VALIDATION AND QUALIFICATION OF ADVANCED THERMAL HYDRAULIC AND SAFETY ANALYSIS TOOLS FOR THE SAFETY ASSESSMENT OF VVER-1000 REACTORS

V. Sánchez, W. Jaeger, M. Boettcher, K. Ivanov, R. Stieglitz Karlsruhe Institute of Technology (KIT),

Institute for Neutron Physics and Reactor Technology (INR) Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany

Corresponding author: victor.sanchez@kit.edu

Abstract

At the Karlsruhe Institute of Technology (KIT), the Institute for Neutron Physics and Reactor Technology (INR) is focused on the further development and qualification of best-estimate coupled neutronic and thermal hydraulic system codes for reactor safety evaluations of LWRs including VVER-type reactors. These research activities are embedded in international programs such as the US NRC CAMP program, OECD/NEA Benchmarks of sub channel, CFD and uncertainty methods for multiphysics analysis for LWR.

In the frame of benchmarks related to coupled codes (PBTT2, VVER-1000 CT-1 and CT-2) the focus was the qualification of the system codes with 3D neutron kinetics models by a plant data. Special attention is paid on the prediction of the spatial asymmetrical core cooling and its effects on the local power distribution. Sensitivity evaluations are necessary to identify the most important phenomena and assumptions affecting the numerical predictions of coupled codes. The V1000-CT-benchmark Phase 2 was focused on both multidimensional thermal hydraulics phenomena inside the reactor pressure vessel of a VVER reactor. KIT participated in this benchmark using CFD, 3D coarse mesh system codes and a combination of 1D system codes with CFD (RELAP5/CFX) to investigate the coolant mixing tests measured at Kozloduy NPP.

This paper summarizes the validation, qualification and application of different codes such as RELAP, TRACE and CFX to predict key phenomena relevant for VVER reactors. The models vary from one dimensional to three dimensional representation of the RPV and whole plant models to describe the multidimensional coolant mixing in the downcomer and the lower plenum. Finally, a simulation of the postulated MSLB-transient was performed with the 3D TRACE model of the VVER-1000 reactor pressure vessel to show the ability of best-estimate coupled codes like TRACE/PARCS to describe complex transients, where the thermal hydraulics and core neutronics are tightly linked. 1. Introduction

The qualification of thermal hydraulic system codes is an important prerequisite for the use of best-

estimate codes to assess the safety features of nuclear power plants. In the last years, many best-estimate codes were improved by the implementation of three-dimensional thermal hydraulic (3D) models, at least for the reactor pressure vessel e.g. RELAP5-3D [1], CATHARE-3D [3] and TRACE [5]. In TRACE, a 3D vessel model in both cartesian and cylindrical geometry is available. These advanced system codes are coupled to 3D neutron kinetic resulting powerful multi-dimensional tools to analyze transients with strong, non-uniform power perturbations.

The main goal of this paper is to present the validation and application of 3D thermal hydraulic and CFD codes using experimental data related to the coolant mixing phenomena measured in the Kozloduy VVER-1000 plant distributed in the frame of the VVER-1000 Coolant Transient Phase 2 benchmark team [6]. In addition, the analysis of the core behavior of a VVER-1000 reactor in case of a postulated

main steam line break transient with a neutronic/thermal hydraulic coupled code (TRACE/PACRS) will be presented. The 3D thermal hydraulic models of the VESSEL-component of TRACE solve the fluid dynamics equations in three directions in coarse 3D nodes.

The development of coarse mesh 3D thermal hydraulic models for a complete reactor pressure vessel is very challenging since a proper strategy needs to be developed for the discretization of the computational domain in axial levels, radial rings and azimuthal sectors. To do so, the constructive peculiarities of the main flow paths such as the downcomer, lower plenum, core, and upper plenum must be taken into account.

On the other hand, the use of CFD codes for the description of thermal hydraulic phenomena inside the reactor pressure vessel e.g. in case of boron dilution transients or coolant mixing situations but also for design optimization is continuously increasing in the nuclear technology and nuclear safety. Due to the available CPU power of parallel computing environments, very detailed and large CFD models of complete primary systems including the RPV of LWR is nowadays feasible.

In the next chapters, the validation work performed for system and CFD codes will be presented and discussed. In this connection, the test conditions and measured data as well as the results obtained with the different thermal hydraulic codes will be presented and discussed hereafter. Finally the application of the validated 3D thermal hydraulic models of TRACE to simulate a postulated MSLB transient will be given.

