05.07.2014 Views

SIMPLE Method on Non-staggered Grids - Department of ...

SIMPLE Method on Non-staggered Grids - Department of ...

SIMPLE Method on Non-staggered Grids - Department of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<str<strong>on</strong>g>SIMPLE</str<strong>on</strong>g> METHOD FOR THE SOLUTION OF INCOMPRESSIBLE<br />

FLOWS ON NON-STAGGERED GRIDS<br />

I. Sezai - Eastern Mediterranean University, Mechanical Engineering <strong>Department</strong>, Mersin 10 -Turkey<br />

Revised in January, 2011<br />

1. Introducti<strong>on</strong><br />

There are usually two kinds <strong>of</strong> grid arrangements used to solve the fluid flow problems:<br />

<strong>staggered</strong> grids and n<strong>on</strong>-<strong>staggered</strong> grids. For the n<strong>on</strong><strong>staggered</strong> grids, vector variables and<br />

scalar variables are stored at the same locati<strong>on</strong>s, while for the <strong>staggered</strong> grids, vector<br />

comp<strong>on</strong>ents and scalar variables are stored at different locati<strong>on</strong>s, shifted half c<strong>on</strong>trol<br />

volume in each coordinate directi<strong>on</strong>. Staggered grids are popular because <strong>of</strong> their ability<br />

to prevent checkerboard pressure in the flow soluti<strong>on</strong> as discussed in Chapter 6.<br />

However, programming <strong>of</strong> the <strong>staggered</strong> grid method is tedious since u and v-momentum<br />

equati<strong>on</strong>s are discretized at different c<strong>on</strong>trol volumes shifted in different directi<strong>on</strong>s from<br />

the main c<strong>on</strong>trol volumes. The programming difficulties increase when <strong>on</strong>e deals with<br />

curvilinear or unstructured grids. As a result, nearly all codes written <strong>on</strong> curvilinear or<br />

unstructured grids use n<strong>on</strong>-<strong>staggered</strong> grid arrangement for the soluti<strong>on</strong> <strong>of</strong> fluid flow<br />

problems.<br />

On the other hand n<strong>on</strong>-<strong>staggered</strong> grids are pr<strong>on</strong>e to produce a false pressure field -<br />

checkerboard pressure if special precauti<strong>on</strong>s is not taken. For this reas<strong>on</strong>, in the early<br />

1980s and before, n<strong>on</strong>-<strong>staggered</strong> were rarely used for incompressible flow. However,<br />

since 1983 the n<strong>on</strong><strong>staggered</strong> grid (or collocated grid) has been used more widely, after<br />

Rhie and Chow (1983) proposed a momentum interpolati<strong>on</strong> method to eliminate the<br />

checkerboard pressure problem.<br />

1. Mathematical Formulati<strong>on</strong><br />

The governing equati<strong>on</strong>s for two dimensi<strong>on</strong>al incompressible laminar heat and fluid flow<br />

in Cartesian coordinates with c<strong>on</strong>stant properties are:<br />

c<strong>on</strong>tinuity equati<strong>on</strong><br />

1


∂ u<br />

∂ x<br />

+<br />

∂ v<br />

∂ y<br />

= 0<br />

(1)<br />

u-momentum equati<strong>on</strong><br />

( ρ ) ( ρ ) ( ρ )<br />

∂ u ∂ uu ∂ vu ∂p ∂ ⎛ ∂u⎞<br />

∂ u<br />

μ<br />

⎛ ∂<br />

+ + =− + ⎜ ⎟+ ⎜μ<br />

⎞<br />

⎟+<br />

s<br />

∂t ∂x ∂y ∂x ∂x⎝ ∂x⎠ ∂y⎝ ∂y<br />

⎠<br />

v-momentum equati<strong>on</strong><br />

( ρ ) ( ρ ) ( ρ )<br />

∂ v ∂ uv ∂ vv ∂p ∂ ⎛ ∂v⎞<br />

∂ v<br />

μ<br />

⎛ ∂<br />

+ + =− + ⎜ ⎟+ ⎜μ<br />

⎞<br />

⎟+<br />

s<br />

∂t ∂x ∂y ∂y ∂x⎝ ∂x⎠ ∂y⎝ ∂y⎠<br />

Energy equati<strong>on</strong><br />

( ) ( ) ( )<br />

∂ ρT ∂ ρuT ∂ ρvT ∂ ⎛ k ∂T ⎞ ∂ ⎛ k ∂T<br />

⎞<br />

+ + = + + s<br />

∂t ∂x ∂y ∂x⎜c ∂x ⎟ ∂y⎜c ∂y<br />

⎟<br />

⎝ p ⎠ ⎝ p ⎠<br />

Equati<strong>on</strong>s (2) – (4) can be expressed in a general form as:<br />

