http://syeilendrapramuditya.wordpress.

com

A Numerical Method for Thermo-Fluid-Dynamics

Analyses of Fast Nuclear Reactors Fuel Assemblies

*** Two-Phase Flow Formulation ***

http://syeilendrapramuditya.wordpress.com

May 2010

1 of 15
 http://syeilendrapramuditya.wordpress.com

This document is freely downloadable in Microsoft Word format

The use of this document is ABSOLUTELY UNLIMITED
Some materials related to this document might be available from the following link:
http://wp.me/p61TQ-rR

2 of 15
 http://syeilendrapramuditya.wordpress.com

FOREWORD

As the title implies, this paper explains a general numerical technique for two phase fluid dynamics

problems, under the Subchannel Analysis approach. Comprehensive explanations can be found on

various text books and scientific papers, in which some of them are listed at the end of this paper.

Any comments and suggestions are always welcomed.

Syeilendra Pramuditya

Tokyo, May 2010

3 of 15
 http://syeilendrapramuditya.wordpress.com

CONTENT

FOREWORD........................................................................................................................................3

CONTENT...........................................................................................................................................4

1. TWO PHASE FLUID DYNAMICS............................................................................................5

1.1. The Six-Equation Model......................................................................................................5

1.2. Discretization in Triangular (Subchannel) Geometry..........................................................7

1.3. The Newton-Rhapson Method and Derivation of Poisson Equation.................................11

2. FURTHER READING...............................................................................................................15

4 of 15
 http://syeilendrapramuditya.wordpress.com

1. TWO PHASE FLUID DYNAMICS

1.1. The Six-Equation Model

We will use the six-equation model, three conservation equations for each phase.

Mass conservation:

  
Eq. 1  j  j    j  j u j    j
t

Energy conservation:

      
 j  j e j    j  j e j u j    p  j    j u j    QWj  QIj  QKj

Eq. 2
t  t 

Momentum conservation:

         
Eq. 3  j  j u j    j  j u j u j    j p  FWj  FIj  Fj   j  j g
t

Here the momentum equation is written in conservative form. For the calculation technique we will

use to solve the above equation system, it is more convenient for us to use the momentum equation

(Eq. 3) in its non-conservative form

Expanding the LHS terms in (Eq. 3):

          
u j   u j  j  j      j  j  u j u j   j  j   u j  u j  u j  j  j u j 
 
jj
Eq. 4 t t
    
  j p  FWj  FIj  Fj   j  j g

           
jj u j   u j  j  j   u j u j  j  j   j  j u j u j   j  j u j u j 
Eq. 5 t t
    
  j p  FWj  FIj  Fj   j  j g

Rearrange:

           
u j  j  j   u j u j  j  j   j  j u j u j   j  j u j    j  j u j u j 
Eq. 6 t t
    
  j p  FWj  FIj  Fj   j  j g

5 of 15
 http://syeilendrapramuditya.wordpress.com

          
u j   j  j    u j  j  j   j  j u j     j  j u j    j  j u j u j 
Eq. 7  t  t
    
  j p  FWj  FIj  Fj   j  j g

        
u j   j  j    j  j u j    j  j u j    j  j u j u j 
Eq. 8  t  t
    
  j p  FWj  FIj  Fj   j  j g

We can substitute (Eq. 1) to the leftmost term of (Eq. 8) as follow:

    
u j  j   j  j u j    j  j u j u j 
Eq. 9 t
    
  j p  FWj  FIj  Fj   j  j g

Re-ordering:

   
jj u j    j  j u j u j 
Eq. 10 t
     
  j p  FWj  FIj  Fj  u j  j   j  j g

Interface jump conditions:


Eq. 11 
j
j  0, Fj
Ij  0, Qj
Ij 0

Where:

QWj  energy exchange between wall and phase j

QIj  energy exchange between interface and phase j
QKj  energy transfer due to heat conduction in phase j

FWj  momentum exchange between wall and phase j

FIj  momentum exchange between interface and phase j

Fj  momentum exchange due to mass transfer

g  gravitational acceleration
j  g (vapor) or f (liquid)
 j  density of phase j
 j  volume fraction of phase j

u j  velocity of phase j
e j  internal energy of phase j
 j  interface mass transfer