2. Short description of the numerical tools

The system code TRACE [3] (TRAC/RELAP Advanced Computational Engine) is the reference

system code of the US NRC for LWR safety assessment. TRACE is a multidimensional, two-phase flow system code, developed to simulate any kind of operational events, transients and design basis accidents of LWR. The component VESSEL allows the 3D simulation of the flow in the reactor pressure vessel.. The in-build point kinetics model, based on the Kaganove-approach, is extended by the coupling or direct incorporation of the three dimensional core reactor simulation tool PARCS [8]. Thanks to the coupling of TRACE with PARCS, a powerful system is created that is appropriate for the simulation of transients and accident scenarios with strong power perturbation within the core. CFX [7] is a commercial CFD tool widely used to simulate 3D flow in complex geometries. Its application in the nuclear reactor safety is rapidly increasing, especially for single phase flow situations. 3. Thermal hydraulic Models of the Reactor Pressure Vessel 3.1 The 3D TRACE Model of the Reactor Pressure Vessel

A detailed 3D model of the RPV of the VVER-1000 reactor representing the most relevant internals was developed for TRACE [9]. The 3D VESSEL component of TRACE was used for the representation of the RPV. To catch the asymmetrical coolant mixing expected to occur mainly in the downcomer and lower/upper plenum a rather fine nodalization of the RPV in azimutal and radial direction is needed. It must be noted that the finer nodalization leads to a higher CPU time. A reasonable compromise between accuracy and CPU cost is here mandatory. According to this, the whole RPV is subdivided in 30 axial levels, six radial rings and six azimutal sectors, Figure 1 and Figure 2. The sizes of the respective nodes depend on the existing flow conditions along the main flow paths within the RPV determined by the constructive peculiarities of the RPV internals. From the 30 axial levels of the RPV, 10 axial nodes belong to the core region while two belong to the lower and upper axial reflector. The azimuthal sectors

(S1 to S6) were defined in a way that the cold legs are connected to sector 4 (cold leg 1), sector 6 (cold leg 2) sector 1 (cold leg 3) and sector 3 (cold leg 3). Figure 9 shows the radial nodalization of the RPV into 6 rings, 3 of them are for the core. For each of the 3D volume elements, the main thermal hydraulic parameters for each direction such as hydraulic diameter, flow area, heated diameter and form loss coefficients, etc. are derived from the detailed plant data. In developing the 3D model using the VESSEL component the following aspects had to be kept in mind:

- Make use of geometrical symmetry (R, θ, Z), - Select the size of cells (radial, axial, angular) as small as necessary (based on underlying

physics), and - Consider the details of flow paths as much as necessary. - Otherwise the 3D model may become unnecessary complex.

Figure 1. TRACE axial nodalization of the RPV

Figure 2. TRACE radial and azimutal subdivision of the core

S4

S1

S6

S3

R1

R2

R3R4

S4

S1

S6

S3

R1

R2

R3R4

Figure 3. TRACE Nodalization of the core and relative position of the cold/hot legs

3.2 Detailed CFX Model for the VVER-1000 primary circuit

Two CFD model versions

of the VVER1000 were developed. The first one considers the RPV, see [8], while the Loop model is extended to the primary loops with all its components described in [9]. In the lower and upper plenum all components with significant influence on coolant mixing are resolved in detail. The core is resolved by a porous media approach representing 163 individual assemblies. At an active core length of 3550 mm the hexa mesh contains 35 layers with about 800000 cells. For the other parts of the RPV an unstructured mesh consisting of 23 million cells with a variable spatial resolution between 5 and 50 mm is used. Ошибка! Источник ссылки не найден. shows a cut through

the RPV model and in Ошибка! Источник ссылки не найден. a part of the mesh from the lower plenum is presented. For the inlet nozzles, the downcomer and some parts of the lower plenum a CAD model in [10] was used. For other parts like the elliptical bottom plate

Figure 4 The detailed CFD model of the VVER-1000

or the upper plenum the construction was performed directly from technical drawings provided within the VVER-1000 coolant transient benchmark [4].

