Plateauing Residual

gpyr · February 21, 2024, 20:59

Hi all,

I am very new to OpenFOAM and am trying to build a model to simulate wind over a ship. The build case is adopted from the motorbike tutorial files. I was able to perform the simulation but noticed that the residuals for the velocity fields are relative high (around 1e-2 to 1e-3 range) and it plateaus. Quick search online indicates two issues: (1) the residual should not plateau, and (2) the residual of in the order of 1e-2 to 1e-3 is too high. In addition to that, I also notice that the checkMesh function concluded with one failure (probably due to skewFaces).

I tried refining the mesh size (from 200k to 1M) but it didn't solve the issue. I am thinking that I may not have used the case file correctly and would like to seek some help please?

Case file uploaded to googledrive (https://drive.google.com/file/d/1NcM...usp=drive_link) and the residual plots shown.

petros · February 22, 2024, 14:31

Hi,

The fact that the residuals appear to flatten out is not a problem; it rather shows that there is convergence which is usually a good sign. Of course convergence doesn't mean necessarily accuracy.

I think in your case you could achieve residuals of at least an order of magnitude lower by using the linearUpwind scheme for the turbulent quantities with cellLimited Gauss linear 1 as a gradScheme.

Hope that helps.
Petros

Alczem · February 26, 2024, 06:51

I agree with the previous comment, and would also add that the residuals are not the only check you should perform to monitor convergence. Try to monitor quantites of interest such as forces and total flowrates and make sure they also are stable. I tend to ignore residuals until the last iteration, except if my physical values are not converging or if the simulation diverges, but you will find plenty of different opinions about monitoring convergence.

Tobermory · February 28, 2024, 06:41

Just to add to the above comments, I have often seen quoted the view that "the solution residuals should drop by 3 or four orders of magnitude". Whilst I applaud the sentiment, this is way too general and as both the previous posters have stated, residual reduction is only part of the story.

For example, think about the following - what if your initial solution guess is extremely good for the majority of the domain? Indeed, what if it is the perfect solution?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
laplacianFoam with source term	Herwig	OpenFOAM Running, Solving & CFD	17	November 19, 2019 14:47
Segmentation fault when using reactingFOAM for Fluids	Tommy Floessner	OpenFOAM Running, Solving & CFD	4	April 22, 2018 13:30
chtMultiRegionSimpleFoam turbulent case	Aditya Patil	OpenFOAM Running, Solving & CFD	6	April 24, 2017 23:13
simpleFoam error - "Floating point exception"	mbcx4jc2	OpenFOAM Running, Solving & CFD	12	August 4, 2015 03:20
calculation stops after few time steps	sivakumar	OpenFOAM Running, Solving & CFD	7	March 17, 2013 07:37

February 22, 2024, 14:31		#2
petros Member Petros Ampatzidis Join Date: Oct 2018 Location: Bath, UK Posts: 64 Rep Power: 8	Hi, The fact that the residuals appear to flatten out is not a problem; it rather shows that there is convergence which is usually a good sign. Of course convergence doesn't mean necessarily accuracy. I think in your case you could achieve residuals of at least an order of magnitude lower by using the linearUpwind scheme for the turbulent quantities with cellLimited Gauss linear 1 as a gradScheme. Hope that helps. Petros

February 26, 2024, 06:51		#3
Alczem Senior Member Join Date: Dec 2021 Posts: 251 Rep Power: 5	I agree with the previous comment, and would also add that the residuals are not the only check you should perform to monitor convergence. Try to monitor quantites of interest such as forces and total flowrates and make sure they also are stable. I tend to ignore residuals until the last iteration, except if my physical values are not converging or if the simulation diverges, but you will find plenty of different opinions about monitoring convergence.

February 28, 2024, 06:41		#4
Tobermory Senior Member Join Date: Apr 2020 Location: UK Posts: 745 Rep Power: 14	Just to add to the above comments, I have often seen quoted the view that "the solution residuals should drop by 3 or four orders of magnitude". Whilst I applaud the sentiment, this is way too general and as both the previous posters have stated, residual reduction is only part of the story. For example, think about the following - what if your initial solution guess is extremely good for the majority of the domain? Indeed, what if it is the perfect solution?