Using Python on the CSF for Post-processing

This content is an addition to the previous section here.

After getting the geom_load.hdf5 file as the result of DAMASK simulation. We can do processing with Python codes on CSF.

Output .vti File from .hdf5

.vti file is used to store series of image data, which can be exported during post-processing and visualised by ParaView. Here is a detailed example can be named postprocessing.py(You may create a new folder and make the file):

import damask                                           # Necessary for .add, .export_VTK commands.
import numpy as np                                      # Necessary for sort out the axes.

f = damask.Result('/path/to/geom_load.hdf5')            # Find the .hdf5 file.

l1 = np.array([1,0,0])                                  # Represent X axis.
l2 = np.array([0,1,0])                                  # Represent Y axis.
l3 = np.array([0,0,1])                                  # Represent Z axis.
for inc in [0, 10, 20, 30, 40]:                         # Depends on which time-points you want to visaulise, e.g. the 5 time-points from step 0 to 40. 
  f_inc = f.view(increments=inc)
  f_inc.add_stress_Cauchy()                             # Calculate Cauchy stress and write into .hdf5 file
  f_inc.add_equivalent_Mises('sigma')                   # Calculate Von-Mises stress and write into .hdf5 file
  f_inc.add_equivalent_Mises('epsilon_V^0.0(F)')        # Calculate equivalent strain and write into .hdf5 file
  f_inc.add_IPF_color(l=l1)                             # Apply IPF colours along X to RVE.
  f_inc.add_IPF_color(l=l2)                             # Apply IPF colours along Y to RVE.
  f_inc.add_IPF_color(l=l3)                             # Apply IPF colours along Z to RVE.
  f_inc.export_VTK(output=['sigma', 'epsilon_V^0.0(F)', 'O', 'IPFcolor_(1 0 0)','IPFcolor_(0 1 0)','IPFcolor_(0 0 1)', 'phi'])   # Export .vti with the parameters added above.
  print(f_inc.list_data())                              # Show existed dataset.

We can modify it by adding more parameters we need to analyse. It is worth to mention that though the command is called .export_VTK, the output file is in the format of ‘.vti’, not .vtk or .vtr. It is a standard output format of the latest damask module.

Don’t .add the same parameters twice otherwise there will be errors. A good way to check if the dataset already existed is print(f_inc.list_data()) where we can check all the parameters we added before.

Output Stress vs Strain Curve from .hdf5

Write another python script or just add the following lines to the postprocessing.py:

import damask
import matplotlib.pyplot as plt

f = damask.Result('/path/to/geom_load.hdf5')                               # Find the .hdf5 file.
f.view(increments=-1).get().keys()                                         # Aanlysis the last step.
sigma = [np.average(s) for s in f.get('sigma_vM').values()]                # Acess the stress.
epsilon = [np.average(e) for e in f.get('epsilon_V^0.0(F)_vM').values()]   # Acess the strain.
plt.plot(epsilon,sigma)                                                    # Plot Stress vs Strain Curve.
plt.savefig('stress_vs_strain_plot.png')                                   # Save the picture.

Submit task to CSF

The postprocessing.py can be run by the command ‘python postprocessing.py’, however, when the .hdf5 file is big in size, it would take ages to complete the processing. Therefore, it is better to use a parallel environment to run the task by create another file ‘postprocessing_jobscript.sh’:

#!/bin/bash --login
#$ -cwd                                                 # Submit in the current working directory
#$ -pe smp.pe 4                                         # Use a parallel environment with 4 cores
#$ -l short                                             # Select type of core, use mem512 when the task is huge.

module load apps/binapps/anaconda3/2021.11              # Load Anaconda
source activate /mnt/eps01-rds/jf01-home01/shared/.conda/damask_v3a7      # Activate the conda environment, can be change to other envs.

python postprocessing.py interactive                   # Run with Python

Then we can type the command qsub postprocessing_jobscript.sh to submit the task to CSF.