Yan Yizhou

DEVELOPMENT OF A COUPLED CFD—SYSTEM-CODE CAPABILITY
(WITH A MODIFIED POROUS MEDIA MODEL) AND ITS APPLICATIONS

TO SIMULATE CURRENT AND NEXT GENERATION REACTORS
BY
YIZHOU YAN
DISSERTATION
Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy in Nuclear Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2011
Urbana, Illinois
Doctoral Committee:
Professor Rizwan-uddin, Chair and Director of Research

Professor Barclay G. Jones
Professor James F. Stubbins
Doctor Nahil A. Sobh, Beckman Institute for Advanced Science and Technology
Abstract
Motivated by recent developments in the field of Computational Fluid Dynamics
(CFD) and recognizing the limitations on computing power, this dissertation is aimed
at combining the desirable features of system codes and CFD codes, thus elevating
the nuclear reactor thermal hydraulics simulation capabilities to address problems
that cannot be addressed with existing computational tools. The goal is achieved
by first implementing improved porous media models in a commercial CFD code
and then by judicious coupling of the CFD code with a coarse nuclear system code.
Computationally intensive CFD is used in spatial domains where the flow is expected
to be three-dimensional; whereas a system code is used to simulate regions where the
flow is expected to be one-dimensional or to simulate components such as pumps, etc.
Work accomplished in this dissertation can be divided into the following five parts.
• Test a commercial CFD code, FLUENT, by solving a nuclear-specific bench-
mark problem. Extend the porous media turbulence model in the CFD code
using User-Defined Functions (UDFs).
• Demonstrate the porous media simulating capability by a nuclear system using
a combined CFD model of clear flow and porous media flow(for core region).
• Propose a hybrid approach to couple a CFD code with a nuclear system code.
Develop the coupled CFD—system-code approach. Verify the coupled code
using a simple flow in a network of pipes.
ii
• Test the large scale application of the coupled CFD—system-code by model-
ing the Nuclear Steam Supply System (NSSS) of a Pressurized Water Reactor
(PWR).
• Demonstrate the potential of the coupled CFD—system-code for next genera-
tion reactors by applying it to a Gas Turbine - Modular Helium Reactor (GT-
MHR).
Part 1 is accomplished by implementing a modified k − ϵ turbulence model
for porous media in a CFD code (FLUENT) using UDFs. Transverse flow through
porous media is simulated with the extended CFD code. Results are compared with
experimental data.
In the second part of this thesis, the International Standard Problem (ISP)
No. 43, rapid boron-dilution experiment, is simulated using FLUENT to verify capa-
bility to model nuclear systems. Australian Replacement Research Reactor (RRR) is
modeled to demonstrate application of CFD, with porous media model for the reactor
core. The parameters for the porous media model are obtained through a series of
assembly level CFD simulations.
RELAP5-3D is introduced in Part 3 as the nuclear system code for coupled
CFD—System-code development. UDF feature of FLUENT is used to develop the
interface for this coupling effort. This innovative coupling approach is verified by
comparing the results of a simple transient flow problem obtained using the cou-
pled codes with the results from the CFD-only simulation and the system-code-only
simulation.
Part 4 is the first large scale application of the coupled CFD—system-code. A
simplified PWR NSSS is modeled by the coupled CFD—system-code approach devel-
oped in Part 3. Time-dependent three-dimensional reactor power profile is calculated
in a PWR transient scenario which investigates the spatial impact of the coolant
thermal mixing by using a specially developed discrete reactor kinetic model.
iii
In Part 5 simulation of reactor coolant system in the GT-MHR vessel is carried
out using the coupled CFD—system-code, demonstrating the potential of the coupled
CFD—system-code approach to Gen IV reactor design and optimization.
Thus, by implementing an improved porous media model in a CFD code, and
combining the best features of a CFD code and a nuclear system analysis code, a sim-
ulation capability has been developed to model three-dimensional effects in complete
integral systems with existing computational resources. The utility of this capability
has been demonstrated by applications to a PWR and to a GT-MHR. This coupled
CFD—system-code capability will be useful in developing better optimized reactor
designs by reducing reliance on conservative models and simulations.
iv
Acknowledgements
First, I would like to thank my advisor, Professor Rizwan-uddin for his continuous
guidance and encouragement throughout my years of study at University of Illinois. I
would also like to appreciate Professor Barclay Jones, Professor James Stubbins, and
Dr. Nahil Sobh for serving on my defense committee.
I wish to extend special recognition to financial support provided by the In-
novations in Nuclear Education and Infrastructure (INIE) program. INIE program
was established in 2001 by the Department of Energy’s Office of Nuclear Energy,
Science, and Technology (http://www.ne.doe.gov), to strengthen the university nu-
clear engineering education programs through innovative use of university research
and training reactors and encouraging partnerships between the universities and the
Department of Energy.
I would like to thank my office mates, Yuxiang Gu, Jianwei Hu, Daniel
Rock, Federico Teruel, and Quan Zhou for making a pleasant working environment,
Dr.Kyungdoo Kim for help in RELAP5-3D code usage, Dr. Nahil Sobh for providing
computation resource at NCSA.
I would also like to thank my wife, Xuhua Qin, for the support and encour-
agement during the hardest years.
v
Table of Contents
List of Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
List of Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2 Computational Fluid Dynamics and Porous Media Model 8

2.1 Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.2 Turbulence Models . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.3 Finite Volume Method . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Porous Media Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Existing Porous Media Model in FLUENT . . . . . . . . . . 17
2.2.2 Turbulence Treatment of Porous Media Model in FLUENT . . 18
Chapter 3 CFD Verification Cases for Nuclear Applications . . . . . . . . . 20

3.1 CFD Simulation of Rapid Boron-Dilution Transient . . . . . . . . . . 21
3.1.1 Rapid Boron-Dilution Transient . . . . . . . . . . . . . . . . . 21
3.1.2 CFD Model and Mesh . . . . . . . . . . . . . . . . . . . . . . 21
3.1.3 Results and Comparison . . . . . . . . . . . . . . . . . . . . . 23
3.2 Porous Media Model and CFD Simulation of Replacement Research
Reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.1 Replacement Research Reactor . . . . . . . . . . . . . . . . . 25
3.2.2 CFD Model and Mesh . . . . . . . . . . . . . . . . . . . . . . 26
3.2.3 CFD Porous Media Model . . . . . . . . . . . . . . . . . . . . 28
3.2.4 Fuel Assembly Level CFD Simulation . . . . . . . . . . . . . . 29
3.2.5 RRR Integral CFD Simulation . . . . . . . . . . . . . . . . . . 37
3.2.6 Parametric Study . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.7 Some Intermediate Conclusions for RRR . . . . . . . . . . . . 41
Chapter 4 Turbulence Porous Media Models . . . . . . . . . . . . . . . . . . . . . . . . 43

4.1 Macroscopic Turbulence Model for Porous Media Flow . . . . . . . . 44
4.1.1 Turbulence Porous Media Model for Nuclear Reactor Core . . 47
4.2 Verification of Turbulence Porous Media Model Implementation . . . 52
4.2.1 Modified k − ϵ Model for Porous Media Flow . . . . . . . . . . 52
4.2.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . 53
vi
4.2.3 CFD Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.4 Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Chapter 5 Nuclear System Code and Coupling Approach . . . . . . . . . . . 58

5.1 RELAP5-3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.1.1 Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1.2 Neutron Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2 Coupled CFD—System-Code Approach . . . . . . . . . . . . . . . . . 65
5.2.1 PVMEXEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2.2 Coupled CFD—System-Code Simulations Review . . . . . . . 70
5.2.3 Large Scale Simulation using Coupled CFD—System-Code . . 71
Chapter 6 CFD—System-Code Coupling using UDF . . . . . . . . . . . . . . . . 73

6.1 User-Defined Function . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2 Coupling Interface in Coupled CFD—System-Code . . . . . . . . . . 76
6.3 CFD—System-code Coupling using UDFs . . . . . . . . . . . . . . . 78
6.3.1 Semi-Implicit Coupling . . . . . . . . . . . . . . . . . . . . . . 78
6.3.2 Coupled Code Structure . . . . . . . . . . . . . . . . . . . . . 80
6.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Chapter 7 Coupled CFD—System-Code Validation . . . . . . . . . . . . . . . . . 82

7.1 CFD-only Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.2 RELAP-only Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.3 Pipe Flow Simulation by Coupled CFD—System-Code . . . . . . . . 85
7.3.1 Boundary Conditions and Coupling Boundary Interface . . . . 86
7.3.2 Time Step Size . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.4 Results and Comparison . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Chapter 8 Large Scale Simulation of Pressurized Water Reactor Using

Coupled Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.1 Pressurized Water Reactor . . . . . . . . . . . . . . . . . . . . . . . . 89
8.1.1 AP1000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.1.2 Large Scale Thermal Hydraulic Simulation for PWR . . . . . 91
8.1.3 RELAP Portion in Coupled Model . . . . . . . . . . . . . . . 91
8.1.4 CFD Portion in Coupled Model . . . . . . . . . . . . . . . . . 94
8.1.5 Coupled CFD—System-Code Model for PWR . . . . . . . . . 98
8.2 Reactor Power Profile Calculation . . . . . . . . . . . . . . . . . . . . 99
8.2.1 Point Reactor Kinetics . . . . . . . . . . . . . . . . . . . . . . 100
8.2.2 Temperature Coefficient of Reactivity . . . . . . . . . . . . . . 101
8.2.3 Spatial Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.2.4 Power Profile Calculation in Coupled Model . . . . . . . . . . 102
8.3 Coupled Transient Simulation . . . . . . . . . . . . . . . . . . . . . . 107
8.3.1 Transient Scenario . . . . . . . . . . . . . . . . . . . . . . . . 107
8.3.2 Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 109
vii
8.4.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.4.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Chapter 9 Coupled Simulation of Gas Turbine - Modular Helium Re-

actor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9.1.1 Gas Turbine - Modular Helium Reactor . . . . . . . . . . . . . 124
9.1.2 Thermal Hydraulic Challenges . . . . . . . . . . . . . . . . . . 128
9.2 Coupled CFD—System-Code Model . . . . . . . . . . . . . . . . . . . 129
9.2.1 CFD Model of Lower Plenum . . . . . . . . . . . . . . . . . . 130
9.2.2 RELAP Portion in Coupled Model . . . . . . . . . . . . . . . 133
9.2.3 Coupled CFDSystem-Code Model . . . . . . . . . . . . . . . . 135
9.3 Steady State Simulation and Results . . . . . . . . . . . . . . . . . . 137
9.3.1 GT-MHR Steady-State Conditions . . . . . . . . . . . . . . . 137
9.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
9.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Chapter 10 Summary, Discussion and Future Work . . . . . . . . . . . . . . . . . . 151

10.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
10.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Appendix A User-Defined Functions for Porous Media Model . . . . . . . 155

A.1 Porous Media Model Expansion UDFs . . . . . . . . . . . . . . . . . 155
Appendix B User-Defined Functions for Coupled CFD—System-Code158

B.1 UDF to Initialize RELAP5-3D . . . . . . . . . . . . . . . . . . . . . . 158
B.2 UDF to Execute RELAP5-3D . . . . . . . . . . . . . . . . . . . . . . 159
B.3 UDF for Coupling Calculation and Communication . . . . . . . . . . 161
B.4 UDF for Time Step Synchronizing . . . . . . . . . . . . . . . . . . . 162
B.5 UDF for Reactor Power Profile Calculation . . . . . . . . . . . . . . . 162
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
viii
List of Tables
3.1 Core pressure drop results from assembly level simulations with struc-
tural details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Comparison of core pressure drop results. . . . . . . . . . . . . . . . . 36
3.3 Natural circulation conditions for two different chimney sizes. . . . . 42
4.1 Model constants for macroscopic turbulence porous media model . . . 51
8.1 Nominal operational conditions of the AP1000 at full power . . . . . 95

8.2 Point reactor kinetics model parameters [Duderstadt and Hamilton,
1976] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
9.1 Coupling interface boundary specifications . . . . . . . . . . . . . . . 137

9.2 Steady-state thermal hydraulic conditions of GT-MHR [LaBar, 2002] 138
9.3 Average swirling number and temperature uniformity in the connecting
duct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
ix
List of Figures
1.1 Commercial nuclear power reactor evolution and naming convention

[US Department of Energy, 2002] . . . . . . . . . . . . . . . . . . . . 2
3.1 Two representations of the CFD model for rapid boron-dilution simu-
lation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Components and flow paths in ISP 43 test series A [NEA, 2001] . . . 23
3.3 Transient temperature and inlet velocity inlet conditions at cold leg A1 23
3.4 The average temperature at level 4 . . . . . . . . . . . . . . . . . . . 24
3.5 Front view of the Replacement Research Reactor [ANSTO, 2001] . . 25
3.6 Diagram of core, chimney, and reactor pool . . . . . . . . . . . . . . 27
3.7 Full scale CFD model of RRR . . . . . . . . . . . . . . . . . . . . . . 28
3.8 CFD model of core and chimney . . . . . . . . . . . . . . . . . . . . 29
3.9 CFD model of a fuel assembly . . . . . . . . . . . . . . . . . . . . . 30
3.10 Boundary conditions on the fuel assembly . . . . . . . . . . . . . . . 30
3.11 Velocity vector plot at assembly inlet (Box A in Figure 3.10) . . . . 31
3.12 Velocity vector plot at assembly outlet (Box B in Figure 3.10) . . . . 32
3.13 Pressure drop along an assembly . . . . . . . . . . . . . . . . . . . . 33
3.14 Pressure drops in assembly level CFD with complete structural details
and using two different sets of porous media parameters . . . . . . . 34
3.15 CFD model for the Replacement Research Reactor . . . . . . . . . . 38
3.16 Reverse flow in the chimney . . . . . . . . . . . . . . . . . . . . . . . 39
3.17 Four configurations for parametric study . . . . . . . . . . . . . . . . 40
4.1 Porous media REV for macroscopic average . . . . . . . . . . . . . . 46

4.2 Side view of channel flow with submerged vegetation . . . . . . . . . 54
4.3 Comparison of mean velocity term between experimental data
and
√ CFD results at x = 15.0 m . . . . . . . . . . . . . . . . . . . . . 55
′ ′
4.4 < u1 u1 > term in Reynolds stress comparison between experimental
data and CFD results at x = 15.0 m . . . . . . . . . . . . . . . . . . 56
4.5 − < u′1 u′3 > component of Reynolds stresses comparison between ex-
perimental data and CFD results at x = 15.0 m . . . . . . . . . . . . 57
5.1 Coupled simulation diagram for PVMEXEC . . . . . . . . . . . . . . 69

5.2 Coupled CFD-TRACE code scheme [Bertolotto et al., 2008] . . . . . 70
6.1 A conceptual coupled CFD—system-code model . . . . . . . . . . . 76

6.2 Coupling scheme by FLUENT UDF programming interfaces . . . . . 80
x
7.1 CFD model for the simple pipe . . . . . . . . . . . . . . . . . . . . . 83
7.2 RELAP nodeliztion for the simple pipe . . . . . . . . . . . . . . . . 84
7.3 Coupled CFD—system-code model for the simple network of pipes . 85
7.4 Mass flow rate comparison between FLUENT, RELAP-3D(ATHENA),
and coupled code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.1 Westinghouse AP1000 nuclear steam supply system [Schulz, 2006] . . 90

8.2 Nodilizaition of a typical pressurized water reactor core [INEEL RELAP5-
3D Group, 1999] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.3 RELAP nodalizaition of one loop of NSSS . . . . . . . . . . . . . . . 94
8.4 Mesh in a vertical plane of the reactor vessel . . . . . . . . . . . . . 95
8.5 Mesh in a horizontal plane of the reactor vessel . . . . . . . . . . . . 96
8.6 CFD model for 1/8 of a PWR assembly . . . . . . . . . . . . . . . . 96
8.7 Pressure loss across a PWR assembly (CFD simulations). . . . . . . 98
8.8 Diagram of coupled large scale PWR hydraulic system model . . . . 99
8.9 Reactor core with hot legs and cold legs(CFD part of the coupled
simulation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.10 Initial power distribution in the reactor core . . . . . . . . . . . . . . 109
8.11 Initial pressure contour in the reactor core . . . . . . . . . . . . . . . 110
8.12 Initial velocity magnitude contour in the reactor core . . . . . . . . . 111
8.13 Initial temperature contour in the reactor core . . . . . . . . . . . . 112
8.14 Temperature contour in the reactor core on the cross sectional plane
aligned with the hot legs, at t =10 sec . . . . . . . . . . . . . . . . . 115
aligned with the cold legs B1/A2, at t=10 sec . . . . . . . . . . . . . 115
aligned with the cold legs A1/B2 , at t=10 sec . . . . . . . . . . . . 116
8.17 Temperature contour in the reactor core on the horizontal cross sec-
tional planes, at t=10 sec . . . . . . . . . . . . . . . . . . . . . . . . 116
8.18 Temperature contours in the reactor core on the horizontal cross sec-
tional plane Z0, at t = 10 sec . . . . . . . . . . . . . . . . . . . . . . 117
tional planes Z3, at t = 10 sec . . . . . . . . . . . . . . . . . . . . . 118
8.25 Power density contours in the reactor core on the several horizontal
planes, at t = 10 sec . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
xi
8.26 Power density contours in the reactor core on the horizontal cross sec-
8.29 Reactor core power density contour on the section aligned with the
cold legs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
9.1 Gas turbine modular helium reactor thermal hydraulic system [LaBar,
2002] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2 Gas turbine modular helium reactor vessel [LaBar, 2002] . . . . . . . 126
9.3 GT-MHR graphite core section [LaBar, 2002] . . . . . . . . . . . . . 126
9.4 GT-MHR core graphite module [General Atomics, 2002] . . . . . . . 127
9.5 Isometric view of the GT-MHR CFD mesh . . . . . . . . . . . . . . 131
9.6 Top view of the GT-MHR CFD mesh . . . . . . . . . . . . . . . . . 131
9.7 Side view of the GT-MHR CFD mesh . . . . . . . . . . . . . . . . . 132
9.8 RELAP5-3D nodalization for the GT-MHR reactor vessel . . . . . . 134
9.9 Coupled model for the GT-MHR . . . . . . . . . . . . . . . . . . . . 135
9.10 Temperature contour at the lower plenum’s upper surface in the CFD
portion of the model. Temperature is in Kelvin. Highest temperature
is in the annular shaped region right below the annular core. . . . . 139
9.11 Velocity contour plot on the coupling interface between the core and
the lower plenum. Highest velocity is approximately 24 m/s. . . . . 140
9.12 Temperature contour plot on a vertical cross sectional plane in the
lower plenum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.13 Velocity magnitude contour plot on a vertical cross sectional plane in
the lower plenum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.14 Temperature contour plot on a vertical sectional plane . . . . . . . . 143
9.15 Velocity magnitude contour on a vertical sectional plane . . . . . . . 143
9.16 Temperature contour on a horizontal sectional plane . . . . . . . . . 144
9.17 Velocity magnitude contour on a horizontal sectional plane . . . . . . 144
9.18 Streamlines colored by temperature in the GT-MHR lower plenum and
the connecting duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
9.19 Streamlines colored by velocity magnitude in the GT-MHR lower plenum
and the connecting duct. . . . . . . . . . . . . . . . . . . . . . . . . 145
9.20 Velocity magnitude contour plot on the plane at the end of the duct
in the 3D CFD model, where the flow then continues in the simpler
RELAP model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
9.21 Swirl intensity contours on planes in the connecting duct and outlet
plane. The five contour planes from left to right on the top are the
planes from the lower plenum to the connecting duct outlet plane, as
shown in the 3D figure. . . . . . . . . . . . . . . . . . . . . . . . . . 147
9.22 Temperature contour on the connecting duct outlet plane . . . . . . 148
xii
9.23 Temperature contours on internal cross sectional planes in the con-
necting duct and the outlet plane. The five contour planes from left
to right on the top are the planes from vessel to the connecting duct
outlet plane, as shown in the 3D figure. . . . . . . . . . . . . . . . . 149
xiii
Chapter 1
Introduction
Nuclear energy is an indispensable source of energy. It also is a potential solution
to the global warming problem since nuclear power stations do not exhaust global
warming gases, such as carbon dioxide, to environment as fossil fuel burning power
plants do. Figure 1.1 shows the evolution of the nuclear power reactor technology.
Most nuclear power reactors in operation are categorized as Gen II reactors. Gen III
reactor technology, such as the AP1000, is being deployed in the new constructions
of nuclear power stations.
Generation IV (GEN IV) reactors are considered to be the nuclear reactor
technologies for the future (20 years out). Next generation, or GEN IV , reactors nor-
mally refer to six conceptual designs considered for the next generation of commercial
reactors. These designs can be divided based on the choice of coolant and neutron
spectrum. There are two thermal neutron spectrum systems (Very-High-Temperature
Reactor (VHTR) and Supercritical-Water-Cooled Reactor (SCWR)) with coolants
and temperatures that enable hydrogen or electricity production with high efficiency.
Three are fast neutron spectrum systems (Gas-Cooled (GFR), Lead-Cooled (LFR),
and Sodium-Cooled (SFR) fast reactors) that will enable more effective management
of actinides through recycling of most components in the spent fuel. The sixth GEN
IV design is the Molten Salt Reactor (MSR) [Marcus, 2000].
1
Figure 1.1: Commercial nuclear power reactor evolution and naming convention [US De-
partment of Energy, 2002]
The development of the next generation of nuclear reactors elevates the simu-
lation requirements to a higher level. Design tools for multi-dimensional, multi-phase,
multi-physics simulations are desirable. The need for safer and more economically
efficient designs of next generation reactors poses many challenges to the thermal
hydraulic engineers who are developing simulation tools for nuclear reactor design,
performance optimization, and safety analysis. In the mean time, thermal hydraulic
simulation capabilities have improved dramatically to meet several aspects of the
new requirements. Significant progress has been made in both software and hard-
ware. Large scale, parallel, Computational Fluid Dynamics (CFD) is also gaining
acceptance by the nuclear industry. The development process of next generation re-
actors needs to take advantage of the large scale multi-dimensional thermal hydraulic
simulation capabilities in the design of the advanced reactor systems.
CFD has been widely used as design and optimization tools in automobile
manufacture industry, aeronautical engineering, chemical engineering, and even bio-
logical engineering. But as a simulation and design tool, CFD has not been widely
2
used by nuclear engineers. Lack of acceptability and use of CFD by the nuclear
community can be traced to three major factors. The first one is the limited valida-
tion of CFD for two-phase flow simulation. Majority of nuclear reactors in operation
are Light Water Reactors (LWRs) which require a significant capability and reli-
ability of two-phase models in the simulation of flow dynamics and heat transfer.
Secondly, due to the regulatory requirements in the industry, CFD codes are not
yet certified for licensing safety analyze of nuclear systems, whereas codes like RE-
LAP5 or GOTHIC, which have been developed exclusively for nuclear applications
and validated extensively, are accepted by the U.S. Nuclear Regulatory Commission
(NRC). Thirdly, CFD simulation of nuclear systems requires extensive computational
resources [Rizwan-uddin and Yan, 2004].
CFD is however slowly gaining acceptance by researchers and engineers in nu-
clear industry as the three aforementioned factors are gradually addressed. Moreover,
only one of the GEN IV reactor designs uses water as coolant. Hence CFD limitation
due to two-phase modelling will no longer be a hurdle in the design and analysis of
reactors using non-water coolant, such as helium. CFD will hence be a powerful tool
for the engineers to develop GEN IV reactors. Second factor favoring CFD is a grad-
ual acceptance by regulatory agencies of CFD-based results for certain applications.
As a result, increasing numbers of CFD simulations of nuclear systems are being car-
ried out. CFD simulations of reactor downcomer, lower plenum, upper plenum, and
coolant channels have been reported in literature. Interest has been growing due to
the recognition that CFD could supply more information on flow and heat transfer
than traditional nuclear system codes do, and hence may provide adequate margins
while avoiding unnecessary conservatism in design. With these merits, nuclear design-
ers as well as regulatory authorities such as the NRC are increasingly recognizing the
importance of CFD. Thirdly, with computing power increasing dramatically in recent
years, especially as the computational cluster technology becomes mature, large scale
3
CFD simulation of future nuclear reactor designs will be an irresistible temptation to
nuclear engineers.
Increasing numbers of CFD simulations for nuclear engineering applications
have been carried out using commercially available CFD codes. There are many
advantages of using commercial CFD codes. Other than the necessary models require
for thermal hydraulic simulation in the nuclear industry, commercial CFD codes have
user friendly interfaces and comprehensive documentation which make the codes easy
to use. The technical support for a code is also a necessary consideration. With
the help of CFD mesh generation software, very complex flow geometry could be
modelled and simulated. CAD designs can be imported into most commercial meshing
software. This can be a very useful capability in streamlining the design and analysis
process since, like other industries, nuclear industry is also already making use of
CAD software tools to design the next generation of nuclear reactors. Moreover
commercial CFD codes have large user groups, which may provide valuable experience
to nuclear applications. After more than almost twenty five years of development,
many commercial CFD codes, such as FLUENT, CFX, and STAR-CD, are available
to the nuclear engineers.
CFD investigations of existing LWRs with different level of complexities have
been reported in literature. For example, CFD validation studies were performed by
simulating the fluid mixing and flow distribution in a primary loop of a Pressurized
Water Reactor (PWR) [Rohde et al., 2007]. A three-dimensional CFD analysis was
performed to study the flow characteristics in the reactor vessel downcomer during the
late reflood phase of a postulated Large-Break Loss-of-Coolant Accident(LBLOCA)
[Kwon et al., 2003]. A CFD simulation process for fast reactor fuel assemblies was
developed by the researchers at Idaho National Lab [Hamman and Berry, 2010].
Commercial CFD code, CFX was used to investigate natural convection in passive
heat removal system in advanced designs of nuclear power plants [Krepper et al., 2002].
4
CFX was also used to model the heat transfer in the fuel bundles of a supercritical
water reactor [Shang, 2009]. A rectangular slit impactor was also simulated using
CFX [Hari et al., 2005]. The high cycle temperature fluctuations problem caused by
thermal stripping in a mixing tee was benchmarked using FLUENT [Hu and Kazimi,
2006]. Yadigaroglu summarized the technical trends and needs in LWR simulation,
and stated the necessary consideration of turbulence in nuclear system codes. He also
stressed the need for coupled multi physics simulations of nuclear system [Yadigaroglu
et al., 2003]. While these recommendations for the CFD applications were made for
LWRs, they are equally valid for GEN IV reactors. Already, for GEN IV reactor
thermal hydraulic simulations, three-dimensional flow and heat transfer modelling for
Pebble Bed Modular Reactor (PBMR) has been done using CFD [Becker and Laurien,
2003]. An attempt at systematic CFD simulation of PBMR has also been made
[du Toit et al., 2006]. CFD will likely play a very significant role in the simulation
and analysis of GEN IV reactors. However, as mentioned earlier, despite significant
progress in CFD, the technology is not yet capable of routinely simulating large
integral systems. Although some attempts have been made to system-level CFD
analysis of nuclear reactors, large scale CFD simulation of nuclear reactor system is
not a task to be done routinely with currently available resources. Hence, a coupled
CFD—system-code approach is necessary.
There are additional limitations that may restrict the application of CFD
codes to GEN IV reactors. For example, the general porous media model in CFD
codes is not suitable to model the turbulent flow in a nuclear reactor core. Due
to extreme flow conditions in reactor cores, turbulence in porous media should be
taken into consideration. Turbulent flows are characterized by fluctuating velocity
fields. These fluctuations mix transported quantities such as momentum, energy, and
chemical concentration, and cause the transported quantities to fluctuate as well.
Therefore, accurate turbulence modeling in the regions of the reactor cores using
5
porous media approximation is important.
Recognizing the limitations of the computational tools and computing power,
this dissertation is aimed at advancing the simulation capabilities in two specific
ways, followed by validation and extensive applications to realistic nuclear power
plant systems including a PWR and a Gas Turbine-Modular Helium Reactor (GT-
MHR). Capabilities of the simulation tools are advanced by first adding models for
porous media flow, and then by developing a methodology to couple the CFD code
with the system code thus allowing the best features of both to be used. This coupled
simulation capability with multiple levels of details could be applied to improve the
thermal hydraulic designs of the existing and Gen IV nuclear reactors.
“Nuclear system codes” usually refer to a family of thermal hydraulic sim-
ulation programs, such as RELAP series, TRAC series, and RETRAN, widely used
by nuclear engineers. RELAP5-3D is coupled in this thesis with a commercial CFD
code, FLUENT. RELAP5-3D is a best estimate simulation code developed at the
Idaho National Engineering & Environmental Laboratory (INEEL) for U.S. NRC.
It has been used extensively, and accepted as a safety analysis tool by the nuclear
industry.
Thermal hydraulic designs of Gen IV reactors are diverse, and a single uni-
versal simulation tool for them is not feasible currently. Therefore, efforts in this
dissertation are focused on a few specific designs. Some of the developments will have
broader applications as well as to non-nuclear applications.
The remainder of this dissertation is organized as follows. The CFD code
used in this thesis is described in the second chapter. The existing porous media
model available in the CFD code is also reviewed in that chapter. The limitation of
the existing porous media model is discussed, and the need for improved turbulence
porous media model for nuclear applications is presented. Two applications of the
CFD code for nuclear systems are reported in Chapter three. The first one is the
6
transient simulation of the Rapid Boron-Dilution (RBD) experiment carried out at
University of Maryland at College Park. The second test case is the CFD study
of the Replacement Research Reactor (RRR) with the reactor core represented as a
porous media. The extension of turbulence modelling in porous media flow in the
CFD code is reported in Chapter four. A test simulation to validate the methodology
of the new turbulence model is also presented. CFD results obtained with turbulence
porous media model are compared with experimental data. The nuclear system code,
RELAP5-3D, is reviewed in Chapter five, and its applications and limitations are
discussed. The concept of coupled CFD—system-code and the coupling methodology
used in RELAP5-3D are also reviewed in that chapter. Currently available coupling
code PVMEXEC in RELAP5-3D package is introduced in Chapter five. User-Defined
Functions (UDFs) feature of FLUENT is introduced in Chapter six. The coupling
methodology based on UDFs is also developed in this chapter. The validation of the
coupled CFD—system-code is reported in Chapter seven. Coupled CFD—system-
code’s results for a simple forced circulation loop problem are compared with results
from CFD-only simulation and system-code-only simulation. Large scale applications
of coupled CFD—system-code to a PWR and a GT-MHR are reported in Chapter
eight and Chapter nine, respectively. This thesis is summarized and recommendations
for future work are presented in the last Chapter.
7
Chapter 2
Computational Fluid Dynamics
and Porous Media Model
2.1 Computational Fluid Dynamics
Computational Fluid Dynamics (CFD) uses numerical methods to solve the governing
equations for fluid flow and heat transfer. CFX, STAR-CD, and FLUENT are the
most widely used commercial CFD codes both in industry and in academia. FLUENT
is used in this thesis work.
Licence is available for parallel computation using FLUENT for up to 128
processors on the supercomputer clusters at the National Center for Supercomputing
Applications (NCSA). In addition to the high performance computational resource at
NCSA, an in-house 16-processor cluster capable of performing parallel FLUENT sim-
ulations has been constructed to ensure computational resources for CFD simulations
with extensive simulation time.
FLUENT can simulate laminar and turbulent flows in complex geometries.
Meshes generated by other meshing softwares can be imported. Complicated models
can be created and meshed using a package called GAMBIT, a meshing pre-processor
in the FLUENT package.
8
Comprehensive modelling capabilities of FLUENT enable researchers to in-
vestigate a wide range of engineering flow problems including Newtonian or non-
Newtonian flows, viscous or inviscous flows, incompressible or compressible flows,
laminar or turbulent flows, single-phase or multi-phase flows, etc. Moreover, both
steady-state and transient simulations can be performed.
Various modelling features are available to apply FLUENT to specific appli-
cations. These include porous media model, lumped parameter models (fan, pump,
and porous jump), and Boussinesq approximation. Heat transfer can also be modelled
with natural, forced, and mixed convection.
Multiphase flow and free surface flow can also be modelled using FLUENT.
These capabilities are very important in the analysis of vapor-liquid flows in the nu-
clear systems which use water as coolant. For multiphase CFD simulations, FLUENT
has the Volume-of-Fluid (VOF), mixture, and Eulerian models, as well as the discrete
phase model (DPM) with liquid drops tracking. The applications of multiphase flows
include boiling, condensation, sprays, separation, and cavitation. Multiphase model-
ing capability using CFD has great potential in nuclear engineering since most reac-
tors in operation are light water reactors, which could have two-phase flow in normal
operation (Boiling Water Reactor (BWR)) or under accident conditions (Pressurized
Water Reactor (PWR)). However, multiphase CFD is very computationally costly
and far from mature due to the limitation of the models in the commercial CFD
codes. Multiphase flow problems in nuclear systems will not be addressed in this
thesis.
FLUENT is a very powerful tool to investigate the multi-dimensional flow
and heat transfer phenomena in nuclear engineering applications.
9
2.1.1 Governing Equations
FLUENT numerically solves the governing equations for fluid flow using finite vol-
ume methods. Mass transport equation and three-dimensional momentum transport
equations are the fundamental governing equations solved in the CFD code. Energy
equation is included for problems involving heat transfer or compressible flow. Tur-
bulence models require the transport equations for the turbulence flow variables in
addition to the Navier-Stokes equations. The chemical species transport equations
can be solved for applications with chemical reaction and species transport. Multi-
phase CFD requires additional governing equations for each phases. The number of
extra governing equations for the other phases depends on which multiphase model
is employed.
Mass Conservation Equation
Equation 2.1 is the time dependant mass conservation equation for both compressible
and incompressible flows,

∂ρ
+ ∇ · (ρ⃗v ) = Sm (2.1)
∂t
where Sm is the mass source term in the flow domain.
Momentum Equation
Equation 2.2 is the momentum conservation equation, or the Navier-Stokes equations,
used in FLUENT,
∂
(ρ⃗v ) + ∇ · (ρ⃗v · ⃗v ) = −∇p + ∇ · (τ ) + ρ⃗g + F⃗ (2.2)
∂t
in which, p is the static pressure, τ is the stress tensor defined in Equation 2.3, ρ⃗g
is the gravitational body force term, and F⃗ is the external body force. The stress
tensor τ is defined as
[ ]
2
τ = µ (∇⃗v + ∇⃗v ) − ∇ · ⃗v · I
T
(2.3)
3
10
where µ is the molecular viscosity, and I is the unit tensor.
The body force term, F⃗ , can also contain other model-dependent source
terms such as porous-media source term and other user-defined source terms. Many
application-specific flow dynamic models are incorporated using the source term (last
term) of Equation 2.2.
Energy Conservation Equation
Energy conservation equation in FLUENT is written as

( )
∂ ∑
(ρE) + ∇ · (⃗v (ρE + p)) = ∇ · keff ∇T − hj J⃗j + (τ eff · ⃗v ) + Sh (2.4)
∂t j
where keff is the effective conductivity (k + kt , where kt is the turbulent thermal
conductivity, defined according to the turbulence model being used), hj is the enthalpy
of species j, and J⃗j is the diffusion flux of species j. The first three terms on the
right-hand side of Equation 2.4 represent energy transfer due to conduction, species
diffusion, and viscous dissipation, respectively. Sh includes all volumetric heat sources
defined by users.
In Equation 2.4, E is defined as
p v2
E =h− + (2.5)
ρ 2
where h is enthalpy.
For the standard k- ϵ model, the effective thermal conductivity is given as
cp µt
keff = k + (2.6)
Prt
where k, in this case, is the fluid thermal conductivity. The second term on the right
hand side is referred to as turbulent thermal conductivity. The default value of the
turbulent Prandtl number is 0.85. FLUENT allows the value of the turbulent Prandtl
number in the turbulence model to be adjusted.
11
2.1.2 Turbulence Models
Turbulence modelling is vital in CFD because most flows in nature are turbulent
flows. It is generally accepted that the turbulence phenomena can be modeled using
the Navier-Stokes equations. Simulation of turbulent flows requires computational
mesh with extremely high resolution to cover the whole range of the characteristic
length scales. Direct Numerical Simulation (DNS) to solve the Navier-Stokes equa-
tions has been carried out only for turbulent flows with low Reynolds numbers in
simple geometries [Moser et al., 1999]. However, using DNS to solve the Navier-
Stokes equations for turbulent flows with high Reynolds numbers is still one of the
most challenging problems in computational science. Solving the Navier-Stokes equa-
tions for engineering applications using DNS method is impossible using currently
available computational resources. Practically, less computational resource demand-
ing turbulence models are employed in engineering CFD applications.
A commonly accepted turbulence modelling approach is to solve the Reynolds-
Averaged Navier-Stokes (RANS) equations. If the velocity in the Navier-Stokes equa-
tions is decomposed into the sum of mean velocity term and turbulence fluctuation
term, the Navier-Stokes equations could be rearranged into the form of the RANS
equations,
[ ( ) ]
∂ ūi ∂ ūj ūi ∂ ∂ ūi ∂ ūj ′ ′
ρ +ρ = −p̄δij + µ + − ρui uj + ρf¯i (2.7)
∂t ∂xj ∂xj ∂xj ∂xi
Equation 2.7 is the RANS, or time-averaged momentum equation of mean flow in
index notation. The turbulent flow velocity, ui , in the Navier-Stokes equations is
denoted as sum of mean velocity, ūi , and turbulence fluctuation velocity, u′i , as
ui = ūi + u′i (2.8)
in which ūi is the mean velocity obtained after time averaging over a time window
which is large enough for velocity fluctuation term, u′i , to vanish statistically. The
12
variables with overhead bar are the mean values obtained after the time averaging.
Equation 2.7 governs the mean velocity field of turbulent flow.
The last term, u′i u′j , in the RANS equations 2.7 is referred to as Reynolds
stress. Although Reynolds stress is in the form of stress, it is not strictly a stress as
molecular viscous stress. The transport equation for Reynolds stress term could be
derived through mathematical transform of the RANS equations. However, higher
order momentum terms of the turbulent fluctuation velocity are introduced into the
equation. Therefore, the governing equation set is still not closed. Different closure
approaches of the RANS governing equations set lead to different turbulence models.
Although different levels of closure of RANS equations could be proposed, in most
engineering CFD simulations Reynolds stress is modelled by the eddy viscosity or
turbulence viscosity, νT .
Turbulence Model in FLUENT
The turbulence models available in FLUENT vary from one-equation turbulence mod-
els, such as Spalart-Allmaras model, to the computationally expensive Large Eddy
Simulation (LES). The options available in FLUENT also span a broad range of ap-
plications, and they can simulate many physical phenomena with turbulent flow, such
as buoyancy, compressibility, and other user-defined turbulence source terms.
Two-equation turbulence models are most commonly used for engineering
applications. Extra effort has been dedicated to accurately model near-wall turbu-
lence by using extended wall functions or enhanced wall treatment models in these
two-equation models. Two-equation models such as k − ϵ model and k − ω model
are commonly used in industry. Two-equation turbulence modelling is a very active
research area, and new two-equation models are still being developed. Turbulence
models with special consideration for nuclear thermal hydraulic systems are available
in literature.
13
By definition, two-equation models include two additional transport equations
to represent turbulence. This approach allows a two-equation model to account for
the effects like convection and diffusion of turbulent kinetic energy.
First variable in the two governing equations is generally the turbulence ki-
netic energy, k. The second transported variable depends on the specific two-equation
model. Common choices are the turbulent dissipation, ϵ, for the k − ϵ model, or the
specific dissipation, ω, for the k − ω model. The second variable is the variable
that determines the scale of the turbulence (length-scale or time-scale), whereas the
turbulence kinetic energy, k, determines the turbulence intensity.
Considering the limitations on the computational resources, the two-equation
k − ϵ turbulence model is commonly used in CFD, and is used through out this thesis.
k−ϵ model is a two-equation turbulence model based on the RANS equations.
It is a semi-empirical model based on transport equations for turbulence kinetic en-
ergy, k, and turbulence kinetic energy dissipation rate, ϵ. The additional transport
equations for turbulence kinetic energy and turbulence energy dissipation rate are
[( ) ]
∂ ∂ ∂ µt ∂k
(ρk) + (ρkui ) = µ+ + Gk + Gb − ρϵ − YM + Sk (2.9)
∂t ∂xi ∂xj σk ∂xj
[( ) ]
∂ ∂ ∂ µt ∂ϵ ϵ ϵ2
(ρϵ)+ (ρϵui ) = µ+ +C1ϵ (Gk + C3ϵ Gb )−C2ϵ ρ +Sϵ (2.10)
∂t ∂xi ∂xj σϵ ∂xj k k
where Gk represents the generation of turbulence kinetic energy due to the mean
velocity gradients. Gb is the generation of turbulence kinetic energy due to buoyancy.
YM represents the contribution of the fluctuating dilatation in compressible turbulence
to the overall dissipation rate. C1ϵ , C2ϵ , and C3ϵ are constants. σk and σϵ are the
turbulent Prandtl numbers for k and ϵ, respectively. Sk and Sϵ are user-defined source
terms.
The turbulent viscosity is modelled as

k2
µt = ρCµ (2.11)
ϵ
14
where Cµ is a model constant. According to the turbulent-viscosity hypothesis,
Reynolds stress in the RANS equations could be modelled as

( )
′ ′ 1
ρui uj = µt Sij − Skk δij (2.12)
3
where Sij is the mean strain rate defined by

( )
1 ∂ ūi ∂ ūj
Sij = + (2.13)
2 ∂xj ∂xi
The production term of turbulence kinetic energy in Equation 2.9 is given by
∂uj
Gk = −ρu′i u′j (2.14)
∂xi
Equation 2.14 could be rearranged as
Gk = µt S 2 (2.15)
√
where S is the modulus of the mean rate-of-strain tensor, defined as S ≡ 2Sij Sij .
The model constants C1ϵ , C2ϵ , Cµ , σk , and σϵ have the following default values
[Pope, 2000] :
C1ϵ = 1.44, C2ϵ = 1.92, Cµ = 0.09, σk = 1.0, σϵ = 1.3 (2.16)
These default values have been determined from experiments with air and
water for fundamental turbulent shear flows including homogeneous shear flows and
decaying isotropic grid turbulence. They have been found to work fairly well for a
wide range of wall-bounded and free shear flows. The default constant values can
be modified to fine tune the k − ϵ model for some particular cases. Equation 2.9
and Equation 2.10 are solved in addition to the mass conservation equation and the
momentum equations.
2.1.3 Finite Volume Method
Finite volume method is a numerical discretization scheme for partial differential
equations. Most commercial CFD codes, including FLUENT, use finite volume
15
method to solve the governing equations numerically. Similar to the finite difference
method, values are calculated at discrete locations on a meshed geometry.
Finite volume refers to the small volume surrounding each node point on a
mesh. In the finite volume method, volume integrals in a partial differential equation
that contain a divergence term are converted to surface integrals, using the divergence
theorem. These terms are then evaluated as the fluxes on the surfaces of each finite
volume.
Finite volume method is inherently conservative. The changing rate in a
control volume is the balance between the flux entering a given volume, the flux
leaving that volume, and the sources in the volume. Boundary conditions can be
applied noninvasively.
Another advantage of the finite volume method is that it does not require
a structured mesh (although a structured mesh can also be used) and can be easily
formulated for unstructured meshes. Using unstructured meshes, complex geometry
could be discretized and modelled with moderate effort.
2.2 Porous Media Model
Porous media in general refers to a material consisting of solid matrix with an inter-
connected void [Nield and Bejan, 1992]. In CFD simulation, porous media model does
not resolve the flow field at every point at the microscopic scale. Instead, equivalent
spatially-averaged flow field is determined. The porous media effects on the macro-
scopic flow (such as pressure drop, turbulence intensity) are incorporated using extra
momentum loss terms in the Navier-Stokes equations and in the turbulence model
equations.
Although CFD simulations of flow domains with hundreds of millions of cells
can be carried out on the most advanced supercomputers, these are not enough to
16
capture the structural details of any GEN IV nuclear reactor, such as Gas Turbine-
Modular Helium Reactor (GT-MHR) which has thousands of flow channels in graphite
blocks, or Pebble Bed Modular Reactor (PBMR) which has approximately 452,000
tennis ball size fuel pebbles in the reactor core. Porous media modelling is needed to
reduce the number of cells needed to model the fine structure and relax the compu-
tational requirement.
Coolant flow through hundreds of channels with transverse cross flow, or
through pebble bed reactor core in large scale nuclear systems is very complicated
and can not be modelled in detail with the computational resource currently available
to average researchers. But a nuclear reactor core can be modelled as a porous media
with special pore structure.
2.2.1 Existing Porous Media Model in FLUENT
FLUENT provides a porous media model to simulate flow through porous media zone.
The effect of the porous media on mean flow is included via an extra flow resistance
term in the momentum equation.
The extended Darcy’s law,

∂p µ −1/2
=− · vj − cf Kij ρf (vk · vk )−1/2 · vj (2.17)
∂xi Kij
relates mean flow velocity and pressure drop to model porous media. Equation 2.17
is also called Forchheimer equation. Permeability Kij is a second-order tensor which
depends on the the geometry of the media only, and cf is a dimensionless form-drag
constant. In FLUENT, default modelling of porous media model is carried out by a
momentum source term in the Navier-Stokes equations

( 3 )
∑ ∑3
1
Si = − Dij µvj + Cij ρvmag vj (2.18)
j=1 j=1
2
where Si is the source term that appears in the ith (x, y, or z) momentum equa-
tion. Dij and Cij are prescribed matrices, called viscous resistance factor and inertial
17
resistance factor. They are defined by the user for specific porous media cases.
The default porous media energy equation in FLUENT is

[ ( ) ]
∂ ∑
(γρf Ef + (1 − γ)ρs Es )+∇·(v(ρf Ef + p)) = ∇· kef f ∇T − hi Ji + (τ · v) +Sfh
∂t i
(2.19)
where Ef is the total fluid energy; Es is the total solid medium energy; γ is porosity
of the porous medium; keff is effective thermal conductivity of the medium, which is
the mass averaged of fluid phase and solid porous medium phase; Sfh is fluid enthalpy
source term; J is mass flux or diffusion flux; and τ is the stress tensor.
The effective thermal conductivity in the porous medium, keff , is computed in
FLUENT as the volume average of the fluid conductivity and the solid conductivity
as
keff = γkf + (1 − γ)ks (2.20)
where γ is porosity of the medium; kf is fluid phase thermal conductivity (including
the turbulence contribution; kt ); and ks is the solid medium thermal conductivity.
2.2.2 Turbulence Treatment of Porous Media Model in FLU-
ENT
The flow in porous media is laminar in most situations. This is largely due to the fact
that most flows in porous media, such as ground water flow, have low flow velocities
and small hydraulic diameters. As a result, the general microscopic turbulence models
such as standard k − ϵ or standard k − ω models are not appropriate to be applied
to the turbulent flow in porous media without modification.
The general microscopic turbulence model could be used in porous media
zones. FLUENT solves the transport equations for turbulence quantities in the porous
zone. In the default approach, turbulence in the porous zone is treated as though the
solid medium has no effect on the turbulence generation or dissipation rates. This
18
assumption may be reasonable if the medium’s permeability is quite large and the
geometric scale of the porous medium does not interact with the mean scale of the
turbulent eddies. The default option implies that turbulence will be computed in
the porous region just as in the bulk fluid flow. However, in other instances, porous
media may suppress the turbulence intensity.
In micro-channel flow applications of porous media model, the flow in porous
media zone is laminar due to the small length scale of the open flow path. If using two-
equation turbulence model, users can also suppress the effect of turbulence in a porous
region by setting the turbulent contribution to viscosity, µt , equal to zero. When
this option is chosen, FLUENT solver will transport the inlet turbulence quantities
through the medium, but their effect on the fluid mixing and momentum will be
ignored within the porous media zone. In addition, the generation of turbulence will
be set to zero in the medium. This modelling strategy is enabled by turning on the
laminar zone option in the panel for fluid zone. Enabling this option implies that
turbulent viscosity, µt , is zero and that generation of turbulence will be ignored in
the porous zone.
The aforementioned turbulent flow treatments are the approaches in the
porous media model available in FLUENT. They are appropriate for cases in which
porosity is either very large (underground flow) or very small (open flow with sparse
obstacles). But these morphological assumptions are not valid for nuclear reactors.
The coolant flow in a nuclear reactor core is turbulent flow in the presence of the core
solid structure. The default porous media model in the commercial CFD code does
not contain a satisfying treatment of turbulence. It is hence desirable to implement a
porous media model with adequate turbulence treatment in the CFD code to simulate
flows through nuclear reactor cores.
19
Chapter 3
CFD Verification Cases for Nuclear
Applications
The CFD code and the default porous media model in FLUENT are tested in this
chapter. Two nuclear thermal hydraulic problems solved using the CFD code are
reported. The first one is the simulation of the 2×4 Rapid Boron-Dilution (RBD)
transient test on University of Maryland 2×4 thermal hydraulic loop (UM 2×4 Loop)
[NEA, 2001]. The results demonstrate that FLUENT is capable of modeling the time-
dependent 3D flows in a large scale nuclear system. The second one is CFD simulation
of the Replacement Research Reactor (RRR) at Australian Nuclear Science and Tech-
nology Organization (ANSTO) [ANSTO, 2001]. The default porous media model in
FLUENT is used to model the reactor core of the RRR system. The methodology
to calculate the porous media model parameters based on detailed heterogeneous
simulations is described in the RRR application. Assembly level CFD results help
determine the porous media model parameters for the porous media simulation of the
RRR integral model.
20
3.1 CFD Simulation of Rapid Boron-Dilution Tran-
sient
Boron, strong neutron absorber, is used in nuclear reactors to control reactivity.
Accurate prediction of flow mixing of different boron concentration streams in the
downcomer and lower plenum is important for nuclear safety analysis. In the set of
International Standard Problems (ISP), a RBD problem has been established to test
the simulation capabilities of computer codes (ISP No. 43) [NEA, 2001]. To test and
establish the methodology for CFD simulations of nuclear specific thermal hydraulic
problems, the transient test A of the RBD problem is solved using FLUENT.
3.1.1 Rapid Boron-Dilution Transient
ISP No. 43 is meant to evaluate the nuclear industry’s capabilities to simulate fluid
dynamics aspects of RBD transients. Four sets of experiments of ISP No. 43 were
carried out on UM 2×4 Loop. Ten organizations from eight countries used CFD to
simulate the experiment. Several commercial as well as in-house codes were used by
different teams to simulate the experiments. Good agreement between simulation
and experimental data was obtained. It was concluded in the ISP No. 43 report that
“the most general finding of ISP No. 43 is that there is no reason to preclude the
use of CFD codes from investigating prototype PWR boron-dilution transients and,
more broadly, transients that involve single phase mixing” [NEA, 2001].
3.1.2 CFD Model and Mesh
The UM 2×4 loop is a 1/7 scale thermal hydraulics model of a commercial Pressurized
Water Reactor (PWR). The CFD model simplified the flow paths and components
in the experimental facility. The CFD model for the reactor vessel in the UM 2×4
loop is illustrated in Figure 3.1. The geometry and mesh of the CFD model is gener-
21
ated from the design draft of the UM 2×4 loop using GAMBIT. The mesh contains
approximately 1 million cells. Transient CFD simulation of experiment test series A
has been carried out.
(a) transparent surface (b) frame wire
Figure 3.1: Two representations of the CFD model for rapid boron-dilution simulation
Test series A is designed to study the mixing of a single flow-front as it enters
the downcomer through cold leg A1 from an external source and leaves the system
through an opening at hot leg A. Flow in test series A is characterized by a single
inlet path and a single outlet path for the downcomer flow. Figure 3.2 shows the top
view of the participating components and flow paths.
In the experiment, temperature is measured and recorded to represent the
boron concentration. The measured temperature and inlet velocity at cold leg A1 are
plotted in Figure 3.3. A single mixing front condition is achieved when the injected
front volume is large enough.
The CFD model geometry was created based on the drawings and dimensions
from [NEA, 2001]. The CFD mesh is predominantly hexahedra while tetrahedra
mesh is also used in transitional mesh zone. The UM 2×4 loop does not have a
22
Figure 3.2: Components and flow paths in ISP 43 test series A [NEA, 2001]
Figure 3.3: Transient temperature and inlet velocity inlet conditions at cold leg A1
complicated reactor core, therefore the CFD model has relatively small mesh size.
k − ϵ model is used. The transient simulation which led to the results presented in
the following section took approximately 24 CPU hours on a work station with single
XEON processor and 2 GB memory.
3.1.3 Results and Comparison
The experimental data and CFD results for the transient profile of average temper-
ature at level 4 are compared. The level 4 thermo couples are located at 616 mm
23
below the inlet and outlet level. There are 24 thermal couples, 15 degrees apart, in
the azimuthal direction at level 4. The average temperature transient plot of reactor
downcomer at level 4 is shown in Figure 3.4. The results from three CFD simulations
with different flow properties are included in Figure 3.4. Good agreement between
CFD results and experimental data is observed. Note that this is a component specific
experiment, which can be simulated using the commonly available CFD capabilities.
350
Experimental Data
CFD results with Boussinesq approximation
CFD resutls with incompressible air
CFD results with incompressible water

