InterPore2021

Tortuosity and permeability of random porous medium using deep learning

4 Jun 2021, 09:40

1h

Poster (+) Presentation (MS15) Machine Learning and Big Data in Porous Media Poster +

Speakers

Dr Maciej Matyka (Faculty of Physics and Astronomy, University of Wrocław)Dr Krzysztof Graczyk (Faculty of Physics and Astronomy, University of Wrocław)

Description

We will present our recent results on porosity (P), permeability (K), and tortuosity (T) of artificial, randomly generated porous medium predicted directly from the geometry images [1]. We will show that convolutional neural networks (CNN) can predict porosity, permeability, and tortuosity based only on the obstacles' picture. The CNN is trained on artificial data samples, for which the permeability and tortuosity are obtained within the Lattice-Boltzmann method. The CNN predicts permeability and tortuosity with about 6% accuracy.

[1] Graczyk, K. M., and Matyka, M., Predicting Porosity, Permeability, and Tortuosity of Porous Media from Images by Deep Learning, Sci Rep 10, 21488 (2020)

References

Graczyk, K. M., and Matyka, M., Predicting Porosity, Permeability, and Tortuosity of Porous Media from Images by Deep Learning, Sci Rep 10, 21488 (2020)

Presentation materials

Poster_Matyka_Graczyk_Deep_Learning_2021.pdf

Video presentation - Wistia

Video presentation - YouTube