Driving digital rock towards machine learning: Predicting permeability with gradient boosting and deep neural networks

    Research output: Contribution to journalArticlepeer-review

    102 Citations (Scopus)

    Abstract

    We present a research study aimed at testing of applicability of machine learn-ing techniques for permeability prediction. We prepare a training set containing. 3D scans of Berea sandstone subsamples imaged with X-ray microtomography and corresponding permeability values simulated with Pore Network approach. We also use Minkowski functionals and Deep Learning-based descriptors of 3D images and 2D slices as input features for predictive model training and pre-diction. We compare predictive power of various descriptors and methods. The; latter include Gradient Boosting, Deep Neural Networks (DNN) and Convo-lutional Neural Networks (CNN). Introduced Deep Learning-based descriptors; outperform previously used alternatives. 3D CNN outperforms the competitors in terms of the percent error and prediction time. The results demonstrate the applicability of machine learning for image-based permeability prediction and open a new area of Digital Rock research.

    Original languageEnglish
    Pages (from-to)91-98
    Number of pages8
    JournalComputers and Geosciences
    Volume127
    DOIs
    Publication statusPublished - Jun 2019

    Keywords

    • Artificial neural networks
    • Digital rock
    • Gradient boosting
    • Machine learning
    • Permeability prediction

    Fingerprint

    Dive into the research topics of 'Driving digital rock towards machine learning: Predicting permeability with gradient boosting and deep neural networks'. Together they form a unique fingerprint.

    Cite this