Does MULES::implicitsolve allow for Courant numbers > 1?

Philip · May 31, 2011, 04:00

Hello,

although there are several threads concerning the time step size for MULES VOF (in interFoam and/or interDyMFoam) I have not found a definitive answer whether you (I/OpenFOAM) can safely increase the time step size, or that anyway the max Courant number should still remain < 1?

thanks,

Philip

akidess · May 31, 2011, 06:09

No, you are still bound by the PISO algorithm to Co < 1. However, using explicit MULES you are bound by the capillary wave length, which usually gives a time step criterion dt << (Co < 1). MULES implicit somewhat relaxes that time step criterion.

Philip · May 31, 2011, 06:37

Hi Akidess,

thanks for this fast reply. I am an openfoam beginner and I did not know that all PISO (so all transient Navier-Stokes) calculations were limited by Co < 1. Do you know why this is so? Apart from losing accuracy for larger steps, I am not aware of stability issues with eg backward Euler time integration and SIMPLE (or PISO).

thanks,

Philip

akidess · May 31, 2011, 06:54

The restriction is not OpenFOAM specific, but tied to the PISO algorithm itself. If you read the derivation of the algorithm, there is a step where it is assumed that the velocity only slightly changes between time steps. This is the reason why PISO does not require relaxation and usually converges within 2-3 iterations. With SIMPLE you don't have the same time step restriction, but you'll have to do a lot more work to get a converged solution for a given time step.

You can read more about this in the original paper of the PISO algorithm:
***
R. I. Issa, Solution of the implicitly discretised fluid flow equations by operator-splitting, Journal of Computational Physics, Volume 62, Issue 1, January 1986, Pages 40-65, ISSN 0021-9991, DOI: 10.1016/0021-9991(86)90099-9.
(http://www.sciencedirect.com/science...21999186900999)

Philip · May 31, 2011, 07:25

Hi again,

Thanks. I will look up the original article of Issa.
I found a similar discussion under the title "Courant number and implicit treatment thread":

http://www.cfd-online.com/Forums/ope...treatment.html

that make things a bit clearer. At this moment I still wonder if this has been tried: Looping (within a time step) over the momentum solver + corrections does not allow for a larger time step?

Philip

May 31, 2011, 04:00	Does MULES::implicitsolve allow for Courant numbers > 1?	#1
Philip New Member Philip Simons Join Date: Mar 2009 Posts: 3 Rep Power: 17	Hello, although there are several threads concerning the time step size for MULES VOF (in interFoam and/or interDyMFoam) I have not found a definitive answer whether you (I/OpenFOAM) can safely increase the time step size, or that anyway the max Courant number should still remain < 1? thanks, Philip

May 31, 2011, 06:09		#2
akidess Senior Member Anton Kidess Join Date: May 2009 Location: Germany Posts: 1,377 Rep Power: 30	No, you are still bound by the PISO algorithm to Co < 1. However, using explicit MULES you are bound by the capillary wave length, which usually gives a time step criterion dt << (Co < 1). MULES implicit somewhat relaxes that time step criterion. PonchO likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
1.7.x -> buoyantPimpleFoam -> hRhoThermo -> incompressible and icoPoly3ThermoPhysics?	will.logie	OpenFOAM Programming & Development	1	February 16, 2011 21:52
CAD -> gMsh -> enGrid -> OpenFOAM Problem	AlGates	OpenFOAM	7	August 6, 2010 13:46
IcoFoam parallel woes	msrinath80	OpenFOAM Running, Solving & CFD	9	July 22, 2007 03:58
HELP!! Prandtl numbers in RNG k-e turbulence model	maoasis	FLUENT	0	April 24, 2006 11:51
Could anybody help me see this error and give help	liugx212	OpenFOAM Running, Solving & CFD	3	January 4, 2006 19:07

May 31, 2011, 06:37		#3
Philip New Member Philip Simons Join Date: Mar 2009 Posts: 3 Rep Power: 17	Hi Akidess, thanks for this fast reply. I am an openfoam beginner and I did not know that all PISO (so all transient Navier-Stokes) calculations were limited by Co < 1. Do you know why this is so? Apart from losing accuracy for larger steps, I am not aware of stability issues with eg backward Euler time integration and SIMPLE (or PISO). thanks, Philip

May 31, 2011, 06:54		#4
akidess Senior Member Anton Kidess Join Date: May 2009 Location: Germany Posts: 1,377 Rep Power: 30	The restriction is not OpenFOAM specific, but tied to the PISO algorithm itself. If you read the derivation of the algorithm, there is a step where it is assumed that the velocity only slightly changes between time steps. This is the reason why PISO does not require relaxation and usually converges within 2-3 iterations. With SIMPLE you don't have the same time step restriction, but you'll have to do a lot more work to get a converged solution for a given time step. You can read more about this in the original paper of the PISO algorithm: *** R. I. Issa, Solution of the implicitly discretised fluid flow equations by operator-splitting, Journal of Computational Physics, Volume 62, Issue 1, January 1986, Pages 40-65, ISSN 0021-9991, DOI: 10.1016/0021-9991(86)90099-9. (http://www.sciencedirect.com/science...21999186900999)

May 31, 2011, 07:25		#5
Philip New Member Philip Simons Join Date: Mar 2009 Posts: 3 Rep Power: 17	Hi again, Thanks. I will look up the original article of Issa. I found a similar discussion under the title "Courant number and implicit treatment thread": http://www.cfd-online.com/Forums/ope...treatment.html that make things a bit clearer. At this moment I still wonder if this has been tried: Looping (within a time step) over the momentum solver + corrections does not allow for a larger time step? Philip