Lecture 20:
The SIMPLE Algorithm (Cont’d)
Last Time …
z Looked into problem of introducing pressure into
continuity equation for incompressible flows
z Introduced SIMPLE algorithm
» Derived the pressure correction equation
This Time …
z Look at the SIMPLE algorithm in detail
z Examine auxilliary issues
» Under-relaxation and convergence
z Boundary conditions
» Nature of pressure in incompressible flows
SIMPLE Algorithm
z Semi-Implicit Method for Pressure-Linked Equations
z Proposed by Patankar and Spalding (1972)
z Idea is to start with discrete continuity equation
z Substitute into it the discrete u and v momentum
equations
» Discrete momentum equations contain pressure
differences
z Hence get an equation for the discrete pressures
» SIMPLE actually solves for a related quantity called
the pressure correction
SIMPLE Algorithm
z Solve momentum equations with guessed pressure
field p* -- resulting velocity fields are u* and v*
» Do not satisfy continuity because p* is wrong
z Propose corrections to velocities and pressure so that
corrected velocities satisfy discrete continuity
z Let the corrected values be:
Pressure
Velocity corrections correction
SIMPLE Algorithm (Cont’d)
z Also require that corrected velocities and pressures
satisfy momentum equations:
z Subtracting starred momentum equations from above:
Velocity Correction Equation
z Make an approximation:
Dropped
∑ nb anbunb
'
and ∑ nb anb vnb
'
Velocity Correction Equations
z Define:
so that
and
Pressure Correction Equation
z Starred velocities do not satisfy discrete continuity
equation:
z However, corrected velocities do:
Pressure Correction Equation (Cont’d)
z Substituting for flow rate corrections:
z Collect terms in pressure correction p’ to create
pressure correction equation
Discrete Pressure Correction Equation
aP p = ∑ anb p + b
'
P
'
nb
z Scarborough criterion
nb satisfied in the
equality
z b term is the amount
by which the starred
velocities do not
satisfy continuity
SIMPLE Solution Loop
1. Guess velocities and pressure p*
2. Discretize and solve u momentum equation to obtain
u* using p* for pressure term
3. Discretize and solve v momentum equation to obtain
v* using p* for pressure term
4. Formulate p’ equation coefficients. In particular, find b
term in p’ equation using u* and v*:
b= F −F +F −F
*
w e
*
s
*
n
*
SIMPLE Solution Loop
5. Solve p’ equation to obtain the pressure correction at
all main control volume cell centroids
6. Correct velocity and pressure:
7. At this point, velocities satisfy continuity but not
momentum
8. Solve for other φ’s
9. Check for convergence. If converged, exit. If not, go to
2.
Discussion
z Pressure correction equation nudges velocity and
pressure fields into satisfying both continuity and
momentum equations through a set of continuity-
satisfying fields
z At step 7, corrected velocities (u,v) satisfy discrete
continuity exactly every single iteration
» However they don’t satisfy momentum
z Note how continuity-satisfying velocity fields are used
to solve for φ’s in step 8
» If this was not done, we wouldn’t get bounded φ
during iteration even with UDS!
Effect of Approximation
z Dropping ∑ nb anbunb
'
, ∑ nb anb vnb
'
in deriving p’
equation does not change the final answer
z At convergence u’ and v’ are zero
z Similarly, p’ becomes a constant
» Can choose arbitrarily to be zero for all-velocity bc
z Thus, approximations to primed equations cannot
change the converged solution
z Approximations can change rate of convergence,
though
Under-Relaxation
z In reality, velocity correction consists of two parts:
Pressure part
Velocity part
z Dropping ∑ nb anbunb
'
, ∑ nb anb vnb
'
places the entire
burden of velocity correction on the pressure
correction
Under-Relaxation (Cont’d)
z Corrected velocities always satisfy continuity,
regardless of approximation
z However, large pressure correction yields poor
pressure iterates
z Under-relax pressure correction in correcting p*:
z Do not under-relax velocity correction or else
corrected velocities will not satisfy continuity !
