Add NamedAdditionalOutput for strain weakened plastic parameters #1763

anne-glerum · 2017-05-23T12:00:42Z

A visualization output that shows the cohesion and friction angle as weakened by the current value of the tracked strain.

@naliboff any comments? I hope this way it doesn't interfere with your plans to separate the viscosity calculation out into several functions too much.

gassmoeller

Only a few comments. Have you checked that the tests give reasonable output? Also please add an entry to the changelog so that people are aware of the new output.

gassmoeller · 2017-05-23T15:01:02Z

source/material_model/interface.cc

+      {
+        std::vector<std::string> names;
+        names.push_back("weakened_cohesions");
+        names.push_back("weakened_friction_angles");


I am not quite happy with the term weakened, what if people implement a healing algorithm that might make the values higher than the initial values? Would you be ok with current_...?

I agree with @gassmoeller here. While the original intent was only to do strain weakening, it would be good to leave the door open for implementing strain hardening and not having to rename these terms. Another term that could be used is modified....

gassmoeller · 2017-05-23T15:02:37Z

source/material_model/interface.cc

+      :
+      NamedAdditionalMaterialOutputs<dim>(make_plastic_additional_outputs_names()),
+      cohesions(n_points, -1.0),
+      friction_angles(n_points, -1.0)


Please replace the -1 by numbers::signaling_nan<double>() in both lines. -1 was a leftover from early days in the seismic velocities, and we would like to get rid of it.

gassmoeller · 2017-05-23T15:07:27Z

source/material_model/interface.cc

@@ -747,6 +747,46 @@ namespace aspect
      return vs;
    }

+    namespace


Which material models will use this property? I am asking, because I would prefer this to be moved in some place other than the interface (it is only relevant for plastic material models). If there is only one (or several, but they are all derived from one base model), then please move all this code into the header and source file of that class (see e.g. the handling of MeltOutputs in melt.h and melt.cc).

gassmoeller · 2017-05-23T15:23:04Z

source/material_model/visco_plastic.cc

+      const double strain_fraction = ( cut_off_strain_ii - start_strain_weakening_intervals[j] ) /
+                                     ( start_strain_weakening_intervals[j] - end_strain_weakening_intervals[j] );
+      const double weakened_coh = cohesions[j] + ( cohesions[j] - cohesions[j] * cohesion_strain_weakening_factors[j] ) * strain_fraction;
+      const double weakened_phi = angles_internal_friction[j] + ( angles_internal_friction[j] - angles_internal_friction[j] * friction_strain_weakening_factors[j] ) * strain_fraction;


Same here, I would prefer current over weakened

gassmoeller · 2017-05-23T15:23:42Z

source/material_model/visco_plastic.cc

@@ -353,6 +365,7 @@ namespace aspect
              // of compositional field viscosities is consistent with any averaging scheme.
              out.viscosities[i] = average_value(composition, composition_viscosities, viscosity_averaging);

+


remove this empty line

gassmoeller · 2017-05-23T15:35:56Z

/run-tests

naliboff · 2017-05-23T16:24:49Z

@anne-glerum - nice, thank you for adding this in! This does not interfere with any plans I had for visco_plastic. Honestly, I think it easier to just make a new material model that is both significantly restructured to accommodate all the features we have discussed (viscoelasticity, new brittle and viscous deformation mechanisms, brittle vs total strain, etc).

gassmoeller · 2017-05-23T17:05:53Z

I agree. visco_plastic has reached a complexity that is ok for users, who have previously used diffusion_dislocation or drucker_prager, but additional complexity should be moved into a new material model (e.g. visco_elasto_plastic ?).

anne-glerum · 2017-05-24T12:01:35Z

I've moved the code in interface to the visco_plastic header and source files. There could be another plastic model for which this output is interesting however, namely the dynamic_friction model, @alarshi?

anne-glerum · 2017-05-24T12:06:56Z

Also, the output of the tests looks good: the cohesion and friction is weakened by the specified initial strain by the right amount in the right area. I'll run a 'real' model over time today to be sure.

alarshi · 2017-05-24T13:55:10Z

@anne-glerum: Cool! I will try this after merge. Thank you!

gassmoeller

Two tiny things I missed yesterday, otherwise good to go.

gassmoeller · 2017-05-24T16:45:15Z

source/material_model/interface.cc

@@ -747,6 +747,7 @@ namespace aspect
      return vs;
    }

+


Please remove this line

gassmoeller · 2017-05-24T16:47:11Z

source/material_model/visco_plastic.cc

-    evaluate(const MaterialModel::MaterialModelInputs<dim> &in,
-             MaterialModel::MaterialModelOutputs<dim> &out) const
+    calculate_weakening(const double &strain_ii,
+                        const unsigned int &j) const


hand over the doubles by-value (i.e. remove the &), for simple data types like double this is equally fast (or even faster), and the chance is lower that you accidentally change them inside of the function.

