CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > General Forums > Main CFD Forum

Dirichlet condition incremental pressure correction scheme

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree1Likes
  • 1 Post By FMDenaro

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   March 2, 2020, 05:22
Default Dirichlet condition incremental pressure correction scheme
  #1
New Member
 
Jack Tattersall
Join Date: Jan 2020
Posts: 13
Rep Power: 6
jacktat7 is on a distinguished road
Hello,

I was wondering to what extent it is / if it is physically corerct to prescribe a pressure dirichlet BC in an IPCS method. Due to the fact that I seem to remember that there is a slight discrepancy because you are prescribing the perssure, however the equation solves for the pressure correction.
So my question is, is it actually correct to prescirbe the pressure in an IPCS method.

Cheers,
Jack
jacktat7 is offline   Reply With Quote

Old   March 2, 2020, 07:46
Default
  #2
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,811
Rep Power: 72
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by jacktat7 View Post
Hello,

I was wondering to what extent it is / if it is physically corerct to prescribe a pressure dirichlet BC in an IPCS method. Due to the fact that I seem to remember that there is a slight discrepancy because you are prescribing the perssure, however the equation solves for the pressure correction.
So my question is, is it actually correct to prescirbe the pressure in an IPCS method.

Cheers,
Jack

Mathematically, the Poisson problem for the pressure is elliptic and the Dirichlet BC is a correct choice. However the problem is more general, you can prescribe physically a pressure value? That depends on the assumption of your problem. For example an outlet in a constant-pressure enviroment.

Do not forget that there is no physical thermodynamic pressure for incompressible flows.
FMDenaro is offline   Reply With Quote

Old   March 2, 2020, 16:55
Default
  #3
New Member
 
Pedro Costa
Join Date: Jul 2017
Posts: 9
Rep Power: 9
pcosta is on a distinguished road
Quote:
Originally Posted by jacktat7 View Post
Hello,

I was wondering to what extent it is / if it is physically corerct to prescribe a pressure dirichlet BC in an IPCS method. Due to the fact that I seem to remember that there is a slight discrepancy because you are prescribing the perssure, however the equation solves for the pressure correction.
So my question is, is it actually correct to prescirbe the pressure in an IPCS method.

Cheers,
Jack
Hi,

You can. There's just one think one should be careful about. What is important in a pressure correction scheme in this regard is that you cannot prescribe pressure and velocity boundary condition (BC) independently:
- If you prescribe a Dirichlet velocity BC, you want that the pressure correction step does not compromise this constrain on the velocity. So the pressure gradient normal to the boundary has to be zero (Neumann BC).
- Similarly, you can prescribe a zero pressure boundary condition (e.g. outflow), but then you should make sure that the velocity BC normal to the boundary is consistent with this condition. For that you can employ the pressure correction step also at the boundary.
pcosta is offline   Reply With Quote

Old   March 3, 2020, 08:25
Default
  #4
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,811
Rep Power: 72
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by pcosta View Post
Hi,

You can. There's just one think one should be careful about. What is important in a pressure correction scheme in this regard is that you cannot prescribe pressure and velocity boundary condition (BC) independently:
- If you prescribe a Dirichlet velocity BC, you want that the pressure correction step does not compromise this constrain on the velocity. So the pressure gradient normal to the boundary has to be zero (Neumann BC).
- Similarly, you can prescribe a zero pressure boundary condition (e.g. outflow), but then you should make sure that the velocity BC normal to the boundary is consistent with this condition. For that you can employ the pressure correction step also at the boundary.



Let me highlight that, the normal derivative of the pressure is not zero but it is expressed according to the Hodge decomposition, that is expressed by the momentum equation projected along the normal direction to the boundary



dp/dn = n.(a* -a )


The homogenous Neumann condition is an approximation that can generate errors if n.a* is not added to the source term of the Poisson equation.
FMDenaro is offline   Reply With Quote

Old   March 3, 2020, 08:37
Default
  #5
New Member
 
Pedro Costa
Join Date: Jul 2017
Posts: 9
Rep Power: 9
pcosta is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Let me highlight that, the normal derivative of the pressure is not zero but it is expressed according to the Hodge decomposition, that is expressed by the momentum equation projected along the normal direction to the boundary



dp/dn = n.(a* -a )


The homogenous Neumann condition is an approximation that can generate errors if n.a* is not added to the source term of the Poisson equation.
Perhaps we are talking about different things, so let me clarify. The boundary condition for the correction pressure has to be homogeneous Neumann, because if you discretise your pressure projection step at the boundary, it is the only boundary condition that will not disturb the velocity BC.
pcosta is offline   Reply With Quote

Old   March 3, 2020, 11:17
Default
  #6
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,811
Rep Power: 72
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by pcosta View Post
Perhaps we are talking about different things, so let me clarify. The boundary condition for the correction pressure has to be homogeneous Neumann, because if you discretise your pressure projection step at the boundary, it is the only boundary condition that will not disturb the velocity BC.



Are we talking about the closure for the pressure equation? If yes, the Poisson problem has the closure defined by the non homogeneous Neuman condition that satisfy the compatibility condition.
At a boundary, the pressure derivative must be never discretized by it is substituted into the Div Grad () operator.
That issue is general and does not depend on the specific incremental pressure, pressur-free or gauge method.

If I am wrong to understand you comment, please write down the equation you are talking about.
FMDenaro is offline   Reply With Quote

Old   March 3, 2020, 11:37
Default
  #7
New Member
 
Pedro Costa
Join Date: Jul 2017
Posts: 9
Rep Power: 9
pcosta is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Are we talking about the closure for the pressure equation? If yes, the Poisson problem has the closure defined by the non homogeneous Neuman condition that satisfy the compatibility condition.
At a boundary, the pressure derivative must be never discretized by it is substituted into the Div Grad () operator.
That issue is general and does not depend on the specific incremental pressure, pressur-free or gauge method.