The only parts inside the lower and upper plenum modelled by a porous medium approach are parts of the core support columns. Each of the 163 support columns is formed by a solid cylindrical lower part and a tube like upper part perforated by 14 rows with together 420 slots and at the 42 peripheral positions by 12 rows with 360 slots at a length of 30mm and a width of 3 mm. An individual CFD model for a single upper support column with a much higher spatial resolution was

created and by the calculated pressure losses loss coefficients were derived and introduced in the RPV and Loop model. Fig. 6 shows the flow situation at standard conditions of 17.9 tons/s and 540 K inlet temperature. Compared with individual model each part of one upper support column is meshed only with 6000 cells (against 250000 for a 12° part), but the flow patterns agree quite well.

Figure 5. Mesh from the lower part of the RPV

As mentioned before the Loop model also considers the primary loops with all its components. At the present computational possibilities the only chance of implementation only can be achieved by a porous media approach and the implementation of source terms for momentum and energy exchange. About 10 million additional cells are used for the representation of the primary loops represented by Fig. 7. The steam generators have an outer shape of a horizontal cylinder. Inside a steam generator the primary loop coolant flows through approximately 11000 horizontally located U-pipes which are placed in a water pool. The 3D void fraction from this pool cannot be described by simple correlations therefore the heat

Figure 6. The flow around a standard support column at the lower plenum

transfer by steam generation can

be only described by a multiple layer model of a system tool. Due to this reason and with the intention to get a simple, numerical stable model the secondary flow part is not considered and the heat removal by steam generation is implemented by a homogeneous volumetric heat sink.

The heat transfer through the u-pipe walls is introduced as a constant volumetric heat source based on plant data and with linear dependency on the difference between the mean fluid temperature at the entrance of the steam generators and a constant temperature on the secondary side. Within Ошибка! Источник ссылки не найден. the flow path within the steam generator of loop 3 is indicated. After the entrance into the steam generator the flow separates into 2 parts with

approximately the same fluxes and moves along an internal wall towards the outlet. The total amount of heat removal Qj,tot from steam generator j is calculated as

Figure 7. The integral CFD Model of the VVER-1000 Primary Loop

sjpj

sjpj

totj TTTT

QQ,0,

,,

0, −

−−= (4),

where Q0 is a given total amount of heat at a reference point, Tj,s is the temperature from the steam generators primary side at a reference point, Tj,p0 is the temperature at the entrance of the steam at reference conditions and pj ,T is the volume-averaged temperature from the RPV outlet until the inlet of the steam generator. This equation implies that the mass flux and pressure at the secondary side are equal or very close to those at the reference point. Moreover the heat transfer is direct proportional to the temperature difference between the primary and secondary side. It has to be mentioned that it is not sufficient to define a heat source independent on the fluid temperature because the Loop model is a closed system without inlets or outlets where the fluid temperature can be specified. As the nuclear sources of the core are specified as absolute values derived from a PARCS simulation (see [11]) and the outer walls of the system are considered as adiabatic, eq. (4) is the only possibility for the solver to adjust the temperature of the fluid.

For the pumps only the shape of the housings and the total flow volume was known. No information concerning the impeller shapes was available. As consequence the interaction of the impellers with the fluid has to be modelled by volumetric momentum source terms derived by measured mass flow rates. Within the pump housing a constant axial pressure gradient was introduced adjusted by the measured mass flows rates of each loop. The implementation follows the momentum equation

Mi

j

iji

ij

iji Sx

gxp

xUU

tU

+∂

∂++

∂∂

−=∂

∂+

∂∂ τ

ρρ )()(

(5),

where is a source term, which is set to 1.3·106 [kg m-2 s-2] inside the pump volume at standard operation conditions.

MiS

In fact the interaction of the impellers with the coolant is a 3D, instationary process. When passing through the pump the fluid suffers a pressure increase between inlet and outlet, additionally a radial and circumferential local instationary momentum transfer through the interaction with the impellers. Finally the impellers are acting as turbulence source. For a detailed quantitative analysis an implementation of the impeller shapes together with a frozen rotor approach or a sliding mesh implementation for consideration of their rotation is necessary. 4. Short description of the heat-up experiment

Before the test, the nuclear power plant Kozloduy was operated at around 9.36 % of the nominal power, i.e. 281 MWth, with all main coolant pumps running. On the secondary side all steam generators were available. The core was loaded with fresh fuel i.e. beginning of cycle conditions (BOC) with a core averaged exposure of 0.4 effective full power days (EFPD) and a boron concentration of 7.2 g/kg. The position of the control rod groups were as follows: group #9 and #10: fully inserted; groups #1-#7: fully withdrawn and the regulating rod group #8 was about 84% withdrawn from the bottom of the core. The coolant temperature at core inlet was 20 K lower than the one at nominal conditions. Finally the steam generator levels were as high as the ones at nominal conditions. The main steam header pressure was 5.07 MPa, meaning 1 MPa lower than the nominal value.

The test was initiated by the isolation of the steam generator of loop 1 due to the closure of the main steam isolation valve. As a consequence, the primary coolant temperature of loop 1 started to increase up to about 14 K higher than the coolant temperature of the other loops. Under such conditions, coolant mixing occurred, first of all in the downcomer region. The resulting mixing pattern propagates through the lower plenum, core and upper plenum. Since the power was relatively low, the feedbacks between thermal hydraulics and core neutronics are negligible according to the recorded data. Due to the mixing, the temperature of the unaffected loops, especially of the loop close to the loop 1 (loop 2) increased. The test lasted for 1800 s. At that time the power increased up to 286 MW. Different data was recorded at the Kozloduy plant during the test.

In Figure 8, the recorded data of the four hot legs are given for the whole test. There it can be seen that the coolant temperature of the affected loop 1 starts to increase very rapidly at around 130 sec. due to the deteriorated heat transfer over the SG 1. From 500 s onward the increasing rate becomes smaller, stabilizing at a value below 556 K. Due to the coolant mixing in the downcomer the temperature of the loop 2 experienced a higher temperature than the one of loop 4, indicating that the mixing pattern is not in clockwise direction. Note that the position of the loops is not symmetrical.

540

542

544

546

548

550

552

554

556

0 300 600 900 1200 1500 1800

time [s]

tem

pera

ture

[K]

Cold leg #1 Cold leg #2 Cold leg #3 Cold leg #4

Figure 8. Measured evolution of the hot legs during the test at the KNPP 5. Selected Results of Simulations 5.1 TRACE Simulation Results

The TRACE post-test calculations of the coolant mixing experiment were performed in two steps. First of all, a steady state calculation was carried out to predict the plant conditions just before the test. Secondly, a transient run was made for 1800 seconds to determine the final state of the plant. In Table 1, a comparison of the TRACE predictions and the plant data is given for the initial plant state is exhibit. It can be seen that the agreement between data and prediction is quite good. At the initial state the coolant temperature at the core inlet/outlet is uniformly since all pumps and steam generators are in operation. This will change drastically during the heat-up test progression. Note that the largest deviation between the prediction and the data is below 4 %.

The transient phase started with the isolation of the main steam isolation valve and lasted for 1800 sec. The final plant state predicted by TRACE is compared to the plant data and shown in Table 2. A comparison of the measured coolant temperature at the fuel assembly outlet with the predicted values by TRACE is given in Figure 11. It can be seen that the TRACE predictions follow qualitatively the trend of the measured data. In some positions, TRACE tends to over predict and in others to under predict the data. But the differences between data and predictions are within the measurement error. In Table 2, it can be observed that the code predictions are close to the plant data. In addition to the hot/cold leg temperatures, the pressure drop is also in good agreement with the data. Since during the test the hot leg temperature of the loop 1 (Figure 9), was continuously increasing while the one of the other loops were not, a considerable coolant mixing took place in the downcomer. The predicted coolant temperature of each fuel assembly at the core outlet for the beginning and end state of the test is given in Figure 9 and Figure 10. There the mixing pattern within the core can be observed. The hotter fluid of the loop 1 get mixed with the one of the sector between the loop 1 and loop 2, i.e. in counter-clockwise direction as observed in the tests.

