Oscillating residuals. Is it okay?

Ana Herrero · September 19, 2020, 07:04

Hi,

I am running some simulations on a car with strong boundary separation and wake. I have realized that I don't get smooth residuals convergence. I know it may be normal if there are strong turbulences, but still, I think it is oscillating a lot, especially when I run a refined mesh. I have attached some pictures. What do you think?? is there any numerical frame (maximum and minimum) the residuals should always keep within?? I understand that the smoother the residuals, the better (less error) the results, right?

arjun · September 19, 2020, 07:26

in the second one the fluctuation is too much but rest of them look normal for starccm (looks like this is what you used).

Ana Herrero · September 19, 2020, 08:02

Hi Arjun,

the second one is the one with the refined mesh. What can it be due to?? may it have any effect to use couple solver instead of segregated?? it is a steady incompressible simulation

arjun · September 19, 2020, 08:41

Quote:

Originally Posted by Ana Herrero

Hi Arjun,

the second one is the one with the refined mesh. What can it be due to?? may it have any effect to use couple solver instead of segregated?? it is a steady incompressible simulation

It is bascially the turbulence model trying to diverge and solver trying to bring it back.

There are few things you can do. First put the under-relaxation for viscosity and momentum down. Third check the maximum viscosity ratio and it bring it to the values that make more sense.

I would just stop the simulation when it was doing good and check the viscosity ratio. Then i would just set the maximum viscosity ratio to 100 times of that.

MikeBravo · September 19, 2020, 11:45

Hi Ana,

I think you should switch to an unsteady scale-resolving simulation, DES or Dynamic LES. The residuals will come down once you switch to unsteady. I wouldn't trust steady or unsteady RANS for simulating the wake behind a vehicle.

FMDenaro · September 19, 2020, 14:13

There is no physical meaning in an oscillating residual. Someone thinks that it could be seen as an unsteady effect due to the physics of the problem (turbulence). No. First, one should consider an unsteady statistical formulation (URANS) but this is flawed if not associated to a full 3D approach.

Ana Herrero · September 19, 2020, 14:36

okay, I am going to give both options a try (redefined the viscosity ratio and run in unsteady. Should I keep the coupled solver (Automatic CFL control) or the segregated would be better?
with implicit unsteady, K-omega SST, time step 0.001 and 2nd order discretization. Does that set up sounds good??
I always thought that a steady simulation with a correct st up was perfectly capable of capturing the turbulences in an aero run.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Oscillating residuals	juan4815	CFX	5	October 11, 2017 12:59
Oscillating residuals in steady flow problem	Mohit Singh	Main CFD Forum	3	February 23, 2016 14:39
Reason for oscillating residuals	fayaazhussain	Main CFD Forum	3	August 21, 2014 10:31
Oscillating residuals and no convergence!	Sherlock_1812	OpenFOAM Running, Solving & CFD	7	August 20, 2014 09:39
Oscillating Residuals in Finer Mesh Case	Jake	FLUENT	3	June 16, 2005 09:34

September 19, 2020, 07:26		#2
arjun Senior Member Arjun Join Date: Mar 2009 Location: Nurenberg, Germany Posts: 1,272 Rep Power: 34	in the second one the fluctuation is too much but rest of them look normal for starccm (looks like this is what you used).

September 19, 2020, 08:02		#3
Ana Herrero New Member Ana Join Date: Sep 2020 Posts: 10 Rep Power: 5	Hi Arjun, the second one is the one with the refined mesh. What can it be due to?? may it have any effect to use couple solver instead of segregated?? it is a steady incompressible simulation

September 19, 2020, 11:45		#5
MikeBravo New Member Mike Join Date: Jan 2018 Posts: 29 Blog Entries: 6 Rep Power: 8	Hi Ana, I think you should switch to an unsteady scale-resolving simulation, DES or Dynamic LES. The residuals will come down once you switch to unsteady. I wouldn't trust steady or unsteady RANS for simulating the wake behind a vehicle.

September 19, 2020, 14:13		#6
FMDenaro Senior Member Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,768 Rep Power: 71	There is no physical meaning in an oscillating residual. Someone thinks that it could be seen as an unsteady effect due to the physics of the problem (turbulence). No. First, one should consider an unsteady statistical formulation (URANS) but this is flawed if not associated to a full 3D approach.

September 19, 2020, 14:36		#7
Ana Herrero New Member Ana Join Date: Sep 2020 Posts: 10 Rep Power: 5	okay, I am going to give both options a try (redefined the viscosity ratio and run in unsteady. Should I keep the coupled solver (Automatic CFL control) or the segregated would be better? with implicit unsteady, K-omega SST, time step 0.001 and 2nd order discretization. Does that set up sounds good?? I always thought that a steady simulation with a correct st up was perfectly capable of capturing the turbulences in an aero run.