Skip to content

Latest commit

 

History

History
297 lines (195 loc) · 13.1 KB

README.md

File metadata and controls

297 lines (195 loc) · 13.1 KB

© Ihor Mirzov, 2019-2024
Distributed under GNU General Public License v3.0

PyPi PyPi downloads
GitHub Github All Releases




Downloads | How to use | Screenshots | Your help | For developers | TODO




CalculiX to Paraview converter (frd to vtk/vtu)

Converts CalculiX ASCII .frd-file to view and postprocess analysis results in Paraview. Generates von Mises and principal components for stress and strain tensors.

Creates separate file for each output interval - it makes possible to animate time history.

Caution! If you have 300 time steps in the FRD, there will be 300 Paraview files. If you need one file - write output only for one step in your CalculiX model.

Hint! If you want/need to only have one file including all of the timesteps, you can easily save everything into one vtkhdf-file in ParaView - either manually in ParaView after loading the .pvd-file or in Python by using ParaView's paraview Package (See section: Create vtkhdf-file).

Converter is tested on CalculiX examples. Here is how some test log looks like.

FRD reader is tested to reduce processing time as much as possible. Now it's quite optimized and fast, but Python itself is slower than C/C++. Here we can do nothing, so, for example, Calmed converter must be faster - another question is if it's able to read and convert any CalculiX results.



How to use

Release Version

Installation

Installation with pip or pipx

To install and run the latest release (version 3.2.0) of of this converter you'll need Python 3 (Python >= 3.9).

# install via pip:
pip install ccx2paraview

VTK needs to be available on your system for ccx2paraview to run, either directly or from ParaView's python package. When you have neither, install VTK as an optional dependency alongside (works also with pipx to install apps, which are exposed on your $PATH and will be run in an isolated environment):

# install via pip:
pip install 'ccx2paraview[VTK]'
# or, with pipx:
pipx install 'ccx2paraview[VTK]' 

Attention! Using vtk and numpy concurrently seems broken in python 3.13. When using pipx on a computer with Python 3.13, install ccx2paraview with a python version < 3.13, e.g.:

pipx install 'ccx2paraview[VTK]' --python 3.12

Installation with a conda environment

You can also use a conda environment to install ccx2paraview:

# Install to a new conda environment: 
conda create -n ccx2paraview_env numpy paraview ccx2paraview

Hint! Don't forget to activate the conda environment before trying to use ccx2paraview:

conda activate ccx2paraview_env

Hint! Installing paraview and ccx2paraview from the conda-forge channel can be achieved by adding conda-forge to your channels with:

conda config --add channels conda-forge
conda config --set channel_priority strict

Usage

Having installed ccx2paraview, run the converter with command (both in Linux and in Windows):

ccx2paraview yourjobname.frd vtk
ccx2paraview yourjobname.frd vtu

Also you can pass both formats to convert .frd to .vtk and .vtu at once.

There are also the following aliases for converting files to a fixed format

ccxToVTK yourjobname.frd
ccxToVTU yourjobname.frd

Using ccx2paraview in your python code

To use the current release of ccx2paraview in your python code (having installed with pip or into a conda environment):

import logging
from ccx2paraview import Converter
logging.basicConfig(level=logging.INFO, format='%(levelname)s: %(message)s')
c = Converter(frd_file_name, ['vtu'])
c.run()

Create vtkhdf-file using ParaView's simple Module

While ccx2paraview cannot convert to vtkhdf directly (yet), you can create such a single file bundling all of the timesteps into one by using ParaView's simple Module. When you are using a conda environment with ParaView, a working version of ParaView's simple Module should be available in the environment.

# use ccx2paraview to convert the file '/Users/ccx/ball.frd' 
# into vtu-files ('/Users/ccx/ball.x.vtu') and write a pvd-file
from ccx2paraview import Converter
c = Converter('/Users/ccx/ball.frd', ['vtu'])
c.run()

# Convert all vtu-files into one vtkhdf file 
# - the .vtkhdf-extension is mandatory for ParaView to write the correct output
# - The Compression level (0-9) is set to 4 here, making the vtkhdf-file 
#   roughly the same size as the input vtus toghether
from paraview.simple import (SaveData, PVDReader)
pvd_proxy = PVDReader(registrationName='ball', FileName='/Users/ccx/ball.pvd')
SaveData('/Users/ccx/ball.vtkhdf', proxy=pvd_proxy, WriteAllTimeSteps=1, CompressionLevel=4)

General Remarks

Please, pay attention that .frd-file type should be ASCII, not binary! Use keywords *NODE FILE, *EL FILE and *CONTACT FILE in your INP model to get results in ASCII format.

It is recommended to convert .frd to modern XML .vtu format - it's contents are compressed. If you have more than one time step there will be additional XML file created - the PVD file. Open it in Paraview to read data from all time steps (all VTU files) at once.

Starting from ccx2paraview v3.0.0 legacy .vtk format is also fully supported - previously there were problems with component names.

