Nuclear Engineering and Design 62 (1980) 1-24

©North-Holland Publishing Company

AN OVERVIEW ON ROD-BUNDLE THERMAL-HYDRAULIC ANALYSIS

William T. SHA
Components Technology Division, Argonne National Laboratory, 9700 South CassAvenue, Argonne, IL 60439, USA

Received July 1980

Three methods used in rod-bundle thermal-hydraulic analysis are summarized. These methods are: (1) subchannel analy-
sis, (2) porous medium formulation with volume porosity, surface permeability, distributed resistance and distributed heat
source (sink) and, (3) bench-mark rod-bundle thermal-hydraulic analysis using a boundary-fitted coordinate system. Basic
limitations and merits of each method are delineated.

1. Introduction

Most heat exchangers and reactor fuel assemblies employed in the power industry have rod-bundle geometry.
Fluid flow and heat transfer in a rod bundle are complex phenomena, and basic understanding of these phenomena
is essential to achieving optimum design performance during normal operating conditions and maintaining struc-
tural integrity during off-normal operations.
The traditional way of obtaining the detailed performance information is to build and test a scaled model of the
heat exchanger under consideration. The heat-giving fluid is circulated through the shell or the tube side of the
heat exchanger, and the heat-receiving fluid through the other side. With a reactor fuel assembly heat sources must
be employed, but no heat-giving fluid is required. In both cases, velocity and temperature measurements can be
made over the desired range of operating conditions. Based on these measurements, correlations of heat transfer
coefficients and pressure drops as functions of Reynolds number, geometry ... etc. are developed, and subse-
quently, these correlations are used for designing the full scale apparatus. This approach, however, suffers from
several drawbacks: for each new geometry considered, a new test model must be constructed, and scaling from
the test model of design size often poses some concern. Furthermore, expenses, instrumentation and measuring
technique usually limit data to global heat transfer and flow rates, while detailed temperature and velocity distri-
butions are needed for an optimum design. Finally, correlations based on these measurements are valid only in
the range of parameters tested; additional experimental measurements are often required when the operating con-
ditions fall outside the range of these correlations.
An alternative approach is to develop direct numerical solutions of conservation equations of mass, momentum,
and energy, using appropriate initial and boundary conditions of the physical system under consideration.
However, a major obstacle to the use of numerical methods in rod-bundle thermal-hydraulic analysis has been
the complex geometry. Recently, this obstacle has been overcome by using the boundary-fitted coordinate trans-
formation. Moreover, with the anticipated improved performance of the next generation of computers and further
advances in numerical solution technique, detailed numerical simulation of rod-bundle thermal-hydraulic behavior
appears practical and attractive. Accordingly, the objectives of this paper are: (1) to summarize pertinent methods
employed in rod-bundle thermal-hydraulic analysis; and (2) to delineate basic limitations in each method.
2 W. 7: Sha / Rod-bundle thermal-hydratdic analysis

2. Methods used in rod-bundle thermal-hydraulic analysis

The rod-bundle thermal-hydraulic analysis discussed in this report is limited to solution methods for conserva-
tion of mass, momentum, and energy equations considered as an initial-value problem in time and a boundary-
value problem in space. It also is limited to descriptions and performance of pertinent computer codes available in
the open literature. In general, these methods can be classified into three broad categories: (1) subchannel analysis;
(2) porous medium formulation; and, (3) benchmark rod-bundle thermal-hydraulic analysis.

2.1. Subchannel analysis

The subchannel method is the most widely used in rod-bundle thermal-hydraulic analysis to date. Historically,
and as originally intended, subchannel analysis has been used primarily for design calculations; it explicitly assumes
that one of the velocity components (axial velocity in z-direction) is dominant, compared to components in the
other directions (crossflows in x- and y-directions). Thus, it is advantageous from both physical and computational
points of view to treat axial and transverse momentum equations separately, so that simplifications can be applied
to the transverse momentum equations.
Untit recently, subchannel analysis also has been employed in reactor safety analysis; e.g., to resolve flow-block-
age problems in a fuel assembly. In this case, the usual assumptions for small cross-flows in transverse directions
are no longer valid, and numerical solution techniques for the initial value problem cannot be applied. A typical
subchannel arrangement of a 19-pin rod bundle, and control volumes for axial and transverse momentum equations
are shown in figs. 1a, 1b, and 1c respectively. The transverse momentum equations used in the THI3D [ 1], COBRA-
IIIC [2], SABREd [3] codes are listed in table 1.

EXACT CONTROLVOCUME
FOR TRANSVERSE
~LI4 EQUATIONS

CONTROL VOLUME FOR

a ~AXlAL MOMENTUMEQUATION

APPROXIMATED CONTROL
VOLLME FOR TR/~ISVERSE
t't3"IENTt~ EQUATION

Fig. l.(a) Typical subchannel arrangement, (b) control volume for axial momentum equation, (c) control volumes for transverse
momentum equations.
 W.T. Sha / Rod-bundle thermal-hydraulic analysis

Table 1
Comparison of transverse momentum equations used in various computer programs a

TH13D
z~w
Ow m.. ~uimij aP
0=- q + - S--+ Fij
~z ax i ~x i

COBRA-IIIC ~--x2,u
ami/ _ Ow*mij S
at
- -
3z
+ ~ (Pi - Pj) - Fij Xl,U
/
SABRE-1 a
~Ouiw* ~pu~ OP 02ui
0= X--
bz bx i ax i + # bz---
~- + R i

COMMIX- 1
OOui O'yzPUiW bq/jPuiu j , ~P ~ 3u i ~ ~u i
"YV 0~- - Og Oxj 7Vi!ox--~+-~Tzt~ O~ + Ox~q'/l~Ox--~ + R i

1 2 3 4 5 6 7

Computer 1 2 3 4 5 6 7
program

COMMIX-1 yes yes yes yes yes yes yes

THI3D-1 no yes yes yes no no yes
COBRA-IIIC yes yes no yes no no yes
SABRE-1 no yes yes yes yes no yes

Note: Fij = force per unit area; h = geometrical factor to account for the difference between the exact control volume and the
approximated control volume for the transverse momentum equation [see fig. lc].
a The latest version of SABRE code has transient capability.

2.1.1. Basic l i m i t a t i o n s
(1) The fine structure o f b o t h velocity and temperature within a subchannel is ignored ( l u m p e d parameter ap-
proach).
(2) The transverse m o m e n t u m equations cannot be treated with the same rigor as the axial m o m e n t u m equa-
tions because of the n o n o r t h o g o n a l characteristics o f subchannel arrangement. Specifically, the finite-difference
representation o f the second derivatives cannot be rigorously evaluated.
(3) T o facilitate calculations, a p p r o x i m a t i o n s are necessary for interfacing the required i n f o r m a t i o n at various
locations b e t w e e n control volumes for the axial and transverse m o m e n t u m equations.

2.1.2. M a t h e m a t i c a l m o d e l
The following governing equations, b o u n d a r y conditions, and solution t e c h n i q u e are r e p r o d u c e d f r o m ref. [ 1 ] ;
extension to transient fornulations is straightforward.

