New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Assertion `i*j < _val.size()' failed when coupling displaced and undisplaced variables with nonzero BC #9659
Comments
I'm looking at this right now. Notes of what I found so far:
Looks like an off diagonal jacobian is computed "unexpectedly". I.e. during setup MOOSE does not know the variables are coupled, but it computes the odj anyways... |
This is not a bug in the TensorMechanics action. If I unfold the action using the extremely useful Input File
Which fails with the same error |
The issue seems coupling a variable that has kernels only on the displaced mesh to a variable that has kernels only on the undisplaced mesh. Minimal example without tensor mechanics:
|
Yeah, bummer, this issue just got forgotten. Let me 🏓 @andrsd, who should know all about the allocation of the Jacobian blocks. Maybe he has some fresh ideas. |
So, you guys have a PDE where volumetric terms are on displaced mesh and its BCs are on undisplaced? Or is this just a simplification of something more complex? BTW: just by looking at the input file, this should work. I suspect where the problem might be... |
@andrsd, I don't exactly remember, but I thought this required coupling the u and v (where u is in the displaced and v is on the undisplaced mesh), and then adding a non zero BC (no idea if the BC is on the displaced or undisplaced mesh though). |
The minimal example that breaks has variable I suspect the real problem is that a kernel is on a displaced mesh - those would be q-point values - plus there is a nodal BC - those would be nodal values - on a non-displaced mesh. And we do not flip the right flag to compute one of those... It is just my working theory for now... |
D'oh, you're right, there is no coupling. I guess you just need variables on both meshes. |
Cool, looks like you guys are tracking it down? I just don't want this to get lost again :) |
So, here is the problem (mostly for me so that I do not forget):
|
Looks like this is close, I'll be glad to see this resolved after all this time! |
I guess my last comment was confusing for some...this isn't closed, but sounds close since @andrsd has an idea on what the problem is |
Looks like this has be solved by something else |
UGGG!!!! This problem still exists. Here is another condensed input file that reproduces the error. As far as I can tell, no line can be removed without clearing the error. This is causing problems with some large BISON runs I'm trying to run through debug and is hindering development. Any ideas @dschwen @andrsd ? Output
|
This latest input has to be run with combined-dbg |
Any new thoughts on this issue ? I've been trying to run some test cases with the peridynamics module and running into the same error in devel mode |
Argh, never mind, the one in your comment from May 31st still triggers the bug. |
Ok, some preliminary observation:
This can be traced further up to The problem also seems to vanish when only using first order elements (just setting order 1 on |
I'm on travel this week |
Whew...I think this was fixed along the way... |
Really? It's not referenced. I hope we have a test case, or if not adding one now would be great! |
There are three different inputs that should prove it in this issue |
Nope! Still a problem!!!!! I closed this prematurely somehow. I keep coming around to this when I have to use dbg on thermo-mechanical problems. Once I get pulled away to something else and I stop pushing this, I forget, @dschwen forgets, @andrsd forgets, everyone else forgets, and it gets consistently lost. This was so close at some point with @andrsd identifying the problem, but this never being solved. I'll open a PR with a test that clearly fails so this hopefully gets more visibility, and this damn thing gets closed finally! |
Gonna rename this to "Topher's bug" - Seems like it's the bane of your existence. |
Glad someone could sense my frustration... :) |
Description of the enhancement or error report
When running a tensor mechanics problem with debug, the following assertion fails:
Output
It seems to happen when the DenseMatrix is formulated for the off-diagonal jacobian, and only when there is a non-zero BC condition preset. It also requires an additional variable beyond just the displacement variables, and for
strain = FINITE
Rationale for the enhancement or information for reproducing the error
The following input file can be used to reproduce the error with
tensor-mechanics-dbg
.Input File
Identified impact
This will allow more complicated input files to be debugged.
The text was updated successfully, but these errors were encountered: