CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Nonphysical flow field while using coodles solver (http://www.cfd-online.com/Forums/openfoam-solving/59193-nonphysical-flow-field-while-using-coodles-solver.html)

ankgupta8um November 24, 2007 18:18

Hello, I am using coodles s
 
Hello,

I am using coodles solver in OpenFOAM 1.4.1 to simulate flow through a cylinder. The computational domain is very simple and consists of an annular pipe attached to a cylinder. This is an expansion problem where the flow through the annular pipe expands into the cylinder (annular jet expanding into a cylindrical pipe). The boundary conditions I am using are:
- fixed value velocity inlet at the annular cross-section; zero gradient for pressure.
- no-slip walls for annlar pipe and cylindrical pipe; zero gradient for pressure.
- waveTrnasmissive pressure BC at the downstream end of the cylinder; inletOutlet BC for velocity.

I am not able to get a reasonable flow field in the computational domain. It seems to me that the unphysical pressure waves in the domain are destroying the numerical solution. Initially, I used the default solver settings as given in the pitzdaily tutorial. Then I tried using recommended convective discretization schemes (eg, fliteredLinear) for LES but that didn't help. I also tried using upwind scheme with the hope that the increased dissipation will dissipate out the spurious pressure waves, but that also didn't help.
I tried different initializations at the start of the run (starting from the RANS solution, starting from a quiscent medium) but all led to unphysical flow field.

I think that the unphysical pressure waves are destroying the solution. The reason being: I ran a case where instead of using walls for the cylindrical surface, I used pressure transmissive boundaries (i.e., no confinement; annular jet expanding into an eclosure with pressure transmissive boundaries at the cylindrical periphery and at the downstream end). For this case, I could get a physical flow field. That suggests that the numerical disturbances in the form of spurious pressure waves exit the computational domain from the peripheral boundary of the cylinder when pressure Transmissibe BC is used there. When wall is used it reflects back and destroys the solution.

I have been trying to resolve this issue in coodles with the use of walls close to the jet for quite some time now but haven't had any success. I am stuck with this issue and need to resolve it in order to make some progress.

I would really appreciate it if someone can throw some light on what the potential causes for this problem are and how to resolve it. People having similar experiences with coodles are kindly requested to share them with the forum. Any information on getting coodles solver to work is also welcome !!

Thanks!
Regards,
Ankur

msrinath80 November 25, 2007 13:55

What time discretization are y
 
What time discretization are you using?

ankgupta8um November 25, 2007 17:20

Hello Srinath, I am using b
 
Hello Srinath,

I am using backward scheme for time discretization.

Regards,
Ankur

ankgupta8um January 22, 2008 01:55

Hello, I am using coodles s
 
Hello,

I am using coodles solver with adaptive time stepping to solve for a low mach number variable density flow through a cylinder with an annular inlet. I was encountering the non-physical pressure field in my computational domain. I ran some test cases and observed the following behavior:
1. The pressure field looks reasonable when I use a large time step (say, corresponding to a maximum courant number of 2). If I reduce the allowed maximum courant number to say 0.75, I see the nonphysical pressure field in my domain.
2. If I increase the number of correctors in the PISO algorithm from a default value of 2 to 5, I could get the physical pressure field.

I am not able to understand the cause of this behavior. I would highly appreciate if some body can throw some light here.

Thanks!!
Regards,
Ankur

lillberg January 22, 2008 16:26

Hi Ankur, Try using Crank-Nic
 
Hi Ankur,
Try using Crank-Nicholson time stepping with some Euler for stability, e.g.

ddtSchemes
{
default CrankNicholson 0.9;
}

And try keeping your maximum Co < 0.5.

Good luck!

//Eric

ankgupta8um January 26, 2008 17:54

Hi Eric and others, I tried
 
Hi Eric and others,

I tried running my case with CrankNicholson 0.9 with maximum Co < 0.5, but it didn't help. That again is resulting in the nonphysical pressure field and thus nonphysical velocity fields.
So far, from all the test runs I have made, I could get the physical pressure and velocity fields only with a higher number of correctors (~5) in the PISO algorithm.
Any comment on how to get reasonable pressure field with lesser number of correctors (~2) in the PISO algorithm is greatly appreciated. The use of a higher number of correctors tremendously increases the computational time.

Thanks!!
Regards,
Ankur


All times are GMT -4. The time now is 06:33.