You are on page 1of 138

Reactor Analysis ANL-90/45

Division
Reactor Analysis
Division
Reactor Analysis
Division
Reactor Analysis
Division
Reactor Analysis C0MMIX-1AR/P:
Division A Three-dimensional Transient
Reactor Analysis
Division Single-phase Computer Program
Reactor Analysis
Division for Thermal Hydraulic Analysis
Reactor Analysis of Single and Multicomponent Systems
Division
Reactor Analysis
Division Volume 1: Equations and Numerics
Reactor Analysis
Division
Reactor Analysis
Division by R. N. Blomquist, P. L. Garner,
Reactor Analysis and E. M. Gelbard
Division
Reactor Analysis
Division
Reactor Analysis
Division
Reactor Analysis
Division
Reactor Analysis rvofSiTT a*
Division ANl-W Technital Ubror/
Reactor Analysis
Division
Reactor Analysis
Division

Argonne National Laboratory. Argonne, Illinois 60439


operated by The University of Chicago
for the United States Department of Energy under Contract W-31 -109-Eng-38

Reactor Analysis
Division
Reactor Analysis
Division
Reactor Analysis
Division
Argonne National Laboratory, with facilities in the states of Illinois and Idaho, is
owned by the United States government, and operated by The University of Chicago
under the provisions of a contract with the Department of Energy.

DISCLAIMER-
This report vi^as prepared as an account of work sponsored by an agency of
the United States Govemment. Neither the United Stales Govemment nor
any agency thereof, nor any of their employees, makes any warranty, express
or implied, or assumes any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, product, or pro-
cess disclosed, or represents thai its use would not infringe privately owned
rights. Reference herein to any specific commercial product, process, or
service by trade name, trademark, manufacturer, or otherwise, does not
necessarily constitute or imply its endorsement, recommendation, or
favoring by the United States Government or any agency thereof. The views
and opinions of authors expressed herein do not necessarily state or reflect
those of the United States Govemment or any agency thereof.

Reproduced from the best available copy.

Available to DOE and DOE contractors from the


Office of Scientific and Technical Information
P.O. Box 62
Oak Ridge, TN 37831
Prices available from (615) 576-8401, FTS 626-8401

Available to the public from the


National Technical information Service
U.S. Department of Commerce
5285 Port Royal Road
Springfield, VA 22161
Distribution Category:
LMFBR--Safety (UC-542)

ANL-90/45

ARGONNE NATIONAL LABORATORY


9700 South Cass Avenue
Argonne. Illinois 60439

COMMIX-lAR/P: A THREE-DIMENSIONAL TRANSIENT


SINGLE-PHASE COMPUTER PROGRAM
FOR THERMAL HYDRAULIC ANALYSIS
OF SINGLE AND MULTICOMPONENT SYSTEMS

Volume 1; Equations and Numerics

raorwTT or
^ ftNL-W Technita! Librar/

R. N. Blomquist, P. L. Garner, and E. M. Gelbard

Reactor Analysis Division

July 1991
PREFACE

The documentation of the computer code COMMIX-lAR/P has been divided into
four volumes, each addressing different aspects of the documentation process and
written from different perspectives. A brief summary of the four volumes is as
follows;

Volume 1: "Equations and Numerics"


Presents the basic equations, assumptions, finite-difference
techniques, and solution methods.

Volume 2; "User's Guide"


A step-by-step description of how to use the code, including
an input guide and a sample problem.

Volume 3; "Programmer's Guide"


A description of each subroutine and variable, showing linkage
among these and their relation to the equations and variables
presented in Volume 1.

Volume 4; "Verification and Validation"


A collection of representative problems run on this computer
code and a comparison of the results obtained to those from
theory, experiments, and other computer codes.

The normal user will primarily use the information in Volumes 1, 2, and 4
and will be using a controlled configuration of the code. Volume 3 Is designed
as a consolidated source of reference material for those involved in the code
development effort.

The development of COMMIX-lAR/P is an ongoing task. Future changes to the


models or coding will be documented as addenda to, or revisions of, these
volumes.

lii
iv
TABLE OF CONTENTS

NOMENCLATURE "

ABSTRACT ^

EXECUTIVE SUMMARY ^

1. INTRODUCTION ^
3
1.1 Overview
1.2 Features ^
1.3 Organization of Report
2. GENERAL FORM OF CONSERVATION EQUATIONS '

3. CONTROL VOLUME ^^

3.1 Construction of a Computational Cell


3.2 Control Volume for Field Variables 1^
3.3 Control Volume for Flow Variables 1^

4. FINITE DIFFERENCE FORMULATION 1^

4.1 Unsteady Term ^'


4.1.1 Main Control Volume 1'
4.1.2 Momentum Control Volume 20
4.2 Convection Term ^^
4.2.1 Main Control Volume 21
4.2.2 Momentum Control Volume 23
4. 3 Diffusion Terra 25
4.3.1 Main Control Volume 25
4.3.2 Momentum Control Volume 26
4.4 Source Term 28
4.4.1 Main Control Volume 29
4.4.2 Momentum Control Volume 29
4.5 Complete Equation ^'^
4.5.1 Main Control Volume 30
4.5.2 Momentum Control Volume 32

5. PRESSURE EQUATION 34

6. FLUID-STRUCTURE THERMAL INTERACTION 38

6.1 Introduction 38
6.2 Geometrical Description 39
6.3 Conduction Heat Transfer 42
6.3.1 Governing Equation 42
6.3.2 Finite-Difference Formulation 43
6.3.3 Initial and Boundary Conditions 46
6.3.4 Solution '*1
6.3.5 Heat Transfer to Adjacent Fluid 48
TABLE OF CONTENTS (Cont'd)
. . • 50
6.4 Radiation Heat Transfer ^^
6.4.1 Governing Equations
6.4.2 Solution
... 56
7. TURBULENCE MODELING
56
7.1 Introduction ,,
7.2 Background on Turbulence Modeling
7.3 Two-Equation (k-c) Model
7.3.1 Transport Equations ^^
7.3.2 Boundary and Initial Conditions °2
7.4 One-Equation (k) Model
7.5 Zero-Equation Mixing-Length Model °°
7.6 Constant Turbulent Viscosity and Conductivity 67

8. PUMP MODELING ^*

8.1 Introduction ^^
8.2 Homologous Pump Model ^^
8.3 Specified Pump Speed Model ^1
8.4 Specified Pressure Increase Model 72
8.5 Interface with Rest of Code 72

9. SUPPLEMENTARY PHYSICAL MODELS 74

9.1 Fluid Properties 74


9.2 Structure Properties 75
9.3 Heat Transfer Coefficient Correlations 76
9.4 Fluid-Structure Momentum Interactions 77

10. BOUNDARY AND INITIAL CONDITIONS 79

10.1 Boundary Conditions 79


10.1.1 Velocity 80
10.1.1.1 Specified Constant Velocity 80
10.1.1.2 Specified Transient Velocity 80
10.1.1.3 Free Slip 81
10.1.1.4 Continuative Mass Flow 81
10.1.1.5 Continuative Mass Flux 82
10.1.1.6 Continuative Velocity 82
10.1.1.7 Uniform Continuative Velocity 83
10.1.1.8 Expansion Cell 83
10.1.2 Pressure 84
10.1.2.1 Specified Constant Pressure 85
10.1.2.2 Specified Transient Pressure 85
10.1.3 Energy 86
10.1.3.1 Specified Constant Temperature 86
10.1.3.2 Specified Transient Temperature 87
10.1.3.3 Recirculation Temperature 87
10.1.3.4 Specified Constant Heat Flux 87
10.1.3.5 Specified Transient Heat Flux 88
10.1.3.6 Adiabatic 88
10.1.3.7 Duct Wall 88

vl
TABLE OF CONTENTS (Cont'd)

10.2 Initial Conditions '°

11. CALCULATIONAL SEQUENCE AND METHODS 92

11.1 Overall Calculational Flow During a Run 92


11.2 Advancement of Variables Across a Time Step 92
11.3 Steady-State versus Transient Mode "
11.4 Time Step Size Selection ^0"
11.5 Inner Iteration Methods ^"2

ACKNOWLEDGEMENTS ^°'*

REFERENCES ^"^

APPENDICES ^°*

A. HEXAGONAL FUEL ASSEMBLY 1°8

A.l Hex-Geometry Option ^^^


A. 2 Wire Wrap Model 1°'
A. 2.1 Introduction ^"9
A. 2.2 Smeared Wire Option 109
A. 2.3 Cell Integrated Option HI
A. 2.3.1 Geometrical Effects HI
A.2.3.2 Wire Drag Effect HI
A. 3 Fuel Pin Resistance H5
A.4 Other Features 117
A.4.1 Heat Source 117
A.4.2 Pressure Boundary Conditions 118
A.4.3 Initialization 118

B. MASS REBALANCING 119

B.l Introduction 119


B.2 Description 119
B.3 Derivation of the Pressure Correction Equation 120
B.4 Application 122
LIST OF FIGURES

13
3.1 Construction of Cell Volumes j^^
3.2 Placement of Variables on Staggered Grid . . • • ' ' : '
3.3 Main Control Volume around Point 0 and Neighbor Numbering . . . . . 15
3.4 Momentum Control Volumes s u • j
4.1 Control Volumes Showing Variable Placement: (a) M a m and ^^
(b) x Momentum , \ \, • j
4.2 Control Volumes Showing Convective Fluxes: (a) M a m and
(b) X Momentum
4.3 Control Volumes Showing Diffusive Fluxes: (a) Main and
(b) x Momentum 25
6.1 Definition of a Cylindrical Thermal Structure
Element Relative to the Flow Domain 39
6.2 Outer and Inner Surfaces of Several Thermal Structure Elements
in the Fluid Grid Structure 40
6.3 Four Quarter-Cylindrical Thermal Structure Elements
each Interacting with One Fluid Cell 40
6.4 Multiple Thermal Structures Interacting with a Single Fluid Cell . . 41
6.5 Material Regions and Gaps for a Typical Thermal Structure Element . 42
6.6 Partitions in Cross Section of a Thermal Structure Element 43
6.7 Energy Balance on Partition { of a Thermal Structure Element . . . . 43
6.8 Radiant Energy Interchange on Surface i 51
10.1 Cells Near Outlet Boundaries 82
10.2 Recommended Surface Arrangements for Pressure Boundary Conditions . 85
10.3 Finite-Thickness Wall Boundary 89
A.l Quarter-Pin Partitioning of a 19-Pin Fuel Assembly
in a Hexagonal Duct ^^0
A.2 Full-Pin Partitioning of a 19-Pin Fuel Assembly
in a Hexagonal Duct ^IQ
A. 3 Typical Arrangement of Helical Wire Wrap 113
A.4 Cross Section of a Helical Wire Wrap around a Fuel Pin 113
B.l Coarse Mesh Rebalancing Regions 120

vlli
LIST OF TABLES

2.1 Source Terms in the Cartesian Coordinate System 10


2.2 Transformations from Cartesian Coordinates to
Cylindrical Coordinates 11
3.1 Notational Convention for Main Control Volumes 15
3.2 Notational Convention for i-Direction Momentum Control Volumes ... 17
3.3 Notational Convention for j-Direction Momentum Control Volumes ... 17
3.4 Notational Convention for k-Direction Momentum Control Volumes ... 17
4.1 Convective Fluxes for Main Control Volume 22
4.2 Convective Fluxes for x-Momentum Control Volume 24
4.3 Diffusion Strengths for Main Control Volume 26
4.4 Diffusion Strengths for x-Momentum Control Volume 28
4.5 Coefficients of General Finite-Difference Equation 31
5.1 Coefficients of Pressure Difference in Velocity Equations 35
5.2 Coefficients of Pressure Equation 36
7.1 Expansion of Summation Terms in Pj, (Eq. 7.6) and G^ (Eq. 7.7) . ... 60
8.1 Quadrant Assignments for X 69
11.1 Sequence of Calculations within a Run 92
11.2 Sequence of Calculations within Time Step n 94

NOMENCLATURE
NOMENCLATURE ( C o n t ' d )

Re : Reynolds number . .„xu.e P - - - ^ ^„,„si.y - ^^^


, 1 ^ o r d i n a t e In c y l i n d r i c a l system (radius) 7v ^.ional surface
r : 1 - d l r e c t i o n s p a t i a l coorom^. y
7x, Y y •^'' directi°"= ge in ( • ' " ^ ,.„drtcal system
: r a d i u s of f u e l pin
: s o u r c e term i n c o n s e„r v„<-ion
a t i o n e q u a t i o n f o r (p A(...)
-P^"^' .„ . and 9 directxons ^^..^sian system
^ f-prm in c o n s e r v a t i o n e q u a t i o n f o r (p .esh spacing - ^ ^^^ ^ ^^^,,.,ons for Car
: c o n s t a n t p a r t of s o u r c e term Ar, Afl
(Sc,^)o ^ t o unknown « i n c e l l f, of s o u r c e term .h spacing 1-"
: d e r i v a t i v e , with r e s p e c t to Ax, Ay
(Sp^)f in c o n s e r v a t i o n equation for V "" clng m ^ -^-^^^'"^
niesn ^v
Az . ^ step size quantity *
; temperature
T ovation equation for q
At 1 of conservacj-
time . residual ot ^^ _^^^ ^ ^^t^erwise 0)
t
(surface) heat transfer coefficient S»
, Kroneclcer
KronecKex '^^^'^\''- ' dissipation
Aissipatto" rate
rate
U
fr^r r) direction «ij . . r b u l,e n t^ .Iclnetic
^ n e t i c eenergy
ner. d . . ^ ^ ^ ^ ^ ^^ ^ ^ ^ ^ ^ ^ ^ , ^ . ....ace
velocity component m x (or ._ turbulent ..^^ation «hicH xs
u e
j,-.-octi°n 1 (general)
velocity in coordinate direct
Ui E (emittance m ^ -^eria (Sec. I D
j^v-ir surface
normal velocity at boundary „^ i n c o n v e r g e n c e c r g^steitv Cangle)
Un constant m c y l i n d r i c a l syst.
r,«>ral momentum equation
constant term m general m «n = ^ l a l c o o r d i n a t e m cy
/,- = Ax-Ay-Az or Ar-r-A9-Az) J.direction spatxal
1-r-if cell volume (i.e. 0
geometric cexj. ttiermal conductivity
V
V L (i.e. rv-'^^
• fluid volume X
V, , • i k-direction
ffusion term in the i-, J - . ^^ <> viscosity
V^: boundary vise
momentum equa A*
density U c t e d by a surface
n y (or e) direction
vfelocity comp< P

velocity compc
n z direction
P
r.rnS»c. "" s.e..»
pump
ing pump operating state
radian value i
coordinate in Cartesian system Tf, Ti,» 'm
i-direction sp-
• direction i (general) . „aii shear stress
wall shear s^^ — -
spatial coordinate m direc ion
^ . m a t e in Cartesian system ^. m fluid energy equation
J.direction spatial coordina . „ , „ , cylindrical viscous dissipation rate . ;. C or h ) i
. oatiat coordinate in Cartesian and cy
k-direction spatial c
systems
xiii
COMMIX-lAR/P: A THREE-DIMENSIONAL TRANSIENT
SINGLE-PHASE COMPUTER PROGRAM
FOR THERMAL HYDRAULIC ANALYSIS
OF SINGLE AND MULTICOMPONENT SYSTEMS

by

R. N. Blomquist, P. L. Garner, and E. M. Gelbard

ABSTRACT

The COMMIX-lAR/P computer code is designed for analyzing the


steady-state and transient aspects of single-phase fluid flow and heat
transfer in three spatial dimensions. This version is an extension of
the modeling in COMMIX-lA to include multiple fluids in physically
separate regions of the computational domain, modeling descriptions for
pumps, radiation heat transfer between surfaces of the solids which are
embedded in or surround the fluid, a k-£ model for fluid turbulence, and
improved numerical techniques. The porous-medium formulation in COMMIX
allows the code to be applied to a wide range of problems involving both
simple and complex geometrical arrangements. The basic equations,
underlying assumptions, and solution techniques are presented for the
entire computer code, covering both old and new features.

EXECUTIVE SUMMARY

The COMMIX (COMponent Mixing) computer codes are designed for analyzing
steady-state and transient aspects of fluid flow and heat transfer in three
spatial dimensions. Code versions in this series having a "1" in the suffix
(i.e. COMMIX-1, -lA, -IB, and -lAR/P) are for single-phase fluids having limited
compressibility.

The COMMIX-lAR/P version is an extension of the COMMIX-lA version to


include:

- allowance for multiple fluids in physically separate regions,

- radiation heat transfer between surfaces of solids embedded in or


surrounding the fluid,

- a k-c model for fluid turbulence, and

- improved numerical methods.

Numerous second-order changes have been implemented to increase the user's


flexibility in specifying boundary conditions and material properties and in
utilizing a wide range of modern mainframe computers.
The three-dimensional modeling in COMMIX is achieved using a porous-medium
formulation. This has been implemented using the concepts of volume porosity
(the fraction of a control volume occupied by fluid), directional surface
porosity (the fraction of a control volume surface through which fluid may flow),
distributed resistances to fluid flow, and distributed heat sources. Either
Cartesian or cylindrical coordinates may be used. The fluid conservation
equations for mass, momentum, and energy and the transport equations for the
turbulence parameters are put into finite-difference form by averaging over the
local control volumes in the system. This process allows simulations of simple
problem geometries (e.g. straight pipe) and complex problem geometries (e.g. fuel
bundle, heat exchanger, reactor vessel). Although primarily developed for
nuclear reactor applications, the models in COMMIX are sufficiently general to
be used for analyzing other engineering systems.

The material properties for the fluids and embedded solids, heat transfer
coefficients, and friction factors are assumed to have various basic forms for
which the user supplies coefficients. In some cases, the user may supply the
coding to override these basic forms. Detailed fluid property packages are
included for liquid sodium, liquid water (both normal and heavy) , and helium gas,
as well as a library of friction factors for specific applications.

The discretized conservation equations are solved as an initial-value


problem in time and a boundary-value problem in space. The user is provided with
a broad range of boundary condition choices, allowing the simulation of the usual
types of boundary conditions encountered in real systems. The boundary
conditions and embedded heat sources may be functions of time. The formulation
of the conservation equations is implicit in time.

This report concentrates on the presentation of the modeling features


(covering the basic equations and underlying assumptions) and the numerical
methods (covering spatial differencing, implicitness of temporal differencing
and Iterative solution methods). Separate reports show how to use the code'
describe the internals of the coding, and provide assessments of the modeling '
1. INTRODUCTION

1.1 Overview

The COMMIX (COMponent MIXing) computer codes are intended for steady-state
and transient analysis of fluid flow and heat transfer in three dimensions.
Although primarily developed for nuclear reactor applications, COMMIX can be used
to analyze the fluid behavior in other engineering systems.

The development of COMMIX began in 1976 and resulted in the release of the
COMMIX-1 computer code (Ref. 1). This version introduced the concepts of volume
porosity and directional surface porosity for modeling fluid flow in complicated
geometries containing distributed solids, defined the basic conservation
equations to be used for single-phase fluids, developed the finite-differencing
techniques to be used in Cartesian geometry, and established the overall
framework of the computer code. Special modeling techniques for hexagonal
bundles of wire-wrapped fuel pins were included. The formulations of the
conservation equations were explicit in time.

Two development paths were followed after COMMIX-1. One of these paths
led to the development of COMMIX-2 (Ref. 2), which extended the COMMIX-1 modeling
techniques to two-phase fluids. The other path continued development of the code
for single-phase fluids and resulted in code versions denoted COMMIX-lx.

The COMMIX-IA version (Refs. 3 and 4) extended COMMIX-1 to allow


cylindrical geometry, provide generalized models for thermal and momentum
interactions between the fluid and the submerged solids, add a basic fluid
turbulence model, provide a formulation which is implicit in time, and introduce
additional options for boundary conditions.

The COMMIX-IB version (Refs. 5 and 6) extended COMMIX-IA by adding a k-c


model for fluid turbulence and several skew-upwind spatial differencing schemes
to reduce numerical diffusion.

The recently documented COMMIX-IC version (Refs. 7 and 8) extended COMMIX-


IB by modifying the conservation equations to be applicable to subsonic
compressible flows, modifying both the k-c turbulence and skew-upwind
differencing schemes, and providing two new user-selectable matrix solution
methods.

The COMMIX-IAR/P version described in the present report is an extension


of COMMIX-IA to allow multiple fluids, include pumps, allow radiation heat
transfer between submerged solids, and provide more efficient numerical
treatments. COMMIX-lAR/P includes a modified version of the k-c turbulence model
from COMMIX-IB. Numerous second-order changes have increased the user's
flexibility in specifying boundary conditions, material properties, and utilizing
a wide range of modern mainframe computers. A complete description of the
features in COMMIX-lAR/P is given in Sec. 1.2.

The continued development of COMMIX over the years has had two major
purposes: (1) the addition of modeling to augment the computer code's range of
applicability to various systems, and (2) the improvement of the numerical
methods, particularly with respect to reducing the computer time required to
perform the simulation of a transient. The first of these two purposes has
generally been driven by the changing needs of the classes of problems for which
three-dimensional analyses are needed. The second purpose has followed naturally
as attempts are made to analyze increasingly more complex geometries in
sufficiently fine spatial detail. Consequently, COMMIX has become a rather
general purpose computer code with a wide range of applications.

There are several distinct purposes to be served by the documentation


package for a computer code; for COMMIX-lAR/P, these are addressed by a set of
reports consisting of four volumes. The present volume describes the code
capabilities, the equations being solved, and the solution methods employed. The
Users's Guide (Ref. 9) presents a step-by-step description of how to actually
set up, run, and interpret output for a problem. The Programmer's Guide
(Ref. 10) presents a detailed description of the internal coding to document the
purpose of each subroutine and variable, facilitating future modifications. The
fourth volume (Ref. 11) presents a collection of results computed using COMMIX
compared to experimental data, theory, or results computed using other computer
codes and forms a basis for validation and verification.

1.2 Features

The present version of COMMIX has evolved over a number of years. In


order to provide some historical perspective, the features of the present code
as presented in this section will be identified with a parenthetical indication
of the version in which that feature was added or significantly modified: [lA]
denotes COMMIX-lA, [IB] denotes COMMIX-IB, and [lAR/P] denotes the code version
described in this report. Absence of this parenthetical indicator means the
feature is part of the original COMMIX-1 code. (There are no references in this
section to COMMIX-IC since it was not a basis for the development of COMMIX-
lAR/P.) Note clearly that the purpose of this section is to give an overview of
the features of COMMIX-lAR/P; it is not intended to show all features of all
COMMIX versions; therefore, features which were in an earlier code version but
not in the current code version are not discussed.

COMMIX is designed, primarily, to solve the thermal hydraulics equations


governing the behavior of the fluid contained in a control volume system Both
steady-state and transient solutions may be calculated. The system volume may
be one two, or three dimensional. The base coordinate system is Cartesian
^!!:L!^L!'!!!r!.:^:_7_^'"'!'^.'."^ ^"•{•^K coordinates may also be used [lA] The
es
ry-
ne
mesh cells. Each mesh cell contains' a'VinVil" °". "^^"^ °': constant
however, several different fluid types may b e . r P . . ? ' t^"Sl^-Ph«" fl"id;
regions of the system volume belnsTdeled'^ [lAR/P] '" Physically separated

The fluid in the system control volume

The control volume may contain distributed soliH<= , •


quasi-continuum fluid domain. The physical =>uu.as, resulting in a
incorporated using a porous-medium approach Th^i"""'!f °^ these solids is
porosity concepts: the volume porositv whirh i^ ..^u ^ " ^ ^^ introducing two
y. nicn IS the fraction of the volume
occupied by fluid, and the directional surface porosities, which are the
fractions of the area open to flow in each of the three directions. The
distributed solids do not move or change size during a calculational history.
Except for the portions occupied by distributed solids, the system volume is
completely filled with fluid at all times.

The fluid behavior is governed by equations for the conservation of mass,


momentum (resolved into component form for each of the three coordinate
directions), and energy plus an equation of state. The fluids are viscous and
Newtonian and may contain volumetric heat sources; a limited treatment of fluid
compressibility is provided; however, COMMIX is not designed for analysis of
pressure-wave propagation. The conservation equations are formulated on a
porous-medium basis; when the volume and directional surface porosities are all
unity, the equations reduce to the more familiar equations for a continuum flow
field. A finite-difference approach is used in formulating the conservation
equations. There are four classes of terms in the conservation equations:
unsteady, convection, diffusion, and source. Full upwind (donor-cell)
differencing is used for convective terms in the equations. The viscosity and
thermal conductivity in the diffusion terms are assumed to be spatially
dependent. The formulation is implicit in time [lA] to avoid the time-step-size
restrictions imposed by methods which are explicit in time.

Various forms of boundary conditions may be specified for the conservation


equations on the boundary surfaces. If the boundary is open to flow, the
velocity may be specified directly (typical for an inlet) as a constant, a
function of time, or a function [lAR/P] of the temperature of a radiation heat
transfer surface; or may be specified indirectly (typical for an outlet) via
continuity of velocity, mass flux, or [lA] mass flow rate. An expansion-cell
velocity boundary condition is provided to facilitate modeling systems which are
intended to have constant fluid mass rather than constant fluid volume [lAR/P].
Both zero-slip (typical for a true wall) and free-slip (typical for a symmetry
boundary) velocity boundary conditions are available for impervious surfaces.
Boundary conditions for the energy equation may consist of a specified
temperature (constant, function of time, or as the current temperature [lAR/P]
of another boundary surface), a specified heat flux (adiabatic, constant, or a
function of time), or a duct wall [lA] model (which may include a heat source and
provides thermal coupling to an external sink temperature). The pressure may be
specified (constant or function of time) on boundaries having continuative-type
velocity boundary conditions [lA].

The presence of embedded solids is allowed to affect the momentum and


energy solutions in the fluids. A distributed resistance model is included for
momentum effects; the user [lA] supplies coefficients for a generalized
friction-factor correlation or selects one of the nongeneral forms from the
built-in friction-factor library. Thermal effects of embedded solids [lA] are
treated using the thermal structure model; these solids may be rectangular,
cylindrical, or spherical, may be composed of multiple layers of different
materials, and may contain internal heat generation; a one-dimensional conduction
heat transfer calculation is performed within each element of the solid with
boundary conditions which allow for energy exchange with the surrounding fluid.
The thermal coupling between fluids and solids may be treated using a
user-selected level of implicitness [lAR/P]. Surfaces of these solids may also
exchange energy with each other via thermal radiation [lAR/P].
The code has the capability to model one or more pumps within the system
volume [lAR/P], which result in a pressure gradient increase source term in the
momentum equation. The pressure gradient increase may be a specified function
of time, calculated from a speclfied-pump-speed correlation, or calculated using
the generalized homologous pump model equations.

A special option is provided in the code for modeling hexagonal bundles


of wire wrapped fuel pins arranged on a triangular lattice. With minimal user
input, the code will automatically generate the geometry (assigning boundary
surfaces, volume porosities, and directional surface porosities) and resistance
terms to account for the hexagonal bundle wall shape, the presence of the fuel
pins, and the presence of the spiralling wire spacers on the fuel pins.

Two options are available for the fluid equation of s'ate: detailed and
simplified. The detailed properties are provided by a group of FUNCTION
subprograms containing high-order fits to physical property data; packages are
available for liquid sodium, liquid normal water [lA], liquid heavy water
[lAR/P], and helium gas [lAR/P], The simplified properties take the form of
linear functions of temperature, to which the user inputs the coefficients [lA].
A suboption within the simplified properties allows the density to be calculated
from the ideal gas law rather than being a linear function of temperature
[lAR/P], In a multifluid problem, only one fluid may use detailed properties;
all other fluids must use simplified properties.

Physical properties for the solids used in thermal structures and duct
walls are second-order polynomials in temperature to which the user inputs the
coefficients for each solid [lA]. The user may, optionally, supply FUNCTION
subprograms to describe the thermal conductivity and specific heat capacity for
solids in order to circumvent the limitations imposed by the basic second-order
form [lAR/P].

Heat transfer between the surfaces of solids (duct walls and thermal
:) and the fluids involves
structures) involve.-! the
the use nf surface
i.oo of o,,,-<;.,„„ heat
i . .c_„
transfer
coefficients. A basic functional form has been assumed for these heat transfer
coefficients which involves the Reynolds [1A,1AR/P] and Prandtl [lAR/P] numbers
of the fluid and a set of user-supplied constants for each fluid/structure
interface The user may, optionally, supply a FUNCTION subprogram to describe
basic form [ m / p " " ' " correlations which are not adequately described by the

The aspects of turbulence are treated by using an effective diffusi


coefficient, which is the sum of laminar (molecular) and turbulent on

ence model.
The formulation in COMMIX uses the same <:ol,,n
in both steady-state and transient mode T t i l e l T ''""'"" ' ° ' calculations
order to advance the equations through a calcul^Mo i ^ ^ " """"' ^^ selected in
The user may specify the time-step size dUect^v v, ^''"''^ '" "^'^^^ " ° ^ " •
internally using a Courant-based [Ul or rsfo „f \.°^ ^^^ '^°'^^ compute it
"J rate-of-change-based [lAR/P] formula.
Within a time step, the fluid conservation equations are solved in blocks.
Within each block, the same type of equation is solved for all fluid cells
simultaneously [lA] to retain the implicitness of the finite-difference
formulation. Each type of equation for each cell has the same general form,
which usually has the quantities in a given cell related to the quantities in the
six nearest neighbor cells (i.e. one cell in the minus and plus direction for
each spatial coordinate direction). These sets of equations are assembled into
a matrix formulation (A)*(x)-(b), where (A) is a (sparse) coefficient matrix of
knowns, (b) is a known vector, and (x) is the vector of unknowns (one unknown for
each mesh cell). These matrix systems are solved using successive-overrelaxation
[1A,1AR/P1 or conjugate-gradient [lAR/P] techniques; both of these techniques are
iterative. The pressure equation, which results from a combination of the mass
and momentum conservation equations, should generally be solved using the
conjugate-gradient technique; if the successive-overrelaxation technique is used,
then the mass rebalancing scheme [lA) should generally be used to accelerate
convergence. The successive-overrelaxation technique is always used to solve the
equation systems which correspond to the conservation equations for energy,
turbulent kinetic energy, and turbulent kinetic energy dissipation rate.

