3D Slice Data

3-D Slice Data

An &SLCF whose XB spans a volume rather than a plane produces a 3-D slice. readSLCF3Ddata reads every mesh's contribution and interpolates it onto one grid.

Reading a volume

import os
import numpy as np
import pyfdstools as fds

workingDir = os.path.join(
    os.path.dirname(fds.__file__), 'examples', 'data', 'case001.zip')
chid = 'case001'

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'TEMPERATURE')

print(grid.shape)     # (25, 11, 25, 3) -> (NX, NY, NZ, xyz)
print(values.shape)   # (25, 11, 25, 6) -> (NX, NY, NZ, NT)
print(units)          # C

Note the argument order: readSLCF3Ddata takes chid before workingDir, unlike query2dAxisValue.

Return value	Contents
grid	array(NX, NY, NZ, 3) of absolute coordinates
values	array(NX, NY, NZ, NT) of slice values
times	timestamp of each frame
units	units of the quantity

Cells outside every mesh are NaN.

When no 3-D slice exists

If the case wrote no 3-D slice of the requested quantity, readSLCF3Ddata prints what the case does contain and returns (False, False, False, False):

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'NOT A REAL QUANTITY')

if grid is False:
    raise SystemExit('quantity not present in this case')

Extracting a plane

Giving axis and value extracts just that plane. This is much cheaper than reading the whole volume and slicing it yourself, because only the meshes intersecting the plane are read:

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'TEMPERATURE', axis=3, value=1.0)

print(grid.shape)     # (25, 11, 2) -> (N, M, two in-plane coordinates)
print(values.shape)   # (25, 11, 6) -> (N, M, NT)

Meshes that do not span the requested plane are skipped, and verbose=True reports which:

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'TEMPERATURE', axis=3, value=1.0, verbose=True)

Slicing a volume you already read

If you already have the whole volume in memory, findSliceLocation takes a plane out of it without going back to disk:

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'TEMPERATURE')

x, z, frame = fds.findSliceLocation(grid, values[:, :, :, -1], axis=3, value=1.0)

fig, ax = fds.plotSlice(x, z, frame, axis=3, clabel='Temperature (C)')

It picks the nearest grid plane rather than interpolating, so the plane you get is the one closest to value.

Extracting a point

history = fds.extractPoint([2.0, 1.5, 1.0], grid, values)

Returns the time history at the nearest grid point, and warns if that point is more than 0.25 m away from the one you asked for.

Dumping the volume to csv

pyfdstools/examples/dump_3d_slice_to_csv.py writes a 3-D slice out as one csv per z-plane per frame:

python dump_3d_slice_to_csv.py \
    --chid case001 --quantity TEMPERATURE \
    --working_dir data/case001.zip

Speed and memory

A 3-D slice over a long run is large: NX × NY × NZ × NT float32 values per mesh, plus the assembled array. Two things help:

• Restrict the time range. Passing time and dt reads only the frames inside the window.
• Extract the plane at read time. Passing axis and value avoids materialising the volume at all.

grid, values, times, units = fds.readSLCF3Ddata(
    chid, workingDir, 'TEMPERATURE',
    time=60, dt=20, axis=3, value=1.0)

saveTimesFile=True caches each slice file's timestamps in a csv next to it, which speeds up repeated reads of a case on disk.

Exporting to ParaView

For interactive 3-D work, export to VTK and open the result in ParaView:

fds.exportSl3dDataToVtk(chid, workingDir, outDir='./vtk')

Requires the paraview extra. See Exporting to ParaView.

Deprecated

readSLCF3DdataXYZ does the same job using the .xyz files FDS writes when WRITE_XYZ=.TRUE. is set on &DUMP. It emits a DeprecationWarning; use readSLCF3Ddata, which reads the mesh grids from the smokeview file and so works on any case.