( ) ( u ) ( v )<br />

∂ ρφ ∂ ρ φ ∂ ρ φ ∂ ⎛ ∂φ ⎞ ∂ ⎛ ∂φ<br />

⎞<br />

+ + = ⎜Γ ⎟+ ⎜Γ ⎟+<br />

sφ<br />

∂t ∂x ∂y ∂x⎝ ∂x ⎠ ∂y⎝ ∂y<br />

(5)<br />

⎠<br />

where u and v are the velocity comp<strong>on</strong>ents, T is temperature and φ is any dependent<br />

variable (i.e. u, v and T), and t, ρ , Γ , and s φ<br />

are time, density, diffusi<strong>on</strong> coefficient, and<br />

source term per unit volume, respectively. Also,s u and s v terms represents source terms<br />

per unit mass, ie. drag forces, buyancy forces etc. Note that for the c<strong>on</strong>tinuity equati<strong>on</strong>,<br />

φ = 1, Γ= 0 , and s φ<br />

= 0 .<br />

In numerical calculati<strong>on</strong>s, the source terms in the above equati<strong>on</strong>s are written in<br />

linearized form as<br />

s = sc + spφ<br />

p<br />

For example, for natural c<strong>on</strong>vecti<strong>on</strong> problems, the right hand side <strong>of</strong> the v-momentum<br />

equati<strong>on</strong> c<strong>on</strong>tain an additi<strong>on</strong>al term ρ g <br />

− β ref<br />

( T − T ∞)<br />

, for which the source terms are<br />

<br />

sc<br />

=−ρrefgβ ( T −T∞<br />

) and s p = 0. Similarly, for problems involving heat generati<strong>on</strong> the<br />

right hand side <strong>of</strong> the energy equati<strong>on</strong> have an additi<strong>on</strong>al term q which is the energy<br />

generati<strong>on</strong> per unit volume. In this case, s c = q , s p = 0.<br />

T<br />

v<br />

u<br />

(2)<br />

(3)<br />

(4)<br />

2


The governing equati<strong>on</strong>s are discretized by using the finite-volume method <strong>on</strong> a<br />

n<strong>on</strong><strong>staggered</strong> grid system in which all variables are stored at the center <strong>of</strong> the c<strong>on</strong>trol<br />

volume (Figure 1).<br />

Integrating Eq. (5) over the c<strong>on</strong>trol volume with bounded cell faces e, w, n, and s<br />

surrounding center P, we have<br />

ρΔxΔy<br />

o<br />

( φP − φP) + ⎡( ρuφ) −( ρuφ) ⎤Δ y+ ⎡( ρvφ) −( ρvφ)<br />

⎤Δ x=<br />

e w n s<br />

Δt<br />

⎣ ⎦ ⎣ ⎦<br />

⎡ Γe Γ ⎤ ⎡<br />

w<br />

Γn Γ ⎤<br />

s<br />

( φE −φP) − ( φP −φW ) Δ y+ ( φN −φP) − ( φP −φS)<br />

Δx<br />

(6)<br />

⎢ ⎥ ⎢ ⎥<br />

⎣δ xe δ xw ⎦ ⎣δ yn δ ys<br />

⎦<br />

+ s ΔxΔ y+ s φ ΔxΔy<br />

c P P<br />

where the superscript o refers to the previous time level and Δ x , Δ y , δ x , δx , δy<br />

and<br />

δ y s<br />

are geometric lengths as shown in Figure 1.<br />

e w n<br />

Figure 1. N<strong>on</strong><strong>staggered</strong> Grid Arrangement<br />

The discretized c<strong>on</strong>tinuity equati<strong>on</strong> for incompressible flow is<br />

( ρue) Δy−( ρuw) Δ y+ ( ρvn) Δx−( ρvs) Δ x= 0<br />

(7)<br />

The cell face values φ e , φ w etc are calculated using QUICK or any other high order<br />

scheme in a deferred correcti<strong>on</strong> manner. Substituting these values into Eq. (6) and<br />

3


earranging, we obtain the final discretized equati<strong>on</strong> for any general variable φ as<br />