The solution methods used for the temperature fields in thermal structures
are somewhat different. Within a thermal structure, the conduction equation
results in a matrix system for which the coefficient matrix (A) is tridiagonal;
this system is solved by a process of forward elimination followed by back
substitution, which is noniterative. The matrix system to be solved for the
radiation heat fluxes does not necessarily have a nice banded structure for the
coefficient matrix (A); the system is solved by direct inversion of the
coefficient matrix.

The equations are advanced across a time step by repeatedly sweeping


through the various blocks of equations until convergence is obtained. The user
has the flexibility to specify the degree of convergence desired for all
Iterative processes. Normally, all equations are solved in a fully coupled
manner at every time step; however, the user has the option to turn off and on
the mass/momentum and energy equations at selected time steps.

The International System of Units (SI) is used COMMIX (i.e. base units of
meter for length, kilogram for mass, and second for time) except that degree
Celsius (rather than kelvin) is used for temperature. The other units used in
COMMIX are derived from the base units and include newton for force, pascal for
pressure, joule for energy, and watt for power.

The coding in COMMIX is all in FORTRAN and conforms [lAR/P] to the


FORTRAN-77 standard (Ref. 12) with the exception of the use of the NAMELIST
construct for processing most card input. (The compilers on many computers
accept the NAMELIST extension to the standard.) The coding has been extensively
tested on IBM 370, 3033, and 3084 computers running MVS/JES3 and [lAR/P] on a
Cray X-MP running UNICOS; less extensive testing has been performed [lAR/P] using
a DEC VAX 8700 computer running VMS. The iterative solvers and other portions
of the code are written to take advantage of the vector processing hardware
available on Cray computers [lAR/P]. When used on a short-word machine (e.g. IBM
and DEC VAX), all floating-point quantities are declared DOUBLE PRECISION
[lAR/PJ. The code is intended for batch, rather than interactive, mode of
operation. The code reads text input from a card-image file and binary input
from a restart file. The code produces printed output and writes two binary
files: restart and plot; the restart file may be used to continue a calculation
in another computer run; either binary file may be used as input to one of
several postprocessor programs [1A,1AR/P] to produce graphical output of the
calculated results.

With the above features, COMMIX may be used for simulations of both simple
and very complex problems. Single component applications have included laminar
and turbulent flow through straight pipes, bundles of fuel pins, and coolant
mixing in plena of nuclear reactor vessels. A typical multicomponent application
is the use of COMMIX to model the complete primary system of an advanced
liquid-metal cooled reactor design, including the fuel assemblies, reactor vessel
plena, intermediate heat exchanger, primary pumps, and passive cooling of the
reactor vessel exterior by natural circulation of air.

1.3 Organization of Aeport

This volume presents the models used in the COMMIX-lAR/P computer code,
covering the basic equations, underlying assumptions, and solution techniques.

Sections 2 through 5 present the majority of the basic fluid modeling.


The conservation equations for mass, momentum, and energy are presented in Sec.
2 in porous-medium form; the equations are cast in a general form which
facilitates the further development of the presentation and the actual coding.
The control volume concepts used in COMMIX are presented in Sec. 3, including the
definition of mesh cells, the placement of variables, and the notation used to
refer to mesh cells on a relative basis. Section 4 presents the
finite-difference forms of the conservation equations, proceeding on a
term-by-term basis. The way in which the finite-difference forms of the momentum
equations are substituted into the continuity equation to form the equation
system to be solved for the pressure field is described in Sec. 5. The various
types of boundary conditions used for the fluid equations are described in
Sec. 10.

Sections 5 through 9 present additional modeling which supplements the


basic fluid modeling. The thermal interactions between the fluid and submerged
solids are described in Section 6, along with the modeling of radiation heat
transfer between the submerged solids. Section 7 presents the several models for
fluid turbulence, which range from a two-equation k-c model down to a model
having constant turbulence quantities. The several models available to describe
pumps are presented in Sec. 8. Section 9 presents several supplementary physical
models, including the fluid equation of state, properties for solids, heat

interfctlont h't''"'
interactions between 'the
r ^ Afluid
' ^ T ' and
'/"' ' '""^ ^'"^"^
submerged resistance model for momentum
solids.

The overall calculational flow in the computer code is presented in


ec. 11, including discussions of the order in which the equations are solved
the concepts of inner and outer iterations, iterative solution methods and
time-step size selection. ^j-un mci-nous, ana

The special geometry and distributed resi <;tar„-o f» •


hexagonal bundle containing wire wrapped fuel p ^ n s " / "'>.". ""'* '° "'°''"^ "
Appendix B presents the ™ass%ebalanci'ng technique used to ace 1 T ^'""''"'''' ^•
of the pressure equations when solved using the successive ^ " ^ ^ ^ " ^ ^ convergence
iiig Lne successive-overrelaxation method.
2. GENERAL FORM OF CONSERVATION EQUATIONS

The conservation equations for mass, momentum, and energy of the fluids
in COMMIX possess a common form (Ref. 3 ) , which, in Cartesian coordinates, is
given by

* V S (2.1)

where * stands for the "variable of conservation"; u, v, and w are the velocity
components in the x-, y-, and z-directions; >„ is the volume porosity (i.e., the
fraction of the physical volume occupied by fluid); 7,, >,, and 7^ are the
directional surface porosities (i.e., the fraction of the areas through which
fluid may flow); p is the fluid density; t is time; r^ is the diffusion
transport coefficient; and S,^ is the source term on a volumetric basis.

The variable of conservation is different for each equation: 1 (a


constant) for the mass conservation equation; velocity components u, v, and w for
the momentum equation in the x-, y-, and z-directions, respectively; and the
fluid enthalpy, h, for the energy equation. Similarly, the diffusion transport
coefficient is different for each type of conservation equation: 0 for the mass
conservation equation; viscosity, H , for the momentum equations; and thermal
conductivity, X, divided by specific heat capacity, Cp, for the energy equation.

The source term also depends on the specific equation; these are shown
symbolically in Table 2.1 for each equation. The mass conservation equation has
no source term. The gravity and pressure-gradient terms in the momentum equation
are straightforward as presented in Table 2.1; the viscous diffusion term is only
nonzero at a boundary surface when one of the nonconstant turbulence models is
being used and is described in Sec. 7.3.2; the distributed resistance term is
described in Sec. 9.4; the pump pressure terra is described in Sec. 8.5. The
internal heat generation in the energy equation is supplied by the user; the
thermal interaction with submerged solids is described in Sec. 6.3.5; the
substantial derivative of pressure and the dissipation term are currently
Ignored.

The same general form of Eq. 2.1 may also be used in a cylindrical
coordinate system, using the transformations in Table 2.2. Generally, x, y, and
z are replaced with r, 9, and z, respectively. Changes in the second spatial
coordinate. Ay, are replaced with rA9, to retain the dimensions of length;
similarly, the partial derivative with respect to y, a()/3y, is replaced with
1/r d()/d6. The additional convection and diffusion terms which arise in the
momentum equations (due to the unit vectors in polar coordinates being a function
of position) are moved to the source term in order to maintain the same base form
for Cartesian and cylindrical coordinates. The components of the gravitational
acceleration vector, g, are a function of angular position and are calculated
internally by the code from the user-input Cartesian components.
Table 2.1 Source Terms in the Cartesian Coordinate System

Diffusion Source
Variable Coefficient Term
Equation Direction

Continuitv Scalar 1 0

Momentum

(i) X u V^ P,.-v.-...(A£)^_-(i)
(ii) y V (t

(ill) z w v^ PSz - V,

Energy Scalar h X/Cp * q p<t

Components of gravitational acceleration (body


force terms).
V ,V ,V Viscous diffusion terms at boundaries.
X' y' z
Distributed resistances due to submerged
solids.

(AP/Dp.x- (AP/L)p,y, (AP/L), Pressure increase supplied by pump.


P Pressure.

Rate of heat liberated from submerged solids


per unit fluid volume.

q' Rate of internal heat generation per unit fluid


volume.
Rate of dissipation of mechanical energy, per
unit mass of fluid, due to viscous effects.
11

Table 2.2 Transformations from Cartesian Coordinates to Cylindrical


Coordinates

Cartesian Cylindrical

General:
(x,y, z) (r, d, z)
A A a A lA A
dx ' dy ' dz
dl ' r dd' dz
Ax,Ay, Az
Ar,rA0,Az
U=Ux
u=u,
V=Uy
V=U9

w=u.
Source Terms in Momentum
Equations:
g^ = C O S (0) g^j + s i n (0)gy (*)
gx
gg= - sin(6)g^ + cos(0)gy (*)
Sy
g^ gz

Vx. V . V Vr . Ve , V,
' "y ' z
R,,. Ry ' Rz Rj , Rj , Rj

S = (Table 2.1 with above substitutions)


s„ = (Table 2.1

S = (Table 2.1) S^, = (Table 2.1 with above substitutions)

S„ = (Table 2.1) S„ = (Table 2.1)

*The components of the gravitational acceleration vector are always defined


in terms of the components in a Cartesian coordinate system.
12

In the forms stated, the general conservation equations describe the time-
dependent behavior of a viscous, Newtonian fluid of limited compressibility
having spatial distributions of its properties in three dimensions, allowing for
thermal and momentum interactions with embedded solids. The conservation
equations given by Eq. 2.1 reduce to the more familiar conservation equations for
a continuum domain when the volume and directional surface porosities are set to
1 and there are no distributed resistances or heat sources, viscous boundary
diffusion terms, or pumps.

The common form of the conservation equations for all fluid variables
simplifies the finite-differencing and other aspects of the solution process.

The set of equations given by Eq. 2.1 and Table 2.1 constitute a set of
5 coupled equations (mass, energy, and three momenta) in 6 unknowns (u, v, w, p,
P, and h ) . The sixth equation needed to complete the set is the equation of
state, having the form

p = f (h,P) . (2.2)

This equation is not directly involved in the finite-differencing process --


there are no convection, diffusion, or source terms, and the equation is applied
on a pointwlse basis. The various forms allowed for the fluid equation of state
are discussed in Sec. 9.1.

For turbulent flow, all quantities in Eq. 2.1 are considered to be


time-averaged values. The diffusion transport coefficient is then an effective
(laminar plus turbulent) diffusion transport coefficient,

r* = W t Mr*)t • (2.3)

The laminar part of this diffusion coefficient comes from the equation of state
information. The turbulent part of this diffusion coefficient is a function of
the fluid state and the velocity field; it may be supplied by the user as input
or calculated as a function of space and time using one of the turbulence models
in the code (cf. Sec. 7 ) .

The user has the option of directing COMMIX to solve additional


conservation equations for the turbulent kinetic energy, k, and turbulent kinetic
energy dissipation rate, c. These equations have the same form as given by
Eq. 2.1. The diffusion transport coefficients and source terms for these
equations and the way in which the solution of these two equations affects
Eq. 2.3 are presented in Sec. 7.
13

CONTROL VOLUME

3.1 Construction of a Computational Cell

The computational cells and grid points may be defined in a number of


different ways in a finite-difference formulation. Figure 3.1 illustrates the
grid definition used for fluids in the COMMIX codes (Refs. 1, 3, and 5 ) . A
Cartesian coordinate system is used for illustrations in this section, but the
same principles apply when using cylindrical coordinates. The computational cell
is defined by the locations of the cell volume faces. The locations of the cell
faces are determined by the user-specified spacings Ax^ (or Ar^), Ayj (or Aftj),
and Az^. These spacings do not need to be uniform, either within a given
direction or from one direction to another. The faces of the computational cells
do not move during a calculational history; therefore, this is an Eulerian or
constant-volume approach. A grid point is placed in the geometric center of each
computational cell volume, which is coordinate location (-n^.y^.z^,) (or (T^.B^.Z^))
where

JAz^^ /2 for k=l (3.1)


^•^ ^ Fk-i * (^z^., + Az^)/2 for k = 2,..,KMAX

and Xi (or r j and yj (or 6^) are defined in the same manner.

4y,

-I A typical cell volume

Fig. 3.1. Construction of Cell Volumes

Most variables (e.g. pressure, enthalpy, density, turbulent kinetic


energy turbulent kinetic energy dissipation rate, etc.) are defined at the grid
points' these are identified as "field" variables. In contrast, the three
velocity components are defined on the different cell faces; these are identified
as "flow" variables. In a similar manner, the volume porosity is associated with
14

a grid point and the three directional surface porosities are associated with the
different cell faces. This placement of variables is illustrated in two
dimensions in Fig. 3.2. The x- (or r-) direction component of velocity (u) is
defined at the center of the cell face which separates grid point location
(i,j,k) from grid point location (1+1,j,k); the y- (or 6-) direction component
of velocity (v) is defined at the center of the cell face which separates grid
point location (i,j,k) from grid point location (i,j+l,k); the z-direction
component of velocity (w) is defined at the center of the cell face which
separates grid point location (i,j,k) from grid point location (i,j,k-t-l).

f
—»- • — —— • — - • - •
1 — i j + 1

f i
1
1
1 1
—^ • —-^ • —-^ • — .
(
y 1
ik —.- • — —^ • -.- • - j - 1
(
1
t
1 - 1
1
*• X
i
1 i + 1

u
V
• Other variables
Fig. 3.2. Placement of Variables on Staggered Grid

The use of velocity component definition points which are displaced or


staggered from the definition point used for field variables is implemented by
using a control volume for each velocity component which differs from the control
volume used for the field variables. These two types of control volumes are
described in the next two subsections. The grid containing all fluid cells must
be completely enclosed by a set of boundary surfaces, which are described in Sec

3.2 Control Volume for Field Variables

The control volume for a field variable is shown in Fig. 3.3 and the
notational convention used to refer to the surrounding control volumes and cell
faces IS given in Table 3.1. These control volumes are denoted the "main control
volumes" in later sections of this report. The control volume is constructed
around grid point 0 at location indices (i,j,k). The faces of this control
15

volume are planes of separation between this control volume and the six nearest
neighbor control volumes. The cell volumes to the west and east contain,
respectively, grid points 1 and 2 at location indices (1-1,j,k) and (i+l,j,k);
the cell volumes to the south and north contain, respectively, grid points 3 and
4 at location indices (i,j-l,k) and (i,j+l,k); the cell volumes below and above
(not shown in Fig. 3.3) contain, respectively, grid points 5 and 6 at cell
location Indices (i,j,k-l) and (i,j,k+l).

i-1 i +1

Fig. 3.3. Main Control Volume Around Point 0 and Neighbor Numbering

Table 3.1 Notational Convention for Main Control Volumes

Digit Cell Center Cell Face Center


Direction Subscript Location Indices Location Indices

center 0 1, i. k (not applicable)


west 1 i-1, i. k i-1/2 j. k.
east 2 i+1, i. k 1+1/2 j. k
south 3 i, i-1. k 1, j-1/2, k
north 4 i, ,i+i. k 1, j+1/2, k
bottom 5 1, 1. k-1 1, j. k-1/2
top 6 i. j. k+1 1, j, k+1/2
16

Having velocity components defined at cell faces is particularly


convenient when the conservation equations for field variables are integrated
over the main control volume (cf. Sec. 4 ) .

If solids are immersed in a fluid cell, the user must input the volume
porosity (0 < 7^ < 1) to indicate what fraction of the cell volume is occupied
by fluid; the default assumption is that the cell volume is completely filled
with fluid (,y„ - 1 ) . Similarly, if the presence of solids obstructs the fluid
flow area of a cell face, the user must input the directional surface porosities
(0 < Ti < 1, for i- X (or r ) , y (or $), z) to indicate what fractions of the cell
face areas are open for fluid flow; the default assumption is that the cell face
areas are completely open to fluid flow (7^ - 1, for i- x (or r) , y (or 8), z ) .
All of these porosities are allowed to be spatially dependent. The thermal and
momentum aspects of injnersed solids are discussed in Sec. 6 and Sec. 9.4,
respectively.

3.3 Control Volume for Flow Variables

The control volume for each velocity component is displaced from the main
control volume in the velocity component direction, extending from grid point to
grid point and encompassing the cell face upon which the velocity component is
defined. The other four faces are coincident with the faces of the main control
volumes; however, each of these four faces is composed of half of the face of two
different main control volumes. This is illustrated for i- and j-direction
control volumes in Fig. 3.4. These control volumes are denoted the "momentum
control volumes" in later sections of this report. Each momentum control volume,
thus, consists of half of two adjacent main control volumes. Grid points 0
through 6 have the same relative meanings (i.e., center, west, east, south,
north, bottom, and top) as for the main control volume; however, the specific
spatial indices are different and are defined in Tables 3.2 through 3.4.

y moinentuni
Control Volune"

X momentum
Control Volume

Fig. 3.4. Momentum Control Volumes


17

3.2 Notational Convention for i-Dlrectlon Momentum Control


Volumes

Digit Cell Center Cell Face Center


Direction Subscript Location Indices Location Indices

center 0 1+1/2, j, k (not applicable)


west 1 1-1/2, j, k i, j. k
east 2 1+3/2, j, k 1+1, j, k
south 3 i+1/2, j-l, k i+1/2, j-1/2, k
north 4 1+1/2, j+1, k i+1/2, j+1/2, k
bottom 5 1+1/2, j, k-1 i+1/2, j , k-1/2
top 6 1+1/2, j , k+1 1+1/2, j , k+1/2

Table 3.3 Notational Convention for j-Direction Momentum


Control Volumes

D igit Cell Center Cell Face Center


Direction Sub script Location Indices Location Indices

center 0 1, j+1/2, k (not applicable)


west 1 i-1, j+1/2, k 1- 1/2 j+1/2. k
east 2 1+1, j+1/2, k i-t1/2 j+1/2. k
south 3 i, j-1/2, k 1 j . k
north 4 i, j+3/2, k 1 j+1. k
bottom 5 i, j+1/2, k-1 i j+1/2. k-1/2
top 6 i, j+1/2, k+1 1 j+1/2. k+1/2

Table 3.4 Notational Convention for k-Dlrection Momentum


Control Volumes

Digit Cell Center Cell Face Center


Direction Subscript Location Indices Location Indices

center 0 1, j , k+1/2 (not applicable)


west 1 i-1, j , k+1/2 i-1/2, j , k+1/2
east 2 1+1, j, k+1/2 i+1/2, j , k+1/2
south 3 1. j-l, k+1/2 i, j-1/2, k+1/2
north 4 1, j+1, k+1/2 i, j+1/2, k+1/2
bottom 5 i, j , k-1/2 1. j, k
top 6 i, j, k+3/2 i, j. k+1
18

Use of these staggered momentum control volumes allows the pressure


difference between adjacent grid points to be used directly in the forms of the
momentum equations which result from integration over the momentum control
volumes (cf. Sec. 4 ) .
19

4. FINITE-DIFFERENCE FORMULATION

The finite-difference equations are derived by integrating the governing


equation, Eq. 2.1, over the control volumes defined in Sec. 3. Each term in the
equation is treated separately for the main and momentum control volumes in the
following subsections. The resulting formulation is implicit In time; i.e., all
variables are assumed to have new-time values (except for the use of the old-time
value in the unsteady term). All spatial differencing of convective terms is
done on a pure-upwind (donor-cell) basis. A Cartesian coordinate system having
spatial coordinates (x,y,z) is used to illustrate the spatial-differencing
aspects; the forms of the equations in a cylindrical coordinate system having
spatial coordinates (r,e,z) are very similar to the forms shown for a Cartesian
coordinate system and are obtained from the Cartesian forms by using the variable
substitutions shown in Table 2.2 (e.g., by replacing Ax with Ar, Ay with rAff,
etc.). Similarly, the finite-difference form of the momentum equation is
Illustrated by showing the form for the x-direction momentum equation; the forms
of the y- and z-direction momentum equations are very similar to the form of the
x-directlon momentum equation and are obtained by interchanging the roles of y
and z respectively, with x and the roles of v and w, respectively, with u. A
more detailed discussion of the details of the derivation process is presented
in Ref. 13.

4.1 Unsteady Term

4.1.1 Main Control Volume


Representation of the unsteady term is obtained by assuming that the
values po and 4,^ prevail over the control volume surrounding point 0 (cf.
Fig. 4.1a). Integration of the unsteady term over the main control volume then

: . . J * 1

J - 1

(b) X Momentum

Fig. 4.1. Control Volumes Showing Variable Placement


20

gives

• d , x> J .. .J ( P 4 > ) O - (P4>)g"^ . .


-^(YvP<t>) dxdydz = r- V, (4.1)
At
where

Vjt, = Yv A x A y A z (4.2)
is the volume of fluid in the main control volume surrounding point 0, At is the
time step size (i.e. t - t""^) and the superscript n-1 refers to the known old
time-step values. The new time-step values are written without a superscript (it
would be n) for notational simplicity in this and the remaining equations.

4.1.2 Momentum Control Volume

The geometrical differences between momentum and main control volumes and
the differences between definition points for field and flow variables leads to
differences in the integrated forms of the terms in the conservation equation for
the two control volumes. The unsteady term is processed by applying the chain
rule of differentiation to the integrand and then performing the integration to
give

/ A ( Y ^ P 4 , ) dxdydz = / ( p - ^ + <I>|^)YV d x d y d z

=[(^(nr)*(*¥^)(^)) (4.3a)

where the bar over a field variable is used to denote that the value to be used
must be for the momentum control volume rather than the main control volume for
which it is nominally defined. The fluid volume and density for a momentum
control volume, thus, must be defined as averages of the fluid volumes and
densities in the two adjacent main control volumes which are intersected by the
momentum control volume. The specific averages used for an x-momentum control
volume (cf. Fig. 4.1b) are

r(Yv,i + Yv,i»i)-^(AXi + AXi,,) A y A z (4.4)


and

Po = Pi. (4.5)
AXH + Ax, ,

Furthermore the density is assumed to be sufficiently slowly varying in time


that the following approximations may be used:

=p (4.6)

J-
At at (4.7)
21

With these assumptions, the Integrated form of the unsteady term for a momentum
control volume may be written as

Po . if^o (4.3b)
/•^(Y,P*) dxdydz-' <l>o- At 2latj
At 2l at

where 4, is specifically u in this equation. The (afpo/at)"'' term in Eq. 4.3b


is based on the change in density during the most recently completed time step.

4.2 Convection Term

4.2.1 Main Control Volume


Integration of the convection terms over the control volume may be
expressed as

/[-|^(YxPU<t>) +-|^ (YyPV*) + ^ (YzPw<l>)]dxdy dz

+ Fj«t)>6 - F5<(t»o - (4.8a)

where subscripts 1, 2, 3, 4, 5, and 6 correspond, respectively, to the west,


r'-.st -louth, north, bottom, and top faces (cf. Fig 4.?a>.

4 »24
j ^ 1 -1-
4 f. U
1

j-l- -
II m4 ^ f ? - '"2

L
i
|_3
(.-•-:-
1 _, i - 1 i i t 1
L . i'-1 i +1

(LI) X ,";nillL-lllUl.l
(0) Main

Fig. 4.2. Control Volumes Showing Convective Fluxes


22

The quantity F is the mass flow rate (i.e., the product of density,
velocity, and flow area) across the surface of the control volume; for example,
the flow rate across the east surface is given by

F , = (u YxAyAz)^,^^^ . ,^<p>^ = ( ^ A , ) i . i / 2 , i , . < P ^ ^ ' ^'-'^


where the flow area on the face in the x-direction is given by
(4.10a)
Ax = Yx A y A z .
The flow areas in the other two directions are given by
(4.10b)
Ay = Yy A>^ A z
(4.10c)
Az = Yz A x A y

and the mass flow rates for all six faces are shown in Table 4.1.

Table 4.1 Convective Fluxes for Main Control Volume

^1 = (^Jl-l/2,^..<P>°

^2 •• i^A.U2.i..'P>°

^^ •• (^^)i,^.l/2.K<P>2

^= •• ("^^)i,3,.-l/z<P>0

^^ •• ^A^)i,3,K.l//P>°

The pure-upwind (donor-cell) difference scheme is used to define a


property value at the surface of a cell; this is expressed mathematically using

,, f F (J),2 , for F > 0 |


F «t» 1 = \ ' } . (4.11a)
* " [ F (t>,i , for F<0'^

Uniformity of t h e f u r t h e r development of t h e equations i s aided by w r i t i n g


Eq. 4.11a i n the following general form

F «i»ll = l i e F||4.,2 - ||0, -F||<t)„ , (4.11b)


23

where the symbol | A , B | designates the greater of A and B. The form given by Eq.
4.11b is also advantageous since its use results in vectorizable coding.

Use of Eq. 4.11b allows Eq. 4.8a to be written as

/ [i^c-p^*)*-|;(''yp''*)"i<^^p''*)l '^^y'^^
= [|0, FJ + lO.Fjj + 10, F J

+ 110, -FJI + 110, -F3II + 110, -F5I ] (t>o

-[10, -Fj|(j)2 + 10, - F j * , + l|0,-Fj|(t>e

+ 110,FJI*! + |0,F3l(J)3 + ilO,F3l|<t.5] . (^-Sb)

4.2.2 Momentum Control Volume


The integrated form of the convective terms for the momentum control
volume is analogous to that for the main control volume and may be written as

/ [-|^(^xP"*)^-|^(''vP''*)*-^(Y^p''*)l ^"^y^^
= [HO, F j + |0, FJI + 10, F j

+ II0, -FJI + 10, -F3I + 10, -F5I ] (t>o

- [||0, - F J * ^ + 10, - F j * , - 10, -FJI*,

+ I C F J I * ! * IO,?,!*, + IIO.FJ*^ ] . (^-12)

As was the case for the unsteady term, a bar over a quantity (in this case, the
mass flow rate) denotes an average of quantities nominally defined for main
control volumes.
The x-direction momentum control volume depicted in Fig. 4.2b will be used
to illustrate several examples of the averaging processes used to define mass
flow rates for momentum control volumes. Since the west and east faces of this
momentum control volume are located between the faces of main control volumes,
the i-direction mass flows rates for these two faces of the momentum control
volume will be defined by averaging the volume flow rates for the two main
control volume faces which are on either side of the momentum control volume face
and multiplying by the density of the main control volume in which the momentum
control volume face is located; for example, the mass flow rate across the west
face of the momentum control volume in Fig. 4.2b is the average of the volume
flow rates across the west and east faces of main control volume 0 multiplied by
the density of main control volume 0, or
(4.13)
h = |[(^)i-i/a,i,k M^xK.,/,,,,,]Po •
24

Since the j- and k-direction faces of this momentum control volume each coincide
with half of two different main control volume faces, the mass flow rates in
these two directions for the momentum control volume will be the sum of the mass
flow rates across the two half faces of the coincident main control volume faces.
For example, the mass flow rate across the north face of the momentum control
volume in Fig. 4.2b is the sum of half of the mass flow rates across the north
faces of main control volumes 0 (F^) and 2 (F24) , or

F4 = I F . + 1^24 • (^-H)
In a similar manner, the mass flow rate across the top face of this momentum
control volume (not shown in Fig. 4.2b) is the sum of half of the mass flow rates
across the top faces of main control volumes 0 (F,) and 2 (F24), or

Fe = -?F, + ^F^e . (4.15)


2 * 2

The mass flow rates appearing on the right-hand sides of these last two equations
are defined using the method given in Sec. 4.2.1. The mass flow rates for all
six faces of an x-direction momentum control volume are given in terms of base
variable definitions in Table 4.2.

Table 4.2 Convective Fluxes for x-Momentum Control Volume

F • 1
^ = 2 [("'^x)i-l/2,i,.M^x),.,,,,,.JP0

^^ •• l[('^^«)i.l/2,i,K^(^x)i.3,,,,,,]P.

^3 = i K)i,i-i/...<P>o + (vA,K.,,,_,,,,,<p>f ]

^. = i K)i,i.i/.,K<p>? Mv\),,,,..,,,,,<P>L]

^' = i K)i,i.K-l/.<P>0 - («A.)i.l,3,..l,.<P>f ]

^^ •• \ [(^^^)i,i,..l/2<P>° - («Mi.l,3,K,l/.<P>26]
25

4.3 Diffusion Term

4.3.1 Main Control Volume

Integration of the diffusion terms over the control volume may be


expressed as

= 02*1*2 -^o) -D?(4>o -4>i) +D*(({), -<t>„)

- D3*((J.o-<l>3)+Df((t),-*o)-D5(4>o-«t>s)

= D ? * ! +Df({>2 +D3*<t>3 +D*(j>, +D5*4)5 + 0 * * 6

(4.16)
- ( D * + D ? + D * + D j + D* + D*)(t)o •

The quantity D is a diffusion strength for the indicated surface and is


defined as the product of the effective diffusivity and the flow area divided by
the distance between the centers of the two adjacent main control volumes. The
diffusivity, TJ,, is assumed to be uniform within each main control volume and a
harmonic Interpolation is used to obtain the value at the surface. For example,
the effective diffusivity for the east face of the main control volume shown in
Fig. 4 ''a is given by

j +1

j -1

(S) X 'lor^entJT
(a) Main

Fig. 4.3. Control Volumes Showing Diffusive Fluxes


26

AXi/2 + AXi.i/2
(r*)e«,2
AXi/2 AXi,i/2 (4.17)
(r*)o (r»)2

The same procedure is used for all six faces.

The diffusion strengths for all six faces of the main control volume are
shown in Table 4.3.

Table 4.3 Diffusion Strengths for Main Control Volvime

D* Ax Ax
'1-1/2,i,k 2r,• / i 2r„

Ax Ax)
°* = (^x)l.l/3,,,. 2r.• Vo 2rJ2

.Ay)
^ * •• (^)i..-l/2
2rJ3 2rJo

°* = (^)i,,.l/2..
tl- .Ay)
2rJ,

D* : (A,i Az Az
^ ^'i,j,k-1/2 2r, 2r„

Az^l Az
'' •• (^^)M,.n/
2rJ„ 2r,

4.3.2 Momentum Control Volu

The integrated form of the diffusion terms for t-h» „ • , ,


is analogous to that for the main c o n t r o l \ " w t d mayTe^^rTttr'a"'™'""^
27

= D**! +D*(J>2 +D?*3 + 0 ! * , +D**5 +D*<t>,

- ^ • + Df + D ? + D t + D t + D*)<t.o . (4.18)

As was the case for the convective terms, the only difference between the forms
of the diffusion terms for the momentum and main control volumes is the use of
a bar over the coefficients to denote that the quantity is for a momentum control
volume.

The diffusion strengths, D , used in Eq. 4.18 are defined as follows,


using the x-momentum control volume depicted in Fig. 4.3b as an example. Since
the west and east faces of this momentum control volume are located at the grid
points of the main control volumes, the diffusivity of the applicable main
control volume may be used directly. (The diffusivity is specifically viscosity,
H, in the momentum equations; however, use of the symbol r^ is retained in this
section to facilitate intercomparison of equations.) The area which it
multiplies is the average of the flow areas at the adjacent faces of the main
control volume. For example, the east face of the momentum control volume shown
in Fig. 4.3b would have a diffusion strength Dj calculated using (r^)2 for the
diffusivity, [(A,)i^i/2,j,k + {Ax)it3/2,j.kl/2 for the area, and Ax^^i for the
separation distance.

A different procedure is used for the other two directions. Since the j-
and k-direction faces of this momentum control volume coincide with half of two
different main control volumes, an average diffusivity is first calculated for
the momentum control volume on each side of the face by averaging the
diffusivities for the main control volumes which are intersected by each momentum
control volume; for the north face of the momentum control volume shown in
Fig. 4.3b, these would be

(^•)o = i (1^0 + r2)


in the shaded momentum cell and

in the adjacent momentum cell to the north. These two diffusivities are then
used In a harmonic interpolation to get the effective diffusivity for the
surface; for the north face of the momentum control volume shown in Fig. 4.3b,
this would give
_ A y , / 2 + Ay,.,/2
(^•)eff,4 - Ay3/2 ^ Ay^.,/2 ' (4-21)

The area in the diffusion strength is half of the areas of the two main control
volume faces which form the momentum control volume face; for the north face of
28

the momentum control volume shown in Fig. 4.3b, this would be [(Ay)ij+i/2,k +
(Ay)i+i,J+1/2,kl/2- The separation distance is the distance between the centers of
the adjacent momentum cells, which is [Ayj + Ayj+i]/2 for the north face of the
momentum control volume shown in Fig. 4.3b.

The resulting diffusion strengths for all six faces of an x-direction


momentum control volume are given in Table 4.4.

Table 4.4 Diffusion Strengths for x-Momentum Control Volume

^ = i [(^x),.,/2.i,k MAx),.,/2.3,k] (^\

^ = i [(Ax),.,/2,i.k * (Ax),,3/2,,,k] (^\

°' '• 1 [(^)i,3-i/2,k * (^)i,l,3-l/2,k] ( T T i v w T v T i r " [{r^)o + '(r,)2]

°* = 1 [(^)i,i.i/2,k * (Ay)i.i,3.i/2,k] ([(r^),+ (r^)2] ^ T o v T ^ l f v U

D* • — r/A ^ + Ik \ 1 / Az^_^ Az^.


• • 2 f •'..!.«-./. I .),.,,i,.-.„J (i(r.i,.(r.i„i * ((r.)..(r,),i,

^ ' ^P-V....».-'--K.........]{TCV7^- ii,;,!!',r.i,.,!

4.4 Source Term

assumed ' to
t conform
c r ^ o ™ ItoT '" ''^ ^^""^^ ^°™ °^ ' ^ ^ c o n s e r v a t i o n equation is

s*'• == (Sc4,)„ + (Sp,j)„(t)o + E [(Sp*),*,] - (4.22)


where (S,,^)^ is assumed to be constant over the control vol,,n,
point 0, and (S,,), and 4, are constant over the control vo r ^"'^"""^^"S grid
point f. The form of Eq' 4.22 allows source terms which ar are a^"'^"""'^^"^ ^'''
function of the
29

variable of solution (4>) to be evaluated at advanced-time values, retaining the


impliclt-in-time nature of the formulation. Each conservation equation has
different source terms, the specifics of which dictate the S terms in Eq. 4.22.

4.4.1 Main Control Volume

The main control volume is used for the mass, energy, turbulent kinetic
energy, and turbulent kinetic energy dissipation rate conservation equations.
The source terms for the mass and energy conservation equations are given in
Tables 2.1 and 2.2.

The mass conservation equation has no source term.

The energy equation (for which * is h) has several source terms. The
Internal heat generation term, q", is part of (5^^)^ since it is not a function
of 0. The structure-to-fluid heat transfer, q^,, contributes to (Sc^)o and to
(Sp*)o »"d (S^)t depending on the solution technique selected (cf. Sec. 6.3.5);
this is the only term in any conservation equation which uses the summation term
in Eq. 4.22, the summation representing the coupling of fluid cell enthalpies due
to heat transfer through a thermal structure. The DP/Dt and « terms are
neglected in the present code version (as was the case in prior versions (Refs.
1, 3, and 5)).
The source terms for the two turbulence equations are discussed in Sec. 7.

The integrated form of the source term for the main control volume, thus,
may be written as

| Y V S » dxdydz = (S,<,)^V„ + (S^^)^ v,„ ^„

(4.23)

4.4.2 Momentum Control Volume

There are several source terms which are common to each direction's
momentum equation (for which 4 is either u, v, or w). The gravitational
acceleration (g) is constant and is multiplied by the density of the momentum
control volume (Eq. 4.5) to become part of (S,^)^. The pressure increase
supplied by a pump ((AP/L)p as defined in Sec. 8.5) is constant over a momentum
control volume and becomes part of (S^,^);,. The distributed resistance (friction)
terms (R,, Ry, Rj as defined in Section 9.4) involve the velocity in the cell;
the velocity in the direction to which the momentum equation is being applied is
factored out of the term and the remainder of the term becomes part of (S^)^
(cf. Sec. 9.4). The special momentum terms arising in hexagonal fuel assemblies
(cf. Appendix A) are treated as distributed resistances. The viscous diffusion
terms (V,, V , V^ as defined in Sec. 7.3.2) only appear for momentum control
volumes which' are adjacent to a zero-slip boundary surface when one of the
nonconstant turbulence models is used; these terms involve the velocity; after
factoring out the velocity, the remainder of the term becomes part of (Sp^)„.

The normal pressure term is treated separately, rather than being part of
either S^s or Spj. After integration over the momentum control volume, a term
30

results which contains the difference in pressures at the grid points in the main
control volumes which intersect the momentum control volume.

The r- and 9-direction momentum equations in cylindrical geometry have


terras which are not present in their x- and y-direction counterparts in Cartesian
geometry. These terms arise from the convection and diffusion terms, since the
spatial unit vectors in polar coordinates are a function of position. These
terms normally become part of (S(,,^)Q; if the term contains the velocity of
solution (i.e., u in the r-direction momentum equation and v in the 9-direction
momentum equation), this is factored out of the term and the remainder of the
term becomes part of (Sp,^)o.

The integrated form of the source term for the x-direction momentum
control volume, thus, i. ay be written as

/ Y V S * d x d y d z = {S,^)V,, + (Sp^)^ V£„ *„

- -. (P2-P0) • (4.24)
-(AXi + Ax,,,)

Similar expressions result for the y- and z-direction momentum equations, with
P2 replaced by P4 and P^ (which are the pressures at the grid points which are,
respectively, north and above grid point 0) and the Ax terms replaced with Ay and
Az for the appropriate positions.

4.5 Complete Equation

The finite-difference form of the full conservation equations may be con-


structed by substituting the individual terms from Sees. 4.1 through 4 4 into
Eq. 2.1. ^ '

4.5.1 Main Control Volume

The individual terms are given by Eqs. 4.1, 4.8b, 4.16, and 4 23 and
substitution into Eq. 2.1 gives the following form

a*(t.„ = a*<i), + a*(t>2+a3*())3+a*(t>,+a|(t>3+a|'<t),

+ E (a.**,) + b * , (4.25)

where the a and b terms are given In Tahlo L <=, TI,„ C


in the c o e f f i c i e n t s for t h e ^ l u i d ^ i r g y t,Ld'r.\lf, t T ' ' i r L r Z l
T.2c.^:\z\TXzr"° 1'"/' ° ^^ '^""p^^^ ^° "^1 '"through a ther::"
selected by the user t h " " ' ' . " ^ """""^^^ described in Sec. 6.3 5 has been
cells f o r ^ < 6 and oV H- "' '^-P^-ent the thermal coupling of adjacent
cells ror I < b and ot nonadiacent cells fni- P >. t TI. . . •'
kinetic energy, and turbulent kinetic e n e « v dissi' ^ " " ' m u i t y , turbulent
have (S,,), = 0 for f = 1 L and af - o T o r f ! '7 " " L " ' ' equations always
Table 4.5 Coefficients of General Finite-Difference Equation

Main Control Volume (Eq. 4.25) Momentum C o n t r o l Volume ( E q . 4.28)


Coefficient

||0,Fj|+Df+(Sp^)^V£„ I|0,FJ|+D*

10, -Fjl+D*
a* 80, -Pj*Dt*{S^)yto

||0,F3l+D*+(Sp^)3V£„
|0,F3l+D*

lie -Fj|+D*+(Sp^)^V£„ 10, -FJI+D*

|10,Fj|+D|'+(Sp^)^V„ |0,F5«+Dt

SO, -Fj|+D*+(Sp^),V,„ 10, -FJI+D*

a* {=7,...,L
not applicable

Dp At (Sc»)o .4(f)-]«-..,s.,.|.
1 0 , - F J + 1 0 , F2 1 + II 0 , - Fj 1 10, - F j + | 0 , F 2 l + 10, -F3II
a*(l)

+ | 0 , FJI + IIO, -F3II + IO, F j + | 0 , F J | + 10, - F j l + 10, F J


(1st form)

E^.* P° - (S ) + E B? + t 2\ dt ) (Sp*)o
1-1

not a p p l i c a b l e
a*(2) E [af-(Sp*),V,o]+ - ^ - ( S p * ) o
(2nd form)
32

Coefficient af has been written in two forms in Table 4.5. The first form
is the direct result following the substitution process described in the
preceding paragraph. The second form results from subtracting the continuity
equation from the first form. The continuity equation comes from Eq. 4.25 with
1^-1 and using coefficients from Table 4.5 with all D-0 and all S-0; with these
substitutions, the continuity equation may be written as

-E^ + IIC-F^I + I C F ^ I + IO,

+ 1 0 , F J | + 10, -F5l + | | 0 , F J |

1 0 , F J | + 10, -F2l| + | 0 , F 3 | |

,n-lT
+ | | 0 , - F J | + 1 0 , F J | + 10, -FJI+ •^\'°} =0 . (4.26
Note that the first density term in the above equation is at the new time and the
second density term is at the old time. The second form of ag'* is then computed
by subtracting this equation from the first form, or

aj(2) = a j ( l ) - {continuity f r o m Eq. 4.26} . (4.27)

This second form of af may be used in formulating the two turbulence equations
and the energy equation, since these three equations are constructed and solved
after the continuity equation has already been satisfied in the solution sequence
(cf. Sec. 11.2) used in COMMIX.

4.5.2 Momentum Control Volume

The individual terms are given by Eqs. 4.3b, 4.12, 4.18, and 4 24 and
substitution into Eq. 2.1 gives the following forms for the x-, y-,' and
z-direction momentum equations:

ao"uo = ai"u^+a2"u2+a3"u3+a4"u,+a5"u5+a6"u5+bo"

- I (Yv,o+Yv,2) Ay^Az^ (Pj-Po) . (4.28a)

aoX = aiX+a2''v2+a3V3+a4X+a5"v5+a6X+bo"

-^<Yv,o + Yv,<,) AXiAz^(P, - P J ,


(4.28b)
and

ai Wj + aj W2 + a3"w3 + a^^w, + a^-w. + a^w^ + b„"

j(Yv,o+Y.,,) Ax,Ay. ( P , - P „ ) , ^^ ^g^,


33

where the a and b terms are given in Table 4.5. The conservation equation for
a momentum control volume, thus, has the same general form as that for the main
control volume except for the appearance of the pressure gradient term. These
pressure gradients are included as separate terras, since the pressure is neither
known beforehand in the solution sequence (cf. Sec. 11.2) nor can it be
eliminated in terras of other known quantities. Thus, pressure must be regarded
as an unknown and will be determined from the constraint that the velocity field
must satisfy the continuity equation (cf. Sec. 5).

Only the first form for a^ is shown for the momentum equations in
Table 4.5. Use of the second form, which is acceptable for the main control
volume equations, could Introduce inconsistencies in the momentum equations,
since the continuity equation may not be satisfied at the point in the
calculational sequence where these coefficients are calculated.
34

5. PRESSURE EQUATION

The pressure appearing in the momentum equations (Eq. 4.28) is an unknown


and will be found by requiring satisfaction of the equation for conservation of
mass.

The equation for conservation of mass (continuity) for the main control
volume around point 0 (cf. Fig. 4.1) is found by using the general
finite-difference form from Eq. 4.25 with the following substitutions: variable
4-1, diffusion strengths D-0, and source S-0. In Cartesian coordinates, these
substitutions result in

^" ( - ^ ^ ) - (^=<")i-l/2,3,/P^° ' (^xU),.,,2.3,k<P>^°

-(M)i,3-l/2.k<P>° M^V),,3.1/2,k<P>°

-(AzW),.^,,.,/2<P>' MAzW),,^,,.,,2<P>° =° ' (5-1)

where the fluid volume (V^Q) , the flow areas (A), the velocities (u, v, and w ) ,
and the upwind density (<p>) are all defined in Sec. 4. This same equation
holds for cylindrical coordinates with the variable substitutions indicated in
Table 2.2.

The velocities appearing in Eq. 5.1 may be replaced with functional forms
involving pressure which result from the momentum equations presented in
Sec. 4.5.2. For this purpose, the momentum equations, Eq. 4.28, may be written
in the following form

(t.=$-a*AP (5.2)
for 0=u, V, and w. The velocities on the six faces of the main control volume
surrounding point 0, thus, may be written as

i) ' (5.3a)

<^2 - d2(P2 - Po) . (5.3b)

^3 - d ^ P o - P3) - (5.3c)

Po) . (5.3d)

P5) ' (5.3e)


and

Wi,J.k.i/2 = W, = 0 , - a^(p, - P„) , (5 3f)


where
35

T (aUt) + bo
= _ ^-i (5.4)
a;

The a and b terras used in Eq. 5.4 are given in Sec. 4.5.2, and the d terms used
In Eq. 5.3 are defined in Table 5.1.

Table 5.1 Coefficients of Pressure Difference In Velocity Equations (Eq. 5.3)

dl : |(7v,i + Tv.olAy^Az^ / (a?)^.i,2,j.k

d2 : -i(7v,o + Tv.zJAy^Az, / (aj),.,,2.j.k

dj : -|(7v.3 + 7v,o)Ax,Az, / (aj), j.i/2.k

ar : -i(7v.o + 7v,.)Ax,Az, / (aj), j.,,2.k

a^ : l(7v.5 +7v,o)Ax,Ay^ / i^o) ,.,.y,.^;z


2^

dr : -i(7v.o +7v.6)Ax,Ayj / (ao),.3.k.

The density is assumed to be sufficiently regular in temporal variation


that the following approximation may be used:

Po - PV . i^Y'^ (5.5)
At I 5t j
This is the same approximation which was used in Sec . 4.1.2 for the unsteady term
in the momentum equation.
Substitution of the velocities given by Eq . 5 . 3 and the density derivative
given by Eq. 5.5 into the continuity equation given by Eq. 5.1 results in the
following pressure equation

ao^Po - afp, - a!P2 - a3^P3 - afP, - ajp^ - a^P, - bJ = 0 . (5.6)


36

where the coefficients a and b are given in Table 5.2. Equation 5.6 has the same
form as the general finite-difference equations presented in Sec. 4.5 for the
main (Eq. 4.25) and momentum (Eq. 4.28) control volumes. The system of these
pressure equations for all main control volumes will be solved simultaneously
using iterative methods. The pressure solution is then substituted into Eq. 5.3
to calculate the velocities. Further details of the solution sequence are given
in Sec. 11.2.

Table 5.2 Coefficients of Pressure Equation (Eq. 5.6)

a' = (Ax)i.,/2,:.k^"<P>o

^^' •• (Ax)i.,,2,3,k^2"<P>2

^' = (^)i.,-l/2,k^3^<P>0

^" •• (^)l,3n/2,k^^^<p>^°

ai : \IA\. . ^ , a5"<p>^
^'1,3.k-1/2 5 r g

^6 = (A,). . . , a"<p>°
^ ^'i.],k*l/2 6 r- 6

..P P p p p p p
ao : ai + aj + a3 + a^ + as + a /

bo^ = -^"(^f'MAxaw,,,k<P>o-(Axfl),,,,2.,.k<P>°

(M)l,,-l/2,k<P>0-(V),,,,.;2,k'<P>5

(^A,,,k-1/2<P>0-(M),,,,,.,,2<P>°

(The upwind d e n s i t y <p> must be chosen" based on t h e s i g n of v e l o c i t i e s given


by Eq. 5.3 r a t h e r than j u s t the p o r t i o n s of the v e l o c i t i e s which appear in the
c o e f f i c i e n t s in t h i s t a b l e . )
37

The system given by Eq. 5.6 defines the pressure field satisfying the
momentum and mass conservation equations. Of course pressures, P*, obtained
during the iterative solution process never satisfy this equation exactly.
Instead they satisfy the identity
ajp„' - alP,' - alP,- - alP,- - aJP," - a^^P^- - afp,' - bJ = «p,o , (5.7)

where 6p o is the mass residual for the main control volume surrounding point 0.

The iterative solution process obtains a solution for the pressure field which
minimizes the mass residuals to a user-specified degree. If the momentum and
mass conservation equations were satisfied exactly, the mass residual would be
zero.
38

6. FLUID-STRUCTURE THERMAL INTERACTION

6.1 Introduction

In many real applications, the presence of immersed solid objects affects


the thermal behavior of the fluid. This thermal interaction between solids and
fluids can not be calculated without also calculating the temperature
distribution within the solids. The present section describes the "thermal
structure" raodel which is used to calculate the teraperature distribution in
solids and the fluid-solid heat exchange. (The model for fluid temperature
distributions was presented in Section 2.)

The modeling of thermal structures has the following features:

- A structure can be planar, cylindrical, or spherical with the axis


aligned with any one of the three coordinate directions. The
intersection of the fluid coordinate grid system with the structures
divides the structures into thermal structure elements.

- Within each thermal structure element, temperatures are computed by


solving the one-dimensional heat conduction equation. The direction in
which this is solved is determined by the axis alignment. Heat
conduction in the other two directions is neglected.

- Each structure may consist of more than one type of material, arranged
in layers and separated by gaps. The thickness of each material may be
subdivided into a number of nodes.

- The various materials in a structure may contain heat sources. The


heat sources may have spatial and temporal variations.

- The spatial variation of thermal conductivity and specific heat


capacity are incorporated through the use temperature-dependent
properties for each material.

- The structure exchanges heat with the fluid, through the use of
convection heat transfer correlations, at either or both of the
structure's surfaces which have the same surface normal vector as the
direction in which the conduction equation is being solved within the
structure.

- Radiation heat transfer may occur among the surfaces of various thermal
structures. The intervening fluid is assumed to be transparent to
thermal radiation.

The user has a rather high degree of flexibility in preparing the input to
describe thermal structures, including the specification of geometry,
composition, fluid-structure and gap heat transfer coefficients, material
properties and radiation heat transfer view factors; the details of this input
are covered in the User's Guide (Ref. 9 ) .

«^r„.^ ^^^ remaining subsections describe the details of the thermal


structure modeling.
39

6.2 Geometrical Description

A thermal structure may be either planar, cylindrical, or spherical. The


surface normal vector of a planar-type thermal structure and the axis of a
cylindrical-type thermal structure must be aligned with one of the three fluid
coordinate axes. By way of illustration. Fig. 6.1 shows a cylindrical thermal
structure whose axis is aligned with the z axis of the fluid grid.

AZ

-ELEMENT Of
A STRUCTURE

Fig. 6.1. Definition of a Cylindrical Thermal Structure Element


Relative to the Flow Domain

Intersection of the fluid grid coordinates with the thermal structure


divides the structure into thermal structure elements. For the example shown in
Fig. 6.1, the intersection of the fluid's axial coordinate grid divides the
structure into thermal structure elements having a circular cross section and
height Az. Each thermal structure element has its own one-dimensional
temperature distribution. Adjacent thermal structure elements of the same
thermal structure are insulated from each other.

Each thermal structure element has two surface elements: inner and outer.
It is through these surfaces that a thermal structure element exchanges heat with
a fluid (by convection) or with another thermal structure element (by radiation).
40

One of these two surface elements may be adiabatic. Various arrangements of


thermal structure surface elements are shown in Fig. 6.2. Each surface element
may Interact with only a single fluid cell; this concept is illustrated in Fig.
6.3, where each quadrant of the cylinder, having been partitioned by the fluid
coordinate grid, is actually a separate thermal structure element. On the other
hand, a single fluid cell may interact with surface elements of several thermal
structures, as illustrated in Fig. 6.4.

Symmetry Line/Plane,

Fig. 6.2. Outer and Inner Surfaces of Several Thermal Strueture


Elements in the Fluid Grid Structure

^/
//)

Fig. 6.3. Four Quarter-Cylindrical Thermal Structure Elements


each Interacting with One Fluid Cell
41

® 0
0

Fig. 6.4. .Multiple Thermal Structures Interacting


with a Single Fluid Cell

Thermal structure locations are specified to COMMIX by stating the fluid


cell locations which are adjacent to each surface element. No further
s p e d ficity is employed which would, for example, locate the three thermal
struc tures shown in Fig. 5.4 at the specific locations shown within the cell; the
formu latlon in COMMIX only needs to know that the surfaces of these three thermal
struc tures all interact with the same cell. If both surfaces of a thermal
struc ture element interact with fluid, then these should be two different fluid
cells (In this regard, the geometry shown in Fig. 6.2b is only allowed if the
inner surface is adiabatic.) The two fluid cells are normally adjacent to each
other in the grid, but this is not a requirement.

Two cross-sectional views of a thermal structure element are shown tn


Fig. 6.5. This cylinder has two surfaces; the outer surface has been designated
surface 1 and the inner surface has been designated surface 2. The thermal
structure is composed of layers of materials, separated by gaps. Each material
region may be subdivided into a number of partitions. This partitioning forms
the spatial grid to be used in the finite-difference formulation of the
conduction equation. A teraperature will be computed at the center of each
material partition of each thermal structure element.
42

FLUID
FLUID

CROSS SECTION A-A

SURFACE 1

Fig. .5. Material Regions jnd Gaps for a Typical Thermal


Structure Element

6.3 Conduction Heat Transfer

6.3.1 Governing Equation

The governing equation for transient, one-dimensional conduction heat


transfer within a thermal structure element is given by
-P ^ _ 1 8 ,,,
(6.1)

where T, p , and Cp are the temperature, density, and specific heat capacity of
the material, r is the position coordinate (and is only truly a radius for
J,l,r,tT \ t = P ^ " ^ = ^1 t^"-"^! structures), A is the area whose unit normal
Iflux^ % X T ' ' ^ ' " " i " " " -^^ q i= the heat transfer rate per unit area
Ulux;, q IS the heat generation rate per unit volume, and t is time
43

6.3.2 Finite-Difference Formulation

Figure 6.6 shows the cross section of a typical thermal structure element
under consideration. Each element is divided into a number of material regions,
and each material region is further divided into a number of partitions.

1 i 1 1
. • • •
T, T, T, \

Fig. 6.6. Partitions in Cross Section of a Thermal Structure Element

Figure 6.7 shows a single material partition f and its two neighbors. The
energy equation when integrated over material partition f becomes

(5.2)
(A,.i/2q(.i/2 - A,.i/2q(-i/2) * Q . v ,

where superscript n-1 denotes a value from the end of the previous time step and
absence of a superscript denotes a value at the end of the current time step (n) .

Fig. 6.7. Energy Balance on Partition I of a Thermal Structure Element


44

The surface area between two material partitions, Af, used in Eq. 6.2
depends on the geometry and location of the thermal structure and is given by

•RODFR * AXji * Axj2 (planar) (6.3a)


RODFR * 27ir, AXj3 (cylindrical) (6.3b)
RODFR • 4nr, (spherical) (6.3c)

where Ax^ is the spacing of the fluid coordinate grid in the i-th spatial
direction at the location of the thermal structure and RODFR allows for multiples
or fractions of the basic thermal structure shape to be included. The exact
directions associated with II, 12, and 13 depend on the orientation of the
thermal structure; further details are in the User's Guide (Ref. 9 ) . The volume
of a material partition, V(, used in Eq. 6.2 is given by

V, = A, A r (6.4)

where Aj is given by Eq. 6.3.

The heat source appearing in Eq. 6.2 is given by

q,'" = O K (14) * q,o"' » f„j (t) , (5.5)

where QK is a spatial shaping function, f„£ is the nf-th transient function, and
q^Q is a reference volumetric heat source. The index 14 refers to a spatial
location in the fluid coordinate grid system; its exact meaning in terms of i,
j , or k depends on the orientation of the thermal structure element; further
details are in the User's Guide (Ref. 9 ) .

The differencing of the two heat flux terms in Eq. 6.2 depends on the
location of the partition boundaries; there are four cases to be considered:
adjacent material partitions (no gap), adjacent partitions separated by a gap,
and the two end partitions.

Adjacent Material ParCiCions: The heat flux to be used in Eq. 6.2 is


written in terms of the temperature difference as

q.*i/2 = U..1/2 (T,-i - T,) (6.6)

The overall heat transfer coefficient, Vf-n/i' ^^ based on conduction heat


transfer through the two half partitions:

U,,,,2 = 11-^^1 + l^^^\ (6.7)

where Ar i s the thickness of a m a t e r i a l p a r t i t i o n and X i s the thermal


c o n d u c t i v i t y . With use of Eqs. 6.5 and 5.7, Eq. 5.2 may be w r i t t e n as
(a, + b,_,/2 + b „ , / 3 ) T, = b,.j/2 T,_i + b „ , / 2 T,.i + d, , (6.8)

where
45

(6.9)
(pCpV), /At

(6.10)
b|»l/2 - U,.i/2 A,.1/2
and

(6.11)
d, = (q"'V), + a.T,"-

Material Partitions Separated by Cap: If material partitions I and f+1


are separated by a gap, the heat flux is still given by Eq. 6.6; however, the
overall heat transfer coefficient is based on conduction heat transfer through
the two half partitions and convection across the gap

(6.12)
I Alii] . _ ^ *i^^\
where U^.p is the coefficient for heat transfer across the gap. The revised
definition of V/^^^^ ^^ 'hen used in Eqs. 6.8-6.11.
End Partition 1: If surface number 1 has a nonadiabatic boundary
condition, the heat flux across the end surface is given by
(6.13)
qi/2 = u 1/2 (T,cool , 1 - T i)' <3r.l

where T , , is the temperature of the fluid in the adjacent cell and q,., is the
net radiation heat flux to this boundary from the surfaces of other thermal
structures. (Radiation heat transfer is discussed in Sec. 6.4.) The overall
heat transfer coefficient is given by a combination of convection through the
fluid to the surface of the material partition and conduction through half of the
material partition
(6.14)
Ucool.l \ ^ 'iJ

where U , , is computed from a user-supplied correlation. The heat flux on the


other (1+1/2) side of the first material partition is computed using Eq. 6.6 or
Eq. 5.17. The resulting form of the energy balance is given by
dl (6.15)
(a^ + bi/ 2)Ti -'1/2 •'cool.l

( I n s t e a d of Eq. 5 . 8 ) , where
(6.15)
d l = (q"'V)3 + a i X r ' + qi",'i Ai/2

and the a and b terms come from Eqs. 5.9 and 6.10, respectively. If surface
number 1 does not interact with a fluid cell, then biy2 is zero in Eq. 6.15.

End Partition L: If surface number 2 has a nonadiabatic boundary


condition, the heat flux across the end surface is given by
46

IT T ) + n""^ (6.17)
qL.l/2 (T, - T^ool.2' + yr,2 ' ^ '

where T^ool 2 is the temperature of the fluid in the adjacent cell and q^ 2 is the
net radiation heat flux to this boundary from the surfaces of other thermal
structures. (Radiation heat transfer is discussed in Sec. 6.4.) The overall
heat transfer coefficient is given by a combination of conduction through half
of the material partition and convection from the surface of the material
partition through the fluid

lAlIl] . ^L- (6.18)


\ >• L Ucocl,

where U^o^i 2 i^ computed from a user-supplied correlation. The heat flux on the
other (L-1/2) side of the last material partition is computed using Eq. 6.5 or
Eq. 5,13. The resulting form of the energy balance is given by

(ai. + \.U2 + bi,,i/2) TL = bi._i/2TL-i + bL.i/2Tcooi,2 + d^ (6.19)

(instead of Eq. 6.8), where

d, = (q"'V), + a . x r ' + q?:l A,.i,2 <^-20)

and the a and b terms come from Eqs. 6.9 and 6.10, respectively. If surface
number 2 does not interact with a fluid cell, then bL+1/2 is zero in Eq. 6.19.

6.3.3 Initial and Boundary Conditions

Boundary conditions must be supplied for both surfaces of each thermal


structure element. Several possibilities are allowed, all of which are in the
form of a heat-flux-type boundary condition and determine which, if any, of the
terms in Eqs. 5.13 and 6.17 are nonzero.

If the surface interacts with fluid, then a convection-type boundary


condition is employed, which is the first term in Eqs. 6.13 and 6.17. The user
supplies a correlation to describe the surface heat transfer coefficient (cf.
Sec. 9.3). If a surface does not interact with fluid, then biy2 or bL+1/2 " i H be
set to zero.

If the surface is part of a radiation surface, then a net heat flux to


this surface from the other radiation surfaces is included, which is the second
term in Eq. 5.13 and 5.17. The details of the radiation model are presented in
Sec. 6.4.

If the surface does not interact with fluid and is not part of a radiation
surface, then the surface is adiabatic. The q given by Eq. 5.13 or 6.17 is set
to zero along with the corresponding biy2 or bLti/2-

A given surface of a thermal structure may have a convection-type boundary


condition, a radiation-type boundary condition, or a combination of these two
types of boundary conditions. At least one surface should have a
fluid-convection-type boundary condition. Only one surface may be adiabatic.
47

There is no provision for the user to directly specify a temperature or


heat flux level (other than through the convection or radiation conditions above)
as a boundary condition for a thermal structure.

