Comparison of temperature distribution at 10106 s.

Two vertical lines (

CFX predictions (line

Contour plot of the temperature at 10106 s.

Stream lines and velocity vector at 10106 s.

It can be seen that the model and boundary conditions are selected symmetric along

For the sake of ensuring the comparability between CFD and COCOSYS, the following should be mentioned here:

The size of the system under consideration makes the computations very time-consuming. The running time (or CPU time) is in the range of 2 months per run to reach thermal equilibrium with CFD if the convective heat transfer coefficient is about 10 W/(m^{2}
^{2}

In order to eliminate the wall dynamic inertia delay, measures of higher structure conductivity, lower density, and heat capacity are taken in COCOSYS. Thin shell and plate model with ignoring wall thickness are used in CFD to separate containment enclosure and environment, while it affects the dynamic process, but the steady state will not be affected principally.

There are 14 layers (R-SUMP, R-SG12, D1, D3, D5, D7, D9, D11, D13, D15, D17, D19, D21, and D23) in COCOSYS model, so 14 points are used for plotting the COCOSYS predicted data (Figure

Three vertical lines (

Comparisons of right part zones between COCOSYS prediction and CFX prediction are shown in Figure

Comparison of temperature distribution.

It can be readily seen that near the heat source zone (R-SG12 in COCOSYS,

CFX predictions (line

Contour plot of the temperature.

CFX predictions (line

This feature can be verified from the comparison of horizontal temperature distribution for the zones of R-ANN12, SG12, SG34, and R-ANN34 (Figure

Horizontal temperature distribution.

On the other hand, the horizontal line along

From Figure

Stream lines and velocity vector.

For comparison purpose of two boundary conditions, the flow configurations and temperature profiles are different; from Figure

Comparison of thermal stratification area.

Adiabatic boundary condition

Convection boundary condition

Main conclusions for separate-effect plume (due to a heat source) simulation between COCOSYS and CFX can be summarized as follows.

Temperature in the upper part of the enclosure is higher for both adiabatic and convection boundary condition; this can be predicted by both COCOSYS and CFX.

Convection intensity affects concentration and temperature stratification; both COCOSYS and CFX can predict that there is no temperature stratification in the upper region of enclosure. Temperature stratification exists in the lower region of enclosure except in the region near the thermal plume.

Boundary condition affects the temperature stratification. Temperature stratification area under adiabatic boundary condition is bigger than the area under convection boundary condition; CFX are able to predict this phenomenon; however, COCOSYS are not able to predict this phenomenon.

CFX can predict local temperature of thermal plume, while COCOSYS cannot predict local temperature of thermal plume at present nodalisation.

Boundary condition affects the predicted average temperature. The average temperature in enclosure predicted by COCOSYS is overestimated compared to that predicted by CFX under the same adiabatic boundary condition; the average temperature in enclosure predicted by COCOSYS is underestimated compared to that predicted by CFX under the convection boundary condition.

The authors declare that they have no competing interests.

This work is supported by the National Science Foundations of China (51176132). The corresponding author gratefully acknowledges the support of K. C. Wong Education Foundation and DAAD. The corresponding author is approved to use COCOSYS code in KIT Germany.