Temperature (K)
300
0 50
Time (sec)
Figure 3.4: The average temperature at level 4
3.2 Porous Media Model and CFD Simulation of
Replacement Research Reactor
A CFD simulation of flow under natural circulation conditions in the ANSTO Re-
placement Research Reactor (RRR, or OPAL)[ANSTO, 2001] has been carried out.
Design optimization studies have also been carried out. A model consisting of the
core, the chimney, and the pool was simulated. In order to test the methodology
of porous media modelling, the reactor core in the integral model was represented
24
as a porous medium. Parameters for the porous media model were obtained from
complete CFD analysis of single fuel assembly. The default porous media model
in FLUENT was employed for this simulation. The methodology developed in this
simulation exercise is employed in the course of this thesis work.
3.2.1 Replacement Research Reactor
Replacement Research Reactor is a multi-purpose open-pool reactor. The cooling
system of RRR, including passive heat removal mechanism, has some unique design
features. The cooling system is composed of reactor pool, submerged coolant flow
channel, coolant pumps, and service pool which cleans the coolant before it is pumped
back to the reactor pool. The cooling system operates under forced or natural circu-
lation. Figure 3.5 shows the front view of the RRR reactor pool [ANSTO, 2001].
Figure 3.5: Front view of the Replacement Research Reactor [ANSTO, 2001]
The 20 MWth reactor core is submerged under approximately 13 meters of
water. Coolant is forced to circulate and remove the heat from the reactor core
25
during normal operating conditions. 90% of the coolant flow goes upward through
the reactor core for cooling purpose, while remaining 10% goes downward through
the upper chimney to keep the upper part of reactor pool free of radioactive material.
In a blackout accident, residual heat is expected to be carried away by nat-
ural circulation. After reactor shutdown, the control valves are switched to allow
the establishment of natural circulation in the core, in the reactor pool, and in the
chimney above the reactor core. The flow in the chimney is upward under natural
circulation conditions, which is opposite of the flow direction under normal forced
cooling operating conditions.
The goal of this exercise is to determine the rise in water temperature as it
passes through the core under natural circulation condition. This is accomplished in
two steps.
a) A set of assembly level CFD simulations with full structural details. Results
are used to calculate the parameters in the porous media model.
b) An integral CFD simulation of the core, the riser and the pool with porous
media model employed to model the core. Values of the parameters used in the porous
media model are those determined in step (a).
3.2.2 CFD Model and Mesh
GAMBIT 2.1 (in FLUENT CFD package) is used to create the geometry, and to gen-
erate the meshes. Simulations were carried out on an IBM pSeries-690 machine and
Tungsten, a Linux cluster at the National Center for Supercomputing Applications
(NCSA) at the University of Illinois at Urbana-Champaign. Though there are many
more processors available, these simulations were carried out on only 4-16 processors.
Natural circulation in the RRR is simulated using the Boussinesq approximation.
k − ϵ model with enhanced wall treatment to describe the near-wall behavior was
chosen for turbulence.
26
The simplified model of the RRR used for the CFD simulation is composed of
the reactor core, the chimney, the heavy water reflector vessel and the reactor pool.
A 2D schematic diagram is shown in Figure 3.6. More detailed 3D, FLUENT models
are shown in Figures 3.7 and 3.8.
Figure 3.6: Diagram of core, chimney, and reactor pool
RRR core has a total of 16 fuel assemblies arranged in a 4× 4 grid. A RRR
assembly has 21 fuel plates with thickness of 1.5 mm, width of 65 mm, and height of
615 mm. Water flows through the 22 narrow flow channels 2.54 mm thick, 75 mm
wide, and 1000 mm long. There are total of 336 fuel plates (1.5 mm 65 mm) and 352
(2.54 mm 75 mm) rectangular coolant flow channels. Average velocity through the
core under normal operating conditions (downward forced flow) is 8.2 m/s. Due to the
extreme aspect ratio of the flow channels, a large number of cells are needed to model
the reactor core. Hence, this is a good candidate for the porous media approximation.
By modeling the core as porous media, the mesh requirement is reduced dramatically.
27
Figure 3.7: Full scale CFD model of RRR
3.2.3 CFD Porous Media Model
Porous media in FLUENT 6 is modeled using additional source terms in the momen-
tum equation given by

( )
∑
3 ∑
3
1
Si = − Dij µvj + Cij ρ|v|vj (3.1)
j=1 j=1
2
where Si is the source term for the ith (x, y or z) momentum equation, vj is the
velocity component in jth (x, y or z) direction, |v| is the velocity magnitude, and µ is
the fluid viscosity. The first term on the right hand side of Equation 3.1 is the viscous
pressure loss term (Darcy term) due to the porous media structure. The second term
(Forchheimer term) represents the pressure loss due to the momentum of the flow
in the porous media zone. The two parameters; tensor Dij , called viscous resistance
factor, and Cij , called inertial resistance factor, should be provided by the user.
The viscous resistance factor and the inertial resistance factor used in this
work are estimated from the results of the fuel assembly level simulations with full
structural details.
Porous media induced turbulence is not considered in the porous media model
28
Figure 3.8: CFD model of core and chimney
of this simulation. This simplification may lead to lower turbulent kinetic energy and
turbulent viscosity, since turbulent kinetic energy equation of the standard k−ϵ model
does not include the term corresponding to porous media induced turbulence. Hence,
the results obtained here may give a higher circulation flow rate. This issue is later
addressed by enhancing the porous media turbulence model in FLUENT.
3.2.4 Fuel Assembly Level CFD Simulation
A RRR fuel assembly has 21 fuel plates with thickness of 1.5 mm, width of 65 mm,
height of 1000 mm with active fuel height of 615 mm. The coolant flow channels are
2.54 mm thick and 1000 mm long, sandwiched between fuel plates. Due to the extreme
aspect ratio and length-to-width ratio of the coolant channels, a single assembly is
meshed using 2.2 million cells. Figure 3.9 shows the three-dimensional CFD model
of a fuel assembly.
The goal of these assembly level simulations with complete structural details
of the assembly is to determine pressure drops for different inlet flow rates, and then
estimate porous media parameters that would lead to the same pressure drop that
would occur if the assembly was modelled using the porous media approximation.
29
Figure 3.9: CFD model of a fuel assembly
Figure 3.10 schematically shows the boundary conditions specified for the
assembly-level simulations. In the assembly level simulation, the local pressure drops
at the entrance (Zone A in Figure 3.10) and the exit (Zone B in Figure 3.10) of the
flow channels are of interest.
Figure 3.10: Boundary conditions on the fuel assembly
The velocity vectors at both, channel entrance and exit regions of the channel
are plotted in Figures 3.11 and Figure 3.12. Flow characteristics agree with expecta-
30
tions.
Figure 3.11: Velocity vector plot at assembly inlet (Box A in Figure 3.10)
Figure 3.13 shows the cross sectional average pressure along the flow direction
in the assembly. It shows that a significant fraction of the pressure drop occurs at the
channel’s entrance (Zone A in Figure 3.10) and exit regions (Zone B in Figure 3.10).
However, fraction of pressure drop that occurs along the assembly is higher than
those normally seen in pipe flow (clear flow). Under normal operating conditions, the
core is cooled with forced circulation. The inlet velocity in the assembly level CFD
simulation (see Figure 3.10) is different from the average velocity used in Equation
3.1 because the average velocity flow area in the core region is smaller than that
at the core inlet. The average velocity in the coolant channels is 8.2 m/s which
corresponds to 5.08 m/s of inlet velocity at the inlet of the ”lower plenum.” The
pressure drop under normal operating condition is reported to be 240 kPa [ANSTO,
2001]. Corresponding CFD simulations show a pressure drop of only 180 kPa. The
difference of 60 kPa is most likely due to local pressure drops because of other core
31
Figure 3.12: Velocity vector plot at assembly outlet (Box B in Figure 3.10)
structures which are not included in the assembly level CFD model.
Mesh refinement exercise was carried out, and number of cells was increased
from 2.2 million to 2.9 million. Similar pressure profiles were obtained in both cases,
indicating that 2.2 million cells were sufficient to obtain stable and accurate results.
The total pressure drops corresponding to different inlet velocity conditions
are listed in column 2 of Table 3.1. Column 3 is the pressure drop in an assembly
without the fuel plates. The difference (column 4) between the pressure drop in
column 2 and column 3 is the pressure drop due to the fuel plate structure. For inlet
velocity ranging from 0.01-5.08 m/s, a quadratic function is used to fit the pressure
drop due to the fuel plates calculated using complete structural details in the fuel
assembly (column 4 of Table 3.1). This is shown as the ’poly’ curve in Figure 3.14.
This completes the CFD modelling of a single assembly with complete structural
details.
Comparing Equation 3.1, where Si is treated as pressure drop due to the
32
Figure 3.13: Pressure drop along an assembly
porous media, with the quadratic fit for the pressure drop due to the structural details
(porous media) shown in Figure 3.14, values of porous media model parameters (valid
over the entire flow rate range) are determined. Average viscous resistance factor and
the inertial resistance factor in the flow direction, z, for the porous media model in
the flow direction were found to be 1.59 × 107 1/m2 and 8.80 1/m, respectively.
33
(a) *using a single set of values for Cz and Dz for the entire velocity
range; ** using separate Cz and Dz for low velocity range and high
velocity range
Figure 3.14: Pressure drops in assembly level CFD with complete structural details and
using two different sets of porous media parameters
34
Table 3.1: Core pressure drop results from assembly level simulations with structural
details.
Inlet Assembly Pressure Channel Pressure Pressure Drop Due
velocity Drop with Structural Drop without Fuel to the Fuel Plates
(m/s) Details (Pa) Plates (Pa) (column 2-column 3)
(Pa)
0.01 131.7 0.062 131.6
0.025 349.1 0.20 348.9
0.05 773.3 0.50 772.8
0.1 1650.3 15.03 1635.3
0.5 6188.5 34.39 6154.1
1.0 14819.0 118.12 14700.9
2.0 45279.9 412.80 44867.1
5.08 180980.8 2259.98 178720.8
35
Table 3.2: Comparison of core pressure drop results.
Pressure
Drop CFD Simulation for CFD Simulation for
Inlet across the Pressure Drop across Pressure Drop across
velocity Assembly the Assembly the Assembly
(with (with Porous Media*) (with Porous Media**)
Structural
Details)
(m/sec) (Pa) Cz (1/m) Dz (1/m2 ) ∆p(Pa) Cz (1/m) Dz (1/m2 ) ∆p(Pa)
0.01 131.7 135.7 132.2
0.025 349.1 340.1 119.3 1.25E+07 357.7
0.05 773.3 683.0 779.9
0.1 1650.3 1388 1422
0.5 6188.5 8.8 1.59E+7 7844 7983
1.0 14819.0 17953 8.58 1.37E+7 16202
2.0 45279.9 44933 45170
5.08 180980.8 184438 187167
* using a single set of values for Cz and Dz for the entire velocity range;
** using separate Cz and Dz for low velocity range and high velocity range.
36
The flow transition in porous media from laminar to turbulent occurs over
the range of Rep (pore diameter based Reynolds number) from 300 to 1000 [Antohe
and Lage, 1997]. For an inlet velocity of 0.1 m/s, based on pore diameter in our
model, Rep = 820. Hence, flow is expected to be laminar for vi = 0.0 ∼ 0.1 m/s
and turbulent for vi = 0.1 ∼ 5.08 m/s. To more accurately capture the pressure
drop over both laminar and turbulent flows in porous media, two pairs of Cz and Dz
are determined; one for low inlet velocity range and the other for the high velocity
range, which are defined by the inlet velocity range of 0.0 ∼ 0.1 m/s and 0.1 ∼ 5.08
m/s, respectively. Table 3.2 shows the parameter values for the single set of Cz and
Dz (column 3 and column 4 in Table 3.2), and for separate Cz and Dz (column 6 and
column 7 in Table 3.2) for laminar and turbulent flows. Pressure drop across a single
fuel assembly are then calculated using the prous media approximation for the fuel
region using the porous media model parameters given in Table 3.2. These are also
shown in Table 3.2 (columns 5 and 8) and plotted in Figure 3.14 for comparison.
3.2.5 RRR Integral CFD Simulation
Using the porous media model parameters parameters Cz and Dz (= 8.80 1/m and
1.59×107 1/m2 ), obtained from assembly level simulations, full scale RRR simulations
were carried out that include the core (as porous media), the chimney and the pool
(Figure 3.15). Number of mesh for these simulations range from 400k to 2.9M cells.
Boundary condition at the top of the pool was set as wall (rigid lid). Temperature
on all outside surfaces is set to be 308 K. In natural circulation simulations, with a
”constant” residual heat of 1 M W (5% of full thermal power) added to the porous
media zone as a uniform thermal source term of 14 M W/m3 , the temperature rise
across the core under natural circulation conditions is found to be around 74.4 K.
Average velocity in the core is around 0.028 m/s.
A typical flow pattern established under natural flow is shown in Figure 3.16.
37
Even if the flow through the core is over estimated due to the nonconservative ap-
proximation of ignoring porous media induced turbulence, a 10 ∼ 20% reduction in
flow rate is not likely to increase the temperature rise to unacceptable level.
YX
Figure 3.15: CFD model for the Replacement Research Reactor
Highest velocity is just above the chimney outlet. It is due to the tempera-
ture difference between the coolant from the chimney and the water in the reactor
pool. The buoyancy due to the temperature difference keeps accelerating the jet flow
through the chimney. Recirculation region is formed in the RRR chimney during nat-
ural circulation. Flow at the chimney exit is both upward and downward. Buoyancy
driven flow leaves the chimney over a part of the exit cross section, while coolant just
above the chimney exit also flows downward into the chimney over the remainder of
the cross section due to the higher density.
This reverse flow builds up the inner circulation as shown in Figure 3.16.
38
Figure 3.16: Reverse flow in the chimney
3.2.6 Parametric Study
The effect of the chimney cross-sectional area on RRR natural circulation was studied
parametrically. A modified system with chimney side length that is 80% of the design
value (0.35 m × 0.35 m) was modelled and simulated. All other model characteristics
and dimensions were kept as in the simulation of the original RRR design. Figure
3.17a and Figure 3.17b illustrate two chimney CFD models used in the parametric
study of the chimney dimension.
The CFD simulation shows that although the smaller chimney dimension
doesn’t eliminate the inner recirculation flow in the chimney, it does reduce the down-
ward flow from the top of chimney. By defining the recirculation factor as the ratio
between mass flow rate entering the chimney at the top to the mass flow rate going
upward through the reactor core, the effects of two RRR chimneys can be compared
quantitatively. A larger recirculation factor indicates less efficient chimney. Mass
39
Figure 3.17: Four configurations for parametric study
flow rate through the reactor core, recirculation factor, temperature rise across the
core, and average velocity in the chimney for the two cases are compared in Table
3.3. The comparison shows that the chimney with smaller cross-sectional area leads
to higher flow rate through the reactor core, and the inner recirculation observed in
the original simulation is now less intense. Thus, residual heat can be better removed
with smaller cross-sectional area chimney.
To further reduce the flow into the chimney at the top, which reduces the
average temperature in the chimney, a cap with a circular hole is attached to the top
of the chimney, as shown in Figure 3.17c and Figure 3.17d. CFD simulations of these
two designs show a significant increase in natural circulation mass flow rate. Results
40
are included in Table 3.3. No reverse flow is observed in the simulations of cases with
chimney caps. This indicates that the reverse flow at the top of the chimney is a
major factor impairing the natural circulation driving force, and chimney caps can
prevent it effectively.
Comparing the two designs with chimney caps, the chimney design with 100%
cross-sectional area yields a higher natural circulation mass flow rate than the 80%
cross-sectional area chimney design does. This is different from the comparison of the
two designs without chimney caps. The reason behind this is that the smaller flow
channel in 80% cross-sectional area chimney design produces more resistance than
original design does, while natural circulation driving heads are similar if there is no
downward flow from the top of the chimney.
From the parametric study it is concluded that, as expected, the reverse flow
and inner recirculation impairs the RRR full-scale natural circulation, thus impairing
the system’s inherent ability to remove the residual decay heat. Any new research
reactor design should be configured to avoid the occurrence of reverse flow and inner
recirculation in the chimney. A chimney with a cap preventing reverse flow from the
top can enhance the driving force of natural circulation.
3.2.7 Some Intermediate Conclusions for RRR
Full-scale simulations of Replacement Research Reactor were carried out with the
porous media model for the reactor core. Parameters for the porous media were
estimated from CFD simulations of flow in a single assembly. These simulations
demonstrate that CFD analysis of research reactors can be carried out using the
computational power currently available. Thus, further optimization of RRR and
other reactors’ thermal hydraulic system design could be carried out with the help of
CFD analysis.
41
Table 3.3: Natural circulation conditions for two different chimney sizes.
Chimney Coolant flow Recirculation Temperature Average
Cross-sectional rate through Factor rise across velocity in
Area core the core the chimney
(m2) (kg/s) (K) (m/s)
0.1225 3.37 1.12 74.4 0.0276
0.0784 3.69 0.59 68.3 0.0472
0.1225 with cap 4.27 0 59.0 0.0349
0.0784 with cap 4.13 0 61.2 0.0528
42
Chapter 4
Turbulence Porous Media Models
Default turbulence treatment of porous media model is used for the ANSTO Replace-
ment Research Reactor (RRR) CFD simulation in the previous chapter. Laminar flow
model in porous media in RRR core was adequate since flow under natural circula-
tion conditions is expected to be laminar. However, flow in the nuclear reactor core
is turbulent under normal operating conditions.
For example, the diameter of the coolant channels in Gas Turbine - Modular
Helium Reactor (GT-MHR) core is ∼ 6 mm. Under normal operating conditions,
the helium mass flow rate through the reactor core is 320 kg/s at 7.0 MPa. The
average coolant flow speed is more than 10 m/s. Reynolds number, based on the
hydraulic diameter of the coolant channels, could be as high as 10,000 under the
normal operating conditions. Due to the extreme complexity, it is unlikely that
the entire nuclear reactor core can be simulated in a CFD-type approach with a
microscopic turbulence model. Porous media modelling is indispensable to model the
nuclear reactor core. It is recommended that turbulence in porous media should be
taken in consideration when Reynolds number based on pore length scale is larger
than 300 ∼ 1000 [Dybbs and Edwards, 1984] [Macdonald, 1979]. The porous media
flow in the prismatic core of GT-MHR is turbulent flow. Pebble Bed Modular Reactor
(PBMR) is another Gen IV Very High Temperature Reactor (VHTR) design. The
43
design flow velocity and flow path parameters are similar in the pebble bed core of
PBMR to GT-MHR. The random coolant flow paths in PBMR make the flow even
more likely to be turbulent.
Turbulence enhances convective heat transfer, and leads to additional pressure
drop across the reactor core. Therefore, it is desirable that CFD codes have the
capability of simulating turbulent flow within the porous media paradigm to allow
nuclear engineers to simulate the turbulent flow in nuclear reactor cores. Equipping
current CFD codes with enhanced macroscopic turbulence models for porous media
is necessary for CFD simulation of nuclear systems.
The general methodology to derive macroscopic turbulence models for porous
media is summarized in the next subsection. A macroscopic turbulence model de-
veloped for porous media similar to nuclear reactor cores is reviewed. A modified
k − ϵ model is implemented in FLUENT using User-Defined Functions (UDFs). The
turbulence model implemented in FLUENT is tested by simulating the conditions in
an experiment and comparing the simulation results with the experimental data.
4.1 Macroscopic Turbulence Model for Porous Me-
dia Flow
Modeling of flow in porous media is generally carried out using Darcy’s law, which
is a semi-empirical equation correlating the flow rate and the pressure drop due to
the presence of porous media structure. With recent developments in CFD, several
investigations for porous media flow have been carried out based on the spatially-
averaged equations, derived from the Navier - Stokes equations. A very comprehensive
review of work on modelling of flow through porous media was given by Antohe and
Lage[Antohe and Lage, 1997]. Masuoka and Takatsu derived a 0-equation turbulence
model using the local volume-averaging technique. They modelled the effective eddy
44
diffusivity as the algebraic sum of the eddy diffusivities estimated from two types of
vortices: the pseudo vortex and the interstitial vortex[Masuoka and Takasu, 1996].
A one-equation k − ϵ model with explicit expression for turbulent energy dissipation
rate and four model parameters was proposed to simulate turbulent flow through a
stack of apples. Experimental data and numerical simulation results were compared
[Alvarez et al., 2003]. In another work for environmental engineering, the standard
k − ϵ model was modified for turbulent flow in submerged vegetation, which was
simulated as porous media. Experimental data from a flow channel with mock-up
vegetation was used to validate the turbulence model in porous media [Lopez and
Garcia, 2001]. A modified k − ϵ model was developed to model nuclear reactor core
as porous media. The simulation results using the new porous media turbulence
model were compared with both Direct Numerical Simulation (DNS) results and
experimental data [Chandesris et al., 2006].
Although the porous media models for turbulent flow are diverse in the ap-
plication areas, the methodologies to develop these models are relatively similar.
Volume-averaging of the Reynolds Averaged Navier-Stokes (RANS) equation is the
general procedure to develop a two-equation macroscopic turbulent flow model in
porous media. In Figure 4.1, box A is the Representative Elementary Volume (REV)
over which volumetric averaging is carried out to derive the macroscopic model. The
approach to closure of the equation set obtained from volume-averaging, and treat-
ment of the additional source terms in momentum equation, turbulent kinetic equa-
tion, and turbulent kinetic energy dissipation equation distinguish one two-equation
macroscopic turbulence model from other porous media models.
Numerous two-equation macroscopic models for turbulent flow in porous me-
dia have been proposed. Antohe and Lage derived a two-equation macroscopic tur-
bulence model by applying the time averaging operator to the extended Darcy Forch-
heimer model [Antohe and Lage, 1997]. Getachew extended this work by taking
45
Figure 4.1: Porous media REV for macroscopic average
into account the higher order closure terms [Getachew et al., 2000]. Following an-
other approach, Nakayama and Kuwahara proposed a two-equation macroscopic tur-
bulence model obtained by spatial averaging the Reynolds-averaged Navier-Stokes
equations[Nakayama et al., 1999]. However, for turbulent flows, the application or-
der of the two operators (time-averaging for turbulence, and volume averaging) is
important. Pedras and de Lemos [Pedras and de Lemos, 2000] showed that the two
approaches lead to similar equations for the mean flow, but that the turbulent kinetic
energies resulting from the two approaches were different. In particular, they showed
that, applying the time-averaging operator first, takes into account the turbulence
inside the porous media. Since nuclear engineers are interested in turbulence inside
the porous media (reactor core), the model derived by the latter approach is cho-
sen in this thesis work. Applying the volume-averaging theory to the microscopic
transport equations of turbulent kinetic energy and its dissipation rate, Nakayama
and Kuwahara, and Pedras and de Lemos, developed a macroscopic two-equation
turbulence model. They obtained a new set of equations for the transport of the
volumetrically averaged turbulent kinetic energy and its dissipation rate. These new
equations involve additional terms which quantify the influence of the porous medium
on the turbulent kinetic energy and turbulent kinetic energy dissipation level. The
challenge in these approaches is generally in developing closure models for these ad-
46
ditional terms. Unfortunately there is no general well-developed closure expression
valid for different kinds of porous media morphologies for these additional terms.
Nakayama and Kuwahara [Nakayama et al., 1999], and Pedras and de Lemos [Pedras
and de Lemos, 2000] proposed different closure models for these additional terms. The
constant in these models are obtained by integrating microscopic results obtained from
numerical experiments over a unit porous structure. Different unit porous structures
have already been considered including regular morphology made of square, circular
or elliptic rods. This method of integrating microscopic results obtained from nu-
merical experiments over a unit porous structure has also been used with success by
Kuwahara and Nakayama [Nakayama et al., 1999] to study thermal dispersion and
interfacial heat transfer coefficient in porous media.
4.1.1 Turbulence Porous Media Model for Nuclear Reactor
Core
As mentioned above, the flow in nuclear reactor cores is generally turbulent under
normal operating conditions. Hence, CFD modelling of nuclear reactor cores with
porous media approximation should take turbulence into account. However, nuclear
reactor cores are different from the general porous media that in general do not have
any regular structure. Most nuclear reactor cores are however composed of hundreds
of bundles, planes, or channels with some regularity. Macroscopic turbulence model
specially developed for these porous media geometries can be developed for nuclear
applications.
Chandesris et al proposed a macroscopic two-equation turbulence model for
turbulent flows through porous media of particular morphologies, such as arrays of
square, or circular rods. The model was developed for macroscopic longitudinal flows
in channels, pipes and rod bundles, which are the simplified geometries of reactor
cores. The additional source terms of the macroscopic k − ϵ equations, which appear
47
as a result of the volume-averaging process, are modelled using the kinetic energy
balance and physical considerations. The two unknown constants of the closure ex-
pression are determined from the spatial averaging of the microscopic k−ϵ simulations
and from numerical and experimental results available in the literature. This model
was successfully tested in various geometries such as channel flow and pipe flow by
Chandesris et al. Good agreement was reported between the model results and an
experiment with decreasing turbulence inside a rod bundle [Chandesris et al., 2006].
A brief description of this macroscopic porous media turbulence model for nuclear
reactor core modelling is given in the following paragraphs.
The porous media turbulence model for nuclear reactor core was developed
by volume-averaging the Reynolds averaged Navier-Stokes equation (2.7) within a
REV. The dimension of the REV should be small enough to preserve the macroscopic
character of the flow in the porous media. Moreover, REV should be larger than the
length scale of a characteristic pore.
After taking the volumetric average of the continuity equation and the mo-
mentum equations over the REV, they can be written as
∂ϕ ⟨ui ⟩f
=0 (4.1)
∂xi
∂ ⟨ui ⟩f ∂ ( )
ϕ + ϕ ⟨ui ⟩f ⟨ui ⟩f =
∂t ∂xi
1 ∂ ( )
− ϕ ⟨p⟩f
ρ ∂xi
( ( ))
∂ ∂ϕ ⟨ui ⟩f ∂ϕ ⟨uj ⟩f
+ ν +
∂xj ∂xj ∂xi
∂ ( )
− ϕ ⟨δui δuj ⟩f + R
∂xi
∂ ( ⟨ ′ ′⟩ )
− ϕ ui uj f (4.2)
∂xi
where <> is the volumetric averaging operator over the REV, and <>f is the volu-
48
metric averaging operator over the liquid phase only in the REV.
∫
1
⟨ψ⟩f ≡ ψdV (4.3)
Vf Vf
δψ is defined as the deviation of ψ from its average ⟨ψ⟩. u′i is the velocity fluctu-
ation term in i direction. Similar to the Reynolds Averaged Navier-Stokes (RANS)
equations, the volume averaged RANS equations also face the closure problem.
The last three unknown terms in Equation 4.2 need to be modelled. The
first term describes the inertial dispersion due to the presence of the porous media
structure in the reactor core. The R term represents the drag force due to the solid
surfaces in the porous media. R is given by

∫ ( ( ) )
1 ∂ui ∂ui p
R= ν + − δij · nj dΣ (4.4)
V Σ ∂xj ∂xj ρ
where Σ is the surface of the solid in the porous media. These two terms could be
interpreted as the frictional force on the mean flow from the porous media. They are
modelled by a lumped friction force term F , which depends on the morphology of the
reactor core.
The third term is the Reynolds stress term that results even if there is no solid
structure (porous media) present in the flow. Similar to the microscopic turbulence
model, the volume averaged Reynolds stress term is to be modelled by an artificially
introduced macroscopic turbulent viscosity νtϕ , defined as
νtϕ ⟨Sij ⟩ = ⟨νt Sij ⟩ (4.5)
where ⟨Sij ⟩ is the strain tensor,
( )
1 ∂ϕ ⟨ui ⟩f ∂ϕ ⟨uj ⟩f
⟨Sij ⟩ = + (4.6)
2 ∂xj ∂xi
The macroscopic Reynolds stress tensor is given by
⟨ ⟩ 2 2
−ϕ u′i u′j f = 2ϕ ⟨νt Sij ⟩f − ϕ ⟨k⟩f δij = 2νtϕ ⟨Sij ⟩ − ϕ ⟨k⟩f δij (4.7)
3 3
49
.
The macroscopic momentum equation with macroscopic turbulent viscosity
can be written as
∂ ⟨ui ⟩f ∂ ( )
ϕ + ϕ ⟨ui ⟩f ⟨ui ⟩f =
∂t ∂xi
( )
1 ∂ 2
− ϕ ⟨p⟩f + ϕρ ⟨k⟩f
ρ ∂xi 3
( ( ))
∂ ( ) ∂ϕ ⟨ui ⟩f ∂ϕ ⟨uj ⟩f
+ ν + νtϕ +
∂xj ∂xj ∂xi
+ϕFi (4.8)
where macroscopic turbulent viscosity is modelled by

⟨k⟩2f
ν tϕ = C µ (4.9)
⟨ϵ⟩f
The macroscopic turbulent kinetic energy and the macroscopic turbulent kinetic en-
ergy dissipation rate equations in the proposed model are

∂ ⟨k⟩f ∂ ( )
ϕ + ϕ ⟨ui ⟩f ⟨k⟩f =
∂t ∂xi
2ϕνtϕ ⟨Sij ⟩f ⟨Sij ⟩f
(( ) )
∂ νtϕ ∂ϕ ⟨k⟩f
+ ν+
∂xj σ̃k ∂xj
−ϕ ⟨ϵ⟩f + ϕSk (4.10)
∂ ⟨ϵ⟩f ∂ ( )
ϕ + ϕ ⟨ui ⟩f ⟨ϵ⟩f =
∂t ∂xi
⟨ϵ⟩f
ϕ(2c1 νtϕ ⟨Sij ⟩f ⟨Sij ⟩f − c2 ⟨ϵ⟩f )
⟨k⟩f
(( ) )
∂ νtϕ ∂ϕ ⟨ϵ⟩f
+ ν+ + ϕSϵ (4.11)
∂xj σ̃ϵ ∂xj
where (c1 , c2 ) are the two model constants in the standard microscopic k−ϵ turbulence
model. σ̃k and σ̃ϵ are macroscopic turbulent Prandtl numbers. The extra terms Sk
and Sϵ are the source of turbulent kinetic energy and its dissipation rate due to the
presence of the solid in the porous media.
50
The closure of the equation sets and the modelling of the unknown terms in
the momentum equation, averaged turbulent kinetic energy equation, and in averaged
turbulent dissipation rate equation are based on the morphology of the reactor core
to be modelled.
Turbulent kinetic energy source term Sk is proposed to be of the form

( √ )
⟨u⟩3f Cf
Sk = 2Cf 1 − ylim
+
(4.12)
Dh 2
+
where Dh is the hydraulic diameter, and ylim corresponds to the limit of the integration
zone expressed in wall units. The geometry of the reactor core is used to determine
the friction coefficient, Cf , in Equation 4.12.
The turbulent dissipation source term Sϵ is modelled as
Sk2
Sϵ = c2 (4.13)
k∞
where Sk is defined in Equation 4.12. A model for k∞ is proposed as
−1/6
k∞ = cp ⟨u⟩2f ReH (4.14)
+
Two model constants, ylim and cp , are estimated [Chandesris et al., 2006]
based on DNS, microscopic turbulence model, and experimental results for channels
and pipe flow. The values of the model constants recommended by [Chandesris et al.,
2006] for nuclear reactor cores’ porous media modelling are listed in Table 4.1.
Table 4.1: Model constants for macroscopic turbulence porous media model
+
Geometry cp ylim
channel 0.0306 8
pipe 0.0367 7
A systematic development of the macroscopic turbulence model developed
specifically for the porous media structure in the nuclear reactor core is reviewed in
51
the previous subsection. The capability and methodology of model implementation
in the CFD code, FLUENT, has been first tested in the following section by using
another turbulence porous media model that was briefly mentioned in Section 4.1.
4.2 Verification of Turbulence Porous Media Model
Implementation
User developed models can be implemented in CFD codes using UDFs. To test
UDFs in FLUENT, a simpler k − ϵ model for porous media is implemented first via
UDFs. This porous media model was originally developed to simulate flow through
underwater vegetation.
4.2.1 Modified k − ϵ Model for Porous Media Flow
Lopez and Garcia proposed a modified k − ϵ model for turbulence in porous media
to simulate flow through underwater vegetation [Lopez and Garcia, 2001]. Source
terms modeling the porous media effects were added to the corresponding clear flow
equations to develop an equivalent porous media flow field model. The continuity
equation in this model is

∂ < u1 >
=0 (4.15)
∂x1
The x-momentum equation equation is
∂ < u1 > ∂ ∂ < u1 >

= gS0 + [(νT + ν)( )] − fx1 (4.16)
∂t ∂x3 ∂x3
The turbulent kinetic energy equation, or the k equation, is

[( ) ]
∂k ∂ νT ∂k
= +ν + Pk − ϵ + Cf k fx1 < u1 > (4.17)
∂t ∂x3 δk ∂x3
The turbulent dissipation rate equation, or the ϵ equation, is

[( ) ]
∂ϵ ∂ νT ∂ϵ ϵ
= +ν + [C1(Pk + Cf ϵ fx1 < u1 >) − C2 ϵ] (4.18)
∂t ∂x3 δk ∂x3 k
52
Compared to the standard k − ϵ model equations, the source terms for the
porous media effects are fx1 , Cf k fx1 < u1 >, and ϵ/kC1 Cf ϵ fx1 < u1 >.
4.2.2 Boundary Conditions
The boundary conditions used in simulating flow through vegetation are imposed at
the river bed and at the free surface. These are
u∗ ( x30 u∗ )
U0 = ln E (4.19a)
κ v
u2∗
k0 = √ (4.19b)
2
C/ mu
u3
ϵ= ∗ (4.19c)
κx30
at the river bed, and
dUf s
=0 (4.20a)
dx3
dkf s
=0 (4.20b)
dx3
dϵf s
=0 (4.20c)
dx3
at the free surface.
This model is implemented in FLUENT using UDFs. Explicit examples of
source terms coded are given in Appendix A.
4.2.3 CFD Model
A 2D simulation of an experiment [Lopez and Garcia, 2001] has been carried out using
the modified k − ϵ model presented above. The model is a simple rectangular flow
channel with dimensions as in the actual experiment. Figure 4.2 shows a schematic
diagram of the open channel with submerged vegetation.
53
Figure 4.2: Side view of channel flow with submerged vegetation
The experiment was conducted under uniform open-channel flow conditions.
The open-channel flow domain is divided into a clear flow region and a vegetation
region. The vegetation (lower) region is 0.12 m high. In vegetation region, modified,
porous media k − ϵ model is used to simulate the flow. In the region above the
vegetation region and in the developing flow region, standard k −ϵ model is employed.
The cross section of the flow channel is a 0.91 m × 0.61 m (width × depth) rectangle.
Water level is maintained at 0.34 m. The length of the flow channel is 19.5 m with a
slope of 0.0036. The volumetric flow rate is 0.179 m3 /s. Velocity measurements were
taken with a acoustic doppler velocimeter at a sampling frequency of 25 Hz [Lopez
and Garcia, 2001].
4.2.4 Comparison
The CFD results are compared with the experimental data. Figure 4.3 shows the
comparison of the x component of the mean velocity obtained using the CFD sim-
ulations and the experimental data [Lopez and Garcia, 2001]. Figure 4.4 shows the
√
comparison of the < u′1 u′1 > component of Reynolds stress. − < u′1 u′3 > component
of Reynolds stress are compared in Figure 4.5.
It is noted that agreement between the CFD results and experimental data
for turbulent kinetic energy and Reynolds stress are not as good as that for the mean
x velocity < u1 >. The assessment of the user-defined turbulence model in porous
54
1.0
0.9 Experimental data
CFD results with modefied porous media model
0.8
Height of Vegetation
0.7
 (m/s)
0.6
1
0.5
0.4
0.3
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
z (m)
Figure 4.3: Comparison of mean velocity term between experimental data and
CFD results at x = 15.0 m
media in this case validate the extension possibility through FLUENT UDFs.
55
0.20
Experimental data
0.18
CFD results with modified porous media model
0.16
0.14
(m/s)
0.12
0.10
1/2
>
2
 term in Reynolds stress comparison between experimental data
and CFD results at x = 15.0 m
56
0.010
0.009
Experimental data
CFD results with modified porous media model

