"MLMG failed" when I use EB #243
Unanswered
ZhaofanZhu
asked this question in
Q&A
Replies: 2 comments 4 replies
-
|
Can you also show the velocity for the case that does work? This seems to be the linear solver issue that @drummerdoc and I hinted might happen in reply to your last question. Based on where it's failing it's probably the MAC projection but might be a diffusion solve. First step is to crank up the verbosity of the output for those solves with |
Beta Was this translation helpful? Give feedback.
4 replies
-
|
You are foresight, Mr.Bruce. Update: I discovered that I misunderstood the unit of the simulation domain size. The actual length of this simulation domain is 2 m (not 2 cm). This may be the reason for the abnormal flow velocity. |
Beta Was this translation helpful? Give feedback.
-
|
The output of Case B is: Click to expand
<\details> |
Beta Was this translation helpful? Give feedback.
-
|
Hi @ZhaofanZhu it seems that AMReX is barely able to handle the geometry on the coarse level. You barely make it past the EB generation in one case (case B) while it fails for case A. You can see that in case B, the increased MLMG verbose writes: which means that a single MG level is available on the coarse level, thus leading to a very inefficient multigrid. Can you try doubling the resolution of the level 0 and lowering the max_level by one ? That should allow a better representation of the geometry on the coarse level and a deeper multigrid hierarchy. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you @esclapez . I changed the resolution of level 0 and max_level, but there are still errors. That's weird. Do you have any other suggestions? |
Beta Was this translation helpful? Give feedback.
-
|
A few more questions:
One thought is that that initial velocity profile you are specifying (uniform x-velocity?) is likely quite different from the projected velocity field that accounts for the rough surface. It may work better to let the simulation gradually evolve rather than making this jump all at once. You could try initializing zero velocity in a thin region along the wall that covers the sinusoidal roughness, and letting the simulation develop from there. As an additional note on this, in the future we're planning to add an option to use alternative linear solvers that are more robust and may help with EB geometries with lots of fine features. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi, I used EB to insert a sine-shaped wall at y=0. However, when I adjust the amplitude and wavelength of the sine function or other parameters related to the shape, an error message of "MLMG failed" may appear.
For example, as shown in the figure below, Case A and Case B have a sine-shaped wall, and the amplitude of the sine function decreases to 0 as it approaches the flow inlet (x=0) by multiplying by a tangent function (std::tanh). The only difference between Case A and Case B is the position where the amplitude starts to decrease. Case A runs as expected, but Case B reports the error:
What bothers me is that this problem seems to appear randomly. I think there should be no difference between Case A and Case B.
Does anyone know the reason for this error and how to avoid it? Thanks in advance!
Beta Was this translation helpful? Give feedback.
All reactions