Residuals oscillating about a fixed value in transient shear layer analysis

rafr894 · September 9, 2022, 09:57

Hi,

I am trying to simulate the flow resulting from the interaction of multiple shear layers in a 2D domain. I have a step velocity and temperature profile at the inlet, and would like to see the effect of dissipation due to the growing shear layers on the resultant velocity and temperature profiles at the outlet. I am running a
a transient DES model with a time step size ~1e-6s.
The contour plot of temperature attached does seem like what would physically happen in the above scenario. However, my residuals oscillate about a fixed value, and do not seem like then will vary any more. What can I make of these residuals? Is there a way to lower them? If not, can I say my results are still physically valid?
Any suggestions/comments to help me understand would be appreciated. Thank you.

LuckyTran · September 9, 2022, 14:35

No picture. But if your residuals oscillate in the sinusoidal fashion then something is very very very wrong. But if it looks like a sawtooth function then it is very very right. Residuals should jump abruptly after each time-step and then decrease as the solution converges for that time-step.

rafr894 · September 14, 2022, 05:35

Thanks for your reply. Sorry about not attaching the image. Please find the residual plot and temperature contours attached below.
As you can see, there is a visible sawtooth, but no decay in the residuals. I have around 50 inner iterations per time-step. Is the behavior normal?
Further, I plotted the net mass flowrate through the flow domain (plot attached), and notice that there is a mass imbalance. There is no warning of reverse flow, but when I ran the case as steady state, the net mass flowrate was zero after a few initial oscillations. I suppose this would prove that my boundary conditions for the problem are well defined?
Thanks.

LuckyTran · September 14, 2022, 13:05

That is the sawtooth you are looking for. They go from 1 to 1e-12 every timestep. Of course there are more checks you can do to ensure your solution is converged every timestep, but yes it is a very normal looking residual plot.

It depends on your initial conditions and boundary conditions what the mass flow plot means. It is a transient calculation. You can have mass imbalance in transient cases, i.e. you can fill a tank with water and empty a tank. You can't have mass imbalance in steady cases because filling up a tank with water forever would imply you have a black hole or a white hole somewhere. But mass imbalances are allowed for transient cases.

In the first 10 timesteps there is a clear horizontal for the mass flowing showing it is converged. However, later timesteps it becomes less noticeable, which is a bit of a concern. It looks like your case is converging less well at later timesteps. Might become an issue eventually.

September 9, 2022, 09:57	Residuals oscillating about a fixed value in transient shear layer analysis	#1
rafr894 New Member KristianRanta Join Date: Oct 2016 Location: Sundsvall Posts: 29 Rep Power: 10	Hi, I am trying to simulate the flow resulting from the interaction of multiple shear layers in a 2D domain. I have a step velocity and temperature profile at the inlet, and would like to see the effect of dissipation due to the growing shear layers on the resultant velocity and temperature profiles at the outlet. I am running a a transient DES model with a time step size ~1e-6s. The contour plot of temperature attached does seem like what would physically happen in the above scenario. However, my residuals oscillate about a fixed value, and do not seem like then will vary any more. What can I make of these residuals? Is there a way to lower them? If not, can I say my results are still physically valid? Any suggestions/comments to help me understand would be appreciated. Thank you.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] Layer coverage at 98% with collapses at corners	oryx	OpenFOAM Meshing & Mesh Conversion	1	December 4, 2023 20:04
transient analysis	a.e.	FLUENT	0	July 19, 2021 17:51
[snappyHexMesh] Prism layer collapsing in refinement transition	rubenparedes	OpenFOAM Meshing & Mesh Conversion	8	May 13, 2021 11:55
[snappyHexMesh] Holes in internal mesh when adding boundary layer snappyHex	otaolafr	OpenFOAM Meshing & Mesh Conversion	3	February 8, 2021 09:19
errors	Fahad	Main CFD Forum	0	March 23, 2004 14:20

September 9, 2022, 14:35		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,761 Rep Power: 66	No picture. But if your residuals oscillate in the sinusoidal fashion then something is very very very wrong. But if it looks like a sawtooth function then it is very very right. Residuals should jump abruptly after each time-step and then decrease as the solution converges for that time-step.

September 14, 2022, 13:05		#4
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,761 Rep Power: 66	That is the sawtooth you are looking for. They go from 1 to 1e-12 every timestep. Of course there are more checks you can do to ensure your solution is converged every timestep, but yes it is a very normal looking residual plot. It depends on your initial conditions and boundary conditions what the mass flow plot means. It is a transient calculation. You can have mass imbalance in transient cases, i.e. you can fill a tank with water and empty a tank. You can't have mass imbalance in steady cases because filling up a tank with water forever would imply you have a black hole or a white hole somewhere. But mass imbalances are allowed for transient cases. In the first 10 timesteps there is a clear horizontal for the mass flowing showing it is converged. However, later timesteps it becomes less noticeable, which is a bit of a concern. It looks like your case is converging less well at later timesteps. Might become an issue eventually.