Table 1 Comparison of TRACE predictions with plant data for the initial state

Parameter Initial State Accuracy TRACE Deviation Thermal power, MW 281 ± 60 281 0 Pressure above core, MPa 15.593 ± 0.300 15.592 0.001 Pressure drop over RPV, MPa 0.418 ± 0.043 0.404 0.014 Coolant temperature at core inlet #1, K 541.75 ± 1.50 541.78 -0.03 Coolant temperature at core inlet #2, K 541.85 ± 1.50 541.88 -0.03 Coolant temperature at core inlet #3, K 541.75 ± 1.50 541.78 -0.03 Coolant temperature at core inlet #4, K 541.75 ± 1.50 541.78 -0.03 Coolant temperature at core outlet #1, K 545.00 ± 2.00 544.63 0.37 Coolant temperature at core outlet #2, K 545.00 ± 2.00 544.70 0.30 Coolant temperature at core outlet #3, K 544.90 ± 2.00 544.61 0.29 Coolant temperature at core outlet #4, K 545.00 ± 2.00 544.62 0.38 Mass flow rate of loop #1, kg/s 4737 ± 110 4749 -12 Mass flow rate of loop #2, kg/s 4718 ± 110 4735 -17 Mass flow rate of loop #3, kg/s 4682 ± 110 4750 -68 Mass flow rate of loop #4, kg/s 4834 ± 110 4737 97

Table 2 Comparison of TRACE predictions with plant data for the final state

Parameter Final State Accuracy TRACE Deviation Thermal power, MW 286 ± 60 286 0.000 Pressure above core, MPa 15.593 ± 0.300 15.591 0.002 Pressure drop over RPV, MPa 0.417 ± 0.043 0.404 0.013 Coolant temperature at core inlet #1, K 555.35 ± 1.50 555.39 -0.04 Coolant temperature at core inlet #2, K 543.05 ± 1.50 543.08 -0.03 Coolant temperature at core inlet #3, K 542.15 ± 1.50 542.18 -0.03 Coolant temperature at core inlet #4, K 542.35 ± 1.50 542.38 -0.03 Coolant temperature at core outlet #1, K 554.85 ± 2.00 555.14 -0.29 Coolant temperature at core outlet #2, K 548.55 ± 2.00 548.66 -0.11 Coolant temperature at core outlet #3, K 545.75 ± 2.00 545.44 0.31 Coolant temperature at core outlet #4, K 546.45 ± 2.00 545.69 0.76 Mass flow rate of loop #1, kg/s 4566 ± 110 4657 -91 Mass flow rate of loop #2, kg/s 4676 ± 110 4693 -17 Mass flow rate of loop #3, kg/s 4669 ± 110 4724 -55 Mass flow rate of loop #4, kg/s 4816 ± 110 4724 92

Figure 9 Predicted coolant temperature at the core outlet at initial state (0 sec)

Figure 10 Predicted coolant temperature at core outlet at final state (1800 sec)

T (K)

T(K)

Final State at 1800 sec.

542

546

550

554

558

562

1 21 41 61 81 101 121 141 161FA_Number

Tout

(K)

Data Trace

Figure 11 Comparison of the predicted coolant temperature at each FA outlet with data

5.2 CFX Simulation Results

The CFD predictions are compared to both thermocouple data at the main RPV inlets/outlets as well as the temperature at the FA outlet including the flow rates [9]. In addition mixing coefficients derived from thermocouple measurements are used. At first a steady state analysis of the temperature distribution at the core outlet is presented. The theoretical average temperature rise can be very easily calculated as approximately 3.3 K by using a flow rate of 17900 kg/s, a thermal heat release of 281 MW and a specific heat capacity from the coolant of 4900 J/kg K. The Figures 12 and 13 show a comparison of the assembly outlet temperatures for the RPV model and the Loop model. The predicted temperatures from both models are quite similar which means that the velocity distribution at the core inlet is more or less invariant against different conditions of velocity and turbulence parameters at the inlets of the downcomer if the total mass fluxes are the same. The differences between simulations and measurements are less than 1.2 K, i.e. within the accuracy of the thermocouple measurements. The

measurements are about 0.5K below the theoretical core outlet temperature of 545.08K, while the predicted averaged assembly outlet temperature are 544.77K (RPV model) and 544.95K (Loop model). The global energy balance of the simulations was fulfilled with an accuracy of 1‰ which reveals excellent convergence behaviour.

543,0

543,5

544,0

544,5

545,0

545,5

546,0

546,5

1 21 41 61 81 101 121 141 161

Assembly No

Tem

pera

ture

[K]

RPV model 2nd orderExperimentT_average, theoret.

Figure 12 Assembly outlet temperatures for the RPV model

543,0

543,5

544,0

544,5

545,0

545,5

546,0

546,5

1 21 41 61 81 101 121 141 161

Assembly No

Tem

pera

ture

[K]

Loopmodel 2nd orderExperimentT_average, theoret.