2.1.2.1. Conservation of mass

N
~mz, i = _ ~ m i i , (1)
~z j=l
4 W.T. Sha / Rod-bundle thermal-hydraulic analysis

where

mz = pwAz = axial flow r a t e ,

/9 = coolant density,
w = axial velocity,
Az = axial flow area,
mq = Puiiu4ij/Az = diversion crossflow rate per unit height between subchannels i and j (wire-wrap sweeping cross-
flow can be included),
uij = transverse velocity in xi-direction,
A 6 = crossflow area between subchannel i and j,
Az = axial increment,
N = number o f neighboring subchannels.

2.1.2.2. Conservation o f momentum. As mentioned before, the m o m e n t u m equation is subdivided into the axial
momentum equation and the transverse momentum equations so that the latter can be treated with simplified
assumptions:

(1) Axial momentum

3P~z,i
3Z j=l
I (m )2(iz
[mijw. + (mii + mij)(wi _ w/)] _ Pig +
, 2DPi
+__
k +_,'-Az"3 1
2PiAz Az, iaz Pi
1 aAzi]~ A
piAz, i ~z' JJ z,i,

where

g = gravitational constant,
f = friction factor,
k = form-loss coefficient,
D = equivalent hydraulic diameter,
P = pressure,
m~i = turbulent crossflows between subchannels i a n d j per unit height, including natural turbulence and turbulence
promoted by the presence of wire-wrap,
Az
Az = f A z(z) ee/ z,
o

w* = ÷ wj)
and the asterisk denotes the quantity associated with the donor subchannel.

(2) Transverse momentum

If the interactions between the transverse momentum in the direction under consideration and the other trans-
verse direction (or directions) are neglected, 1 the transverse momentum equation m a y be written as

amqwi + ~milui. = ( _ _3P

_ + Fxj~ =
3z 3x i 3xj V )Sj, j 1,2 ..... N , (3/

1 An attempt to include both the transverse momentum flux gradients and their interactions in principal directions has been for-
mulated and presented in THI3D-1 [ 1].
 W.T. Sha / Rod-bundle thermal.hydraulic analysis 5

where
x / = transverse direction along the centroids of subchannels i and j,
V = elementary volume,
S/= crossflow width between subchannels i and j.
The general surface force term Fxj may be approximated by
~cl mij I mij V ,
Fx/- 2piS ? A~xi
where
= crossflow resistance,
&xj = centroid distance between subchannels i and j in the xj-direction.

2.1.2.3. Conservation energy

N t

~hi- 1~ ~ F SJ ~(Tj - Ti)+8(mij+m~j-c}/)(hi-hi)]+(q 7' 8lqd'ilAz'i (4)

~z mz, i j=l t_z~j ~ ]mz, i '
where
h = coolant enthalpy,
T = coolant temperature,
k = thermal conductivity of coolant,
ci[t ' = turbulent-exchange coefficient per unit height,
q = heat generation per unit volume per unit time,
q~ = heat loss per unit height per unit time through the duct wall,
A d = heat-transfer area through the duct wall,
6 = 0 for crossflows leaving channel i to j,
= 1 for crossflows entering channel i from j,
8 ~ = 0 for channels located away from the duct wall,
--- 1 for channels located next to the duct wall.
Both m~j and c~j can be combined into a single turbulent exchange coefficient (i.e., am~j = m~)"- c~j, where a is a
constant, since mi7 and c~j cannot be distinguished through experimental measurements),

2.1.2.4. Boundary conditions. The pressure-drop boundary conditions is used. Inlet pressure, coolant densities, and
temperatures (or enthalpies) of all subchannels are assumed to be known. Inlet pressures can be specified as either
uniform or nonuniform for all subchannels, but outlet pressures must be uniform. To satisfy the uniform outlet
pressures, inlet velocities of all subchannels must be adjusted, using the following iterative solution technique.

2.1.3. Solution technique

Coolant parameters are calculated simultaneously for all subchannels, by starting at the bottom of the core and
moving upward, step-wise. Coolant enthalpies (temperatures), pressures, densities at the inlet, and heat-flux distri.
butions in all subchannels are known. Inlet velocities are first assumed to be known, and then solved iteratively.
Crossflows are computed based on coolant properties at the bottom of the step length. Knowing the crossflows
within the step and the heat addition into the step, the change in coolant enthalpy (temperature), pressure, density,
and velocity can be determined successively for each step length. To account for the effect of local pressure drop
on coolant properties, an iteration within each step length must be performed. Thus, a consistent set of local
thermodynamic properties of the coolant can be maintained.
6 W.7~ Sha / Rod-bundle thermal-hydraulic analysis

The resulting enthalpy, pressure, density, and velocity at the top of the step length are used as input to the next
step length. This procedure is continued until the top of the core is reached. Accordingly, the pressure at the top
of the core may be expressed as a function of inlet pressures, densities, and velocities of all subchannels:
• • i " " "
P~=P~/(w],w~,...,wM;p],p~ ,...,p/l;PI,P~,...,P~if), j= 1, °-, .... M , (5)

where superscripts i and o correspond to inlet and outlet values; P, w, and p are pressure, axial coolant velocity,
and density, respectively; and M is the number of subchannels under consideration.
The total derivative of eq. (5) can be written as

de 7 = k=-,\3
C k dw , + -Opik
- doj, + O--~'kdPk)" (6)

For practical purposes in reactor design, both the inlet pressure and the coolant-density distribution are assumed
to be known. A set of inlet velocities must be determined to satisfy the uniform pressure at the exit of all subchan-
nels. Basic equations for conservation of mass, energy, and momentum are used to compute the outlet pressure dis-
tribution for a given set of inlet coolant velocity conditions. Interaction between subchannels is taken into account
by allowing for crossflow through adjacent subchannels. The amount of crossflow is expressed as a function of
pressure and of temperature or enthalpy gradient of the adjacent subchannets at the same elevation. Eq. (6) can
now be expressed in the matrix form
[dP °] = [B] [dwi], (7)
where
[dP °] = column matrix with elements dP~,
[B] = M X M Jacobian matrix with elements OP~/Ow~,
and
[dw i] = column matrix with elements dw~.
If [B] is a nonsingular matrix, eq. (7) can be written as

[,',w ~1 = [B]-' [ r e °] . (8)

Eq. (7) couples with the following boundary conditions:

At the top of the core:
P~=constant, j=l,2 ..... M .
At the bottom of the core:

(1) P) = constant, or a known distribution, j = 1,2 ..... M ,

(2) T} = inlet temperature or h} (inlet enthalpy) are known at j = 1,2 ..... M ,
and N

(3) ~ p}Ai,lw~ = constant,

j=l
where A ~,j i s the axial coolant flow a r e a a t the inlet of subchannel j .
Boundary condition (3) can be considered a constraint of eq. (8), which can be written:
[Aw i] = [B -1 ] [AP O] + ~k[B-1 ] [I] , (9)
 W. T. Sha / Rod-bundle thermal-hydraulic analysis 7

where [I] is a unit column matrix, and )~ is a scalar or Lagrangian multiplier to be determined.
Multiplying eq. ( 9 ) b y [A] and [p] gives

X= [A] [p] [B-'] lAP°] (10)

[A] [p] [B-'] [11 '
where [A ] is a row matrix of the axial density flow area at the inlet, [p] is a square diagonal matrix of the coolant
density at the inlet, and [.4] [p] [Aw'] = 0.
In practice, the inlet velocity of the first subchannel is changed by a preassigned small magnitude, and resulting
changes in outlet pressure [AP°] for all subchanne]sare then determined. Elements of the first column of the
Jacobian matrix [B] are determined from this calculation. The originalvelocity distribution is restored, and the
procedure is repeated for each subchannel. The inverse of matrix [B] is then determined, and X is calculated using
eq. (I0). Knowing ~., [Awi] in eq. (9) is then determined. The new inlet velocity distribution can now be obtained
from

1
INPUT ~ , Tj, hj, p1,
j HEATFLUX
i
DISTRIBUTION AND INITIAL VALUES OF w.
]
SUCH THAT E w! 0 i. A i. = C
j J ] J

Z= 0

Z = Z+AZ

PEHFOP&I CALCULATIONS OF CONSERVATION OF

['lASS, ENERGY, AND MO[,[ENTIJ~.[ A t ~ TRANSVERSE

INTERACTIONS BE~4EEN SUBCHANNELS

NO

~ YES

Fig. 2. Simplified computational logic diagram of subchannel analysis.

8 W. rE. Sha / Rod-bundle thermal.hydraulic analysis

W~EW= [Wi] + [Awi] , (1l)

where wN{'W and [w i] are column matrices of the new and original inlet axial velocities, respectively.
A simplified computational logic diagram of subchannel analysis is presented in fig. 2.

2.2. Porous medium formulation

A set of quasi-continuum governing equations for conservation of mass, momentum and energy for a finite con-
trol volume is derived from both integral and differential approach. The integral approach is presented here, the
differential approach can be found in ref. [4]. The system consists of a singlephase fluid with discrete stationary
solid structures. Volume-porosity, surface permeabilities, distributed resistance and distributed heat source (or
sink) are systematically included in the derivation.
The concept of volume-porosity and distributed-resistance and heat-source arises naturally in the local volume
averaging process [5,6]. The size of control volume used in the local volume average must be small, compared to
the large-scale phenomena of interest. Distributed resistance alone is normally used to characterize anisotropy of a
porous medium. However, when local flow area (or path) changes abruptly, and high resolution of local tetaapera-
ture and velocity distribution is needed, additional delineation of the anisotropic characteristics of the medium is
necessary. Thus, a new approach with volume porosity, surface permeability, and distributed resistance and heat
source is developed. The concept of surface permeability is new in porous medium formulation [4,7], and it greatly
facilitates the modeling of anisotropic effect of a medium. Now, the anisotropic characteristics of rod bundles can
be modelled by using appropriate local surface permeability and disstributed resistance. It is important to note that
the concept of surface permeability, when properly applied, will reduce the inaccuracy in the calculated volume
averaged velocity and temperature due to replacing the discrete fluid resistance and heat source (or sink) of the
physical system by an idealized distributed system. The surface permeability is usually known, while the distributed
resistance is, in general, not well defined.
Recently, the COMMIX-1 code, which employs the porous medium approach with volume porosity, surface
permeability, distributed resistance and distributed heat source was developed. The typical mesh structure for a
rod bundle used in the COMMIX-1 code is shown in fig. 3. The transverse momentum equations used in the
COMMIX-1 code are presented in table 1 along with those of the other codes for comparison.

Fig. 3. Typical transverse positions of a rod-bundle used in the porous medium formulation.
 141.T. Sha / Rod-bundle thermal-hydraulic analysis 9

' Vi

Fig. 4. Physical system consisting of a total volume V of solid and liquid, enveloped by a surface A ; some of the dispersed solids
may be cut through A.

2.2.1. Basic limitations

The unique advantages of the volume-porosity, surface-permeability, and distributed resistance and distributed
heat source approach presented in this section are the use of orthogonal coordinates and geometrically similar con-
trol volume. Thus, limitations (2) and (3) employed in the subchannel analysis are eliminated; however, limitation
(1) is still retained.

2.2.2. Derivation o f quasi-continuum governing equations

A region R of finite volume V in space with enveloping surface A is considered. There are a finite number of
dispersed, fixed heat-generating solids inside V, some of which may be cut through by A as illustrated in fig. 4.
Clearly, V = Vf + Vs, where Vf is the total fluid volume and Vs is the total solids volume. In general, only a frac-
tion of the enveloping surface A is unobstructed to fluid flow.
To facilitate further discussion, the following variables are defined:
ni = outdrawn unit normal vector to an arbitrary surface element dA.
vi = local fluid velocity vector, either interior or on surface.
p = fluid density.
7v = local volume porosity; i.e., fraction of the local infinitesmal volume element dV that is occupied by the fluid.
It may take on values between 0 and 1. If dV is completely inside a dispersed solid, 7v = 0; if it is completely
in the fluid, 7v = 1. If it is partly in a dispersed solid and partly in fluid, then 0 < 7v < 1. Hence, in general,
0<%<1.
")'A = local surface permeability; i.e., fraction ot r the local surface element dA that is unobstructed to fluid flow.
Following the reasoning similar to that given for 7v, we conclude that 0 ~< 7A ~< 1.
Af = unobstructed area in A that is available for fluid flow.
Clearly,

Af = fA'~A dA,
V f = f v ' ) ' v dV, and
A f s = total solid-fluid interface area in R. Note that it includes portions of the solids surfaces cut through by A
andA :/=Af + A f s .
I0 W.T. Sha / Rod-bundle thermal-hydraulic analysis

2.2.2.1. Conservation of mass. Tire principle of mass conservation states that the rate of increase of fluid mass in 1/,
plus tire net mass outflux through A must vanish (in the absence of mass source). Hence,

at
V
3'vdV+ f
A
Pvini'), A dA --0, (12)

which is the continuity equation in the integral form.

