Pnextract Guide
pnextract_guide
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 3
1 Input file 1
Network extraction code – pnextract
pnextract extracts a conventional pore network from a microCT image. The
algorithm is a rewrite of the Dong and Blunt (2009) code. There are major
differences though. First, the pore and throat detection algorithm is revised;
see Stages 1 and 2 described in Raeini et al. (2017) https://doi.org/10.1103/
PhysRevE.96.013312. Raeini et al. (2017) is an extension of this code. The
shape of pores in this code are deduced from shape factors, the shape-factor
equation is changed compared to the old definition, see Bultrys et al (2018,
currently under-review).
1 Input file
The input file for the network extraction code is a mhd header file compatible
with paraview and Fiji (ImageJ with plugins) with additional optional keywords
specific to network extraction algorithm. See the file Image.mhd for a sample
input.
1.1 Format specifications:
1. The order of the first 6 keywords should not be changed for compatibility
with third-party software (ImageJ and Paraview)
2. Use ”#” for comments
3. All keyword and its data should be given in a single line
Important keywords:
1 to 3rd keywords (should not be changed):
1. ObjectType = Image
2. NDims = 3
3. ElementType = MET_UCHAR
4. keyword: DimSize – used to assign the dimensions of the image: Nx, Ny
and Nz
5. keyword: ElementSpacing – used for assigning voxel size: δx,δy and δz
should be equal
6. keyword: ElementDataFile – specifies the name of binary 8bit data file
(.raw), ascii (.dat), .raw.gz, and .tif files are supported too.
1 Input file 2
ObjectType = Image
NDims = 3
ElementType = MET_UCHAR
DimSize = 400 400 400
ElementSpacing = 5.345 5.345 5.345
Offset = 0 0 0
ElementDataFile = Berea.raw
Fig. 1: Sample input header file
Medial-surface settings:
The medialSurfaceSettings is an optional technical keyword which can be used
for sensitivity analysis, for instance.
medialSurfaceSettings 0.1 0.9 0.7 0.5 1.5 1.21 7 0.25 1.6;
where the keyword arguments are clipROutx clipROutyz midRFrac RMedSurfNoise
lenNf vmvRadRelNf nRSmoothing RCorsf RCors, respectively.
The pnextract code produces few lines showing the settings being used.
something like:
medialSurfaceSettings: 0.05 0.98 0.7 2.75 0.6 1.1 3 0.15 1.75
medialSurfaceSettings:
clipROutx : 0.05
clipROutyz : 0.98
midRFrac : 0.7
RMedSurfNoise : 2.75
lenNf : 0.6
vmvRadRelNf : 1.1
nRSmoothing : 3
RCorsf : 0.15
RCors : 1.75
The first line is the keyword and its parameters and the rest are short names
for each of the parameters and their values. In case you want to do a quick
evaluation, you can copy the first line into the pnextract input, the .mhd file,
and change the parameters and re-run the code. Here is a short explanation for
these parameters:
clipROutx is used to limit the size of maximal-spheres extending outside the
rock image in the x direction.
1 Input file 3
clipROutyz is used to limit the size of maximal-spheres extending outside the
rock image in the y and z directions.
midRFrac is the relative size of the distance-map of the voxel between two
maximal-spheres, for the spheres to be considered part of the same pore.
RMedSurfNoise is a measure of noise amplitude. Decreasing this will likely
increase the number of pores, but it also affects the number of corners
per throat.
lenNf is a relative distance for merging adjacent pores which are too close to
each other.
vmvRadRelNf is the relative size of the throat between the two pore considered
for merging, the contraction should be less than this to merge the nearby
pores (that are less than lenNf apart), otherwise the pore will not be
merged. Decreasing these two will increase the number of pores.
nRSmoothing applies a small amount of Gaussian-like smoothing on the com-
puted distance map, which in turn affect the rest of the computations.
Decreasing this will probably increases the number of pores.
RCorsf controls the distance between the maximal spheres. This is a sensitive
parameter, changing it may need changing other parameters to get good
results.
RCors controls the minimum distance between (small) maximal-spheres. This
is a sensitive parameter, changing it may need changing other parameters
to get good results.
Contact:
For any queries please email:
Ali Q. Raeini : a.qaseminejad-raeini09@imperial.ac.uk
References:
See http://www.imperial.ac.uk/earth-science/research/research-groups/
perm/research/pore-scale-modelling/publications/
