Implement Aiden's single phase microstructure workflow

[gcapes]

I met with Aiden yesterday and discussed recent changes to his workflow.
The dual phase microstructure is no longer needed, so he can probably use the task generate_volume_element_from_statistics, which also uses Dream3d.

@aidenmha has sent me his current, simpler workflow (below) and I'll try to implement that.
There are actually two workflows because there is a manual editing step of the HDF5 volume element file (from dream3D) to cluster orientations. This clustering is done in a jupyter notebook. The HDF5 file is then read in, in the second workflow.

I'm not sure how to do that - perhaps with the (unused?) script https://github.com/hpcflow/matflow-new/blob/develop/matflow/data/scripts/import_VE.py?

It seems that it should be possible to implement the code from the jupyter notebook as a MatFlow task schema and have this as a single, longer workflow, but in the shorter term, implementing this as-is would be the starting point.

The workflow files:

name: grain_growth_from_VE_dream3D_nucleus_with_MTEX_texture_binned
run_options:
  l: short
  
#archive: dropbox

tasks:

    # rotated cube texture for the sub-grain matrix:
  - name: sample_texture
    method: from_model_ODF
    software: mtex
#    context: phase_1
    base:
      num_orientations: 2300
      crystal_symmetry: cubic
      specimen_symmetry: orthorhombic
      ODF_components:
        - type: unimodal
          component_fraction: 1.0 
          modal_orientation_HKL: [0, 0, 1]
          modal_orientation_UVW: [1, 1, 0]
          halfwidth: 5


  - name: visualise_orientations
    method: pole_figure
    software: mtex
#    context: phase_1
    base:
      crystal_symmetry: cubic
      pole_figure_directions:
        - [0, 0, 1]
        - [1, 0, 1]
        - [1, 1, 1]
      use_contours: false


  - name: generate_volume_element
    method: from_statistics
    software: Dream3D
    base:
      grid_size: [512, 512, 1]
      resolution: [1, 1, 1] # side length is 512
      periodic: false
      phase_statistics:
        - type: primary
          name: sub-grain-matrix
          crystal_structure: cubic
          volume_fraction: 1.0
          size_distribution:
            ESD_mean: 7.03431
            ESD_log_stddev: 0.04
            num_bins: 15
#        - type: precipitate #81-101
#          name: nuclei
#          crystal_structure: cubic
#          volume_fraction: 0.0
#          size_distribution:
#            ESD_mean: 2
#            ESD_log_stddev: 0.1
#            num_bins: 15
#          number_fraction_on_boundary: 1
#          radial_distribution_function:
#            min_distance: 10
#            max_distance: 80
#            num_bins: 50
#            box_size: [100, 100, 100]
#      precipitates:
#        - phase_number: 2
#          position: [256, 256, 0.5]
#          major_semi_axis_length: 45
#          mid_semi_axis_length: 45
#          minor_semi_axis_length: 45
#          euler_angle: [0, 0, 0]

  - name: visualise_volume_element
    method: VTK
    software: damask

name: grain_growth_from_VE_dream3D_nucleus_with_MTEX_texture_binned
run_options:
  l: 
import:
 - parameter: volume_element
   from: 
     workflow: "/scratch/j36293ah/nonuc/hw_7_with_seeds/runs_final/setup/2pcf/10000/workflow.hdf5" 
#archive: dropbox

tasks:

    # rotated cube texture for the sub-grain matrix:
#  - name: sample_texture
#    method: from_model_ODF
#    software: mtex
 #   context: phase
#    base:
#      num_orientations: 1000
#      crystal_symmetry: cubic
#      specimen_symmetry: orthorhombic
#      ODF_components:
#        - type: unimodal
#          component_fraction: 1.0 
#          modal_orientation_HKL: [0, 0, 1]
#          modal_orientation_UVW: [1, 1, 0]
#          halfwidth: 7


#  - name: visualise_orientations
#    method: pole_figure
#    software: mtex
#    context: phase_1
#    base:
#      crystal_symmetry: cubic
#      pole_figure_directions:
#        - [0, 0, 1]
#        - [1, 0, 1]
#        - [1, 1, 1]
#      use_contours: false


#  - name: generate_volume_element
#    method: from_statistics
#    software: Dream3D
#    base:
#      grid_size: [512, 512, 1]
#      resolution: [1, 1, 1] # side length is 512
#      periodic: false
#      phase_statistics:
#        - type: primary
 #         name: sub-grain-matrix
 #         crystal_structure: cubic
#          volume_fraction: 1.0
#          size_distribution:
#            ESD_mean: 10
#            ESD_log_stddev: 0.03
#            num_bins: 15