2.2.2.2. Conservation of momentum. The principle of conservation of linear momentum states that the rate of in-
crease of linear momentum of the fluid mass in V, plus the net momentum outflux through A equal the sum of the
body force and surface force acting on the fluid mass.
Thus,

at
V A V Af+A fs A f+Afs

where X i is the body force per unit mass of the fluid. The solid surface is assumed to be impervious to fluid flow
and the surface force is decomposed into components due to pressure and due to viscosity. It is important to have
proper limits assigned to the various integrals.
The sum

.... f pn ida + f rjinj dA

A fs A fs

represents the total drag force acting on the fluid mass in V due to the presence of the solid. Hence, it is convenient
to define an equivalent dispersed resistance per unit volume of the fluid Ri, such that

-f PnidA+ f rjinidA-- fRiTvdV. (14)

Afs Afs V

The negative sign on the right hand side of eq. (14) is introduced in agreement with the convention that such resis-
tance retards the motion.
Using the relationship in eq. (14), eq. (13) can be rewritten as

o_
at
f,v.uvdV+ fov.vjnjvA dA= f,x,VvdV- f pn, dA+ f ..n. aA - fRiv dV
V A V Af Af V

= f p x i T ~ d V - fyAPnidA+ fyAr/inzdA- fRiYvdV, (15)

V A A V

which is the desired integral linear momentum equation for a fixed control volume V with enveloping surface A
containing fixed, dispersed solids.
For turbulent flow, all velocities should be replaced by their temporal means and all stresses by their effective
values "/'eff-

2.2.2.3. Conservation of energy. The K.E. equation in differential form is applicable everywhere in the continuum
fluid mass:

Vi a(viJrji ) av i
+ pviXi + • (16)
at axi axi +p ~ - '
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 11

where
q2 = ViVi'
n/i =pS/i + r/i (6ji = Kronecker delta),
Ovi
d~ = rii Ox--~"
In eq. (16), qb is the dissipation function; it represents the rate of irreversible dissipation of mechanical energy into
heat per unit volume of the fluid as a result of fluid friction.
Multiplying eq. (16) by dVf and integrating over the entire fluid volume Vf gives

f \2 ]dgf+ f - - dgf = f , v , X , d g f + f O(ViTr/i)dgf+ f p ~~vi

x i d V f - f ¢~dVf, (17)
Vf
Ot Vf
Oxi Vf Vf
Oxj Vf Vf
where dVf = 7v dV, dAf = 7A dA. Since Vf is enclosed by A f + Afs ,

f Oxi dVf=f ¢q___zvlnidA=

2 f ~f-vinidAf (SinceAfsisimpervioustoflow)
Vf A f+A fs Af

_Fpq2
+ ff)--~- vini7A dA (Since.TA = 0 for the portion of A which is notAf)
A

and

f ~(viTr/i)dVf = f rift/in ] dh = f rift~in/dZf = f VitiTA dA (t i =--7rjinj, t i is the stress vector).

Vf ~Xj A f +A fs Af A
Consequently,

f pq2
~fvj -~-7~dV+f ~pq2 vin,vAdA=f ,v,X,7~dV+f vitivAdA+f paV' dv- f~v~ dV, (18)
A V A V axi V

which is the kinetic energy equation in integral form for the fluid mass in V.

2.2.2.4. Conservation of energy

(1) Total energy equation
The third basic principle of mechaniscs is that of conservation of energy: the rate of increase of the total energy,
i.e., sum of the internal and the kinetic energy of the fluid mass inside a fixed volume V, plus their outflow rates
through A, equal the rate of work done by the body and surface forces on the fluid inside V, plus the influx of heat
to the fluid and any extraneous internal heat source (or sink) such as that due to joule heating, chemical reaction,
radioactive decay, etc.
Denoting the sum of the internal and the kinetic energy per unit fluid mass by E (= e + q2/2, e being the inter-
nal energy), and applying the principle of energy conservation enables us to write:

~-f E%dV+foEviniYAdA= fpviXiT~dV+fvitiTAdA - f q,n, aA+ f O v ~ d V , (19)

A V A Af+Afs V
12 W.T. Sha / Rod-bundle thermal-hydraulic analysis

where (~ denotes the extraneous internal heat source (or sink) per unit volume of the fluid. In writing eq. (19), the
following relationships have been used:
f o E dVf = fpE~'vdV (Since dVf = 7v dV, and 7v = 0 for the portion of V which is not Vf),
vf v

. l pEvini dAf = IpEviniT.4 dA (Since 7.4 = 0 for the portion of A which isnot A f),
Af A

vitidA = f vitidA (sincev i---0OnAfs) = fvitiT.4 dA.

Af+Afs .4f A
(2) Internal energy equation
Subtracting eq. (18) from eq. (19) gives
_ ¢ ~vi
fpe~/vdg+ fpeviniT.4dA=- j p - z - - y v d V - f qinidA + f ( Q + @ ) ~ / v d V . (20)
Ot v A v OXi Af+Afs V
The surface integral f qini dA can be decomposed into:
Af+Afs

(a) f qini d A f = - f ~c -or

-n i dAf = - ni~"A dA , (21)
Af Af Oxi A t

where the Fourier law of isotropic conduction is used and k is the thermal conductivity of the fluid.
(b) - fAfsqin i dAfs represents the rate of heat release at all fluid-solid interface inside V since n i is the out-
drawn unit normal vector from the fluid into the solid. In reactor cores or heat exchangers with rod bundle, the
integral merely denotes all of the heat released or removed at the immersed rod bundle surface. Hence, it is con-
venient to define an equivalent, dispersed heat source (or sink) per unit volume of the fluid; Orb, such that

-- f A f s qini dmfs ~ f v QrbTv d V . (22)

Using eqs. (21) and (22), eq. (20) can be rewritten as

-a- foe'r dV+ a.,

fpevini'~.4 dA=- fp ~xiVvdV + ori ni7.4 dA + f(Q,b
~x + Q +*)%dV. (23)
Ot V A V A V

The dissipation function (I) is given by vii(Ovi/Oxi) , ~4hich for Newtonian flow, is:

+\Oy
2
Oz! +~oz Ox] - 3 \ 3 x +oy--+ Oz! J
t2t
An alternate expression is:

+3L\ax ~y ] \ ax - ~z /
wi 71
\ ~y - ~z / J J
(3) Enthalpy equation
The energy equation may be alternatively given in terms of enthalpy: h = e + (p/p), which is again defined on
the basis of unit fluid mass. Substituting this relationship into eq. (23) gives

a fphTv dV+ lphviniTA dh - ~~fPTv dV- f p v i n i T A dA

3t v A A
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 13

£ 3vi dV+ C Or aA + f(orb + 0 + ®) dV. (24)

= - vj p o x i % aA oxiniTA V 7v

Since,

(a)-~fpTvdV= fTvOP dv,

v V
Ot

(b) fpvini7 AdA = f Pvin , dA (Since (i) 7A = 0 for that portion of

A Af+Afs A which is not Af; (ii) v i = 0 on Afs )

(Since Af +Afs is the

Vf 0 i enveloping surface of Vf)

= J-~-" 7v dV.
v Oxi
ap op O(PVi)
= Ot + vi Oxi -- Ot + 3x i P~.
It follows then,

~t foh, dV+ Afloh ,n,'rAaA= f-V v dV+ aA ~aT (25)

v i niTA dA + f(Orbv + 0 + qs) % dV,

which is the integral energy equation in terms of enthaply.

2.2.3. Initial and boundary conditions

Initial conditions are obtained either by specifying the fluid velocity, temperature and pressure distribution
throughout the interior points of the space under consideration, or by continuing a previous computer run via a
restart capability. Once the fluid temperature and pressure are specified, as in the former case, the corresponding
density and enthalpy are calculated from equations of state. Boundary conditions are specified with appropriate
temperature, velocity, and heat flux distributions at all boundaries.

2. 2. 4. Numerical technique
The quasi-continuum governing equations, eqs. (12), (15) and (25) are finite-differenced in a staggered mesh sys-
tem. The pressure (p), temperature (T), enthalpy (h), density (p), and volume porosity (Tv) are defined at cell
centers. The velocity v and surface permeability (Tx, 7y, and 7a in x-, y- and z directions, respectively) are defined
at the center of each cell surface. The finite-difference equations are solved by using ICE (Implicit Continuous-fluid
Eulerian) technique. A simplified computational logic diagram of COMMIX-1 is shown in Fig. 5. Some pertinent
numerical results obtained from the COMMIX-1 code are presented in the paper by Domanus, Shah, and Sha in
the current special issue.

2.2.5. Unique features

The COMMIX-1 computer program has two unique features: one is that the model is capable of treating both
continuum (e.g., reactor plenum, piping mixing) and quasi-continuum (e.g., rod bundle or fuel assembly) systems
by introducing volume porosity and surface permeability, as well as distributed resistance and heat source. This
capability represents the most cost-effective way of handling two classes of problems by a single computer program.
Furthermore, it is to be noted that the concept of surface permeability is new in porous medium formulation, and
14 W. T. Sha / Rod-bundle thermal-hydraulic analysis

<
t = 0 ]

[INITIALIZATION]

I SP~CIEYB. C. I
• I t = t+ ~~ ]

EVALUATE EXPLICIT TE~MS IN MOMENTUM EQS.

l SOLVE ENERGY EQ. FOR ENTHALPY;

COMPUTE TEMPERATURE & DENSITY

I ° I
I ITER = ITER + i I
J
+
I COMPUTE MASS RESIDUE FROM CONTINUITY EQ.

l I
COMPUTE VELOCITIES FROM MOMENTUM EQS.

Fig. 5. Simplified computational logic diagram of COMMIX-1.

it greatly facilitates accounting for the anisotropic effects in a medium. The other is the inclusion of the local mass
residue effect [5] in the energy and momentum equations. As a result of this inclusion, both the convergence rate
and accuracy are greatly improved,

2.3. Benchmark rod-bundle thermal-hydraulic analysis

The fine structure of both velocity and temperature within a subchannel or a computational cell is ignored in
the subchannel-analysis and the porous medium approach. However, this fine structure can be accounted for through
the specification of rods as internal boundary conditions. Recently, BODYFIT-1, a three-dimensional, transient,
single-phase component computer program, [8,9] has been developed. It utilizes boundary-fitted coordinates to
transform a complicated rod-bundle geometry into a rectangular coordinate mesh system as shown in fig. 6. Note
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 15

C/ ~--'-A
PHYSICAL PLANE AND DIIIENSIONS
'{

' q

i
IC IRANSFORMED PLANE

Fig. 6. Physical and transformed coordinate systems.

that all internal fuel rods are transformed into slits, slabs, or squares. The governing equations of interest are trans-
formed accordingly into the new variables (transformed space) before being approximated by finite-difference
equations which are solved in the usual manner [7].
Accurate representation of boundary conditions is best accomplished when the boundary is coincident with a
coordinate line, so that the boundary can be made to pass through the points of a finite difference grid constructed
on the coordinate lines. Finite-difference expressions at or adjacent to the boundary may then be applied, using
only grid points on the intersections of coordinate lines. Interpolation between grid points is not required, and this
is particularly important for boundaries with strong curvature, such as fuel rods.
In the Navier-Stokes equations, the boundary conditions are the dominant influence on the character of the
solution; therefore, the use of grid points that are not coincident with the boundaries places the most inaccurate
difference representation in precisely the region of greatest sensitivity. The generation of a curvilinear coordinate
system with coordinate lines coincident with all boundaries is thus an important aspect of a general numerical solu-
tion of the Navier-Stokes equations.
As mentioned before, applications to the rod bundle or reactor fuel assembly, as shown in Fig. 6, is the primary
interest of coordinate transformation. Thus, the transformation is basically two-dimensional. However, the trans-
formation outlined below is very general; it can apply to any boundary shape and can be readily extended to three-
dimensions. (A general treatment of the time-dependent movable boundary problem and the control of coordinate
system is discussed in ref. [10].)
16 W.T. Sha / Rod-bundle thermal.hydraulic analysis

2.3.1. Mathematical preliminaries

Two-dimensional elliptic boundary value problems are considered. The general transformation from the physical
plane (x, y ) to the transformed plane (~, r/) is
~ = ~(x,y), n=n(x,y) t~_~,)
and the inverse transformation is
x = x(~, r~), y =y(~, r/). (27)
The Jacobian of the transformation is

J = J ( ~ , Y )= x~y n - x n y ~ -~ O, (28)

where
~x = y ~ / J , ~y = - x n / J , ~x = - Y U J , ~Ty = x U J , (29)
fx = (Ynf~ Y~fn)/J, f y = (x~fn - x n f O / J , (30)
f x x = (VnJ~
2 , _ 2y~ynf~n +Y~fnn)/J2 + [(YnY~
2 - 2y~YnY~n + y ~ y n n ) ( x n f ~ _ x~fn )

+ ( y ,2x ~ - 2y~ynx~n + y ~ x n n ) ( y f f n - y n f ~ ) l / j 3 ,
and

fyy = ( x ~ f ~ _ 2x~xnf~n + x~fnn)/j2 + [ ( x 2 y ~ _ 2x~xny~n +x~ynn)(xnf~ _ x~fn )

+ (x~x~ 2x~xnx~n + X~X,r~)(y2~f. - ynf~)]/J 3 • (31)
Higher derivatives can be obtained by repeating the above operations.
Two tasks are involved in the transformation: One is to find the interior physical points after specifying the
physical boundary at a number of discrete points. The other is to transform the partial differential equations of
interest into the new variables before being approximated by the finite-difference equations.

2.3.2. Mapping o f physical domain [10]

The choice of this mapping is largely dependent on its simplicity and effort required for a desired accuracy.
Without loss of generality, a simply-connected region will be considered first, and a doubly-connected region will
be discussed in subsequent sections.

2.3.2.1. Simply-connected region. The boundary of.the physical domain (x, y ) is specified at discrete points cor-
responding to boundary points in the transformed plane (~, r/). It is desirable to have a prescribed, convenient
mesh in the (~, r/) plane, therefore, (~, r/) must be used as independent variables. Their values are governed by any
suitable elliptic partial differential equations as a boundary value problem. The simplest choice appears to be that
~, ~7must satisfy the Laplace equation in the physical plane:
V2~ = O, (32)
Vz~ = 0 , (33)
where
32 ~2
- 0x 2 + 0y2 •

The dependent and independent variables can be interchanged by applying eqs. (29)-(31). The transformed
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 17

r"2

F3
r4

Physical Plane
,q
r;
: % /4 i
I
I"4* t .eI ',t r;

Transformed Plane
Fig. 7. Field transformation: single body.

equations are:
ax~ - 2fJx~n + ")'xnn = 0 , (34)
o c y ~ - 213y~n + 7 Y n n = 0 , (35)
where
a = x~ + y ~ , 13 = x ~ x n + y ~ Y n , 7 =x2
n +y~. (36)

Eqs. (34) and (35) are clearly coupled quasi-linear elliptic equations, and only in special cases (x~ = Y n and
x n = y ~ ) can they be reduced to Laplace equations for which mapping is conformal.
Eqs. (34) and (35) can be conveniently solved by the finite-difference method, using second-order central dif-
ferences with successive over-relaxation (SOR) of the dependent variables. Discrete values of (x, y) at the cor-
responding point (~, 7) are thus determined.
In general, the numerically calculated boundary may not coincide completely with the given boundary at inter-
mediate points. This is not an essential limitation since only discrete points are needed in the basic principle of the
finite difference method. The finer the mesh, the smaller is the numerical error.

region. Consider the transformation of a two-dimensional, doubly-connected region D

2.3.2.2. D o u b l y - c o n n e c t e d
bounded by two simple, closed, arbitrary contours onto a rectangular region D* as shown in fig. 7. The basic trans-
formation is discussed here, assuming that the body contour and outer boundary are transformed, respectively,
to the constant q-lines forming the bottom and top sides of the transformed region. The more general case of seg-
mented body contours transforming to any side of the transformed region follows analogously. Let F 1 map onto
P~, F2 map onto F~, P3 onto P~, and P4 onto P~. For identification purposes, region D will be referred to as the
physical plane, D* as the transformed plane, and P~ and P2 as the body contours. Note that the transformed
boundaries (P~ and P~) are made constant coordinate lines (r~-lines) in the transformed plane. The contours Fa and
P4, which connect the contours F~ and F2, are coincident in the physical plane and thus constitute re-entrant
boundaries in the transformed plane.
18 W. T. Sha / Rod-bundle thermal-hydraulic analysis

Let ~(x, y) and rl(x, y ) be continuous functions in D, then

V~ = 0, (37)
V2r~ = 0 , (3S)
with the Dirichlet boundary conditions

i+lF+.1
=

L~J L~
, [x,y]~rl, (39)

ill:Lv2 = , [x, y] E F 2 , (40)

where/)1 and rh are different constants (72 > 7/i), and ~l(x, y ) and ~2(x, y ) are specified monotonic functions on
Fa and F2, respectively, varying over the same range. The arbitrary curve joining F~ and 1-'2 in the physical plane,
which transforms to the right and left sides of the transformed plane, specifies a branch cut for the multiple-valued
function ~(x, y). Thus, the values of the physical coordinate functions x(~, r/) and y(~, 77) are the same on F3 as on
F4, and these functions and their derivatives are continuous from Fs to ['4. Therefore, boundary conditions are
neither required nor allowed on F3 and F4.
Since it is desired to perform all numerical computations in the uniform rectangular transformed plane, the
dependent and independent variables must be interchanged in eqs. (37) and (38), similar to eqs. (34) and (35):
eoc~ 2/~oc~ + Txnn = 0 , (41)
a y ~ - 2 3 y ~ +TYnn = 0 , (42)
where c~,/3, and 7 are defined as before, with the transformed boundary conditions

71= F!f2(~
,71)-,
(~,~,) [~, rhl e F ; , (43)

= , [~, rh ] E P~. (44)

The functions/l(~, rh), f2(~, rh), gl(~, 72), and g2(~, r/2) are specified by the known shape of the contours FI and
F2 and the specified distribution of ~ thereon. As noted, boundary conditions are neither required nor allowed
along the re-entrant boundaries F; and F~.
The system given by eqs. (41) and (42) is a quasi-linear elliptic system for the physical coordinate functions
x(~, r~) andy(~, r/) in the transformed plane. As pointed out before, although this system is considerably more
complex than that given by eqs. (37) and (38), the boundary conditions, eqs. (43) and (44), are specified on straight
boundaries, and the coordinate spacing in the transformed plane is uniform. The boundary fitted coordinate sys-
tem generated by the solution of eqs. (41)-(44) has a constant r/-line coincident with each boundary in the physi-
cal plane. The G-constant lines may be spaced as desired around the boundaries since the assignment of the ~-values
to the [x, y] boundary points via the functions fl, f2, gl, and g2 in eqs. (43) and (44) is arbitrary. Numerically, the
discrete boundary values [x, y] are transformed to equally spaced discrete ~k-points on both boundaries. As illus-
trated in fig. 7, the left and right boundaries of the transformed plane are re-entrant boundaries, which imply that
both solutions, x(~, r/) andy(~, r/), are required to be periodic in the region {[~, 77] I - oo< ~ < oo, r~l ~< 77~< rh }.
The analogous cylindrical coordinates used in the work [8] fall into the category of this section. In this case, the
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 19

fuel assembly center line plays the role of the curve ['1 in fig- 7. This curve is thus collapsed to a point. Extension
to a multiple-connected region is straightforward.

2.3.2.3. Preservation o f equation type. When coordinate transformations are utilized as an aid in solving a given
partial differential equation, it is imperative that the equation not change type under the transformation. Such an
invariance will now be demonstrated for the transformation of the system equations of interest. Consider the gen-
eral, second order, quasi-linear partial differential equation

A(x, y, f) fxx + B(x, y, f) fxy + C(x, y, JO fyy + E(x, y, JO fx + F(x, y, JO fy + G(x, y, f) = 0 , (45)
where F =f(x, y ) is a twice continuously differentiable scalar function, and A, B, C, E, F, and G are continuous
functions. Recall that the equation type is determined by the coefficient functions A, B, and C as follows:
Elliptic if B 2 - 4A C < 0 ,
Parabolic if B 2 - 4 A C = 0 ,
Hyperbolic if B 2 - 4.4 C > 0 .
It can readily be shown that eq. (45) transforms to

A * f ~ + B*f~ n + C*fnn + E*f~ + F*fn + G* = 0 , (46)

where
A* =-A ~2x + B~x~;y + C~2y,
B* - 2,4 ~xr~x + B(~xrly + ~yrlx ) + 2C~yrly ,
C* - Arl2x + B~x~y + Crl2y ,
E* = A~xx + B~xy + C~yy + E~x + e~y ,
F* =-Arlxx + Br~xy + CT~yy + E77x + F~ly ,
F* =- F .
Now, consider (B*) 2 - 4A'C*:
2 2
(B*) 2 - 4A'C* = [2A~xrl x + B(~xr~y + rlx~y) + 2C~yr/y] 2 - 4(A~2x + B~x~y + C~y)(Ar~x + Br~xrly + Crl2y)
= (B 2 + 4A C) (~xr~y - ~yrlx) 2
= (B 2 - 4A C)/J 2 .

Since j2 > O, B 2 - 4AC and (B*) 2 - 4A *C* are either both positive, both negative, or both zero. This implies that
eqs. (45) and (46) are of the same type.