0.008
0.007
- (m /s )
2
2
0.006
0.005
3
1
0.004
0.003
0.002
0.001
0.000
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
z (m)
Figure 4.5: − < u′1 u′3 > component of Reynolds stresses comparison between experimental
data and CFD results at x = 15.0 m
57
Chapter 5
Nuclear System Code and
Coupling Approach
5.1 RELAP5-3D
RELAP5-3D is the latest released version of RELAP code. It is an outgrowth of the
one-dimensional RELAP5/MOD3 code developed at the Idaho National Engineering
and Environmental Laboratory (INEEL) for the U.S. Nuclear Regulatory Commission
(NRC).
RELAP5/MOD3 code is a best estimate thermal hydraulic analysis tool for
light water reactors. It uses two-phase six-equation model for the nuclear thermal
hydraulic system modelling [INEEL RELAP5/MOD3 Group, 1990]. Semi-empirical
models for the components, such as pumps, valves, pressurizer, and accumulators, are
available to simulate the transient behavior of the entire nuclear thermal hydraulic
systems. RELAP series codes can be used to model the entire thermal hydraulic
system of nuclear power plant using a relatively coarse nodalization. They have
widespread acceptance in the nuclear industry.
RELAP5-3D has many enhancements in addition to the extension from one-
dimensional treatment in RELAP5/MOD3. The most prominent one is that the code
58
has fully integrated, multi-dimensional thermal hydraulic and neutron kinetic mod-
elling capability. Although the multi-dimensional capability is available in RELAP5-
3D, it is fundamentally different from the three-dimensional CFD modelling. It pro-
vides the option to model the cross flow in reactor core and flow distribution using a
network of nodes. Other improvements in RELAP5-3D include the new matrix solver
for multi-dimensional problems, the updated water and other coolants properties,
and improved time advancement for better robustness [INEEL RELAP5-3D Group,
1999]. RELAP5-3D development has also benefitted from the extensive experimental
data generated for the severe accident scenarios.
With additional working fluids added, RELAP5-3D can be used not only to
simulate the water reactors, both Light Water Reactor (LWR) and Heavy Water
Reactor (HWR), but also to study the next generation reactor designs such as fast
breeder reactor, gas cooled reactor, and even the conceptual space reactor. Helium
is a new working coolant in RELAP5-3D. This extension enables RELAP5-3D to
analyze the Gen IV VHTR, which uses helium as reactor coolant.
RELAP5-3D is chosen to develop the coupled CFD—system-code in this the-
sis. Author has earlier used RELAP5/MOD3 to study the Passive Residual Heat
Removal (PRHR) system in the conceptual design of the 600 MWe Advanced Chinese
Pressurized Water Reactor (ACP600) [Yan, 2001]. The source code of RELAP5-3D
is available to be modified for the coupled CFD—system-code development.
5.1.1 Field Equations
The RELAP5-3D thermal hydraulic model is based on eight field equations for the
eight primary dependent variables of the water/vapor two phase system. The eight
primary dependent variables are pressure(P ), phasic specific internal energies (Ug ,
Uf ), vapor/gas volume fraction (void fraction, αg ), phasic velocities (vg , vf ), noncon-
densable gas quality (Xn ), and boron density (ρb ). The independent variables are time
59
(t) and location (x) along the flow path in the one-dimensional field equations. For the
multi-dimensional cases, the independent variables are time (t) and three-dimensional
location (x, y, z for cartesian coordinate; r, θ, z for cylindrical coordinate).
In RELAP5-3D, the phasic velocities are along the predefined flow path di-
rection. The transverse directions to the mean flow are not solved by the code.
Although RELAP5-3D has multi-dimensional capability, it is still fundamentally dif-
ferent from the multi-dimensional capability of the CFD codes. The flow domain
simulated by RELAP5-3D is a network which consists of one-dimensional flow path.
Cross flow may be modelled by multi-dimensional RELAP5-3D. But flow phenomena
like multi-dimensional mixing and flow distribution is still beyond the scope of the
traditional nuclear system codes. As a result, even with multi-dimensional capability,
RELAP5-3D is still categorized as a one-dimensional code.
RELAP5-3D includes the noncondensable gas in the hydrodynamic model.
The noncondensable gas model could be used to take the entrained nitrogen and air
into consideration. The noncondensable gas quality (Xn ) is defined as the fraction
of the noncondensable gas mass to the total vapor and gas phase mass. Mn is the
mass of noncondensable gas in the vapor/gas phase and Ms is the mass of the vapor
in the vapor/gas phase. The secondary dependent variables used in the equations
are phasic densities (ρg , ρf ), phasic temperatures (Tg , Tf ), saturation temperature
(Ts ), and noncondensable mass fraction in noncondensable gas phase (Xni ) for the
i-th noncondensable species. Mni is the mass of the i − th specie of noncondensable
in the vapor/gas phase, Mn is the total mass of noncondensable gas in the vapor/gas
phase, and n is the number of noncondensable species.
The basic two-fluid differential equations form the framework of the hydrody-
namic model in RELAP5-3D. The basic field equations for the two-fluid nonequilib-
rium model consist of two-phasic continuity equations, two phasic momentum equa-
tions, and two phasic energy equations. The time dependant conservation of mass
60
equations for vapor phase and liquid phase respectively are
∂ 1 ∂
(αg ρg ) + (αg ρg Vg A) = Γg (5.1)
∂t A ∂x
∂ 1 ∂
(αf ρf ) + (αf ρf Vf A) = Γf (5.2)
∂t A ∂x
where αg +αf = 1. For the flow without external mass sources or sinks, overall
continuity consideration yields the requirement that the liquid generation term Γf be
the negative of the vapor generation Γg , that is, Γf = −Γg . The interfacial mass
transfer model assumes that total mass transfer can be partitioned into mass transfer
at the vapor/liquid interface in the bulk fluid Γig and mass transfer at the vapor/liquid
interface in the thermal boundary layer near the walls Γw ; that is, Γg = Γig + Γw .
The momentum equations for the vapor phase and the liquid phase are
∂vg 1 ∂vg2
α g ρg A + αg ρg A =
∂t 2 ∂x
∂P
−αg A + αg ρg Bx A − (αg ρg A)F W G(vg )
∂x
+Γg A(vgI − vg ) − (αg ρg A)F IG(vg − vf )
∂(vg − vf ) ∂vg ∂vf
−Cαg αf ρm A[ + vf − vg ] (5.3)
∂t ∂x ∂x
∂vf 1 ∂vf2
α f ρf A + αf ρf A =
∂t 2 ∂x
∂P
−αf A + αf ρf Bx A − (αf ρf A)F W F (vf )
∂x
−Γg A(vf I − vf ) − (αf ρf A)F IF (vg − vf )
∂(vf − vg ) ∂vf ∂vg
−Cαg αf ρm A[ + vg − vf ](5.4)
∂t ∂x ∂x
The phasic conservation of momentum equations are one-dimensional for the
phasic primitive velocity variables vg and vf . The spatial variation of momentum
61
term, or convection term, is expressed in terms of vg2 and vf2 . This form has the desir-
able feature that the momentum equation reduces to Bernoulli equation for steady,
incompressible, and inviscous flow. The frictional resistance due to the fluid viscosity
and the flow turbulence is lumped as the flow resistant factors in the momentum
equations. The force terms on the right sides of Equations 5.3 and Equation 5.4 are,
respectively, the pressure gradient, the body force (i.e., gravity and pump head), wall
friction, momentum transfer due to interface mass transfer, interface frictional drag,
and force due to virtual mass.
The momentum equations are based on the following simplifications:
i The Reynolds stresses are neglected.
ii The phasic pressures are assumed to be in equilibrium.
iii The interfacial pressure is assumed to be equal to the phasic pressures.
iv The covariance terms are universally neglected.
v The interfacial momentum storage is neglected.
vi The phasic viscous stresses are neglected.
vii The interface force terms consist of both pressure and viscous stresses.
The terms FWG and FWF are part of the wall frictional drag, which are linear
in velocity, and are products of the friction coefficient, the frictional surface area per
unit volume, and the magnitude of the fluid bulk velocity. The FWG and FWF
normally are determined through experiments or CFD simulations. The coefficients
FIG and FIF are part of the interface frictional drag.
Conservation of momentum at the interface requires that the force terms
associated with interface mass and momentum exchange add up to zero, and is given
62
as
[ ]
∂(vg − vf )
Γg AvgI − (αg ρg A)F IG(vg − vf ) − Cαg αf ρm A
∂t
[ ]
∂(vf − vg )
+Γg Avf I − (αf ρf A)F IF (vf − vg ) − Cαf αg ρm A =0 (5.5)
∂t
where the spatial derivatives have been eliminated. This particular form for interface
momentum balance results from consideration of the momentum equations in unex-
panded form. The force terms associated with virtual mass acceleration in Equation
5.5 add up to zero identically as a result of the particular form chosen. In addition, it is
usually assumed (although not required by any basic conservation principle) that the
interface momentum transfer due to friction and due to mass transfer independently
add up to zero, that is, vgI = vf I = vI and αg ρg F IG = αf ρf F IF = αg ρg αf ρf F I.
The phasic thermal energy equations are
∂ 1 ∂
(αg ρg Ug ) + (αg ρg Ug vg A) =
∂t A ∂x
∂αg P ∂
−P − (αg vg A) + Qwg + Qig
∂t A ∂x
+Γig h∗g + Γw h′g + DISSg (5.6)
∂ 1 ∂
(αf ρf Uf ) + (αf ρf Uf vf A) =
∂t A ∂x
∂αf P ∂
−P − (αf vf A) + Qwf + Qif
∂t A ∂x
−Γif h∗f − Γw h′f + DISSf (5.7)
These equations are based on the following simplifications:
i The Reynolds heat flux is neglected.
ii The covariance terms are universally neglected.
iii The interfacial energy storage is neglected.
63
iv The internal phasic heat transfer is neglected.
In the phasic energy equations, Qwg and Qwf are the phasic wall heat transfer
rates per unit volume. These phasic wall heat transfer rates satisfy the equation
Q = Qwg + Qwf , where Q is the total wall heat transfer rate to the fluid per unit
volume.
The phasic specific enthalpies (h∗g , h∗f ) associated with bulk interface mass
transfer in Equations 5.6 and 5.7 are defined in such a way that the interface energy
jump conditions at the liquid-vapor interface are satisfied.
The phasic energy dissipation terms, DISSg and DISSf , are the sums of
wall friction, pump, and turbine effects. The dissipation effects due to interface mass
transfer, interface friction, and virtual mass are neglected. This is an acceptable
assumption since these terms are small in magnitude relative to the other terms in
the energy equation. In the mass and momentum equations, interface mass transfer,
interface friction, and virtual mass are important, and are not neglected. The wall
friction dissipations are defined as
DISSg = αg ρg F W Gvg2 (5.8)
DISSf = αf ρf F W F vf2 (5.9)
The phasic energy dissipation terms satisfy the relation
DISS = DISSg + DISSf (5.10)
where DISS is the total energy dissipation. When a pump component or a turbine
is present, the associated energy dissipation is also included in the total dissipation
terms.
5.1.2 Neutron Kinetics
Neutron kinetics capabilities in RELAP5-3D have been expanded from a best estimate
thermal hydraulic code. Latest version of RELAP5-3D code has two options available
64
for the computation of the reactor power.
The first option for neutron kinetics in RELAP5-3D is the point reactor ki-
netics model. This option had been available in earlier versions of RELAP5. It was
sufficient for a one-dimensional thermal hydraulics code like RELAP5/MOD3. But
with the multi-dimensional thermal hydraulics capabilities in RELAP5-3D, multi-
dimensional neutron kinetic model was desirable. The second option for reactor
physics in RELAP5-3D is a multi-dimensional neutron kinetics model based on the
NESTLE code developed at North Carolina State University. This option computes
the reactor fission power in either Cartesian or Hexagonal geometry. The NESTLE
code solves the two or four group neutron diffusion equations. The Nodal Expansion
Method (NEM) is used to solve the neutron diffusion equations for the neutron flux.
A flexible neutron cross section model and a control rod model have been imple-
mented to allow for the complete modelling of the reactor core behavior. The decay
heat model developed as part of the point reactor kinetic model has been modified
to compute the decay power for both the point reactor kinetics model and for the
multi-dimensional neutron kinetics model.
Three, two, or one-dimensional models may be used. Several different core
symmetry options are available including quarter, half, and full core options for Carte-
sian geometry and 1/6, 1/3, and full core options for hexagonal geometry. Zero flux,
non-reentrant current, reflective, and cyclic boundary conditions are available. The
steady-state eigenvalue and time dependent neutron flux problems can be solved by
the NESTLE code as implemented in RELAP5-3D.
5.2 Coupled CFD—System-Code Approach
Nuclear system codes like RELAP5 are still the predominant simulation tools in
nuclear industry. But using CFD to model the nuclear system is becoming more
65
attractive to nuclear engineers. Due to the size and complexity of nuclear system
geometry, it is not feasible to use CFD alone to solve the kind of problems currently
addressed using the system codes like RELAP5. Hence, until CFD reaches the level
where complete nuclear thermal hydraulic systems can be simulated using CFD alone,
an interim solution is to use a ”coupled approach,” in which CFD is used where it is
necessary and system code is used in the remaining portions of the system.
Design calculation and safety analysis of the complex thermal-fluid systems
of GEN IV reactors, such as GT-MHR or PBMR require different levels of simula-
tions, including detailed component-level CFD simulations to integrated system-level
simulations. Coupled simulation tools are expected to optimize the design and as-
sist in design of experiments to fill design gaps, as well as possibly eliminate some
experimental tests [Diamond, 2003].
As mentioned above, system codes are very crude, however, CFD along can-
not be currently used to replace the system codes. The solution to this problem is
to couple a system code with a three-dimensional CFD code, and apply the CFD
code to simulate flow in parts and components where system codes do not perform
adequately, such as lower plenum, downcomer, and possibly even reactor core while
taking advantage of system codes to simulate flow in components where flow can be
adequately simulated by one-dimensional flow.
Two options are generally available for code coupling. In the first one, dif-
ferent codes are merged into an integral one. This method integrates the solution
matrices of the codes and solves the problem with semi-implicit numerics. In the
second approach, both codes execute independently and couple across the computa-
tional domain boundary externally. The advantage of the first approach is that the
numerical scheme of the merged code does not suffer from the kind of stability prob-
lems faced by codes coupled via an interface. The disadvantage of the first option
is that access is needed to the source of both codes. Moreover, due to significantly
66
varying code structure, the task of merging codes can be horrendously complicated.
The fact that source code of one of the two codes to be coupled is not accessible,
makes it necessary to follow the second approach. [Weaver et al., 2002].
RELAP5-3D and FLUENT are respectively chosen as the system code and
CFD code for this development. The geometry or domain to be analyzed is divided
into regions that must be modelled using the CFD approach and regions that can be
reasonably well simulated using a system code. This division identifies the interfaces
at which fluid is transferred from the system-code-portion to the CFD-code-portion
and vice versa. System variables must be transferred at these coupling interfaces
from one code to the other. Different definitions of some of the variables, different
meshes and different time steps used by the two codes pose interesting challenges. In
this coupling paradigm, there are two solvers running simultaneously over separate
parts (system-code-portion and CFD-code-portion) of the same problem, so that in-
formation must be exchanged either with an explicit or semi-implicit interface. This
requires considerable programming effort to maintain consistency between the two
codes.
Development, implementation and application of the coupled FLUENT and
RELAP5 code to large integral system is a significant challenge. It requires a very
good understanding of the ”guts” of the two codes. Due to the coupling paradigm
chosen, two matrices from two separate codes are to be solved independently.
Below, in the remainder of this chapter, a software package used for code
coupling as well as some past work on code coupling are reviewed.
5.2.1 PVMEXEC
Idaho National Laboratory laid the ground work to couple RELAP5-3D with FLU-
ENT [Schultz and Wieselquist, 2001]. Specifically, a general purpose framework called
PVMEXEC, based on Parallel Virtual Machine (PVM) [Geist and et al.., 1994], was
67
developed to allow coupling of RELAP5-3D with another simulation code by [Weaver
et al., 2002]. It is claimed that the coupling interface is fairly generic and any code that
implements the coupling application program interface (API) can be used in a cou-
pled simulation [Weaver, 2005]. This methodology was verified for RELAP/RELAP
coupling, and RELAP5/CFD coupling with a simple and specific RELAP5 bench-
mark problem (Edwards’ Pipe Problem) [Aumiller et al., 2001]. A coupled system
code–CFD analysis of the high temperature helium flow in the outlet plenum of a
VHTR using PVMEXEC has been recently reported [Anderson et al., 2008]. The
results of pressure, mass flow rate, and temperature across the coupled boundary
showed a differences of less than 5 percent in every location except for one channel
of the total eleven simplified reactor core coolant channels.
The PVMEXEC program was developed at Idaho National Laboratory to
facilitate the coupled simulation of a nuclear thermal hydraulic system using several
different computer programs (such as COBRA2, RELAP5-3D, TRAC-P, FLUENT5
and TRACE6 for the fluid systems, CONTAIN7 and MELCOR8 for the containment
systems, NESTLE9 and PARCS10 for reactor kinematics) to investigate the steady
state or transient behavior of the system.
The PVMEXEC program and the code coupling methodology enable the use
of different codes for the simulation of different portions of the system in a unified
analysis of the transient behavior of the system. Each code in the coupled simulation
needs data from some of the others codes in the simulation. The data is passed among
the codes using the PVM message passing methodology. The PVMEXEC program
was originally developed to couple the RELAP5-3D code to other thermal hydraulic
codes, however it is claimed that any code that implements the PVMEXEC API may
be used in a coupled simulation. The coupling system that the PVMEXEC program
implements was designed as a batch system. The PVMEXEC program is executed
by the user, and the simulation is expected to proceed to the end without additional
68
user interaction with the PVMEXEC program.
The PVMEXEC program has several tasks. These tasks roughly correspond
to the several sections of its input file. These include: 1) configuring the virtual
machine, 2) executing several simulation codes on the virtual machine, 3) manage
the data exchange among simulation codes, 4) manage the time advancements of the
simulation codes participating in the coupled simulation, and 5)writing the output
file. Another responsibility of the PVMEXEC program is to terminate the coupled
computation gracefully if any of the simulation codes terminate unexpectedly or en-
counter problems from which they cannot recover. The PVMEXEC program writes
information to its output file. It also writes the status information to the terminal
screen so that the user may monitor the progress of the simulation.
Figure 5.1: Coupled simulation diagram for PVMEXEC
PVMEXEC was validated using a coupled case including two RELAP5-3D
processes [Aumiller et al., 2001]. But using PVMEXEC to couple RELAP5-3D and
FLUENT requires source code level of programming in both codes. As shown in Fig-
ure 5.1, PVMEXEC runs as the master code and controls the processes of RELAP5-
3D and FLUENT. Both slave codes need to communicate with their master code,
PVMEXEC. Unfortunately, access to the source code of a commercial CFD code
like FLUENT is impossible without the code developer’s support. As a result, an-
other coupling methodology is employed in this thesis. Before the introduction of
the coupling methodology in Chapter 6, other coupled codes simulations in nuclear
69
application are briefly reviewed in the next subsection.
5.2.2 Coupled CFD—System-Code Simulations Review
Coupled CFX and TRACE simulations were carried out in a collaboration among
Paul Scherrer Institute (PSI), Ecole Polytechnique Federale de Lausanne (EPFL), and
Eidgenosische Technische Hochschule Zurich (ERHZ) [Bertolotto et al., 2008]. Single-
phase mixing studies were carried out by the coupled CFX-TRACE. The validation
of the coupled code was through the comparison between results from coupled code,
CFD code and system code. A simple mixing experiment was used to validate the
coupled code. Figure 5.2 is the coupled code scheme used in the CFX-/TRACE
coupling. PVM was the communication framework in that coupled code.
Figure 5.2: Coupled CFD-TRACE code scheme [Bertolotto et al., 2008]
Another coupled code development was reported using CATHARE and TRIO U
by Commissariat al Energie Atomique at Grenoble, France. The whole reactor was
modelled with the CATHARE system code and the three-dimensional upper plenum
was modelled with the TRIO U CFD code. CATHARE provided TRIO U with flow
rates and inlet temperatures, and received the temperature of the flow exiting the
plenum. The coupled code is applied to the gas cooled fast reactor. [Perdu and
70
Vandroux, 2008]
Most coupled simulation efforts in literature are performed either on single-
phase flow system or on gas cooled reactor. This is due to the fact that multi-
phase models in CFD codes are not as mature as the multi-phase models in nuclear
system codes. The latest multi-phase models available in commercial CFD codes,
such as FLUENT, are Volume of Fraction (VOF) model, mixture model, and Eulerian
multi-phase model. The multi-phase model in CFD can simulate the phenomena like
noncondensible gas injection or bubble transport in flow field. The incompatibility
of multi-phase models between nuclear system code and CFD code makes it very
difficult to couple them in multi-phase applications.
Multi-phase CFD (MCFD) enhancement is beyond the scope of this thesis
work. Therefore, the objective of this thesis is limited to the large scale coupled
thermal hydraulic simulation to the VHTR reactor, GT-MHR, which has a single-
phase reactor coolant system.
5.2.3 Large Scale Simulation using Coupled CFD—System-
Code
To perform large scale simulation which takes advantage of both types of computa-
tional tools is the reason to develop the coupled CFD—system-code capability. CFD
results provide the three-dimensional flow field. Extensive 3D thermal hydraulic in-
formation could be obtained for nuclear engineers. The detailed flow conditions can
be valuable to develop certain components in an advanced reactor. Therefore, CFD
modelling in certain regions of the reactor vessel could be a desirable enhancement
to the traditional large thermal hydraulic simulation using nuclear system codes. In
addition to the multidimensional flow simulation, coupled CFD—system-code simu-
lation potentially is able to include other physics, such as reactor neutronics codes
to incorporate the three-dimensional reactor kinetics module into the thermal hy-
71
draulic model. Reactor physics simulation with coupled MCNP and CFD code has
been found in literature ([Seker et al., 2007]). Working with Finite Element Analysis
(FEA) stress analysis tools, CFD can also investigate the thermal stress and ther-
mal fatigue for structure integrality analysis in a nuclear system. In this thesis the
large scale simulation using coupled CFD—system-code approach is demonstrated by
applying it on a PWR and the GT-MHR.
72
Chapter 6
CFD—System-Code Coupling
using UDF
A coupled CFD—system-code simulation capability is developed in this thesis based
on FLUENT and RELAP5-3D. The methodology to couple FLUENT and RELAP5-
3D is determined by the availability of the source code. The source code of RELAP5-
3D is accessible. But as a commercial CFD code, FLUENT’s source code is not
available to general researchers. The development of the coupled CFD—system-code
is hence accomplished by User-Defined Function (UDF) in FLUENT. In the coupled
CFD—system-code structure, FLUENT is the master code, while RELAP5-3D is
executed as the slave code, controlled and synchronized by the master code. FLUENT
UDF is the programming interface to couple the nuclear system-code and the CFD
code in this thesis work.
6.1 User-Defined Function
User-Defined Function in FLUENT is an advanced application tool which allows the
FLUENT users to significantly enhance the CFD code’s modelling capabilities. Using
the C language programming interface, UDFs allow the user to customize the CFD
73
code for special applications, and to simulate the conditions that are otherwise not
possible to be modelled using FLUENT. UDFs may be used to realize a variety of
applications, some of which are listed below[FLUENT Inc., 2005a]:
i Customization of boundary conditions, material property definitions, sur-
face and volume reaction rates, source terms in transport equations, source
terms in User-Defined Scalar (UDS) transport equations, diffusivity func-
tions, etc. For instance, a velocity or temperature profile could be provided
as boundary condition if it is known or is being calculated by other codes.
ii Adjustment of computed values on a once-per-iteration basis. External
subroutines or functions could be included in the FLUENT framework
using UDFs.
iii Initialization of a solution. Accurate initialization is not only crucial for
convergence in steady state problem, but also very important to obtain
correct transient simulation result.
iv Asynchronous execution of a UDF.
v Post-processing enhancement. Values which are not provided by standard
FLUENT user interface could be calculated by UDFs. For example, the
recirculation ratio reported in replacement research reactor CFD study is
not defined by FLUENT. Without the help of UDFs, the values could not
be calculated.
vi Enhancement of existing models (e.g., turbulence model, porous media
turbulence model, phase-change model, etc.). This is probably the most
valuable function of FLUENT UDFs. It makes FLUENT a very powerful
platform suitable for many CFD related research.
74
Testing a new model by programming from scratch requires significant effort.
While using UDFs is quicker and more flexible. It allows the researcher to concen-
trate on the physics instead of numerical and programming issues in the process.
Implementing new models for nuclear applications can take advantage of UDFs.
It is only through UDFs that a commercially available code can be a research
tool. In fact, by releasing the researcher from the elementary details of fairly time
consuming but well understood steps, and allowing the researcher to focus on ”the
next step,” these commercial codes may even help in speeding up the research (as
opposed to only development) enterprise. As an example, a researcher who develops
a new porous media model can now, using UDF, test the new model on very realistic
applications without having to write a computer code to solve the Navier-Stokes
equations.
FLUENT UDFs programming interface provides C types, functions, and pre-
processor macros to facilitate the programming of UDFs, and the use of CFD objects
as defined inside FLUENT. Certain macros allow the UDFs to access data in a FLU-
ENT solver such as cell variables (e.g., cell temperature, cell density, centroid), face
variables (e.g., mass flux, face velocity components, area), or connectivity variables
(e.g., adjacent cell thread and index) that can be used in a computation in the UDFs.
The solver variables are stored in the FLUENT data structure called domain.
Domain is further divided into threads, which correspond to certain boundary con-
ditions and internal volumes in the CFD model. Thread is a group of cells or faces
in the CFD computational mesh. A special set of macros commonly used in UDFs,
that return such values as the thread ID pointer (an internal FLUENT structure)
when passed the Zone ID (the number assigned to a zone in a boundary conditions
panel), is provided. Some UDF macros that enable the users to loop over nodes, cells,
and faces in a thread or domain in order to retrieve and set values are also provided.
Another useful macro that enables the UDF to set a boundary condition value in the
75
solver is also available.
Practically, FLUENT UDFs enable users to access each element in the CFD
mesh and most variables in the solver matrix. UDFs provide the loop-function to
access the cells and faces in a thread. The thread ID could be obtained through the
standard graphical user interface of FLUENT. So UDFs can be used to customize
some calculations and model certain boundary zones or volumes zones in the CFD
model.
In Chapter 4 UDFs were employed to add in FLUENT a macroscopic tur-
bulence model for porous media zones. UDFs were used to test and validate the
volume-averaged k − ϵ turbulence model for porous media. UDFs are used in this
chapter to develop the coupled simulation capability using a system code (RELAP)
and a CFD code (FLUENT).
6.2 Coupling Interface in Coupled CFD—System-
Code
Figure 6.1 shows the sequential coupling scheme. The overall computational domain
is divided into CFD portions and system portions, which are interconnected at the
interface boundaries in the coupled simulation.
Figure 6.1: A conceptual coupled CFD—system-code model
The CFD portions and nuclear system code portions may have one or more
than one interfaces. Each CFD or system code portion may have more than one inlet
or outlet interfaces. The flow direction across the interface is predetermined based
76
on the overall thermal hydraulic system. Current implementation does not simulate
flow reversal at interfaces correctly. The simulation variables on both sides of the
interfaces are exchanged between the system code and the CFD code.
The upstream domain provides the mass flow rate ṁ, the static pressure p,
and the temperature T to the inlet boundary of the downstream domain. The inlet
boundary at the downstream domain is a ghost cell which overlaps with the compu-
tational cell next to the outlet boundary in the upstream domain and is updated by
the upstream domain after each time step during the coupled simulation.(Figure 7.3
in the next chapter explains these data exchange schemes on the coupling interface.)
Computational domain downstream also provides the outlet boundary con-
ditions to the upstream domain. The outlet boundary condition in the upstream
domain is also a ghost cell updated by the downstream computation result after each
time step. The variables obtained from downstream are static pressure p and tem-
perature T . If flow reversal takes place, the properties of the backward flow at the
outlet of the upstream domain are calculated based on the downstream results.
RELAP5-3D and FLUENT have differences in their governing equations, for-
mulation system, and variable definition. For instance, the pressure variable ex-
changed between the system and CFD code is absolute static pressure. CFD code
provides gauge static pressure when UDFs access the solver variables. The value
should be adjusted to absolute static pressure before it is passed to RELAP5-3D.
UDFs also take care of the conversion of the variables with different definition in the
two component codes.
Turbulence modelling methodologies are fundamentally different between the
nuclear system code and the CFD code. In the nuclear system code, turbulence is
not explicitly modelled. The friction loss coefficients in the momentum equations
of RELAP5-3D take into account the effects of different pressure loss phenomena,
including turbulence. The extra pressure loss incurred by turbulence is lumped into
77
the model parameters for each control volume. The heat transfer enhancement by
turbulence is included in the empirical correlation used in the heat transfer coefficient
calculation. On the other hand, CFD codes model turbulence explicitly. The differ-
ence makes turbulence variables across the interface of system code and CFD code
incompatible.
The coupled system-CFD code does not exchange the turbulence variables
across the interfaces. This is adequate for nuclear system code since no turbulent
boundary condition is required. However, the turbulence boundary condition for
CFD portion needs turbulence intensity at the inlet boundary and outlet boundary
to be specified. The turbulence boundary condition used at CFD interface boundary
is specified based on past experience with CFD simulations. The turbulence intensity
in most engineering thermal hydraulic system practically never exceeds 10%. The tur-
bulence intensity on the CFD portion boundary in coupled simulations is arbitrarily
set as 4%.
6.3 CFD—System-code Coupling using UDFs
6.3.1 Semi-Implicit Coupling
Mass conservation and energy conservation are the primary requirements while cou-
pling CFD and system codes. The mass flow rate is passed from the upstream com-
putational domain to the downstream domain. And pressure calculated in the down-
stream code is passed back to the upstream code across the coupling interface. The
information exchange takes place at the end of each simulation time step. Using
this coupling scheme, mass and energy are conserved. But momentum may not be
conserved in the coupled CFD—system-code.
In general, to conserve momentum at the interfaces and hence in the coupled
code, explicit coupling scheme should be used. In explicit coupling, the solution for
78
new time step depends only on the system state at the previous time step. The
only new variables that appear in the matrix of the governing equations are in the
representation of the temporal derivative. Since sonic phenomena (i.e. pressure wave
propagation and resultant velocities) are not treated implicitly in the spatial terms,
explicit coupling requires that the sonic Courant limit
∆x
∆t < (6.1)
vsonic
be satisfied. Application of fully explicit numerics results in impractical execution
times for realistic problems due to the time step restriction. In order to apply coupled
CFD—system-code to large scale applications a solution to the limitation on time step
must be found.
The coupled CFD—system-code in this work is based on a semi-implicit
scheme. In this scheme both codes run in transient mode and exchange informa-
tion at coupling interfaces after each time step. The pressure boundary condition
used at the outlet of the upstream code is based on the solution at the previous time
step. The flow rate information passed to the downstream code is current time step
information. As a result, the solution for new time step depends on both, previous
and current, time step values. The time step limit for semi-implicit coupling scheme
is thus based only on the velocity at the interface
∆x
∆t < (6.2)
vf low
Using the semi-implicit method enables the coupled code to overcome the sonic
Courant limit. However, the time step necessary to overcome numerical instabil-
ities in the coupled code simulations is still much smaller than the time step for
simulations using the CFD-code only or system-code only. As a result, the coupled
CFD—system-code in its current state is not very practical for slow transient sce-
narios like Small-Break Loss-of-Coolant Accident (SBLOCA) in a PWR, which may
take several hours to evolve.
79
6.3.2 Coupled Code Structure
In the coupled CFD—system-code, FLUENT is the master code, while RELAP5-3D
runs as slave code under the control from FLUENT. UDFs provide the code coupling
interface between FLUENT and RELAP5-3D. Figure 6.2 is the diagram of coupling
method of the CFD—system-code.
Figure 6.2: Coupling scheme by FLUENT UDF programming interfaces
Written by the user in the C language, UDFs provide the flexibility to cus-
tomize the FLUENT code for specific applications.
An UDF macro, which is executed at the end of every iteration for steady state
CFD simulations or at the end of each time step for transient CFD simulations, is the
container function of the slave (RELAP5-3D) code. The RELAP5-3D source code
is modified and compiled into a set of dynamic link libraries (RELAP5-3D DLLs),
initialized (Appendix A), and executed within the container functions. Appendix
B shows a representative portion of the container function. RELAP5-3D is loaded
by the UDF using the C and FORTRAN hybrid programming. UDFs also control
the data exchange process between the CFD and the system codes. The UDF that
specifies the interface boundary condition of the CFD portion using the information
from the system code portion is given in Appendix C. The UDF to synchronize the
80
CFD code by the time step size from RELAP5-3D is given in Appendix D.
6.3.3 Discussion
The advantages of the coupling methodology employed in this thesis include availabil-
ity, flexibility, and expansibility. First of all, the tools used in this thesis are available
to general public as commercial software. The coupling methodology demonstrated
could be duplicated without prohibitive limitation. Secondly, the methodology used
in the coupled CFD—system-code requires a unique set of UDFs for each coupled sim-
ulation model. Finally, it is relatively convenient to include other, such as neutron
kinetics, models into the coupled thermal hydraulic simulation.
The next three chapters cover the validation of the coupled code; the appli-
cation to a PWR; and the application to a Gen IV reactor, (GT-MHR), respectively.
81
Chapter 7
Coupled CFD—System-Code
Validation
A simple pipe flow simulation is performed to verify the coupled CFD—system-code
capability which is developed in the previous chapter. The coupled simulation results
are compared with CFD-only and RELAP5-only results for validation.
As the proof-of-principal simulation for the coupled CFD—system-code ca-
pability, an artificial flow problem is selected. The flow system consists of a network
of pipes with helium as the working fluid. Figure 7.1 is the isometric view of the pipe
network. In the piping system helium from the inlet splits into two parallel vertical
pipes, which then connect back through a tee branch near the top. The helium flow
then turns 180 degrees and exits at the outlet. The inner diameter of the pipes is 60
mm. The piping network is 2000 mm in height. The helium in the pipe network is
at 550 K, 0.5 MPa.
7.1 CFD-only Simulation
CFD-only simulation is carried out using the model shown in Figure 7.1. The oper-
ating pressure in the CFD model is set at 0.5 MPa. Inlet helium temperature is set
82
Figure 7.1: CFD model for the simple pipe
at 550K. The ideal gas model is employed for helium state calculation. The k − ϵ
turbulence model with the standard wall function for near wall treatment is used in
the CFD model.
Inlet boundary gauge pressure (pressure beyond the operating pressure) is
100 Pa, while outlet gauge pressure is 10 Pa. Flow in the piping system is driven by
a 90 Pa differential pressure between inlet and outlet. The CFD-only simulation is a
transient analysis in which the flow starts from stationary.
Mesh refinement study has been carried out to obtain the transient mass flow
rate used for the coupled CFD—system-code validation. The study has been done
by comparing results from two sets of computational mesh with 160k cells and 530k
cells respectively.
7.2 RELAP-only Simulation
RELAP-only simulation uses the nodalization shown in Figure 7.2. Two time depen-
dent volumes specify the inlet and outlet boundary conditions for the RELAP5-3D
83
model. Inlet boundary absolute pressure is set at 500,100 Pa, while outlet absolute
pressure is 500,010 Pa. Fluid property database in RELAP5-3D package is used
for helium state calculation. The initial condition of the RELAP5-3D model is also
stationary.
Figure 7.2: RELAP nodeliztion for the simple pipe
The RELAP5-3D simulation and the CFD simulation have identical bound-
ary conditions and initial conditions. But as a one-dimensional code, RELAP5-3D
does not explicitly solve turbulence model as a CFD code does. The turbulence effect
is lumped in the pressure loss model in RELAP5-3D implicitly. Therefore, the com-
ponent (such as bends and branches) flow resistance factors along the flow path in
RELAP5-3D model need be tuned to predict the correct pressure drop and flow rate.
The steady-state results from the CFD simulation are used to tune the resistance
factors in RELAP5-3D model.
84
7.3 Pipe Flow Simulation by Coupled CFD—System-
Code
In coupled CFD—system-code simulation the piping system is broken up artificially
into a system-code portion and a CFD portion. The coupled model diagram is shown
in Figure 7.3. The left (colored) portion, with upward flow, is modelled using the
CFD code, and the right hand (nodalized) portion is modelled using the system code.
Figure 7.3: Coupled CFD—system-code model for the simple network of pipes
The CFD portion model setup in coupled CFD—system-model is the same
as in the CFD-only model. The tuned resistance factors in the RELAP5-3D model
are also used in the RELAP5-3D portion of the coupled simulation. Inlet (in CFD
portion) gauge pressure is 100 Pa, while outlet absolute pressure (in system-code
portion) is 500,010 Pa. The flow in the transient simulation performed using the
coupled CFD—system-code also starts from rest.
85
The components of the coupled code, RELAP5-3D and FLUENT, have some
fundamental differences. Before the coupling between the two codes, the variables
exchanged between the component codes should be determined.
7.3.1 Boundary Conditions and Coupling Boundary Inter-
face
Accurate prediction of coolant inventory in the reactor coolant system is the primary
requirement for nuclear safety analysis. Temperature and pressure in nuclear ther-
mal hydraulic systems are also critical for performance evaluation. Therefore, mass
conservation and energy conservation have higher priority for a nuclear thermal hy-
draulic simulation than momentum conservation. Because there are many significant
momentum sources (pumps) and sinks (valves and flow meters) in any nuclear thermal
hydraulic systems normally, momentum conservation is of secondary importance.
As a result, mass flow rate, pressure, and temperature are the variables se-
lected to pass across the coupling boundary interface in the coupled CFD—system-
code. The mass flow rate going into the downstream component code is the mass flow
rate that exits upstream component code at every time step. This setup ensures the
mass conservation in the coupled CFD—system-code simulation. Passing pressure
and temperature between two component codes guarantees the energy conservation.
But momentum conservation requirement may not be strictly met. Firstly,
the two component codes use different calculation methods for working fluid property.
RELAP5-3D uses helium property database. FLUENT employs the ideal gas model.
Identical mass flow rates across the coupling boundary may not guarantee that flow
velocities are identical. Secondly, the coupling interfaces across the coupling boundary
may not be conformal (see the coupled model for GT-MHR in Chapter 9) which leads
to inevitable momentum added or lost during coupling simulations.
Note that RELAP5-3D does not have any explicit modeling of turbulence.
86
Therefore, turbulence model variables are not passed between the CFD portion and
the RELAP5-3D portion during coupled simulation. Constant turbulence intensity is
used as the boundary condition for the turbulence model on the interface boundary
of the CFD portion.
The system variables exchanged across the coupling interface are indicated in
Figure 7.3. The outlet helium temperature, pressure, and mass flow rate calculated in
the CFD portion are passed to the downstream RELAP5-3D portion as inlet boundary
conditions. Helium temperature and pressure are passed from the system-code portion
to the upstream CFD portion as outlet boundary conditions. These flow field variables
are passed back and forth across the CFD portion and system-code portion at the
end of each time step.
7.3.2 Time Step Size
The time step employed in the simple pipe coupled simulation is 2.5 × 10−4 sec. Time
steps larger than this value did not lead to a converged solution. This time step size
confirms the interface velocity Courant limit on the time step, Equation 6.2, for the
semi-implicit coupling scheme.
7.4 Results and Comparison
The time dependent mass flow rates obtained from CFD-only, RELAP-only, and
coupled code simulations are compared in Figure 7.4. CFD results from two sets
of computational mesh for mesh refinement study are included. Good agreement is
obtained for the steady state mass flow rates from CFD-only, RELAP5-3D-only, and
coupled code simulations. The transient mass flow rate ramps up to the steady state
flow rate in approximately 0.3 sec, and there is a good agreement even during the
rapid transient that lasts about 0.3 sec.
87
Figure 7.4: Mass flow rate comparison between FLUENT, RELAP-3D(ATHENA), and
coupled code
Recall that momentum is not conserved across the interface between the two
code systems. To quantify the gain/loss of momentum across such interfaces, mo-
mentum on both sides of the interface is evaluated at convergence. It is noted that a
momentum loss of 0.7 % occurs across the coupling boundary in the verification case
at the steady state conditions. This momentum loss impacts only the coupled sim-
ulation in which CFD code models the upstream flow domain of the nuclear system
portion. Therefore, the PWR application (Chapter 8) and the GT-MHR application
(Chapter 9) do not experience this momentum loss.
7.5 Conclusion
This coupled CFD—system-code simulation verifies the innovative coupling method-
ology implemented by using FLUENT UDFs. The flow rate comparison validates that
the coupled CFD—system-code simulation is capable of predicting both the system
steady state conditions as well as the transient process.
88
Chapter 8
Large Scale Simulation of
Pressurized Water Reactor Using
Coupled Code
The coupled CFD—system-code simulation capability, which has been verified in the
previous chapter, is used to simulate the Nuclear Steam Supply System (NSSS) of a
Pressurized Water Reactors (PWR) as the demonstration of the large scale applica-
tion.
8.1 Pressurized Water Reactor
There are 104 commercial nuclear power reactor units in the United States which
generate 20% of the electricity consumed in the country. Sixty nine of them are
PWRs. The rest are Boiling Water Reactors (BWR).
PWR uses light water as reactor coolant and neutron moderator. The high
pressure (15.5 MPa) in the reactor coolant system (RCS), also called the primary
loop, keeps light water coolant from boiling. Under normal operating conditions,
two-phase flow does not occur in the primary loop except in the pressurizer. This
89
characteristics of PWR allows the application of CFD to simulate the RCS in Reactor
Pressure Vessel (RPV) without the recourse to still evolving multiphase CFD.
8.1.1 AP1000
Coupled CFD—system-code simulation of a PWR is performed in this chapter. The
AP1000 is selected as the prototype to develop the coupled PWR model.
The AP1000 is a design initiative from Westinghouse Electric Company LL to
develop an advanced generation III+, two-loop PWR design. The United States Nu-
clear Regulatory Commission (U.S.NRC) certified the AP1000 design in January 2006.
The combined Construction and Operating Licenses (COLs) to build the AP1000
power plants have been applied for several sites in the United States. In the mean-
time four AP1000 units are under construction at Haiyang and Sanmen, China. The
first AP1000 unit will be connected to the power grid in 2016.
Figure 8.1: Westinghouse AP1000 nuclear steam supply system [Schulz, 2006]
90
Figure 8.1 is the NSSS of the AP1000. The AP1000 plant generates approx-
imately 1100 MWe during normal operation. With simplification compared to older
PWR designs, total pipe length, valves, and pumps are reduced significantly in the
AP1000. Modular features expedite the construction of the AP1000 plant. Passive
safety systems are also included. Since the operational experience of the AP1000 is
not yet available, the evaluation of the thermal hydraulic system performance depends
heavily on the system level simulation with necessary details and adequate models.
The coupled simulation models a simplified AP1000. The dimensions and
thermal hydraulic conditions used in the model development are obtained from the
AP1000 engineering reports open to public in the NRC online library [Schulz, 2006].
8.1.2 Large Scale Thermal Hydraulic Simulation for PWR
A large scale thermal hydraulic simulation normally includes the complete NSSS and
other thermal hydraulic systems of interests. Instrument and Control (I&C) systems,
heat conduction in solid reactor core structure, and nuclear kinetics module could
also be incorporated into large scale simulations. System level response of a nuclear
power plant can be evaluated through a large scale simulation. For the accidental
scenarios which are difficult to study through experimental investigations, the large
scale thermal hydraulic simulation is the preferred option. Therefore, the large scale
thermal hydraulic simulation is an indispensable tool for the nuclear engineers.
8.1.3 RELAP Portion in Coupled Model
PWR nodalization in the RELAP5-3D code package is utilized to build the RELAP
portion in the coupled CFD—system-code model. The RELAP nodalization includes
the complete RCS and partial secondary loops of a typical PWR.
Figure 8.2 shows the pressure vessel in the RELAP nodalization. Reactor
coolant flow path is simplified to a one-dimensional flow with constant resistance
91
coefficients and predefined flow directions. The reactor pressure vessel in the RELAP
nodalization is replaced by a three-dimensional CFD model in the coupled CFD—
system-code simulation. Although the RELAP nodalization in Figure 8.2 is not used
in the coupled model, it assists in the development of the CFD model for the RPV.
In the pressure vessel nodalization shown in Figure 8.2, reactor coolant flow
enters from the four cold legs into the reactor vessel downcomer annulus in branch
300. The primary reactor vessel flow path is downward through branch 305 and
annulus 315 to the lower plenum, components 322 and 323. A portion of the inlet
flow is diverted around the downcomer through bypass pipe 320. This bypass is
a large-volume but low-flow region between the core former plates and core barrel.
Another portion of the inlet flow is diverted upward through pipe 310 and through
the upper reactor vessel bypass nozzles into the upper head, branch 355. Core inlet
branch 330 recombines the downcomer and bypass flows before entering the reactor
core that is represented by pipe 340. The upper plenum is represented by branches
355 and 356, and by pipe number 350. Branch 350 represents the guide tubes that
route a portion of the core exit flow from the upper plenum to the upper head.
Figure 8.3 is the RELAP nodalization of a RCS loop of the PWR NSSS.
When hot coolant leaves the reactor pressure vessel though hot legs, it exits the
control volume 345. The pressurizer (150) are connected to the primary side close to
the hot leg. The U-tubes in the steam generator (SG) is modeled by pipe 108 with
heat structure material attached. Some other components on the secondary side, such
as the main feedwater (182), auxiliary feedwater (184), main steam outlet (186), and
SG relief valves (188) are also included in the RELAP nodalization. After the coolant
returns from the steam generator, it is pumped back to the reactor pressure vessel
through the cold legs. There are two reactor coolant pumps in each loop. In RELAP
nodalization, they are lumped into one pump; component number 113.
The RELAP nodalization shown in Figure 8.3 (excluding the RPV) is inte-
92
Figure 8.2: Nodilizaition of a typical pressurized water reactor core [INEEL RELAP5-3D
Group, 1999]
grated in the coupled CFD—system-code simulation as the RELAP5 portion. It is
setup using the thermal hydraulic parameters for the AP1000 listed in Table 8.1.
93
Figure 8.3: RELAP nodalizaition of one loop of NSSS
8.1.4 CFD Portion in Coupled Model
Reactor pressure vessel is modeled by CFD in the coupled CFD—system-code simu-
lation. The RPV is simplified significantly for the CFD simulation. Only the major
flow paths in the reactor pressure vessel are retained in the CFD model. The complete
AP1000 reactor core is too complicated to be modelled accurately using a CFD code.
Therefore, the reactor core is modeled as porous medium. The diverted flow to the
upper plenum (pipe 310 in the RELAP nodalization) is included in the CFD model.
The downward bypass flow (pipe 320 in the RELAP nodalization) is not included in
the CFD portion because it is a low flow region.
The CFD computational mesh is built using GAMBIT in the commercial CFD
package. The mesh used in the CFD model has 1,053,936 hexahedral and tetrahedral
cells. Figure 8.4 and Figure 8.5 show the computational mesh of the reactor pressure
vessel on a vertical section and a horizontal section, respectively.
Standard k − ϵ turbulence model with the standard wall function is employed
in the CFD portion. The porous media model parameters for the reactor core region
are obtained through a set of assembly level CFD simulations. The porous media
94
Figure 8.4: Mesh in a vertical plane of the reactor vessel
methodology to model the reactor core has been demonstrated in the Replacement
Research Reactor CFD work reported in Chapter 3.
The reactor core power profile is included as volumetric heat source term.
A demonstrative reactor kinetics model is proposed and implemented in the CFD
Table 8.1: Nominal operational conditions of the AP1000 at full power
Primary Coolant System
AP-1000 Core Power 3,400 MWt
Reactor Coolant Mass Flow Rate 18,000 kg/sec
Reactor Cold Leg Temperature 280.7 ◦ C
Reactor Hot Leg Temperature 321.1 ◦ C
Reactor Core Temperature Increase 40.4 ◦ C
Reactor Cold Leg Pressure 15.71 MPa
Reactor Hot Leg Pressure 15.29 MPa
Cold Legs Flow Rate 19.87 m3 /s
Reactor Core Pressure Drop 430 kPa
95
Figure 8.5: Mesh in a horizontal plane of the reactor vessel
portion. The model is described in section 8.2.
Assembly Level CFD Simulations
The AP1000 reactor core is 3.04 meter in diameter, and consists of 157 fuel assemblies.
PWR fuel assemblies are composed of fuel rods which are held together by 4 or 6
spacers. The 17 × 17 reactor fuel assembly grid design has 264 fuel rods per assembly.
Assembly pitch is 215 mm. Fuel rod is 4.267 meters (14 feet) in length, 9.5 mm in
diameter. Fuel rods pitch is 12.6 mm with 0.57 mm cladding thickness. High speed
coolant flows along the bundle in the flow channels in between the fuels rods. Cross
flow exists in a PWR fuel assembly.
Figure 8.6: CFD model for 1/8 of a PWR assembly
By taking advantage of the symmetry, one eighth of a PWR assembly is
96
modelled using CFD to estimate the reactor coolant pressure loss through the AP1000
core. Figure 8.6 shows the CFD model for 1/8 of a assembly. A flow restriction curve
is obtained based on the assembly level CFD results.
The assembly level simulation uses the standard k −ϵ turbulence model. Mass
flow rate is set as inlet boundary condition, while pressure is set as outlet boundary
condition. A set of coolant mass flow rates are used to obtained the pressure drop
versus flow curve for fuel assembly in the reactor core.
Porous Media Model for Reactor Core
Porous media in FLUENT is modelled using additional source terms in the momentum
equation. Equation 2.18 or 3.1 is reproduced here
( 3 )
∑ ∑
3
1
Si = − Dij µvj + Cij ρvmag vj (8.1)
j=1 j=1
2
where Si is the source term for the j-th (x, y or z) momentum equation, vj is
the velocity component in j-th (x, y or z) direction, vmag is the velocity magnitude,
and µ is the fluid viscosity. The first term on the right hand side of Equation 8.1
is the viscous pressure loss term (Darcy term) due to the porous media structure.
The second term (Forchheimer term) represents the pressure loss due to the flow
disturbing by the solid structure in the porous media. The parameter tensor Dij ,
called viscous resistance factor, and Cij , called inertial resistance factor, should be
provided by the microscopic level CFD simulations. The viscous resistance factor
and the inertial resistance factor used in this work are estimated from the results of
detailed fuel assembly level simulations.
Figure 8.7 shows the pressure drop across the assembly as a function of average
coolant flow velocity. The velocities span the range in a reactor core under operational
conditions. Porous media model parameters are obtained from these assembly level
simulations by curve fitting. Parameter values are found to be, Dz = 8.17 × 107 1/m2
97
Figure 8.7: Pressure loss across a PWR assembly (CFD simulations).
for the viscous term and Cz = 5.37 1/m for the inertia term. These porous media
model parameters are subsequently used in the CFD portion of the coupled code.
8.1.5 Coupled CFD—System-Code Model for PWR
The coupled CFD—system-code model diagram for PWR NSSS is shown in Figure
8.8. The coupling interfaces of CFD portion and system portion are at the inlet to
the cold legs (300), and at the outlet of the hot legs (345) in the RELAP nodaliza-
tion for reactor pressure vessel. The reactor pressure vessel is modelled by the CFD
portion of the coupled model, while the rest of the NSSS and some supporting sys-
tems, including the pressurizer, the coolant pumps, feedwater system, and two steam
generators, are modelled as lumped-parameter components using RELAP5-3D. The
coupling methodology, which has been verified in the previous chapter, is used for the
data exchange across the interfaces of the CFD portion and the system-code portion.
98
Figure 8.8: Diagram of coupled large scale PWR hydraulic system model
8.2 Reactor Power Profile Calculation
Coolant temperature variation in the reactor core impacts the reactor power profile.
Cold coolant enhances the neutron moderation, raises the thermal neutron flux, and
increases the local fission reaction rate. Fission reaction rate is determined by fission-
able material density, fission cross section, and thermal neutron flux. More than 95%
of the reactor thermal power is produced directly from the fission reactions. Reactor
power profile is affected by the coolant thermal mixing in the reactor core.
The power profile change is then expected to impact the temperature distri-
bution. To capture the coupled nature of neutronics and thermal hydraulics inside
the reactor core, it is necessary to include a distributed reactor kinetics model (for
example via time dependent multigroup neutron diffusion equations in which cross
sections vary based on local temperature). Such models have been included in ther-
99
mal hydraulics codes such as RELAP. However, our focus here has been to model the
core using CFD, and hence the ability to couple core thermal hydraulics with core
neutronics available in system codes such as RELAP, is no longer available . One
option to resolve this would be to develop a detailed 3D reactor kinetics model and
link it using UDFs with the CFD code. This however is a major undertaking. There-
fore, an ad hoc extension of the point reactor kinetics model has been developed here
that does take the spatial effects into account. This ad hoc model is simply the point
reactor kinetics model applied to several nodes in the reactor core. While spatial
diffusion of neutrons is not taken into account, it is however possible to include the
temperature feedback effect (via coolant density and hence the moderation effect) at
discrete set of nodes in the core. A set of FLUENT UDFs (see Appendix B) are de-
veloped to incorporate the discrete point reactor kinetics equation model for multiple
nodes in the core, and coupled to the core thermal hydraulics. The model is described
in the next section.
8.2.1 Point Reactor Kinetics
Governing equations for the reactor kinetics model are well known [Lamarsh and
Baratta, 2001].
dn(t) ρ(t) − β ∑ N
= n(t) + λi Ci (t) (8.2)
dt Λ i=1
dCi (t) βi
= n(t) − λi Ci (t) (8.3)
dt Λ
where β is the delayed neutron fraction, Λ is defined as the mean generation time
between birth of neutron and subsequent absorption inducing fission, and λi is the
decay constance for i-th group of delayed neutrons precursors and i = 1, 2, · · · N . N
is the number of the delayed neutron groups. Classical reactor kinetic equations have
six groups of delayed neutron.
100
The set of seven coupled equation above describes both the time-dependent
neutron population n(t) in the reactor and the delayed neutron precursors population
Ci (t). ϕ(t) is thermal neutron flux density with assumption that v is the thermal
neutron speed. The total thermal power P (t) of a reactor could be calculated by
P (t) = Qf · ψ(t) (8.4)
where ψ(t) is the fission reaction Σf ϕ(t) rate and Qf is the average power released
per fission, which is approximately 200M eV in a PWR.
8.2.2 Temperature Coefficient of Reactivity
The feedback in the point reactor kinetics equations is via the reactivity as it responses
to changing temperatures and densities etc. The effect is usually taken into account
via an expression of the form ∆ρ = αT (T − Tref ). The change in reactivity per degree
change in the average core temperature is defined as the temperature coefficient of
reactivity, αT
dρ
αT = (8.5)
dT
where ρ is the reactivity of the core, and T is the average core temperature.
By introducing the definition of ρ(≡ 1 − k1 ) into the definition of temperature
coefficient, it can be written as

