@Maciek @AnnaFless
Thank you for the information. It is clear now!
Hi, I have a question about surface refinement level that we have to use for this homework, i have done a first simulation with refinement levels 7-8 following the tutorial and the results achieved convergence in about 20 minutes (32cores) with mesh created in about 11 minutes. After i have setted the same simulation with finer mesh (refinement levels 8-9), and the time for create mesh and to achieve convergence increased respectively to 33 minutes and 77 minutes! I had a look at the results in terms of force plots and seems similar in their final values, another thing is that, with refinement parameters setted like in the tutorial, the simulation time in the first run stops at about 550 seconds whereas in the second run, with finer mesh, it stops at the setted end (1000seconds).
How can I evaluate in post processing if the benefits of finer mesh balance the increase of execution time and if is it necessary to use finer mesh?
Thanks, Best regards
Here the link to my project
1 Like
Well, without geting too deep in the theory, I can tell you that:
-
Increasing the refinement level is equivalent to specify smaller cells, so the total number of elements increases, if I am not wrong in this particular case it doubles from 2+ million to 4+ million. That explains the increase in computation time (and memory usage, tough I think you didn’t notice that).
-
I think the convergence is expected to be faster for the model with the less number of cells. That is because the bigger cells capture less variations in the flow. You can think of this as some kind of ponderation over the flow variables. But you risk losing accuracy.
-
If the simulation reaches the 1000 steps (be careful with this, it is not seconds), this means it hasn’t converged. This is, at least to the control level indicated in the numerics section (default residual of 1e-5). If it reaches this level for all variables before the maximum number of steps, the simulation stops (in your case at step 550). Anyway, this doesn’t assure that the result is correct.
-
To asses the benefits of a finer mesh, the usual procedure is to simulate with many meshes (a minimum of 3), and compare the results of interest variables. Specially see if they converge to some value. You got it the right way by comparing the forces, but you can share with us the variation and we can discuss. If it is small enough (say for example less than 2%), I guess the higher computational cost is not worth it. I think it could be in this case, as the pressure field is what dominates this forces. But be aware that this is not general, as for example if the interest variable were the velocioties or formation of vortexes, I think it won’t be so close.
-
Also be aware that I raised the question of the refinement levels because the geometry was not being correctly modelled. In my opinion, this is not acceptable.
Greetings, hope I could shed some light on your doubts.
4 Likes
if i try to go to my dashboard I receive the following message:
This XML file does not appear to have any style information associated with it. The document tree is shown below.
NoSuchBucket
The specified bucket does not exist
https://www.simscale.com
B40F9CD44E9B7D82
FAgzYEaKRI7n31dm8JTqJcfvc98fSj+JYQvu8Zex1Gd0kOyiGnWapQxmAjN3MlZToZhEej+S2ho=
What can I do?
@aceci - it’s a technical problem from our end, we will get the platform up and running as soon as we can!
I’ve followed the steps to mesh the geometry as described and although the log states there are no errors it has failed the mesh quality check.
Does it make sense to run the simulation with the poor mesh? Seems a strange outcome for a tutorial
Hi,
this the my mesh log file:
Checking final mesh …
Checking faces in error :
non-orthogonality > 70 degrees : 0
faces with face pyramid volume < 1e-13 : 0
faces with concavity > 80 degrees : 0
faces with skewness > 4 (internal) or 20 (boundary) : 0
faces with interpolation weights (0…1) < 0.02 : 0
faces with volume ratio of neighbour cells < 0.01 : 0
faces with face twist < 0.01 : 0
faces on cells with determinant < 0.001 : 0
Finished meshing without any errors
Finished meshing in = 214.58 s.
End
Finalising parallel run
There are no errors regarding the mesh quality. If you are talking about this one:
Don’t worry. This is a different kind mesh quality test which is not mandatory for CFD simulations.
Cheers,
Milad
2 Likes
Yes that is the one i see, thanks
Thanks for the answers man 
, yes of course i know about the increasing of computational time and memory usage due to number of elements increase. About the values of forces i see that there is roughly a difference in a range of 2% to 6% in the results final value (considering only the pressure forces and moment) between the two simulations.
Hi
I got for the F1 aerodynamics Workshop Homework 2 a Mesh Error for the Rear Wing geometry (the one with loures). The Rear Wing _2 geometry (withouth loures) meshed without problems
These are the Error messages in my Mesh Operation Event Log:
The tesselated surface is not closed. There could be a problem with the CAD geometry (such as self-intersections). Please inspect your geometry. Trying to proceed anyway.
Illegal triangles were found after surface tesselation. There could be a problem with the CAD geometry. Trying to proceed anyway.
Mesh quality check failed. The mesh is not OK.
This is the project’s link
Please advice
Thanks for your help
Regards
Jorge
Interesting, the variations seem to be low (2-6%), but I think this has to be judged depending on the purpose of the simulation. If for example you are using this to validate design changes, and it shows an advantage of 5% in downforce (I think this is a lot), how do you know it doesn’t come from the mesh?
That sounds like exactly the same problem I had. See earlier response that it is ok to use the mesh as is for processing
Ok thanks for the help . i just wanted to confirm
For the simulation without the louvres, do we have to upload another CAD?
It’s in the project. its name is Rear_Wing_2
1 Like
Hi, in the following link you will find the results of my homework 2 with some images and the downloaded cases for 80 m/s, both meshes.
I have done the meshes using the final layer thickness equal to 0.9 like in the modified simulation of homework 1.
The fact is that I experience very small changes in the drag in the second case. What can cause such this behavior?
Thanks a lot for the help
Alessandro
1 Like
Is the problem in the CAD or in the mesh? If it’s in the mesh, don’t worry and go ahead, I have the same problem ( I saw your mesh log)
Hi @ggiraldo, I just wanted to say that I really appreciated your response to this post. It was definitely helpful to understand what was going on, and also a great reminder of vetting simulations and not trusting every pretty fringe plot that’s generated. Thanks.
2 Likes
Oh, this is a really kind answer from your part. I’m so glad I could help a little bit!
2 Likes