Under-Relaxation (Cont’d)
z That is, do not use:
u = u* + α u '
z Because of non-linearity, it is necessary to under-relax
momentum equations:
ae (1 − α )
aeue*
α α
Nature of Pressure in Incompressible
Flows
z Consider domain with all-velocity bc:
•Pressure does not
appear except as
gradient
Wall
•Absolute value of
pressure does not
matter
Inflow Outflow •Only differences of
Wall pressure are
meaningful
Pressure in Incompressible Flows
(Cont’d)
z What about pressure boundary conditions?
•Say Pin =100, Pout
=50.
Wall •Compute velocity
Pout
field
Pin
•Would velocity field
change if Pin =200,
Wall Pout =150 ?
Pressure in Incompressible Flows
z What about mixed conditions?
• Say Vin =10, Pout =10
•Would computed
Wall velocity be different if
Vin Pout Vin =10, Pout =100 ?
Wall
Discussion
z Pressure does not change density
» Absolute level of pressure does not matter
z Only pressure differences matter in incompressible flows
z When all bc are velocity bc, pressure level is indeterminate
» p and p+c are solutions
z When at least one pressure bc is present
» Pressure level is fixed (not indeterminate)
» But only differences of pressure determine solution
» Changing pressure bc while keeping pressure
differences the same does not change velocity solution
Velocity Boundary Conditions
z Continuity equation for near-boundary cell P:
z Here,
z Incorporate boundary mass flow rate directly, without
corrections.
z Corresponding p’ equation would have aw=0
Boundary P’ Equation for Velocity BC
aP pP' = ∑ anb pnb
'
+b • Scarborough criterion satisfied
nb in the equality
a E = ρ e d e ∆y • p’ and p’+c are both solutions
to this equation, just as in the
aW = 0 interior
a N = ρ n d n ∆x
• Thus, at convergence,
a S = ρ s d s ∆x p’=constant is a solution
a P = a E + a N + aS • Pressure level is thus
b = Fb − Fe* + Fs* − Fn* determined only up to an
additive constant
Use boundary
value directly
Velocity BC (Cont’d)
z Consider domain with all-velocity bc
» BC must satisfy overall mass balance
z N main control volumes
» N discrete continuity equations
» Only N-1 are linearly independent since Nth control
volume balance can be deduced from overall mass
balance
» Thus only N-1 pressure differences can be found
» Pressure level thus arbitrary
Pressure Boundary Conditions
z Will not do in detail
z At given-pressure boundaries, p’ =0
z This provides a “Dirichlet” boundary condition for
pressure
Boundary P’ Equation for Pressure BC
aP pP' = ∑ anb pnb
'
+b • Scarborough criterion satisfied
nb in the inequality
aE = ρe d e ∆y • p’=0 at pressure boundary.
ab = ρb db ∆y Thus p’ and p’+c are NOT
both solutions to this equation.
aN = ρ n d n ∆x
aS = ρ s d s ∆x • At convergence, p’=0 is the
solution
aP = aE + ab + aN + aS
• Pressure level is thus NOT
b= F −F +F −F
*
b e
*
s
*
n
*
arbitrary
•However, only differences of
Derived from boundary pressure matter
momentum equation
Example Problem 1
Problem 6.4 from Patankar
A one-dimensional flow through a porous material is
governed by:
dp
c u u+ =0
dx
where c is constant. The continuity equation is
d
( uA) = 0
dx
where A is the effective area for the flow.
Problem 1 (Cont’d)
Use the SIMPLE procedure to compute p2 , uB, and uC
from the following data:
∆x = 2, c B = 0.25, cC = 0.2
AB = 5, AC = 4, p1 = 200, p3 = 38
As an initial guess, set
u B = uC = 15 and p2 = 120
1 uB 2 uC 3
Closure
In this lecture:
z We presented the SIMPLE solution loop
z Found that the SIMPLE algorithm takes the iteration
through a set of continuity-satisfying fields to
convergence
z Considered the nature of pressure in incompressible
flows
z Saw how to implement boundary conditions
