homnath
commented
1 year ago Hi Haiyang Liao,
How does it look like when you visualize all processor files, proc000000_vs.vtk, proc000001_vs.vtk, proc000002_vs.vtk, etc., altogether? The individual mesh may look irregular due to partitioning but altogether, they should match your original model.
Best, Hom Nath
From: liaohaiyang1534 @.> Sent: Saturday, May 27, 2023 11:15 PM To: SPECFEM/specfem3d @.> Cc: Subscribed @.***> Subject: [SPECFEM/specfem3d] models may not match (Issue #1609)
Hi,
i use a python scipt to make a model (given at the end), and it looks like this in the software CoreForm Cubit 2022.11:
[)EP3~DHA8G8%TPP{AN03@{6]https://user-images.githubusercontent.com/104192749/241474009-81f180e1-06e3-4825-8374-0acb5c092df9.png
when I set
save AVS or OpenDX mesh files to check the mesh
SAVE_MESH_FILES = .true.
and check part_array.vtk from OUTPUT_FILES/DATABASES_MPI, it looks like this:
and I think this is alright, it matches the size of my model.
but when I check proc000010_vs.vtk:
It looks a little strange.
Should I ignore it? Or is there some potential bug?
the python script that I use:
` from future import print_function
import os import sys
elementsize = 4
default directories
SEMoutput='MESH' CUBIToutput='MESH_GEOCUBIT'
os.system('mkdir -p '+ SEMoutput) os.system('mkdir -p '+ CUBIToutput)
import cubit try:
cubit.init([""])
cubit.init(["-noecho","-nojournal"])except: pass
version = cubit.get_version() version_major = int(version.split(".")[0]) version_minor = int(version.split(".")[1]) print("cubit version: ",version)
cubit.cmd('reset') cubit.cmd('brick x 300 y 150 z 100')
This seems to conflict with boundary_definition.py
cubit.cmd('volume 1 move x 150 y 75 z -50')
create vertices for discontinuity
cubit.cmd('split curve 9 distance 20') cubit.cmd('split curve 10 distance 20') cubit.cmd('split curve 11 distance 20') cubit.cmd('split curve 12 distance 20')
create surface for interface
cubit.cmd('create surface vertex 9 10 12 11')
cubit.cmd('section volume 1 with surface 7 keep normal') cubit.cmd('section volume 1 with surface 7 reverse')
create vertices for auxiliary interface to allow for refinement
cubit.cmd('split curve 29 distance 30') cubit.cmd('split curve 31 distance 30') cubit.cmd('split curve 32 distance 30') cubit.cmd('split curve 36 distance 30')
create surface for buffer interface to refine BELOW the discontinuity
cubit.cmd('create surface vertex 25 26 28 27')
cubit.cmd('section volume 3 with surface 19 keep normal') cubit.cmd('section volume 3 with surface 19 reverse')
cubit.cmd('delete volume 2 4')
cubit.cmd('merge all') cubit.cmd('imprint all')
Meshing the volumes
cubit.cmd('volume 3 size '+str(elementsize)) cubit.cmd('mesh volume 3')
cubit.cmd('volume 1 size '+str(elementsize)) cubit.cmd('mesh volume 1')
cubit.cmd('volume 5 size '+str(elementsize)) cubit.cmd('mesh volume 5')
End of meshing
obsolete:
import boundary_definition
import cubit2specfem3d
new:
from geocubitlib import boundary_definition from geocubitlib import cubit2specfem3d
This is boundary_definition.py of GEOCUBIT
..... which extracts the bounding faces and defines them into blocks
boundary_definition.entities=['face'] boundary_definition.define_bc(boundary_definition.entities,parallel=True)
Define material properties for the 3 volumes
cubit.cmd('#### DEFINE MATERIAL PROPERTIES #######################') cubit.cmd('block 1 name "elastic 1" ') # elastic material region cubit.cmd('block 1 attribute count 7') cubit.cmd('block 1 attribute index 1 1 ') # material 1 cubit.cmd('block 1 attribute index 2 1200 ') # vp cubit.cmd('block 1 attribute index 3 400 ') # vs cubit.cmd('block 1 attribute index 4 2000 ') # rho cubit.cmd('block 2 attribute index 5 9999.0') # Q_kappa cubit.cmd('block 2 attribute index 6 9999.0') # Q_mu cubit.cmd('block 2 attribute index 7 0 ') # anisotropy_flag
cubit.cmd('block 2 name "elastic 2" ') # elastic material region cubit.cmd('block 2 attribute count 7') cubit.cmd('block 2 attribute index 1 2 ') # material 2 cubit.cmd('block 2 attribute index 2 2000 ') # vp cubit.cmd('block 2 attribute index 3 700 ') # vs cubit.cmd('block 2 attribute index 4 3200 ') # rho cubit.cmd('block 2 attribute index 5 9999.0') # Q_kappa cubit.cmd('block 2 attribute index 6 9999.0') # Q_mu cubit.cmd('block 2 attribute index 7 0 ') # anisotropy_flag
cubit.cmd('block 3 name "elastic 3" ') # elastic material region cubit.cmd('block 3 attribute count 7') cubit.cmd('block 3 attribute index 1 3 ') cubit.cmd('block 3 attribute index 2 2500 ') # vp cubit.cmd('block 3 attribute index 3 800 ') # vs cubit.cmd('block 3 attribute index 4 3600 ') # rho cubit.cmd('block 3 attribute index 5 9999.0') # Q_kappa cubit.cmd('block 3 attribute index 6 9999.0') # Q_mu cubit.cmd('block 3 attribute index 7 0 ') # anisotropy_flag
cubit.cmd('export mesh "' + CUBIToutput + '/top.e" dimension 3 overwrite') cubit.cmd('save as "' + CUBIToutput + '/top.cub" overwrite')
Export to SPECFEM3D format using cubit2specfem3d.py of GEOCUBIT
cubit2specfem3d.export2SPECFEM3D(SEMoutput)
screen shot
(could crash version < 16.4)
if version_major >= 16 and version_minor >= 4: cubit.cmd('view top') cubit.cmd('hardcopy "' + CUBIToutput + '/waterlayered.png" png')
all files needed by SCOTCH are now in directory MESH`
I'm very appreciated to have any advice and suggestions for this. (If you need any more clarifications, please let me know)
Regards
Haiyang Liao @.**@.>
— Reply to this email directly, view it on GitHubhttps://github.com/SPECFEM/specfem3d/issues/1609, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABMCQ4XI3Q3GTV3BTOBUE5DXIK7MZANCNFSM6AAAAAAYRR4LDE. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hi,
i use a python scipt to make a model (given at the end), and it looks like this in the software
CoreForm Cubit 2022.11:when I set
and check
part_array.vtkfromOUTPUT_FILES/DATABASES_MPI, it looks like this:and I think this is alright, it matches the size of my model.
but when I check
proc000010_vs.vtk:It looks a little strange.
Should I ignore it? Or is there some potential bug?
the python script that I use:
I'm very appreciated to have any advice and suggestions for this. (If you need any more clarifications, please let me know)
Regards
Haiyang Liao haiyangliao@smail.nju.edu.cn