Vous êtes sur la page 1sur 4

School of Mechanical Aerospace and Civil Engineering Overview

Advanced Modelling & Simulation: CFD


◮ Up to now, we have considered discretization schemes, and how they are
used for a single transport equation.
◮ In practical flow solvers, we have a set of coupled pde’s. In 3-D this
involves determining U, V , W and P, as well as any additional quantities
(temperature, enthalpy, turbulence energy, etc) that may be required.
Pressure-Velocity Coupling
◮ We have already examined how to discretize the momentum equations to
T. J. Craft obtain U, V and W . Pressure gradients appear in these, so we need a
George Begg Building, C41 method to determine the pressure distribution.
◮ However, the 4th equation we have is the continuity equation – which
does not actually contain the pressure explicitly.
Reading:
Contents:
◮ Review of basic FV method
J. Ferziger, M. Peric, Computational Methods for Fluid ◮ In this lecture we thus consider a strategy for handling the coupling
Dynamics between the velocity and pressure fields in a flow computation.
◮ Pressure-velocity coupling H.K. Versteeg, W. Malalasekara, An Introduction to
◮ Body-fitted coordinate systems Computational Fluid Dynamics: The Finite Volume ◮ We restrict attention to 2-D steady-state flow, and work within a finite
◮ Unsteady problems Method
S.V. Patankar, Numerical Heat Transfer and Fluid Flow volume framework on a simple rectangular mesh. The same methods are
◮ Turbulence and other physical
Notes: http://cfd.mace.manchester.ac.uk/tmcfd readily extended to 3-D and more complex grids.
modelling - People - T. Craft - Online Teaching Material
Pressure-Velocity Coupling 2008/9 2 / 16

◮ The equations we need to solve, are then Storage Arrangements


∂ (ρ UU) ∂ (ρ UV ) ∂P ∂ ∂U ∂ ∂U
   
+ =− + µ + µ + Su (1a) The U momentum equation can be integrated over the control volume
∂x ∂y ∂x ∂x ∂x ∂y ∂y ◮
shown below, resulting in a discretized equation of the form
∂ (ρ UV ) ∂ (ρ VV ) ∂P ∂ ∂V ∂ ∂V
   
+ =− + µ + µ + Sv (1b) aup Up = ∑ aui Ui + Sup + Su (2)
∂x ∂y ∂y ∂x ∂x ∂y ∂y
where Sup is source terms arising from integrating the pressure gradient
∂ (ρ U) ∂ (ρ V )
+ =0 (1c) over the cell, and Su represents any other source terms.
∂x ∂y

where Su and Sv represent any other source terms that may be present
◮ On the grid shown, Sup would be given by N

(from buoyancy, rotation, turbulent stresses, etc).


(Pw − Pe )
Sup = (∆x ∆y ) = (Pw − Pe )∆y (3)
∆x W w P e E
◮ Before considering how to obtain the pressure field, we first examine
alternative storage arrangements that can be used for the discretized flow ◮ A convenient storage arrangement is to use a
variables, since these have a bearing on how source terms are single set of control volumes, with all variables
represented, and thus on how the schemes to be described may be stored at the same locations (collocated). S

implemented.
◮ An advantage of this is that all geometrical data is only stored once.
◮ However, this arrangement does also have some disadvantages.
Pressure-Velocity Coupling 2008/9 3 / 16 Pressure-Velocity Coupling 2008/9 4 / 16
◮ With collocated storage we must interpolate between PP and PE to get ◮ A much stronger coupling between the velocity and pressure field nodal
Pe , the pressure at the east face of the cell (similarly for the west face). values is obtained by using a staggered grid.
PN
◮ On a uniform grid, this interpolation gives ◮ In this arrangement the velocity components
are stored at the centres of the faces of the Vn

Sup = (Pw − Pe )∆y = 0.5(PW − PE )∆y (4) pressure control volume.


P Uw PP Ue PE
W

so there is a relatively weak linkage between the velocity and local ◮ A disadvantage of this is there are separate Vs
pressure field (as PP is not used). control volumes for U, V and P, so more
geometrical information has to be stored. PS

P
◮ This can lead to chequerboarding: ◮ Such overheads become particularly cumbersome in non-orthogonal and
an unphysical pressure field that 3-D grid arrangements.
oscillates from node to node can
◮ The pressure gradient term in the equation for Ue is now simply
appear as uniform to the discretized
momentum equation. Sup = (PP − PE )∆y (5)
i-2 i-1 i i+1 i+2
giving a direct coupling between adjacent velocity and pressure values.
◮ Methods are available to overcome this problem. However, for now we ◮ For now, we work within a staggered grid arrangement. In the Advanced
consider an alternative arrangement that results in a stronger linkage CFD course we consider how schemes can be adapted to handle the
between adjacent pressure and velocity values. case when all quantities are stored at the same locations.
Pressure-Velocity Coupling 2008/9 5 / 16 Pressure-Velocity Coupling 2008/9 6 / 16

Pressure Correction Schemes The SIMPLE Scheme

◮ The problem in obtaining the pressure field arises because, although we ◮ A widely-used pressure-velocity coupling scheme is the SIMPLE (Semi
have 3 equations for U, V and P, the continuity equation does not Implicit Method for Pressure Linked Equations) scheme (Patankar, 1980),
explicitly contain P. which is outlined in the following.