Figure 13 Assembly outlet temperatures for the Loop model

In Figure 14 the steady state temperature distribution from the Loop model together with the flow

situation at loop 1 illustrated by streamlines is exhibited. The temperature distribution within the core shows hot spot regions which are a consequence of fuel assemblies with higher enrichment. The coolant flows into the downcomer with a temperature of about 542 K. By passing through the core it is heated up by 3.3 K in average. On its way to the RPV outlets through the upper plenum with guide tube structures two perforated walls have to be passed. When leaving the vessel the temperature gradients have decreased but temperature stratifications are still visible in the pipe connections to the steam generators.

Finally the transient heat-up of loop 1 and 2 compared with the measured data are compared in Figure 15 and Figure 16. There the results of the old and new RPV-model are also given. For all transients a time step of 12 s is used. While in the old model version (CFX 5.7.1, 1st order results) a standard k-ε turbulence model was used, in the actual version a SST (Shear Stress Transport) model was applied, also for the steady state case. Both models predict the temperature rise of the disturbed loop quite well but the new RPV model shows an over prediction of about 2 K. This refers to an underestimation of the mixing process between the disturbed loop and its neighbours due to other influences not taken into account. The time delay observed in the experimental data is reproduced now

much better by the new model. The fluctuations of the measurement data can be also seen in the simulations, Figure 15 and Figure 16. Even the fluctuations with a lower frequency can be in principle reproduced now by the new model. One exception seems to be given by loop 4, where the older results are much closer to the observations. The interaction with the disturbed loop is strongly over predicted by the new model.

Figure 14: Steady state temperature distribution from the RPV model

542

544

546

548

550

552

554

556

558

0 200 400 600 800 1000 1200 1400 1600 1800

Time [s]

Tem

pera

ture

[K]

experimentCFX5 coarse grid 1st orderCFX11 fine grid 2nd orderCFX11 fine grid 1st orderLoopmodel 2nd order

Figure 15: Comparison of predicted and measured temperature of loop 1 at RPV outlet

544

544,5

545

545,5

546

546,5

547

547,5

548

0 200 400 600 800 1000 1200 1400 1600 1800

Time [s]

Tem

pera

ture

[K]

experimentCFX5 coarse grid 1st orderCFX11 fine grid 2nd orderCFX11 fine grid 1st orderLoopmodel 2nd order

Figure 16: Comparison of predicted and measured temperature of loop 2 at RPV outlet

It can be seen that the 1st order method is much closer to the measured data. The additional

numerical diffusion by the coarser mesh seems to compensate unconsidered mixing effects. A reason might be the existence of flow structures which can only be captured by modelling the primary loops with all their components. Additional tests with different turbulence models or with shorter time steps have led to very similar results.

First transient simulations with the Loop model (red curves in Figure 15 and Figure 16) show that the heat-up from loop 1 is significantly underestimated. Due to the much stronger swirl in the downcomer the temperature rise at loop 2 is significantly above the measurements. Without implementation of the pump impellers or (and) additional artificial tasks like damping of the flow swirl at the cold legs between the pumps and the RPV inlets an accurate reproduction of the plant behaviour is not possible. 6. Calculation of a main steam line break transient

The core behaviour under a postulated MSLB transient was investigated with TRACE/PARCS

using the validated 3D thermal hydraulic TRACE model.

6.1 Description of the MSLB Scenario The MSL-break occurs outside the containment between the steam generator (SG) and the steam

isolation valve (SIV) of loop #4 leading to an unsymmetrical core overcooling. The main purpose of this investigation is the possible return to power after reactor scram due to overcooling. In the MSLB scenario a reactor scram with a “stuck-rod” assumption immediately after the break initialization and the shut down of the main coolant pump (MCP) of the affected loop are assumed. Due to the MCP shut down a reverse flow established in the faulted loop in around 16 sec. A detailed description of the MSLB is given in [3]. Figure 17 shows the temperatures in the cold legs and Figure 18 shows the mass flow rates during the transient. These values will serve as boundary conditions for the TRACE/PARCS investigations.