Thermal structure temperatures are initially set to zero internally by the


code at the beginning of a steady-state calculation. Using the fluid
temperatures and velocities initialized by the user, the code makes one pass
through the thermal structure equations, with an Infinite time step size and
Ignoring all thermal radiation, to Initialize the thermal structure temperatures
to nonzero values. A transient calculation should normally be started from a
converged steady-state calculation, so that the thermal structure temperatures
are properly initialized from the Restart file.

6.3.4 Solution
With the fluid temperatures and radiation heat fluxes taken as knowns, the
equations in the preceding section are a set of L linear equations in L unknowns
(the temperatures) for each thermal structure element. Equations 6.15, 5.8, and
6.19 may be summarized as follows:

Surface Partition (-1:


(ai + bi/2 + b3/2 ) Ti = b3/2T2 + (dl + bi/2T,,,i,i) (6.21a)

(If surface 1 does not interact with fluid, bi/2-0.)

General Interior Partition f-2,3 L-1:


(a, + b,.i/2 + b,.i/2) T, = b,_i/2 T,.i + b,.i,2 T,.i + d, (6.21b)

Surface Partition f-L:


(a, + b,.i/2 + b,.i/2)T, = b,.i/2T,.i + (d, + b,.i/2 T_i,2) (6.2lc)

(If surface 2 does not interact with fluid, bi.,i/2-0.)

These equations may be viewed as a matrix system [A1(T]-[D], to be solved for


[T], where the coefficient matrix, [A], is tridiagonal.

This system of equations may be solved by a combination of forward


elimination and backward substitution. First, rewrite Eq. 6.21 as

ci T. = b,.i,2 T,.i + A; (f = l L-1) ^^•"^>

a-L) , ^'•''^'

where the new coefficients are given by


K (f-l) (6.23a)
Cl = ai + bi/2 + b3/2 ('-1'
48

(5.23b)
b,.i/2 - b,li/2 / C.'.i (1=2.....h)

and
A; = d i + b i , 2 T _ , , i (c=i) (6-2^-)

A ; = d, + b,.i/2 A;.I / C;.i ({=2,...,L-1) (6-24b)

A L = dL + b,_i/2 A;_I / C;.i +b,.i/2 T,„„i,2 (f=L) . (6.24c)

Sweep through in the forward direction (t-l L) calculating the coefficients


C, and A', using Eqs. 6.23 and 6.24. Then sweep backwards (f-L,...,l)
calculating the new temperatures T, from Eq. 6.22.

6.3.5 Heat Transfer to Adjacent Fluid

Once the temperature distribution in the thermal structure element has


been obtained by solving the conduction equation, the net heat transfer rate to
the fluid may be obtained. If surface 1 interacts with fluid, then the heat
transfer rate from the structure to the fluid through that surface is given by

Qi =Ui/2 Ai/2 (Ti -T,„„i,i) . (6.25a)

If surface 2 interacts with fluid, then the heat transfer rate from the structure
to the fluid through that surface is given by

Q2 = U1..1/2 A1..1/2 (T, -T,„„i,2) • (6-25b)

The heat transfer coefficients used in Eq. 6.25 are defined by Eqs. 5.14 and
6.18. If one of the surfaces does not interact with fluid, then the
corresponding heat flux is set to zero. These heat transfer rates are summed
over all thermal structure elements, j , interacting with a given cell and then
divided by the fluid volume in the cell to obtain the source terra appearing in
the fluid enthalpy equation

qbs = (pSA I Vj„ , (6.25)

where Vfo is the fluid volume in the computational cell (e.g. >„ Ax Ay Az) .

The structure and fluid energy equations are solved in separate blocks of
the computation, rather than all equations being solved truly simultaneously.
The coupling is provided by the structure-to-fluid heat transfer rate. This heat
transfer rate must be irapllcitized in time in order for the coupling to be
numerically stable for all tirae step sizes. This is accoraplished by writing the
heat transfer rate at the end of an outer iteration, 1, in terras of the heat
transfer rate at the end of the prior iteration and the change in heat transfer
rate across the iteration. There are two recommended methods for doing this in
COMMIX: limited and full coupling.
49

In the limited coupling approach, the heat transfer rate at the surface
of a thermal structure element is assumed to be a function of the enthalpy hj
(which is related to temperature T^„„i) in the fluid cell which interacts with
that surface; therefore, the heat transfer rate (per unit fluid volume) at the
end of outer iteration 1 is written as

-i -i-1 3qt,3 i i_i (6.27a)


qbs = <3b8 + -g^ (ho - ho ) .

The definition of q^, given by Eq. 5.26 results in both terms on the right-hand
side of Eq. 6.27a actually being summations over all thermal structure surface
elements which interact with a given fluid cell. Each derivative needed for the
right-hand side of Eq. 5.27a is computed numerically using

aq^s _ q^a(h;"^+Ah) - q^^tho'^ (6.28)


aho " Ah

where Ah is a fixed enthalpy increment (currently 100 J/kg). The terras in Eq.
6.27a are allocated as follows in the finite-difference formulation of the source
term for the fluid energy equation (Eq. 4.23): the term qt,'"' + Oqbs/^ho) (-hV )
becomes part of (S.Jo and the terra dqljdhc becomes part of (Sph)o- This approach
accounts for the fact that the temperatures of the thermal structure and the
adjacent fluid are not independent and is exact if only one surface of a thermal
structure element Interacts with fluid.

In the full coupling approach, the heat transfer rate at the surface of
a thermal structure element is assumed to be a function of the enthalpy ho in the
fluid cell which interacts with that surface and the enthalpy h„ in the fluid
cell which interacts with the other surface (if there is such interaction);
therefore, the heat transfer rate at the end of outer iteration 1 is written as

-i -i-1 3q^s ,^,1 y.i-U . y ^^^" (h' - h'"M (6.27b)

where the summation is over all of the m fluid cells which are coupled to a given
fluid cell through thermal structures and the derivatives are computed
numerically using Eq. 6.28. The terms in Eq. 6.27b are allocated as follows in
the finite-difference formulation of the source term for the fluid energy

equation (Eq. 4.23): the term q^,'-! + (3qL/3ho) (-ho'"^) + E [ (3qb»/3h») ("hi,"') 1

becomes part of (S,h)o. the term flq^/dho becomes part of (S^Jo. ^nd each dqljBK
term becomes part of (S,0. f°r m - 1,2,...,L; each (SpJ„ term ultimately becomes
oart of a" for ra - 0,1, .,L. This approach accounts for the temperatures in
two differe"nt fluid cells being coupled through a thermal structure as well as
the Interdependence of structure and fluid temperatures. If the fluid cells
interacting with the two surfaces of a thermal structure element are physically
adjacent to each other (e.g., thermal structure surface 1 interacts with the
fluid cell at spatial location (I,J,K) and thermal structure surface 2 interacts
with the fluid cell at spatial location (I-1,J,K)), then this approach conforms
to the six-neighbor-cell coupling already provided in the basic finite difference
form of the fluid energy equation in Sec. 4. The derivatives appearing in Eq.
6.27b become part of the a* (m - 0,1,.,6) terms.
50

The fluid cells coupled through a thermal structure are not, however,
restricted to being physically adjacent; for example, thermal structure surface
1 may interact with the fluid cell at spatial location (I,J,K) and thermal
structure surface 2 may Interact with the fluid cell at spatial location
(1-3,J,K); this flexibility in thermal structure specification is not used
frequently but is extremely convenient for some cases. The basic
six-neighbor-cell finite difference form of the fluid energy equation is modified
to allow proper numerics for this nonadjacent cell coupling by adding terms like
a* h,„ (m-7,8,..) for those coupled cells which are not physically adjacent, where
af contains the derivative of the heat transfer rate with respect to the non-
adjacent cell's enthalpy. (The maximum number of nonadjacent cells allowed to
be coupled through thermal structure elements to another cell is limited by the
value of variable NTSCUP, which is set by a PARAMETER statement in the source
code at compile time; see User's Guide (Ref. 9).) When full coupling is
selected, the two surfaces of a thermal structure element must not both be
coupled to the sarae fluid cell and the same pair of fluid cells must not be
coupled by more than one thermal structure element.

Both the limited and full coupling approaches give the same ultimate
result for end of step temperatures and heat transfer rates during a transient;
the difference is in the amount of numerical work performed. The full coupling
approach requires more CPU time per iteration than the liraited coupling approach
but usually results in obtaining convergence in a sraaller number of iterations.
The tradeoff is such that the overall CPU tirae for a calculational history
usually goes down when the full, rather than limited, coupling is used, except
when the time step size is relatively small during a transient calculation.
Naturally, the full coupling approach is most helpful when the thermal coupling
is strongest (e.g. a sodium-to-sodlum heat exchanger). For cases where the
coupling is weak (e.g. where heat transfer coefficients, heat transfer areas, or
structure thermal conductivity are small), the liraited coupling approach is
adequate.

(A third method in the code is the original COMMIX technique, which


considers the structure and fluid temperatures to be independent in computing the
derivative of structure-fluid heat transfer rate. Use of this technique is not
recommended.)

Regardless of the assumption employed, the computation is arranged so that


the structure-fluid heat transfer rate and its derivative(s) with respect to
fluid cell enthalpy are calculated as part of the thermal structure block and
passed to the fluid energy calculation block.

5.4 Radiation Heat Transfer

6.4.1 Governing Equations

The radiation model allows for calculating radiant energy interchange


among the surfaces of thermal structures. Use of this raodel is optional; the
user declares via input specifications which thermal structure surfaces are to
be considered as radiation surfaces.

The process of radiative exchange occurring within an enclosure is very


complex. Radiant energy leaves a surface and travels through an intervening
medium toward other surfaces. The intervening medium may absorb and reflect the
radiation, as well as be an emitter of radiation itself. Upon reaching other
51

surfaces, the incident radiation is partially absorbed and partially reflected;


a portion of the radiation could be transmitted through the solid without
absorption. The reflected part of the energy, combined with other energy leaving
this surface, then travels to other surfaces, undergoing further absorption,
reflection, and transmission interactions. In general, the absorption,
reflection, and transmission processes are a function of the wavelength of the
radiation. The mathematical formulation required to track these beams of
radiation on a discretized basis would be very complex. The formulation is
simplified by using the "net radiation method." The net radiation method was
first devised by Hottel (Ref. 14) and later developed in a different manner by
Poljak (Ref. 15). An alternate approach was given by Gebhart (Ref. 16). All of
these methods are basically equivalent.

A number of assumptions have been invoked to make the modeling tractable.


All participating structures are gray, i.e., the emittance and absorptance are
not a function of wavelength. Reflections and emissions at surfaces are diffuse,
ie uniform in all directions. Participating surfaces are opaque, i.e.,
transmittance is zero. The intervening fluid separating radiation surfaces is
transparent, i.e. nonemitting, nonabsorbing, and nonscattering. The radiosity
Is uniform over an emitting surface.

Theraodelconsiders a set of N radiating surfaces. Figure 6.8 illustrates


the energy balance for the 1-th surface in the system. The total radiation
leaving the surface (radiosity) per unit area and per unit time is the sum of the
reflected incident energy and the energy emitted by the surface:

/ / / / / / / y y / / / / / / / / / / ^ / / / / - ^ - ^
Surface i at temperature, i^

Fig. 5.3. Radiant Energy Interchange on Surface i


52

Ji = PiGi + SiEb (6.29)

where p Is the reflectance, G is the incident energy flux (Irradiation), e is


the emittance, and E), is the black-body radiation emmlslve power, given by

E = oT*. (^'^C)
l^bi "•'li '

where o is the Stefan-Boltzman constant (5.67*10"' W/m^-K*) and T^ Is the absolute


temperature of the surface (i.e., in kelvlns).

An opaque gray surface has the following relationships between absorptance


(a), reflectance, and emittance:
a + p = 1 (6-31)
and
a = 6 . (6-32)

The irradiation G^ is the total radiation energy arriving at surface 1


from all other surfaces j, or

A.G. = t F.. A.J. , (6.33)


j-l

where Fji is the fraction of energy leaving surface j which arrives at surface i
(view factor) and A is the surface area. Use of the reciprocity relation,
AjFji = Ai Fij , (6.34)

allows Eq. 6.33 to be written as

SF.
G, = E F,,J, . (6.35)
3-1 '' '

Substituting Eqs. 5.30, 6.31, 6.32, and 6.35 into Eq. 6.29 gives

Ji = (l-Ei) t Fij Jj + EiOT^i . (6.36)

As illustrated in Fig. 6.8, the net radiation leaving surface 1 is the


difference between the radiosity and irradiation, or
<3rl = 'Ji - Gi • (6.37)

The irradiation Gj may be removed from this equation by rearranging Eq. 6.29 and
using Eq. 6.30-5.32, whereupon Eq. 6.37 may be rearranged to show

Ji = oT,^ - IZh q^^ . (6.38)


^1
53

Finally, substitution of Gj from Eq. 6.35 and J^ from Eq. 6.38 into Eq. 5.37
removes both the irradiation G^ and the radiosity J^ from the balance equation
and results in

(6.39)
t (F,, -».,)oTV
£(Fi, t ^^^'"^
- 8,,)oT^ = E - ^ F i-, --^Jq,,

as the energy b a l a n c e e q u a t i o n for s u r f a c e 1, where 6^ i s the Kronecker d e l t a


defined by
, fl if i=j\ (6.40)
^ij = (o if i o j / •

Since there are N surfaces to be considered, Eq. 6.39 is actually a set


of N equations having the following matrix form:
(6.41)
[B] [T,'] = [C] [q,] .

Matrices [B] and (C) are square (N*N); their elements are given by
(6.42)
Bi3 = (Fi3 - Sij)"
and
(6.43)
Cij = [(l-ej)Fij - 6i^] /e^

which are known and constant for a given set of radiation surfaces. Column
vectors [T*] and [q,] contain the fourth power of absolute temperature and the
net radiation heat flux, respectively, for each radiation surface. If the
temperatures of the radiation surfaces are known, the systera of equations may be
solved for the net radiation heat fluxes (and vice versa).

6.4.2 Solution
The solution of the equations governing radiation heat transfer among the
thermal structures is coupled to the solution of the equations governing
conduction heat transfer within thermal structures (which is, in turn, coupled
to the solution of energy equations for the fluid). All of these computations
are performed in separate blocks of the code, with temperatures and heat fluxes
from one block of the calculation being supplied to the next block. The
specifics of the coupling for radiation heat transfer are described below.

A radiation surface in COMMIX is defined as a collection of surface


elements of specific thermal structures, as specified by the user input. A given
thermal structure element may have both, one, or neither of its - - ^ f ; " / ! ^ " ^ " "
declared as a part of a radiation surface. Surface elements f"™ " ^ " ^ ^
different thermal structure elements may be combined to form one radiation
surface. Each radiation surface is assigned an index "^''^//""^"/./.'^"p;^^,
(The maximum number of radiation surfaces allowed is set by the PARAMETER PNRAD
in the code at compile time; see User's Guide (Ref. 9).)

The user must also provide an emittance for each radiation surface and the
complete set of view factors, F^^, which specify the fraction of radiation
54

leaving radiation surface i which is incident upon radiation surface j . These


view factors are a function of the shape and size of the two surfaces, as well
as their relative locations and orientations. In general, these view factors are
the result of a double surface integral; for many combinations of finite-size
surfaces, this integral can not be evaluated analytically. The wide range of
possible surface arrangements precludes automating the view factor calculation
process within COMMIX. View factors for many geometries are usually given in
heat transfer texts; a particularly extensive tabulation is given in Ref. 17.
For each radiation surface 1, these view factors should sum to 1, i.e.,

f; F.. = 1 , (6.44)
jTi '^^

otherwise energy will not be conserved in the systera. The input view factors
must also conform to the reciprocity law given by Eq. 6.34. The diagonal term
in the set of view factors, F n , is not necessarily zero, since a concave surface
can see (at least a part of) itself.

At the end of a converged time step, the thermal structure temperatures


are updated using the equations in Sec. 6.3 (i.e. Eq. 6.22 and associated terms).
The temperature of each radiation surface 1 is then computed to be the area
averaged temperature of all thermal structure surface elements k which form the
radiation surface i using

.r> ^^-^ ^^'^ (5.45)

The definitions of A and T used in Eq. 5.45 depend on which surface of a thermal
structure element forms part of radiation surface 1. The area is

Ai/2,k f°^ surface 1 (5.45)


AL.i/2,k foi surface 2

which uses the surface areas given by Eq. 5.3. The temperature is

_ 1^1,1; for s u r f a c e 1 (5.47)


^•^ " 1 TT'IC for s u r f a c e 2

which uses the temperatures at the center of the first or last material partition
in absolute units (kelvlns). (COMMIX does not track a true surface temperature
for thermal structure elements; for this reason, the first and last material
partitions of a thermal structure should be relatively thin when the surfaces
they adjoin form part of a radiation surface. Additionally, all teraperature
values in COMMIX are stored in units of degrees Celsius; since the temperatures
within the radiation heat transfer model must be in absolute units, conversions
are made from degrees Celsius to kelvlns and from kelvlns to degrees Celsius as
needed.) The use of the fourth power of temperature in this averaging process
results in an average temperature which properly characterizes the total emissive
power of the elements which form the surface.
55

With the temperatures of the radiation surfaces known, the net radiation
heat fluxes may be computed by solving the systera of equations given by Eq. 6.39.
This system of equations is solved using a set of routines from the LINPACK
(Ref 18) linear algebra package. The net radiation heat fluxes are converted
to a net radiation heat transfer rate for each surface by multiplying by the area
of the radiation surface (which is the denominator of Eq. 6.45). Provision is
also made for averaging the heat transfer rate obtained from solving Eq. 6.39
with the rate from the previous tirae step. The end of step net radiation heat
transfer rates are thus given by
(6.48)
Q,"i = (i-n,)Or":' + Qr qn Al ,
where Ai is the denominator of Eq. 6.45. The quantity n, is a user-input
parameter, in the range 0 to 1, which controls the averaging process. In theory,
fl should always be 1. In practice, the beginning of a steady-state calculation
usually needs n, « 1, since the radiation heat transfer rates are zero during
the first time step and are very sensitive to the radiation surface temperatures
as they begin to change during the first few tirae steps. Larger values of fl^ raay
be used in later stages of the calculation.

The values of the radiation heat transfer rate, as calculated at the end
of the time step, are held constant during the next tirae step for use in
computing thermal structure temperatures. These radiation heat transfer rates
are used in the boundary conditions for the thermal structure energy equation
(Eq 6 1 ) . The radiation heat flux is assumed to be uniform for all thermal
structure surface elements which compose a given radiation surface. Thus, the
radiation heat flux q,,, appearing in Eqs. 6.13 and 6.17 is given by Q,, from Eq.
6.48 divided by the denominator of Eq. 6.45.
56

7. TURBULENCE MODELING

7.1 Introduction

For turbulent flow, the diffusion transport coefficient, r, in the general


conservation equation is considered as an effective time-averaged value. In
COMMIX, this is accomplished by writing the viscosity, fl, to be used as the
diffusion coefficient in theraoraentumequations as

V^ = J*e£t = 1*1 + l^c (^•'•)

and writing the thermal conductivity, X, to be used in the diffusion coefficient


for the energy equation as

where the subscripts t and t stand for laminar (molecular) and turbulent
properties. The laminar values are provided directly by the equation of state
(cf. Sec. 9.1); the turbulent values are a function of the flow field. These
effective viscosity and thermal conductivity values are only used in the
diffusive terms in Eq. 2.1; friction and heat transfer correlations always use
Uf and Xf. There are five models in the present version of COMMIX for providing
values of the turbulence quantities ^l^^ and X^:

- User-input constant turbulent viscosity and conductivity,

- User-input spatially variable turbulent viscosity and turbulent


conductivity,

- Zero-equation mixing-length model,

- One-equation (k) turbulence model, and

- Two-equation (k-£) turbulence model.

After a brief background on turbulence modeling, the details of these


models will be presented.

7.2 Background on Turbulence Modeling

The subject of turbulence has attracted countless researchers over a


period of more than 90 years. In 1895, Reynolds (Ref. 19) proposed that a fluid
particle in turbulent flow is in randomly unsteady motion. To attempt a
solution.^he expressed the quantities as the sura of a mean and fluctuating value,
i.e. u - u + u', and then he averaged the Navier-Stokes equation over a time
scale that Is long compared with the turbulent time scale and derived the
equations that describe the mean turbulent motion. In spite of the long tirae
span and large research effort since Reynolds averaged the Navier-Stokes
equation, the problem of turbulence has not been completely resolved for the
following reasons. •

The appearance of the time-averaged correlations, such as pu'v' in the


governing equations, gives rise to the so-called "closure" problem (more unknowns
57

than equations available for the solution of unknowns). Here p denotes fluid
density, u' and v' are the fluctuating velocity components in the coordinate
directions x and y, and the overbar denotes the time averaging. The correlations
u'v' are known as Reynolds stresses.

Another difficulty is that the smallest turbulent eddies may be extremely


small compared to the whole flow domain. As a result, the number of mesh nodes
required to corapute the turbulent flow field will, generally, far exceed the
raaxiraum number allowed by storage limitations. Even if this were not true,
running times for problems with the required number of nodes would be
prohibitive.

An alternative approach to resolve these difficulties is to employ some


form of turbulence raodeiing in which only the time-averaged equations of motion
are solved along with a set of transport equations for turbulence quantities,
e.g., k the turbulence kinetic energy, c the rate of dissipation of k, etc. Even
this approach requires a significant amount of numerical computation. This
alternative turbulence raodeiing approach has becorae more feasible in the last 25
years, with the recent advances in computer power.

Many turbulence models have been proposed to resolve the above-mentioned


difficulties by providing solvable equations for computation of turbulent flows.
The central idea in most of the turbulence models, except the Reynolds-stress
model or algebraic stress raodeiing, is the eraployment of an artificial turbulent
viscosity Ut to account for the additional diffusional flux due to the turbulent
motion. To do that, the Reynolds stress terra is expressed as

- p U'V -
^A^'^)-i'^'^'''^''
" ' — ^ ' - II III - .f V T w I . V . /

Note that the turbulent viscosity H^ is a property of the local state of


turbulence and not a physical property of the fluid. A turbulence model in this
category is therefore generally referred to as a viscosity raodel.

The siraplest araong these viscosity raodels is Prandtl's mixing-length


hypothesis (Ref. 20). The mixing-length hypothesis is identified as a "zero-
equation" model because it does not require solution of any transport equation
of turbulence parameters.

In 1945 Prandtl (Ref. 21) suggested a more general approach than the
mixing-length hypothesis. His new approach is generally referred to as a one-
equation turbulence raodel. In this raodel, the turbulent viscosity is assumed to
be a function of the square root of the turbulence kinetic energy k. To
determine the value of k, its transport equation must be solved. Since then,
many one-equation turbulence models have been proposed. The transport equation
for the shear stress developed by Bradshaw et al. (Ref. 22) and the transport
equation for the turbulent viscosity developed by Nee and Kovasznay (Ref. 23) are
typical turbulent viscosity models.

Undoubtedly, one-equation raodels generally produce more reliable results


than the mixing-length hypothesis produces for most computations. However, a
need to obtain a more accurate estimate of the length scale distribution,
especially in a separated flow region, leads to the suggestion of two-equation
turbulence models.
58

There are several two-equation turbulence models (k-c model, k-f model,
k-U model, etc.) Here, the symbol k is the kinetic energy of turbulence, c is
the dissipation rate of turbulence energy, f is a macroscopic length scale of
turbulence, and W is interpreted as the time-averaged square of the velocity
fluctuations. Among the two-equation models, the k-c model, as proposed by
Harlow and Nakayama (Ref. 24) and Jones and Launder (Ref. 2 5 ) , is the most widely
used.

The next level in turbulence modeling is represented by the complex


Reynolds stress models (Ref. 26-28). These models are still in the development
stages; therefore, only the constant, 0, 1, and 2 equation turbulence models have
been programmed in COMMIX for the analysis of turbulent flows. These models are
described in this section.

As the level of turbulence modeling increases from 0 to 1 equation, from


1 to 2 equations, and so forth, the complexity in the turbulence modeling
increases and, therefore, computer cost increases. In the selection of a
turbulence raodel, the increase in accuracy raust be weighed against the increased
cost of coraputing.

7.3 Two-Equation (k-c) Model

The general approach in a k-e turbulence model is to solve two transport


equations: one for k, the turbulent kinetic energy, and one for c, the turbulent
kinetic energy dissipation rate. The specifics of how this is done, both in
terms of the transport equations and the boundary conditions, vary from
implementation to implementation. The transport equations presented are usually
restricted to being valid for either low- or high-Reynolds-number flow. This
presents difficulties in a general application code such as COMMIX, since part
of the flow field may be at low Reynolds number and part may be at high Reynolds
number in complex geometry problems; programming the code to make a transition
from one equation set to another could induce severe numerical instabilities.
Implementations of boundary conditions depend on whether the equations are solved
"all the way up to the wall" or whether the effect of the wall is brought in
through some sort of "wall function" treatment.

The two-equation turbulence raodel in the present version of COMMIX has


been constructed as follows. The starting point was the base transport equations
frora COMMIX-IB (Ref. 29), which are the equations for high-Reynolds-number flow
presented by Jones and Launder (Ref. 25) . These equations were then modified for
use in the full range of flows by incorporating terms for low-Reynolds-number
flows using works by Jones and Launder (Refs. 25 and 30) and Nagano and Hishida
(Ref. 31). The use of a "wall function treatment" for boundary conditions at
rigid walls was retained frora COMMIX-IB; however, the forra used to approximate
the velocity gradient was changed to incorporate low and high flow behaviors.

Most of these modifications involve multiplying certain terms in the


transport and boundary condition equations by coefficients which are smoothly
varying functions of a turbulence Reynolds number; these functions are created
in such a manner that the resulting equations have proper limiting forms for very
low and very high flows and a smooth transition between these two limits. The
forms of these coefficient functions have been adjusted to give good agreement
between calculations and experiments over a wide range of Reynolds numbers for
simple geometry problems. This involves a tradeoff between accuracy and
stability; the transition needs to be fairly abrupt to match experiraent results
59

but will cause numerical difficulties if too abrupt. More work is needed to
improve the current transition treatment.

Most validation studies in the literature presenting k-c turbulence models


utilize simple geometry (e.g. a circular pipe) and a fine grid structure (e.g.
20 to 100 nodes across the radius of a pipe); however, many practical
applications of COMMIX involve rather complex geometries, for which use of such
a fine grid structure is not affordable. A rather thorough investigation is
needed to demonstrate the validity of results from the present turbulence model
when used in conjunction with the porous-media formulation (y^^l) and a coarse
grid structure.

7.3.1 Transport Equations


The turbulent kinetic energy and turbulent kinetic energy dissipation
transport equations conform to the general transport equation given by Eq. 2.1
with * replaced by k and c, respectively. The diffusion coefficients, r^, and
source terms, S,^, are given below for each equation.

The diffusion coefficient in the turbulent kinetic energy transport


equation is given by
(7 4)
TR = H, + ^t/Ok '

where o^ is the effective turbulent Prandtl number for k. The source term is
given by
S, = Pk + G , - P t + E, . ^^-5)

The source due to mean shear is defined in Cartesian coordinates by

aui (d\ii au3^ (7.6)

where u, is the velocity component in the coordinate direction Xj. The source
due to thermal buoyancy is defined in Cartesian coordinates by

(7.7)
pP Oh
Oh ar l^axj 'j

where o^ is the effective turbulent Prandtl number for energy, T is teraperature,


and g, is the component of the gravitational acceleration vector in coordinate
direction x, The usual summation convention is employed in the last two
equations, where repeated occurrence of a spatial subscript (1 and j m this
case) indicates summation over all three spatial directions. The expanded forms
of the summation terms used in Eqs. 7.5 and 7.7 are shown in Table 7.1 for
Cartesian and cylindrical coordinates.
60

Table 7.1 Expansion of Summation Terms In P^ (Eq. 7.6) and G^ (Eq. 7.7)

Ul = u , U2 = V , and U3 = w

Cartesian Coordinates:

Xi = X , X2 = y , X3 = z

aui aui auj


ax. axi ax, MitfMirMir
/du av\2 Idu dw\2 Idv dwV^
[dy ax) laz dxj iaz ay)

„ ar „ar^„aT^„aT
arr = 9x 3ax
9: ax, ='>' 3::
: : + gy ay + Sz az

Cylindrical Coordinates:

Xi = r , Xj = 9 , and X3 = z

g^ = cos (6) • g,; + sin (6) • gy

ge = -sin (6) • g^ + cos (6) • g

aui dUi aUj /.^\^ + /1 av u\2 /aw\2


= 2
axj axi \ar) U ae l) lai)

+ / l i y + ^ - •^l^ 4. /au ^ aw\2 ^ /av i aw\2


U ae ar ' 7) ("a? a?) la^ 7 ae)

^^ ax. = Sx f + ge1 ar ^ „ ar
r ae az
61

The diffusion coefficient In the turbulent kinetic energy dissipation


transport equation is given by
(7.8)

where a is the effective turbulent Prandtl number for c. The source term is
given by
S. = tCi(P^ + G^ - fjpele/k + E. , (7.9)

where the turbulence coefficient function is defined by

fj = Cj [1 - 0.3 exp ( - Ret')] (7.10)

in terms of a turbulence Reynolds number

Re, = pkV (^,e) (^-1^^

and Cl and C2 are turbulence constants.

Components E^ and E^ in the source term definitions (Eqs. 7.5 and 7.9) are
both neglected in the current implementation of the k-c turbulence model. Their
omission is only strictly correct for high Reynolds-number flow. There Is
limited theoretical basis for including these terms in the equations for low
Reynolds-number flow; Instead, various researchers have added various forms of
these terras in order to obtain agreement between computational raodels and
experimental measureraents. Several forms for these terras are summarized in
Ref 31' as a specific example, Jones and Launder (Ref. 30) suggest
E^ = -2Mak'^V3y)^ and E. = (2\i.\i^/p) (d'u/dy')' for low Reynolds-number flow.
Further work is needed to ascertain the proper forms for these terms and to
Implement them in such a way as to result in a set of equations which provides
a correct and numerically stable solution for all flow regimes.

