CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM (http://www.cfd-online.com/Forums/openfoam/)
-   -   Running stably with a large Courant number using pimpleFoam (http://www.cfd-online.com/Forums/openfoam/69466-running-stably-large-courant-number-using-pimplefoam.html)

mlawson October 23, 2009 16:14

Running stably with a large Courant number using pimpleFoam
 
2 Attachment(s)
I am attempting to simulate laminar incompressible pulsatile flow in a complex geometry. To adequately resolve my geometry I need at least 15 million cells. I am driving the flow by setting the pressure and the outlet and oscillation pressure at the inlet.

Currently I am using pimpleFoam with automatic timestep adjustment to maintain a max Courant number of 2. Using a higher Courant number causes the solution to become unstable.

My problem is that at a max Courant number of 2 my timesteps are between 1E-5 and 1E-6 seconds. I need to simulate at least 1 second of time, which would require at least 100,000 timesteps at my current timestep. Given the size of my model, solving this many timesteps is computationally prohibitive.

I'm using fvSolution and fvSchemes dictionaries directly from the pimpleFoam tutorial in OF-1.6 (see attached).

Does anyone have any suggestions on how I could increase my timesteps by a couple orders of magnitude to make this simulation feasible? Should I be using a different solver? Should I be using a different discretization scheme?

Any help would be appreciated,
-Mike

Djub July 13, 2012 06:04

Hi Mike,

Did you solve your problem?
I am trying to run a case with a very large Reynolds number (2.5e8). So I would like also to run with a large Co (Courrent flow) to limit my time calculation.
Does anyone has a solution? Or maybe an explanation that what I am looking for does not exist (and cannot exist)? Maybe the Co cannot be greater thant 1 in an unsteady flow?
In that case, is there in OpenFoam any Explicit algorithm for unsteady, incompressible, turbulent, newtonian flow ?

crixman November 24, 2014 10:27

Dear all, any news on this topic?
I was wondering if there was a way to detach Co from the pimple solver, being able to use fixed (and larger!) time steps, as I think it's done in Fluent or other solvers.
Any ideas or suggestions?

alexeym November 24, 2014 11:25

Use first-order schemes. With upwind scheme I was able to run a case (quite simple - von Karman vortex street) with Co up to 50.

crixman November 24, 2014 13:05

Interesting to know, did you use turbulence schemes as well?
I could give it a try to start the simulation then switch to second order schemes, even if I'm not sure that the Co jump will be too big to be handled by the solver then!

alexeym November 24, 2014 15:11

No, simulation was at rather low Re (100).

crixman November 24, 2014 17:32

I'm making an extreme test, I'll keep you updated if I could reach an interesting value of Co (turbomachinery with kOmegaSST)!

Tobi November 24, 2014 18:13

Hi, you should Check the pimple stuft at the Wikipedia Site.

crixman November 24, 2014 18:28

Hi Tobi, sure, wanted to learn more and try out to see how far I can push it!

Tobi November 24, 2014 18:33

It Is also in my blog.
And before I ment the open foam wiki

crixman November 24, 2014 18:38

I will look it through thanks and post some updates as soon as I got time!

crixman November 29, 2014 11:37

Hi Tobi, I read through your post about Pimple and I am now running at Co=12.5 and everythign seems to converge well!
As I am using upwind schemes for U,k and omega, when switching to linear schemes, should I drop my Co below 1?


All times are GMT -4. The time now is 06:49.