follows:<br />

o o<br />

aPφP = aEφE + aWφW + aNφN + aSφS + aPφP<br />

+ b+ pterm<br />

(8)<br />

where<br />

aE<br />

=<br />

ΓΔ<br />

e<br />

y<br />

+ max ⎡−( ρu ) Δy, 0⎤<br />

e<br />

Δx<br />

⎣<br />

⎦<br />

e<br />

aW<br />

=<br />

ΓΔ<br />

w<br />

y<br />

+ max ⎡( ρu)<br />

Δy, 0⎤<br />

w<br />

Δx<br />

⎣ ⎦<br />

w<br />

aN<br />

=<br />

ΓΔ<br />

n<br />

x<br />

+ max ⎡−( ρv)<br />

Δx<br />

, 0⎤<br />

n<br />

Δy<br />

⎣<br />

⎦<br />

n<br />

aS<br />

=<br />

ΓΔ<br />

s<br />

x<br />

+ max ⎡( ρv)<br />

Δx<br />

, 0⎤<br />

s<br />

Δy<br />

⎣<br />

⎦<br />

s<br />

o ρΔΔ<br />

x y<br />

aP<br />

=<br />

Δt<br />

o<br />

aP = aE + aW + aN + aS + aP − Sp<br />

+ΔF<br />

Δ F = ( ρu) Δy−( ρu) Δ y+ ( ρv) Δx−( ρv)<br />

Δx<br />

e w n s<br />

b= ( sc )<br />

eqn<br />

ΔxΔ y+ b1 b1<br />

= Sdc + ( Sc)<br />

bc<br />

Sp = ( sp) eqnΔxΔ y+<br />

( Sp)<br />

bc<br />

(9)<br />

Sdc = −max ⎡<br />

⎣( ρu) Δy, 0⎤( ) max ( ) , 0 ( )<br />

e ⎦ φe − φP + ⎡<br />

⎣<br />

− ρu Δy<br />

⎤<br />

e ⎦ φe −φE<br />

−max ⎡<br />

⎣<br />

−( ρu) Δy, 0⎤( w P) max ( ) , 0 ( w W)<br />

w ⎦ φ − φ + ⎡<br />

⎣ ρu Δy<br />

⎤<br />

w ⎦ φ −φ<br />

−max ⎡<br />

⎣( ρv) Δx , 0⎤( φn φP) max ( ρv) x, 0 ( φn φN)<br />

n ⎦<br />

− + ⎡<br />

⎣<br />

− Δ ⎤<br />

n ⎦<br />

−<br />

−max ⎡<br />

⎣<br />

−( ρv) Δx, 0⎤( φs − φP) + max ⎡( ρv) Δx, 0⎤( φs −φS)<br />

s ⎦ ⎣ s ⎦<br />

⎧=−( pe<br />

−pw) Δy<br />

for x-momentum equati<strong>on</strong><br />

⎪<br />

pterm= ⎨=−( pn<br />

− ps) Δx<br />

for y-momentum equati<strong>on</strong><br />

⎪<br />

⎩ = 0 for all other equati<strong>on</strong>s<br />

( s ) , ( s ) = source terms per unit volume in the differential equati<strong>on</strong> (buoyancy, drag for<br />

p eqn c eqn<br />

( Sp) bc, ( S<br />

c) bc=<br />

source c<strong>on</strong>tributi<strong>on</strong>s <strong>of</strong> near-boundary points<br />

where b represents all discretized source terms, excluding the pressure term. The term S dc<br />

is the source c<strong>on</strong>tributi<strong>on</strong>, which results from the adopti<strong>on</strong> <strong>of</strong> the deferred-correcti<strong>on</strong><br />

o<br />

procedure in which the face values <strong>of</strong> the independent variable, φ<br />

P<br />

is the value <strong>of</strong> φ<br />

P<br />

from the previous time step. φ e , φ w , φ n , φ s are calculated from a suitable high order<br />

4


scheme such as QUICK. The term S bc is the source c<strong>on</strong>tributi<strong>on</strong> <strong>of</strong> the near boundary<br />

points. For example, for a west boundary, if a Dirihlet boundary c<strong>on</strong>diti<strong>on</strong> φ = φ w exists,<br />

then, for a point near a west boundary, the source c<strong>on</strong>tributi<strong>on</strong> is S bc = a W φ w , where a W<br />

would be the coefficient if that point were an interior point. Similarly, (S p ) bc is the source<br />

c<strong>on</strong>tributi<strong>on</strong> <strong>of</strong> the near boundary points.<br />

Also, note that for source terms small letters corresp<strong>on</strong>d to per unit mass <strong>of</strong> the source.<br />

That is, S = sΔV, S c = s c ΔV, and S p = s p ΔV, where ΔV is the volume <strong>of</strong> the c<strong>on</strong>trol<br />

volume.<br />

It can be observed that the coefficients <strong>of</strong> the discretized x- and y-momentum equati<strong>on</strong>s<br />

are the same in collocated grid system (excluding the near boundary points), provided<br />

that the diffusi<strong>on</strong> coefficients Γ are the same in x and y-directi<strong>on</strong>s.<br />

In order to slow down the changes <strong>of</strong> dependent variables in c<strong>on</strong>secutive<br />

soluti<strong>on</strong>s, an underrelaxati<strong>on</strong> factor is introduced into the discretized equati<strong>on</strong> as follows:<br />

( 1−αφ<br />

)<br />

−1<br />

aP<br />

φ = a φ + a φ + a φ + a φ + a φ + b+ pterm+ a φ<br />

α<br />

(10)<br />

φ<br />

o o n<br />

P E E W W N N S S P P P P<br />

αφ<br />

where the superscript (n-1) refers to the previous iterati<strong>on</strong>. In that case, the terms in Eq.<br />

