Extracting Exodus information with netcdf-python
In an earlier post I demonstrated how to use a set of ctypes wrappers of Sandia's Exodus database format, i.e. the exodus.py script that is distributed with Trilinos, to extract information from an Exodus database. Exodus is a finite element format library that is build upon NetCDF4 and HDF5. While the use of exodus.py makes for a straightforward and intuitive interface to the database, it requires that Trilinos (at least the SEACAS packages) be installed as shared libraries with an $ACCESS environment variable in your $PATH pointing to the library root. This setup is cumbersome if you don't have any other reason to have Trilinos installed. Thankfully, since Exodus files are NetCDF files are HDF files, we can use other tools to access the data. One of those, that is mature and available in the Python Package Index is netcdf4-python. It can easily be installed via pip i.e.
pip install netcdf4
or if your using the popular Anaconda Python distribution then
conda install netcdf4
should do the trick. This is all we need to have installed to read from an Exodus database. The simple 2D finite element mesh I will be using in the following example can be downloaded here
First we'll import the netCDF4 module and of course Numpy.
import netCDF4
import numpy as np
To open the Exodus file for reading we use the following command.
nc = netCDF4.Dataset('../files/s_curve.g')
Here we are using a "Genesis" file which is a Exodus file used for finite element model input. It typically contains information about the nodal coordinates, mesh connectivity, nodesets for boundary condition application, etc. A common source of creation for these forms of Exodus files is the mesh generation tool Cubit. We could access the output variant of an Exodus file (i.e. a *.e file) in the same manner.
Unless you take a lot of time carefully naming all your variables, nodesets, blocks, etc. during the creation of the database, the variables will be given default names which we can use to access that data. The default names are intuitive enough so that with a little investigation of thier properties, you can usually figure out what they refer to. To see all of the information of the database we can simply
print nc
Or to get even more information about the variables
nc.variables
I won't attempt to decipher all the information in the database as you can use the Exodus API reference for that. Below, I'll show a short example of how we might do something useful. First we will extract the X and Y nodal coordinate locations from and the connectivity array. The first 10 entries of the connectivity array are shown.
X = nc.variables['coordx']
Y = nc.variables['coordy']
connect = nc.variables['connect1']
print connect[0:10]
Now we'll use the nodal locations and the connectivity array to plot a visualization of the mesh. First, we upload the required libraries from matplotlib.
%matplotlib inline
from matplotlib.patches import Polygon
from matplotlib.collections import PatchCollection
import matplotlib.pyplot as plt
import matplotlib
Now we'll create an array that contains the $x,y$ nodal coordinate location pairs and extract, via Numpy fancy indexing, the coordinates for each quadralateral in the mesh creating polygons with them for visualization.
xy = np.array([X[:], Y[:]]).T
patches = []
for coords in xy[connect[:]-1]:
quad = Polygon(coords, True)
patches.append(quad)
fig, ax = plt.subplots()
colors = 100 * np.random.rand(len(patches))
p = PatchCollection(patches, cmap=matplotlib.cm.coolwarm, alpha=0.4)
p.set_array(np.array(colors))
ax.add_collection(p)
ax.set_xlim([-4, 4])
ax.set_ylim([-4, 4])
ax.set_aspect('equal')
plt.show()
The colors have no meaning other than to give each element a random color. Now we'll close the Exodus file, completing our analysis.
nc.close()
Comments
Comments powered by Disqus