Issue385 explicit by einola · Pull Request #423 · nansencenter/nextsim

einola · 2020-04-14T14:54:53Z

A huge merge adding the explicit solver and different ways of solving the internal stress that I've been working on.

Can you please check if the "old" setup still works the way we expect? I haven't changed the defaults.

…extsim into issue385_explicit Conflicts: model/finiteelement.cpp

…ing send and recv instead of isend and irecv is less efficient.

… solver, since the difference is just in the sigma term.

… damage.

… called pseudo_recursive (because that's what it is). Also: In the recursive scheme cohesion scales like ice thickness squared.

…odel writes a set of output files called "flip" and exits.

…sion_factor for the P* of MEB-EP. It appears the appropriate value is something like 150 kPa - at least at 10 km. This needs a lot more testing and polishing.

…nto issue385_explicit

… different branch (issue issue413_thickening).

…etter.

tdcwilliams

had a first pass, recommended some changes. Will need another look though. What did you mean by testing the "old setup"?

tdcwilliams · 2020-04-14T20:02:37Z

+        // We need to update the mesh every time step
+        M_surface[cpt] = this->measure(M_elements[cpt],M_mesh,M_UM);
+        std::vector<double> const shapecoeff = this->shapeCoeff(M_elements[cpt]);
+        // TODO: Put the B0T code in a seperate function


might be a good idea. B0T is calculated in FETensors also (in a nicer way).

tdcwilliams · 2020-04-15T10:23:02Z

Hi, I guess 'old' setup is no pressure? Also what is the time step for the explicit code for a 10km mesh?

einola · 2020-04-15T11:06:43Z

I just mean to check that the implicit solver still works mostly the same in whatever setup you were using before, whatever the time step, pressure values, etc.

…

Sent from my iPhone On 15 Apr 2020, at 12:23, Timothy Williams <notifications@github.com> wrote: Hi, I guess 'old' setup is no pressure? Also what is the time step for the explicit code for a 10km mesh? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#423 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AGABRFEJQMF7KTGW2LXBRV3RMWDJJANCNFSM4MHZYZ2Q>.

Co-Authored-By: Timothy Williams <tdcwilliams@gmail.com>

tdcwilliams · 2020-04-27T08:22:58Z

@einola, I am getting some strange results using the implicit solver with this .cfg file:

see
fram.sigma2.no:/cluster/work/users/timill/nextsim-stand-alone/medium_arctic_10km/fcpaper/ec2/expt_109-no-pressure-explicit/outputs/20181101/Moorings.nc

[thermo]
diffusivity_sss=0
diffusivity_sst=0
ocean_nudge_timeS=1296000
ocean_nudge_timeT=1296000
[mesh]
filename=medium_arctic_10km.msh
partitioner-fileformat=ascii
[solver]
mat-package-type=mumps
[setup]
ice-type=cs2_smos
ocean-type=topaz_forecast
atmosphere-type=ec2
bathymetry-type=etopo
[simul]
spinup_duration=1
timestep=20
time_init=2018-11-01
duration=21 #finish 20190901
[restart]
#write_restart=true
#output_time_step=10
write_final_restart=true
write_interval_restart=true
output_interval=10
start_from_restart=true
input_path=outputs/20181101/restart
basename=final
type=extend
[output]
datetime_in_filename=true
exporter_path=outputs/20181101
[dynamics]
time_relaxation_damage=15
pressure_nu=0 #nu=0
#p_coef_type=1 #D_unit
#C_lab=6161850#.9*6.8465e+6
#C_lab=5477200#.8*6.8465e+6
#C_lab=4792550#.7*6.8465e+6
#C_lab=3423250#.5*6.8465e+6
compression_factor=6000
exponent_compression_factor=3
[damage]
clip=0.95
#[drifters]
#use_osisaf_drifters=true
#[forecast]
#true_forecast=false
#time_init_ocean_fc=2018-11-01
#time_init_atm_fc=2018-11-01
[moorings]
use_moorings=true
spacing=5
output_timestep=1
output_time_step_units=days
variables=conc
variables=thick
variables=snow
variables=damage
#variables=ridge_ratio
variables=velocity
[debugging]
log-level=debug

…enough results.

…nto issue385_explicit

…ue if reset==true (as it should be).

…previous placement may have been problematic)

einola · 2020-04-28T07:00:39Z

I've addressed all of Tim's comments now ... and made some other modifications. A first look at my test using the semi-implicit solver indicates that it still works ok. @tdcwilliams can you re-run your test?

tdcwilliams

hi @einola, a few more comments this time.
Looks good code though (with the exception of calling the shape coefficients dx...).
I've launched my test of the implicit MEB so we'll see how that goes.

Co-authored-by: Timothy Williams <tdcwilliams@gmail.com>

…i (it's still the default)

…ate_sigma option instead of the reset option

tdcwilliams

run looks fine now - you can merge once you've applied the suggestions you want

…ebi and meb_explicit instead of mebe)

einola added 30 commits November 25, 2019 11:08

updateGhosts and friends from Abdoulaye r546beb4

147b2dc

updateGhosts and friends from Abdoulaye r546beb4

1808374

Merge branch 'issue385_explicit' of https://github.com/nansencenter/n…

00225ec

…extsim into issue385_explicit Conflicts: model/finiteelement.cpp

Merge remote-tracking branch 'origin/develop' into issue385_explicit

098ca1b

Debugging things for updateGhosts.

c0fb7e9

Fix to updateGhosts (also removed some LOG calls in there).

1bee445

More fixes to updateGhosts and initUpdateGhosts. It works now, but us…

b162364

…ing send and recv instead of isend and irecv is less efficient.

An EVP implementation. It will act as the foundation for the explicit…

efe6fe7

… solver, since the difference is just in the sigma term.

Grounding scheme and a few cosmetic changes.

dff30ef

An explicit solver for the MEB that works. TODO: Recalculate sigma on…

e0388bf

… damage.

Bugfix!

d565415

We now adjust M_sigma as we damage.

6cbc5a9

A new recursive damaging scheme. Véro's old "recursive" scheme is now…

e7ab6e6

… called pseudo_recursive (because that's what it is). Also: In the recursive scheme cohesion scales like ice thickness squared.

Turn off the sub-stepping when there are flipped triangles. Now the m…

7a7f88e

…odel writes a set of output files called "flip" and exits.

Two minor bug-fixes.

0a08bd9

Minor clean up.

94be2c6

bugfix

4028168

Initial implementation of MEB-EP. NB I'm still using dynamics.compres…

e94c089

…sion_factor for the P* of MEB-EP. It appears the appropriate value is something like 150 kPa - at least at 10 km. This needs a lot more testing and polishing.

Bugfix

6007967

An improved MEB_EP version.

902fb21

A big fat commit with all the different things I've been testing lately.

705e6f0

Undo changes not related to the work in this branch.

769853d

Simplifications of BMEB.

afbf30d

Bug fix

7bdf54e

For BMEB D_dcirt is the distance to the plastic criterion.

e125d98

Merge branch 'issue385_explicit' of github.com:nansencenter/nextsim i…

5a42ec7

…nto issue385_explicit

Merge remote-tracking branch 'origin/develop' into issue385_explicit

9f35082

Clean up and simplification. I've moved the BMEB and MEBEP tests to a…

6bfb4b0

… different branch (issue issue413_thickening).

Improved calculation of dt (on element basis this time).

5f9b5bb

Move the calculation of coef_P to one place

0ab28cb

Just move the updateSigma function so that the graphical diff looks b…

fecd407

…etter.

einola requested review from akorosov and tdcwilliams April 14, 2020 14:54

tdcwilliams requested changes Apr 14, 2020

View reviewed changes

einola and others added 2 commits April 27, 2020 10:19

Update model/finiteelement.cpp

2b8f67a

Co-Authored-By: Timothy Williams <tdcwilliams@gmail.com>

Update model/finiteelement.cpp

b025802

Co-Authored-By: Timothy Williams <tdcwilliams@gmail.com>

einola added 10 commits April 27, 2020 13:10

One-liner from Tim.

119ce01

Remove the Direct and Recursive schemes as they don't give realistic …

5068b3d

…enough results.

Fix to a comment

1f856c7

Merge branch 'issue385_explicit' of github.com:nansencenter/nextsim i…

56bbf3c

…nto issue385_explicit

Better help for some options

df4e0e6

Bugfix: updateSigma only recalculates M_sigma with the new damage val…

ffab557

…ue if reset==true (as it should be).

Small bug-fix

d27c2a4

Only call initUpdateGhosts when it's actually needed

b042094

Move the updateSigmaCoefs call in assamble to fit original code (the …

dc64ff0

…previous placement may have been problematic)

Remove left over crud.

64ba012

tdcwilliams requested changes May 6, 2020

View reviewed changes

einola and others added 5 commits May 8, 2020 14:45

Update model/finiteelement.cpp

17eebde

Co-authored-by: Timothy Williams <tdcwilliams@gmail.com>

Change the name of the semi-implict dynamics-type from default to meb…

66864a0

…i (it's still the default)

Extra documentation

680a657

The function updateSigma is now called updateDamage and it has an upd…

ad91efa

…ate_sigma option instead of the reset option

dx is now dxN

3d99aa7

tdcwilliams approved these changes May 9, 2020

View reviewed changes

More explicit names for dynamics-type (meb_semi_implicit instead of m…

f6e1b0a

…ebi and meb_explicit instead of mebe)

einola merged commit 5676d32 into develop May 11, 2020

einola deleted the issue385_explicit branch May 11, 2020 04:54

Conversation

einola commented Apr 14, 2020

Uh oh!

tdcwilliams left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tdcwilliams Apr 14, 2020

Choose a reason for hiding this comment

Uh oh!

Uh oh!

tdcwilliams commented Apr 15, 2020

Uh oh!

einola commented Apr 15, 2020 via email

Uh oh!

tdcwilliams commented Apr 27, 2020

Uh oh!

einola commented Apr 28, 2020

Uh oh!

tdcwilliams left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tdcwilliams left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants