SIMPLE method small time step

dsr_cfd · June 29, 2016, 08:20

Hi, I have created a CFD code following the SIMPLE method. It is a co-located grid and uses Rhie-Chow interpolation. The problem is, I need to take rediculously small time steps for it to be stable.

Has anyone ever had this issue? Fixes?

Any response will be much appreciated!

FMDenaro · June 29, 2016, 10:18

Quote:

Originally Posted by dsr_cfd

Hi, I have created a CFD code following the SIMPLE method. It is a co-located grid and uses Rhie-Chow interpolation. The problem is, I need to take rediculously small time steps for it to be stable.

Has anyone ever had this issue? Fixes?

Any response will be much appreciated!

what discretization are you using in time? first order explicit? How about the Re_h?

dsr_cfd · June 29, 2016, 10:45

First order implicit. I've tried with multiple Re, but it seems the time step is always limited by a CFL of 1.0. Which seems very low for an implicit solver.

arjun · June 29, 2016, 22:33

Quote:

Originally Posted by dsr_cfd

Hi, I have created a CFD code following the SIMPLE method. It is a co-located grid and uses Rhie-Chow interpolation. The problem is, I need to take rediculously small time steps for it to be stable.

Has anyone ever had this issue? Fixes?

Any response will be much appreciated!

Your program has a bug and it could be anywhere. Instablity in cfd solver could creep in from anything from small to big.

It is not possible to point out bug from your information.

dsr_cfd · June 30, 2016, 08:30

But is it safe to say that there is definitely a bug? That I should be able to run with much higher time steps?

flotus1 · June 30, 2016, 09:05

Quote:

First order implicit. I've tried with multiple Re, but it seems the time step is always limited by a CFL of 1.0

With his kind of behavior, the first place to look for the error is your implementation of the temporal discretization.

dsr_cfd · June 30, 2016, 09:43

Could it be related to the rhie chow or pressure correction methods which both use the coefficient value from momentum (which depends on the time step)? Or any other suggestions? My temporal discretization method comes straight from the Versteeg textbook.

mprinkey · July 1, 2016, 22:30

What is the nature of the instability? flow variables going to infinity? If so, which one starts it? Or do linear systems fail to converge? If so, which field? Or does the nonlinear iteration fail to converge?

dsr_cfd · July 4, 2016, 07:44

The pressure correction equation goes unstable first. If I just do heat conduction, I can use a large time step. The instability seems to occur during velocity-pressure coupling.

FMDenaro · July 4, 2016, 07:58

Quote:

Originally Posted by dsr_cfd

The pressure correction equation goes unstable first. If I just do heat conduction, I can use a large time step. The instability seems to occur during velocity-pressure coupling.

just check after one time step if your pressure solution appears regular and if the gradients added to the partial velocity field satisfy the continyuity equation.

dsr_cfd · July 4, 2016, 08:08

If my time step is too big (not really even that big), my pressure correction equation diverges (first time step, first loop).

FMDenaro · July 4, 2016, 10:11

Quote:

Originally Posted by dsr_cfd

If my time step is too big (not really even that big), my pressure correction equation diverges (first time step, first loop).

that demonstrates you have some bugs in the code..check the BC.s for the pressure

dsr_cfd · July 4, 2016, 10:53

Ok thanks. What should I expect for time step (based on cell size, velocity, etc.)?

FMDenaro · July 4, 2016, 12:13

Quote:

Originally Posted by dsr_cfd

Ok thanks. What should I expect for time step (based on cell size, velocity, etc.)?

After a single time step do you satisfy the continuity equation?

mprinkey · July 5, 2016, 00:54

If you are using the SIMPLE method, then the algorithm is completely implicit and (in principle) there is no timestep limit. In fact, taking the limit of dt->inf should lead you to the normal steady SIMPLE formulation. This, of course, assumes that the SIMPLE algorithm (outer iterations) converge at each timestep.

I am aware of no theory...or even heuristic...that gives time-step guidance with regard to the convergence of SIMPLE. There are requirements on pressure and velocity underrelaxation factors for the steady state case under assumptions of simple models...no reactions, no buoyancy, etc.