1 dk
αT = (8.6)
k 2 dT
Under normal operating conditions, k is very close to unity. Equation 8.6 is often
simplified to
1 dk
αT = (8.7)
k dT
Reactor temperature coefficient can be divided up between that due to the change in
fuel temperature and that due to the the change in moderator temperature.
101
The change in reactivity per degree change in moderator temperature is called
the moderator temperature coefficient of reactivity. A negative moderator tempera-
ture coefficient is desirable because of its self-regulating effect.
The fuel temperature has a greater feedback effect than the moderator tem-
perature for some reactors. The fuel temperature coefficient is the change in reactivity
per degree change in fuel temperature. This coefficient is also called the ”prompt”
temperature coefficient because an increase in reactor power causes an immediate
change in fuel temperature. A negative fuel temperature coefficient is generally con-
sidered to be even more important than a negative moderator temperature coefficient
because fuel temperature immediately increases following an increase in reactor power.
The overall temperature coefficient in a PWR is negative under normal oper-
ating conditions. In this thesis, a temperature coefficient, αT , of -50 pcm/◦ C [Duder-
stadt and Hamilton, 1976] is used.
8.2.3 Spatial Effect
Dimension of a light water reactor is about 200 neutron diffusion lengths. Therefore,
in a PWR, power profile has strong spatial distribution which is affected by boron
density, control rod position, temperature profile et al. It may take appreciable time
for a locally introduced reactivity to alter the spatial shape of power density through
out the reactor core. Point reactor model is not capable of evaluating this spatial
effect.
8.2.4 Power Profile Calculation in Coupled Model
The reactor power profile calculation in this coupled CFD—system-code model uses a
discrete reactor kinetics model. This model enables the coupled CFD—system-code
simulation to calculate the space- and time-dependent power profile in the reactor
core.
102
Discrete Reactor Kinetics Model
In general, one should solve the space- and time-dependent neutron diffusion equa-
tions with temperature dependent microscopic cross sections and temperature de-
pendent number density to couple the neutron diffusion equations with the energy
equation, and solve them simultaneously (probably iteratively). This is however a
major undertaking. Since the goal in this work is to demonstrate the feasibility of a
coupled neutronics-thermal hydraulics calculation using commercial CFD code (and
not to obtain necessarily the most quantitatively accurate results), it will rely on a
simple model to include the effects of thermal feedback in reactor kinetics.
This is achieved by assuming that the spatial diffusion of neutrons during the
transient follows the same behavior as that at steady-state. Hence, it can be ignored
during the transient phase. Spatially varying temperature and its feedback effect in
neutron kinetics are included via a set of discrete “point reactors models”.
The discrete reactor kinetics model extends the concept of the point reactor
model to a set of discrete locations in the reactor core. The reactor core is divided
into a set of ”nodes.” Point reactor kinetics equations are written for each node.
Initial conditions for these nodes are determined from the steady state power profile.
Though the model has its limitations, it does allow spatial and temporal evolution of
the reactor power to be coupled with the CFD code.
The assumptions made in the reactor power profile calculation are:
i Reactor core is assumed as a homogeneous core without reflector or core
blanket. The nonuniform nuclear fuel loading pattern is not considered.
ii Reactor core is modelled as a porous media for the thermal hydraulic
calculations. The heat generated in the fuels transferred to the coolant
instantly.
iii The temperature coefficient of reactivity is constant.
103
iv Control rods and reactivity controlling chemicals are ignored.
v Each node has its own ”reactivity.” This ”nodal reactivity” evolves as
a function of local nodal temperature and the temperature coefficient of
reactivity.
To further reduce the computational demands, the delayed neutrons precur-
sors in the coupled model are lumped into one effective delayed group, characterized
by a yield fraction
∑
6
β= βi (8.8)
i=1
and an averaged decay constant

( )−1
1 ∑ βi
6
λ= (8.9)
β i=1 λi
As a result, the point reactor model for each node j reduces to
dPj (t) ρj (t) − β

= Pj (t) + λCj (t) (8.10)
dt Λ
dCj (t) β
= pj (t) − λCj (t) (8.11)
dt Λ
where j denotes the j-th node.
Equation 8.10 and Equation 8.11 can be solved by assuming a solution of the
form
Pj (t) = Pj0 · eωj t (8.12)
Cj (t) = Cj0 · eωj t (8.13)
Pj0 and Cj0 are constants for node j determined by the initial conditions of the
reactor.
104
Equation 8.12 and Equation 8.13 are substituted in Equation 8.12 and Equa-
tion 8.13, and lead to the inhour equation for ωj
ωj Λ ωj β
ρj = + · (8.14)
1 + ωj Λ 1 + ωj Λ ωj + λ
Equation 8.14 is a quadratic equation for ωj . It has two roots, ωj1 and ωj2 .
For positive ρj , ωj1 is positive and ωj2 is negative. If ρj is negative, both ωj1 and ωj2
are negative. Time evolution of nodal Pj then can be written as
Pj (t) = Pj0,1 eωj1 t + Pj0,2 eωj2 t (8.15)
in which
λρj
ωj1 = (8.16)
β − ρj
β − ρj
ωj2 = − (8.17)
Λ
The power density Pj (t) for small change in reactivity becomes
( λρj
)
β t ρj β−ρj
Pj (t) = Pj0 · e β−ρj − · e− Λ t (8.18)
β − ρj β − ρj
This well known expression for power derived for the point reactor kinetics
model is here used in a discrete representation of the reactor core for each node, j. The
reactivity for each node j is a function of time, and depends upon the temperature
and/or density.
Reactor Kinetic Parameters
The parameters listed in Table 8.1 are used in the discrete reactor kinetics model in
the coupled simulation.
105
Table 8.2: Point reactor kinetics model parameters [Duderstadt and Hamilton, 1976]
Typical PWR Kinetics
αT temperature coefficient 2 × 10−4 1/◦ C
Λ averaged life time of neutron 10−3 sec
λ averaged decay constant of delay neutron 0.08 sec
β delay neutron yield fraction 0.0075
Local Neutron Multiplication Factor
A local neutron multiplication factor k(⃗x, t) is defined as the ratio between two gen-
erations of neutron density at location ⃗x. As discussed previously, the neutron mul-
tiplication for node j is affected by local coolant temperature and fuel temperature.
Equation for the local multiplication factor at time t + dt is as following
ρj (t + dt) = ρj (t)(1 + αT (Tj (t + dt) − Tj (t))) (8.19)
Multiplication factor, ρj (t + dt), at time t + dt is used to get reactor periods
ωj1 and ωj2 . Following equation calculates neutron density at time t + dt

( )
Pj (t + dt) = Pj (t) eωj1 (t)dt + eωj2 (t)dt (8.20)
In the transient coupled simulation, the temperature of node j in the reactor
core is provided by CFD result. A used defined memory (UDM) is created to store
the temperature at the previous time step. The temperature difference and reactivity
temperature coefficient are used to calculate the reactivity for the next time step
based on discrete point reactor model. Then the power density in node j can be
obtained from current power density and discrete reactivity. The heat source from the
discrete point reactor model will be passed to CFD simulation for thermal hydraulics
at the next time step. A spatial-dependent reactor power density in the discrete
point reactor model is solved using the CFD computational mesh at every time step.
106
Another UDM is defined for discrete reactivity as an additional field variable in CFD
solver. See Appendix B for the actual UDFs to perform the calculation for the discrete
point reactor kinetics and transient power density.
Initial Power Profile
With the homogeneous core assumption, the power profile is given by