(9)are redefined as<br />

⎧ aP<br />

n 1<br />

b ← b + (1 −<br />

−<br />

α)<br />

φP<br />

⎪<br />

α<br />

⎨<br />

⎪ aP<br />

aP<br />

←<br />

⎪⎩ α<br />

Equati<strong>on</strong> (10) can be written as<br />

α<br />

α pterm<br />

1<br />

( a a a a b a<br />

o o ) ( 1 ) n φ<br />

φ = φ + φ + φ + φ + + φ + − αφ<br />

φ − + (11)<br />

a<br />

or<br />

φ<br />

P E E W W N N S S P P P P<br />

aP<br />

α<br />

αφpterm<br />

φ = ( a φ + a φ + a φ + a φ + B ) + (12)<br />

a<br />

φ<br />

P E E W W N N S S P<br />

aP<br />

where<br />

( 1−αφ<br />

)<br />

−1<br />

B = b + a φ + a φ<br />

(13)<br />

o o n<br />

P P P P P P<br />

αφ<br />

P<br />

P<br />

5


4. Momentum Interpolati<strong>on</strong><br />

4.1. Momentum interpolati<strong>on</strong> for steady-state problem<br />

BP = bP + ⎡⎣ 1−αu αu⎤⎦ aPuP<br />

[Eq.(13)]. For the<br />

n−1<br />

Note that for the steady problems ( )<br />

velocity comp<strong>on</strong>ent u at nodes P and E, Eq. (12) can be written as<br />

u<br />

u<br />

P<br />

E<br />

( )<br />

( a )<br />

P<br />

P<br />

( )<br />

( a )<br />

αu ∑<br />

i<br />

au<br />

i i<br />

+ Bp α<br />

P uΔy pe − pw<br />

P<br />

= − (14)<br />

( )<br />

( a )<br />

P<br />

E<br />

P<br />

P<br />

( )<br />

( a )<br />

αu ∑<br />

i<br />

au<br />

i i<br />

+ Bp α<br />

E uΔy pe − pw<br />

E<br />

= − (15)<br />

Mimicking the formulati<strong>on</strong> <strong>of</strong> u<br />

E<br />

interface velocity at the cell face e.<br />

u<br />

e<br />

( )<br />

( a )<br />

P<br />

e<br />

P<br />

E<br />

and u<br />

P<br />

, we can obtain the following expressi<strong>on</strong> for the<br />

( )<br />

( a )<br />

αu ∑<br />

i<br />

au<br />

i i<br />

+ Bp α<br />

e uΔy pE − pP<br />

= − (16)<br />

P<br />

e<br />

where the terms <strong>on</strong> the right-hand side with subscript e should be interpolated in an<br />

appropriate manner. The interface velocity at cell faces w, n, and s can be obtained<br />

similarly.<br />

In Rhie and Chow’s momentum interpolati<strong>on</strong>, the first term and 1 ( aP ) in sec<strong>on</strong>d<br />

e<br />

term <strong>of</strong> the Eq.(16) are linearly interpolated from their counterparts in Eqs.(14) and (15):<br />

⎛∑ au + B ⎞ ⎛∑ au + B ⎞ ⎛∑ au + B ⎞<br />

⎜ ⎟ ⎜ ⎟ ( 1 fe<br />

) ⎜ ⎟<br />

⎝ ⎠ ⎝ ⎠ ⎝ ⎠<br />

i i i p + i i i p + i i i p<br />

= fe<br />

+ −<br />

aP a<br />

e P<br />

a<br />

E P P<br />

(17)<br />

1 + 1 + 1<br />

= f + − (18)<br />

a a a<br />

e<br />

( ) ( )<br />

( 1 fe<br />

) ( )<br />

P e P E P P<br />

where f + e<br />

is a linear interpolati<strong>on</strong> factor defines as<br />

+ Δx<br />

P<br />

f<br />

e<br />

= (19)<br />

2δ x<br />

e<br />

In order to have a better understanding <strong>of</strong> Eq. (16), substituting ( au B a )<br />

Eq. (17) and ( ∑ au + B a ) , ( au B a )<br />

i i i p P P<br />

o o<br />

(16) and omitting the term au , we obtain<br />

P<br />

P<br />

i i i p P E<br />

∑ + from<br />

i i i p P e<br />

∑ + from Eqs. (14) and (15) into Eq.<br />

6


( − p ) +<br />

fe<br />

( aP)<br />

e<br />

α y( p p<br />

+<br />

)<br />

e<br />

( a )<br />

( − p )<br />

( a )<br />

⎧ αuΔy p<br />

αuΔy p<br />

E P e w<br />

⎫<br />

E<br />

⎪− +<br />

⎪<br />

+ + ⎪<br />

P E ⎪<br />

ue = ⎡fe uE ( 1 fe ) u ⎤<br />

⎣<br />

+ −<br />

P⎦ +⎨ ⎬<br />

⎪<br />

uΔ e<br />

−<br />

w P<br />

⎪<br />

⎪<br />

+ ( 1−<br />

f )<br />

⎪<br />

⎩<br />

P P<br />

⎭<br />

Equati<strong>on</strong>s (16) and (20) are essentially equivalent. However, Eq. (20) separates<br />

the interfacial velocity into two parts: a linear interpolati<strong>on</strong> part and the additi<strong>on</strong>al <strong>on</strong>e.<br />