The finite differencing of all terms in the transport equations for


turbulent kinetic energy and turbulent kinetic energy dissipation rate proceeds
as described for the main control volume in Sec. 4. The unsteady, convective,
and diffusive terms are straightforward; the treatment of the source term depends
on the equation. The source term in the turbulent kinetic energy transport
equation, S. as given by Eq. 7.5, Is not directly a function of k; therefore,
this term becomes (S.Jo and (Sp,)o is zero. The source term in the turbulent
kinetic energy dissipation rate transport equation, S^ as given by Eq. 7.9, is
a function of c; therefore, the terra (S^/c) becomes (Sp£)o and (S,c)o is zero
These two transport equations are then solved in the sarae manner as the fluid
enthalpy equation (cf. Sec. 11).

After both equations have been solved, the turbulent viscosity is computed
from

\^t
f„c„pkVe
f
. (^12)

where the turbulent viscosity coefficient function is defined by


62

f^ = [1 - e x p (-Ret/26. 5)] 2 (7.13)

and C is a turbulence constant for viscosity. The turbulent thermal


conductivity is then calculated from

Xt = Cpjit/a^ , (7.14)

where Cp is the specific heat capacity of the fluid.

The preceding model presentation introduced six symbolic constants: C


Cl, C2, 0)(, Oj, and Oh. Various values for these constants have been recommended
by various researchers who invoke both theoretical and empirical arguments;
however, the constants are generally assumed to be "universal", rather than being
specific to a particular fluid or geometry. The following default values are
provided in the code: C^ - 0.09, Cj - 1.44, C2 - 1.92, o,, - 1.0, o,. = 1. 3, and o^
- 0.9; these values may be overridden by user input.

7.3.2 Boundary and Initial Conditions

There are several types of boundaries to be considered: symmetry, outlet,


inlet, and rigid boundary. The type of boundary is identified by the velocity
boundary condition specified for the momentum equation for each boundary surface
(cf. Sec. 10.1.1).

For a symmetry boundary, the velocity normal to the surface is zero and
velocity gradients in the direction of the normal are zero. Similarly, the
gradients of the turbulence quantities k and c are both zero. The actual value
at the boundary is not required in this case. This treatment is also used for
a free-slip rigid-wall boundary.

For any of the several outlet boundary conditions, the gradients of the
turbulence quantities k and c in the direction of the normal are assumed to be
zero. The actual value at the boundary is not used for outward flow of fluid in
the donor-cell convection formulation used in COMMIX.

For an inlet boundary condition, the velocity is specified by the user.


Ideally the user should specify the turbulence quantities at the inlet based on
experimental data; at present, there is only limited allowance for this. The
only provision is to set the inlet turbulent kinetic energy as a function of
inlet velocity using

•^in = Ck,i„ ufn , (7.15)

where C^ ;„ is a user-supplied coefficient. The inlet value of the turbulent


kinetic energy dissipation rate is then computed frora the inlet turbulent kinetic
energy using

^in = C,_in kin / (7.15)

where C^, j^ is a user-supplied coefficient.


63

The default value for Ci,,i„ is 0.001. The quantity C^ i„ has units of 1/m
and there is no default value in the code; in absence of other information, the
user may estimate this quantity from

C..i„ = C,°''V(0.17 5 D^K) , (7.17)

where « Is the von Karman constant and D^ is the hydraulic diameter at the inlet.
The value of it is set internally by the code to .23/C^ " for use in other parts
of the turbulence model and is approximately 0.42.

The final type of boundary is a no-slip rigid wall. This type of boundary
Is Identified by having a specified velocity boundary condition with u„o - 0 and
Invokes the most complex of the turbulence boundary conditions. In the immediate
vicinity of this type of boundary, there are large spatial gradients in the
velocity and the turbulence quantities. A fine mesh structure would be required
to explicitly represent the details of these steep gradients. To avoid this, a
wall-function treatment Is used which accounts for the effects of the gradients
on an Integrated basis. This procedure affects the two turbulence transport
equations as well as the momentum transport equations.

The wall function treatment is based on a definition of the wall shear


stress
t„=p|u.|u. , (71«)

where U. is an equivalent friction velocity. Define a dimenslonless velocity


and spatial position using

fl = u-/u. (7.19)

and

X,/[H/(P u. )] , (^-2(5)
p

where point p is at the center of the cell which is adjacent to the wall, Up is
the velocity parallel to the wall at point p, and Xp is the distance from point
p to the wall. The corresponding Reynolds number at point p is given by

Re = P^""- = m . (7-21)
p ^l

The dimenslonless velocity is assumed to be a function of the dimenslonless


position, having a forra corresponding to the theory of the "law of the wall :
a ( X ) = A: [1 + f>i(X) K X] / K , (7.22a)

where the velocity coefficient function is given by

f ,,, _ 1 + (0.44 5t)'-°^^ E/K (7.23)


"' " 1 + (0.44 x ) ' ° "
64

and the constant E is set internally by the code using

E = exp (20 C ° - " K) / (20 C ° " ) (7.24)

and has a value of approximately 9.0. The general forms used for Eqs. 7.22a and
7.23 have been chosen since they are asymptotically equivalent to the classical
near-wall velocity profiles for flows at both very low (laminar) and very high
(turbulent) Reynolds numbers; i.e., the profile is linear for low flow
u(x) - x as x - 0 , (7.22b)

and the profile is logarithmic for high flow


a(X) - to(Ex)/K as x - <» . (7.22c)

The specific constants In Eq. 7.23 (coefficient 0.44 and exponent 6.055) have
been chosen to provide a blending between these two asymptotic behaviors which,
when used in COMMIX, results in friction factors which fit the Moody curve (Ref.
32) as well as possible without destabilizing the numerics of the calculation.

The calculation proceeds as follows. For a given point p, Up and Xp are


known. Use these to calculate Rep using the left part of Eq. 7.21. Substitute
Eq. 7.22a into Eq. 7.21 and solve (numerically) for X. Then a, U. , and r, may
be computed from Eqs. 7.21, 7.19, and 7.18, respectively. The diffusive term in
the turbulent kinetic energy transport equation is zeroed for this cell face and
the source term due to mean shear is replaced with
au„
Pk = X „ - ^ ' (^-25)

where the derivative is found by differentiation of a rearranged forra of


Eq. 7.19. One can infer that this is the limiting form for Eq. 7.5 near the
wall, where the velocity gradient in the direction perpendicular to the wall is
much larger than the gradients in the two directions parallel to the wall and the
wall shear (T„) is /JdUp/aXp. Since this source term is not directly a function
of k, it becomes (S...^)^ in the finite-difference formulation.

The no-slip rigid-wall boundary condition for the turbulent kinetic energy
dissipation rate equation is imposed by setting the value for the wall cell to

Cp = C;-" k ^ - V (KXp) (7.26)

rather than solving the transport equation at node p.

Activation of the turbulence model also affects the boundary condition


treatment in theraoraentumequation. This specifically affects the treatment of
diffusion for the four faces of the momentum control volume which are parallel
to the direction of the momentum control volume (e.g. , the south, north, bottom,
and top faces for an i-direction momentum control volume using the notation
presented in Sec. 3.3). (The treatment for the other two faces (e.g., the west
and east faces of an i-direction momentum control volume) is unaffected ) If one
of these four faces is part of a zero-slip boundary surface, then the normal
65

diffusion term for that face (e.g. Dj"(U3-u„) for the south face of an i-
directlon raoraentum control volume) is dropped from the equation and the effect
of the wall shear is Incorporated into the equation as part of the source term.
The addition to the source term is in the forra of a viscous shear force per unit
volume, and these terms are identified as V,, Vy, and V^ in Table 2.1. Each such
face contributes a term which is the wall shear stress ( t, as given by Eq. 7.18
with Up being the velocity component at the center of the momentum control volume
and Xp being the distance from that point to the wall) multiplied by the area of
the cell face and divided by the volume of the momentum control volume. Since
this source terra for each coordinate direction involves the velocity component
in that direction, the finite-difference form of the momentum equation source
term (Eq. 4.24) will contain the term -V,/u in (Sp„)o, -Vy/v in (Sp^)o, and -V,/w
in (Sp„)o rather than these contributions being contained in (Sj^)o.

Activation of the turbulence raodel also affects the boundary conditions


for the therraal energy equation. Conduction heat transfer between a boundary and
the adjacent cell will normally be calculated using the effective thermal
conductivity given by Eq. 7.2; this is independent of the type of velocity
boundary condition. In the special case of boundary normal velocity being zero
and the fluid Prandtl number (Pr - fif C^ / \,) being greater than 0.5, then the
effective fluid conductivity will be taken from

_ i-p P >-ii "-p .^p (7 27)


" " 5 n ( E X p p C ° " k ° - V n , ) o , / K + f,

where the coefficient function is given by


f, = 9.24 (Pr/o, -l)(o,/Pr)°" . (7.28)
If a heat transfer coefficient correlation is input for a boundary, this will be
used instead of conduction through the fluid.

There is limited flexibility for specifying initial values of turbulence


quantities. The turbulent kinetic energy is initialized using a variant of
Eq. 7.15:
TT2 +
ko = CK,in(Uo * ^ + ^\
VS + , (7.29)

where U , Vg , and w^ are cell-center averages of the initial velocity


components in the three directions. The initial turbulent kinetic energy
dissipation rate is then set using a variant of Eq. 7.16
«. - r V^-^ (7.30)
So = C, in Ko .
These initial values are only set by the code at the beginning of a steady-state
calculation and, as such, only serve as guesses; subsequent continuations of a
calculation in either steady-state or transient mode obtain their initial
conditions frora the Restart file (cf. discussion in Sec. 10.2 for the other fluid
equations) .

7.4 One-Equation (k) Model

In this raodel the transport equation for turbulent kinetic energy is


solved as described in Section 7.3. Rather than solve a transport equation for
66

turbulent kinetic energy dissipation rate, this quantity is computed from the
turbulent kinetic energy
e = C ° - " k ^ - V (KC) (7.31)

for each mesh point. The length scale, t, is the sraaller of (a) the distance
from the raesh point to the closest no-slip rigid-wall boundary and (b) 0.175 D^.
The hydraulic diameter, D^, is input by the user and is currently constant
spatially.

After k and c are computed, the turbulent viscosity and turbulent thermal
conductivity are computed using Eqs. 7.12 and 7.14, respectively. Use of the
one-equation turbulence model invokes the special treatments for zero-slip
boundary surfaces in the momentum and energy equations as described in
Sec. 7.3.2.

This model provides turbulence quantities in less computational time, but


with less accuracy, than the two-equation k-c model described in Sec. 7.3, since
the turbulent kinetic energy dissipation rate is obtained frora an analytical
expression rather than by solution of a system of conservation equations. This
treatment of turbulence was originally implemented in code version COMMIX-lA
(Ref. 3) as an approximation for interim use prior to the developemnt and
implementation of the k-c model. Its use raay be justified for some probleras when
the entire flow field is turbulent and recirculation is minimal; however, a safer
approach is to use the k-c model, since it provides a more complete treatment of
turbulence modeling.

7.5 Zero-Equation Mixing-Length Model

In this model, the transport equations for turbulent kinetic energy and
turbulent kinetic energy dissipation rate are not solved. Instead, the turbulent
viscosity is computed from the velocity gradients using

p (Kt)'
aun / au. au. (7.32)
dx. I ax, dx.

where the characteristic length, t, has the sarae definition as in Section 7.4 and
the expanded form of the summation-notation terra in square brackets is given in
Table 7.1. The turbulent thermal conductivity is calculated from Eq. 7.14. The
special treatments for zero-slip boundary surfaces in the momentum and energy
equations as described in Sec. 7.3.2 are not used.

This model provides turbulence quantities in less computational time, but


with less accuracy, than either the k-c (Sec. 7.3) or k (Sec. 7.4) models, since
both the turbulent viscosity and the turbulent thermal conductivitiy are obtained
directly frora analytical expressions rather than indirectly by solving the
systems of conservation equations for the turbulent kinetic energy and the
turbulent kinetic energy dissipation rate. This treatment of turbulence was
originally implemented in code version COMMIX-lA (Ref. 3) as an approximation for
Interim use prior to the development and implementation of the k and k-c models.
Its use may be justified for some problems when the entire flow field is
turbulent and recirculation is minimal; however, a safer approach is to use the
k-c raodel, since it provides a more complete treatment of turbulence modeling.
67

7.6 Constant Turbulent Viscosity and Conductivity

This is the simplest turbulence model. The user inputs constants HK, and
X^o which become the turbulent viscosity and thermal conductivity, respectively,
for all spatial locations. The default for both constants is zero, which
corresponds to no turbulence effects. The user has the option to override the
base values given by /i^o and ^to on a cell-by-cell basis with other input values.
The special treatments for zero-slip boundary surfaces in the momentum and energy
equations as described in Sec. 7.3.2 are not used.

The turbulent viscosity and thermal conductivity should be based on


experiment data. In absence of these data, the following equations (Ref. 29) may
be used. The turbulent viscosity raay be estiraated using
^to = 0.007 C^o p u ^ { , (7-33)

where

(7.34)
Cjio

with u„„ - max(u,v,w), Re„„ - max(Re,,Rey,Re,) , and t - 0.4 D^. The turbulent
conductivity may be estiraated using
1 ^ ^P^*to ^ (7.35)
" ' 0 . 8 / [l - e x p (-6*10-= Re,H Pr^'h'")]

where Re^h and Pr.j, are characteristic Reynolds and Prandtl numbers, respectively.
The user must input /!„; the user raay input X^o directly or by inputing Re.^ and
a characteristic temperature, T,h, at which to evaluate Pr,h and the code will
calculate X^o internally.

This treatment of turbulence was implemented in the original COMMIX-1 code


version (Ref. 1) as an approximation for interira use prior to the development of
the other turbulence models described in this section. The code will execute the
fastest when using this turbulence treatment, since the turbulent viscosity and
thermal conductivity are provided by the user rather than being computed by the
code at every tirae step. There are limited uses for this raodel unless turbulence
effects are to be Ignored (in which case the user inputs values of zero for Mto
and Xto); when turbulence effects are expected to be iraportant, one of the other
turbulence treatments described in this section should be used.
68

8. PUMP MODELING

8.1 Introduction

The use of pumps in a COMMIX calculation is entirely optional. Inclusion


of models for pumps was driven by the need to analyze certain situations relative
to liquid-metal-cooled nuclear reactors, e.g., the thermal hydraulic behavior
following loss of power to the pump motors. In fact, the use of a pump raodel is
one of only two ways in which the therraal hydraulic behavior of a closed-loop
systera raay be analyzed in a forced convection mode using COMMIX; the other method
is described under Boundary Conditions in Sec. 10.1.3.3.

Several modeling options have been provided: homologous, specified speed,


and specified pressure. These models are described in Sees. 8.2 through 8.4.
A certain amount of user flexibility has been provided within the framework of
each of the models. The primary end result from any of the several pump models
is a pressure increase supplied by the pump; the way in which this pressure
increase affects the rest of the code is described in Sec. 8.5,

8.2 Homologous Pump Model

There are generally four variables of interest for a pump: the head
supplied, H (m) , the volumetric flow rate, Q (m'^/s) , the shaft torque, T (N'm),
and the speed, N (revolutions/s) . At steady state, any two of the four
quantities may be considered independent; these will be the flow rate and speed
in the present analysis; the other two variables are taken from the pump
characteristic curves, which are a set of cross plots of the four variables
against one another. In the strictest sense, the pump characteristic curves must
be known for the specific pump to be modeled.

Homologous pump theory provides a functional means by which to estimate


the characteristics of a specific pump for which the detailed characteristic
curves are not available. The theory is based upon the fact that the
characteristic curves for different pumps have similar shapes if the pumps are
geometrically and dynamically similar. The similarity condition is satisfied
when the specific speeds (N-Q°^/H°") of the pumps are similar. This allows data
taken for one specific pump to be used to characterize a class of similar pumps.

The development is aided by defining normalized values for flow rate, pump
speed, head, and motor torque as

Q = 0/QR , (8.1)
N = N/NR , (8.2)
H = H/HR , (8.3)
and
^ = T/TR , (8.4)

where the subscript R indicates a reference state (typically, the "rated"


conditions). An early approach (Ref. 33) then stated the homologous conditions
as

(H/N')I = (H/N')2 (8.5)


69

and

(Q/N)i =(Q/N)2 - ^^-^^

or these ratios are equal for two pumps (indicated by subscripts 1 and 2) which
are geometrically similar and are being operated in a dynamically similar manner.
This particular form of the homologous conditions has little practical use when
one needs to allow for very low or zero pump speed.

A later approach (Ref. 34 or 35) overcomes the low speed difficulty by


introduction of the variable

X = 7t + tan-'{Q/^ . ^^-^^

By utilizing the signs of both 0 and N the definition of X given by Eq. 8.7
covers all four quadrants of the 0-N plane as shown in Table 8.1. The pump
characteristic curves are assumed to be representable using the dimenslonless
forms

(8.8)
H / ( N ^ + Q ' ) = W H (X)
and

(8.9)
T/{N'+Q')=WT(X) .

Table 8.1 Quadrant Assignments for X (Eq. 8.7)

Operating Mode

<0 <0 0<X<ir/2 reverse speed


<0 >0 Tr/2<X<n dissipation
>0 >0 rr<X<3ir/2 normal
>0 <0 37r/2<X<2ir reverse speed dissipation

The above description works for steady state but raust be combined with
other equations for use in analysis of transient behavior. The equation
governing the change in rotational speed of a mechanical pump is based on the
Change in angular momentum being proportional to the sum of the applied torques.
This may be expressed as

2 It Ip - ^ - ^m^^h ^f '
70

where Ip is the moment of inertia of the rotating elements of the pump (kg.m^),
N is the rotational speed (revolutions/s), and the 2-n factor converts from
revolutions to radians. The r terms are the various torques (Nm) : T„ is the
torque supplied by the motor, t,, is the hydraulic torque due to motion of the
fluid against the impeller, and T, is the total frictional torque which includes
losses to the shaft seal, thrust radial bearings, motor windage, and shaft to
fluid viscous losses. (At steady state, the torque supplied by te motor exactly
balances the sum of the hydraulic and friction torques, resulting in the pump
operating at constant speed.) The pump characteristics, Eqs. 8.8 and 8.9, are
assumed to be valid for each operating point during a transient.

The following computational procedure is used in COMMIX to implement the


homologous pump raodel. Knowing the pump speed and the volumetric flow rate
through the pump, the normalized values of these two quantities are computed from
Eqs. 8.2 and 8.1, respectively, and the quantity X may then be computed from
Eq. 8.7. The hydraulic torque is calculated from the characteristic curves using
Eq. 8.9:

l^h = T:hR(N'+0')*WT(X) , (8-11)

where T^,, is an input reference hydraulic torque. Similarly, the frictional


torque is assumed to be given by

Tf = ThR'*WF (N) , . (8.12)

where WF (N) is the relative friction torque as a function of relative pump speed
from the pump characteristic curves. The motor torque is assumed to be given by

X. = t ^ * f n f ( t ) , (8.13)

where T„R is a user-supplied reference motor torque and fnf(t) is the nf-th
transient function supplied by the user. Knowing the pump speed and flow rate,
the pump head is calculated from the characteristic curves using Eq. 8.8:

(8.14)

where HR is a reference pump head. The pressure increase to be supplied to the


fluid is then calculated from the head using

APp = p g H , (8.15)

where p is the fluid density and g is the acceleration due to gravity.

During a transient, the pump speed is advanced across the time step, from
t"-^ to t", by direct integration of Eq. 8.10, which gives

N " = N - i +(T„-Tn-Tf)(t'>-t"-i)/(27tIp) , (8.15)

where the three torques come from Eqs. 8.13, 8.11, and 8.12.
71

The user must provide the pump inertia (Ip), the reference quantities (QR,
HR, NR, TM, , and x^ ), and the various pump functions (WT(X) , WH(X) , WF(S) ,
and f(t)). The functions are assumed to have the following polynomial forms:

(8.17)
WH (X) = g C,,iX-^
9
(8.18)
vn:(x) = J2 c,,iX'-^
• 1

3
Y, Cf,i N^-^ for N^Cf,,„t,i (8.19a)
i=l
3
(8.19b)
WF(N) = { E <=f.i'3 N'"' for Cj,^^t,i<NsCj,<,„t,2
i-l

E ^£.i-6 N'"' for Cj,,,t.2<N (8.19c)

where the Cj are constants input by the user. The transient function is input
as a series of data pairs.

A suboption within the homologous pump model prevents reverse rotation,


in order to simulate a locked rotor situation. If this option is selected, then
a different set of coefficients (C,,i.,) is used in Eq. 8.18, and when the speed
of the pump is calculated to be negative it is set to zero.

The above formulation results in the pressure increase supplied by the


pump being proportional to the square of the flow rate at low pump speed and
flow. An option in the impleraentation forces this relationship to become linear
at low pump speed and flow, to simulate the expected laminar behavior. On the
first time step during a transient in which the relative pump speed (N) and the
relative pump flow rate (Q) both decrease to below user-specified cutoff values
( N and 0^^^ ), then the current values of the relative head and relative flow
rate^'are saved as H,,,, and Q,„, , respectively. During the remainder of the
transient, the pump head will be calculated frora

(8.20)
st/Ot

Instead of using Eq. 8 .14 as long as N < N^„t and S < O^ut-

More specific horaologous pumpraodelsmay be coded directly in the code by


the user. One specific case has already been included. Further details are in
the User's Guide (Ref. 9).

8,3 Specified Pump Speed Model

The pump speed, N (revolutions/s), in this model is specified by the


user as
(8.21)
N = NR * f„f (t)
72

where NR is a reference speed and f„£(t) is the nf-th transient function, both
specified by the user.

Having the pump speed and the current volumetric flow rate through the
pump, Q (raVs), the pressure increase due to the pump in this model is given by

APp = a ( N ^ + b N O + c |0|^) , (^•22)

where t h e normalized speed and normalized v o l u m e t r i c flow r a t e for the pump are
d e f i n e d by Eqs. 8.2 and 8 . 1 , r e s p e c t i v e l y , and QR i s a u s e r - s p e c i f i e d r e f e r e n c e
flow r a t e . The c o n s t a n t s a, b , c , and d a r e a l l s p e c i f i e d by t h e u s e r . The u s e r
may i n p u t a f l o w - r a t e dependent v a l u e for d, as

d, for Re i RScrl (8.23a)


'^ ^ ^ d^ for Re > Re„j ' <8-23t,'

where the Reynolds number i s d e f i n e d by

Re=p|u|L/n, (8-2M

and u is the fluid velocity in the pump, L is a characteristic length for the
pump, and /Xf is the molecular fluid viscosity. The quantities df, d^, Retr, and
L are all input by the user.

8.4 Specified Pressure Increase Model

The pressure increase supplied by the pump in this model is given by

A P p = APp,,, * f„j(t) , (8-25)

' p, as
model in Sec. 8.2 or 8.3 and fn£(t) is the nf-th user-supplied transient
function.

8.5 Interface with Rest of Code

There may be several pumps (up to 5) in a calculation. All pumps in a


single calculation must use the same type raodel (i.e. as given by Sec. 8.2, 8.3,
or 8.4). Each pump raay have different reference values for speed, flow rate,
head, and torque and different transient functions. Limited provision is made
for the use of different characteristic curves for the different pumps.

Each pump must be aligned with one of the three fluid coordinate axes and
is associated with a particular momentum control volume. The faces of the
momentum control volume containing a pump should be blocked to fluid flow in the
two directions which are transverse to the alignment of the pump. There may not
be more than one pump in a momentum control volume.
73

The pump calculation is performed as a part of the fluid momentum


calculation. A separate routine is called to calculate pump quantities if a
momentum control volume is found to contain a pump during setup of the momentum
equations for an outer Iteration. In general, the pump routine is provided with
the current volumetric flow rate based on the current value of the velocity times
the flow area. The pump routine, depending on the pump model, must calculate
torques, advance the pump speed across the tirae step, and return a value for the
pressure Increase supplied by the pump. The pressure increase, APp (as
calculated from Eq. 8.15, 8.22, or 8.25), is then divided by the length of the
appropriate momentum control volume to obtain (AP/L)p, which is a source term in
the momentum equation (cf. Table 2.1) in the direction corresponding to the pump
alignment; this part of the source terra is added to (S„^)o in Eq. 4.24 when
finite differencing the raoraentum equation. This pressure increase is applied,
essentially, at the center of the momentum control volume.
74

9. SUPPLEMENTARY PHYSICAL MODELS

A number of supplementary physical models have been Incorporated in COMMIX


to account for phenomena that affect therraal-hydraulic simulations. A number of
these models require the user to provide parameters, which leaves flexibility.
Discussed here are physical properties for fluids and structures, heat transfer
correlations, and force structure modeling and correlations.

9.1 Fluid Properties

The present version allows multiple fluids to be present in the


computational domain. (Theraaxiraumnumber of fluids allowed is set at compile
time by the value of PARAMETER PIFHTX; see User's Guide (Ref. 9).) The region
containing one fluid raust be physically separated from regions containing other
fluids; the separation is accoraplished either (a) by completely enclosing each
fluid region with a set of boundary surfaces or (b) by specifying a directional
surface porosity of zero for each cell face which separates two adjacent cells
containing different fluids. (The former method is preferred and must be used
when one of the nonconstant turbulence models is used.) There is no provision
for more than one fluid to be present in the same raesh cell. Each fluid is
assigned an identification number, starting with 1.

The fluid transport equations contain numerous state and transport


properties. Only two of the state properties raay be considered independent; in
COMMIX these are the enthalpy, h, and the pressure, P. An equation of state must
be provided in order to calculate other dependent state properties (e.g. density,
p , and temperature, T) frora the independent state properties. The values of the
laminar (molecular) transport properties (e.g. viscosity, /if, and thermal
conductivity, Xf) must also be provided. The current formulation also requires
certain derivatives of properties (e.g. derivative of enthalpy with respect to
teraperature, Cp, and derivative of density with respect to enthalpy) and the
saturation pressure, Ps.f (The saturation pressure is used within the raain part
of COMMIX only in connection with an option which checks to see whether
saturation conditions have been achieved in Fluid #1.) This fluid property
information must be provided by the user.

There are two ways to describe the physical properties of fluids: detailed
and simplified. The user selects whether detailed or simplified properties are
to be used for Fluid #1. All other fluids being used in a calculation must use
simplified properties.

The detailed fluid properties are specified in a set of FORTRAN FUNCTIONS


containing fits to the available physical property data. These routines must
conform to the following functional dependencies:

p = p(h or T , P) , (9.1)

T = T(h, P) , (9.2)

(1, = n, (h or T, P) , ^ (9.3)

X, = 1, (h or T) , (9.4)
75

C„
"p
= C-p (h or T,P) , <'-5)

d£ = d£ (^^^ T pj ^ (9.6)
on cm
and
(T) (9.7)
The terminology "h or T" means that both enthalpy and temperature appear as
arguments; the correlation may be based on either one. Additionally the inverse
of the functional relationship given by Eq. 9.2, i.e.

h = h (T, P) , ('•*)
must be available. Detailed property packages have been prepared for liquid
sodium, liquid normal water, liquid heavy water, and helium gas; the specific
functional forms used in each of these fluid packages are given in the User's
Guide (Ref. 9 ) . The user may use these as a guide for creating detailed fluid
property packages for other fluids. In any event, the user may include only one
detailed fluid property package when forming an executable code version.

The simplified fluid properties are provided by a set of STATEMENT


FUNCTIONS having the following forms:
h = C„H„ + C i ^ T , <9-')

l:
C + C T , or (9.10a)
P = \ p / [(8314/C2P,) (T + 273.15)] , (9.10b)

X - C + C T ('-^^^
and
u - C + C T , ('-^2)
where the constant coefficients Cij„ are input by the user for each fluid number,
m, which is to use simplified properties. The choice of density functions is
made based on the value of C2,„, which corresponds to the fluid's molecular
weight- if C2^-0, the linear forra given by Eq. 9.10a is used; otherwise, the
Ideal-gas-law forra given by Eq. 9.10b is used. The input constants are used to
generate a T(h) function and to compute the various derivatives needed.

9.2 Structure Properties


In many real applications, the presence of solid boundaries and immersed
solid objects affects the thermal behavior of the fluid. Models for the thermal
effects of these structures are included in the code and are described in
Sections 6 and 10.1.3. Each structure is composed of one or more solid
materials The user must assign an index number,ra,to each solid raaterial being
used in a calculation. (The number of different materials is currently limited
to 20.) The same material may be used in more than one structure. The structure
models require the following physical properties for each material: density, p ,
specific heat capacity, Cp, and therraal conductivity, X.
76

The physical properties of solid materials are nominally assumed to have


the following forms:
r2 (9.13)
P = C Q P ^ + Cjp^T + C2p„T

C T2 (9.14)

and
^ = Cox. + C u . T + C2i,T^ , (5-15)
where the constant coefficients C^p^ are input by the user for each solid
material number,ra,which is being used. The specific heat capacity and therraal
conductivity are actually computed by FUNCTION subprograms which use these
coefficients; the user raay modify these routines to specify different functional
forms for materials whose properties are not well described by the basic
quadratic form.

9.3 Heat Transfer Coefficient Correlations

The heat transfer rate between a fluid cell and the associated surface of
a solid (either a boundary or an immersed object) is generally computed using
Q = U„A(T3-T) , (9.16)
where U„ is a heat transfer coefficient, A is the interfacial surface area, T^
is the temperature of the solid at the surface, and T is the fluid temperature.
The user must input all heat transfer coefficients to be used in a calculation;
each is assigned an index number, n. Then, for each fluid-solid thermal
interaction, the user specifies which heat transfer coefficient, by index n, is
to be used at the surface. The same heat transfer coefficient index may be used
on more than one surface.

The heat transfer coefficients are nominally computed frora


Un = (^./Dh)/NUn , (9.17)
where Dj, is a characteristic length input by the user for each surface. The term
NUn is a Nusselt-number correlation, having the form

Nu = C + C Re^^^" Pr ^*''" ('i'^^)


The fluid Reynolds number

Re = p D j I ? + v ^ + w Y V n , (^-1^'
uses a fluid speed based on the spatial averages of the three velocity
components, u, v, and w. The fluid Prandtl number in Eq. 9.18 is given by
Pr = \i, Cp/X, (9.20)
and the constant coefficients Ci^n are input by the user for each of the n
correlations to be used. (The user may input up to 20 such correlations.)

