4.7 Article

Stochastic modeling of coal fracture network by direct use of micro-computed tomography images

期刊

INTERNATIONAL JOURNAL OF COAL GEOLOGY
卷 179, 期 -, 页码 153-163

出版社

ELSEVIER SCIENCE BV
DOI: 10.1016/j.coal.2017.06.002

关键词

Coal bed methane; Unconditional reconstruction; Stochastic modeling; DFN; X-ray micro-CT

向作者/读者索取更多资源

Characterization of coalbed methane reservoirs is a challenging task because of complex petrophysical properties of coal. Coal cleat system has a key role in permeability of gas through coalbed. Previous computational methods for characterization and modeling in coal formations do not account for the actual complexities in cleat systems as they commonly rely on simple statistical properties for describing the fractures. In this study, unlike the previous methods that try to extract some of the spatial statistical properties, the 2D/3D micro computed tomography images are used directly without any simplifications and assumptions. The generated models are compared to discrete fracture networks as one of the widely-used method for the modeling of such complex systems of coal cleats. Results show that the utilized algorithm produces visually satisfactory realizations of both coal matrix and cleat system. To quantify such similarities, autocorrelation functions, connectivity (with two distinct indices), average fracture length and orientation are computed. Moreover, permeabilities and porosities of the reconstructed samples are calculated and compared with the original sample. It is demonstrated that the proposed reconstruction method reproduces samples with similar statistical and petrophysical properties, but with different patterns of both coal porous region and fracture system. Finally, the proposed method and the DFN realizations are also compared extensively. The results of this study can be used for characterization of coal samples with any degree of complexity and heterogeneity by producing several realistic stochastic models. Consequently, petrophysical properties and their corresponding uncertainties can be evaluated more accurately.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.7
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

Article Computer Science, Interdisciplinary Applications

Improving remote sensing classification: A deep-learning-assisted model

Tsimur Davydzenka, Pejman Tahmasebi, Mark Carroll

Summary: In many industries and applications, obtaining and classifying remote sensing imagery is crucial. This study explores a solution of using a stochastic method to generate variations of training images, thereby improving the accuracy of machine learning classification. By increasing the training set with additional realizations, consistent improvements in classification accuracy can be achieved, providing an opportunity to enhance prediction accuracy when sufficient data is not available.

COMPUTERS & GEOSCIENCES (2022)

Article Entomology

Low radiodensity μCT scans to reveal detailed morphology of the termite leg and its subgenual organ

Travers M. Sansom, Sebastian Oberst, Adrian Richter, Joseph C. S. Lai, Mohammad Saadatfar, Manuela Nowotny, Theodore A. Evans

Summary: Termites rely on their subgenual organs (SGOs) in their legs to sense tiny vibrations. However, little is known about the structure and properties of these SGOs. This study used high-resolution micro-computed tomography (mu CT) and staining techniques to examine the SGOs of Australian termites. The results provide insights into the morphology and coverage of the SGOs, highlighting their potential role in amplifying vibrations.

ARTHROPOD STRUCTURE & DEVELOPMENT (2022)

Article Geochemistry & Geophysics

Dependence of electrical conduction on pore structure in reservoir rocks from the Beibuwan and Pearl River Mouth Basins: A theoretical and experimental study

Xiaojun Chen, Luong Duy Thanh, Chengfei Luo, Pejman Tahmasebi, Jianchao Cai

Summary: The relationship between electrical conduction and pore structure in reservoir rocks was analyzed through theoretical development, petrophysical experiments, error analysis, core-scale displacement experiments, and pore-scale numerical simulations. The electric formation factor was found to be a function of porosity, tortuosity fractal dimension, and pore fractal dimension. The model provided satisfactory predictions for reservoir rocks when the ratio of minimum to maximum pore radius was suitable. Porosity-based formation factor models had high errors at high formation factors, but our model improved predictions with an error factor of +/- 10. Hydraulic and electrical conductions showed different dependencies on pore structure, with hydraulic conductance being influenced by pore size, dominant flow channels, and threshold pressure, while electrical conduction had no dominant channel and did not reflect pore size information at the same porosity.