6 of 15
 http://syeilendrapramuditya.wordpress.com

1.2. Discretization in Triangular (Subchannel) Geometry

In order to be able to numerically solve the above equation system, we need to discretize our

equations on appropriate geometry. For this purpose, we will use triangular geometry, which

represents typical subchannel geometry of Liquid Metal Fast Reactors (LMFR), as shown as follow:

Figure 1. Subchannel geometry

Where:

V  volume of the appropriate control volume

w  axial velocity
u  transverse velocity
i  subchannel number
l  axial plane number
k  gap number

7 of 15
 http://syeilendrapramuditya.wordpress.com

Figure 1 shows “staggered” control volume model, where mass and energy conservation equations

will be discretized by using Control Volume I (CV-I), and momentum equation will be discretized

by using the other two control volumes, which are staggered half mesh respect to CV-I.

The discrete form of (Eq. 1) is as follow:

n 1 n
 j  j    j  j 
i ,l i ,l

t
 n 1 n 1 n 1 
  A j  j  i ,l 1/ 2  w j  i ,l 1/ 2   A j  j  i ,l 1/ 2  w j  i ,l 1/ 2    k  A j  j  k .l u j  k ,l 
n n n

Eq. 12  k

 Vi ,l 
 
 
n 1
  j 
i ,l

With terms evaluated at time n  1 are:  j ,i ,l  j ,i ,l w j ,i ,l 1/ 2 w j ,i ,l 1/ 2 u j , k ,l  j ,i , l

The discrete form of (Eq. 2) is as follow:

Eq. 13
n 1 n
 j  j e j    j  j e j 
i ,l i ,l

t
 n 1 n 1 n 1 
  A j  j e j  i ,l 1/ 2  w j  i ,l 1/ 2   A j  j e j  i ,l 1/ 2  w j  i ,l 1/ 2    k  A j  j e j  k ,l u j  k ,l 
n n n

 k

 Vi ,l 
 
 
  n 1 n 1 n 1  

n n n
n 1
  j    j 
n
 A    w    A    w     A    u 
 k ,l   k ,l  
n 1 
  i ,l   i ,l   j  i ,l 1/ 2   i ,l 1/ 2 
j j  i ,l 1/ 2   i ,l 1/ 2
j k  j j
 pi ,l  k

 t  Vi ,l 
   
 
n 1 n 1 n
 QWj   QIj   QKj 
i ,l i ,l i ,l

With terms evaluated at time n  1 are:

 j ,i ,l  j ,i ,l e j ,i ,l w j ,i ,l 1/ 2 w j ,i ,l 1/ 2 u j , k ,l pi ,l QWj ,i ,l QIj ,i ,l

8 of 15
 http://syeilendrapramuditya.wordpress.com

The discrete form of (Eq. 10) for axial direction is as follow:

Eq. 14

n 1 n
n
 w j    w j 
i ,l 1/ 2 i ,l 1/ 2
 j  j  
i ,l 1/ 2 t

 j  j 
n



 w j 
n

i ,l 1/ 2 
 Aw j    Aw j     k u j 
n

i ,l 1
n

i ,l
k
 n

i ,l 1/ 2
 Au j 
n 
k ,l 1/ 2 

i ,l 1/ 2
 Vi ,l 1/ 2 
 
 
n  Ai ,l 1 pi ,l 1  Ai ,l pi ,l 
n 1 n 1
n 1/ 2 n 1/ 2 n 1/ 2
   j    FWj , z    FIj , z    Fj , z 
i ,l 1/ 2 V i ,l 1/ 2
i ,l 1/ 2 i ,l 1/ 2 i ,l 1/ 2

n n 1 n
  w j      j  j  g
i ,l 1/ 2  j  i ,l 1/ 2 i ,l 1/ 2 z

With terms evaluated at time n  1 are: w j ,i ,l 1/ 2 pi ,l 1 pi ,l FWj , z ,i ,l 1/ 2 FIj , z ,i ,l 1/ 2 Fj , z ,i ,l 1/ 2

The discrete form of (Eq. 10) for transverse direction is as follow:

