SIMPLE Algorithm for Two-Dimensional

Channel Flow

— Fluid Flow and Heat Transfer

by Professor Jung-Yang San

Mechanical Engineering Department
National Chung Hsing University
 Two-dimensional, transient, incompressible
laminar flow neglecting body force:
∂u ∂v
(1) Continuity: + =0
∂x ∂y
∂(ρu) ∂ ( ρ u 2 ) ∂ ( ρ uv ) ∂Pe ∂ 2u ∂ 2u
(2) Momentum: x -direction: +[ + ]= − + µ( + )
∂t ∂x ∂y ∂x ∂x 2 ∂y 2
∂( ρ v) ∂ ( ρ uv ) ∂ ( ρ v 2 ) ∂Pe ∂ 2v ∂ 2v
y -direction: +[ + ]= − + µ( + )
∂t ∂x ∂y ∂y ∂x 2 ∂y 2

∂T ∂T ∂T 1 ∂ 2T ∂ 2T
(3) Energy : + [u +v ]= ( + )
∂t ∂x ∂y α ∂x 2 ∂y 2

For example: the geometry of the original problem (2-D channel) is

y v
u∞ Ly
u
x

Lx
• The SIMPLE (semi-implicit method for pressure-linked
equations) algorithm can only be used to solve the
velocity distribution of steady flow (穩態流).
• It can be arranged into a finite-difference scheme or a
finite-volume scheme.
• In this class, a finite-difference scheme is developed
for the SIMPLE algorithm.
• A large number of iterations involved in the calculation,
thus the method is quite time-consuming.
• In the method, the Navier-Stokes equations for unsteady
flow was used. Time progressing in calculation is
treated as iteration steps.
• As every grid point in the flow field is satisfied with
the continuity equation and the momentum equations,
the acquired velocity distribution is the true velocity
distribution.

• The intermittent result during the iterations does not

possess any physical meaning.

• The SIMPLE algorithm can also be used to solve the

velocity distribution of three-dimensional flow
whether the flow is laminar or turbulent.
Finite-Difference Approach (Staggered Grid - 交錯網格)
y
(i+1/2, j+1)

pt. a
(i, j+1/2) (i+1, j+1/2)

(i-1, j) (i-1/2, j) (i, j) (i+1/2, j) (i+1, j) (i+3/2, j)

dy
(i, j-1/2) (i+1, j-1/2)
pt. b

(i+1/2, j-1)
dx

x
Note: pressure grid ponts
x-dir. velocity (u) grid points
y-dir. velocity (v) grid points
y
(i-1/2, j+1) (i, j+1) (i+1/2, j+1)

pt. c pt. d
(i-1, j+1/2) (i+1, j+1/2)
(i, j+1/2)

(i-1/2, j) (i, j) (i+1/2, j)

dy
(i, j-1/2)
dx

x
1
at point "a": vi +1/ 2, j +1/ 2 ≡ (vi , j +1/ 2 + vi +1, j +1/ 2 )
2
1
at point "b": vi +1/ 2, j −1/ 2 ≡ (vi , j −1/ 2 + vi +1, j −1/ 2 )
2
1
at point "c": ui −1/ 2, j +1/ 2 ≡ (ui −1/ 2, j + ui −1/ 2, j +1 )
2
1
at point "d": ui +1/ 2, j +1/ 2 ≡ (ui +1/ 2, j + ui +1/ 2, j +1 )
2
centered around point (i + 1/ 2, j ), the x-direction momentum equation
is solved (J.D. Anderson, JR., Computational Fluid Dynamics,
McGraw Hill 1995).

( ρ u )in++1/1 2, j − ( ρ u )in+1/ 2, j
∆t
( ρ u 2 )in+3/ 2, j − ( ρ u 2 )in-1 / 2, j ( ρ uv )in+1/ 2, j +1 − ( ρ uv )in+1/ 2, j -1 pin+1, j − pin, j
= −[ + ]−
2∆x 2∆y ∆x
uin+3/ 2, j − 2uin+1/ 2, j + uin−1/ 2, j uin+1/ 2, j +1 − 2uin+1/ 2, j + uin+1/ 2, j −1
+ µ[ + ]
(∆x) 2
(∆y ) 2

∆t n
or ( ρu) n +1
i +1/ 2, j = ( ρu) n
i +1/ 2, j + A∆t − (pi +1, j − pin, j ) (1)
∆x

( ρ u 2 )in+3/ 2, j − ( ρ u 2 )in-1/ 2, j ( ρ uv )in+1/ 2, j +1 − ( ρ uv )in+1/ 2, j -1

where A ≡ −[ + ]
2∆x 2∆y
uin+3/ 2, j − 2uin+1/ 2, j + uin−1/ 2, j uin+1/ 2, j +1 − 2uin+1/ 2, j + uin+1/ 2, j −1
+ µ[ + ]
(∆x) 2
(∆y ) 2
similarly, centered around point (i , j + 1/ 2), the y-direction momentum
equation is solved.

