Division of volume in smaller axial volumes, like slices, and plot overline about pyvista-support

[banesullivan]

Hi @byba1990, thanks for posting your question! This sounds like a sophisticated use case and I'll do my best to help you complete these two tasks.

First, would this dataset be something you are willing to attach here so that I could work directly with the data you have in mind?

Below are some first passes at addressing your questions - please let me know if I misinterpreted any of your questions or if these are/aren't helpful.

Part 1

Is there a way to divide axially a volume, in order to compute total and average variables over this smaller volumes obtained?

Perhaps the .clip filter would work well for you. This filter cuts/clips a mesh/volume along a user defined plane. Here's an example to clip a mesh at the center normal to the Y-axis:

import pyvista as pv
from pyvista import examples

# Load and clip a mesh
mesh = examples.load_channels()
clipped = mesh.clip(normal='y')

# Compare clipped domain to whole domain
p = pv.Plotter()
p.add_mesh(mesh.outline(), color='white')
p.add_mesh(clipped)
p.show()

# Now compute the total and average variables over the clipped domain
variable = clipped.active_scalar_name

average = np.average(clipped[variable])
total = np.sum(clipped[variable])
print('Average: {}'.format(average))
print('Total: {}'.format(total))

Average: 0.27347072
Total: 1709192

Alternative clipping/sub-meshing

You could also leverage:

• the .clip_box() filter to extract specific volumetric regions of your mesh
• the .extract_cells() filter under UnstructuredGrids via the cast_to_unstructured_grid() filter to extract specific non-voxel cells in your mesh.
• the .threshold() filter to extract regions of the mesh that have data values in a give range to understand distributions of variable within specific ranges.

Part 2

Perhaps the best alternative would be to create a high resolution line source (maybe we can use @akaszynski's brand-new Spline feature in PyVista?) to sample a dataset much like what you have in the code snippet above then plot the variables in 2D via matplotlib with X-axis being the distance along the line and the Y-axis being the variable of interest's values. Here's an example with an arbitrarily placed line at a high-ish resolution:

import pyvista as pv
import numpy as np
import matplotlib.pyplot as plt
from pyvista import examples

# Example dataset with some interesting data values
mesh = examples.download_kitchen()
mesh.plot()

# Make two points to construct the line between
a = [mesh.bounds[0], mesh.bounds[2], mesh.bounds[4]]
b = [mesh.bounds[1], mesh.bounds[3], mesh.bounds[5]]
line = pv.Line(a, b, resolution=100)

# And preview line's extent
p = pv.Plotter()
p.add_mesh(mesh, style='wireframe', color='w')
p.add_mesh(sampled, color='b')
p.show()

# Sample the mesh's variables onto the line
sampled = line.sample(mesh)

# Compute distances of every point along line
#  NOTE: this only works for linear lines
compute = lambda p0, p1: np.sqrt(np.sum((p1 - p0)**2, axis=1))
distance = compute(a, line.points)

# Grab a variable of interest
variable = mesh.active_scalar_name
values = sampled[variable]

# Plot it in 2D
if values.ndim > 1:
    for i in range(values.shape[1]):
        plt.plot(distance, values[:, i], label='Component {}'.format(i))
    plt.legend()
else:
    plt.plot(distance, values)
plt.xlabel('Distance')
plt.ylabel(variable)
plt.title('Plot Variable Along Line')
plt.show()

[banesullivan]

Please note that this inspired a new .plot_over_line() filter in pyvista/pyvista#250 - @byba1990, I'd greatly appreciate your feedback on this new filter

[banesullivan]

Hi @byba1990 - checking in to see if you found these details helpful. Please let us know if you have any further questions or if this issue can be closed.

[byba1990]

Dear @banesullivan,

thank you very much for your time and kindness! Your suggestions has been very helpful! Sorry for my late reply.

Howerver, to sum up:

1. filter to select volumes with clip. It seems that it works but from your suggestion:

# Now compute the total and average variables over the clipped domain
variable = clipped.active_scalar_name

average = np.average(mesh[variable])
total = np.sum(mesh[variable])
print('Average: {}'.format(average))
print('Total: {}'.format(total))

the average and total should be referred to clipped (only for values over this subregion), right?
During my own tests, I checked the cells for clipped and mesh but, if my clipped is a subregion that corresponds to half of my initial mesh:

total cell number of mesh=7560
total cell number of clipped=11340

Why? I suppose that the number of clipped cells is half of mesh cells, i.e. 3780. Moreover, with

sizeClipped=clipped.compute_cell_sizes()

I have negative values for volume.
However, checking the total volumes of clipped and mesh:

sizeClipped=clipped.compute_cell_sizes()
sizeMesh=mesh.compute_cell_sizes()
TotVolClip=np.abs(np.sum(sizeClipped['Volume'])) # abs to have positive volumes
TotVolMesh=np.sum(sizeMesh['Volume'])
fraction=TotVolClip/TotVolMesh

my fraction results 0.500, correctly. So I think that total and average values are correct for clipped.

2. Plot overline works perfectly, thanks. Thanks also for your appreciation, in relation to the inspiration of new plot_overline() filter for pyvista.

Thank you very much for your time Bane!
Regards,

Christian

[banesullivan]

Hi @byba1990:

Typo

the average and total should be referred to clipped (only for values over this subregion), right?

Ah yes, my mistake - those should indeed be called on clipped, not mesh. I just edited my original post so this doesn't confuse newcomers

Number of Cells in Clipped Mesh

total cell number of mesh=7560
total cell number of clipped=11340

Why? I suppose that the number of clipped cells is half of mesh cells, i.e. 3780. Moreover, with

This is an artifact of the clip filter. When clipping the input mesh, it is tessellated into triangles or tetrahedrals. So if the input mesh has voxels or non-triangular cells, then it will have to be regenerated as several triangular cells put together - thus a clipped mesh could have more cells than the input mesh which is what you are seeing. You could pass the show_edges=True argument when plotting to see the tetrahedrals.

This is an issue that has been greatly bugging me for a while that I do not have an elegant solution for at the moment. See also: https://discourse.vtk.org/t/is-it-possible-to-turn-off-triangles-on-vtkclipdataset-filter/279

On this note, the slice filter allows us to not generate the triangles so it should be possible to do this with the clip filter as well.

Negative Volume

sizeClipped=clipped.compute_cell_sizes()

I have negative values for volume.

Hm... that's really strange - would you mind uploading your mesh here in a zip archive so I could toy with it? (mesh.save('mymesh.vtk'))

[banesullivan]

pyvista/pyvista@a68b45b fixes the clip triangulation issue