If the heat transfer coefficient can not be adequately represented by the


above general forra, the user may supply the coding for (up to 49) additional
specific forms. Further details are provided in the Users' Guide (Ref. 9 ) .
77

9.4 Fluid-Structure Momentum Interactions

The presence of solid structures in or near the fluid domain influences


the fluid flow by providing increased flow resistance. These effects are
incorporated in COMMIX using the distributed flow resistance terms, Rj, in the
momentum equations shown in Table 2.1. Specification of nonzero values for these
terms is not automatic, except for the wire-wrap and pin effects under the
hex-geometry option (cf. Appendix A ) . (The effects of zero-slip rigid-wall
boundaries under one of the nonconstant turbulence models enters the momentum
equations through V,, as discussed in Sec. 7.) Instead, the user is left with
the flexibility (and responsibility) to specify which cells have distributed
resistance(s), the direction of the resistance, and the parameters describing the
resistance. Each momentum control volume may have only one such user-specified
distributed resistance.

The distributed resistance in COMMIX is a pressure drop per unit length


in the coordinate direction j having the following general form
Rj = Ctn f„(l/Dni„)p|Uj|u3 , ^'-21)
where C,„ is a dimenslonless constant, \,„ is a characteristic length, and f. is
a friction correlation; Ct„, D^,„, and the friction correlation index m are input
by the user for each force structure number n. Since the distributed resistance
for each coordinate direction involves the velocity coraponent in that direction
the finite-difference forra of theraoraentumequation source term (Eq. 4.24) will
contain the terra -R,/u In (S,J,. -Ry/^ in (Sp.)o, and -R,/w in (Sp„)o, "here R is
Riven by Eq. 9.21 for j - x, y, and z. (The user may define parameters for up
to 99 force structure numbers (n); each of these force structure numbers may then
be used in one or more momentum control volumes.)

The friction correlation is normally computed from the general forms


C (9.22)
CaP,„Re *="" + C,p,„ , for Res Ret
and
C (9.23)
C.pt™ Re"-"^" ^ C,pt. . for Re>Ret„ ,
where
I 1/ ('-24)
R e = p Dh2n|Uj|/H,
and the six constant dimenslonless coefficients C,P,„ and transition Reynolds
n l e r R e i must be input by the user for each of the m ^.^.^"-^ " ^ ; ; ^ ^ / - ° ^
functions to be used. (Up to 20 such functions raay be specifled.) The "ser must
also input the characteristic length D,2„ to be used in the Reynolds number for
each of the n distributed resistances.

genei
an
beer
in
Guide
user may .» j — - ,^ ..v.
functions, assigning unused correlation index numbers to thera.
78

The user must input which cell's momentum equations are to use a specific
one of the n force structure correlations. The resulting distributed flow
resistance is applied over the length of the corresponding momentum control
volume, which is offset by a half a cell from the main control volume.
79

10. BOUNDARY AND INITIAL CONDITIONS

Boundary conditions must be specified in order to solve any system of


differential or finite-difference equations. Initial conditions must also be
specified for equations to be solved in a tirae dependent raanner. This section
describes the boundary and initial conditions available in COMMIX for the finite-
difference forras of the fluid raass, momentum, and energy conservation equations.
The similar conditions for the structure energy equations and for the fluid
turbulence equations were presented along with the models in Sees. 6 and 7,
respectively.

10.1 Boundary Conditions

A physical fluid systera to be modeled in COMMIX must have the collection


of fluid cells completely enclosed by a set of rigid nonmoving walls. (There is
no provision for modeling a true free surface.) These walls may have openings
which provide paths through which fluid can enter and leave the volume. The
impervious parts of the walls may be energy transfer paths. If symraetry
arguments may be invoked to avoid raodeiing the full georaetry, then the cuts
perforraed to produce a sector georaetry are syraraetry boundaries. Each of these
walls, openings, and symmetry cuts is identified as a "boundary surface".

There are two types of boundary surfaces: "regular" and "irregular".


A regular boundary surface has its surface normal vector (which always points
toward the fluid) aligned with one of the three fluid coordinate grid directions
(I.e. X (or r ) , y (or 9), or z ) ; usually, most boundary surfaces are regular.
An Irregular boundary surface is not restricted to having its surface normal
vector aligned with one of the fluid grid coordinate directions. Using a
hexagonal fuel assembly as an example (cf. Appendix A ) , four of the six assembly
walls are Irregular boundary surfaces; the other two walls and the inlet and
outlet planes are regular boundary surfaces.

A boundary surface is composed of "boundary surface elements", which are


the faces of the fluid cells which are adjacent to the boundary surface and
toward which the boundary surface normal points. Boundary surface elements are,
likewise regular and irregular depending on whether they are part of a regular
or irregular boundary surface. Not all cells faces lie on boundary surfaces;
a cell face may be part of, at most, one boundary surface. Boundary conditions
are specified on these boundary surfaces.

Each boundary surface may have a different combination of boundary


conditions- however, all elements of a single boundary surface raust have the sarae
combination of velocity, pressure, and energy boundary conditions. This means
that a given physical surface may need to be divided into several boundary
surfaces in using COMMIX. (Example: A wall with an opening raust be divided into
at least two boundary surfaces: one for the opening and one (or more) for the
impervious part.) Some physically based rules govern how the various boundary
condition types may be used to surround the fluid. (Examples: if one boundary
surface is an inlet, then (at least) one boundary surface must be an outlet; one
boundary surface must provide a teraperature reference value; one boundary surface
must provide a pressure reference value.) Care raust be taken that the combination
of specified boundary conditions is sufficient (in a numerical sense) for a
unique solution to be obtained. (Example: Teraperature can not be uniquely
deterrained if all energy boundary conditions are of the heat flux type.) The
structure of COMMIX also places some restrictions on how boundary conditions may
80

be combined. (Example: A speclfied-pressure-type pressure boundary condition


may only be used on boundary surfaces which also use a continuative-type velocity
boundary condition.)

The remaining subsections present the forms allowed for the velocity and
pressure boundary conditions, which are associated with the momentum equation,
and for the energy equation boundary conditions. The User's Guide (Ref. 9)
explains how to define boundary surfaces and selectively specify the desired
boundary conditions.

10.1.1 Velocity

Several types of velocity boundary conditions have been provided to


describe the typical physical boundaries used to enclose fluid systems. In all
cases, what is being specified (either directly in value or indirectly by a
relational formula) is the "normal velocity", u^, which is the velocity at the
boundary surface in the direction of the surface normal vector.

The specific details of the various velocity boundary conditions are


discussed in the following subsections.

10.1.1.1 Specified Constant Velocity

The normal velocity is constant in tirae for this boundary condition.


The value, u^o, is specified by the user, and there may be a spatial distribution
of values across the boundary surface. Once these velocity values are
initialized by the user, they are never changed by the code.

This boundary condition is useful for an inlet where the velocity is


constant in tirae (i.e. U^Q = nonzero constant) and for a rigid wall with no slip
(i.e. u^o - 0 ) . Selection of a constant velocity boundary condition with U^Q "
0 activates a special shear term in the momentum equations if one of the
nonconstant turbulence model options has been selected (cf. Sec. 7.3.2).

10.1.1.2 Specified Transient Velocity

The normal velocity is a function of time for this boundary condition,


as given by

Un = Un„ • fn£(t) , (10.1)


where u„o is a user-supplied reference value and fnf(t) is the nf-th transient
function supplied by the user. The velocity is spatially uniform across the
surface. This boundary condition is useful for an inlet where the velocity is
a known function of time.

An alternate form of this boundary condition allows the normal velocity


to be a function of the temperature of a radiation surface and is given by

Up = ^no * fnfCTi.nf) - (10.2)


where the "transient" function has temperature (Instead of tirae) as the
independent variable and T^ „f is the temperature (in degrees Celsius) of the
nf-th radiation surface. This boundary condition is useful for simulating
natural convection cooling of a surface when the complete hydraulics of the flow
path are not modeled.
81

10.1.1.3 Free Slip


The normal velocity and the shear stress at the surface are both zero
for this boundary condition. This boundary condition is useful for a symmetry
boundary. When the cylindrical geometry option is used, the z-axis (at r-0) is
a boundary surface having zero surface area and should be identified by the user
as a symmetry boundary.

10.1.1.4 Continuative Mass Flow


The normal velocity is set to provide a conservation ofraassflow rate
across the cell adjacent to the boundary surface in the direction of the surface
norraal vector.
The notation needed to describe this boundary condition depends on
whether the boundary surface is on therainusor plus side of the cell. For the
two outlet boundary surfaces shown in Fig. 10.1, the continuative mass flow
boundary condition may be expressed as

-(A.UiW2<P>"-^t^|f)„ (10.3a)
^n,m*l/2 /A n\
i^i P)m.l/2
for the boundary associated with cell m and

(A.-i)..i.<P>-t^l?), (10.3b)
""-•-^^^ ' (Af P),_i/2
for the boundary associated with cell (, where p is the fluid density. A, is
fluid flow area for a cell face, V, is the fluid volume of a cell and u, is the
fluid velocity in the coordinate direction 1 which is the same direction as the
surface normal vector. The density, <p>. appearing in the numerator of these
equations is for the upwind cell based on the sign of the velocity at the
indicated cell face, as described in Sec. 4.2. The density m the denominator
is associated with the boundary surface and is defined in terms of boundary
enthalpy and pressure. The arrangement of signs in these equations is diffe"nt
due to the normal velocity (u„) being defined as positive for flow into the fluid
domain.
The condition given by Eq. 10.3 is applied individually to each cell
face which is a part of the boundary surface. The ""^^^^^^''! f "/^^^;; ^^
Eq. 10.3 essentially forces there to also be, on a cell-wise basis for each of
the cells on the boundary surface, zero net mass flow for the sum of the two
directions which are transverse to the surface normal vector.

This boundary condition is the primary boundary condition used for the
outlet of a problem. It may also be used at an inlet when pressure is prescribed
as the inlet boundary condition.
82

Outlet Cutlet
Boundary Boundary

i+l n-1

Inlet

Fig. 10.1. Cells Near Outlet Boundaries

10.1.1.5 Continuative Mass Flux

The norraal velocity is set to provide a conservation of mass flow rate


per unit area across the cell adjacent to the boundary surface in the direction
of the surface normal vector. This boundary condition is implemented in a manner
identical to the continuative mass flow boundary condition except that the two
areas appearing in Eq. 10.3 are assumed to be equal.

This boundary condition has limited practical application unless the two
areas involved are truly equal, in which case the coding will execute slightly
faster than if the continuative mass flow boundary condition were used;
otherwise, raass conservation will be difficult to obtain.

10.1.1.6 Continuative Velocity

The normal velocity is set to provide a conservation of velocity across


the cell adjacent to the boundary surface in the direction of the surface normal
vector. This boundary condition is implemented in a manner identical to the
continuative mass flux boundary condition except that the areas and densities
appearing in Eq. 10.3 are assumed to be constant in space and time.

This boundary condition has liraited practical application unless the


areas and densities involved are truly constant, in which case the coding will
83

execute slightly faster than if the continuative mass flow boundary condition
were used; otherwise,raassconservation will be difficult to obtain.

10.1.1.7 Uniforra Continuative Velocity

The normal velocity is set to provide a uniform velocity across the cell
faces which forra the boundary surface. This boundary condition is implemented
in araanneridentical to the continuative mass flux boundary condition except
that the numerator and denominator of Eq. 10.3 are both replaced with summations
over all surface elements on the boundary surface. The resulting average
velocity is then assigned to all elements of the boundary surface.

This boundary condition should only be used at an exit which has


multiple boundary surface elements where the velocity is known to be spatially
uniform; otherwise, mass conservation will be difficult to obtain. A safer
approach is to use the continuative raass flow boundary condition.

10.1.1.8 Expansion Cell


A fluid region in COMMIX should not, in general, be surrounded
completely by Impervious walls. This is particularly true if one of the detailed
liquid equations of state is used for the fluid. Since the volume available for
fluid is constant, heating of the fluid would lead to a rather large pressure
Increase. Similarly, cooling of the fluid would attempt to create void spaces
as the fluid contracts. Although these are both physically real processes the
pressure increase generally leads to numerical problems in COMMIX and no
provision is normally made for time-dependent void formation in COMMIX.

An early approach to circumventing these problems was to put an opening


on (or near) the top boundary, containing one cell face and applying a
continuative raass flow velocity boundary condition. As the fluid expands
(contracts) under heating (cooling), mass crosses this boundary leaving
(entering) the fluid system. The fluid mass in the system is thu^' " ° ^
constant This is not a good representation of the physical constraint for a
fluid system which is intended to be of constant mass.

The expansion cell boundary condition has been provided as a way to


approximate the boundary constraint of a constant mass fluid system. As in the
'^'^ - . . . _ _ . _ . boundary surface element should be

s
tnis surrace snouiu uc aui=Lj = ^^-. -- - , i ui i j ,-„ i-v,o f 1 nu
designated the "expansion cell." This cell must be totally blocked to the flow
of m!ss and energy in the directions which are transverse ^^;'^^%\l\';ilZT'n
vector and should not interact with any thermal structures. The fl"^d volume in
the expansion cell is allowed to change during a transient calculation, as
described below; all other fluid cells have a constant '^-^''.^°^'^'^.Jll°^''l[
boundary surfaces of the fluid region should have u„ - 0 (with either zero- or
free-slip) to model a constant raass system.
During a transient calculation, this boundary condition provides the
same constraint as provided by the continuative mass flow ''-"^-/ "^^^J^"^
until
itil the
the time
time step
step has
has converged.
converged. The
The code,_
code, therefore,
thererore, _iis
^s_ allowed^to
• ^ - " — ;" believe —-•-
lat mass is actually flowing across the boundary surface element.
that (The
tocity u„ on this boundary surface is, in fact, never set to zero.) Following
ity, u„.
84

convergence, the fluid volume of the expansion cell is adjusted in proportion to


the amount of mass which tried to cross the boundary surface element during the
time step. The volume is increased (decreased) so that the raass which appeared
to leave (enter) the systera can be contained within (removed from) the expansion
cell without a change in expansion cell teraperature and pressure.

Recall that the fluid volume is defined by


(10.4)
where y„ is the volume porosity and V is the cell volume (not just fluid volume)
in terras of the grid coordinates (e.g. Ax * Ay * A z ) . The fluid volume appears
in the finite-difference forms of all fluid conservation equations. Since the
cell volume (V) is constant in time, the change in fluid volume may be expressed
as an equivalent change in the volume porosity. Thus, at the end of the time
step, the volume porosity of the expansion cell (m) is adjusted according to

(pAfUn),(t"-t"-^)
v" = v""^ - '" - - (10 5)
(PV)„
where the subscript b refers to the boundary surface element (i.e. m+1/2 or
m-1/2). The porosity of the cell, thus, increases and decreases as the fluid in
the rest of the region expands and contracts. The porosity of the expansion cell
is allowed to exceed 1 to accommodate the volume change; however, the coding in
COMMIX does not allow for negative values of volume porosity; warnings are issued
and the code will stop if the porosity gets close to 0. (Warnings are issued by
the code when the volume porosity drops below 0.07; the code will stop if the
porosity is below 0.02 or is projected to be below 0.02 at the end of the next
time step.) In performing global balances for edit purposes, the mass and energy
flows across this boundary surface are excluded (since, in actuality, nothing was
ultimately allowed to cross the boundary).

During a steady-state calculation, this boundary condition provides the


same constraint as provided by the continuative raass flow boundary condition --
the volume porosity of the expansion cell is not altered. This, of course, does
not conserve the total fluid mass of the system, but this is not important since
the proper fluid mass is not known until steady state has actually been achieved.

This boundary condition is only meant to handle a moderate degree of


fluid expansion and contraction during a transient. Its use avoids the pressure
and void-formation problems encountered when all walls are impervious and avoids
the mass-conservation problem encountered when an opening is modeled using one
of the continuative boundary conditions. A side benefit of using this over the
continuative mass flow boundary condition is that the system raass and energy
balances printed by the code are a more useful measure of overall convergence
during a transient.

10.1.2 Pressure

The formulation in COMMIX is such that pressure boundary conditions may


only be specified at inlet and outlet surfaces which use one of the continuative
velocity boundary conditions. Using a fuel assembly as an example, the inlet may
have a specified velocity boundary condition and no pressure boundary condition
combined with the outlet having a continuative-type velocity boundary condition
and a specified pressure boundary condition; alternately, both the inlet and the
85

outlet may have a continuative-type velocity boundary condition and a specified


pressure boundary condition; pressure boundary conditions should not be specified
for the other boundary surfaces of the fuel assembly.

The pressure specified on the boundary, P;,, is uniform across the


surface and is actually used to set the cell-center pressure in the fluid cells
which are adjacent to the boundary surface. The boundary surfaces using pressure
boundary conditions should either be (1) a surface containing one surface element
or (2) a surface whose surface normal vector is parallel to the gravity vector
and which is expected to have flow parallel to the surface normal vector, as
shown in Fig. 10.2.

DESIRED PRESSURE BOUNDARY-

« • •
1 1
> 1 1 1 1

Fig. 10.2. Recommend Surface Arrangements for Pressure


Boundary Conditions

Two types of pressure boundary conditions are allowed, as discussed in


the following subsections.

10.1.2.1 Specified Constant Pressure

The pressure is constant in time for this boundary condition. The


value, Pbo, is specified by the user and is used to initialize the pressure in
the cells adjacent to the boundary surface.

10.1.2.2 Specified Transient Pressure

The pressure is a function of time for this boundary condition. The


boundary pressure is calculated from
86

Pb = PbO * fnt (t) , (10-6)


where P^o is a user-supplied reference value and f„£(t) is the nf-th transient
function supplied by the user. This pressure is evaluated at every tirae step and
used to set the pressure In the cells which are adjacent to the boundary surface.

10.1.3 Energy

Several types of energy boundary conditions have been provided to


describe the typical physical constraints imposed on fluid systems. The allowed
conditions fall into two categories: specified temperature, T^, or specified
heat flux, qt. (Temperature is used as a matter of user convenience and is
always converted internally by the code to enthalpy, which is the primary
independent variable in the energy equation.) The heat flux referred to in this
section is the component of the energy transfer rate per unit surface area which
is not due to fluid motion through an opening in the boundary; the energy
transported by fluid motion through an opening is included directly in forming
the convective terms in the finite-difference formulation of the energy equation.
The teraperature or heat flux is specified directly or a method is selected for
indirect specification. The code will calculate the resulting boundary heat flux
if the boundary temperature is specified and vice versa. The sign convention is
such that heat transfer from the boundary to the fluid is positive.

The specific details of the various boundary conditions allowed for the
fluid energy equations are discussed in the following subsections.

10.1.3.1 Specified Constant Teraperature

The boundary temperature is constant in time for this boundary


condition. The value, Tto, is specified by the user, and there may be a spatial
distribution of values across the boundary surface. Once these values are
initialized by the user, the values are never changed by the code.

The boundary temperature is used in conjunction with the temperature


(Tf) in the adjacent fluid cell to calculate a heat flux from the boundary to the
fluid using
Qb = Ut(Tb-T£) , (10.7)
where Uj, is a heat transfer coefficient given by

Uv, (10.8)
Uf I ^ )w
The user has the flexibility to input a correlation number from which the fluid
convection heat transfer coefficient (Uf) is to be calculated (cf. Sec. 9.3); if
a correlation number is not input, then this terra will be computed by the code
based on conduction through a half thickness of the fluid cell

Ut = A.,t/(0.5*Vf/At,) , (10.9)
where X,f is the molecular thermal conductivity of the fluid, Vf is the fluid
volume of the cell, and Aj, is the area of the boundary surface element. In a
similar manner, the user has the flexibility to input a material number for the
wall; if this is nonzero, then the thermal conductivity of the wall, X„, is
evaluated using the structure property correlations (cf. Sec. 9.2) at the
87

boundary temperature and L, is the input wall thickness; otherwise, the solid
conduction term is neglected.
This boundary condition is normally associated with an inlet and is
sometimes useful for an impervious wall. For an inlet, a fluid heat transfer
correlation and a wall material number would not usually be specified.

10.1.3.2 Specified Transient Temperature

The boundary temperature is a function of time for this boundary


condition, as given by
T -T * f (t) (10.10)
Tb - IbO * '^nf ^'^' '
where T.o is a user-supplied reference value and f„j(t) is the nf-th transient
function supplied by the user. The temperature is spatially uniform across the
surface The boundary heat flux will be calculated in the same manner as was
described for a specified constant teraperature boundary condition. This boundary
condition is useful for an inlet or an impervious surface where the temperature
is a known function of tirae.

10.1.3.3 Recirculation Temperature


This boundary condition allows the teraperature at one surface to be set
to the current value of the average temperature at another reference surface,
which is given by

„ i£ ^"^ ^"^ (10.11)


^^ - J: A,i '
ietn
where the summation is perforraed over all surface eleraents (1) of the
user-specified surface number (m) . (Currently, the T,, in this equation are
actually the fluid temperatures in the cell adjacent to the boundary rather than
the actual boundary temperatures.) The average temperature given by Eq_ l O ^ U
is computed at every time step and then used to set '^e teraperature of every
eleraent on the boundary surface. The heat flux for the boundary surface will
be zero.
This boundary condition is sometimes useful at the "inlet" of a
closed-loop fluid geometry, in which case the reference surface ra should be the
"exit" of the systera and should have an adiabatic energy boundary condition.

10.1.3.4 Specified Constant Heat Flux

The heat flux frora the boundary to the fluid is constant in tirae for
this boundary condition. The value, q,o, i^ specified by the user and is
spatially uniform across the boundary surface.

The boundary heat flux is used in conjunction with the temperature (T,)
In the adjacent fluid cell to calculate a boundary teraperature based on
conduction through the fluid, which is given by
/T7 (10.12)
Tb = Tf + qb/Uf .
where Uf is given by Eq. 10.9.
88

This boundary condition is useful for Impervious walls where the heat
flux is known. If a zero heat flux is desired, the adiabatic boundary condition
described in Sec. 10.1.3.6 is generally preferred rather than using the specified
constant heat flux boundary condition with q,, - 0.

10.1.3.5 Specified Transient Heat Flux

The heat flux from the boundary to the fluid is a function of time for
this boundary condition, as given by

Qb = q b o * fnt(t) ' (10-")


where q^o is a user-supplied reference value and f„£(t) is the nf-th transient
function supplied by the user. The heat flux is spatially uniforra across the
surface. The boundary teraperature will be calculated in the sarae manner as was
described for a specified constant heat flux boundary condition. This boundary
condition is useful for an impervious surface where the heat flux is a known
function of time.

10.1.3.6 Adiabatic

The heat flux from the boundary to the fluid is zero for this boundary
condition. The boundary temperature will be set to the teraperature of the
adjacent fluid cell for each boundary surface element. The results obtained
using the adiabatic boundary condition are equivalent to, but are obtained in
less computational time than, the results obtained using a specified heat flux
boundary condition with q^ = 0.

This boundary condition is useful at exit, symmetry, and some irapervious


wall boundaries.

10.1.3.7 Duct Wall

This boundary condition allows the transient therraal response of a


finite-thickness wall to set the energy boundary conditions for the fluid. Each
boundary surface eleraent has its own wall segraent and it treated separately. The
temperature within a wall segment is uniform and a lumped-heat-capacity method
is employed. With the geometry shown in Fig. 10.3, the energy balance for a wall
segment (on a per unit area basis) is given by

(PCpL)„ ^ = Ub(Te-T,) - U,(T,-T,) + q j L„ , (10.14)

where p„ and Cp„ are the density and specific heat capacity of the wall, L„ is
the wall thickness, Tf is the fluid temperature in the adjacent cell, and T^ is
a user-input constant temperature for the surroundings on the other side of the
wall. The heat transfer coefficient on the fluid side (U^) comes from Eq. 10.8
and the coefficient on the other side (U^) is a user-input constant. The heat
source in the wall, if used, is assumed to be in the following form

qC = QIN(i,j) • QK (k) • qJo * fn£ (t) , (10.15)


where QIN(i,j) is the normalized transverse power distribution at location (i,j)
for all axial locations k, QK(k) is the normalized axial power distribution at
axial location k, q"„Q is a reference volumetric heat generation rate, and fnf(t)
is the nf-th transient function; all of these quantities are user input.
89

/ /

4/

Tf
SURRC
/ /

^ Lw
\ / /

Fig. 10.3. Finite-Thickness Wall Boundary

If the fluid temperature and heat source are assumed constant over a
time step,
. , ,then Eq. 10.14 may be integrated across the time step (frora t"" to
lai temperature at the end of the tirae step
t") to give the boundary
(10.16)
= T
(TS"^-T^) * exp [-o(t
where

(10.17)
a^Tj + a,T, + P

(10.18)
ffb = Ub/ (PCpL)^
(10.19)
tts = Us/ (pCpL)^
(10.20)
a = Ob + ag ,
and
(10.21)
P = ql/ (pCp)^ .
All therinophyslcal properties of the wall are evaluated at teraperature T^ using
the correlations in Sec. 9.2 for the user-specified material.
90

After the temperature of the wall has been computed, the heat flux from
the boundary to the fluid is calculated using Eq. 10.7.

This boundary condition is useful for irapervious walls which are thin,
such as the duct wall of a fuel asserably. In raany cases, the wall bounding a
fluid region is sufficiently complex (in terms of composition or boundary
conditions) that it raust be raodeled using the thermal structure treatment (cf.
Sec. 5 ) ; when the thermal structure model is used to simulate a boundary
condition, then the normal fluid energy equation boundary condition should
usually be adiabatic.

10.2 Initial Conditions

Initial values raust be assigned to all fluid variables before a


calculation may be started; these include the state variables (enthalpy,
pressure, temperature, and density) and the three components of velocity. The
transport properties (specific heat capacity, thermal conductivity, and
viscosity) are computed from the state variables as needed in the calculation.
These initial values are merely guesses to start a steady-state calculation; a
good set of initial values will shorten the time required by the code to find
steady state. On the other hand, these values should be true initial conditions
at the beginning of a transient calculation; these values should generally be the
result of a steady-state COMMIX calculation. The way in which these values are
input, as well as the values theraselves, depends on the type of calculation being
performed.

A total calculational history for a simulation with COMMIX usually


requires a sequence of computer runs. At the end of each run, a "restart file"
is written. This file contains sufficient information to allow restoration of
all calculational variables to their prior values and is input to the next run
in the sequence. The next run may continue a calculation in the same mode (i.e.
steady state or transient) or begin a transient from a converged steady-state
calculation. The restart file, thus, provides the initial conditions for all but
the first run In a sequence.

For the first run in a sequence, the code performs a base level
initialization as follows. The user specifies reference values for temperature
(Tjjf) and for the pressure (Pj^f) at one location (x^ef ,yref, Zret) for each fluid
type. This temperature becomes the initial uniform value for all cells
containing that fluid type. The reference teraperature and pressure are used to
establish a reference fluid density p^^f frora the equation of state for each
fluid. Using these values and the three components of the gravitational
acceleration vector (g,), the code will establish a reference state pressure for
all fluid cells of each fluid type, using

Po = Pre£ + Pref [SA^-^xet) + gy(y-Yref) + ^zi^'-^x^t)] ' (10.22)


where x, y, and z are the Cartesian coordinates of the centers of the main (as
opposed to momentum) control volumes. (The components of the gravitational
acceleration vector and the coordinates of the reference point are always
specified on a Cartesian-coordinate basis, even when the cylindrical geometry
option is selected.) The uniform temperature and spatially varying pressure are
then used to establish initial values for enthalpy and density using the equation
of state. The initial values for fluid velocities at all cell faces are zero
91

except for faces which lie on boundary surfaces with specified-value-type


boundary conditions.

After this base level initialization has been perforraed, the user has
three ways to provide alternate values: restart, pointwlse, and one-dimensional
calculation, as discussed in the next three paragraphs.

If the fluid conditions are expected to be similar to those in a prior


run, the restart file written by the prior run raay be input to the new run. In
this case, the teraperature, pressure, and velocity distributions in the fluids
will be initialized to the values read frora this file. This option raay only be
used if the geometry is the same for the old and new runs. This method of using
a restart file as a guess differs from the normal use of a restart file to
continue a calculational history: in the guess mode, only the temperature,
pressure, and the three velocity components are read; in the continuation mode,
all variables are restored to their prior values.

The user also has the option to specify values on a pointwlse basis
using records in the input stream. Values may be specified for fluid
temperature, pressure, and the three components of velocity for as raany cell
locations as desired. These values are read after the optional reading of a
guess from the restart file. After all of these pointwlse values are read, an
enthalpy and density for each fluid cell will be calculated from the equation of
state using the temperature and pressure.

If the hexagonal fuel assembly option (cf. Appendix A) is being used,


the user may request that the code perform a one-dimensional initialization of
fluid variables. This initialization overrides all values read from a restart
file and the pointwlse card input. Crossflow velocities are assumed to be zero
during this initialization; the axial raass flow rate is constant. The user raust
specify a nonzero velocity and the temperature at the inlet. The code then
sweeps up the asserably from inlet to exit, working on one axial level at a time.
Assuming steady state, the energy added to the fluid at an axial level is taken
to be the sum of the energy generation in all fuel pins for that axial level^
Based on the enthalpy at the next lowest axial level, the energy added at this
axial level, and the mass flow rate, an average enthalpy for this axial level is
computed and assigned to all fluid cells at this axial level The gravitational
and frictional terms are evaluated to estimate the level-to-level pressure
change. The density and teraperature are established through the ^ q - t i o n of
state using the enthalpy and pressure, which are constant for all cells at this
axial level. The mass flow rate is divided by the density and flow area to get
an average velocity which is assigned to all cell faces at this axial level.

Further details on the use of restart files, pointwlse input and


selection of the one-diraensional initialization are given m the User s Guide
(Ref. 9).
92

11. CALCULATIONAL SEQUENCE AND METHODS

11.1 Overall Calculational Flow During a Run

A complete simulation of a physical thermal hydraulic problem with COMMIX


generally requires a sequence of individual computer runs. Within each computer
run, the overall calculational flow is shown in Table 11.1.