Eq. 15

n 1 n
n
u j   u j 
 j  j  k ,l k ,l

k ,l t

n 
 j  j  
k   k ,l 
j j 
  u  n  Au  n   Au  n   w  n  Aw  n
 m ,l  j  i ,l   k ,l 
j j  
 k ,l 1/ 2  Aw j

n
 
 k ,l 1/ 2  
k ,l Vk ,l 
 
 
n  Am pm ,l  Ai pi ,l 
n 1 n 1

   j      F  n 1/ 2   F  n 1/ 2   F  n1/ 2  u  n   n1

 Wj , x  k ,l  Ij , x  k ,l  j , x  k ,l  j  k , l  j  k ,l
k ,l V k ,l

With terms evaluated at time n  1 : u j ,k ,l pm ,l pi ,l  j ,i ,l FWj , x ,k ,l FIj , x ,k ,l Fj , x ,k ,l

In total, variables which contain terms evaluated at time n  1 are:

 j ,i ,l  j , i ,l w j ,i ,l 1/ 2 w j ,i ,l 1/ 2 u j , k ,l e j ,i , l
pi ,l 1 pi ,l pm,l  j ,i ,l QWj ,i ,l QIj ,i ,l
FWj , z ,i ,l 1/ 2 FIj , z ,i ,l 1/ 2 Fj , z ,i ,l 1/ 2 FWj , x ,k ,l FIj , x,k ,l Fj , x , k ,l

9 of 15
 http://syeilendrapramuditya.wordpress.com

We choose four main variables, they are:

1. Pressure  p 
n 1

2. Void fraction   g 
n 1

3. Liquid temperature  Tl 
n 1

4. Vapor temperature  Tg 
n 1

All other variables evaluated at time n  1 are related to the main variables through various

constitutive equations. Therefore, now the two momentum equations, (Eq. 14) and (Eq. 15), only

contain two variables evaluated at time n  1 , they are velocities and pressures.

We can re-arrange (Eq. 14) into the following form:

n 1
Eq. 16  w j   C1  C2  pin,l11  pin,l1 
i ,l 1/ 2

Where C1 and C2 are the sums of terms evaluated at time n in (Eq. 14).

We can also obtain the following equation:

n 1
Eq. 17  w j   C3  C4  pin,l1  pin,l11 
i ,l 1/ 2

Where C3 and C4 are the sums of terms evaluated at time n in (Eq. 14).

And (Eq. 15) becomes:

n 1
Eq. 18 u j   C5  C6  pmn ,l1  pin,l1 
k ,l

Where C5 and C6 are the sums of terms evaluated at time n in (Eq. 15). We can then eliminate

velocity terms in (Eq. 12) and (Eq. 13) by substituting (Eq. 16), (Eq. 17), and (Eq. 18).

Therefore, we now have 4 unknowns  p,  g , Tl , Tg  and 4 equations to solve, they are mass and

energy equation for each phase. Note that by using this scheme, spatial coupling is accomplished

only through pressures, while the neighboring  g , Tl , Tg are not involved. In actual discretized

10 of 15
 http://syeilendrapramuditya.wordpress.com

equations above, the pressure terms that must be evaluated for each calculational cell are denoted as

pi ,l , pi ,l 1 , pi ,l 1 , pm1,l , pm 2,l , pm 3,l . This equation system is to be solved by using the Newton-Rhapson

method, which is explained as follow.

1.3. The Newton-Rhapson Method and Derivation of Poisson Equation

Consider a single variable non-linear equation:

Eq. 19 f ( x )  x 3  54

Determination of x is equivalent to find the roots of (Eq. 19):

Eq. 20 f ( x )  x3  54  0  x 3  54

The variable x can be determined by using iterative technique known as the Newton-Rhapson

method as follow:

1
r 1
d 
Eq. 21 x  x   f ( x)
r
  f (x )
r

 dx 
x  xr 

Or we can write:

d 
   x  x    f ( x )
r 1 r r
Eq. 22  f ( x)
 dx x  xr 


Eq. 23  D f ( x)     x   f ( xr )
 x 
x  xr 

Where r denotes Newton-Rhapson iteration step. The iteration is repeated until  x and f ( x )
r

6 9
become sufficiently small, ~ 10 for  x and ~ 10 for f ( xr ) .

The multi-dimensional version of the Newton-Rhapson method is as follow:

Eq. 24 J r   X r 1  X r    F r  J r   X  F r

Where the symbol J denotes a Jacobian matrix.