#  - name: visualise_volume_element
#    method: VTK
#    software: damask

  - name: generate_phase_field_input
    method: from_volume_element
    software: cipher
    run_options:
      l: short
      num_cores: 2
    base:
      materials:
        - name: sub-grain-matrix
          properties:
            chemicalenergy: none
            molarvolume: 1e-5
            temperature0: 500.0

      interfaces:
        - materials: [sub-grain-matrix, sub-grain-matrix] 
          properties:
            width: 6.0
            energy:
              e0: 1.0e+8
            mobility:
              m0: 3.333e-11

      components: [ti]
      outputs: [phaseid, matid, interfaceid, 0_phi]
      solution_parameters:
        abstol: 0.0001
        amrinterval: 25
        initblocksize: [1, 1]
        initcoarsen: 9
        initrefine: 9
        interpolation: cubic
        maxnrefine: 9
        minnrefine: 0
        outfile: out
        outputfreq: 100
        petscoptions: -ts_adapt_monitor -ts_rk_type 2a
        random_seed: 1579993586
        reltol: 0.0001
        time: 1000_000
      interface_binning: # specify one or both of energy_range/mobility_range:
        base_interface_name: sub-grain-matrix-sub-grain-matrix
        theta_max: 50
        bin_width: 1
        energy_range: [0.1e+8, 0.555e+8]
        mobility_range: [2.105e-11, 10.0e-11]
        n: 7 # mobility parameter
        B: 625 # mobility parameter

  - name: simulate_grain_growth
    method: phase_field
    software: cipher
    base:
      num_VTU_files: 22
      derive_outputs:
        - name: num_voxels_per_phase 
      save_outputs:
        - name: phaseid 
          time_interval: 5_000
        - name: matid
          number: 4
        - name: interfaceid
          number: 4
        - name: 0_phi
          number: 4
        - name: num_voxels_per_phase
          time_interval: 5_000
      delete_VTUs: true
    run_options:
      num_cores: 8

[gcapes]

I've done the first one, but it's a bit of a mess because it needs to be on the cipher branch as it depends on that software, but I've had to comment out the previous task schema/demo workflow file in order to run this without errors (matflow won't run with errors in the demo workflows/task schemas).

[gcapes]

Maybe for loading the VE after modification I can reuse matflow-new/matflow/data/scripts/dream_3D/parse_dream_3D_volume_element.py as it will need to load from an hdf5 file?

[gcapes]

Hi @aidenmha,
Could you please share with me your jupyter notebook that does the clustering of orientations? I think I'll try to implement it as a task schema.
Cheers!

[gcapes]

Note from Adam: this can be a generic VE modifying task - it's not specific to dream3D

[gcapes]

It might be more complicated than that. The notebook loads a geom.vtr file from Damask to get a material variable.

[gcapes]

Should be able to get "material" info from the VE itself rather than having to load geom.vtr.

The volume element has a key something like element_material_idx, which is a 3d array. This corresponds to material variable from from hdf5 file.
Load the first part of the workflow into a notebook and poke around.

[gcapes]

I can't see how to do this. The volume_element doesn't look to be a dictionary or object that I can easily inspect, so I can't see the element_material_idx key.
I've asked for help on slack, but this is what I've tried:

[gcapes]

The answer is wk.tasks.generate_volume_element_from_statistics.elements[0].outputs.volume_element.value['element_material_idx']

[gcapes]

Before I try to implement that, I think it would be useful to complete the workflow with the "step 2" tasks, before finally adding the (yet to be written) clustering task. That way I'll know the "plumbing" works correctly with existing tasks.

[gcapes]

Well it works in so far that it doesn't. I (somewhat obviously) get the same error from the same task that I did in the dual phase workflow:
#432 (comment)

But given the workflow submits and runs, I think the syntax is all ok.

So ... back to implementing the clustering task schema.

[gcapes]

@aidenmha, have you already sent me these files: alpha\goldilocks_ori.npy and gamma\gamma_mix.npy which are loaded by the notebook?
Thanks

[gcapes]

I've been trying to build this up from the simplest possible task which is to return the volume element unaltered. I get this error, which I think is because the variables are stored as zarr arrays, and to be saved as output variables and return from the task schema, need converting to numpy arrays, and reassembling into the original data format of the volume_element.

ValueError: Parameter data with type <class 'zarr.core.Array'> cannot be serialised into a ZarrStoreParameter: <zarr.core.Array '/parameters/arrays/param_27/arr_0' (3,) int64 read-only>.

[gcapes]

I've used validate_volume_element from damask_parse to convert these to numpy arrays. This looks very close now, but I'm unable to import Shuffle from Sam's subsurface package. He's doing something of a overhaul which has temporarily removed this, so it looks like I need an older version?

[gcapes]

Try using the master branch from https://github.com/samuel-hg-engel/subsurface/commits/master/