2.3.3. Governing equations in non-dimensional form

ap + a(puj) = 0, (47)
at axj

a(pui) + a(ouiufl _ arii

at ax] ax/ P&' (48)

aE a(ujE3 a(uirij)_ aq/ +

-- + = D Q, (49)
at ax] ax] axj
The variables in the conservation equations are non-dimensionalized with respect to the reference velocity (Vo),
length (Lo), density (Po), enthalpy (ho), gravitational constant (go), viscosity (tto), thermal conductivity (~o), and
specific heat (Cpo), as follows:
20 W.T. Sha / Rod-bundle thermal-hydraulic analysis

(Ui)NI) = Ui/Vo, (t)ND = t/(Lo/Vo), (P)ND = P/Po ,

(Xi)ND = xi/Lo, (E)ND = E/(poho) , (g)ND = g/go ,
(C)N D = e/ha, (P)ND = P/(Poho), (T)ND = T/(ho/Cpo),,
(/2)ND = ]'//U0 , (K)ND = t~/t~0 , ('r/j)Nl) = "rij/(Poho),
(qi)ND = qi/(PohoVo), (Q)ND = Q/(hoPoVo/Lo),

where subscript ND denotes non-dimensional and is deleted in eqs. (47)-(49) for simplicity, and

E =p e + D 2 =p +D -p,

1 1 [au i + a u / l
"r/j = ~P~ij + ~ld~x] aXi ]'

1 aT
qi = - - ~ KaXi,

with the non-dimensional numbers

VoLopo ~oCpo vl
R _ - Reynolds n u m b e r , P = = Prandtl n u m b e r , D = -v.
~to ~o ho

2.3.4. Transformed governing equations

The boundary fitted curvilinear coordinates are taken such that ~ and rt lie in the plane normal to the fuel assem-
bly axis, with ~"parallel to this axis. The planes are assumed equally spaced along the axis so that, with Z along the
axis,

clZ
Zf = ~ = constant, (50)
us

The Cartesian coordinates x and y lie in the plane normal to the fuel assembly axis.
After transformation to the ~, 77, ~"system, the governing equations in non-dimensional form become:
ap + 1 a(pfi) l a(p~) + d~ a(pw)
(51)
at z a~ J an dZ a~
) 1 a(pu~) 1 a(pu~) d[ a(puw) 1 a _ +1 3_~
a(pu__+___ + -- +. . . . (Yn'rll Xn'r21) (X+'r21 -- Y+'r11)
at J a~ J an dZ Of Ja~ J an
d~" a('r31)
(52)
dZ af '

O(pv) 10(pvft) 10(pvO) d~ 8(pvw) 1 0 1a d~" O('rx2)

(52a)
at g a+ J arl dZ a+ J O+ aorl dZ a~

O(pw) 10~wa) 1 a(pw~) dfa(Oww) 1 a +la

- ~ ( v ~ r 1 3 - xn'r23) an(X~'r23 -y~r13)
at J ~ J ~n dZ a~ J J

d~ a(r33)
+ pg, (52b)
dZ a~
 W.T. Sha / Rod-bundle thermal-hydraulic analysis 21

and

-8E
- + - - 1 8(Eta)+ - -1 8(E~) d~" 8(Ew)
+ - - -
1 8(pa) 1 8(p~) d~"0(pw)
8t J 8~ J ~r/ dZ ~)~" J ~ J or/ dZ O~
D8
+ -J ~" [v,#~ ~u + rl:V + r~3w) - x , # : i u+ r~:v + r:~w)]

D0
+-)- ~ [xdr:lU + r : : v + r:3w) - y d ' c ~ u + r~2v + r~w)]

d~'~ 1
+ D ~ , ~-~t.'c3,u + r32v + r33w) + j ~ - ~ ( Y n ~ T - x r ~ r y )

+
1 8--(x~rr-y~rx)+
1 d~ ~(~Tz)+Q_. (53)
JRP 8~7 RP dZ
The stress terms in the momentum equations are:
1
y n r 1 1 - xnr21= - ~ C a p + Cgu~ - C 1 3 u n - C16v~ + C l a V n - - ~ C a - ~ w ~ ,

1
x~ r: , - y~K ,1 = - - ~ C~p + -~
/a(C 13u~ - 2 d~" )
G ,u,7 - G ~v~ + G ~v,7 - -~ c~ ~ , w r ,

r31 R IdZ Uz + 2 Cl(C8w ~ - CTw ~ ,

1C ~( 2 d~" )
Yr/rl2 - xrfl'22 = - ~ 6P +~- C16tt~ - C17ur/- CloV~ + ClaV~ - - ~ C6 ~ w~ ,

x~r22 --Y~T12 = -- Csp + ClaU~ - C15url - C14]2~ + CI2F n - 3 C 5 ~ w~- ,

/a ['d~"
r32 =~[-~v~ + 2G(Csw, - C6w~
)],
y,r~3 x,r23 ---~(Gw~ - C3w,

X/57"23 --Y~713 = u- ~ [ C 4 w n - C a w ~ +d_~(Csu, - CTv0] ,

r3a -
1
b p
+/aV4 d~"
RI_3 - ~ w~ -
2 2
-~( C1)(Cau~ - CTu. + Csv. - C6v~
)].
The stress terms in the energy equation are:

yn(r, ,u + r,2v + r 13w) - X.o(T2IL/ + T221y + T23W) = ~_(Cgu - C161,') u~ + (C 171~- C 13~/) L/.0

+ c~w ~ u~ + (C, ov - c, ou) v~ + (c,~u - c , . . ) v. - cow ~ ~ + C.ww~ - c ~ w w . - a ~ w~ ,

x~(r2 ~u + r22v + r2 aw) - y~(r~ lU + rl 2v + rl 3w) = ~ C~ 8v - C~ au) u~ + (C~ ~u - Ci sv) u n

22 W.T. Sha / Rod-bundle thermal-hydraulic analysis

ut w(C17u - C141,)v~ +(C12l,_ C15lJ)]y~+Csw_~P~ +C4ww ._C3ww~ 2 d~w ~ ,

3 dZ
and

u [4 4 d~" 4 4
T31U + T3212+ T33W =-R L--3ClC8wu~ +-3C1CTwLtn + u - ~ u f +-3 C1C6WF~ ~ClCswv~

+ v d ~ v ~ + 2 C ~ f t w ~ + 2 C ' ~ w ~ + 4 w d ~ w3~ l ' " dZ

The conduction terms are:

KO'~T x - x ~ T y ) = K(C2T~ - C3Tn) ,

K ( x ~ T y - y ~ T x ) = ~¢(C4T n - C3TO,

and

tc Tz = tc d~ T~ .

The coefficients in these relations are defined

1 a /3 7
C1 -- C2 - C3 - C4 - - Cs -~ x~ , C6 - x,~, C7 - y ~ ,
2J' J' J' J'

Cs=Yn, C 9 = ~ ( 3 a + Y Zn , Cao- (3a + x~), C,, =-3-~(37 + y~) ,

1 1
C,2 = ~ ( 3 7 + x~), Cx3 = ~ ( 3 3 + y ~ y n ) , C14 - 1(3/3 + x ~ x n ) ,

1 1
Cl 5 --
: x~y~ C16 = x n Y n c~=~(x~ + 2J), c,~ - ~ ( x ~ - 2J),
3J ' 3J ' 3d
with

J = x~y n - xnY ~ , Lt = ldy~ -- 1~Xll , = vx~ - uy~ ,

a = x~ + y ~ , /3 = x ~ x n + y ~ y n , 3' = x~ + yZ .

