Deep Learning Convolutional Neural Networks to Predict Porous Media Properties

Abstract

Digital rocks obtained from high-resolution micro-computed tomography (micro-CT) imaging has quickly emerged as a powerful tool for studying pore-scale transport phenomena in petroleum engineering. In such frameworks, digital rock analysis usually carries the problematic aspect of segmenting greyscale images into different phases for quantifying many physical properties. Fine pore structures, such as small rock fissures, are usually lost during segmentation. In addition, user bias in this process can lead to significantly different results. An alternative approach based on deep learning is proposed. Convolutional Neural Networks (CNN) are utilized to rapidly predict several porous media properties from 2D greyscale micro-computed tomography images in a supervised learning frame. A dataset of greyscale micro-CT images of three different sandstones species is prepared for this study. The image dataset is segmented, and pore networks are extracted to compute porosity, coordination number, and average pore size for training and validating our model predictions. The greyscale images (input) and the computed properties (output) are uploaded to a deep neural network for training and validation in an end-to-end regression scheme. Overall, our model estimates porosity, coordination number, and average pore size with an average error of 0.05, 0.17, and 1.8μm, respectively. Training wall-time and prediction error analysis are also discussed. This is a first step to use artificial intelligence and machine learning methods for the robust prediction of porous media properties from unprocessed image-driven data.

Upscaling the porosity–permeability relationship of a microporous carbonate for Darcy-scale flow with machine learning

“…Over the past decade numerous open-source computer vision tools have enabled in-depth analysis and vectorization of these immense datasets. Several recent studies have used convolutional neural networks to predict permeability from micro-CT images at a single scale 37 , 38 . However, these deep learning models are highly computationally expensive and are black boxes, providing little insight into the underlying structural features that ultimately control the permeability and making it difficult to apply them to other rock types or upscale effectively.…”

A Comprehensive Review for Breast Histopathology Image Analysis Using Classical and Deep Neural Networks

“…In addition, this review discussed the deep ANNs method not only can be applied in the field of breast histopathological image analysis, but also in the field of other closed microscopic image analysis, such as: Cervical histopathological analysis [167], [168], [169], cervical cytopathological analysis [170], [171], [172], stem cell analysis [173], [174], microbiological image analysis [175], [176], [177], sperm quality analysis [178], [179], [178], web-based platform for computer assisted diagnosis [180], [181], and rock microstructural analysis [182], [183]. No matter from the aspects of image pre-processing, feature extraction and selection, segmentation, and classification, or from the aspects of deep ANN model design and proposed framework idea, the methods of deep ANN summarized in this review can bring a new perspective to the research in other fields.…”

An Application of Deep Neural Networks for Segmentation of Microtomographic Images of Rock Samples

“…In [24], end-to-end CNN for prediction of pores media characteristics from images is described. Three samples of different sandstones are considered.…”