GEOPHYSICS (2023)

Article Green & Sustainable Science & Technology

Insights into wettability alteration during cyclic scCO2-brine injections in a layered Bentheimer sandstone

A. L. Herring, C. Sun, R. T. Armstrong, M. Saadatfar

Summary: Residual trapping is essential for the security and sustainability of geologic sequestration operations. Recent experiments indicate that cycles of scCO2 and brine injections can cause surface chemistry reactions, enhancing residual trapping. This study uses X-ray microcomputed tomography to investigate the alteration mechanism and provides new insights into the conditions under which wettability alteration affects scCO2 flow and trapping.

INTERNATIONAL JOURNAL OF GREENHOUSE GAS CONTROL (2023)

Article Engineering, Civil

NMR-data-driven prediction of matrix permeability in sandstone aquifers

Xiaojun Chen, Xiaobo Zhao, Pejman Tahmasebi, Chengfei Luo, Jianchao Cai

Summary: A data-driven model based on nuclear magnetic resonance (NMR) was developed to predict matrix permeability using nine machine learning models. The type of input data had a strong influence on the machine learning modeling. By using cumulative T2 relaxation data instead of the original T2 data, the gradient boosting decision tree model tuned by GridsearchCV showed a stronger agreement between experimental results and NMR estimates of matrix permeability. The correlation coefficient reached 0.92 with the lowest MSE of 0.12. Evaluation: 9 out of 10.

JOURNAL OF HYDROLOGY (2023)

Article Engineering, Civil

Graph neural network for groundwater level forecasting

Tao Bai, Pejman Tahmasebi

Summary: In this study, a graph neural network (GNN) is used for accurate prediction of groundwater dynamics. The model incorporates spatial relationships between wells using graph convolution layers and temporal features using gated temporal convolutional networks. The proposed model outperforms two baseline models in terms of evaluation metrics, even when the spatial dependencies are unknown.

JOURNAL OF HYDROLOGY (2023)

Article Chemistry, Physical

Mixing properties of bi-disperse ellipsoid assemblies: mean-field behaviour in a granular matter experiment

F. M. Schaller, H. Punzmann, G. E. Schroeder-Turk, M. Saadatfar

Summary: This paper explains the observations of a study on the X-ray computed tomography of granular ellipsoidal packings using a fundamental theoretical relationship for mixture distributions. The main finding is that the bi-disperse ellipsoidal packings studied can be interpreted as a mixture of two uncorrelated mono-disperse packings, insensitive to the compaction protocol.

SOFT MATTER (2023)

Article Engineering, Chemical

Effect of Wettability on Two-Phase Flow Through Granular Porous Media: Fluid Rupture and Mechanics of the Media

Mehryar Amir Hosseini, Serveh Kamrava, Muhammad Sahimi, Pejman Tahmasebi

Summary: The wettability of porous media significantly impacts the spatial distribution of fluid phases. Computer simulations show that contact angle affects particle dynamics, fluid velocity, and rupture in the pore space. Additionally, increasing contact angle reduces inter-particle interactions and increases drag force, leading to larger particle displacement.

CHEMICAL ENGINEERING SCIENCE (2023)

Article Mechanics

Drafting, Kissing and Tumbling Process of Two Particles: The Effect of Morphology

Xiaoming Zhang, Pejman Tahmasebi

Summary: In this study, the DKT process of irregular particles was investigated numerically, and it was found that the particle shape plays an important role. Particles with low sphericity are more sensitive to orientation, and lower roundness accelerates the separation of particle pairs. The vertical velocity decreases when roundness or sphericity is smaller.

