Poor converegence of plasiticity models in solver #651
Unanswered
superflyer11
asked this question in
Q&A
Replies: 1 comment 2 replies
-
|
The fact that explicitly derived von Mises does not converge is very strange. Have you tried with an elastic material (this can be achieved by setting the YieldStrength to a very high value) ? I suspect that the tangent operator is not correctly taken into account by |
Beta Was this translation helpful? Give feedback.
2 replies
-
|
Hi @thelfer , Linear elasticity converges immediately, i.e. All plasticity models can converge after I further refine my mesh but it is still just much slower than I expected. This may very well just be a FE issue but I am not sure yet, so putting here just in case. Tolerances: For all plasticity models, a homogenously strained tridimensional cube converges easily (but not quite quadratic) and number of iterations does not increase, example with explicit von-Mises Implicit von Mises: Regularised Drucker-Prager: But the more complex problem, where there are 2 models used (linear elasticity + one of the elastoplastic models), and plasticity increasingly appears, converges slowly, and number of iterations increase by a lot. Explicit von Mises: Implicit von Mises Implicit regularised Drucker-Prager is much worse Which I don't understand given the roughly constant iterations in mtest / the homogenous 3D problem, and why this would happen even with the explicit von Mises. I have no control over |
Beta Was this translation helpful? Give feedback.
-
The simple von Mises law shall not lead to any convergence issue. Would you try to make a simple shear test on a unit test ? The main issue related to the conversion of the consistent tangent operators are:
|
Beta Was this translation helpful? Give feedback.
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.
-
Hello, I am trying to investigate why my analysis is converging very slowly in the
MoFEMsolver, particularly with my von Mises and Drucker-Prager models.There are 4 models that I have been actively trying so far: 1) von Mises in the Default DSL, 2) von Mises in Implicit, 3) Drucker-Prager in Implicit, and 4) a regularised Drucker-Prager with hyperboloid form in Implicit
VMHardening_DefaultDSL.mfront.txt
VMHardening_implicit.mfront.txt
DP_negative.mfront.txt
DP_hyperboloidal_negative.mfront.txt
with mtest I have tested with these conditions:
vMDefault.mtest.txt
vM.mtest.txt
DP_simple.mtest.txt
DP_hyperboloidal.mtest.txt
with the output:
vMDefault_output.txt
vM_output.txt
DP_output.txt
DP_hyperboloidal_output.txt
The models working as stress is pulled back onto the yield surface. Example:
In the outputs the
orderin all the steps is less than 2, so does that tell that is not quadratic convergence?Nonetheless since solving for one point was cheap, I moved on to using the
MoFEMsolver. Solving for a simple problem was okay, however when I moved to a relatively complex problem convergence became a problem.These are tolerences I am using:
A sample analysis with von-Mises implicit:
With the analysis eventually failing
And even worse with Drucker-Prager, especially when approaching the tip (not at the tip)
Example of steps:
Same issue happens with the other two models as well more or less. Even with the explicitly derived von Mises I cannot get a good covergence. With linear elasticity the problem converges immediately so I do not think it is the problem of the mesh.
So from the perspective of mfront, at the moment I am thinking of
Any ideas?
Beta Was this translation helpful? Give feedback.
All reactions