510

520

530

540

550

560

57 0

0 50 100 150 200 250 300 350 400 450 500 550 600Time [s]

Tem

pera

ture

[K]

C old leg #1 C old leg #2 C old leg #3 C old leg #4

-2000

-1000

0

1000

2000

3000

4000

5000

6000

0 50 100 150 200 250 300 350 400 450 500 55

Time [s]

Mas

s flo

w ra

te [k

g/s]

Cold leg #1 Cold leg #2 Cold leg #3 Cold leg #4

Figure 17 Cold leg temperature distribution during the transient

Figure 18 Cold leg mass flow rate distribution during the transient

6.2 TRACE/PARCS Simulation

The used TRACE model is the same like the one for the coolant mixing problem. A mapping file was created [10] to link the thermal hydraulic nodes with the neutronic nodes taking into account the provided cross section libraries, since no automatic mapping exists for hexagonal geometry. Eigenvalue calculations with TRACE/PARCS were performed for different control rod positions to check if the cross sections are being read correctly by the PARCS simulator. 6.3 Selected MSLB Simulation Results

A steady state simulation was performed first of for the plant conditions just before the postulated

transient. The predicted results are also in good agreement to the benchmark specification. Figure 19 shows the calculated values for the temperatures in the hot legs during the transient. It can be seen that the temperature decreases in all hot legs after the MSLB and the reactor scram respectively.

460

470

480

490

500

510

520

530

540

550

560

570

580

590

0 50 100 150 200 250 300 350 400 450 500 550 600

Time [s]

Tem

pera

ture

[K]

Hot leg #1 Hot leg #2 Hot leg #3 Hot leg #4

-14

-12

-10

-8

-6

-4

-2

0

2

0 50 100 150 200 250 300 350 400 450 500 550 600

Time [s]

Rea

ctiv

ity [$

]

0,0E+00

5,0E+07

1,0E+08

1,5E+08

2,0E+08

2,5E+08

3,0E+08

3,5E+08

4,0E+08

4,5E+08

5,0E+08

Ther

mal

Pow

er [W

]

Reactivity Thermal power

Figure 19 Hot leg temperature distribution during the transient

Figure 20 Comparison of the reactivity and the thermal power during the transient

The biggest impact shows hot leg #4. In the first seconds after the MSLB the temperature in hot leg #4 as the same trend like the other legs.

Just before the transient the reactor operates at normal conditions. The core heat up is about 30 K. Due to the MSLB and the occurrence of boil-off on the secondary side of SG #4 the heat transfer ratio rises. This leads to a cool down of loop #4 and to coolant mixing in the downcomer. The temperature of cold leg #1 shows a slight impact because of mixing (0 s < t ≤ 16 s). A MCP trip in loop #4 is initialized to mitigate the overcooling of the core. The measured temperature in cold leg #4 is now nearly the same like the temperature in cold leg #1 due to coolant mixing (t ≈ 16 s). The shut down of MCP #4 yields a reverse flow through the faulted loop. The coolant from the unaffected loops, mainly loop #1 feeds the inlet of loop #4. Thus the coolant flows backward through SG #4. Due to the increasing of the reverse flow in loop #4 the temperature decreases quickly [20 s < t ≤ 45 s). After approx. 50 s the reverse flow rises slightly and the outlet temperature of loop #4 decreases slightly. This coolant enters the reactor at the upper plenum and mix there with coolant from the core due to the big temperature difference. This yields a decrease of the temperature in the unaffected loops. After approx. 100 s the feed water supply of the secondary SG side is shut down [9]. That leads to a faster evaporation of the secondary SG side due to the broken steam line and a temperature increase of hot leg #4. This has an influence to the unaffected loops again (120 s < t ≤ 600 s).

The main reason for this investigation was the possible return to power due to overcooling. Figure 20 shows the development of the reactivity and the thermal power during the transient. It is visible that the reactivity increases in the first 200 s of the transient despite the reactor is scrammed. The temperatures in the hot legs decrease in the same period. But it can also be seen that the power decreases after the scram during the whole transient. No return to power has been predicted. Hence this accident will not lead to re-criticality and to an undesirable return to power. 7. Summary and conclusions

In this paper, the investigations performed to validate the prediction capability of TRACE and CFX using coolant mixing data gained in a real nuclear power plant were presented and discussed. Detailed models for a 3D coarse mesh system code (TRACE) and fine mesh CFD model of the reactor pressure vessel of a VVER reactor were developed. From the comparison of the calculated parameters by TRACE with the available measurement data the following conclusions can be drawn:

- TRACE is able to qualitatively describe the coolant mixing phenomena inside the RPV in case of non-symmetrical thermal hydraulic flow conditions using a coarse 3d core discretisation.

- The chosen 3D thermal hydraulic nodalization scheme using the VESSEL component of TRACE seems to be appropriate to catch the underling physics of the coolant mixing process of VVER-1000 reactors.

- Based on this validated model a postulated MSLB transient was analyzed with TRACE/PARCS. The obtained results seem to be physically sound.

Different CFX models (RPV and Loop) of the VVER-1000 primary circuit with a different representation of components were developed to simulate the coolant mixing tests of the Kozloduy plant. The loop model includes a simplified representation of the primary circuit and its components like pipes, pumps and steam generators. Due to the complexity and the lack of geometrical detail information the pumps and steam generators are modelled by simplified outer shapes and homogeneous volumetric source terms for momentum and energy exchange. The improved version of the RPV model by CFX is

able to predict the coolant mixing inside the RPV with acceptable accuracy. The counter-clockwise rotation of flow patterns is slightly underestimated by the RPV model and significantly overpredicted by the Loop model. In the future, the detailed geometry of the pump impellers and housing needs to be considered in the modeling to be able to simulate the observed swirl of flow patterns during the tests.

List of Acronyms ATHLET Analysis of Thermal-hydraulics of Leaks and Transients ATWS Anticipated Transients Without Scram BWR Boiling Water Reactor BOC Begin of Cycle CATHARE Code for Analysis of Thermal-hydraulics during an Accident of Reactor and Safety

Evaluation CFD Computational Fluid Dynamics EFPD Effective Full Power Days INR Institute of Neutron Physics and Reactor Technology KNPP Kozlody Nuclear Power Plant LANL Los Alamos National Laboratory MSLB Main Steam Line Break PARCS Purdue Advanced Reactor Core Simulator PWR Pressurized Water Reactor RPV Reactor Pressure Vessel SG Steam Generator SNAP Symbolic Nuclear Analysis Package TRACE TRAC/RELAP Advanced Computational Engine TRAC Transient Reactor Analysis Code VVER Water-Water Energy Reactor

References [1] R. Riemke “RELAP5 Multi-Dimensional Constitutive Models,” RELAP5/TRAC-B International Users

Seminar, Baton Rouge, LA, November 4-8, 1991. [2] P. Bazin, M. Pelissier, 2006, CATHARE 2 V25_1: Description of the base revision 6.1 physical laws

used in the 1D, 0D and 3D modules. CEA/DER/SSTH/LDAS/EM/2005-038. [3] F. Odar. et al; TRACE V4.0 User´s Manual. US NRC 2005. [4] N. Kolev. S. Aniel, E. Royer, U. Bieder, D. Popov and Ts. Topalov; The OECD VVER-1000 Coolant

Transient Benchmark Phase 2 (V1000CT-2) Volume I: Specifications of the Coolant mixing problem. NEA/NC DOC (2004).

[5] H. Joo, D. Barber, G.. Jiang, and T. Downar; PARCS: A multidimensional two-group reactor kinetics code based on the nonlinear analytical nodal method. Purdue University, School of Nuclear Engineering. July 2002.

[6] ANSYS CFX, Release 11.0 (2007). Reference Guide. [7] W. Jaeger, W., Lischke, V. Sánchez Espinoza, Safety Related Investigations of the VVER-1000 Reactor

Type by the Coupled Code System TRACE/PARCS“. Proceedings 15th International Conference on Nuclear Engineering (ICONE 15), Nagoya, Japan, April 22-26, 2007.

[8] Böttcher, M. Detailed CFX-5 study of the coolant mixing within the reactor pressure vessel of a VVER-1000 reactor during a non-symmetrical heat-up test. Nuclear Engineering and Design 238 (2008), 445-452.

[9] Böttcher, M., Krüßmann, R. Primary loop study of a VVER-1000 reactor with special focus on coolant mixing. Nuclear Engineering and Design 240 (2010), 2244-2253.

[10] Kolev, N., Aniel, S., Royer, E., Bieder, U., Popov, D., Topalov, T. (2004). Volume 2: Specifications of the VVER-1000 Vessel Mixing Problems, Commissariat a l’Energie Atomique and OECD Nuclear Energy Agency, VVER-1000 Coolant Transient Benchmark (V1000CT).

[11] V. Sánchez, V. et al; Analysis of The VVER-1000 coolant Transient Benchmark Phase 1 with RELAP5/PARCS. Progress of Nuclear Energy 48 (2006), 865-879.