CFL and CFL_ADAPT with Unsteady Simulations

tjim · April 11, 2019, 01:51

Hi,

I have a few questions on the Unsteady implementation.

1. Are unsteady simulations affected by CFL and CFL_ADAPT flags?

2. If running UNST_TIMESTEP=TIME_STEPPING (not 2nd order), are UNST_TIME and UNST_INT_ITER ignored?

Thanks and regards,
Tim

atelcikti1 · April 11, 2019, 03:41

Hi,

I never run unsteady simulation on SU2 but I am very used to unsteady simulations.

SU2 uses dual-time step algorithm as used in Fluent. There are two time levels in one total iteration. First one takes place in pseudo time which is converged in few iterations (generally 5 to 30) according to your real time step, CFL etc. The second one is marching in time simply.

So I think this is your answer. CFL is important for pseudo time level.

I strongly recommend to look at Blazek book, chapter 6, temporal discretization.

tjim · April 11, 2019, 03:51

Hi,

Thanks for your reply and tip. Sorry, I was actually asking about the implementation in SU2 for single time stepping (not dual-time stepping) - I wanted to find out which flags are ignored and which are used when the UNSTEADY flags are given.

Kind regards,
Tim

atelcikti1 · April 11, 2019, 04:02

Hi,

In single time step, I think it uses only real time step (not cfl etc.) and march in time explicitly. I hope, someone can clear this subject.

Nevertheless, I can say if time accuracy is not important for you, use dual time stepping method. You can go with larger time steps that speed up your run.

Best regards.

April 11, 2019, 01:51	CFL and CFL_ADAPT with Unsteady Simulations	#1
tjim New Member Tim Jim Join Date: Sep 2017 Posts: 25 Rep Power: 9	Hi, I have a few questions on the Unsteady implementation. 1. Are unsteady simulations affected by CFL and CFL_ADAPT flags? 2. If running UNST_TIMESTEP=TIME_STEPPING (not 2nd order), are UNST_TIME and UNST_INT_ITER ignored? Thanks and regards, Tim

April 11, 2019, 03:41		#2
atelcikti1 Member Join Date: Mar 2017 Posts: 61 Rep Power: 9	Hi, I never run unsteady simulation on SU2 but I am very used to unsteady simulations. SU2 uses dual-time step algorithm as used in Fluent. There are two time levels in one total iteration. First one takes place in pseudo time which is converged in few iterations (generally 5 to 30) according to your real time step, CFL etc. The second one is marching in time simply. So I think this is your answer. CFL is important for pseudo time level. I strongly recommend to look at Blazek book, chapter 6, temporal discretization.

April 11, 2019, 03:51		#3
tjim New Member Tim Jim Join Date: Sep 2017 Posts: 25 Rep Power: 9	Hi, Thanks for your reply and tip. Sorry, I was actually asking about the implementation in SU2 for single time stepping (not dual-time stepping) - I wanted to find out which flags are ignored and which are used when the UNSTEADY flags are given. Kind regards, Tim

April 11, 2019, 04:02		#4
atelcikti1 Member Join Date: Mar 2017 Posts: 61 Rep Power: 9	Hi, In single time step, I think it uses only real time step (not cfl etc.) and march in time explicitly. I hope, someone can clear this subject. Nevertheless, I can say if time accuracy is not important for you, use dual time stepping method. You can go with larger time steps that speed up your run. Best regards.