Attention! While developing this converter I'm using latest Python3, latest VTK and latest ParaView. If you have problems with opening conversion results in ParaView - update it.

Hint! When using the conda environment, a working version of ParaView should be available in the environment already.

Python Compatibility

Installation of the latest release via pip was tested with a fresh install of vtk and numpy and:

  • Python 3.9: works!
  • Python 3.10: works!
  • Python 3.11: works!
  • Python 3.12: works!
  • Python 3.13: ERROR: No matching distribution found for vtk

Using a conda-environment (with numpy and paraview):

  • Python 3.9: works!
  • Python 3.10: works!
  • Python 3.11: works!
  • Python 3.12: works!
  • Python 3.13: works!

Paraview programmable filter

A snippet for Paraview programmable filter to convert 6 components data array to full tensor:

import numpy as np 
res = np.array([])
pd = inputs[0].PointData['S']
for xx,yy,zz,xy,yz,xz in pd:
    t = np.array([[xx,xy,xz],[xy,yy,yz],[xz,yz,zz]])
    res = np.append(res, t)
tensor = dsa.VTKArray(res)
tensor.shape = (len(pd), 3, 3)
output.PointData.append(tensor, 'S_tensor')

A snippet for Paraview programmable filter to calculate eigenvalues and eigenvectors:

import numpy as np
eigenvalues = np.array([])
eigenvectors = np.array([])
pd = inputs[0].PointData['S']
for xx,yy,zz,xy,yz,xz in pd:
    t = np.array([[xx,xy,xz],[xy,yy,yz],[xz,yz,zz]])
    w, v = np.linalg.eig(t)
    w_ = np.absolute(w).tolist()
    i = w_.index(max(w_))
    eigenvalues = np.append(eigenvalues, w[i]) # max abs eigenvalue
    eigenvectors = np.append(eigenvectors, v[i]) # max principal vector
eigenvectors = dsa.VTKArray(eigenvectors)
eigenvalues = dsa.VTKArray(eigenvalues)
eigenvectors.shape = (len(pd), 3)
eigenvalues.shape = (len(pd), 1)
output.PointData.append(eigenvectors, 'S_max_principal_vectors')
output.PointData.append(eigenvalues, 'S_max_eigenvalues')



Development Version

Installation from github

To install this converter from github you'll need Python 3 and optionally conda:

# install vtk first
pip install vtk
pip install git+https://github.com/calculix/ccx2paraview.git

# or, with conda (paraview has vtk, so no need to install it seperately):
conda create -n ccx2paraview_devel python numpy paraview
conda activate ccx2paraview_devel
pip install git+https://github.com/calculix/ccx2paraview.git

Attention! Currently, installing vtk via pip seems to break ParaView's pvpython. When using the conda environment, ParaView's included vtk will be used (alongside having a working ParaView in the environment).



Screenshots

Converted von Mises stress field with Turbo colormap:
baffle

Converted translations field with Viridis colormap:
blades



Your help

Please, you may:



For developers

PyPI pyversions Visual Studio Code

CalculiX-to-Paraview Converter

To install and use ccx2paraview-package: see above.

To test ccx2paraview from local sources after cloning from github, you'll find yaml-files in the ./tests-folder. Using the VSCode extension Conda Wingman they can easily be built and activated from within VSCode.

The binaries are created automatically when installing with pip from github via the project scripts in pyproject.toml:

[project.scripts]
ccx2paraview = "ccx2paraview.cli:main"
ccxToVTK = "ccx2paraview.cli:ccx_to_vtk"
ccxToVTU = "ccx2paraview.cli:ccx_to_vtu"

If you have Python version >= 3.8 create binary with nuitka:

pip3 install nuitka

In Windows:
set CC=C:\\MinGW64\\mingw64\\bin\\gcc.exe
python3 -m nuitka --follow-imports --python-flags=-m ccx2paraview

In Linux:
python3 -m nuitka --follow-imports --python-flags=-m ccx2paraview

If you have Python version < 3.8 create binary with pyinstaller:

pip3 install pyinstaller
pyinstaller __init__.py --onefile

Read how to create packages for pypi.org:

python3 -m pip install --upgrade build twine
python3 -m build
python3 -m twine upload dist/*

Read about VTK file formats and VTK unstructured grid. Remember that FRD file is node based, so element results are also stored at nodes after extrapolation from the integration points.



TODO

Test CALMED binary.

Log memory consumption.

Read binary .frd files: https://github.com/wr1/frd2vtu

Read DAT files: it would be a killer feature if Paraview could visualize results in Gauss points. Use CCXStressReader.

Contribute to meshio. FRD writer. Use meshio XDMF writer: #6

Add element’s material tangent stiffness tensor. Easiest for the paraview user would be to provide it in the (deflected) global cartesian frame. This dataset is useful for checking input data for anisotropic materials, as well as for the stuff with inverse design of fields of this tensor. But it’s a lot more work to produce, especially with nonlinear materials. It’s almost as useful to see the highest principal value of the stiffness, as a scalar or a vector. (but for the vector you need to do the transformation)