2.3.5. Initial a n d b o u n d a r y c o n d i t i o n s a n d n u m e r i c a l t e c h n i q u e
The initial and boundary conditions and numerical technique employed in the Benchmark rod-bundle thermal-
hydraulic analysis are similar to those used in the porous medium formulation with volume porosity, surface per-
meability, distributed resistance, and distributed heat source outlined in the previous section, except that the mod-
ified staggered mesh system is used in which pressure, density, temperature or enthalpy are specified at cell center,
and all velocities are specified at the intersections of grid lines. Some recent numerical results obtained from
BODYFIT-1 are presented in the paper by Chen, Vanka, and Sha in the current special issue.

3. C o n c l u s i o n

Three pertinent methods used in rod-bundle thermal-hydraulic analysis are presented. These methods are (1)
subchannel analysis, (2) porous medium formulation, and (3) Benchmark rod-bundle thermal-hydraulic analysis
 w.T. Sha / Rod-bundle thermal-hydraulic analysis 23

using a boundary fitted coordinate system. Basic limitations of methods (1) and (2) are clearly delineated. In sub-
channel analysis, the transverse momentum equation cannot be treated with the same rigor as the axial momentum
equations, due to subchannel arrangement. Furthermore, approximations must be employed for interfacing infor-
mation between two different control volumes, e.g., one for the transverse momentum equation and another for
the axial momentum equation. These basic limitations are inherent; therefore, validity of subchannel analysis to
situations with large flow disturbance poses serious concern, The porous medium formulation with volume poros-
ity, surface permeability, distributed resistance and distributed heat source is used in the COMMiX code, and it
eliminates some of the limitations employed in subchannel analysis. The concept of surface permeability is new in
porous medium formulation, and it greatly facilitates characterization of an anisotropic medium. Furthermore, the
porous medium formulation provides a greater range of applicability. As an example, the COMMIX code is capable
of treating both continuum (reactor plenum mixing, piping analysis, etc.) and quasi-continuum (rod bundle, fuel
assembly, etc.) problems. This capability represents a most cost-effective way of developing a computer code.
However, accurate determination of distributed resistance in a complex rod-bundle geometry requires fine struc-
ture of both temperature and velocity. The most rigorous rod-bundle thermal-hydraulic analysis, of course, uses a
boundary-fitted coordinate transformation. Each rod is explicitly represented as an internal boundary; thus, ap-
propriate boundary conditions can be specified. After the coordinate transformation, the complex rod-bundle
geometry reduces to a multiple connected continuum in a rectangular region with all boundaries coincident with
the grid lines. However, the system of equations to be solved in the transformed plane is more complicated than
the original one. Most important of all, the empirical mixing coefficients and cross flow resistances normally asso-
ciated with a rod-bundle thermal-hydraulic analysis are eliminated, thereby enhancing computational stability and
accuracy. However, one drawback of benchmark rod-bundle thermal-hydraulic analysis using a boundary fitted
coordinate system is that it requires large computer storage and long running time, and thus, is limited to relatively
small rod-bundle sizes or a local region of interest in a large rod bundle.
Recently, the effect of distorted geometry on thermal-hydraulic performance in a fuel assembly has received a
great deal of attention. Once the fuel elements are distorted (in an arbitrary way), the geometrical flow paths
become very complex, making their accurate representation in any calculation procedure very difficult. In sub-
channel analyses, it is difficult to represent distortions in the axial directions. The difficulties encountered are the
following:
(a) The subchannel representations with an arbitrarily distorted configuration in a rod bundle are very difficult.
(b) Because of the distortion, the mixing coefficients, hydraulic resistances, and conduction lengths used in the
flow and heat transfer calculations need to be recalibrated. This is a difficult task for various distorted configura-
tions. This difficulty is compounded by the inherent approximations used in the transverse momentum equations.
(c) The fuel rod boundaries are inclined with the normals to the subchannel, and the velocities are at an angle
with the fuel rods. Because the finite-difference grid does not align with the boundaries, representation of the
boundary conditions will be difficult and inaccurate.
Therefore, subchannel analysis for distorted rod bundle geometries can be subject to significant difficulties and
error. The porous medium approach has the advantage over subchannel analysis in that the distorted geometry can
be represented more conveniently and accurately by varying the volume porosity and surface permeability. It also
removes the inherent assumptions used in the transverse momentum equations associated with subchannel analysis.
The advantage of using correct transverse momentum equation is significant when the flow is not dominant in the
axial direction. It should be noted that empirical calibration of the distorted resistance with experimental data
corresponding to the distorted geometry is necessary. The distributed resistance of a given distorted configuration
may be estimated by using the benchmark rod-bundle thermal-hydraulic analysis with the three-dimensional bound-
ary fitted coordinate transformation.
24 W.T. Sha / Rod-bundle thermal.hydraulic analysis

Acknowledgement

The author is indebted to Professors B.T. Chao, J.C. Slattery and S.L. Soo for stimulating discussions on the
quasi-continuum formulation; to Professor J.F. Thompson for his contribution on benchmark rod-bundle thermal-
hydraulic analysis; to my co-workers Drs. H.M. Domanus, V.L. Shah, B.G-J. Chen, S.P. Vanka and R.C. Schmitt,
for their general support; to Mrs. S.A. Moll for her excellent typing of this paper, and most importantly, to Drs.
R.T. Curtis, C.N. Kelber, and Mr. P.M. Wood of the U.S. Nuclear Regulatory Commission for their support of
this work under contract No. A 2045 and to Dr. B.R. Sehgal o f the Electric Power Research Institute for his par-
tial support of the BODYFIT code development efforts under Contract No. 1383-1.

References

[1 ] W.T. Sha and R.C. Schmitt, ANL-77-15 (July 1977).

[2] D.S. Rowe, BNWL-1695 (March 1973).
[3] R. Potter et al., AEEW-R 1057 (July 1976).
[4] W.T. Sha and B.T. Chao, NUREG-CR-0945, ANL-CT-79-42 (July 1979).
[5] J.C. Slattery, Interfacial transport phenomena, Northwestern University, Evanston, IL (1979), to appear.
[6] W.J. Wnek et al., ANCR-1207 (Nov. 1975).
17] W.T. Sha, H.M. Domanus, R.C. Schmitt, J.J. Oras and E.I.H. Lin, NUREG-0415, ANL-77-96 (Jan. 1978).
[ 8] W.T. Sha and J.F. Thompson, NUREG-CR-0001, ANL-78-1 (Jan. 1978).
[91 W.T. Sha, B.C-J. Chen, Y.S. Cha, S.P. Vanka, R.C. Schmitt, J.F. Thompson and M.L. Doria, Benchmark rod-bundle thermal-
hydraulic analysis using boundary fitted coordinates, Trans. ANS (Nov. 1979).
[10] J.F. Thompson, F.C. Thames and C.W. Mastin, NASA-CR-2729 (1977).