The term in first set <strong>of</strong> brackets <strong>of</strong> Eq. (20) is the arithmetic-averaged values (linear<br />

interpolati<strong>on</strong> method) <strong>of</strong> two neighbor nodal velocities. The term in braces can be<br />

regarded as a correcti<strong>on</strong> term, which has the functi<strong>on</strong> <strong>of</strong> smoothing the pressure field, and<br />

it is this term that may remove the unrealistic pressure field.<br />

Equati<strong>on</strong>s (16) with (17) and (18) or equati<strong>on</strong> (20) c<strong>on</strong>stitute the Rhie and Chow’s<br />

Original Momentum Interpolati<strong>on</strong> <str<strong>on</strong>g>Method</str<strong>on</strong>g> (OMIM). Majumdar (1988) reported that<br />

soluti<strong>on</strong>s <strong>of</strong> steady-state problems from Rhie and Chow OMIM are dependent <strong>on</strong> the<br />

underrealxati<strong>on</strong> factor. To eliminate this underrelaxati<strong>on</strong> factor dependency, an iterati<strong>on</strong><br />

algorithm was proposed by him to calculate the cell-face velocity for steady-state<br />

problem as follows:<br />

( )<br />

( )<br />

( a )<br />

α ∑ au + b α Δy p − p<br />

u u f u f u<br />

( 1 α ) ⎡<br />

( 1 )<br />

u i i i p<br />

e u E P n− 1 + n− 1 + n−1<br />

e<br />

= − + −<br />

u e e E e P<br />

( aP)<br />

⎣<br />

+ − −<br />

e<br />

P e<br />

where superscript n-1 refer to previous iterati<strong>on</strong> values. This iterative implementati<strong>on</strong><br />

algorithm can achieve a unique soluti<strong>on</strong> that is independent <strong>of</strong> underrelaxati<strong>on</strong> factors.<br />

⎤<br />

⎦<br />

(20)<br />

(21)<br />

4.2. Momentum interpolati<strong>on</strong> for unsteady problems<br />

Choi (1999) reported that, the soluti<strong>on</strong> using the original Rhie and Chow scheme<br />

is time step size dependent. He proposed a modified Rhie and Chow scheme for an<br />

unsteady problem as follows, which is quite similar to Majumdar’s scheme for a steady<br />

problem:<br />

( )<br />

( )<br />

⎛∑ au + b ⎞ α Δy p − p α a<br />

u u u<br />

o<br />

i i i p u E P n−1<br />

u e o<br />

e<br />

= αu⎜<br />

⎟ − + ( 1− αu)<br />

e<br />

+<br />

e<br />

⎝ aP ⎠ aP ( aP)<br />

e e e<br />

with<br />

a<br />

o<br />

e<br />

ρδ xeΔy<br />

=<br />

Δt<br />

(22)<br />

(23)<br />

7


It is to be noted that body-force term is neglected for simplicity <strong>of</strong> presentati<strong>on</strong>. By a<br />

similar substituti<strong>on</strong> process as the <strong>on</strong>e for the Majumdar’s interpolati<strong>on</strong> in Eq. (21), the<br />

Eq. (22) can be written equivalently as<br />

( 1 )<br />

u ⎡f u f u ⎤<br />

⎣<br />

⎦<br />

( − p ) +<br />

fe<br />

( aP)<br />

e<br />

α Δy( p − p<br />

+<br />

)<br />

e<br />

( a )<br />

⎧ α Δy p<br />

⎪− +<br />

⎪<br />

⎪<br />

⎪+ ( 1−<br />

f )<br />

⎪<br />

( − p )<br />

( a )<br />

u E P u e w E<br />

u e w P<br />

α Δy p<br />

+ +<br />

P P<br />

e<br />

=<br />

e E<br />

+ −<br />

e P<br />

+⎨ ⎬<br />

n− 1 + n− 1 + n−1<br />

( 1 α ) ⎡<br />

u<br />

ue fe uE ( 1 fe ) u ⎤<br />

⎣<br />

P<br />

⎦<br />

o<br />

o<br />

⎡α<br />

α ( a )<br />

+ +<br />

⎢<br />

( 1 )<br />

⎪+ − + − −<br />

⎪<br />

⎪<br />

o<br />

a α ( a<br />

⎪ u f u f<br />

)<br />

+ − − −<br />

u<br />

⎪ ⎢( a ) a a<br />

⎩ ⎣<br />

u e o u P<br />

E o u P<br />

P o<br />

e e E e P<br />

P ( P)<br />

( P)<br />

e E P<br />

P<br />

E<br />

⎫<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎤⎪<br />

⎥⎪<br />

⎥⎪<br />

⎦⎭<br />

(24)<br />

According to Yu et al. (2002), soluti<strong>on</strong>s by using this scheme are still time step<br />

size dependent, though the dependence is quite small. They proposed a different<br />

interpolati<strong>on</strong> technique for the terms appearing in equati<strong>on</strong> (22) which appears to be both<br />

under relaxati<strong>on</strong> factor and time step size independent. In this method the first term <strong>on</strong><br />

the right-hand side <strong>of</strong> Eq. (22) is interpolated as follows:<br />

