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
Change material model logic for determining initial time step number #2529
Conversation
Can we create a function in SimulatorAccess instead? Maybe something like is_initial_time_step()? |
Exactly what I had in mind. Unfortunately, this now fails several tests -- but I can't see why because @tjhei 's server doesn't want us to see the output diffs :-( |
There is also a merge conflict you'll have to address. |
@tjhei - I think having that something like is is_initial_time_step() is a good idea. Would this apply to only after the model setup or during the first set of set of solves (e.g., time step "0" now)? Until then, any objections to the logic introduced here? @bangerth - I think I know what the issue is with the tests. I'll rebase and see if that fixes the problem. |
3154f31
to
3671904
Compare
Ok, got it down to two tests failing. @tjhei - When you have a chance, can you take a look at the tester permissions? The tests diffs are currently not accessible. |
that means that the upload to the webserver hasn't completed yet. It is available now but it might take up to an hour to complete. |
On 06/28/2018 12:29 PM, Timo Heister wrote:
Can we create a function in SimulatorAccess instead? Maybe something like
is_initial_time_step()?
It's not the initial time step, but instead *before* the initial time step. I
could definitely store a flag for this, as well as an accessor function. Would
you like that?
|
I do not think we need to store a separate flag, we could just move the logic that is already present in many plugins into a separate function |
agreed. |
@naliboff and @anne-glerum: We discussed this before :-) I just looked and there is no such |
3671904
to
c504220
Compare
@MFraters - The visco_plastic Newton derivative tests are failing with the proposed changes. Can we take a look at this together at some point? |
hmm, this is interesting. So I think the reason it is failing has to do with how the test is set up. For the test I needed to trick the material model in using not the reference strain-rate, but the strain-rate which I provided in the test. So we need a way to set |
oofta ;) Is the strain-rate in the test constant throughout the model domain? |
no, I have a different strain-rate tensor for each 'quadrature' point. |
hmm, I will have to look into this a bit more tomorrow. Adding a function to set the variable |
Can you explain why the test now fails? Are you saying that the branch you end up in in the computation of |
yes, it should compute second invariant of the strain-rate tensor instead of using the reference strain-rate. |
c504220
to
211bd76
Compare
211bd76
to
049cdee
Compare
@MFraters - Any chance we can come back and take a look at how to fix these tests? As I understand it, the issue is that current logic in Could we somehow trick |
Closing this as there does not seem to be an easy path for adjusting the Newton test setup. |
ah, sorry. I totally missed your last reply. If you still want to discuss this at some point or want me to have a look at it again let me know. |
Related to #2362.
@bangerth - Let me know if this is what you had in mind. A similar logic was already in
drucker_prager
.diffusion_dislocation
did not have a reference to the tilmestep at all in setting the strain rate during time step 0 or initiation stage. Rather, it just checksin.strain_rate.size()
before calling a function to compute the viscosity.