( ) ( πz )
2.405r
P0 (r, z) = p0 · J0 · cos (8.21)
R H
where H is the height of the reactor core and R is the radius of the reactor core. The
power amplitude factor p0 is determined by the initial reactor total power.
8.3 Coupled Transient Simulation
The PWR transient scenario described in the next subsections is simulated as the
proof-of-principle large scale application using the coupled CFD—system-code ap-
proach. The reactor power profile calculation is incorporated as the user-defined heat
source term in the CFD portion of the coupled model.
8.3.1 Transient Scenario
The transient scenario starts from the PWR normal operation conditions. The coolant
supply temperature of a cold leg (at t = 0 sec) drops 20 ◦ C from the normal temper-
ature, while the rest of the cold legs still receive the coolant at normal temperatures.
The time dependent behavior of the coolant thermal mixing in the reactor core are
studied for this transient scenario, and the coolant mixing with cold front propagation
in the reactor vessel is simulated using the coupled model.
This abnormal operational transient is specifically triggered by an event that
leads to coolant temperature entering cold leg A1 in Figure 8.9 being lowered to
107
Figure 8.9: Reactor core with hot legs and cold legs(CFD part of the coupled simulation)
260.7 ◦ C, while coolant temperature through the other reactor vessel cold legs remains
at 280.7 ◦ C. Though the temperature drop magnitude used in this simulation may
be excessive, similar operational transient could result from the malfunction of the
secondary side feedwater heaters or fault start of high pressure safety injection system.
8.3.2 Initial Conditions
The steady state operating conditions of the PWR constitute the initial conditions
of the transient simulation. Thermal hydraulic initial condition are based on the
AP1000 design parameters in Table 8.1.
Figure 8.10 is the initial reactor core power density contour on the vertical
plane passing through the center lines of the hot legs. The highest power density is
located at the geometric center of the reactor core.
Pressure drop across the AP1000 reactor core is 430 kPa under normal op-
erational conditions. Figure 8.11 is the initial pressure contour on a vertical section
through the reactor core center. The core porous media resistance parameters are
calculated from the assembly level CFD results. The coolant pressure drop across the
reactor core is around 400 kPa which is consistent with the hydraulic specification in
Table 8.1 under steady-state operational conditions. The small discrepancy is because
the bundle grid and spacers are not explicitly modelled in the assembly level CFD
simulations.
108
Total coolant mass flow rate is 18,000 kg/s through the reactor core, which
is evenly distributed among the four cold legs of the reactor vessel. Figure 8.12
shows the velocity magnitude contour at normal operational conditions. The highest
velocity is observed in the hot legs.
Figure 8.13 shows the initial temperature contour on a central vertical section
through the reactor core. Temperature reaches the highest value before it exits the
reactor fuel bundles.
8.4 Results and Discussion
The coolant temperature drop scenario is simulated using the coupled CFD—system-
code. Note that the heat generation rate in this transient analysis is also time-
dependent. Using the discrete reactor kinetics model developed earlier, the heat
generation rate (or the power level) of each node is updated as the temperature in
that node is updated by CFD results. The only feedback effect taken into account is
that due to the changes in the moderator temperature. The coolant flow rate is kept
Figure 8.10: Initial power distribution in the reactor core
109
at the same level as that under normal operational conditions.
8.4.1 Results
Temperature and power density profile in the reactor core are obtained in CFD results.
They are presented in the following subsections.
Temperature Profile in Reactor Core
The extent to which the fluid coming down through the four cold legs mixes in the
lower plenum before it enters the core can be estimated by the temperature contour
shown at t = 10 s in Figure 8.14. This is soon after the cold water front has passed
through the reactor core. As can be seen by the thermal asymmetry, the coolant is
not well mixed in the downcomer and in the lower plenum. Temperature contour in
the active core region is also asymmetric. Though the results may not be surprising,
the tool developed here will help quantify the extent to which such asymmetries may
result, which may have consequences on thermal stresses in the core and may also
impact core neutronics through temperature coefficient of reactivity.
Figure 8.11: Initial pressure contour in the reactor core
110
Figure 8.15 and Figure 8.16 show temperature contours on vertical sections
aligned with cold legs A1/B2 and B1/A2, respectively. The temperature asymmetries
are obvious on these contours. Figure 8.17 shows a set of temperature contours on
several horizontal planes in the reactor vessel. The horizontal temperature contours in
Figure 8.17 are shown in Figures 8.18 ∼ 8.24 individually. Based on these temperature
contours, it is obvious that the coolant at lower temperature has not fully mixed with
the rest of the coolant at normal temperature.
Figure 8.12: Initial velocity magnitude contour in the reactor core
111
Power Density in Reactor Core
Figure 8.25 shows power density contours on several horizontal planes in the reactor
vessel. Figures 8.26 to 8.28 are the 2∼4 power density contours from top in Figure
8.25. Figure 8.28 shows that the highest power density shifts to the quadrant through
which the colder coolant is passing through the core. The negative temperature
coefficient of the reactivity leads to this shift.
Figure 8.27 shows that the peak power density shifts to the quadrant opposite
to the peak quadrant in Figures 8.26.
The temperature and power generation distributions suggest that the colder
coolant as it passes through the core leads to a higher power density in that quadrant,
and as a result the coolant temperature becomes even higher than the other three
quadrants by the mid (axial) plane. This then leads to an opposite effect in the upper
part of the core; that is, as the hotter coolant passes through the upper part of the
core, negative temperature coefficient of reactivity leads to a drop in power density
in that quadrant.
Figure 8.13: Initial temperature contour in the reactor core
112
This implies the instability which may be incurred by the introduction of low
temperature coolant to one of the reactor cold legs. Figure 8.29 shows this power
density profile characteristic on a vertical section through the reactor core.
113
8.4.2 Discussion
A PWR reactor core is highly heterogeneous with nuclear fuel, control rods, moder-
ator, and structural materials. The proposed reactor kinetic model is only approxi-
mately predicting the reactor power profile during the coolant thermal mixing. The
solid structures in the reactor core are not explicitly modelled in the porous media
model in the CFD portion. The heat conduction within fuel rods and the convective
heat transfer on the fuel rod surface are not included either. The heat generated
in the reactor core is assumed to be deposited into the local coolant instantly. Fis-
sion neutron diffusion is not taken into account in the power profile calculation. All
these simplifications introduce uncertainties in the reactor power profile calculation.
However, the propose of including reactor power profile calculation is to demonstrate
the potential of the coupled CFD—system-code simulation with a three-dimensional
reactor kinetic model. This goal is here demonstrated adequately. A more detailed
reactor kinetics model can be coupled via the UDFs with the CFD code following the
same approach as used here, if needed.
114
Figure 8.14: Temperature contour in the reactor core on the cross sectional plane aligned
with the hot legs, at t =10 sec
with the cold legs B1/A2, at t=10 sec
115
with the cold legs A1/B2 , at t=10 sec
Figure 8.17: Temperature contour in the reactor core on the horizontal cross sectional
planes, at t=10 sec
116
Figure 8.18: Temperature contours in the reactor core on the horizontal cross sectional
plane Z0, at t = 10 sec
117
planes Z3, at t = 10 sec
118
119
Figure 8.25: Power density contours in the reactor core on the several horizontal planes,
at t = 10 sec
120
Figure 8.26: Power density contours in the reactor core on the horizontal cross sectional
121
Figure 8.29: Reactor core power density contour on the section aligned with the cold legs
122
Chapter 9
Coupled Simulation of Gas Turbine
- Modular Helium Reactor
The flexibility of the coupled approach implemented in Chapter 6 allows it to be used
to investigate the thermal hydraulic challenges in other reactor designs. Thermal hy-
draulic designs of Gen IV nuclear reactors are diverse. A single universal simulation
tool for all designs is not currently feasible. The coupled CFD—system-code simula-
tion tool is here applied to simulate the primary coolant system in the Gas Turbine
- Modular Helium Reactor (GT-MHR).
9.1 Introduction
A brief review of the GT-MHR is given. It is largely based on the GT-MHR design
documents [General Atomics, 2002] and an introductive report [LaBar, 2002]. It is
included here because some of the information is relevant to the model developed
later in the following section.
123
9.1.1 Gas Turbine - Modular Helium Reactor
GT-MHR is a Gen IV very high temperature reactor (VHTR), under development
by a consortium of General Atomics and several other organizations. GT-MHR is a
helium-cooled, graphite-moderated reactor which offers passive safety, high thermal
efficiency, environmental advantages, and competitive electricity generation cost.
The GT-MHR design combines a modular helium-cooled reactor vessel with
a gas turbine cycle as shown in Figure 9.1. MHR has a prismatic core composed of
graphite modular blocks. Compared to the metallic structural material in a water
reactor core, graphite modular blocks in GT-MHR core may stand higher tempera-
ture. Helium, flowing as a coolant through the GT-MHR core, can be heated to very
high temperature, thus making it possible to generate hydrogen through the thermo-
chemical process. High operating temperature and advanced gas turbine technology
improves the thermal efficiency of the GT-MHR power plant to as high as 48 %.
The high thermal efficiency reduces the nuclear fuel consumption and the thermal
discharge to the environment.
From safety point of view, GT-MHR design significantly reduces the possibil-
ity of core melting which is a key requirement for all next generation nuclear reactors.
GT-MHR safety is achieved through a combination of inherent safety designs that
take advantage of the gas cooled thermal hydraulic characteristics, high density car-
bon coated particle fuel, helium coolant, and graphite reactor core. Due to the passive
safety design, GT-MHR is capable of remaining at safe temperatures after an accident
even without operator intervention for a certain period of time. Triso fuel particle
is a robust nuclear fuel which contains the fission products within the high density
carbon coating. The reactor vessel and power conversion systems are located in an
underground concrete containment. All of these innovative safety features provide
GT-MHR large safety margin.
In Figure 9.2 the GT-MHR reactor vessel of the prismatic graphite core is 8.4
124
Figure 9.1: Gas turbine modular helium reactor thermal hydraulic system [LaBar, 2002]
m in diameter and 31.2 m in height. GT-MHR supplies helium coolant from power
conversion unit to the reactor vessel. Helium is driven from the upper plenum into the
prismatic reactor core. Helium flows downward through the coolant channels in the
graphite core modules without significant cross flow, and mixes in the lower plenum.
The hot helium returns from the reactor vessel through the hot gas duct, as shown
in Figure 9.1, to the power conversion unit. GT-MHR plants can use air cooling and
soil conduction as ultimate heat sink, which allows more flexible plant siting options.
It reduces the heat discharged to the hydro surroundings.
Figure 9.3 is the top view of the GT-MHR prismatic core. The cylindrical
reactor core is composed of closely packed hexagonal graphite fuel blocks and reflector
elements, called block modules. The 600 MWt GT-MHR design has 10 axial levels
of graphite block modules. GT-MHR core can be divided into several regions. The
active core region has an annular shape (red region in Figure 9.3). It has 102 fuel
125
Figure 9.2: Gas turbine modular helium reactor vessel [LaBar, 2002]
Figure 9.3: GT-MHR graphite core section [LaBar, 2002]
126
modules at each level. The regions inside and outside of the active region are the
graphite reflectors. The inner and outer reflectors at each level are consisted of 61
and 156 columns respectively. Additional layers of graphite block modules above and
below the active core function as reflector.
Figure 9.4 is a diagram of a graphite module in the active core region. Coolant
channels and fuel channels penetrate the graphite block vertically. Each hexagonal
block module contains 108 coolant channels and 216 fuel channels.
Figure 9.4: GT-MHR core graphite module [General Atomics, 2002]
The whole GT-MHR prismatic core is supported by a matrix of graphite
cylindrical structure in the lower plenum. Hot gas duct connects the lower plenum of
GT-MHR vessel and the power conversion unit. The returning cold helium annulus
duct is concentric with hot helium circular duct. The helium coolant flows through the
annulus duct outside the hot gas duct. The flow pattern is asymmetric in GT-MHR
lower plenum.
The modular helium cooled reactor vessel and the power conversion unit are
located in an underground concrete silo which is 25.9 m in diameter and 42.7 m
deep. This silo serves as the containment structure. The GT-MHR reactor vessel is
127
surrounded by a reactor cavity cooling system which provides totally passive safety-
related decay heat removal by natural draft air circulation. The shutdown cooling
system located at the bottom of the reactor vessel provides forced helium circulation
for decay heat removal for refueling and maintenance activities.
9.1.2 Thermal Hydraulic Challenges
Several thermal hydraulic challenges could be identified in the GT-MHR design. Dur-
ing normal operation, significant local variations in power density may be observed
in radial direction due to the non-uniform arrangement of the reflectors, control rods,
burnable poison assemblies, and fuel loading. Temperature offset from power varia-
tions in the hot channels is amplified due to the helium thermal expansion. The peak
◦
coolant temperature can therefore be 200 higher than the average coolant tem-
perature. The flow velocities also display a significant variation in different coolant
channels.
Efficiency of a gas turbine is very sensitive to the thermal equilibrium and
the swirl in the incoming flow to the gas turbine. Unfortunately, the helium has more
than 200 ◦ C temperature variation when it leaves the reactor core. Mixing in the
lower plenum and in the hot gas duct is hence crucial since the flow should be fully
mixed before it reaches the gas turbine inlet nozzle.
Helium jets at different temperatures (exiting from different coolant channels)
mix in lower plenum. Adverse thermal stresses can be expected in these areas. The
effectiveness of flow mixing in the lower plenum and the quality of the hot helium
flowing to the power conversion unit need to be investigated. Three-dimensional
capability is required to evaluate the level of mixing in the lower plenum.
Post-accident residual decay heat in the GT-MHR is removed by dissipating
the heat by natural convection and thermal radiation from the reactor pressure vessel
outer wall to the passive reactor cavity cooling system. During the loss of coolant
128
accident (LOCA), up flow in the hot channels and down flow in the cooler channels is
supposed to build up an inner circulation within the core and will result in hot plumes
in the upper plenum. The hot and cold channel flow distributions and the upper
plenum mixing are uncertain. Low Reynolds number flow with turbulent, transitional,
and/or laminar flow, buoyancy effects, and gas property variations should be modeled
to investigate this kind of thermal hydraulic scenario in GT-MHR. The methodology
of the lumped-parameter coolant channels used in nuclear system has very limited
applicability in modelling this phenomena. The performance of the reactor cavity
cooling system should be studied using advanced tools such as CFD.
All of these challenges suggest that the nuclear thermal hydraulic simulations
be raised to a new level. Traditional nuclear system codes alone are not appropriate
tools for these thermal hydraulic challenges. CFD is an indispensable technology for
design calculation and safety analysis of GT-MHR.
However, the complete helium coolant system of GT-MHR is too complex
to be modeled by a CFD code alone. Moreover, helium flow through the graphite
module coolant channels is mostly one-dimensional channel flow, without significant
cross flow or mixing. The helium flow in GT-MHR core can therefore be adequately
modeled by nuclear system codes alone. Therefore, the lower plenum portion of the
integral helium coolant system can be simulated using CFD while the reactor core
portion of the system can be modeled using the nuclear system code. The coupled
CFD—system-code is an ideal approach for the analysis of integral GT-MHR coolant
system.
9.2 Coupled CFD—System-Code Model
GT-MHR steady state operational conditions are simulated using the coupled CFD—
system-code which was developed in Chapter 6 and verified in Chapter 7. In this
129
coupled-codes approach, a nuclear system code and a CFD code are used to model
different regions of the flow loop. The CFD code is used to model the regions where
3D effects are important, while the system code is used to model all those regions
where the flow is expected to remain largely 1D. The two code systems are used to
solve for the flow in their respective regions, while the boundary conditions at the
interface between regions modeled by the CFD code and system codes are transferred
back and forth between the two code systems.
First large scale application of the coupled code system for a PWR was re-
ported in Chapter 8. The flexibility of the coupling approach enables the code to
easily model and simulate the Gen IV nuclear reactor, GT-MHR. The CFD portion
and system portion of the coupled model are described in the following two subsec-
tions, respectively.
9.2.1 CFD Model of Lower Plenum
To investigate the flow mixing phenomena identified in GT-MHR lower plenum, a
three-dimensional CFD model was developed for the lower plenum and for the con-
necting duct. The CFD model is created and meshed using GAMBIT [FLUENT Inc.,
2005b]. The geometry and dimensions for the CFD model are from General Atomic
GT-MHR conceptual design description report submitted to the U.S. Nuclear Regu-
latory Commission (NRC) in August 2002 [General Atomics, 2002]. There are 736 k
cells. It is a hybrid mesh with prism and tetrahedral cells. Majority of the cells are
prism. A layer of tetrahedra cells are used along the transitional regions.
Figure 9.5 is the isometric view of the CFD model for GT-MHR lower plenum
and connecting duct. Since CFD model only retains flow domain, the graphite sup-
porting cylinders are not included. Seventy eight circles visible on the surface are the
edge view of the solid graphite cylindrical support structures in the lower plenum.
The flow is through the meshed region. Figures 9.6, 9.7 are top view and side view
130
of the CFD model, respectively.
Figure 9.5: Isometric view of the GT-MHR CFD mesh
Figure 9.6: Top view of the GT-MHR CFD mesh
The upper surface in the CFD model is the surface through which hot helium
flows is discharged from the core coolant channels. Helium jets of different temper-
atures mix in the lower plenum after leaving the prismatic reactor core. The hot
helium then turns to the hot gas duct and enters the power conversion unit. In the
CFD model of the mixing region, the upper surface is defined as the inlet velocity
131
Figure 9.7: Side view of the GT-MHR CFD mesh
boundary condition. The end of hot gas connecting duct is set as the outlet pressure
boundary condition. Adiabatic boundary boundary conditions are imposed on the
surface of the graphite cylindrical structures since no substantial heat source or heat
sink is located in the graphite supporting structure. Adiabatic surface boundary con-
dition is also applied to the side walls and bottom of the lower plenum since heat lost
to the environment is not significant at normal operational conditions.
Standard k − ϵ turbulence model is used in the CFD portion of the GT-MHR
model. It is realized that other more elaborate models available in the CFD code are
more appropriate for turbulent flow. However, standard k- model is still used here as
it is more computationally efficient. Ideal gas model is used to calculate the density
of the compressible helium coolant.
The CFD model of the GT-MHR lower plenum is then coupled with the
RELAP5-3D model for the reactor core and upper portion of the GT-MHR vessel.
RELAP5-3D model is described in the next subsection.
132
9.2.2 RELAP Portion in Coupled Model
RELAP5-3D model used in the coupled simulation was initially developed for the
complete GT-MHR thermal hydraulic systems by Idaho National Laboratory (INL)
[INEEL RELAP5-3D Group, 1999]. The geometry and the thermal hydraulic spec-
ifications used in the model are based on the 600 MWt GT-MHR. The original
RELAP5-3D model includes three coupled thermal hydraulic systems: the reactor
cooling system (helium); the part of the containment cooled by air; and the reactor
cavity cooling system (cooled by air). It was developed to investigate the thermal hy-
draulic behavior at the system level. The structural details inside the GT-MHR vessel
are not explicitly modeled. The one-dimensional flow path is a lumped-parameter flow
path based on the total flow area and the hydraulic diameter of the coolant channels.
Coupled CFD—system-code model only uses the RELAP5-3D nodalization
for the helium part of the reactor coolant system. The nodalization is shown in Figure
9.8. The reactor core nodalization consists of nine lumped-parameter coolant channels
and two bypass helium flow channels, which are grouped by the distance from the
vessel center. Time dependent volumes and junctions of RELAP5-3D components
provide the inlet and outlet conditions of the helium coolant system.
The RELAP5-3D nodalization shown in Figure 9.8 models the helium coolant
flow into the reactor vessel, upward to the upper plenum, down through the prismatic
graphite core, and out of the vessel. The inlet temperature to reactor core is 490 ◦ C.
The inlet flow rate is controlled to maintain the desired average outlet temperature at
1000 ◦ C. The inner and outer graphite core reflectors are modeled, as well as are the
reactor vessel and containment wall. The inlet flow is assumed to occupy the entire
region between the core barrel and the reactor vessel.
The flow through the active reactor core region is modeled as nine parallel
lumped-parameter channels in Figure 9.8. The active core region is divided into three
zones of annular shape. Four channels in the nodalization are used to model the inner
133
Figure 9.8: RELAP5-3D nodalization for the GT-MHR reactor vessel
fuel zone. Four channels are for the outer fuel zone. The RELAP5-3D nodalization
lumps the first two rows of the coolant flow paths from the reflectors as channel 151,
the next four rows as channel 152, the next four rows as channel 153, the remaining 11
rows as channel 154 in the inner fuel zone. Another four channels (171, 172, 173, and
174) use similar parametric configuration to model the outer fuel zone. The middle
fuel zone is sandwiched between the inner fuel zone and outer fuel zone. It is lumped
into channel 160 in the nodalization. Two core bypass channels (182, 185) represent
the helium flow paths through the reflectors. One represents the bypass coolant flow
paths through the inner reflector, and the other represents the flow paths through
the outer reflector. The nominal bypass flow rate is approximately 10% of the total
helium flow rate through reactor core.
The reactor vessel structure outside the core region is simplified in the RELAP5-
3D nodalization. The vessel below the core extends as a cylinder to half the depth
of the hemispherical lower head. The entire upper head hemisphere is modeled, as
is a hemisphere in the inlet plenum separating the upper plenum from the up-flow
annulus.
134
The coupled CFD—system-code simulation utilizes the reactor core and upper
plenum nodalization in the RELAP5-3D model and the CFD model of the lower
plenum to generate the complete GT-MHR vessel model. The flow path geometry
and heat source specification of the original RELAP5-3D model are preserved in the
coupled model.
9.2.3 Coupled CFDSystem-Code Model
The approach used in this GT-MHR coupled CFD—system-code simulation was in-
troduced in Chapter 6. Figure 9.9 is the diagram of the coupled-code model for the
reactor coolant system in the GT-MHR vessel. The lower plenum is modeled using
the CFD code, FLUENT, while the rest of the system is modeled using the system
code, RELAP5-3D. A set of User-Defined Functions (UDFs) are written to perform
the interface data exchange during the coupled simulation.
Figure 9.9: Coupled model for the GT-MHR
The lower plenum in the RELAP5-3D nodalization (volume 195 in Figure 9.8)
135
is replaced by the 3D CFD mode1 (in the coupled model). The interface between the
CFD portion and the RELAP5-3D portion is the upper surface of the lower plenum.
The 11 helium coolant channels from the RELAP5-3D portion discharge hot helium
jets through the upper surface into the lower plenum (the CFD portion). The inlet
boundary of the lower plenum CFD model is divided into 11 zones to receive the 11
helium jets from RELAP5-3D nodalization. There are a total of nine coolant channels
and two bypass channels in the RELAP5-3D portion which exchange the mass flow
rates and temperature data to the corresponding zones on the CFD inlet boundary.
Meanwhile, the 11 zones of the CFD inlet boundary feedback the pressure conditions
to RELAP5-3D portion. The cross-sectionally averaged inlet pressure of the eleven
boundary zones from the CFD results are used since the GT-MHR RELAP5-3D
model uses a single time dependent volume for the reactor core outlet boundary
condition with the assumption that the pressure at the reactor core exit is uniform.
The pressure outlet boundary at the end of the hot gas duct is imposed on the CFD
model as shown in Figure 9.9. The flow areas and diameters of the eleven channels in
the RELAP5-3D model are listed in Table 9.1. The dimensions used to divided the
inlet boundary of the CFD model are also provided in Table 9.1.
The turbulence intensity is not explicitly calculated in the system code. There-
fore, turbulence intensity and turbulence kinetic energy dissipation rate boundary
conditions required by the CFD k −ϵ turbulence model are not provided by RELAP5-
3D portion. CFD portion uses constant estimated values for the inlet boundary con-
ditions for turbulence intensity and dissipation rate.
The velocity and pressure data are exchanged at the interfaces between the
CFD region and the system code region at the end of each time step during the
coupled CFD—system-code execution.
136
Table 9.1: Coupling interface boundary specifications
RELAP5-3D region flow area outer rim radius
ID in GT-MHR (m2 ) (m)
182 inner reflector bypass 0.0667 1.580
151 inner fuel rows 1-2 0.0697 1.624
152 inner fuel rows 3-6 0.140 1.709
153 inner fuel rows 7-10 0.126 1.782
154 inner fuel rows 11-21 0.261 1.925
160 middle core ring 0.738 2.281
171 outer fuel rows 1-2 0.0930 2.322
172 outer fuel rows 3-6 0.190 2.404
173 outer fuel rows 7-10 0.174 2.476
174 outer fuel rows 11-21 0.266 2.583
185 outer reflector bypass 0.161 3.613
9.3 Steady State Simulation and Results
9.3.1 GT-MHR Steady-State Conditions
The coupled CFD—system-code is used to model and simulate GT-MHR normal
(steady-state) operational conditions. The thermal hydraulic design specifications of
the 600 MWt GT-MHR are listed in Table 9.2. The coupled model incorporates these
core thermal hydraulic inputs. Helium temperature and mass flow rate are imposed in
RELAP5-3D portion as inlet boundary conditions. The reactor vessel outlet pressure
is specified as 7.0238 MPa (7.0238 MPa = 7.07 MPa - 46.2 kPa) [LaBar, 2002] in the
CFD portion. Since the coupled CFD—system-code simulation is time-dependent,
hence the steady-state is determined by simply marching in time. CFD portion
137
Table 9.2: Steady-state thermal hydraulic conditions of GT-MHR [LaBar, 2002]
Helium Coolant System
GT-MHR Core Power 600 MWt
Reactor Inlet Temperature 491 ◦ C
Reactor Outlet Temperature 1000 ◦ C
Mass Flow Rate 226 kg/s
Bypass flow 10%
Reactor Inlet Pressure 7.07 MPa
Reactor Core Pressure Drop 46.2 kPa
and system portion exchange the conditions across the interface at the end of each
time step during the transient execution. Identical time step size are used in both
component codes during a coupled simulation. No inner iteration between the CFD
code and the system code is performed within a time step. When the variables in one
time step change by less than a pre-specified level, the solution is assumed to have
reached the steady-state. For the GT-MHR model, it takes approximately 6 seconds
of the problem marching time to reach the steady state. The results presented in the
next section are those at the end of 6 second of simulation time.
9.3.2 Results
The interface condition on the surface between the CFD and system code regions
are described in the next subsection. The results of the coupled simulations of GT-
MHR coolant system in the reactor vessel are presented and discussed in the following
subsections.
138
Coupling Interface Conditions
The flow velocities and temperatures at the interface for the CFD model are provided
by RELAP5-3D results. Figure 9.10 shows the temperature contour on the interface
between the core and the lower plenum (it is the upper surface of the lower plenum
CFD model). Inset on the lower right corner indicates the location of the contour in
the 3D CFD model. The high temperature region in the contour plot is ring-shaped
because the active core of GT-MHR is annular-shaped between the inner graphite
reflector and the outer graphite reflector. The helium flowing through the central
coolant channels of the annular core is heated up to more than 400 ◦ C above the
average coolant temperature, which is about 1000 ◦ C.
Figure 9.10: Temperature contour at the lower plenum’s upper surface in the CFD portion
of the model. Temperature is in Kelvin. Highest temperature is in the annular shaped region
right below the annular core.
Figure 9.11 shows the velocity contour on the upper surface of the CFD model
of the lower plenum. Figure in the inset on the lower right indicates the contour
location in the CFD model. The velocity magnitude on the interface is provided
by RELAP5-3D portion. The RELAP5-3D mass flow rates result is mapped to the
139
RELAP-CFD interface with mass conservation. Since, the flow areas in RELAP5-3D
and CFD are not identical, the mass conservation is maintained across the interface
by sacrificing the momentum conservation at the interface.
Velocity contour plot is similar to the temperature contour plot which has a
ring-shaped high temperature region. The high velocity region in Figure 9.11 is also
ring-shaped due to the acceleration of helium because of thermal expansion. The
inlet temperature is uniform for every coolant channel from the upper plenum of
the reactor vessel. Therefore, the coolant channels which have the highest coolant
temperature, have the highest coolant velocity.
Figure 9.11: Velocity contour plot on the coupling interface between the core and the
lower plenum. Highest velocity is approximately 24 m/s.
Mixing in the Lower Plenum
The coupled GT-MHR simulation is intented to evaluate the thermal mixing in the
lower plenum. The effectiveness of the mixing in the lower plenum has significant im-
pact on the thermal performance of downstream components. The connecting duct
between the lower plenum and the power conversion unit delivers the high tempera-
140
ture helium to drive the gas turbine. The temperature variation of the helium flow
through the reactor core is as high as 600 ◦ C (as shown in Figure 9.10). To minimize
the thermal stress and optimize the gas turbine thermal performance, the helium
temperature contour on the cross sectional plane at the end of the connecting duct
should be as uniform as possible.
The flow mixing in the lower plenum and in the hot gas duct helps in achieving
thermal uniformity in helium. However, the coolant flow swirl is created through the
intensive three-dimensional mixing in the lower plenum. Swirl enhances the thermal
mixing. On the other hand, because modern gas turbines have very strict requirement
on the level of swirl at the inlet, the swirl in helium flow must be also minimized
before the flow reaches the turbine inlet nozzle. A balance between the temperature
uniformity and flow swirling should be maintained.
CFD portion of the coupled simulation provides the three-dimensional solu-
tions for the flow field and the temperature field in the GT-MHR lower plenum during
normal operation. The 3D graphical representation of the CFD results may be used
by the nuclear engineers to understand the three-dimensional structure of the flow in
the GT-MHR lower plenum, and to help design better optimized structures to strike
the balance between thermal mixing and the level of swirl that can be introduced.
Figure 9.12 shows the temperature contour plot on a vertical cross sectional
plane. White space is the region occupied by the graphite supporting structure. The
temperature scale in Figure 9.12 is from 908 K (535 ◦ C) to 1660 K (1387 ◦ C). Figure
9.13 shows the velocity magnitude contour plot on the same vertical cross sectional
plane as in Figure 9.12. High velocity flow can be seen in regions below the annular
core region.
Figure 9.14 shows the temperature contour plot on the vertical plane (shown
in the inset at lower right) including a segment of the connecting duct(shown in the
insert on the bottom right corner). The contour plot shows that the temperature is
141
Figure 9.12: Temperature contour plot on a vertical cross sectional plane in the lower
plenum.
Figure 9.13: Velocity magnitude contour plot on a vertical cross sectional plane in the
lower plenum
not uniform after the gas enters the connecting duct. The maximum and minimum
temperatures at the end of the duct are 1410 K (1137 ◦ C) to 1210 K (937 ◦ C),
respectively. Figure 9.15 is the velocity contour plot on the same cross section of the
duct. The flow is not fully developed along the segment shown in Figure 9.15.
142
Figure 9.14: Temperature contour plot on a vertical sectional plane
Figure 9.15: Velocity magnitude contour on a vertical sectional plane
Figure 9.16 shows the temperature contour plot on the central horizontal
plane including a segment of the connecting duct. The contour plot shows that
the temperature variation is high in the lower plenum. Figure 9.17 is the velocity
magnitude contour plot on including when a segment of the duct is included in the
model and the boundary condition is specified at the end of this piece of duct. The
143
Figure 9.16: Temperature contour on a horizontal sectional plane
Figure 9.17: Velocity magnitude contour on a horizontal sectional plane
flow is not fully developed in the connecting duct as shown in Figure 9.17.
Figures 9.18, 9.19 depict the flow stream lines in the GT-MHR lower plenum.
The stream lines in Figure 9.18 are colored by temperature. It clearly indicates
that the temperature is far from uniform after the mixing in the lower plenum. The
temperature varies by as much as 200 ◦ C on the plane at the end of the duct, as
144
Figure 9.18: Streamlines colored by temperature in the GT-MHR lower plenum and the
connecting duct.
Figure 9.19: Streamlines colored by velocity magnitude in the GT-MHR lower plenum
and the connecting duct.
shown in the next subsection. The stream lines in Figure 9.19 are colored by velocity
magnitude. Swirling could be observed in the stream lines. Quantitative level of swirl
can be deduced from the streamlines and their color.
145
Flow Quality in Connecting Duct Flow
Path lines in the GT-MHR lower plenum displays swirling flow in the connecting duct
between the reactor vessel and the energy conversion unit. The swirling enhances the
thermal mixing of the coolant to achieve thermal equilibrium. However, it may have
adverse effect on the turbine performance in the power conversion unit.
Figure 9.20 shows the velocity contour plot on the plane at the end of the
connecting duct. The pattern of the velocity magnitude contour indicates that the
flow is not fully developed.
Figure 9.20: Velocity magnitude contour plot on the plane at the end of the duct in the
3D CFD model, where the flow then continues in the simpler RELAP model.
To quantitatively evaluate the flow swirl intensity in the connecting duct, a
dimensionless parameter, swirl number, S, is defined as

