Review of Data Science Trends and Issues in Porous Media Research With a Focus on Image‐Based Techniques. Issue 10 (26th October 2021)

• Record Type: Journal Article
• Authors: Rabbani, A.; Fernando, A. M.; Shams, R.; Singh, A.; Mostaghimi, P.; Babaei, M.
• Is Part Of: Water resources research. Volume 57:Issue 10(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-10-26
• Subjects: data science -- porous materials -- machine learning -- neural networks; Hydrology -- Periodicals; 333.91
• DOI: 10.1029/2020WR029472
• Languages: English
• ISSNs: 0043-1397

Abstract

Abstract: Data science as a flourishing interdisciplinary domain of computer and mathematical sciences is playing an important role in guiding the porous material research streams. In the present narrative review, we have examined recent trends and issues in data‐driven methods used in the image‐based porous material research studies relevant to water resources researchers and scientists. Initially, the recent trends in porous material data‐related issues have been investigated through search engine queries in terms of data source, data storage hub, programing languages, and software packages. Subsequent to a diligent analysis of the existing trends, a review of the common concepts of porous material research and data science are presented through six categories comprising big data, data regression, classification, image segmentation, geometry reconstruction, and image data resolution. We provide: (a) a focus on image‐based and pore scale methods which has not been presented previously, (b) a detailed search engine research for trend investigation, and (c) practical examples and comparison of data storage in porous media image‐based research. By reading this review article, an overall image of the active and popular interdisciplinary research domains can be obtained. Readers will also be informed of the latest data‐driven efforts and recommended research directions for tackling the image‐based porous material problems relevant to water resources research. We concluded that porous material image reconstruction and resolution improvement techniques are unique means to reveal unprecedented details of micro‐structures that may have been missed in a medium quality tomography image. Key Points: The porous material data‐science concepts are portrayed and categorized into six subdomains Image‐based deep learning is becoming a unique tool to characterize, simulate, and classify porous materials Machine learning in porous material research studies is still dominated by MATLAB, however, Python developments are becoming a trend