∑ au + b<br />

( ∑ +<br />

1) + ( 1− )( ∑ + )<br />

E<br />

1<br />

+ +<br />

fe ( sc) ( 1 fe )( sc)<br />

⎤δ<br />

x<br />

E<br />

P e<br />

y<br />

( ∑ ) + ( 1− )( ∑ )<br />

f au b f au b<br />

+ +<br />

e i i i e i i i<br />

+ ⎡ + − Δ<br />

⎛ i i i p ⎞ ⎣ ⎦<br />

⎜ ⎟ =<br />

+ +<br />

⎝ aP ⎠ f<br />

e e i<br />

ai f<br />

E e i<br />

ai<br />

P<br />

where b<br />

1<br />

is defined in Eq. (9).<br />

( ) ( 1 )( )<br />

− ⎡f s + − f s ⎤δ<br />

x Δ y+<br />

a<br />

⎣<br />

⎦<br />

+ +<br />

o<br />

e p<br />

E<br />

e p<br />

P<br />

e e<br />

Also, the denominator <strong>of</strong> the sec<strong>on</strong>d and third terms in Eq. (22) is interpolated as follows:<br />

+ +<br />

( a ) = f ( ∑ a ) + ( 1− f )( ∑ a )<br />

P e e i i E e i i P<br />

( ) ( 1 )( )<br />

− ⎡f s + − f s ⎤δ<br />

x Δ y+<br />

a<br />

⎣<br />

⎦<br />

+ +<br />

o<br />

e p<br />

E<br />

e p<br />

P<br />

e e<br />

Equati<strong>on</strong> (22) combined with Eq. (25) and Eq. (26) is Yu et al.’s (2002) new<br />

scheme. Substituting Eq. (25) into Eq. (22) the following equati<strong>on</strong> is obtained:<br />

P<br />

(25)<br />

(26)<br />

8


⎧ + +<br />

⎡fe ( aP) uE+ ( 1−<br />

fe )( aP)<br />

u ⎤<br />

P<br />

⎪⎣<br />

E<br />

P ⎦<br />

⎪ ⎡ + + +<br />

( fe ( sc) ( 1 fe )( sc)<br />

) δ xe y fe ( sc)<br />

x<br />

E P E E<br />

y⎤<br />

⎪<br />

+ − Δ − Δ Δ<br />

α ⎢<br />

⎥<br />

⎪+ u ⎢ +<br />

( 1 fe )( sc)<br />

x<br />

P P<br />

y<br />

⎥<br />

⎪ ⎣<br />

− − Δ Δ<br />

⎦<br />

1<br />

⎪<br />

u = ⎨+ α ⎡ y p p f y p p f y p p<br />

E<br />

( aP<br />

) ⎣<br />

−Δ − + Δ − + − Δ −<br />

e ⎪<br />

n− 1 + n− 1 + n−1<br />

⎪+ ( 1−αu) ⎡( aP) ue − fe ( aP) uE −( 1−<br />

fe )( aP)<br />

u ⎤<br />

e E P<br />

P<br />

⎪ ⎣<br />

⎦<br />

⎪ o o + o o + o<br />

+ α ⎡<br />

o<br />

u<br />

au<br />

e e<br />

− fe ( aP) uE −( 1−<br />

fe )( aP)<br />

u ⎤<br />

⎪ ⎣<br />

E<br />

P<br />

P⎦<br />

⎪<br />

⎩ ⎪<br />

+ +<br />

( ) ( ) ( 1 ) ( )<br />

e u E P e e w e e w<br />

where ( a )<br />

P<br />

e<br />

is found from equati<strong>on</strong> (26). It should be noted that the term in the<br />

parenthesis, which is multiplied by (1–α u ) is incorrect in the paper <strong>of</strong> Yu et al.’s (2002).<br />

It should also be noted that the cell face velocities found from the momentum<br />

interpolati<strong>on</strong> method are used to determine the mass fluxes across the cell faces.<br />

They should not be used for the cell face value <strong>of</strong> the independent variable φ in the<br />

deferred correcti<strong>on</strong> term b dc in eq. (9) in the case φ stands for u or v in the x- or y-<br />

momentum equati<strong>on</strong>s. The face values <strong>of</strong> the independent variable φ are calculated<br />

using a suitable c<strong>on</strong>vecti<strong>on</strong> scheme, such as UPWIND or QUICK.<br />