Table 11.1 Sequence of Calculations within a Run

1. Read card-image input and Restart file.

2. Establish geometry and initialize calculational variables.

3. Select time step size.

4. Advance variables across a tirae step (cf. Table 11.2)

5. Print results and write results to Plot file.

6. If the requested number of tirae steps or raaxiraum problem time have not
been reached, then go to Step 3 to begin a new time step.

7. End of the run; write a Restart file and stop.

The code reads user supplied information from two files: card-image input
and binary restart input. Based on the information in these two files, the code
establishes the geometry and initializes all calculational variables. The order
in which Information from these two files is used during the initialization
process is discussed in Sec. 10.2. Following selection of a tirae step size
(which will be discussed in Sec. 11.4), the code advances the solution across one
tirae step; the details of Step 4 are discussed in Sec. 11.2. After all variables
have been advanced, there are options for the user to selectively print various
variables and write certain variables to a binary plot file. If more
calculational history is desired, the code loops back to Step 3 to begin a new
time step. The norraal end of a coraputer run raay be based on completing the
requested number of time steps or problem simulation time. At the end of a run,
the code will write a new binary restart file if requested, so that the
calculation may be continued tn a subsequent computer run.

The details of how input is supplied to the code and output is returned
to the user are presented in the User's Guide (Ref. 9 ) . The User's Guide also
explains how the plot file may be passed to various postprocessing prograras to
obtain graphical output.

11.2 Advancement of Variables Across a Time Step

The bulk of the coding in COMMIX is directed toward advancing variables


across a single tirae step. This entails solving all conservation equations for
93

the fluids plus the thermal structure energy equations for all nodes in the
calculational domain.
The formulation is Implicit in time; i.e., all variables appearing in the
equations are at the new (end of step) time except for the use of the old
(beginning of step) time value in forming the unsteady terra of the difference
equations. This allows the time step size to be larger than would be the case
in an explicit formulation (all variables at old tirae except for the new time
value in the unsteady term), where the time step size is limited by Courant,
Fourier, and other such restrictions.

In the ideal case, all equations would be solved simultaneously in the


implicit formulation. The large number of equations (threeraoraentum,two
turbulence, plus one continuity, energy, and state equation for each fluid cell),
the nonllnearity of the equations, and the coupling of variables (both the
appearance of the same variable type for multiple cells and several variable
types for each cell in each equation) makes it impractical to solve all equations
truly simultaneously. Implicitness in COMMIX, therefore, is achieved by a
combination of equation grouping, linearization, and iteration. Each type ot
equation is solved in a separate block; within the block, the equations of that
type are linearized in the variable of solution and solved simultaneously for all
cells- the simultaneous solution within a block typically involves an "inner
iterative process. Results obtained within one block of the calculation are
passed to subsequent blocks, which solve other types of equations. The process
of sweeping through all calculational blocks is repeated, which is an outer
iteration process.
The division of the equations into blocks and the ordering of these blocks
are shown in Table 11.2. (This corresponds to the "SIMPLEST-ANL" treatment
described in Ref. 5; the other calculational schemes of Ref. 5 are no longer
fully functional in the current code version.)

the equations are processed as DIOCKS in LUC loi-^-w^..^ >..-^. . - ^^„l„;^


one block for each direction to generate the coefficients), pressure, turbulent
kinetic energy, turbulent kinetic energy dissipation rate, and energy. Most
b ock consist of two steps: calculation of the coefficients for ^^e Unearized
conservation equations for all cells followed by an iterative solution for the
unknots in the resulting equation systera. Theraoraentume'ju^ions are an
exception- the coefficients calculated for theraoraentumequations are used to
construct the pressure equations, which are solved for the pressures using an
Iterative process- the pressures are substituted into the momentum equations
raakS use of the coef'ficients from Step 5) to calculate the new velocities
without Iteration. The newly updated values f"™ ; " % " - \ "^^^,"„^,tties
subsequent blocks; for example, the latest estimates of end-of-step velocities
from Step 7 are used to form the coefficients for the energy equations in
Step 13.
Linearization of the conservation equations generally " = " 1 " >" * ^
coefficients containing values from the prior outer iteration ""^"Pl/^^S^^"^
solution variable value for the current outer iteration; for example the general
convective and diffusive terms in the momentum equations are formulated using
94

Table 11.2 Sequence of Calculations within Time Step n

1. Save values frora end of the last time step (e.g. u""' , P""' , etc.) as
"old" values; these are new the beginning-of-step values for time step n.

2. Set any pressures given by boundary conditions:

P " = f (t") .

3. Begin outer iteration ra.

4. Construct the coefficients for the raoraentum equations for all cell:

a , a" , -(^ , d" , « , d".


The velocity boundary conditions are applied and the purap equations are
solved during this process.

5. Using the coefficients from Step 4, construct the coefficients for the
pressure (i.e. continuity) equations for all cells:

a,^ ({ = 0,1,...,6) , bo^

6. Solve (iteratively) the system of equations from Step 5 to get the new
pressures:
pn.m

Save the maximum (in absolute value) mass residual per unit fluid volume

in (6p/Vf) for use in Step 15.

7. Substitute the pressures into the momenta equations to get the new
velocities:
Ijn.in ^n.m ^n,m

Save the maximum relative change of each velocity coraponent during this
outer iteration, e.g. (u"'"" - u"''""^ |/[(u^ + V^ + W^)''*]^^, as (Au)„^,
( ^'^)nax' and (Aw)„^ for use in Step 16.

8. Construct the coefficients for the turbulent kinetic energy equations for
all cells:

aj" («=0,1,...,6) , bo''.


95

Table 11-2 (Cont'd)

9. Solve (Iteratively) the system of equations frora Step 8 to get the new
turbulent kinetic energies:

10. Construct the coefficients for the turbulent kinetic energy dissipation
rate equations for all cells:

a,' (f = 0,l,..,6) , bJ.

11. Solve (iteratively) the system of equations frora Step 10 to get the new
turbulent kinetic energy dissipation rates:

(.n,m_

12. Set values for turbulent viscosity and thermal conductivity:

, n , ni\
X?'" = f(ti?'

Construct tVie coefficients for the fluid energy equations for all cells:
13
a,** (C = 0,1, ..,L) , bo'.
The energy boundary conditions are applied and the heat transfer rates to
the fluid from the boundaries and the therraal structures are evaluated
during this process. The value of L is greater than 6 for cells coupled
to nonadjacent cells through thermal structures.

14. Solve (iteratively) the system of equations from Step 13 to get the new
enthalpies:

hn.m.

Save the maximum relative change of enthalpy during this outer iteration,

i.e. Ih"'"" - h"''"-^/ Ih"'"! , in ( A h / h ) ^ for use in step 16.

Set the fluid density and temperature from the equation of state for all
15.
cells:
pn.m = f (pn.m _ j^n.m)

tpn.m = f (pn.m _ h"'") .


96

Table 11-2 (Cont'd)

An outer iteration is complete. If the maximum (a) velocity and enthalpy


16.
changes during this outer iteration and (b) mass residual for this outer
iteration satisfy the following conditions:

(AQ)^ < t, . (A<^)^^ < e, , (Aw),,^ < t,.

i m < e3 , and {h\ < (|)

then this tirae step has converged; go to Step 17 to complete the time
step. If convergence has not been achieved, then go to Step 3 to begin
a new outer iteration.

17. Update conditions for boundaries having adiabatic and duct-wall energy
boundary conditions:
Ti n u n rr n .n
Pb , hb , Tb , Pb-

18. Calculate new therraal structure and radiation surface temperatures:

Tan '
mH
.•• r •

19. Solve for new radiation heat fluxes:

qt •

20. Adjust volume porosities for expansion cells:

Yv

Fr''""M«J»?)"''" if «!»? = (|>o (11.la)


F,«t»?
F,"'""' («!)>?)"•'""' if «t»,° = *, (11.lb)
and

D, (<t), - (J>„) = D,"'"""' ((t)?'"""' - (J)?'"") , (11.2a)

where (> stands for one of the velocity components (i.e. u^ or u, v, or w ) ; these
same terms in the energy, turbulent kinetic energy, and turbulent kinetic energy
dissipation rate equations are formulated using
97

F, «t.>? = ?,"••"( «J»?)"''" (11. Ic)

and

(<t), - (Jio) = Dr-"""" ((J)?'" - (J)?-") . (11.2b)

where ^ stands for one of the computational unknowns (i.e. h, k, or c). In the
above equations, superscript n denotes the current time step number, and
superscripts m and m-1 denote the current and previous outer iteration numbers;
on the first outer iteration for a time step, the "previous outer iteration" is
the final outer iteration (M) of the previous time step (i.e. superscript "n,m-l"
at m-1 is replaced by "n-l,M"). The forms for F^ and D, are given in Sec. 4.
Within Ff, the velocity is Uj"-'""^ for the momentum equations but is Ui"-" for all
other equations since the new velocity is available. The density in ?/ and the
laminar part of the transport coefficient in D, are always p "•'"'' and r,"'""^ since
the equation of state is not invoked until the end of an outer iteration. The
turbulent part of the transport coefficient in Uf depends on whether the
calculational block in which the transport coefficient is needed is before or
after Step 12 of Table 11.2; therefore, it is Pt"-""' for the momenta, turbulent
kinetic energy, and turbulent kinetic energy dissipation rate equations and Ft"-"
for the energy equation. Source terms must also be linearized. The type of
linearization applied depends on the importance of the term and is liraited by the
form of the term. Most source terras are linearized by a separation process
sirallar to that used for the convective and diffusive terras; for example, the
fluid-structure momentum transfer (friction) term is calculated in the momentum
equations using

^fpKiUi = ^(fpiUii)'--ur . (^^-^^


Other terras are sufficiently important or sensitive to changes that an
extrapolation technique must be used; for example, the structure-to-fluid heat
transfer source term (cf. Sec. 6.3.5) Is calculated in the fluid energy equation
using

UA (r,-T) = [UA (T,-T)f--^ " ^ (J{ P (T.-T)]p""'(h"---h"-"-M . (11.4)

Use of advanced-iterate values in the above equations insures that the


formulation is implicit in time, obviating the need for imposing the classical
constraints on maximum time step size (e.g. Courant and Fourier) which are
required for formulations which are explicit in tirae. The choice between
separation and extrapolation involves evaluating a tradeoff in the use of
computing resources: extrapolation generally provides convergence in fewer outer
iterations than separation but requires more computing resources per outer
iteration to evaluate, store, and use the derivative involved in the
extrapolation; the decision is case dependent. In this regard, there are
indications that the body force source term in the momentum equations should be
calculated using an extrapolation technique, e.g.

(11.5)
pn.m-l^ J ^ r " " ' (uJ'^-U?'""-')
giP = 91

in order to improve the convergence rate in buoyancy dominated problems; this has
not been fully impleraented in the current code version.
98

There are four blocks of the calculation (Steps 6, 9, 11, and 14) which
perform an Iterative solution of a coupled system of linear equations. The
different iterative processes used in these blocks are described in Sec. 11.5.
The Important thing to note at this point is that the user specifies a relative
convergence criterion and a maximum number of inner iterations allowed for each
one of these inner Iterative processes.

At the end of an outer iteration, the code checks to see whether the
relative change in values between successive outer iterations and the mass
residuals are less than user-specified convergence criteria. The changes between
successive outer iterations are evaluated using

J^n.m _ j^n.m-l
< e. (11.6a)
hn.m

and
^n,m _ ,|)n.m-l
< e. (11.6b)
[(^ + v^ ... ;:T2\TAI
w^)^l
where 4 stands for each of u, v, and w. The term [...]„„ in the denominator of
Eq. 11.6b is the maximum fluid speed, calculated by sweeping over all fluid main
control volumes and computing the cell-center spatial average of all three
velocity components; theraaxiraumfluid speed is used for norraalization purposes
so that fluctuations of very small velocity components are ignored for the
purposes of assessing convergence. The mass residuals are evaluated using

5 ? ' V V £ < (6/V)p,,,f (11.7)

where the mass residual 5p is given by Eq. 5.7, Vf is the fluid volume for a
cell, and

p-Ai-Ui
(11.8)
(6/V)p,„f = «i

The user supplies the values of the constants Cj. The term (. . . )„ax is the
maximum raass flow rate per cell fluid volume; the quantity is maxiraized by
sweeping over all faces (the subscript 1 denoting the three spatial directions)
of all fluid cells. If Eqs. 11.5 and 11.7 are satisfied for all cells, then the
outer iteration has converged. If convergence has not been achieved, then outer
iterations are continued until a user-specified raaxiraum number of outer
iterations has been completed. If the inner iterations do not converge, then the
outer iterations will probably not converge. The actions taken by the user and
by the code in the event of nonconvergence depend on whether the case is running
in steady-state or transient mode (cf. Sec. 11.3) and whether the change-based
automatic time-step-size selector is being used (cf. Sec. 11.4).

After the outer iteration has converged, the code performs certain
computations associated with therraal structures, radiation surfaces, and
expansion-cell boundary conditions. At this point, all variables have been
advanced across the time step and control is transferred to Step 5 of Table 11.1.
99

11.3 Steady-State versus Transient Mode

Most of what has been presented in this report is directed toward an


explanation of the solution of the various thermal hydraulic equations as a
function of time to simulate the transient behavior of a system. The COMMIX
code however, may also be used to obtain steady-state solutions. These
steady-state solutions may be an end result or may be used to begin a transient
solution. (In this regard, a transient calculation should almost always be
started using initial conditions obtained from a converged steady-state
calculation.)

When running in the steady-state mode, the code solves the same equations
in the same order as for a transient calculation. A time-step size raust be
provided; this generally needs to start small but raay usually be increased as the
equations move toward convergence. This time-step size is used in the unsteady
term of all fluid conservation equations. The time is advanced as steps are
taken during steady state. Due to the nature of the calculation, the meaning of
this time is only partially physical. The solution is advanced somewhat just by
virtue of there being multiple tirae steps; however, a significant problem time
(not just number of time steps) must elapse for the solution to converge. During
a steady-state calculation, the energy equations for the thermal structures and
duct wall energy boundary conditions use an artificially large tirae-step size
(I.e. l.*10''° s) in order to keep these solids in therraal equilibrium with the
fluids during the progression toward steady state. The value of all transient
functions should be 1.0 in the steady-state mode.

The inner iteration pararaeters for a steady-state case should be chosen


in the same manner as for a transient case; i.e., small Cj and sufficient inner
iterations for this convergence to be obtained. On the other hand, the maximum
number of outer iterations per tirae step should be set to 1 in the steady-state
mode; rather than taking multiple outer iterations per time step, steady state
is approached by taking multiple time steps with each having only one outer
Iteration. During the approach to steady state, the code is continually
evaluating the changes noted in Step 16 of Table 11.2; however, the changes are
step-to-step rather than outer-to-outer. When these conditions on velocity,
fluid enthalpy, and raass residual are satisfied, the code will print a message
that "steady state has been achieved"; since these checks do not cover all
calculational variables (e.g. therraal structure teraperatures are not included),
the user should verify from a detailed examination of the code output that steady
state has, indeed, been achieved.

When the code is running in the transient mode, the maximum number of
outer iterations allowed should be selected such that the convergence test in
Step 16 of Table 11.2 can be satisfied. This iteration limit is dependent on the
time step size and the convergence criteria Cj and («/V)p,„f. As the time step
size increases and the £j values decrease, the number of outer iterations needed
to achieve convergence will increase. Transient calculations in which the outer
iterations are routinely not converging should be discontinued and restarted
using a smaller tirae step size or allowing raore outer iterations to achieve the
desired level of convergence.
100

11.4 Time Step Size Selection

There is a need during most calculations to vary the time step size for
computational efficiency. Small time step sizes are generally needed at the
beginning of both steady-state and transient calculations and during the portions
of a transient calculation when the boundary conditions are changing rapidly.
Large time step sizes may be used after the initial part of a steady-state
calculation and during the portions of a transient calculation when the boundary
conditions are changing slowly. There are three ways for the user to control the
tirae step size: manual, Courant-based automatic, and change-based automatic.

Under the manual mode, the user supplies two time step sizes, Atj and At2,
and a time step number for switching, Nto- The first of these two tirae step
sizes is used for all time steps through tirae step number Ntol the second of
these two time step sizes is used thereafter.

Under the Courant-based automatic mode, the time step size is adjusted
based on the Courant condition, which relates the time step size to the number
of mesh cells a mass particle would traverse in one time step. This is
implemented as follows:

At J_
A,uA (11.9)
V,
where p is a user-input constant and (...)„ax is the same as the quantity used
in Eq. 11.8 but without the density. The tirae step size is adjusted using this
forraula at the time step numbers specified by the user. This mode of time step
size selection has a strong theoretical basis when calculations are performed
using a formulation which is explicit in time, which dictates P <1 and, thus,
restricts a particle of mass from completely traversing a cell during a time
step. There is less theoretical basis for choosing p for the current
impliclt-in-time formulation in COMMIX, although arguments can be made
for P » 1 .

Under the change-based automatic mode, the tirae step size is adjusted
based on the relative rate of change of certain variables. Specifically, the
step-to-step changes of the kinetic energy, fluid teraperature, and turbulent
viscosity for each fluid cell are evaluated using the following relational
expressions:

|E" - E"-M < e (11.10)


"•Cmax
max (E")

I T H _ rpn-i j
< e (11.11)
rp n
and

K u , + >.t)''- (H, + U t ) " " ! , , (U 12)


et,„in< (^^, + ^ , ) n ' ^ ' — • ^ ' ' • ' ' '

The kinetic energy of a fluid cell is used as an aggregate raeasure of fluid


velocity and is defined by
101

E = pV, (iP+v^+w^) , (11.13)


where u, v, and w are center-of-raass-cell averages of the velocity components in
the three spatial directions. The maximum cell kinetic energy, found by sweeping
over all fluid cells, is used for normalization purposes in Eq. 11.10 so that
fluctuations of very small velocities are Ignored for the purposes of adjusting
the time step size. The turbulent viscosity is a measure of both the turbulent
kinetic energy and the turbulent kinetic energy dissipation rate (cf. Eq. 7.12);
the total viscosity is used in Eq. 11.12 since this is the main term that
provides coupling between the normal conservation equations and the turbulence
transport equations; if the turbulent viscosity is small relative to the laminar
viscosity, then Eq. 11.12 will not be the controlling factor in tirae-step size
selection; if the constant turbulent viscosity model (cf. Sec. 7.6) is being
used, then Eq. 11.12 ir not used for time-step size selection. The quantities
c„i„ and c„„ are user-input constants.

Under the change-based automatic mode, the relational expressions given


by Eqs. 11.10 through 11.12 are evaluated after the outer iteration converges;
thus, this is part of Step 16 in Table 11.2. If the relational expressions are
all satisfied, then the tirae step size remains unchanged for the next time step.
If all of the changes are less than c„i„ for Nt2 consecutive time steps, then the
time step size to be used on the next tirae step will be increased using

At"-^ = min (P-At", A t ^ ) , (11.14)


where P is a user-input constant which is greater than 1 and At„„ is a
user-input raaxiraum tirae step size; the quantity Nt2 is set by the code and
initially has a value of 5. If any of the changes exceeds c„„, then the tirae
step size is reduced using

At=AtVP , '''•'''
where P is the same constant used in Eq. 11.14. If the resulting tirae step size
is greater than a user-input minimum time step size (At„j„) , all variables are
restored to their beginning of tirae step values and the current time step Is
repeated; otherwise, the calculation is terminated with an error message. During
a steady-state calculation, the logic controlling time step size decrease is
bypassed in the code until after the tirae step size has undergone at least one
increase (i.e. the changes are less than c„i„ for Nt2 consecutive steps); this
avoids needlessly decreasing the tirae step size early in the calculation where
quantities are expected to be changing by rather large araounts; during this early
stage of the calculation, the time step size is set by the code using what was
described above for the manual mode (i.e. switching from Ati to At2 at step
number Nto). During a transient calculation, the step-repeat path is also taken
if the outer iteration does not converge in the allowed maximum number of
iterations. If, after any time step size increase, the code during the next time
step finds that the time step must be repeated using a reduced time step size,
then the code will double the current value of Nt2; this will slow down the rate
at which the tirae step size can be increased and avoid sorae of the step repeats
encountered in ill-behaved cases.
iU2

11.5 Inner Iteration Methods

There are four blocks of the calculation where an iterative method is used
to solve a system of linear equations. These are the solutions for pressure (P),
turbulent kinetic energy (k), turbulent kinetic energy dissipation rate (c), and
fluid enthalpy (h), which are Steps 6, 9, 11, and 14 in Table 11.2. Each of
these is denoted an "inner" iterative process, since they are all contained
within the outer iteration which sweeps through all blocks of equations multiple
times to complete a single time step.

Two different methods are used within the code for inner iterations:
successive overrelaxation and conjugate gradient. The successive overrelaxation
method is always used for the two turbulence equations and the fluid energy
equation; these equation systems have a coefficient matrix which has positive
diagonal elements and negative off-diagonal eleraents, is strongly or irreducibly
diagonally dominant, and is nonsymmetric. The coefficient matrix for the
pressure equation system has the same properties except that it is symmetric;
recognition of this symmetry property led to the addition of the conjugate
gradient method to COMMIX; the user may select either conjugate gradient or
successive overrelaxation as the method of solution for the pressure equation;
the former is recommended.

Brief descriptions of the two methods are given below. The impleraentation
of both methods is predicated on the fact that all of the coefficient matrices
are sparse; this is due to the physical restriction that a cell is only coupled
to its six closest neighbors in forming the difference equations. (Although
allowance is made in the fluid energy equation for coupling of nonadjacent fluid
cells via thermal structures, the number of such couplings is usually small and,
thus, the coefficient raatrix is still sparse.) The details of implementation are
deferred to the Programmer's Guide (Ref. 10).

The term "successive overrelaxation" is used here to denote what is more


precisely called the "point successive over- or under-relaxation iterative
method" (Ref. 35). In this method, a sweep is made through all cells (points)
m to calculate the residual of the conservation equation at inner iteration
1 frora

6j,m = a*,, K-' - t (at,Ah - b*,„ , (11.16)


••1
where 4 stands for P, k, c, or h and the superscripts denoting tirae step nuraber
and outer iteration number (which are constant during the inner iteration
process) have been omitted for compactness of notation. The coefficients a,*„
and bj,„ are defined in Tables 4.5 and 5.1 and are constant during an outer
iteration. The upper limit of the summation is normally the number of adjacent
cells (-6); the limit is higher for the fluid energy equation to include coupling
to nonadjacent cells through thermal structures. The superscript j is i if the
value for cell t has already been updated during inner iteration 1; otherwise,
it is i-1. The quantity ((i/ is the appropriate boundary value if the point
referred to by f is a boundary rather than an adjacent cell. After the residual
is calculated for a cell, it is immediately used to calculate a new value of the
solution variable for cell ra frora

•m = <l>m"' - 0)^ •Sj.m/ao*. , (H-l?)


103

where w^ is a relaxation constant input by the user. The process is technically


underrelaxation when u^<l and overrelaxation when (Ji,p>l; a proper choice of u^ can
accelerate convergence. Since the conservation equations for energy, turbulent
kinetic energy, and turbulent kinetic energy dissipation rate (i.e., 4
corresponding to h, k, and c) are nonsyraraetric, convergence of this solution
technique cannot be guaranteed unless u^ - 1, which reduces this to the "point
Gauss-Seidel" technique. (Another way of accelerating convergence of the
successive overrelaxation solution of the pressure equations is use of the
rebalancing technique described in Appendix B.) After updating the value for
cellra,the process of using Eqs. 11.16 and 11.17 is repeated for the next cell,
m+1. (The order in which cells are processed is explained in the Programmer's
Guide (Ref. 10) and is not necessarily "geometric".) This process continues
until the values for all cells have been updated, which completes inner iteration
1.

The term "conjugate gradient" is used here to denote what is more


precisely called the "preconditioned conjugate gradient method with ILU
(Incomplete Cholesky) factorization and no fill-in's" (Ref. 37). The method may
only be used for systems where the coefficient matrix is positive definite and
symmetric; therefore, within the context of COMMIX, this solution methodraayonly
be applied to the pressure equation. Like the successive overrelaxation method,
an inner Iteration process is involved, A relaxation factor (e.g. up) is not
used. New ILU factors are computed (1) at the beginning of a computer run,
(2) during the third outer iteration after they were last calculated, and
(3) during the outer iteration immediately following an outer iteration in which
the pressure did not converge in the allowed nuraber of inner iterations.
(Convergence is defined in the last paragraph of this section.) Mass rebalancing
should not be invoked when the conjugate gradientraethodof solution is selected.

The basis for convergence depends on which set of equations is being


solved. For 4 - Vi, c and h, the convergence is evaluated using

<t>i -<t>r < e.


(11.18)
<t>i
where c. are user-input constants. For 4 - ?, the convergence (within both the
successive overrelaxation and conjugate gradient methods) is evaluated using

5km/V, <(6/V)p,„, , ^"'^^


where (6/V)p „f is given by Eq. 11.8; thus, the convergence of the pressure
equation is based on attaining the desired degree of minimization in the mass
residual. Inner Iterations are continued until the convergence criterion is
satisfied or until a user-specified maximum number of inner iterations has been
exceeded. Control then transfers to the next block of the calculation (cf.
Table 11.2).
104

ACKNOWLEDGEMENTS

As COMMIX-IAR/P is an extension to earlier versions of COMMIX, we


acknowledge the liberal use of the prior COMMIX documentation in preparing this
report and thank the developers of the prior COMMIX versions for their time in
discussing the implementation of the new raodeiing described herein.

We acknowledge the contributions by others in developing certain new


features in COMMIX-IAR/P: Jon Beitel, Felix Chen, and Don Malloy for their work
on the multifluid, pump, and radiation heat transfer models and Greg Greenman for
his work on the conjugate gradient solution method.

We deeply appreciate the diligent work of Rosemary Barnes, Joyce Driver,


Eileen Johnson, Lucia Johnson, Betty Kinney, Jill McGregor, Ken Miles, Trang
Nguyen, and Karen Stephenson in typing the manuscript and preparing the figures.

This work was supported by the United States Department of Energy.


105

REFERENCES

1. W. T. Sha, H. M. Domanus, R. C. Schmitt, J. J. Oras, and E. I. H. Lin,


"COMMIX-1: A Three-Dlmensional Transient Single-Phase Component Computer
Program for Thermal-Hydraulic Analysis", NUREG/CR-0785 (or ANL-77-96).
Argonne National Laboratory, Argonne, IL (September 1978).

2. H. M. Domanus, W. T. Sha, V. L. Shah, J. G. Bartzls, J. L. Krazinski, C,


C. Mlao, and R. C. Schmitt, "COMMIX-2: A Steady/Unsteady
Single-Phase/Two-Phase Three-Dlmensional Computer Prograra for
Thermal-Hydraulic Analysis of Reactor Components", NUREG/CR-1807 (or
ANL-81-10), Argonne National Laboratory, Argonne, IL (October 1980).

3. H. M. Domanus, R. C. Schmitt, W. T. Sha, and V. L. Shah, "COMMIX-lA: A


Three-Dlmensional Transient Single-Phase Computer Program for Thermal
Hydraulic Analysis of Single and Multicomponent Systems; Volume I: Users
Manual", NUREG/CR-2896 (or ANL-82-25), Vol. I, Argonne National
Laboratory, Argonne, IL (December 1983).

4. H. M. Doraanus, R. C. Schraitt, W. T. Sha, and V. L. Shah, "COMMIX-IA: A


Three-Dlraensional Transient Single-Phase Computer Prograra for Thermal
Hydraulic Analysis of Single and Multicomponent Systems; Volume II:
Assessment and Verification", NUREG/CR-2896 (or ANL-82-25), Vol. II,
Argonne National Laboratory, Argonne, IL (December 1983).

5. "COMMIX-IB: A Three-Dlmensional Transient Single-Phase Computer Program


for Thermal Hydraulic Analysis of Single and Multicomponent Systems;
Volume I: Equations and Numerics", NUREG/CR-4348 (or ANL-85-42), Vol. I,
Argonne National Laboratory, Argonne, IL (September 1985).

6. "COMMIX-IB: A Three-Dlmensional Transient Single-Phase Coraputer Prograra


for Thermal Hydraulic Analysis of Single and Multicomponent Systems;
Volume II: User's Manual", NUREG/CR-4348 (or ANL-85-42), Vol. II, Argonne
National Laboratory, Argonne, IL (September 1985).

7. H. M. Doraanus, Y. S. Cha, T. H. Chien, R. C. Schmitt, and W. T. Sha,


"COMMIX-IC: A Three-Diraensional Transient Single-Phase Computer Program for
Thermal-Hydraulic Analysis of Single-Coraponent and Multicoraponent
Engineering Systeras; Equations and Numerics" , NUREG/CR-5649 (or ANL-90/33) ,
Vol. 1, Argonne National Laboratory (November 1990).

8. H. M. Domanus, Y. S. Cha, T. H. Chien, R. C. Schmitt, and W. T. Sha,


"COMMIX-IC: A Three-Dlmensional Transient Single-Phase Computer Program for
Therraal-Hydraulic Analysis of Single-Coraponent and Multicoraponent
Engineering Systems; User's Guide and Manual", NUREG/CR-5649 (or ANL-
90/33), Vol. 2, Argonne National Laboratory (November 1990).

9. R. N. Blomquist, P. L. Garner, and E. M. Gelbard, "COMMIX-lAR/P: A


Three-Dlmensional Transient Single-Phase Coraputer Program for Thermal
Hydraulic Analysis of Single and Multicomponent Systems; Volume 2: User's
Guide", Argonne National Laboratory, Argonne, IL (to be published).

10. R. N. Blomquist, P. L. Garner, and E. M. Gelbard, " COMMIX-lAR/P: A


Three-Dimensional Transient Single-Phase Coraputer Prograra for Therraal
106

REFERENCES (Cont'd)

Hydraulic Analysis of Single and Multicomponent Systems; Volume 3:


Prograraraer's Guide", Vol. 3, Argonne National Laboratory, Argonne, IL (to
be published).

11. R. N. Bloraquist, P. L. Garner, and E. M. Gelbard, "COMMIX-lAR/P: A


Three-Dimensional Transient Single-Phase Computer Program for Thermal
Hydraulic Analysis of Single and Multicoraponent Systems; Volume 4:
Verification and Validation", Argonne National Laboratory, Argonne, IL
(to be published).