11 of 15
 http://syeilendrapramuditya.wordpress.com

Let Fm,l , Fm, g , Fe,l , Fe , g stand for mass equations for liquid and vapor, and energy equations for liquid

and vapor, respectively (e.g. Eq. 12 and Eq. 13), and they are non-linear functions of the main

variables  p,  g , Tl , Tg  . For the sake of clarity, we can write:

Fm,l  Fm,l  pin,l1 ,  gn,i1,l , Tl ,ni,l1 , Tgn,i,1l , pin,l11 , pin,l11 , pmn 1,1l , pmn 2,1l , pmn 3,1l   0
Fm, g  Fm, g  pin,l1 ,  gn,i1,l , Tl ,ni,l1 , Tgn,i,1l , pin,l11 , pin,l11 , pmn 1,1l , pmn 2,1l , pmn 3,1l   0
Eq. 25
Fe,l  Fe ,l  pin,l1 ,  gn,i1,l , Tl ,ni,l1 , Tgn,i,1l , pin,l11 , pin,l11 , pmn 1,1l , pmn 2,1l , pmn 3,1l   0
Fe, g  Fe , g  pin,l1 ,  gn,i1,l , Tl ,ni,l1 , Tgn,i,1l , pin,l11 , pin,l11 , pmn 1,1l , pmn 2,1l , pmn 3,1l   0

Our task now is to determine pi ,l ,  g ,i ,l , Tl ,i ,l , Tg ,i ,l , pi ,l 1 , pi ,l 1 , pm1,l , pm 2,l , pm3,l , which is equivalent to

n 1 n 1 n 1 n 1 n 1 n 1 n 1 n 1 n 1

find the roots of non-linear equation system shown in (Eq. 25), by using Newton-Rhapson method.

For simplicity of notation, we drop the " n  1" superscripts. For this equation system, the matrices

in (Eq. 24) are:

 Fm,l Fm,l Fm ,l Fm ,l Fm,l Fm,l Fm,l Fm,l Fm ,l 

 
 pi ,l  g ,i ,l Tl ,i ,l Tg ,i ,l pi ,l 1 pi ,l 1 pm1,l pm 2,l pm3,l 
 F Fm, g Fm, g Fm, g Fm, g Fm, g Fm, g Fm , g Fm , g 
 m, g
 pi ,l  g ,i ,l Tl ,i ,l Tg ,i ,l pi ,l 1 pi ,l 1 pm1,l pm 2,l pm 3,l 
Eq. 26 Jr   
 Fe,l Fe,l Fe,l Fe,l Fe,l Fe,l Fe,l Fe,l Fe,l 
 p  g ,i ,l Tl ,i ,l Tg ,i ,l pi ,l 1 pi ,l 1 pm1,l pm 2,l pm3,l 
 i ,l
 F Fe , g Fe, g Fe, g Fe, g Fe, g Fe, g Fe , g Fe, g 
 e, g 
 pi ,l  g ,i ,l Tl ,i ,l Tg ,i ,l pi ,l 1 pi ,l 1 pm1,l pm 2,l pm3,l 

Where all the derivatives are evaluated at Newton-Rhapson iteration step r .

Eq. 27
T
 X   pi ,l  g ,i ,l  lr,i ,l Tl ,i ,l  gr ,i ,l Tg ,i ,l  pi ,l 1  pi ,l 1  pm1,l  pm 2,l  pm 3,l 

Where the temperatures are multiplied by their volume fractions evaluated at iteration step r , in

order to avoid singularities of the Jacobian matrix.

12 of 15
 http://syeilendrapramuditya.wordpress.com
T
Eq. 28 F r   Fmr,l Fmr, g Fer,l Fer, g 

The matrix system, (Eq. 24), (Eq. 26), (Eq. 27), (Eq. 28), can be re-arrange as follow:

 pi ,l 1 
 a11 a12 a13 a14    pi ,l   a15 a16 a17 a18 a19     Fmr.l 
      p  r 
a29  
i ,l 1 
 a21 a22 a23 a24    g ,i ,l   a25 a26 a27 a28 F
Eq. 29   pm1,l    mr. g 

 a31 a32 a34    l ,i ,l Tl ,i ,l   a35 a39     Fe.l 
r
a33 a36 a37 a38
  r     pm 2,l   r 
 a41 a42 a43 a44   g ,i ,l  Tg ,i ,l   a45 a46 a47 a48 a49     Fe , g 
  p m 3, l 

Or in simple notation:

Eq. 30 A x  B y  z

Where aij are elements of the Jacobian matrix (Eq. 26). Now we multiply both sides of (Eq. 30)

with the inverse of matrix A as follow:

Eq. 31 A 1
A  x  A 1
B  y   A z  1

Eq. 32 xC y  g

Where C and g are new matrices as a result of the above matrix operation. Now we write (Eq. 32)

as follow:

 pi ,l 1 
  pi ,l   c11 c12 c13 c14 c15     g1 
     pi ,l 1   
Eq. 33   g ,i ,l   c21 c22 c23 c24 c25   g
 pm1,l    2 
  lr,i ,l Tl ,i ,l   c31 c32 c33 c34 c35     g3 
 r     pm 2,l   
 g ,i ,l  Tg ,i ,l  c41 c42 c43 c44 c45     g 4 
 p m 3, l 

Eq. 33 implies that spatial coupling is accomplished only through pressures, which means that if the

pressure correction terms   p 's are known, the remaining three main variables are readily

obtained. Therefore, to solve (Eq. 33), first we evaluate the pressure correction terms, so we need to

solve the following equation for each calculational cell:

13 of 15
 http://syeilendrapramuditya.wordpress.com

Eq. 34  pi ,l  c11 pi ,l 1  c12 pi ,l 1  c13 pm1,l  c14 pm 2,l  c15 pm3,l  g1

Note that the pressure correction formula (Eq. 34) is a discrete formulation of the Poisson equation

in terms of the pressure correction  p :

Eq. 35 2   p   

Also note that Poisson equation is of type elliptic Partial Differential Equations (PDEs), where

disturbance introduced at any interior point, influences all other point in the domain, which is

consistent with the incompressible model we use here. The ability to influence all other points from

interior point implies that boundary conditions are required on all boundaries.

Eq. 34 is the pressure correction formula for a single computational cell, and our task now is to

solve it for the entire computational domain. If we write (Eq. 34) in matrix format (for all cells), we

would have a banded matrix, representing a linear algebraic equation system, as shown as follow:

 A1,1 A1,2  P   B 
  1   1 
 A2,1 A2,2 A2,3   P2   B2 
    
 A3,2 A3,3 A3,4   P3   B3 
    
          
Eq. 36     
 Al ,l 1 Al ,l Al ,l 1   Pl   Bl 
         
    
 AL 1, L  2 AL 1, L 1 AL 1, L   PL 1   BL 1 
    

 AL , L 1 AL , L   PL   BL 

T
Pl   p1,l  p2,l   pi ,l   pI ,l 

T
Bl  b1,l b2,l  bi ,l  bI ,l 

Where I is total number of subchannels per axial plane, L is total number of axial planes, Al ,l is

an I  I sparse matrix, Pl is the solution vector, and Bl is collection of remaining terms that are

not associated with the pressure correction terms. The structure of this pressure correction matrix is

14 of 15
 http://syeilendrapramuditya.wordpress.com

called Block Tridiagonal, and many algorithms are available to solve it (e.g. Block LU

factorization, Thomas algorithm, Gaussian elimination, SOR, etc).

2. FURTHER READING

Some useful reading materials:

1. http://scannapieco.asu.edu/fluids.html

2. http://www.mpia-hd.mpg.de/~dullemon/lectures/fluiddynamics/index.html

3. http://www.amath.unc.edu/Faculty/minion/class/puckett/C_P_Notes.pdf

4. http://preview.tinyurl.com/2ef72hn

5. http://preview.tinyurl.com/2a85qn6

6. http://preview.tinyurl.com/2dk8afc

7. Advances in two-phase flow modeling for LMFBR applications. A.L. Schor, M.S. Kazimi

and N.E. Todreas. (http://tinyurl.com/2bm3v65)

8. A semi-implicit method for two-phase fluid dynamics. D.R Liles and Wm.H Reed.

(http://tinyurl.com/2ejkxeb)

9. SABENA: Subassembly boiling evolution numerical analysis. Hisashi Ninokata, Taka-aki

Okano. (http://tinyurl.com/2alr6om)

10. Analysis of low-heat-flux sodium boiling test in a 37-pin bundle by the two-fluid model

computer code SABENA. Hisashi Ninokata. (http://tinyurl.com/27qjttx)

15 of 15