INTERNATIONAL JOURNAL OF MULTIPHASE FLOW (2023)

Article Computer Science, Interdisciplinary Applications

On the influence of the natural shape of particles in multiphase fluid systems: Granular collapses

Mehryar Amir Hosseini, Pejman Tahmasebi

Summary: This paper investigates the influence of particle morphology on granular collapse behavior and wave generation in multiphase fluid systems. The study establishes a clear relationship between particle morphology and important characteristics such as displacement, velocity, inter-particle forces, and kinetic energy. The findings demonstrate that as irregularity increases, interlocking between particles becomes more prominent, leading to reduced particle travel distances. Additionally, interlocking also influences particle-fluid interactions, resulting in significant alterations in the formation of generated waves.

COMPUTERS AND GEOTECHNICS (2023)

Article Geochemistry & Geophysics

Explainable machine learning for labquake prediction using catalog-driven features

Sadegh Karimpouli, Danu Caus, Harsh Grover, Patricia Martinez-Garzon, Marco Bohnhoff, Gregory C. Beroza, Georg Dresen, Thomas Goebel, Tobias Weigel, Grzegorz Kwiatek

Summary: This study pioneers in time to failure (TTF) prediction based on machine learning using acoustic emission (AE) records from laboratory stick-slip experiments. The regression voting ensemble of Long-Short Term Memory (LSTM) networks predicts TTF with a coefficient of determination (R2) of 70% on the test dataset. Feature importance analysis reveals that AE rate, correlation integral, event proximity, and focal mechanism-based features are the most important features for TTF prediction.

EARTH AND PLANETARY SCIENCE LETTERS (2023)

Article Thermodynamics

Modeling the physical properties of hydrate-bearing sediments: Considering the effects of occurrence patterns

Yuqi Wu, Pejman Tahmasebi, Keyu Liu, Chengyan Lin, Serveh Kamrava, Shengbiao Liu, Samuel Fagbemi, Chang Liu, Rukuai Chai, Senyou An

Summary: This study proposes a novel hybrid modeling approach integrating X-ray CT imaging technology, morphological operation algorithm, and quartet structure generation set method to investigate the dependence of the physical properties of hydrate-bearing sediments (HBS) on hydrate occurrence patterns and saturation levels. The findings suggest that different hydrate types have varying heterogeneity in the distribution of pore and throat radii.

ENERGY (2023)

Review Energy & Fuels

Minireview on Porous Media and Microstructure Reconstruction Using Machine Learning Techniques: Recent Advances and Outlook

Hossein Mirzaee, Serveh Kamrava, Pejman Tahmasebi

Summary: This article reviews the most promising studies in machine learning-assisted reconstruction of porous media, categorizing the approaches and discussing their characteristics, advantages, and disadvantages. It also provides information on various methods for evaluating algorithm performance. Furthermore, the article explores the current research status and challenges in ML-assisted porous media reconstruction in energy-related applications and suggests potential areas for future studies.

ENERGY & FUELS (2023)

Review Materials Science, Multidisciplinary

A state-of-the-art review of experimental and computational studies of granular materials: Properties, advances, challenges, and future directions

Pejman Tahmasebi

Summary: Modeling of heterogeneous materials and media plays a crucial role in various phenomena and systems, including condensed matter physics, soft materials, composite media, porous media, biological systems, geosystems, ceramic engineering, pharmaceutical science, and space discoveries. This review paper examines recent developments in experimental and computational methods, such as neutron and nanometer-scale tomography, magnetic resonance imaging, digital image correlation, and 4D techniques. It also explores the shift towards micro-scale and the development of multiscale approaches in modeling, as well as the exploration of coupled or multiphysics systems.

PROGRESS IN MATERIALS SCIENCE (2023)

Article Geochemistry & Geophysics

Applicability of 2D algorithms for 3D characterization in digital rocks physics: an example of a machine learning-based super resolution image generation

Sadegh Karimpouli, Rail Kadyrov, Mirko Siegert, Erik Hans Saenger

Summary: Digital rock physics involves imaging, segmentation, and numerical computations of rock samples. Traditional 2D algorithms face challenges in handling the third dimension of 3D samples. This study suggests generating 3D samples in different directions and utilizing averaging or binary combination strategies to improve the accuracy. Numerical computations of rock physical properties provide insights into the effectiveness of these models.

ACTA GEOPHYSICA (2023)

暂无数据