12. "American National Standard: Programming Language FORTBIAN" , ANSI


X3.9-1978, American National Standards Institute, Inc., New York, NY
(1978).

13. W. T. Sha and B. T. Chao, "Local Volume-Averaged Transport Equations for


Single-Phase Flow in Regions Containing Fixed, Dispersed Heat-Generating
(or Absorbing) Solids", NUREG/CR-1969 (or ANL-80-124), Argonne National
Laboratory, Argonne, IL (April 1981).

14. Hoyt C. Hottel, "Radiant-Heat Transmission", in Williara H. McAdaras (ed.).


Heat Transmission, 3rd edition, McGraw-Hill Book Company, New York, NY,
55-125 (1954).

15. G. Poljak, "Analysis of Heat Interchange by Radiation between Diffuse


Surfaces", Tech. Phys. USSR, 1, 555-590 (1935) as presented in Max Jakob,
Heat Transfer. Vol. II, John Wiley 6< Sons, Inc., New York, NY, 21-34
(1957).

16. B. Gebhart, "Unified Treatment for Thermal Radiation Transfer Processes --


Gray, Diffuse Radiators and Absorbers", Paper no. 57-A-34, ASME (Dec.
1957).

17. Robert Siegel and John R. Howell, Thermal Radiation Heat Transfer. 2nd
edition, McGraw-Hill Book Company, New York, NY, 822-831 (1981).

18. J. J. Dongarra, C. B. Moler, J. R. Bunch, and G. W. Stewart, LINPACK


Users's Guide. Soc. for Ind. and Appl. Math., Philadelphia, PA (1979).

19. 0. Reynolds, "On the Dynamical Theory of Incompressible Viscous Fluids and
the Determination of the Criterion", Philos. Trans. R. Soc. Lon., Ser. A,
Math. Phys. Sci.. 185, 123 (1895).

20. L. Prandtl, "Bericht uber Untersuchungen zur ausgebildeten Turbulenz", Z.


Angew. Math. Mech., 5, 136-139 (1925).

21. L. Prandtl, "Uber eln neues Formelsystera fur die Ausgebildete Turbulenz",
Nachrichten Akaderaie der Wlssenschaften. Gottinpen,
Matheraatisch-Phvsikallsche Klas.qe 6-19, (1945).

22. P. Bradshaw, D. H. Ferriss, and N. P. Atwell, "Calculation of


Boundary-Layer Development using the Turbulent Energy Equation", J. Fluid
Mech., 28, 593-517 (1957) .
107

REFERENCES (Cont'd)

23. Victor W. Nee and Leslie S. G. Kovasznay, "Simple Phenomenological Theory


of Turbulent Shear Flows", Physics of Fluids, H , 473-484 (1969).

24. F. H. Harlow and P. I. Nakayama, "Transport of Turbulence Energy Decay",


LA-3854, Los Alamos Scientific Laboratory, Los Alamos, NM (1958).

25. W. P. Jones and B. E. Launder, "The Prediction of Laminarization with a


Two-Equation Model of Turbulence", Int. J. Heat Mass Trans.. 15, 301-314
(1972).

26. P. Y. Chou, "On Velocity Correlations and the Solutions of the Equations
of Turbulent Fluctuations", Quart. Appl. Hath.. 1, 38-54 (1945).

27. K. Hanjalic and B. E. Launder, "A Reynolds Stress Model of Turbulence and
its Application to Thin Shear Flows", J. Fluid Mech., 52, 609-538 (1972).

28. B. E. Launder, G. J. Reece, and W. Rodl, "Progress in the Development of


a Reynolds-Stress Turbulence Closure", J. Fluid Mech. . 68, 537-566 (1975).

29. OE- cit. (Ref. 5 ) , pp. 39-49.

30. W. P. Jones and B. E. Launder, "The Calculation of Low-Reynolds-Number


Phenomena with a Two-Equation Model of Turbulence", Int. J. Heat Mass
Transfer, 16, 1119-1130 (1973).

31. Y. Nagano and M. Hishida, "Improved Form of the k-c Model for Wall
Turbulent Shear Flows", Trans. ASME, J. Fluids Engineering. Ij09, 156-150
(1987).

32. Lewis F. Moody, "Friction Factors for Pipe Flow", Trans. ASME. 61, 671-684
(1944).

33. Victor L. Streeter and E. Benjamin Wylie, Hydraulic Transients.


McGraw-Hill Book Company, New York, NY, 151-158 (1957).

34. M. Marchal, G. Flesch, and P. Suter, "The Calculation of Waterhammer


Problems by Means of the Digital Computer", Proceedings of the
International S-ymposium on Waterhammer in Pumped Storage Projects, ASME
Winter Meeting, Chicago, IL, 168-188 (November 7-11, 1965).

35. E. Benjamin Wylie and Victor L. Streeter, Fluid Transients. FEB Press, Ann
Arbor, MI, 102-117 (1983).

36. Richard S. Varga, Matrix Iterative Analysis. Prentice-Hall, Inc.,


Englewood Cliffs, NJ, 55-61 (1962).

37. David S. Kershaw, "The Incomplete Cholesky - Conjugate Gradient Method for
the Iterative Solution of Systems of Linear Equations", J. Coraput. Phys.,
26, 43-65 (1978).
108

APPENDIX A
HEXAGONAL FUEL ASSEMBLY

In the Initial COMMIX development period, major emphasis was on the


analysis of hexagonal fuel assemblies. Consequently, several features and raodels
have been iraplemented in COMMIX (Refs. 1 and 3) that are specific to hexagonal
fuel assemblies. These are described briefly in this section.

A.l Hex-Geometry Option

The hex-geometry option is available in COMMIX for the calculation of


all required geometrical pararaeters for hexagonal fuel assemblies. The
subroutines for this option have been developed so that a minimura araount of
inforraation is required as input, relieving the user frora the tedious work of
preparing georaetrical data. The user has to provide only the following input
data:

Pins Number of pins, pin diameter, distance between pin


centers, and clearance between pin and wall.

Wire wrap Diameter of wires next to wall, diameter of wires


away from wall, and type of wire wrap option
desired.

Partitioning Number of axial partitions, size of each axial


partition, and type of cross-sectional partitioning.

With this rainiraum inforraation, the code calculates all required geometrical
parameters--grid sizes in x and y directions, directional surface porosities,
volume porosities, wetted perimeters, hydraulic diameters, surface areas, etc.

The following restrictions are imposed when the hex-georaetry option is


selected:

Axial length is along the z direction.


One flat side of the asserably lies on the x axis, and
Eight boundary surfaces of the assembly are numbered as follows;

Surface No. Locations

1 Lower left diagonal in x-y plane


2 Upper left diagonal in x-y plane
3 Lower right diagonal in x-y plane
4 Upper right diagonal in x-y plane
5 Lower flat along x axis
6 Upper flat parallel to x axis
7 Entrance plane normal to z axis at z=0
8 Exit plane normal to z axis at z-z^^.

Two types of cross-sectional partitioning are available: Quarter-pin


and Full-pin. To illustrate the differences, a quarter-pin partitioning and a
full-pin partitioning of a 19-pin hexagonal fuel assembly are shown in Figs. A.l
109

and A.2, respectively. (Unlike what is shown in Figs. A.l and A.2, the six
corners of the bundle are represented in COMMIX as being sharp, not rounded.)
The quarter-pin and full-pin partitionings derive their names from how much of
a fuel pin is associated with each fluid cell in a constant-z plane in the bundle
interior; for quarter-pin partitioning, this is one quarter of one pin and for
full-pin partitioning this is one half of one pin and one quarter of two adjacent
pins (for a total of one full pin); there are, of course, exceptions raade for
fluid cells near the assembly wall.

The hex-geometry option can also be used even if the hexagonal fuel
asserably under analysis has some deviations from a regular hex-georaetry or
contains some internal structures (e.g., blockage) that affect the values of
volume porosity and directional surface porosities. The only difference is that
the user now has to input the values of volume porosity and directional surface
porosities that are different from those for a regular hex-georaetry. The user-
prescribed values override the code-calculated values.

If the fuel bundle georaetry is very irregular (e.g., different fuel-pin


diameters, nonuniform spacing with different pitches, etc.), then the user should
bypass the hex-geometry option and use the norraal box-geometry option, i.e. the
basic Cartesian coordinate grid for which the user specifies all mesh spacings,
boundary surface locations, porosities, force structures, and therraal structures.

The hex-geometry option is only applicable when the full cross section
of the bundle is raodeled. If only a sector (e.g., 1/12) is desired, then the
normal box-georaetry option raust be used.

A.2 Wire Wrap Model

A.2.1 Introduction

The presence of a helical wire wrapping around a fuel pin has two
effects on fluid flow.

The geometrical effect, where the presence of the wire wrap


influences the fluid flow by reducing the available flow space (via
modification of the volume porosity and directional surface
porosities in the appropriate fluid cells), and

The physical effect, where the presence of the wire produces


additional drag on the fluid flow (via inclusions of additional
resistance terms in the raoraentum equation).

The presence of wire wraps may be included by invoking either the smeared or
cell-integrated option, which are described below. There is no automated way for
including the effects of spacer grids.

A. 2. 2 Smeared Wire Option

In this option, the volume porosity and directional surface porosities


are modified uniformly across the section. This is done by distributing the
total wire volume equally over all cells and the total wire-wrap cross-sectional
area equally over all cells in each axial plane. The "Physical" effects are
neglected.
110

Fig. A.l. Quarter-Pin Partitioning of 19-Pin Fuel Assembly


in a Hexagonal Duct

Fig. A.2. Full-Pin Part'.tioning of 19-Pin Fuel Assembly


in a Hexagonal Duct
Ill

A. 2. 3 Cell Integrated Option

A.2.3.1 Geometrical Effects

The geometrical effects due to the presence of wire wraps are introduced
by modifying volume porosities and directional surface porosities. This is done
using the relations

Yv = Yv P A," dz , (A.l)
(AxAyAz)

Y;:,i.i/2 = Yx,i.i/2 - j ^ ^ / ; ; dA:.i.i/2. (A-2)

Yy,j.i/2 " Yy.j.i/2 (AxAz) r dA;,i.i/. , (A. 3)

and

(A.4)
Y2,k»i/2 ' Yz,k;,i/2 (AxAy)

where the superscript w refers to the wire wrap and A is the cross-sectional area
of the wire wrap. The first terra on the right-hand side of these four equations
is the value of the porosity as set by the presence of the fuel pins and the
irregular boundary surfaces. The integrals on the right-hand sides of Eqs. A.l-
A 3 are evaluated nuraerically. At each axial position, A is coraputed by
determining the wrap's proper location in a cell. Jhe step size for numerical
integration is taken to be the axial distance over which the wrap spirals around
the pin by three degrees, i.e.

, wire pitch (A.5)


uz = "^ •
120

A.2.3.2 Wire Drag Effect

This resistance force per unit volume due to the wire wrap is raodeled
as

e = •PI'^I'^ '"^.' . (A.6a)


^" (AxAyAz) '
112

where p is the fluid density and w is the axial velocity. The resistance force
raay be written in component form

f^l + fyj + f,k , (A. 5b)

where f^ is the coraponent of force in spatial direction 1 and becoraes part of the
distributed resistance terra Rj in the momentum equations (cf. Table 2.1). The
quantity (C"A) is a combination of a wire force multiplier and the projected
area of the wrap

(CA) = C , ( C ^ J + CyAyJ + C,A,k) , (A. 7)

where Cj is the user-input wire-force multiplier in spatial direction 1, Aj is


the projected area in spatial direction 1, and Cf is a user-input wire-force
multiplier that allows for a differentiation between the interior and wall
regions of the bundle. The calculation of A is briefly described here.

Figure A.3 shows a typical wire wrap arrangement. Consider the wire
wrap as a spiral ring of width d^, attached to the fuel pin and located at
position S(x,y,z) as shown in Fig. A.4. The projected area is defined in terms
of a vector cross product as

d A = (dS X n)d„ , (A.8a)

where the unit normal vector is

n = 1 COS a. + jsin a, (A.9)

the wire wrap position vector is

S = lip COS o + Jr^ sin « + ic (z,, + y - P„) , (A. 10)

tp is the radius of the fuel pin, ZQ is the reference axial position where wrap's
angular position a is zero, and P„ Is the wrap pitch. The change in the wire
wrap position vector is
113

-i^.ph

Fig. A.3. Typical Wire W'ip


Arrangement

H 2-^P r

WIRE WRAP

I '- FUEL PU^

Fig. A.4. Cross Section of a Helical Wire Wrap around a


Fuel Pin
114

dS = [ i (-Ip sin a) + J tp cos a + E P„/27i] da . (A.11)

Substituting Eqs. A.9 and A.11 into Eq. A.8a gives

P d
dA = " " [i(-sin a) + J cos a - K tan 8] da , (A.8b)
2ll

where

e = tan-^ I " '^Pl (A. 12)

is the angle of Inclination between the wire wrap centerline and the fuel pin
centerline. Integrating Eq. A.8b between two z planes [k-1/2 and k+1/2] for a
given cell gives

- P d,., -
A = — j - ^ [i(cos a^-cos a.,)+j (sin a^-sin a.,)-9.(a^-a.,) tan Q] , (A.13)

where

_ 271 (Zi,.i^2-Zo) ,^ .,, .


«! r , (A. 14a)

2ir{ z^,i,2 - ZQ)


"2 = ^ . (A. 14b)

and

02 - « ! = ^"'^IC^l/Z ~ ^l^-l/2>
(A.15)
115

The components of the projected wire area (A,, Ay, and Aj) raay be
extracted from Eq. A.13 for use in Eq. A.7. Substitution of this form of Eq. A.7
into Eq. A.6a defines the components of the wire force vector (f^, fy, and f,)
in Eq. A.6b:

P„d„
C, C ^ ( c o s O , - cos O.) , (A. 1 5 a )
AxAyAz 2n

fy = A^^^^'l ^ C, Cy (Sin a, - sin a,) . (A.15b)


' AxAyAz 2it ^

and

f = Ph^k M H C, C , ( a i - a,)tanQ . (A.16c)


^ AxAyAz 2ii ' ^ ^ ^

These components are part of the distributed resistance source terms (R,, R,, and
R ) shown in Table 2.1 for the respective component momentum equations. The way
in which these wire force terms are incorporated into the finite-difference form
of the momentum equation source term (Eq. 4.24) is different for the axial and
transverse directions, since the expressions given by Eq. A.15 contain the axial
velocity component (w) but not the transverse velocity components (u and v) . The
term -f^/w becomes part of (Sp„)o for the z-direction equation and the terms -f,
and -fy become part of (S,Jo and (S,y)o, respectively, for the x- and y-dlrection
equations.

A.3 Fuel Pin Resistance

Additional components of the distributed resistance forces R,, Ry, and


R, (defined in Table 2.1) due to fuel pins areraodeledusing

Rj = C,,j fj (Wp/A) p|Uj|Uj , (A. 17)

where R is the distributed resistance force per unit volume, Cf is a user-input


coefficient, f is the friction factor, (Wp/A) is the wetted perimeter per unit
cross-sectional area, u is the local fluid velocity, and the subscript j refers
to spatial directions (x, y, and z).

When a rod bundle is aligned along the z axis, the cross-flow friction
factor fx is given by (Ref. 3) the largest of the following three expressions:
116

Py - 2^P Re, (A.18a)


„ - 1.86r,

0.6 0.25 + 0.118 Re' (A.18b)


2r, (0.5 PjlJ'"'

and

Py - 2 r p 4-32rp
f, = 3 ii/Re. 1 + (A.lBc)
Py - 1 . 8 6 r p P„ - 2 r ,

where Py i s the rod p i t c h (spacing) in the y d i r e c t i o n , tp i s the rod r a d i u s ,


p,f i s the laminar (molecular) f l u i d v i s c o s i t y , and p, i s the t o t a l (laminar plus
t u r b u l e n t ) f l u i d v i s c o s i t y . The Reynolds numbers appearing i n t h e s e e x p r e s s i o n s
are

p(Py-2rp) |uj
Re„ (A.19)

and

. _ P 2 r p |u^|
Re (A. 20)

Analogous e x p r e s s i o n s are used to define the c r o s s - f l o w f r i c t i o n f a c t o r


in the other d i r e c t i o n , fy, by r e p l a c i n g u, with Uy, Py with P,, and Re, with Rey
in Eqs. A.18 through A.20.

The a x i a l f r i c t i o n f a c t o r f^ i s given by

S/RBj f o r Re^ < 9 37 (A.21b)


0.0007 + 0.07/Re°''^ f o r Re, > 937 (A.21a)

where the Reynolds nuraber i s


117

__ P DbluJ (^22)

and Dj, is the local hydraulic diameter.

The distributed resistance due to the fuel pins for each coordinate
direction involves the velocity component in that direction; therefore, the
finite-difference form of the raoraentum equation source terra (Eq. 4.24) will
contain the term -R,/u in (Sp„)o , -Ry/v in (Spy)o , and -R^/w in (Sp,)o , where Rj
is given by Eq. A.17 for j - x, y, and z.

A.4 Other Features

A.4.1 Heat Source

One way of modeling the heat source frora fuel pins requires minimal user
input. When this is selected, the code calculates the heat source in a cell
using the relation

0(i,j,k) = C I N d , j) •OK(k) »(2KrpAz)«PINF.0FLUX.f„,(t) , (A.23)

where QIN(l,j) is the normalized transverse power distribution at location (i,j)


for all axial locations k, QK(k) is the normalized axial power distribution at
axial location k, QFLUX is the average heat source per unit surface area of fuel
pin, PINF is the fraction of a pin in the cell under consideration, (2irrpAz) is
the pin surface area, f is the transient function to account for the variation
of heat source with time, and the subscript nf is the transient function number.

Equation A.23 assumes that all axial planes have the sarae normalized
transverse power distribution QIN, and all points in a transverse plane (cells
with sarae (i,j) locations) have the sarae normalized axial power distribution QK.

The power Q calculated in Eq. A. 23, is divided by the fluid cell volume
(Vf„) and added to the heat source terra (q\,) in the fluid-energy equation.
Therraal Inertia of fuel pins is not accounted for in this calculation. The use
of the easy heat-source input described here is therefore recommended only for
steady-state analysis and slow transients. For fast transients, fuel pins should
be considered as therraal structures (described in Sec. 6 ) .

When the heat source has nonuniform distributions in all three


directions, there are two other alternatives:

. Prescribe a volumetric heat source (q") for each cell using array
QSOUR, or

Treat fuel pins as thermal structures and prescribe a heat source


using thermal structure input.
118

A.4. 2 Pressure Boundary Conditions

Most COMMIX simulations need a specified velocity boundary condition at


the inlet and a pressure condition coupled with a continuative mass flow boundary
condition at the exit. Some fuel assembly problems are raore appropriately
described by having pressure boundary conditions at both ends. The velocity and
pressure boundary conditions are discussed in Sees. 10.1.1 and 10.1.2.

A.4.3 Initialization

An option is selectable which will perform a one-diraensional


initialization of fluid conditions, rather than the default uniform
initialization. This procedure sweeps from the inlet to the exit, one axial
level at a time, calculating the heat added and uses this in conjunction with the
mass flow rate to set enthalpy, teraperature, density, pressure, and axial
velocity. Further details are in Sec. 10.2.
119

APPENDIX B
MASS REBALANCING

B.l Introduction

Several iterative procedures may be used to obtain a solution to the set


of algebraic equations which comprise the pressure equations given by Eq. 5.5.
Two such solution methods were presented in Section 11.5:
successlve-over-relaxatlon and conjugate-gradient. The solution technique must
be as efficient as possible in solving the equations in order to minimize the CPU
time required to run a given problem history.

Before the conjugate-gradient raethod was implemented in COMMIX, a scheme


was developed (Ref. 3) to accelerate convergence of the successive-over-
relaxation-based solution of the pressure equations. The scheme is called raass
rebalancing because it results in adjusting velocities such that raass
conservation is obtained on a coarse-mesh basis. This scheme proved to be
extremely effective in reducing the nuraber of iterations required to achieve mass
convergence when using the successive-over-relaxation solution technique.

In more recent tirae, the conjugate-gradient technique has been implemented


in COMMIX to solve the pressure equations. It has generally been found to be
more efficient than the successive-over-relaxation technique and should be the
user's normal choice. Mass rebalancing is not required with the
conjugate-gradient solution procedure, and its use may, in fact, increase the
computer CPU time.

The remaining subsections describe the raass rebalance scheme.

B.2 Description

In the mass rebalancing scheme, a coarse mesh is developed by combining


several fine mesh cells. Each grouping of fine mesh cells is called a
"rebalancing region". A pressure correction equation is derived for each
rebalancing region by summing the pressure equations for all fine mesh cells in
the rebalancing region. The following conditions are applied in forming and
working with the summations:

- A uniform pressure correction is applied to all cells within a


rebalancing region, and

- The pressure correction for each rebalancing region is determined such


that the sum of the mass residuals over all cells in that region
is zero.

Adjusting the pressures by the pressure correction adjusts the velocity field and
the mass flow rates, since pressure and velocity are related through Eq. 5.3.

The pressure corrections obtained frora the solution of these equations,


when applied to all cells in a flow domain, help resolve large-scale
distributions and, hence, reduce the number of iterations required for final
solution of the pressure equation.
120

B.3 Derivation of the Pressure Correction Equation

The flow domain is divided into N rebalancing regions; an exaraple of this


division is shown in Fig. B.l. The rebalancing regions are chosen such that any
region n has neighboring cells only in the neighboring rebalancing regions n-1
and n+1. (Rebalancing region 1 must be defined such that it only has neighboring
cells in rebalancing region 2.) The faces of the cells separating rebalancing
regions n and n+1 form rebalancing surface n. A pressure correction will be
computed for each one of these N rebalancing regions. No pressure correction is
calculated for the cells (if any) which are outside of the N rebalancing regions.

>N 1 i * t - ^ - Cells Outsi


i 1
1 Rebalancing
1 1
1 1 1 1
1 1
1 1 1 1
__1 J- _ x _
i-
— 1 1
1
1
j 1
1 1 1
' ! 1 1 ;
Inn' 1 1
I ' l l
1 *
Rebalancing -i _ Rebalancing
Surface n
1 n 1 1
H1 1 '
I

|n-l 1 1 1
Surface n

1 1 '
1 t 1 1 1 '
1
i 1 1 ' 1 1
1 1
— 1 - - - 1 —
1
' i 1
1 1 1 1
1 1
—1-- -_4--4--i--^
1
1
' ' 1
1 1
' ! 1
1 1 1
1 ; 2
1
' 1 2 1
1 ! 1
f- -
1
1 RE GION| I 1 1
1

Fig. B.l. Coarse Mesh Rebalancing Regions

The pressure equation for cell ra is (frora Eq. 5.7) given by

^ 0 P; - I (a„,Pr) - b„„ = S; , • (B.l)


121

where P„* is the pressure, 6„* is the mass residual, a„ and b„ are the
coefficients of the pressure equation (cf. Table 5.2), and the summation is over
the six cells which are the neighbors of cell m. The uses of P as a superscript
on a„ and b„ and as a subscript on S„* have been oraitted for notational
simplicity. At this point, P„* has not converged; thus, the continuity equation
is not satisfied and the mass residual is not zero. These pressure equations are
summed for all cells in rebalancing region n to give

(B.2)
amo Pm - ^ {a„jP;)-l=m E 6:
The right hand side of this equation is the net raass residual for rebalancing
region n.

The rebalancing scheme assumes that there exists a new pressure


distribution (P„) which when substituted into Eq. B.2 for P„* will result in zero
net mass residual for each rebalancing region, or

(B.3)
jPm - .^ (a^P.)-bmO

The new pressure is assumed to be the sum of the old pressure and a pressure
correction (APn)

Pm = Pm > A P „ ,
where the pressure correction is a constant for all cells m in rebalancing
region n. Substitution of Eq. B.4 into Eq. B.3 gives

AP„
E ,Pm - ^(am.P<') - K ten
in€n

(B.5)
AP„_i - E E a,, AP„.i = 0 ,
E Vte(n-l) J [m6n\<€(n«l) /.
men
where the summations over f involving pressure corrections have been split into
three partial summations
(B.6)
S (^-^•) = J-U (^'"'^ ' £ ("-'•) ' ,^lu^^'''^
to recognize that the cell referred to by t is in one of three rebalancing
regions n-1, n, or n+1. Equation B.5 holds for all of the N rebalancing
regions with two exceptions: (1) rebalancing region 1 has no AP„_i term, since
there is no rebalancing region 0; (2) rebalancing region N has no AP„,i terms,
since cells outside of the N regions do not participate m rebalancing.

After some rearrangement, Eq. B.5 may be written


(B.7)
- Ai,„*AP„.i + Ao,„ * A P „ - A2.„*AP„.i = B„ ,

where

for n = l
(B.8)
for n=2,...,N
[ men Lie (n-1)
122

E for n=l,...,N-l (B.9)


m€n for n=N
0
Al.n + Aj (B.IO)
and

E «: (B.ll)

where use has been made of the relationship between the coefficients in the
pressure equation (cf. Table 5.2)

^t (B.12)

Since the terms A^ „ and B„ are known, the set of N equations given by Eq. B.7
form a linear system of the form

[A] [AP] = [B] (B.13)


which may be solved for the pressure corrections. Since the coefficient matrix
[A] is tridiagonal, this system of equations may be solved without iteration
using forward elimination followed by backward substitution (which is identical
to the technique eraployed in the thermal structure temperature solution shown in
Sec. 5.3.4). Computation of the terras in the coefficient matrix is simplified
by recognizing that A,„-k2 n-i, rather than having to evaluate the summation given
by Eq. B.8.

B.4 Application

There are two ways in which rebalancing regions may be formed:


user-specified and plane-by-plane. Theseraethodsraaybe used Individually or raay
be combined.

The user-specified rebalancing regions correspond to the exaraple depicted


in Fig. B.l; the user specifies in corapleteness which cells are in each
rebalancing region n and which faces of which cells form rebalancing surface n.
The user-specified regions must conform to the restriction that cells in
rebalancing region n are only coupled to cells in rebalancing regions n-1, n, and
n+1.

In the plane-by-plane option, the user specifies one direction in which


rebalancing is to occur, either I, J, or K. Rebalancing regions and planes are
then defined by the code: rebalancing region n consists of all cells in the n-th
plane in the direction of rebalancing (i.e. cell location indices (n,j,k),
(i,n,k), or (i,j,n), where i-1,..., IMAX, j-l,..., JMAX, and k-1 ,,..,KMAX) ; rebalancing
surface n will be the faces of the cells in the plane which separate region n
from region n+1.

For both ways of specifying rebalancing regions, the progression of


rebalancing regions or the rebalancing direction should normally be chosen such
that region-to-region pressure changes are larger than the cell-to-cell pressure
changes within any of the rebalancing regions. This choice norraally coincides
with the direction of flow: region 1 should be at the problera inlet and
123

successive regions are defined as you move through the problem geometry toward
the outlet.

Multifluid problem geometries should only use user-specified regions and


the N regions should contain only one fluid.

The pressure field P obtained here frora Eq. B.4 is not the final solution,
since this pressure only satisfies net mass conservation for each rebalancing
region -- not for each cell within the rebalancing region. Mass conservation for
each cell must be obtained by solving the pressure equation for each cell. As
rebalancing makes large-scale pressure corrections rapidly by direct solution,
it reduces the number of iterations required to solve the norraal pressure
equations on a cellwise basis. In this regard, rebalancing is perforraed at the
beginning of every IREBIT-th inner Iteration, where the value of IREBIT is
supplied by the user. Since rebalancing requires some work, there is a tradeoff
involved in selecting a value for IREBIT: rebalancing too frequently may
decrease the CPU time spent on inner iterations at the expense of excessive CPU
tirae spent in actually performing the rebalancing.
124

Distribution for ANL-90/45

Internal:

c. H. Adams P. L. Garner (50) M. J. Tan


G. Birgersson E. M. Gelbard A. M. Tentner
R. N. Blomquist K. C. Gross J. H. Tessier
G. L. Bordner Kalimullah C. E. Till
M. Bottoni K. J. Miles R. B. Turski
J. E. Cahalan M. Minkoff C. P. Tzanos
T. R. Canfield D. Mohr J. A. Verklan
Y. S. Cha E. E. Morris R. B. Vilim
Y. W. Chang H. M. Park D. C. Wade
T. H. Chien H. P. Planchon D. K. Warinner
G. H. Chisholm F. G. Prohammer T. Y. Wei
D. H. Cho J. Reifman R. A.
Wigeland
E. V. Depiante R. C. Schmitt B. S.
Yarlagadda
H. M. Domanus W. T. Sha ANL Patent Department
F. E. Dunn Y. W. Shin ANL Contract File
E. E. Feldman J. J. Sienicki TIS Files (3)
E. K. Fuj ita R. M. Singer

External:

DOE-OSTI for distribution per UC-542 (68)


ANL-TIS Libraries
Chicago Operations Office, DOE:
A. L. Taboas
J. L. Hooper
DOE Headquarters, Washington, DC:
P. J. Dirkmaat
P. B. Hemmig
J. D. Nulton
J. Berkow, Bechtel Power Corporation, San Francisco, CA
P. Kroeger, Brookhaven National Laboratory
G. Van Tyle, Brookhaven National Laboratory
D. Grand, CEA-Centre d'Etudes Nuclealres de Grenoble, France
M. Parent, CEA-Centre d'Etudes Nuclealres de Grenoble, France
T. Charlton, EG&G Idaho Inc., Idaho Falls (5)
S. Z. Rouhani, EG&G Idaho Inc., Idaho Falls
E. Brega, ENEL-Thermal and Nuclear Research Center, Milan, Italy
M. McDowell, Energy Technology Engineering Center
A. Shenoy, General Atomics, San Diego, CA
P. MaGee, General Electric Company, San Jose, CA
J. A. Turner, Los Alamos National Laboratory
N. E. Todreas, Massachusetts Institute of Technology, Cambridge
M. Oszewski, Oak Ridge National Laboratory
S. Aslam, University of Illinois, Urbanna
S. Gallopoulos, University of Illinois, Urbana
M. Berzins, University of Leeds, England
N. Aslara, University of Michigan, Ann Arbor
J. Doming, University of Vlrglna, Charlottesville
Rizwan-uddin, University of Virginia, Charlottesville
09283
V-

You might also like