When a walker hits a wall it is bounced back to where it came from then continues randomly under the assumption of Brownian motion. It would be possible to fix the trajectory after hitting the wall for a certain number of steps to increase the effects of wall interactions as opposed to particle-particle interactions. This could be done either by setting a fixed number of steps or adjusting the probability of moving in a particular direction along a particular axis. The second way is probably better from a physics perspective but might be trickier to implement as it would require storing movement probabilities for each walker.

@TomTranter Regarding the idea of doing Knudsen tortuosity with this package, how hard would it be to replace the random step with a predetermined step based on a mean free path value?

The example does run due to the failure of parallelpython to run when not in a main loop. We should either update the readme, or fix the code.

Hi all,

I have one question related to direction-dependent tortuosity. I calculated tortuosity using 3D image. It gives me msd tortuosity, asd0 tortuosity, asd1 tortuosity, asd2 tortuosity, my question is does asd0, asd1, asd2 means the tortuosity in x,,y,z direction of the 3D image? It bothers me a lot and it would be great if this question can be solved. Thanks a lot！

Wenhui

Hi there,

Sorry if my question is stupid but I tried to fix it by myself and I just can't. When I use plot_walk_2d, I have a purple border around walkers path as well as around the solid in the image, as you can see in the following pictures.

I would like to know if it is possible to remove this purple border, because in the case of a porous media, the resulting image is really saturated.

Thanks in advance, and have a nice day.

We've had some great success with dask for simple single-machine parallelization. I think the current version uses something a bit problematic right, requiring the script to be in a main section right? Dask does not require any of this.

It is possible to assign a probability that a movement will take place based on the grey scale value of an image. This is useful for tomography data for materials with variable microporosity that is below the resolution of the image. Whether the grey scale value can always be linked to microporosity is another matter as different materials attenuate differently so composite materials may require some additional segmentation. This concept is proven in the literature and would be useful to implement. Fiedler et. al. have published a few papers on this
https://www.sciencedirect.com/science/article/pii/S0167577X08008446

When I run the code below on a 2D image, the resulting msd vs nsteps lines drop below the straight line at high nsteps. I suspect this is because the code doesn't handle steps in the z direction properly, but am not sure. Like, perhaps steps up and down are added to the step count, but not to the msd? The tortuosity it calculates is also unreasonably high.

import porespy as ps
import pytrax as px
im = ps.generators.blobs(shape=[600, 600])
im= ps.filters.fill_blind_pores(im)
rw = px.RandomWalk(im)
rw.run(nt=100000, nw=1000, stride=10, num_proc=1)
rw.plot_walk_2d()
rw.calc_msd()
rw.plot_msd()

Hey @TomTranter,

I'm trying to run pytrax in parallel. I already put my script body inside a block:

if __name__ == "__main__":
    # Body

However, I don't get significant speed up when changing the num_proc argument. The image I'm running the simulation on is roughly 200^3 voxels and I use 100,000 walkers and 1,000 time steps. Here are the run times for num_proc = [1, 2, 4, 8] (the machine has 8 physical cores):

Elapsed time in seconds: 33.01
Elapsed time in seconds: 33.29
Elapsed time in seconds: 27.83
Elapsed time in seconds: 25.13

This type of behavior can be used to simulate NMR data
https://www.sciencedirect.com/science/article/pii/S1090780707001644
Would be relatively simple to implement but would mess up the MSD calculations so needs some thought might need new metrics

A default of 0 is probably the best option so that it will always work on every machine, not depending on on parallel python and so on.

It should probably detect a 2D image and adjust the plot accordingly.