( ρ v)in,+j1+1/ 2 − ( ρ v)in, j +1/ 2

∆t
(ρv ) 2 n
i , j + 3/ 2 − ( ρ v 2 )in, j -1/ 2 ( ρ vu )in+1, j +1/ 2 − ( ρ vu )in−1, j +1/ 2 pin, j +1 − pin, j
= −[ + ]−
2∆y 2 ∆x ∆y
vin, j +3 / 2 − 2vin, j +1/ 2 + vin, j −1/ 2 vin+1, j +1/ 2 − 2vin, j +1/ 2 + vin−1, j +1/ 2
+ µ[ + ]
(∆y ) 2
( ∆x ) 2

∆t n
or ( ρ v) n +1
i , j +1/ 2 = ( ρ v) n
i , j +1/ 2 + B∆t − (pi , j +1 − pin, j ) (2)
∆y

( ρ v 2 )in, j +3/ 2 − ( ρ v 2 )in, j -1/ 2 ( ρ vu )in+1, j +1/ 2 − ( ρ vu )in−1, j +1/ 2

where B ≡ −[ + ]
2∆y 2 ∆x
vin, j +3 / 2 − 2vin, j +1/ 2 + vin, j −1/ 2 vin+1, j +1/ 2 − 2vin, j +1/ 2 + vin−1, j +1/ 2
+ µ[ + ]
(∆y ) 2
( ∆x ) 2
Assume a set of initial quessed (p* ) n , ( ρ u * ) n and ( ρ v* ) n values, equations (1)
and (2) can be calculated to yield a new set of ( ρ u * ) n +1 and ( ρ v* ) n +1 values,

∆t * n
( ρ u * )in++11/ 2, j = ( ρ u * )in+1/ 2, j + A*∆t − [(p )i +1, j − ( p* )in, j ] (3)
∆x
∆t * n
( ρ v* )in,+j1+1/ 2 = ( ρ v* )in, j +1/ 2 + B*∆t − [(p )i , j +1 − ( p* )in, j ] (4)
∆y

let u ' ≡ u − u* ; v ' ≡ v − v* ; p ' ≡ p − p*

A' ≡ A − A* ; B ' ≡ B − B*

Subtracting eq.(3) from eq.(1) and eq.(4) from eq.(2), it yields

∆t ' n
( ρu )' n +1
i +1/ 2, j = ( ρu )
' n
i +1/ 2, j + A ∆t −
'
[(p )i +1, j − ( p ' )in, j ] (5)
∆x
∆t ' n
( ρ v )i , j +1/ 2 = ( ρ v )i , j +1/ 2 + B ∆t −
' n +1 ' n '
[(p )i , j +1 − ( p ' )in, j ] (6)
∆y

where p ' , u ' , v ' are corrections (修正) of pressure and velocities respectively.
Patankar 建議先將 A' , B ' , ( pu ' ) n 與 ( pv ' ) n 設為 0, eq. (5) 與 eq. (6)
則會分別變為:

∆t ' n
( ρ u ' )in++1/1 2, j = − [(p )i +1, j − ( p ' )in, j ] (7)
∆x
∆t ' n
( ρ v ' )in,+j1+1/ 2 = − [(p )i , j +1 − ( p ' )in, j ] (8)
∆y

∆t ' n
亦即 ( ρ u )in++1/1 2, j = ( ρ u * )in++1/1 2, j − [(p )i +1, j − ( p ' )in, j ] (9)
∆x
∆t ' n
( ρ v)in,+j1+1/ 2 = ( ρ v* )in,+j1+1 / 2 − [(p )i , j +1 − ( p ' )in, j ] (10)
∆y

同理對不同之點上列之關係亦存在,
∆t ' n
( ρ u )i −1/ 2, j = ( ρ u )i −1/ 2, j −
n +1 * n +1
[(p )i , j − ( p ' )in−1, j ] (11)
∆x
∆t ' n
( ρ v)i , j −1/ 2 = ( ρ v )i , j −1/ 2 −
n +1 * n +1
[(p )i , j − ( p ' )in, j −1 ] (12)
∆y
 點 (i, j) 之質量守恆
y
(i, j+1)

(i, j+1/2)

(i-1, j) (i-1/2, j) (i+1, j)

(i, j) (i+1/2, j)
dy
(i, j-1/2)
dx
(i, j-1)
x
∂ ( ρ u ) ∂ ( ρ v)
Continuity equation: + =0
∂x ∂y

( ρ u )i +1/ 2, j − ( ρ u )i −1/ 2, j ( ρ v)i , j +1/ 2 − ( ρ v)i , j −1/ 2

差分, + =0 (13)
∆x ∆y
將 eqs. (9)-(12) 代入 eq. (13),可得

∆t ' n ∆t ' n
{( ρ u * )in++1/1 2, j − [(p )i +1, j − ( p ' )in, j ]} − {( ρ u * )in−+1/1 2, j − [(p )i , j − ( p ' )in−1, j ]}
∆x ∆x +
∆x
∆t ' n ∆t ' n
{( ρ v* )in,+j1+1/ 2 − [(p )i , j +1 − ( p ' )in, j ]} − {( ρ v* )in,+j1−1/ 2 − [(p )i , j − ( p ' )in, j −1 ]}
∆y ∆y
=0
∆y

整理後,亦即,

a ( pi' , j ) + b( pi' +1, j ) + b( pi' −1, j ) + c( pi' , j +1 ) + c( pi' , j −1 ) + d = 0 (14)

∆t ∆t ∆t ∆t
其中, a = 2[ + ] ; b = − ; c = −
(∆x) 2 (∆y ) 2 (∆x) 2 ( ∆y ) 2
1 1
d = [( ρ u * )in++1/1 2, j − ( ρ u * )in−+1/1 2, j ] + [( ρ v* )in,+j1+1/ 2 − ( ρ v* )in,+j1−1/ 2 ]
∆x ∆y
= mass source term
 Step by Step Procedure for the SIMPLE Algorithm

1. Guessed values of ( p* )n at all the "pressure" grid points. Also,

arbitrarily set values of ( ρ u * ) n and ( ρ v* ) n at the "velocity"
grid points.

2. Using equations (3) and (4), solve for ( ρ u * ) n +1 and ( ρ v* ) n +1

at all corresponding internal grid points.

3. Substitute these values of ( ρ u * ) n +1 and ( ρ v* ) n +1 into equation (14),

and solve for p ' values at all appropriate internal grid points.

4. Calculate new " p n +1 " values at all internal grid points.

[p n +1 = ( p* ) n + p ']
5. Designate the p n +1 , ( ρ u * ) n +1 and ( ρ v* ) n +1 values obtained above
as the new values of ( p* ) n , ( ρ u * ) n and ( ρ v* ) n and solve
the momentum equations [eqs. (3) and (4)] again.

6. Repeat steps 2-5 until convergence ('d' values at all grid points
approach zero) is achieved.

7. When all the 'normalized' mass source terms (normalized 'd'

values at all grid points) less than 10-5 , the solution is considered
to be convergent.

8. The larger the selected ∆t value, the faster the convergence of the
solution. But as the selected ∆t value exceeds a certain level,
the solution of the iteration might be divergent.
 Dimensionless Governing Equations
Dimensionless groups (無因次化參數):
u* ≡ u/u∞ ; v* ≡ v/u∞ ; x* ≡ x/Ly ;
y* ≡ y/Ly ; t* ≡ t/(Ly / u∞ ) ; Pe* ≡ Pe / ( ρu∞2 ) ;
T − Tw
Re ≡ ρu∞ ( Dh ) /μ = ρu∞ (2Ly ) /μ ; T* ≡
T∞ − Tw

Substituting back into the original equations, it yields

∂u* ∂v*
(1) Continuity: + =0
∂x * ∂y *

(2) Momentum:
∂u* ∂ u * ∂ u * ∂ P * 2 ∂ 2u * ∂ 2u *
x-direction: + u* + v* =− e
+ ( + )
∂t * ∂x * ∂y * ∂x * Re ∂x *2 ∂y *2

∂v* ∂v * ∂v * ∂P * 2 ∂ 2 v* ∂ 2 v*
y -direction: + u* + v* =− e
+ ( + )
∂t * ∂x * ∂y * ∂y * Re ∂x *2 ∂y *2
the geometry of the scaled problem (2-D channel) is

y*
1
x*

Lx/ Ly
Energy equation (Neglecting axial conduction effect):

∂T ∂T ∂ 2T k
(u +v ) = α where α ≡
∂x ∂y ∂y 2 ρcp

T∞

y
Tw Ly
x

Lx
 T − Tw
Define: T* ≡
T∞ − Tw

∂T * ∂T * 2 ∂ 2T *
(u* + v* )= ( )
∂x* ∂y* Re Pr ∂y*2
ν µ/ρ
where Pr ≡ = = heat transfer Prandtl number
α α

y*
T* = 0 1
x*

Lx/ Ly