Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Mechanical contact algorithms #10950

Closed
lindsayad opened this issue Mar 5, 2018 · 3 comments
Closed

Mechanical contact algorithms #10950

lindsayad opened this issue Mar 5, 2018 · 3 comments
Assignees
@lindsayad
Labels
C: Modules P: normal A defect affecting operation with a low possibility of significantly affects. T: task An enhancement to the software.

Comments

@lindsayad
Copy link
Member

Rationale

Mechanical contact often displays difficult convergence.

Description

Contact introduces non-smooth character into the residual function which poses a challenge to Newton's method. A smart line search may reduce some of the deleterious effects.

Impact

Improve solve behavior for problems involving contact.
lindsayad added a commit to lindsayad/moose that referenced this issue Mar 5, 2018
@lindsayad
Protect against not having petsc. Remove bad line searches. Refs idah…
57269ff 
…olab#10950
@lindsayad lindsayad mentioned this issue Mar 5, 2018
Merged
@permcody permcody added T: task An enhancement to the software. P: normal A defect affecting operation with a low possibility of significantly affects. C: Modules labels Mar 5, 2018
lindsayad added a commit to lindsayad/moose that referenced this issue Mar 20, 2018
@lindsayad
Protect against not having petsc. Remove bad line searches. Refs idah…
231aa8c 
…olab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Apr 5, 2018
@lindsayad
Protect against not having petsc. Remove bad line searches. Refs idah…
d762d9e 
…olab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Apr 6, 2018
@lindsayad
Implement custom contact line search
5fcb5ee 
Closes idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Apr 11, 2018
@lindsayad
Implement custom contact line search
2b96969 
Closes idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Apr 17, 2018
@lindsayad
Implement custom contact line search
f1ecf7c 
Closes idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue May 3, 2018
@lindsayad
Implement custom contact line search
eec114e 
Closes idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue May 4, 2018
@lindsayad
Implement custom contact line search
ed674eb 
Closes idaholab#10950
@lindsayad lindsayad changed the title Make contact solves more robust Mechanical contact algorithms Jan 6, 2020
@lindsayad
Copy link
Member Author

I'm going to reopen this issue and make it a catch-all for mechanical contact algorithm development

@lindsayad lindsayad reopened this Jan 6, 2020
@lindsayad lindsayad self-assigned this Jan 6, 2020
lindsayad added a commit to lindsayad/moose that referenced this issue Jan 6, 2020
@lindsayad
Add ping-ponging tests
9869353 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Jan 6, 2020
@lindsayad
Add ping-ponging tests
76fe09f 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
@lindsayad
Copy link
Member Author

Copying over from #14546 (comment)

Sadly, the node-face RANFS algorithm is susceptible to ping-ponging, just as kinematic is, as shown in 76fe09f. Mortar appears to be more robust since it does not show ping-ponging at any time step for that particular test input while RANFS and kinematic node-face do.

So despite my optimism in #14241 (comment), node-face RANFS is not the magic bullet. Your ping-ponging test in #14241 also shows that while mortar may-be more robust than node-face with respect to ping-ponging, it is still not immune. In fact my favorite mortar contact paper shows an example of ping-ponging mortar. They appear to have been able to fix that case by using a Petrov-Galerkin approach. So that's my newest golden idea 😄

lindsayad added a commit to lindsayad/moose that referenced this issue Jan 6, 2020
@lindsayad
Add ping-ponging tests
a065e02 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Jan 7, 2020
@lindsayad
Add ping-ponging tests
d46c014 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Jan 24, 2020
@lindsayad
Add ping-ponging tests
99bc377 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Jan 25, 2020
@lindsayad
Add ping-ponging tests
a9aab36 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 6, 2020
@lindsayad
Add ping-ponging tests
1b30ada 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 6, 2020
@lindsayad
Add ping-ponging tests
4463e66 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 6, 2020
@lindsayad
Add ping-ponging tests
0f1872b 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 10, 2020
@lindsayad
Add ping-ponging tests
17b5d88 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
aeslaughter pushed a commit to aeslaughter/moose that referenced this issue Feb 13, 2020
@lindsayad @aeslaughter
Add ping-ponging tests
a2dc764 
This batch of tests illustrates the susceptiblity of pure
node-face mechanical contact algorithms to ping-ponging. In this
case the ping-ponging comes from oscillation of a slave node
between two different master faces. A hybrid scheme in which
LM residuals are computed at slave nodes while primal residuals
are computed on mortar integrals does not display oscillations.
This does not necessarily mean that the hybrid scheme is immune
to oscillations, but it at least performs better on this test
case.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 17, 2020
@lindsayad
Add (non)singular frictional mortar test
cfce1c7 
Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 17, 2020
@lindsayad
Make std::abs derivative non-zero always
4474d9f 
This prevents us from having singular Jacobians

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 24, 2020
@lindsayad
Add (non)singular frictional mortar test
f7f4116 
Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 24, 2020
@lindsayad
Make std::abs derivative non-zero always
111ed1a 
This prevents us from having singular Jacobians

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 25, 2020
@lindsayad
Add tangential tolerance option to NormalNodalLMMechanicalContact
2ed8974 
Can use this option to ensure slave node projection onto master
surfaces.

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 25, 2020
@lindsayad
Add tangential_tolerance mortar test
b0cf90c 
Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Feb 25, 2020
@lindsayad
Make std::abs derivative non-zero always
fc179e4 
This prevents us from having singular Jacobians

Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Mar 14, 2020
@lindsayad
Add tangential_tolerance mortar test
f3188b9 
Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Mar 25, 2020
@lindsayad
Increase floors for module tests diffing with mac/conda
6232b2d 
Refs idaholab#10950
lindsayad added a commit to lindsayad/moose that referenced this issue Mar 25, 2020
@lindsayad
Increase floors for module tests diffing with mac/conda
146da33 
Refs idaholab#10950
WilkAndy pushed a commit to WilkAndy/moose that referenced this issue Apr 19, 2020
@lindsayad @WilkAndy
Add tangential_tolerance mortar test
27fefbd 
Refs idaholab#10950
WilkAndy pushed a commit to WilkAndy/moose that referenced this issue Apr 19, 2020
@lindsayad @WilkAndy
Increase floors for module tests diffing with mac/conda
dcc9a7c 
Refs idaholab#10950
@lindsayad
Copy link
Member Author

This is now covered by more specific issues for remaining work, and we have made great progress with mortar

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@lindsayad lindsayad

Labels
C: Modules P: normal A defect affecting operation with a low possibility of significantly affects. T: task An enhancement to the software.
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@permcody @lindsayad