Raster Mesh to VTK

[hem-28]

Hello,

I am familiar with the raster mesh .txt format written out by MicroStructPy. Pardon my ignorance for that. I wanted to convert this Voxelated data to a VTK file which is similar to the output from Dream3D, so I wanted some clarification.

1. Does 'Element Attributes' correspond to the grain number/seed number?
2. In which order is this list written? In this order probably?
   for (i=0; i<nx; i++)
   for (j=0; j<ny; j++)
   for (k=0; k<nz; k++)
3. The phase/material information is printed in 'rastermesh.png', but this information is not available in rastermesh.txt right? Do I have to correlate the grain id to phase id using the seeds.txt file or is there an alternate direct solution?
   Thanks!

[hem-28]

Okay so I figured out that it is possible to export the raster mesh as vtk (including phase IDs) using the python version. The data is however written out as an unstructured grid format as points data, which is a bit surprising for a regular grid rastermesh. Would there be any possibility to export vtk as Structured grid and cell data ?

[kip-hart]

Hi @hem-28 , MicrostructPy was initially written to generate unstructured meshes so there may have been some re-use of the export functions when I added the RasterMesh class. The VTK format section of RasterMesh.write is:

MicroStructPy/src/microstructpy/meshing/trimesh.py

Lines 1165 to 1216 in 44b6a65

	elif fmt == 'vtk':
	n_kp = len(self.elements[0])
	mesh_type = {4: 'Pixel', 8: 'Voxel'}[n_kp]
	pt_fmt = '{: f} {: f} {: f}\n'
	# write heading
	vtk = '# vtk DataFile Version 2.0\n'
	vtk += '{} mesh\n'.format(mesh_type)
	vtk += 'ASCII\n'
	vtk += 'DATASET UNSTRUCTURED_GRID\n'
	# Write points
	vtk += 'POINTS ' + str(len(self.points)) + ' float\n'
	if len(self.points[0]) == 2:
	vtk += ''.join([pt_fmt.format(x, y, 0) for x, y in
	self.points])
	else:
	vtk += ''.join([pt_fmt.format(x, y, z) for x, y, z in
	self.points])
	# write elements
	n_elem = len(self.elements)
	cell_fmt = str(n_kp) + n_kp * ' {}' + '\n'
	cell_sz = (1 + n_kp) * n_elem
	vtk += '\nCELLS ' + str(n_elem) + ' ' + str(cell_sz) + '\n'
	vtk += ''.join([cell_fmt.format(*el) for el in self.elements])
	# write cell type
	vtk += '\nCELL_TYPES ' + str(n_elem) + '\n'
	cell_type = {4: '9', 8: '12'}[n_kp]
	vtk += ''.join(n_elem * [cell_type + '\n'])
	# write element attributes
	try:
	int(self.element_attributes[0])
	att_type = 'int'
	except TypeError:
	att_type = 'float'
	vtk += '\nCELL_DATA ' + str(n_elem) + '\n'
	vtk += 'SCALARS element_attributes ' + att_type + ' 1 \n'
	vtk += 'LOOKUP_TABLE element_attributes\n'
	vtk += ''.join([str(a) + '\n' for a in self.element_attributes])
	# Write phase numbers
	if seeds is not None:
	vtk += '\nSCALARS phase_numbers int 1 \n'
	vtk += 'LOOKUP_TABLE phase_numbers\n'
	vtk += ''.join([str(seeds[a].phase) + '\n' for a in
	self.element_attributes])
	with open(filename, 'w') as file:
	file.write(vtk)

I'll dig into the VTK file format docs and see if I can use the structured grid type.

[kip-hart]

Now that I'm reading the docs from VTK (http://www.princeton.edu/~efeibush/viscourse/vtk.pdf) it looks like the RECTILINEAR_GRID option would be better suited than STRUCTURED_GRID.

[hem-28]

Yes True , Rectilinear_grid would be a better option.

[kip-hart]

Hi @hem-28 I have just opened PR #75 to make this modification. I'll let the tests run over night and update the docs tomorrow. New release with the update should be available in the next couple of days.

[hem-28]

@kip-hart Thanks for taking it as a feature request!

[kip-hart]

Hey @hem-28, I included this feature as of version 1.5.5, released a few minutes ago. Try pip install --upgrade microstructpy and see if that works for you. Let me know how it goes, I'm here to help.

[hem-28]

Hi @kip-hart . Do you probably have your test script? I tried with mine , and I get the following error. I use Python 3.9.7
File "/isi/w/jana/User_Softwares/Anaconda_2021/lib/python3.9/site-packages/microstructpy/meshing/trimesh.py", line 1274, in write e_ind = np.where(np.all(elems == el, axis=1))[0][0] IndexError: index 0 is out of bounds for axis 0 with size 0

This is my script.

  <input>
     <material>
         <fraction> 1 </fraction>
         <volume>
             <dist_type> lognorm </dist_type>
             <scale> 1 </scale>
             <s> 0.95 </s>
         </volume>
         <shape> sphere </shape>
         <name> Phase 1 </name>
     </material>
 
     <material>
         <fraction> 3 </fraction>
         <volume>
             <dist_type> lognorm </dist_type>
             <scale> 0.5 </scale>
             <s> 1.01 </s>
         </volume>
         <name> Phase 2 </name>
     </material>
 
     <domain>
         <shape> cube </shape>
         <side_length> 7 </side_length>
         <corner> (0, 0, 0) </corner>
     </domain>
 
     <settings>
         <directory> Seeds </directory>
         <verbose> True </verbose>
         <filetypes>
             <seeds> vtk </seeds>
             <poly> vtk </poly>
 			<tri> vtk </tri>
             <seeds_plot> png </seeds_plot>
             <poly_plot> png </poly_plot>
 			<tri_plot> png </tri_plot>
             <verify_plot> png </verify_plot>
         </filetypes>
         <mesher> raster </mesher>
 		<mesh_size> 0.5 </mesh_size>  <!-- this value sets the edge length of each voxel -->
         <seeds_kwargs>
             <linewidths> 0.2 </linewidths>
         </seeds_kwargs>
         <poly_kwargs>
             <linewidths> 0.2 </linewidths>
         </poly_kwargs>
     </settings>
 </input>

[hem-28]

I installed MicroStructPy in a new conda environment, including dependencies. But still end with the same error.

[hem-28]

To resolve the issue, I did the following changes to the write function in RasterMesh class :

        coords = [list(set(ax)) for ax in zip(*self.points)]
        coords = np.sort(coords)   # This was important, else the data was not in ascending order, ParaView threw up an error
        .
        .
        .
        el = np.where(mask)
        try:
              e_ind = np.where(np.all(elems == el, axis=1))[0][0]
              .
              .
              .
             # PhasesInfo
        except :
              continue
        with open(filename, 'w') as file:
        file.write(vtk)

[kip-hart]

ah good catch, looks like I used too many built-in functions. I replaced those line in this commit:

MicroStructPy/src/microstructpy/meshing/trimesh.py

Lines 1173 to 1174 in 4937a69

	pts = np.array(self.points)
	coords = [np.unique(ax) for ax in pts.T]

if the CI pipelines pass I'll push an updated version tonight. Sorry for the error!

[kip-hart]

Hopefully v1.5.6 will solve the problem. Let me know if it doesn't, the script works fine for me.

[hem-28]

Hi @kip-hart Yes it works fine in the new version. Thanks again for the new feature!