P<br />

⎫<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎤<br />

⎪<br />

⎦⎬<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪<br />

⎪ ⎭<br />

(27)<br />

The <str<strong>on</strong>g>SIMPLE</str<strong>on</strong>g> Algorithm<br />

For a guessed pressure field p * the corresp<strong>on</strong>ding face velocity<br />

Eq. (22) as<br />

( )<br />

( )<br />

* * *<br />

o<br />

* u i i i p<br />

e u E P n−1<br />

u e o<br />

e ( 1 αu)<br />

e e<br />

( aP)<br />

( aP)<br />

( aP)<br />

e e e<br />

*<br />

u<br />

e<br />

can be written using<br />

au b y p p a<br />

u α ∑ + α Δ −<br />

= − + − u + α u<br />

(28)<br />

*<br />

A similar equati<strong>on</strong> can be written for the face velocity v<br />

n<br />

. Now we define the correcti<strong>on</strong><br />

p' as the difference between the correct pressure field p and the guessed pressure field p *<br />

so that<br />

*<br />

p = p + p′<br />

(29)<br />

Similarly we define velocity correcti<strong>on</strong>s u' and v' as<br />

u = u + u′<br />

(30)<br />

*<br />

e e e<br />

9


v = v + v′<br />

(31)<br />

*<br />

n n n<br />

Subtracting Eq. (28) from Eq. (22) gives<br />

e<br />

( ′ )<br />

( a )<br />

P<br />

e<br />

( − )<br />

( a )<br />

αu ∑<br />

i<br />

au<br />

i i<br />

+ bp α<br />

e uΔy p′ E<br />

p′<br />

P<br />

u′ = − (32)<br />

P<br />

e<br />

As an approximati<strong>on</strong>, in <str<strong>on</strong>g>SIMPLE</str<strong>on</strong>g> method the first term in the above equati<strong>on</strong> is neglected<br />

giving<br />

( )<br />

u′ = d p′ − p′<br />

(33)<br />

u<br />

e e P E<br />

where<br />

d<br />

α A<br />

= (34)<br />

u u e<br />

e<br />

( aP<br />

)<br />

e<br />

where A e = Δy is the area <strong>of</strong> the c<strong>on</strong>trol volume at the east face. Similarly,<br />

( )<br />

v′ = d p′ − p′<br />

(35)<br />

v<br />

n n P N<br />

where<br />

d<br />

α A<br />

= (36)<br />

v v n<br />

n<br />

( aP<br />

)<br />

n<br />

Then the corrected velocities become<br />

( ′ ′ )<br />

u = u + d p − p<br />

(37)<br />

* u<br />

e e e P E<br />

( ′ ′ )<br />

v = v + d p − p<br />

(38)<br />

* v<br />

n n n P N<br />

Substituting the corrected face velocities such as that given by Eq.s (37)and (38) into the<br />

discretized c<strong>on</strong>tinuity equati<strong>on</strong> (7) gives<br />

a p′ = a p′ + a p′ + a p′ + a p′<br />

+ b<br />

(39)<br />

P P W W E E S S N N<br />

where<br />

a = ( ρ Ad) a = ( ρAd) a = ( ρAd) a = ( ρAd)<br />

E e W w N n S s<br />

* * * *<br />

( ρ ) ( ρ ) ( ρ ) ( ρ )<br />

b= u A − u A + v A − u A<br />

w e s n<br />

After solving the p' field from Eq. (39) the face velocities are corrected using Eq.'s (37)<br />

and (38) and the pressure field is corrected by using<br />

= + α ′<br />

(41)<br />

*<br />

p p<br />

p<br />

p<br />

where α p is the pressure under-relaxati<strong>on</strong> factor which is chosen to be between 0 and 1.<br />

(40)<br />

10


Similarly the nodal velocities are corrected using<br />

( ′ ′)<br />

u = u + d p − p<br />

(42)<br />

* u<br />

P P P w e<br />

( ′ ′)<br />

v = v + d p − p<br />

(43)<br />

* v<br />

P P P s n<br />

where<br />

d<br />

u<br />

P<br />

αuAe v αvAn<br />

= and d = (44)<br />

P<br />

( a ) ( a )<br />

P<br />

P<br />

P<br />

P<br />

The pressure correcti<strong>on</strong>s at the cell faces in Eqs. (42) and (43) are calculated by linear<br />

interpolati<strong>on</strong> from the nodal values as<br />

+ +<br />

p′ = f p′ + (1 − f ) p′<br />

(45)<br />

w w P w W<br />

+ +<br />

p′ = f p′ + (1 − f ) p′<br />

(46)<br />

e e E e P<br />

+ +<br />

p′ = f p′ + (1 − f ) p′<br />

(47)<br />

s s P s S<br />

+ +<br />

p′ = f p′ + (1 − f ) p′<br />

(48)<br />

n n N n P<br />

Boundary C<strong>on</strong>diti<strong>on</strong>s for Pressure<br />

Since there is no equati<strong>on</strong> for the pressure, no boundary c<strong>on</strong>diti<strong>on</strong>s are needed for the<br />

pressure at the boundary points. The pressure values at the boundary points can be<br />

calculated by linear extrapolati<strong>on</strong> using the two near-boundary node pressures.<br />

Boundary C<strong>on</strong>diti<strong>on</strong>s for Pressure Correcti<strong>on</strong> Equati<strong>on</strong><br />

When the velocities at the boundaries are known, there is no need to correct the velocities<br />

at the boundaries in the derivati<strong>on</strong> <strong>of</strong> the pressure correcti<strong>on</strong> equati<strong>on</strong>. For example if the<br />

velocity at the west boundary is known then for a c<strong>on</strong>trol volume near the west boundary:<br />

( ′ ′ )<br />

u = u + d p − p<br />

(49)<br />

u<br />

* u<br />

e e e P E<br />

w<br />

= u<br />

(50)<br />

wall<br />

( ′ ′ )<br />

v = v + d p − p<br />

(51)<br />

* v<br />

n n n P N<br />

( ′ ′ )<br />

v = v + d p − p<br />

(52)<br />

* v<br />

s s s S P<br />

Substituting equati<strong>on</strong>s (49)-(52) into the discretized c<strong>on</strong>tinuity equati<strong>on</strong> (7) we obtain the<br />

following pressure correcti<strong>on</strong> equati<strong>on</strong> for a c<strong>on</strong>trol volume near the west boundary<br />

a p′ = a p′ + a p′ + a p′ + a p′<br />

+ b<br />

(53)<br />

P P W W E E S S N N<br />

11


where<br />

a = ( ρ Ad) a = 0 a = ( ρAd) a = ( ρAd)<br />

a = a + a + a + a<br />

E e W N n S s<br />

P W E S N<br />

* * *<br />

( ρ ) ( ρ ) ( ρ ) ( ρ )<br />

b= uA − u A + v A − u A<br />

wall e s n<br />

Comparing Eq.'s (53)-(54) with (39)-(40) for a near boundary c<strong>on</strong>trol volume the same<br />

definiti<strong>on</strong> <strong>of</strong> the coefficients as used for the interior points can be used for a near<br />

boundary c<strong>on</strong>trol volume by setting the corresp<strong>on</strong>ding coefficient (a w in this case) to zero<br />

and using u wall in the b term.<br />

The above formulati<strong>on</strong> corresp<strong>on</strong>ds to Neumann boundary c<strong>on</strong>diti<strong>on</strong> ∂p´/∂n = 0, where n<br />

is normal to the boundary. As a result no value <strong>of</strong> pressure correcti<strong>on</strong> at the boundary<br />

( p′<br />

w<br />

) is involved in this formulati<strong>on</strong>. However, the value <strong>of</strong> the pressure correcti<strong>on</strong> is<br />

needed for correcting the nodal velocities near boundaries. For example, for correcting<br />

the u-velocity at a nodal point P near a west boundary, p′<br />

w<br />

at the west boundary is needed<br />

in accordance with equati<strong>on</strong> (42). This value can be obtained by using ∂p´/∂n = 0 at the<br />

boundary, that is using p′ (1, j) = p′<br />

(2, j)<br />

.<br />

References<br />

Choi S. K., "Note <strong>on</strong> the Use <strong>of</strong> Momentum Interpolati<strong>on</strong> <str<strong>on</strong>g>Method</str<strong>on</strong>g> for Unsteady Flows",<br />

Numerical. Heat Transfer Part, A, vol. 36, pp. 545-550, 1999.<br />

Majumdar S., "Role <strong>of</strong> Underrelaxati<strong>on</strong> in Momentum Interpolati<strong>on</strong> for Calculati<strong>on</strong> <strong>of</strong><br />

Flow with N<strong>on</strong><strong>staggered</strong> <strong>Grids</strong>", Numerical Heat Transfer, Part B, vol. 13, pp. 125-132,<br />

1988.<br />

Rhie C. M. and Chow W. L., "Numerical Study <strong>of</strong> the Turbulent Flow Past an Airfoil<br />

with Trailing Edge Separati<strong>on</strong>", AIAA Journal, vol. 21, no 11, pp. 1525-1535, 1983.<br />

Yu B., Tao W., and Wei J., "Discussi<strong>on</strong> <strong>on</strong> Momentum Interpolati<strong>on</strong> <str<strong>on</strong>g>Method</str<strong>on</strong>g> for<br />

Collocated <strong>Grids</strong> <strong>of</strong> Incompressible Flow", Numerical. Heat Transfer Part B, vol. 42, pp.<br />

141-166, 2002.<br />

(54)<br />

12

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!