3D Collection of Binary Images

• https://www.sciencedirect.com/science/article/pii/S0309170819311145
• https://doi.org/10.1016/j.advwatres.2020.103539
• Download

Abstract — We present the PoreFlow-Net, a 3D convolutional neural network architecture that provides fast and accurate fluid flow predictions for 3D digital rock images. We trained our network to extract spatial relationships between the porous medium morphology and the fluid velocity field. Our workflow computes simple geometrical information from 3D binary images to train a deep neural network (the PoreFlow-Net) optimized to generalize the problem of flow through porous materials. Our results show that the extracted information is sufficient to obtain accurate flow field predictions in less than a second, without performing expensive numerical simulations providing a speed-up of several orders of magnitude. We also demonstrate that our model, trained with simple synthetic geometries, is able to provide accurate results in real samples spanning granular rocks, carbonates, and slightly consolidated media from a variety of subsurface formations, which highlights the ability of the model to generalize the porous media flow problem. The workflow presented here shows the successful application of a disruptive technology (physics-based training of machine learning models) to the digital rock physics community.

• https://link.springer.com/article/10.1007/s11242-021-01617-y
• https://doi.org/10.1007/s11242-021-01617-y
• Download

Abstract — The permeability of complex porous materials is of interest to many engineering disciplines. This quantity can be obtained via direct flow simulation, which provides the most accurate results, but is very computationally expensive. In particular, the simulation convergence time scales poorly as the simulation domains become less porous or more heterogeneous. Semi-analytical models that rely on averaged structural properties (i.e., porosity and tortuosity) have been proposed, but these features only partly summarize the domain, resulting in limited applicability. On the other hand, data-driven machine learning approaches have shown great promise for building more general models by virtue of accounting for the spatial arrangement of the domains’ solid boundaries. However, prior approaches building on the convolutional neural network (ConvNet) literature concerning 2D image recognition problems do not scale well to the large 3D domains required to obtain a representative elementary volume (REV). As such, most prior work focused on homogeneous samples, where a small REV entails that the global nature of fluid flow could be mostly neglected, and accordingly, the memory bottleneck of addressing 3D domains with ConvNets was side-stepped. Therefore, important geometries such as fractures and vuggy domains could not be modeled properly. In this work, we address this limitation with a general multiscale deep learning model that is able to learn from porous media simulation data. By using a coupled set of neural networks that view the domain on different scales, we enable the evaluation of large (>5123) images in approximately one second on a single graphics processing unit. This model architecture opens up the possibility of modeling domain sizes that would not be feasible using traditional direct simulation tools on a desktop computer. We validate our method with a laminar fluid flow case using vuggy samples and fractures. As a result of viewing the entire domain at once, our model is able to perform accurate prediction on domains exhibiting a large degree of heterogeneity. We expect the methodology to be applicable to many other transport problems where complex geometries play a central role.

• https://onepetro.org/SJ/article/15/01/125/192097/Investigating-Matrix-Fracture-Transfer-via-a-Level
• https://doi.org/10.2118/116110-PA

Abstract — Multiphase flow and transport phenomena within fractures are important because fractures often represent primary flow conduits in otherwise low-permeability rock. Flows within the fracture, between the fracture and the adjacent matrix, and through the pore space within the matrix typically happen on different length and time scales. Capturing these scales experimentally is difficult. It is, therefore, useful to have a computational tool that establishes the exact position and shape of fluid/fluid interfaces in realistic fracture geometries. The level set method (LSM) is such a tool. Our progressive quasistatic (PQS) algorithm based on the level set method finds detailed, pore-level fluid configurations satisfying the Young-Laplace equation at a series of prescribed capillary pressures. The fluid volumes, contact areas, and interface curvatures are readily extracted from the configurations. The method automatically handles topological changes of the fluid volumes as capillary pressure varies. It also accommodates arbitrarily complicated shapes of confining solid surfaces. Here, we apply the PQS method to analytically defined fracture faces and aperture distributions, to geometries of fractures obtained from high-resolution images of real rocks, and to idealized fractures connected to a porous matrix. We also explicitly model a fracture filled with proppant, using a cooperative rearrangement algorithm to construct the proppant bed and the surrounding matrix. We focus on interface movement between matrix and fracture, and snap-off of nonwetting phase into the fracture during imbibition in particular. The extent to which nonwetting phase is trapped in fracture/enclosed gaps is very sensitive to the direction of the displacement. Simulated drainage curves in matrix differ systematically from drainage curves in fracture and matrix with transfer between them. In a reservoir simulation, the latter might serve as an upscaled drainage curve input for a fractured medium.

• https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020WR027943
• https://doi.org/10.1029/2020WR027943

Abstract — Fractures are conduits for fluid flow in low-permeability geological formations. Multiphase flow properties of fractures are important in natural processes and in engineering applications such as the evaluation of the sealing capacity of caprocks and productivity of hydrocarbon-bearing tight rocks. Investigations of flow and transport through fractures typically focus on the effects of fracture geometric and mechanical factors such as aperture, roughness, and compressibility. The wettability of the fracture surfaces and its influence on microscale interfacial phenomena and macroscale effective transport properties are seldom studied. Here, we investigated the effect of heterogeneous wetting properties on the displacement of water by supercritical CO2 through a series of lattice Boltzmann method simulations. The results show the evolution of the CO2 plume within a fracture is controlled by both the roughness of the aperture field and the wetting distribution. We combined these factors into a capillary pressure map that can be related to the macroscopic flow behavior of the fracture. We observed that heterogeneous wetting distributions promote the residual trapping of water where lower capillary pressures allowed for isolated water pockets in higher capillary pressure zones. Analysis of fracture unsteady relative permeability shows the effect of wetting on permeability evolution and provides support for the viscous-coupling relative permeability model. Finally, analysis of the steady-state relative permeability and saturation demonstrates a strong correlation between permeability and the standard deviation of the capillary pressure field. Thus, characterizing the distribution of wetting properties of fractures is crucial to understanding multiphase fracture flow and transport properties.