◮ The momentum equations provide us with a set of discretized equations


◮ The discretized momentum equations for Ue and Vn are written as:
which can be solved for U and V , if we know the pressure field. However, aue Ue = ∑ aui Ui + Sup + Su avn Vn = ∑ avi Vi + Svp + Sv (6)
we cannot use the continuity equation directly to obtain P.
◮ Because of the staggered grid, Ue is a nodal value of U, and Vn a nodal
◮ Instead, we consider how an iterative procedure can be used to adjust value of V . The summations on the right hand sides are contributions
the pressure field in order to ensure that the resulting velocity field does from surrounding nodal values; Sup and Svp are pressure-related source
satisfy continuity. terms, and Su and Sv any other source terms.
◮ Such schemes are generally referred to as Pressure Correction schemes. ◮ After dividing through by the diagonal coefficient, the discretized
momentum equations can be written in the form
◮ At each iteration a ‘correction’ to the pressure distribution is calculated,
aui Ui avi Vi
designed to drive the local velocity field towards one that satisfies both Ue = ∑ +Du (PP −PE )+su Vn = ∑ +Dv (PP −PN )+sv (7)
momentum and continuity equations. aue avn
where Du = ∆y /aue , Dv = ∆x /avn , su = Su /aue , and sv = Sv /avn .
Pressure-Velocity Coupling 2008/9 7 / 16 Pressure-Velocity Coupling 2008/9 8 / 16
◮ Starting off with some initial values for the pressure field, the above ◮ Subtracting equations (7) from these, gives
equations can be solved to obtain U and V . aui Ui′ avi Vi′
Ue′ = ∑ + Du (PP′ − PE′ ) Vn′ = ∑ + Dv (PP′ − PN′ ) (10)
◮ However, these will not, in general, satisfy the continuity equation. aue avn
which provide relations between the pressure and velocity corrections
◮ Now suppose we add corrections U ′, V ′, P′ to the velocities and pressure ensuring the momentum equations are still satisfied.
so that the corrected variables
◮ To simplify the analysis, we now split the velocity and pressure
U∗ = U + U′ V∗ = V +V′ P∗ = P + P′ (8) corrections into two parts: U ′ = U1′ + U2′ etc. Then we can write
satisfy both the momentum and continuity equations. aui Ui′

Ue1 ′
+ Ue2 =∑ ′
+ Du (PP1 ′
− PE1 ′
) + Du (PP2 ′
− PE2 ) (11a)
aue
◮ Substituting into the discretized momentum and continuity equations, we
attempt to solve for the corrections U ′ , V ′ , and P ′ . av V ′

Vn1 ′
+ Vn2 = ∑ i v i + Dv (PP1
′ ′
− PN1 ′
) + Dv (PP2 ′
− PN2 ) (11b)
an
◮ Substituting these corrected values into the discretized momentum
equations gives: ◮ In the SIMPLE scheme, the two correction parts are chosen such that
aui aui Ui′
(Ue + Ue′ ) = ∑ (U + Ui′ ) + Du (PP − PE ) + Du (PP′ − PE′ ) + su (9a)

Ue1 ′
= Du (PP1 ′
− PE1 ) ′
Ue2 =∑ ′
+ Du (PP2 ′
− PE2 ) (12a)
aue i aue
a v avi Vi′
(Vn + Vn′ ) = ∑ vi (Vi + Vi′ ) + Dv (PP − PN ) + Dv (PP′ − PN′ ) + sv (9b) ′
Vn1 ′
= Dv (PP1 ′
− PN1 ) ′
Vn2 =∑ ′
+ Dv (PP2 ′
− PN2 ) (12b)
an avi
Pressure-Velocity Coupling 2008/9 9 / 16 Pressure-Velocity Coupling 2008/9 10 / 16

◮ This choice of split ensures a simple linkage between U1′ , V1′ and P1′ ◮ After collecting up terms the above equation can be written in the form
values. The more complicated term involving velocity corrections at
′ ′ ′ ′ ′
neighbouring nodes is put into the second part of the correction. ap PP1 = ae PE1 + aw PW 1 + an PN1 + as PS1 + Su (16)