If I am wrong to understand you comment, please write down the equation you are talking about.

I understand that perhaps I was a bit confusing. Let's not call this variable pressure, but the scalar potential \psi that projects the prediction velocity u* into a divergence-free space u^{n+1}.



When we solve the Poisson equation for this scalar \psi, often denoted correction pressure, we need to prescribe BCs for it. The question is: which BC shall we prescribe when we impose no-slip/no-penetration?


If we look at the correction step, we that:


u^{n+1} = u^* - dt grad \psi



(dt is the time step)



since at the boundary we want to impose a no-penetration BC: u^{n+1} = u* = u_b, we do not want this gradient to change the flow at the wall, so the gradient of this potential normal to the wall should to be zero at the boundary.
pcosta is offline   Reply With Quote

Old   March 3, 2020, 11:48
Default
  #8
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,811
Rep Power: 72
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by pcosta View Post
I understand that perhaps I was a bit confusing. Let's not call this variable pressure, but the scalar potential \psi that projects the prediction velocity u* into a divergence-free space u^{n+1}.



When we solve the Poisson equation for this scalar \psi, often denoted correction pressure, we need to prescribe BCs for it. The question is: which BC shall we prescribe when we impose no-slip/no-penetration?


If we look at the correction step, we that:


u^{n+1} = u^* - dt grad \psi



(dt is the time step)



since at the boundary we want to impose a no-penetration BC: u^{n+1} = u* = u_b, we do not want this gradient to change the flow at the wall, so the gradient of this potential normal to the wall should to be zero at the boundary.



Yes, that is what I am talking, too. According to your notation, project along the normal to the boundary (a wall but also possible to be inflow/outflow) your equation





un^{n+1} = un^* - dt d(psi)/dn



As you see, this is the projection of the Hodge decomposition. You have to use it as it is, just substitute the derivative into the Div Grad psi at the boundary. Then, you will see that un^* is exactly as same as in the source term, but with opposite sign. Thus, the source term has no longer un^* on the boundary but directly the physical velocity un^n+1.
Apparently, that could be seen as the application of the homogeneous Neumann condition but you have, correspondigly to modify also the source term close to the boundary. That will satisfy the compatibility condition and a solution exists apart a function of time.

I worked for some years on the theory of projection methods, if you are interested to the details of what I am talking you can read here in Remark 1

https://www.researchgate.net/publica...ary_conditions
FMDenaro is offline   Reply With Quote

Old   March 3, 2020, 12:06
Default
  #9
New Member
 
Pedro Costa
Join Date: Jul 2017
Posts: 9
Rep Power: 9
pcosta is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Yes, that is what I am talking, too. According to your notation, project along the normal to the boundary (a wall but also possible to be inflow/outflow) your equation





un^{n+1} = un^* - dt d(psi)/dn



As you see, this is the projection of the Hodge decomposition. You have to use it as it is, just substitute the derivative into the Div Grad psi at the boundary. Then, you will see that un^* is exactly as same as in the source term, but with opposite sign. Thus, the source term has no longer un^* on the boundary but directly the physical velocity un^n+1.
Apparently, that could be seen as the application of the homogeneous Neumann condition but you have, correspondigly to modify also the source term close to the boundary. That will satisfy the compatibility condition and a solution exists apart a function of time.

I worked for some years on the theory of projection methods, if you are interested to the details of what I am talking you can read here in Remark 1

https://www.researchgate.net/publica...ary_conditions

Thanks for your comment and reference. I agree with your point. I guess that the difficulty is that in a staggered grid the source term is not defined at the boundary, and since it will cancel out in the pressure correction step, it is simpler to impose a homogeneous Neumann BC for solving the Poisson equation.
pcosta is offline   Reply With Quote

Old   March 3, 2020, 12:19
Default
  #10
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,811
Rep Power: 72
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by pcosta View Post
Thanks for your comment and reference. I agree with your point. I guess that the difficulty is that in a staggered grid the source term is not defined at the boundary, and since it will cancel out in the pressure correction step, it is simpler to impose a homogeneous Neumann BC for solving the Poisson equation.



The key is that you can indeed set the homogenoeus Neumann condition (see Remark 1 in the paper) but when you look at the source term it has the Div v* expression so that on a staggered grid the component n.v* appears. But actually you do not need at all to use it as the Hodge decomposition is used and you simply cancel it. But doing that the source term must be computed using n.v^n+1 on the boundary. In conclusion, if you prescribe dpsi/dn= 0 you must also change the source term accordingly. This way the similarity between the twoi BCs is mathematically correct.

This point is often not fully understood and the consequence is that the compatibility condition is not satisfied and people are forced to set an arbitrary value for the pressure to get a convergent solution.
pcosta likes this.
FMDenaro is offline   Reply With Quote

Reply

Tags
boundary condition, correction, ipcs, pressure

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Wind tunnel Boundary Conditions in Fluent metmet FLUENT 6 October 30, 2019 12:23
how to replace a shock tube high pressure port by a boundary condition? immortality OpenFOAM Pre-Processing 0 May 2, 2013 13:43
Fluent natural ventilation pressure boundary condition pierresandre FLUENT 24 November 8, 2011 14:32
Neumann pressure BC and velocity field Antech Main CFD Forum 0 April 25, 2006 02:15
Hydrostatic pressure in 2-phase flow modeling (CFX4.2) HB &DS CFX 0 January 9, 2000 13:19


All times are GMT -4. The time now is 21:05.