same for the unsigned int of course

anne-glerum · 2017-05-25T12:45:42Z

If the dynamic_friction material model and the new, say visco_elasto_plastic, material model would also use this, should the PlasticAdditionalOutput not stay in interface.cc?

gassmoeller · 2017-05-25T15:41:51Z

I do not feel very strongly about it. Usually we try to only put stuff in the interface class that is of importance for all (or a very significant fraction of all) material models. The fact that SeismicOutputs is in there is mostly a relic, because it was always there. If dynamic_friction and visco_elasto_plastic include the header of visco_plastic the current form should work as well, and it also represents a relationship between them. Even though they are not derived from each other, they do similar things, which is model visco_plastic behavior. I just try to avoid putting every new Output into the interface, which would grow uncontrollably over time, and it is very hard to get rid of anything in the interface (it took us years to get rid of the reference_... functions).

naliboff · 2017-05-25T17:17:34Z

I do not feel strongly on this point either, but one reason to keep these functions within visco_plastic is there a multitude of permutations on how to implement strain weakening. These permutations can be based on the type of tracked strain (brittle vs total, etc) the yield mechanism (Drucker Prager, velocity-weakening, ..) etc. In many cases each material model may have a type of strain weakening similar to the current implementation, but slightly different nonetheless.

anne-glerum · 2017-05-30T16:43:24Z

Alright, then I think it's done :) Just checked it on a rift model and calculating the current values in paraview gives the same results.

Create weakening function Move PlasticAdditionalOuput to interface Implement weakening function Also use new function in viscosity calculation Correct averaging and unweakened values Add new output to test Reformulate after rebase on master Add new output to weakening tests Add function description

gassmoeller · 2017-05-31T21:19:47Z

Looks good.

gassmoeller reviewed May 23, 2017

View reviewed changes

gassmoeller mentioned this pull request May 23, 2017

Be more conservative with regexes. #1762

Merged

gassmoeller reviewed May 24, 2017

View reviewed changes

anne-glerum force-pushed the add_strain_weakening_visu branch from e5d847e to 6a2fae1 Compare May 30, 2017 15:07

anne-glerum force-pushed the add_strain_weakening_visu branch from 6a2fae1 to a6f7556 Compare May 31, 2017 10:26

anne-glerum force-pushed the add_strain_weakening_visu branch from a6f7556 to 960880d Compare May 31, 2017 11:24

gassmoeller merged commit 97b6687 into geodynamics:master May 31, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add NamedAdditionalOutput for strain weakened plastic parameters #1763

Add NamedAdditionalOutput for strain weakened plastic parameters #1763

anne-glerum commented May 23, 2017

gassmoeller left a comment

gassmoeller May 23, 2017

naliboff May 23, 2017

gassmoeller May 23, 2017

gassmoeller May 23, 2017

gassmoeller May 23, 2017

gassmoeller May 23, 2017

gassmoeller commented May 23, 2017

naliboff commented May 23, 2017

gassmoeller commented May 23, 2017

anne-glerum commented May 24, 2017

anne-glerum commented May 24, 2017

alarshi commented May 24, 2017

gassmoeller left a comment

gassmoeller May 24, 2017

gassmoeller May 24, 2017

gassmoeller May 24, 2017

anne-glerum commented May 25, 2017

gassmoeller commented May 25, 2017

naliboff commented May 25, 2017

anne-glerum commented May 30, 2017 •

edited

gassmoeller commented May 31, 2017

		@@ -353,6 +365,7 @@ namespace aspect
		// of compositional field viscosities is consistent with any averaging scheme.
		out.viscosities[i] = average_value(composition, composition_viscosities, viscosity_averaging);

Add NamedAdditionalOutput for strain weakened plastic parameters #1763

Add NamedAdditionalOutput for strain weakened plastic parameters #1763

Conversation

anne-glerum commented May 23, 2017

gassmoeller left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gassmoeller commented May 23, 2017

naliboff commented May 23, 2017

gassmoeller commented May 23, 2017

anne-glerum commented May 24, 2017

anne-glerum commented May 24, 2017

alarshi commented May 24, 2017

gassmoeller left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

anne-glerum commented May 25, 2017

gassmoeller commented May 25, 2017

naliboff commented May 25, 2017

anne-glerum commented May 30, 2017 • edited

gassmoeller commented May 31, 2017

anne-glerum commented May 30, 2017 •

edited