◮ Since we want the corrected velocities to satisfy continuity, we now where


examine the discretized continuity equation. Integrating the continuity
equation over the pressure control volume leads to ae = ∆y [ρ Du ]e aw = ∆y [ρ Du ]w an = ∆x [ρ Dv ]n as = ∆x [ρ Dv ]s

(ρe Ue∗ − ρw Uw∗ )∆y + (ρn Vn∗ − ρs Vs∗ )∆x = 0 (13) ap = ae + aw + an + as Su = −Sm

or (ρe Ue′ − ρw Uw′ )∆y + (ρn Vn′ − ρs Vs′ )∆x = −Sm (14)
◮ This set of linear equations for the pressure correction P1′ can be solved,
where Sm = (ρe Ue − ρw Uw )∆y + (ρn Vn − ρs Vs )∆x is simply the mass using the same methods employed for solving the discretized momentum
imbalance arising from the original U, V field. equations.
◮ As a first approximation, we assume the second part of the corrections ◮ The corresponding corrections to the velocities, U1′ and V1′ , are then
(U2′ , V2′ , P2′ ) may be neglected. Substituting equations (12) into the obtained from equations (12), and the pressure and velocities are thus all
discretized continuity equation then results in updated.
∆y [ρ Du ]e (PP1 ) − [ρ Du ]w (PW
 ′ ′ ′ ′

− PE1 1 − PP1 )

+ ∆x [ρ Dv ]n (PP1 ) − [ρ Dv ]s (PS1
 ′ ′ ′ ′

− PN1 − PP1 ) = −Sm (15)

Pressure-Velocity Coupling 2008/9 11 / 16 Pressure-Velocity Coupling 2008/9 12 / 16


◮ Since the SIMPLE scheme is typically embedded within an iterative flow The PISO Scheme
solver, the whole algorithm is:
◮ One problem with the SIMPLE scheme as described is that it can be
1. Start with initial values
rather slow to converge.
2. Solve momentum equations to get U, V
◮ One reason for this is the neglect of the corrections U2′ , V2′ and P2′ .
3. Calculate coefficients and source terms for the pressure correction
equation (15) ◮ There are a number of variants on the SIMPLE scheme, designed to
improve convergence speeds.
4. Solve for the pressure corrections P1′
◮ One such variant, which is very similar in structure, and may sometimes
5. Calculate velocity corrections U1′ , V1′ from equations (12)
lead to better convergence, is the PISO (Pressure Implicit solution by
6. Update P, U, V Split Operator method) scheme proposed by Issa (1982).

7. Repeat from step 2 until solution has converged ◮ In the PISO scheme, the same decomposition of velocity and pressure
corrections is made as in the SIMPLE scheme, and U1′ , V1′ and P1′ are
computed as described earlier.

◮ However, a second corrector stage is now added, in an attempt to


account for the neglected U2′ , V2′ and P2′ in the first stage.

Pressure-Velocity Coupling 2008/9 13 / 16 Pressure-Velocity Coupling 2008/9 14 / 16

◮ In this second corrector stage the second parts of equations (12) are ◮ This can again be written in the generic form
approximated by
′ ′ ′ ′ ′
ap PP2 = ae PE2 + aw PW 2 + an PN2 + as PS2 + Su (20)
au U ′

Ue2 = ∑ i u i1 + Du (PP2
′ ′
− PE2 ) (17a)
ae where the ai coefficients are the same as those in the equation for the P1′
avi Vi1′ corrections, but the source term Su is now given by the right hand side of

Vn2 =∑ ′
+ Dv (PP2 ′
− PN2 ) (17b) equation (19) above.
avi
◮ These expressions are then substituted into the discretized continuity ◮ In the PISO scheme, therefore, between steps 6 and 7 of the SIMPLE
equation, which now becomes algorithm, the source terms Su of equation (20) are calculated, the
(ρe Ue2

− ρw Uw2

)∆y + (ρn Vn2

− ρs Vs2

)∆x = 0 (18) equation for P2′ is solved and the result used as a second correction to
the pressure.
◮ We thus get

∆y [ρ Du ]e (PP2 ) − [ρ Du ]w (PW
 ′ ′ ′ ′

− PE2 2 − PP2 )
+ ∆x [ρ Dv ]n (PP2 ) − [ρ Dv ]s (PS2
 ′ ′ ′ ′

− PN2 − PP2 ) =
aui Ui1
′ aui Ui1
′ 
  
∆y ρ ∑ u − ∆y ρ ∑ u
ap w ap e
av V ′ av V ′
   
+ ∆x ρ ∑ i v i1 − ∆x ρ ∑ i v i1 (19)
ap n ap s

Pressure-Velocity Coupling 2008/9 15 / 16 Pressure-Velocity Coupling 2008/9 16 / 16

Vous aimerez peut-être aussi