I agree with the others advice. Verify that the face fluxes after pressure correction application results in conservation of mass. If that is not true both globally and in every cell, then your first problem to solve is there. There may be other bugs, but you will not be able to check because the fluxes driving the convection of other properties will be garbage.

And remember that because this is colocated, the cell values of velocity are (almost) irrelevant. The face fluxes are the primary metric with regard to pressure correction and mass conservation.

June 29, 2016, 08:20	SIMPLE method small time step	#1
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	Hi, I have created a CFD code following the SIMPLE method. It is a co-located grid and uses Rhie-Chow interpolation. The problem is, I need to take rediculously small time steps for it to be stable. Has anyone ever had this issue? Fixes? Any response will be much appreciated!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[Other] Contribution a new utility: refine wall layer mesh based on yPlus field	lakeat	OpenFOAM Community Contributions	58	December 23, 2021 03:36
AMI speed performance	danny123	OpenFOAM	21	October 24, 2020 05:13
Stuck in a Rut- interDyMFoam!	xoitx	OpenFOAM Running, Solving & CFD	14	March 25, 2016 08:09
How to write k and epsilon before the abnormal end	xiuying	OpenFOAM Running, Solving & CFD	8	August 27, 2013 16:33
pisoFoam with k-epsilon turb blows up - Some questions	Heroic	OpenFOAM Running, Solving & CFD	26	December 17, 2012 04:34

June 29, 2016, 10:45		#3
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	First order implicit. I've tried with multiple Re, but it seems the time step is always limited by a CFL of 1.0. Which seems very low for an implicit solver.

June 30, 2016, 08:30		#5
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	But is it safe to say that there is definitely a bug? That I should be able to run with much higher time steps?

June 30, 2016, 09:43		#7
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	Could it be related to the rhie chow or pressure correction methods which both use the coefficient value from momentum (which depends on the time step)? Or any other suggestions? My temporal discretization method comes straight from the Versteeg textbook.

July 1, 2016, 22:30		#8
mprinkey Senior Member Michael Prinkey Join Date: Mar 2009 Location: Pittsburgh PA Posts: 363 Rep Power: 25	What is the nature of the instability? flow variables going to infinity? If so, which one starts it? Or do linear systems fail to converge? If so, which field? Or does the nonlinear iteration fail to converge?

July 4, 2016, 07:44		#9
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	The pressure correction equation goes unstable first. If I just do heat conduction, I can use a large time step. The instability seems to occur during velocity-pressure coupling.

July 4, 2016, 08:08		#11
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	If my time step is too big (not really even that big), my pressure correction equation diverges (first time step, first loop).

July 4, 2016, 10:53		#13
dsr_cfd New Member DSR Join Date: Sep 2014 Posts: 11 Rep Power: 12	Ok thanks. What should I expect for time step (based on cell size, velocity, etc.)?

July 5, 2016, 00:54		#15
mprinkey Senior Member Michael Prinkey Join Date: Mar 2009 Location: Pittsburgh PA Posts: 363 Rep Power: 25	If you are using the SIMPLE method, then the algorithm is completely implicit and (in principle) there is no timestep limit. In fact, taking the limit of dt->inf should lead you to the normal steady SIMPLE formulation. This, of course, assumes that the SIMPLE algorithm (outer iterations) converge at each timestep. I am aware of no theory...or even heuristic...that gives time-step guidance with regard to the convergence of SIMPLE. There are requirements on pressure and velocity underrelaxation factors for the steady state case under assumptions of simple models...no reactions, no buoyancy, etc. I agree with the others advice. Verify that the face fluxes after pressure correction application results in conservation of mass. If that is not true both globally and in every cell, then your first problem to solve is there. There may be other bugs, but you will not be able to check because the fluxes driving the convection of other properties will be garbage. And remember that because this is colocated, the cell values of velocity are (almost) irrelevant. The face fluxes are the primary metric with regard to pressure correction and mass conservation.