vtangential · vradial
S = 2

(9.1)
vaxial
Swirl number is a measure of the tangential velocity component perpendicular
to the axial velocity component (which is along the axis of the connecting duct). A
146
field function to calculate the velocity components (vtangential and vaxial ) and the local
swirl intensity is programmed into the CFD code(FLUENT).
Figure 9.21: Swirl intensity contours on planes in the connecting duct and outlet plane.
The five contour planes from left to right on the top are the planes from the lower plenum
to the connecting duct outlet plane, as shown in the 3D figure.
The contours of the swirl number on several cross sectional planes along the
connecting duct are shown in Figure 9.21. The planes in Figure 9.21 are numbered
1 through 5 from left to right, with plane 1 identified as the plane where the duct
intersects with the lower plenum. Plane 5 at the end of the duct is called the outlet
plane. The area averaged swirl number on a cross sectional plane is defined as
∫
A
SdA
Save = total (9.2)
Atotal
Average swirl numbers, Savg , on planes 1-5 in Figure 9.21 are calculated and
147
listed in Table 9.3. As expected, swirl intensity reduces along the connecting duct
from 0.0829 on plane 1 to about 0.0069 on the outlet (second column in Table 9.3).
Figure 9.22 shows the temperature contour on the outlet boundary of the
connecting duct. The temperature scale in Figure 9.22 is from 1210 K (937 ◦ C) to
1410 K(1137 ◦ C). Near 200 ◦ C temperature variations can be observed in the gas at
the connecting duct outlet. Figure 9.23 shows the temperature contour on the cross
sectional planes along the connecting duct.
Figure 9.22: Temperature contour on the connecting duct outlet plane
Temperature uniformity on the planes is measured quantitatively by area
averaged temperature variation, β, which is defined as
∫
Atotal
|T − Tavg |dA
β= (9.3)
Atotal
Averaged temperature variation is calculated on different cross sectional planes
along the duct. The values of β are listed in the third column of Table 9.3. The coolant
temperature variation shows gradual reduction of β along the connecting duct. But
even at the outlet of the duct, the hot helium flow still has a β of 56.1 ◦ C.
148
Figure 9.23: Temperature contours on internal cross sectional planes in the connecting
duct and the outlet plane. The five contour planes from left to right on the top are the
planes from vessel to the connecting duct outlet plane, as shown in the 3D figure.
Table 9.3: Average swirling number and temperature uniformity in the connecting duct
Helium Flow in the Connecting Duct
Plane Average Swirl Number Average Temperature Variation
Plane 1 0.0829 77.3◦ C
Plane 2 0.0358 68.7◦ C
Plane 3 0.0091 64.0◦ C
Plane 4 0.0075 59.4◦ C
Outlet Plane 5 0.0069 56.1◦ C
149
Based on the trend of average values along the flow direction, additional pipe
length will further reduce the swirl intensity and improve the thermal uniformity of
the flow. But extra pipe will also incur the additional pressure loss which is not
desirable.
9.4 Conclusions
Coupled CFD—system-code simulations of reactor coolant system in the GT-MHR
vessel have been carried out, thus demonstrating the potential of the coupled CFD—
system-code approach to Gen IV reactor design and optimization. The coupling
methodology, based on UDFs, is very flexible thus allowing for easy implementation
for new reactor designs. Valuable design and optimization related information can
be obtained by coupled simulations for the designers of next generation of nuclear
reactors. The results obtained and presented in this paper could be used to further
optimize the GT-MHR thermal hydraulic design. Simple extension of this specific
application will include the complete model of the helium loop using RELAP5-3D
and possible inclusion of neutronics feedback.
150
Chapter 10
Summary, Discussion and Future
Work
10.1 Summary
This thesis work elevates the thermal hydraulic simulation by expanding the porous
media modelling capability in the scope of a commercial CFD code, and developing
then applying an innovative coupled CFD—system-code simulation methodology to
more efficiently use the resource which are available to the general researchers. This
dissertation has five major components. First, a commercial CFD code is verified
for nuclear specific application. A new porous media model is then implemented in
this CFD code for improved modeling flows in nuclear systems. Third, a coupling
methodology is developed to link a CFD code and a system code to simulate large
scale nuclear systems with the more appropriate kind of simulation capability. This
coupled capability is first tested and then applied to a PWR as well as to a high
temperature gas cooled reactor.
Commercial CFD code FLUENT is first used to simulate the ISP No. 43,
Rapid Boron-Dilution experiment, to verify its capability to model nuclear systems.
The Australian Replacement Research Reactor is then simulated using a porous media
151
model in CFD for the reactor core. The parameters for the porous media model are
obtained through a series of assembly level CFD simulations.
A modified k − ϵ turbulence model for porous media is implemented in FLU-
ENT using UDFs. Transverse flow through porous media is simulated with the ex-
tended CFD code. The results are compared with experimental data reported in
literature for flow through vegetation.
A coupled CFD—system-code is developed by linking a commercial CFD code
(FLUENT) with a system code (RELAP5-3D). This allows parts of the system to be
modeled using the system code, and other parts that require 3D simulations to be
modeled using the CFD code. In the coupled code FLUENT is the master code, while
RELAP5-3D runs as slave under the control of FLUENT. UDF feature of FLUENT
is the programming interface to develop the coupled simulation capability. This
innovative approach is verified by comparing the results obtained from the coupled
simulations with the results from the CFD-only simulation and the system code-only
simulation.
The coupled CFD—system-code is applied to a simplified PWR NSSS to
demonstrate its large-scale application. Time-dependent three-dimensional reactor
power profile is calculated in a PWR transient scenario which investigates the spacial
impact of the coolant thermal mixing by using a discrete reactor kinetic model. This
model takes the thermal feedback into account. The coupling methodology, based on
UDFs, is very flexible thus allowing for easy implementation for new reactor designs.
A coupled simulation of reactor coolant system in the GT-MHR vessel has also been
carried out. The simulation showed the extent to which mixing occurs in the lower
plenum of the GT-MHR. Thus, The potential of the coupled CFD—system-code
approach to Gen IV reactor design and optimization has been demonstrated.
The time scale of the coupled CFD—system-code simulation ranges from a few
seconds to tens of seconds due to the time step limitation. The coupled simulations
152
could be used to improve the thermal hydraulic design but not to help the operator
for operational process transients, such as Loss of Coolant Accident (LOCA) or Loss
of Off-site Power (LOOP), which normally lasts a few hours to a few days. Therefore,
the coupled CFD—system-code capability developed in this thesis is a tool for the
nuclear system designer.
10.2 Discussion
The methodology followed in this work was motivated by the computational resources
available to average nuclear engineers. The coupling methodology encountered some
difficulties in the application of the coupled CFD—system-code. The limitations that
have not been overcome in this thesis are listed below:
The time step of the coupled CFD—system-code simulation is limited by the
Courant limit which is very small for the large scale safety analysis. The limitation
of the small time step needs to be relaxed if slow transients are to be simulated using
the coupled approach. The time step size limitation results from the independence
of the component softwares in the coupled code. The limitation on time step size
may be relaxed by using implicit code coupling scheme. However, if the two com-
ponent softwares are coupled using an implicit scheme using their source codes and
by coupling the guts of the two codes, then the coupling capability developed here
will not automatically be available to any future upgrades of these two component
codes. Each time a new version of these codes is released, the coupling will have to
be re-developed for the new version.
Second major limitation of this work is that it has not yet been tested for
multi-phase flow simulations. Multiphase CFD (MCFD) has potential application
in the nuclear industry. MCFD may extend the coupled simulation to the BWRs
which have two phase flow in the core under normal operational conditions. MCFD
153
may also be used to simulate the safety injection and core reflooding in PWR LOCA
scenario, during which the coolant front in the reactor coolant system are of interest
to the nuclear safety analysts. Nuclear system code, RELAP5, can calculate the
void fraction in the reactor coolant system. Unfortunately, multiphase model in
commercial CFD codes are only now becoming mature enough for these applications.
Finally, the large scale coupled CFD—system-code approach has not been
validated against any experimental data (only CFD code, FLUENT, is validated
through the comparison with the rapid boron-dilution experimental data). This is
due to lack of 3D data in large integral systems — the kind of systems for which the
coupled system code capability has been developed.
10.3 Future Work
To overcome the aforementioned limitations will be a major undertaking. There are
several potential future work in addition to these challenges. One of these pertains
to the user interface developed to couple the two codes. It could be made more user
friendly; and its functionality can be improved. For example, in the existing coupled
code, a case specific data-exchange function in is embedded in RELAP-3D source
code. The RELAP portion therefore requires recompilation for every new application
of the coupled code. Encapsulation of the RELAP-3D source code is desirable to
provide even more flexibility to the coupled CFD—system-code. Additional capabil-
ities now becoming available in the commercial CFD codes, such as the structural
and stress analysis and fluid-structure interaction capabilities, can also be fruitfully
exploited in the future.
154
Appendix A
User-Defined Functions for Porous
Media Model
A.1 Porous Media Model Expansion UDFs
DEF IN E SOU RCE specifies custom source terms for the different types of solved
transport equations in FLUENT. The expansion of the macroscopic turbulence model
for porous media in FLUENT is realized using this macro. The additional momentum,
turbulent kinetic energy, and turbulent kinetic energy dissipation rate source terms
are defined in UDFs below.
/∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗
UDFs t h a t add momentum , k , and e p s i l o n s o u r c e terms
∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗/
#i n c l u d e ” udf . h”
#d e f i n e Cd 1.13
#d e f i n e a 1.09
#d e f i n e Cfk 1 . 0
#d e f i n e C1 1.44
155
#d e f i n e C2 1.92
#d e f i n e mu 1 e−6
#d e f i n e VonK 0 . 4 1
#d e f i n e Cnu 0.09
#d e f i n e x30 0 . 0 1
#d e f i n e rho 998
r e a l Cfe=C2/C1∗ Cfk ;
DEFINE SOURCE( cell x momentum , c e l l , thread , dS , eqn ) {
real source ;
r e a l u=C U( c e l l , t h r e a d ) ;
r e a l f x 1 =0.50∗Cd∗a∗u∗ f a b s ( u ) ;
s o u r c e=−rho ∗ f x 1 ;
return source ;
DEFINE SOURCE( c e l l x k , c e l l , thread , dS , eqn ) {
r e a l c o e f f =1; // added t o match th e e x p e r i m e n t a l data
real source ;
s o u r c e=rho ∗ Cfk ∗ f x 1 ∗u∗ c o e f f ;
156
return source ;
DEFINE SOURCE( c e l l x e p s i l o n , c e l l , thread , dS , eqn ) {
r e a l c o e f f =1; // added t o match th e e x p e r i m e n t a l data
real source ;
r e a l k=C K( c e l l , t h r e a d ) ;
r e a l e p s i l o n =C D( c e l l , t h r e a d ) ;
s o u r c e=rho ∗ e p s i l o n /k∗C1∗ Cfe ∗ f x 1 ∗u∗ c o e f f ;
return source ;
157
Appendix B
User-Defined Functions for
Coupled CFD—System-Code
B.1 UDF to Initialize RELAP5-3D
DEFINE ON DEMAND( i n i t r e l a p ) {
/∗ I n i t i a l i z e RELAP5 ∗/
RELAP5 MAIN( i c t r l , varstr );
R53D DATA( i c t r l , varstr , voidf );
DEF IN E ON DEM AN D is a general-purpose macro that can be used to
specify a UDF that is executed as being required by FLUENT user, rather than having
FLUENT call it automatically during the calculation. DEF IN E ON DEM AN D
macro is used to define the UDF to initialize the RELAP5-3D portion in the coupled
CFD-system code. The functions defined in the macro will be executed immediately
after the DEF IN E ON DEM AN D macro is activated.
158
UDF name is the only argument to be supplied to DEF IN E ON DEM AN D.
The UDF name defined in the coupled CFD—system-code is init relap. After the
UDF defined using DEF IN E ON DEM AN D is compiled, the UDF name of the
argument supplied as the first macro argument will become visible and selectable in
the “Execute On Demand ” panel in FLUENT.
The UDF, init relap, executes manually right after the CFD portion model
and UDF being loaded into FLUENT. RELAP5-3D is compiled to a DLL which
available to FLUENT UDFs. RELAP 5 M AIN and R53D DAT A are the two func-
tions in RELAP5 DLL. After the initialization the RELAP5 portion of the coupled
simulation runs in parallel with the CFD portion.
B.2 UDF to Execute RELAP5-3D
DEFINE EXECUTE AT END( e x e c a t e n d ) {
Domain ∗domain ;
domain=Get Domain ( 1 ) ;
// D e f a u l t domain
tf inlet = Lookup Thread ( domain , z o n e I D i n l e t ) ;
// ID 7 f o r FLUENT i n l e t
t f o u t l e t = Lookup Thread ( domain , z o n e I D o u t l e t ) ;
// ID 8 f o r FLUENT o u t l e t
/∗ e x e c u t e r e l a p ∗/
i c t r l [0]=1;
TRAN( i c t r l , varstr );
159
R53D DATA( i c t r l , varstr , voidf );
Message ( ”RELAP i s ru nning ” ) ;
d t f l u e n t=d t f r o m r e l a p ;
DEF IN E EXECU T E AT EN D is a general-purpose macro that is exe-
cuted at the end of a time step in a transient run or at the end of iteration in a
steady-state run. Coupled CFD-system code use DEF IN E EXECU T E AT EN D
to define the UDF to execute RELAP5-3D portion. The average flow rate, pressure,
and temperature on coupling interface are also calculated at the end of each time step
in the CFD part in this UDF. This UDF is also used to exchange data between the
CFD portion and the RELAP portion in the coupled code. It is the most important
UDF in the coupled code.
DEF IN E EXECU T E AT EN D does not have to specify whether the execute-
at-end macro gets executed at the end of a time step or at the end of an iteration.
This is done automatically when the steady or unsteady time method is selected in
the FLUENT model.
Name is the only argument to be supplied to DEF IN E EXECU T E AT EN D.
In the coupled code, the UDF is named exec at end. FLUENT may include several
UDFs defined by DEF IN E EXECU T E AT EN D macro. Tasks could be sepa-
rated into different UDFs to maintain a good code structure. In the coupled code,
the UDF name is exec at end.
The domain pointer is not explicitly passed as an argument to DEF IN E ON DEM AN D.
Therefore, if UDF uses the domain variable in any on-demand function, it needs
to retrieve the domain pointer using the Get Domain provided by FLUENT. In
coupled CFD—system-code simulation, the pointer of CFD domain is obtained by
Get Domain(ID). And the pointer to a coupling boundary which is a thread of cells
160
is obtained by function Lookup T hread(domain, zoneID). The RELAP subroutines
T RAN and R53DD AT A are executed in this UDF.
After the UDF being defined using DEF IN E EXECU T E AT EN D is in-
terpreted or compiled, the name of the argument is supplied as the first DEFINE
macro argument becomes visible and selectable in the UDFs panel in FLUENT. Note
that user can activate multiple end-iteration functions.
B.3 UDF for Coupling Calculation and Communi-
cation
DEFINE PROFILE( s e t i n t , t f , i ) {
face t f ;
begin f loop ( f , tf )
F PROFILE( f , t f , i )= t f r o m r e l a p i n l e t ;
end f loop ( f , tf )
DEF IN E P ROF ILE can be used to define a custom boundary profile that
varies as a function of spatial coordinates or time. Some of the variables can be cus-
tomized at an appropriate boundary. The variables include velocity, pressure, temper-
ature, turbulence kinetic energy, turbulence dissipation rate, mass flux, mass source
term, heat source term, etc. There are three arguments for DEF IN E P ROF ILE:
name, t, and i. Users supply the name of the UDF. t and i are variables that are
passed by the FLUENT solver to the UDF. While DEF IN E P ROF ILE is usu-
ally used to specify a profile condition on a boundary face zone, it can also be used
161
to specify, or fix, flow variables that are held constant during the computation in
a cell zone. For these cases, the arguments of the macro will change accordingly.
The coupling boundaries are updated by the UDFs defined by this macro during the
coupled simulation. The example UDF is named set in t which updates the CFD
inlet temperature boundary condition by the corresponding coupling boundary from
RELAP5-3D result.
Note that unlike source term and property UDFs, profile UDFs (defined using
DEF IN E P ROF ILE) are not called by FLUENT from within a loop on threads in
the boundary zone. The solver passes only the pointer to the thread associated with
the boundary zone to the DEF IN E P ROF ILE macro. The specific UDF loops
over all of the faces in the thread.
B.4 UDF for Time Step Synchronizing
DEFINE DELTAT( mydeltat , domain ) {
return dt fluent ;
DEF IN E DELT AT is a general-purpose macro that can be used to control
the size of the time step during the solution of a time-dependent problem. Note that
this macro can be used only if the adaptive time-stepping method option has been
activated in the iterate panel in FLUENT.
The coupled CFD—system-code synchronizes the CFD portion and the sys-
tem portion by using the time step from RELAP5-3D. RELAP5-3D determines the
variable time step size by the Courant condition. The above UDF dictates FLU-
ENT portion to use the time step from RELAP5-3D at the beginning of a time step.
dt f luent is obtained in the UDF which is executed at the end of the last time step.
B.5 UDF for Reactor Power Profile Calculation
162
DEFINE SOURCE( s e t p o w e r , c , t , dS , eqn ) {
r e a l s o u r c e , rho ;
r e a l a , b , cc ;
r e a l omega1 , omega2 ;
rho=(C UDMI( c , t ,2) −1)/C UDMI( c , t , 2 ) ;
a=l p ∗( rho −1);
b=rho+l p ∗lamda ∗( rho −1)−be t a ;
c c=lamda∗ rho ;
omega1=(−b+s q r t ( b∗b−4∗a∗ c c ) ) / ( 2 ∗ a ) ;
omega2=(−b−s q r t ( b∗b−4∗a∗ c c ) ) / ( 2 ∗ a ) ;
s o u r c e=C UDMI( c , t , 1 ) ∗ exp ( omega1∗CURRENT TIMESTEP ) ;
C UDMI( c , t ,1)= s o u r c e ;
return source ;
// Message ( ” S et power d e n s i t y \%1 f w/m3 i n r e a c t o r c o r e
a t time $\%$ 1 f s e c . \n ” , s o u r c e , CURRENT TIME ) ;
DEF IN E SOU RCE specifies custom source terms for the different types of
transport equations solved in FLUENT. It is used to specify the nuclear reactor core
power profile during the PWR transient simulation. The UDF, named set power,
calculates the time-dependent reactor power density profile which is stored in a User-
Defined Memory (UDM).
Three UDM variables should be initialized for local multiplication factor k,
temperature field at last time step, and reactor power density respectively before
this UDF is loaded. After the UDF that is defined using DEF IN E SOU RCE is
163
compiled, the name of the argument supplied as the first DEFINE macro argument
becomes visible and selectable in the fluid zone boundary condition panel in FLUENT.
164
References
G. Alvarez, P. E. Bournet, and D. Flick. Two-dimensional ssimulation of turbulent

flow and transfer through stacked spheres. International Journal of Heat and Mass
Transfer, 46:2459–2469, 2003.
N. Anderson, Y. Hassan, and R. Schultz. Analysis of the hot gas flow in the outlet
plenum of the very high temperature reactor using coupled RELAP5-3D system
code and a CFD code. Nuclear Engineering and Design, 238:274–279, 2008.
ANSTO. Summary of the Preliminary Safety Analysis Report (PSAR) for the ANSTO
Replacement Research Reactor Facility. Australian Nuclear Science and Technology
Organization, 2001.
B. V. Antohe and J. L. Lage. A general two-equation macroscopic turbulence model

for incompressible flow in porous media. International Journal of Heat. Mass Trans-
fer, 40:3012–3024, July 1997.
D. L. Aumiller, E. T. Tomlinson, and R. C. Bauer. A coupled RELAP5-3D/CFD

methodology with a proof-of-principle calculation. Nuclear Engineering and Design,
205:83–90, March 2001.
S. Becker and E. Laurien. Three-dimensional numerical simulation of flow and heat

transport in high-temperature nuclear reactors. Nuclear Engineering and Design,
222:189–201, 2003.
D. Bertolotto, A. Manera, S. Frey, H. M. Hrasser, and R. Chawla. Single-phase mixing

studies by means of a directly coupled CFD/system-code tool. In Preceedings of
Physor 2008, Interlaken, Switzerland, September 2008. Am. Nucl. Soc.
M. Chandesris, G. Serre, and P. Sagaut. A macroscopic turbulence model for flow in

porous media suited for channel, pipe and rod bundle flows. International Journal
of Heat. Mass Transfer, 49:2739–2750, July 2006.
D. J. Diamond. Generation-IV Nuclear Energy Systems —How They Got Here and
Where They Are Going. US NRC, 2003.
C. G. du Toit, P. G. Rousseau, G. P. Greyvenstein, and W. A. Landman. A sys-

tems CFD model of a packed bed high temperature gas-cooled nuclear reactor.
International Journal of Thermal Sciences, 45:70–85, January 2006.
165
J. J. Duderstadt and L. J. Hamilton. Nuclear Reactor Analysis. John Wiley and
Sons, New York, 1976.
A. Dybbs and R. V. Edwards. A new look at porous meida fluid mechanism - darcy to
turbulent. Fundamentals of Transport Phenomena in Porous Media, pages 199–254,
1984.
FLUENT Inc. FLUENT 6.3 UDF Manual. FLUENT USA Inc, Lebanon, NH, 2005a.
FLUENT Inc. GAMBIT 2.3 USER Manual. FLUENT USA Inc, Lebanon, NH,
2005b.
A. Geist and et al.. PVM (Parallel Virtual Machine) User’s Guide and Reference
Manual. Oak Ridge National Laboratory, 1994.
General Atomics. Gas Turbine-Modular Helium Reactor (GT-MHR) Conceptual De-

sign Description Report. General Atomics, 2002.
D. Getachew, W. J. Minkkowycz, and J. L. Lage. A modified form of the k − ϵ model

for turbulent flows of an incompressible flid in porous media. International Journal
Heat Mass Transfer, 43:2909–2915, 2000.
K. D. Hamman and R. A. Berry. A CFD simulation process for fast reactor fuel
assemblies. Nuclear Engineering and Design, 240(9):2304–2312, 2010.
S. Hari, Y. A. Hassan, and A. R. McFarland. Computational fluid dynamics simu-

lation of a rectangular slit real impactor’s performance. Nuclear Engineering and
Design, 235:1015–1028, April 2005.
L. W. Hu and M. S. Kazimi. Les benchmark study of high cycle temperature fluctu-

ations caused by thermal striping in a mixing tee. International Journal of Heat
and Fluid Flow, 27:54–64, February 2006.
INEEL RELAP5-3D Group. RELAP5-3D Code Manuals, Vol. I, II, IV, and V. Idaho
National Engineering & Environmental Laboratory, 1999.
INEEL RELAP5/MOD3 Group. RELAP5/MOD3 Code Manuals. Idaho National

Engineering & Environmental Laboratory, 1990.
E. Krepper, E. F. Hichen, and H. Jaegers. Investigation of natural convection in large

pools. International Journal of Heat and Fluid Flow, 23:359–365, 2002.
T. S. Kwon, C. R. Choi, and C. H. Song. Three-dimensional analysis of flow character-

istics on the reactor vessel downcomer during the late reflood phase of a postulated
LBLOCA. Nuclear Engineering and Design, 226:255–265, 2003.
M. P. LaBar. The gas turbine-modular helium reactor: A promising option for near
term deployment. http://gt-mhr.ga.com/, April 2002.
166
J. R. Lamarsh and A. J. Baratta. Introduction to Nuclear Engineering. Prentice Hall,
New York, 2001.
F. Lopez and M. H. Garcia. Mean flow and turbulence structure of open-channel

flow through non-emergent vegetation. Journal of Hydraulic Engineering, 127(5):
392–402, March 2001.
I. F. Macdonald. Flow through porous media-ergun equation revisited. Industrial

Engineering and Chemical Fundamentals, 18:199–208, 1979.
G. H. Marcus. Considering the next generation of nuclear power plants. Progress in

Nuclear Energy, 37(1-4):5–10, 2000.
T. Masuoka and Y. Takasu. Turbulence model for flow through porous media. Inter-
national Journal of Heat Mass Transfer, 39:2803–2809, 1996.
R. D. Moser, J. Kim, and N. N. Mansour. Direct numerical simulation of turbulent

channel flow up to ReT = 590. Physics of Fluids, 11 (4):943–945, 1999.
A. Nakayama, F. Kuwahara, and T. Hayashi. A macroscopic turbulence model for

flow in a porous media. Journal of Fluids Engineering, 121:427–433, 1999.
NEA. International Standard Problem (ISP) No.43 Rapid Boron-Dilution Transient

Tests for Code Verification. Nuclear Energy Agency, 2001.
D. A. Nield and A. Bejan. Convection in Porous Media. Springer-Verlag, New York,

1992.
M. H. J. Pedras and M. J. S. de Lemos. On the definition of turbulent kinetic energy

for flow in porous media. International Communication of Heat Mass Transfer,
27(2):211–220, 2000.
F. Perdu and S. Vandroux. System/CFD coupling for reactor transient analysis - an

application to the gas fast reactor with CATHARE and TRIO U. In Preceedings
of ICAPP 08, Anaheim, CA, June 2008. Am. Nucl. Soc.
S. B. Pope. Turbulence Flow. Published by Press Syndicate of the Univerisity of

Cambridge, Cambridge, United Kingdom, 2000.
Rizwan-uddin and Y. Yan. Cluster and nuclear engineering. In Preceedings of ANS

meeting, Washington, DC, November 2004. Am. Nucl. Soc.
U. Rohde, T. Hohne, S. Kliem, B. Hemstrom, M. Scheuerer, T. Toppila, A. Ascodi,

I. Boros, I. Farkas, P. M. uhlbauer, Vyskocil L, J. Klepac, J. Remis, and T. Dury.
Fluid mixing and flow distribution in a primary circuit of a nuclear pressurized
water reactor - validation of CFD code. Nuclear Engineering and Design, 237(15-
17):1639–1655, 2007.
R. Schultz and W. Wieselquist. Validation & verification: FLUENT/RELAP5-3D

coupled code. In RELAP5 User’s Seminar, Sun Valley, ID, 2001.
167
T. L. Schulz. Westinghouse AP1000 advanced passive plant. Nuclear Engineering
and Design, 236:1547–1557, 2006.
V. Seker, J. W. Thomas, and T. J. Downar. Reactor physics simulation with cou-

pled Monte Carlo calculation and computational fluid dynamics. In Preceedings of
ICENES 07, Istanbul, Turkey, June 2007.
Z. Shang. CFD inverstigation of vertical rod bundles of supercritical water-cooled

nuclear reactor. Nuclear Engineering and Design, 239(11):2562–2572, 2009.
US Department of Energy. A Technology Roadmap for Generation IV Nuclear Energy

Systems. US Department of Energy, 2002.
W. L. Weaver. The Applicatoin Programming interface for the PVMEXEC Program

and Associated Code Coupling System. Idaho National Engineering & Environmen-
tal Laboratory, March 2005.
W. L. Weaver, E. T. Tomlinson, and D. L. Aumiller. A generic semi-implicit coupling

methodology for use in RELAP5-3D. Nuclear Engineering and Design, 211:13–26,
2002.
G. Yadigaroglu, M. Andreani, J. Dreier, and P. Coddington. Trends and needs in

experimentation and numerical simulation for lwr safety. Nuclear Engineering and
Design, 221:205–223, 2003.
Y. Yan. Thermal hydraulic simulation of ACP600 passive heat removal system by

using RELAP5. Master’s thesis, Tsinghua University, Beijing, China, 2001.
168

Yan Yizhou

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Yan Yizhou

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Yan Yizhou

Uploaded by

Copyright:

Available Formats

DEVELOPMENT OF A COUPLED CFD—SYSTEM-CODE CAPABILITY

(WITH A MODIFIED POROUS MEDIA MODEL) AND ITS APPLICATIONS

Submitted in partial fulﬁllment of the requirements

Professor Rizwan-uddin, Chair and Director of Research

Motivated by recent developments in the ﬁeld of Computational Fluid Dynamics

the nuclear reactor thermal hydraulics simulation capabilities to address problems

by ﬁrst implementing improved porous media models in a commercial CFD code

to be three-dimensional; whereas a system code is used to simulate regions where the

ﬂow is expected to be one-dimensional or to simulate components such as pumps, etc.

• Test a commercial CFD code, FLUENT, by solving a nuclear-speciﬁc bench-

using User-Deﬁned Functions (UDFs).

• Demonstrate the porous media simulating capability by a nuclear system using

Develop the coupled CFD—system-code approach. Verify the coupled code

using a simple ﬂow in a network of pipes.

• Demonstrate the potential of the coupled CFD—system-code for next genera-

tion reactors by applying it to a Gas Turbine - Modular Helium Reactor (GT-

Part 1 is accomplished by implementing a modiﬁed k − ϵ turbulence model

bility to model nuclear systems. Australian Replacement Research Reactor (RRR) is

assembly level CFD simulations.

RELAP5-3D is introduced in Part 3 as the nuclear system code for coupled

CFD—System-code development. UDF feature of FLUENT is used to develop the

Part 4 is the ﬁrst large scale application of the coupled CFD—system-code. A

simpliﬁed PWR NSSS is modeled by the coupled CFD—system-code approach devel-

oped in Part 3. Time-dependent three-dimensional reactor power proﬁle is calculated

thermal mixing by using a specially developed discrete reactor kinetic model.

CFD—system-code approach to Gen IV reactor design and optimization.

Thus, by implementing an improved porous media model in a CFD code, and

ulation capability has been developed to model three-dimensional eﬀects in complete

has been demonstrated by applications to a PWR and to a GT-MHR. This coupled

CFD—system-code capability will be useful in developing better optimized reactor

designs by reducing reliance on conservative models and simulations.

guidance and encouragement throughout my years of study at University of Illinois. I

Dr. Nahil Sobh for serving on my defense committee.

I wish to extend special recognition to ﬁnancial support provided by the In-

novations in Nuclear Education and Infrastructure (INIE) program. INIE program

was established in 2001 by the Department of Energy’s Oﬃce of Nuclear Energy,

Science, and Technology (http://www.ne.doe.gov), to strengthen the university nu-

clear engineering education programs through innovative use of university research

computation resource at NCSA.

agement during the hardest years.

Chapter 2 Computational Fluid Dynamics and Porous Media Model 8

Chapter 3 CFD Veriﬁcation Cases for Nuclear Applications . . . . . . . . . 20

Chapter 4 Turbulence Porous Media Models . . . . . . . . . . . . . . . . . . . . . . . . 43

Chapter 5 Nuclear System Code and Coupling Approach . . . . . . . . . . . 58

Chapter 6 CFD—System-Code Coupling using UDF . . . . . . . . . . . . . . . . 73

Chapter 7 Coupled CFD—System-Code Validation . . . . . . . . . . . . . . . . . 82

Chapter 8 Large Scale Simulation of Pressurized Water Reactor Using

Chapter 9 Coupled Simulation of Gas Turbine - Modular Helium Re-

Chapter 10 Summary, Discussion and Future Work . . . . . . . . . . . . . . . . . . 151

Appendix A User-Deﬁned Functions for Porous Media Model . . . . . . . 155

Appendix B User-Deﬁned Functions for Coupled CFD—System-Code158

4.1 Model constants for macroscopic turbulence porous media model . . . 51

8.1 Nominal operational conditions of the AP1000 at full power . . . . . 95

9.1 Coupling interface boundary speciﬁcations . . . . . . . . . . . . . . . 137

1.1 Commercial nuclear power reactor evolution and naming convention

4.1 Porous media REV for macroscopic average . . . . . . . . . . . . . . 46

5.1 Coupled simulation diagram for PVMEXEC . . . . . . . . . . . . . . 69

6.1 A conceptual coupled CFD—system-code model . . . . . . . . . . . 76

8.1 Westinghouse AP1000 nuclear steam supply system [Schulz, 2006] . . 90