Visual characterization of displacement processes in porous media

This dataset correlates to the submitted article to IEEE VIS 2023, entitled “Visual Analysis of Displacement Processes in Porous Media using Spatio-Temporal Flow Graphs”, by Straub et al. 2023. More specifically, this data set is the one used to create the graphs shown in all Figures, except Figure 2, of the article. In this work, 11 experiments were carried out in a Poly-Di-Methyl-Siloxane (PDMS) micromodel for a variety of capillary numbers and viscosity ratios between the phases involved. More specifically, the used viscosity ratios between the wetting and the non-wetting phase were 0.2, 1, 10. The corresponding capillary numbers (on log scale) were -2, -3, -4, -5. Two additional experiments carried out in two micromodels having the same distribution of grains but different grain shape, and another one with a pore structure based on Delaunay triangulation. The objective of the work was to establish a visualization workflow which can facilitate the extraction of information in a generic way, independently of the boundary flow conditions, or the physical properties of the fluids, and the geometry of the pore space. The micromodels used, which served as the porous medium, were of four types, as mentioned before. The micromodels were produced by following the standard optical and soft lithography techniques, meaning that a silicon wafer was prepared with the features of each flow network on it, and then the PDMS elastomer base was mixed with the curing agent at a ratio of 1:10, degassed, poured on top of the wafer, degassed, thermally cured, and then bonded with corona treatment. The dimensions of the pore space of two of the models were 4 mm × 6 mm in the planar view, and a depth of 115 μm, constant throughout the entire pore space. The pore network was based on a statistical distribution of points in the pore space. One design had 76 non-overlapping circular grains in planar view (Network 1), with a distribution for the radius, and the other one had as grains the inscribed to the circles octagons (Network 2). The grain diameter in both cases ranged from 275 to 575 μm, with a mean size of 380 μm. The third type of micromodels (Network 3) had a pore network with overall dimensions of 15 mm × 20 mm and a depth of 100 μm. The pore network was again based on a statistical distribution of points in the pore space, with 85 cylindrical grains with a mean pore size of 420 μm, and a mean grain diameter of 1.63 mm. Finally, the last network (Network 4) was based on 5000 points generated by Delaunay triangulation, and their connections. The total dimensions of the network were 5 mm x 30 mm, but we could visualize an area of 5 mm x 6 mm, and this is what we account for as the domain size. The mean pore size was 50 μm, as well as the depth of the flow network. The microfluidic chips were not treated in any way to tune or alter their surface electrostatic properties, and in their natural state they are hydrophobic. They were treated as such in the experiments too, with the two fluid phases involved being water dyed with a water-based ink, to create the contrast for visual evaluation of the flow, as the non-wetting phase, and Fluorinert FC-43, which served as the wetting phase. For the tuning of the viscosity of the non-wetting phase, 99.5% glycerol was used to achieve the desired viscosity. The addition of the ink in water, and of glycerol when applicable, did not alter its physical properties, meaning that the density and the viscosity of the final mixture was practically the same as those of pure water at the same temperature (20 degrees Celsius). During each experiment there was a fixed value for the volumetric flux of the non-wetting phase, corresponding to a fixed capillary number. The wetting phase was already present in the pore space, so drainage scenarios were realized. The introduction of the non-wetting phase in the pore space was done with one neMESYS mid pressure syringe pump 1000N. The control of the syringe pump was done with QmixElements© and a personal computer. During the displacement events, pictures from the pore space were recorded with the use of StreamPix 8.0©. The acquisition frame rate would vary between 1 and 15 frames per second, depending on the speed of the process. The data files are the pictures taken during each event, and are structured accordingly based on the pore network, the capillary number, the viscosity ratio, and frames per second.本数据集关联于已提交至IEEE VIS 2023的论文，标题为《使用时空流图可视化分析多孔介质内驱替过程》（Visual Analysis of Displacement Processes in Porous Media using Spatio-Temporal Flow Graphs），作者为Straub等，2023年。更具体而言，本数据集为该论文中除图2外所有配图所用的原始数据。 本研究采用聚二甲基硅氧烷（Poly-Di-Methyl-Siloxane, PDMS）微流控芯片开展了11组实验，覆盖了不同毛细管数与相间粘度比条件。具体而言，润湿相与非润湿相的粘度比设置为0.2、1、10；对应的对数刻度毛细管数分别为-2、-3、-4、-5。此外，本研究还开展了2组额外实验：两组采用晶粒分布一致但晶粒形状不同的微流控芯片，以及1组基于德劳内三角剖分（Delaunay triangulation）孔隙结构的实验。 本研究的核心目标是构建一套通用化可视化工作流，可脱离边界流动条件、流体物理属性以及孔隙空间几何形态的限制，实现孔隙内流动信息的高效提取。 本次实验所用的作为多孔介质的微流控芯片共分为四类，详述如下： 1. 网络1：平面视图下孔隙空间尺寸为4 mm × 6 mm，孔隙深度恒定为115 μm。孔隙网络基于孔隙空间内点的统计分布构建，平面视图内包含76个非重叠圆形晶粒，晶粒半径服从特定分布。 2. 网络2：与网络1采用相同的孔隙空间尺寸与深度，但其晶粒为内接于圆形的八边形。两类网络的晶粒直径范围均为275 μm至575 μm，平均粒径为380 μm。 3. 网络3：孔隙网络整体尺寸为15 mm × 20 mm，深度为100 μm。孔隙网络同样基于孔隙空间内点的统计分布构建，包含85个圆柱形晶粒，平均孔径为420 μm，平均晶粒直径为1.63 mm。 4. 网络4：基于5000个通过德劳内三角剖分生成的点及其连接关系构建。该网络整体尺寸为5 mm × 30 mm，但实际可视化区域为5 mm × 6 mm，此即为本次实验的域尺寸。其平均孔径与流道深度均为50 μm。 微流控芯片采用标准光学与软光刻技术制备：首先制备带有各流道网络结构的硅片，随后将聚二甲基硅氧烷（PDMS）基体与固化剂按1:10的比例混合，脱气后浇注于硅片之上，再次脱气后进行热固化，最后通过电晕处理实现芯片键合。 所有微流控芯片均未经过任何表面静电属性调控处理，天然状态下为疏水表面，实验中也保持了这一特性。实验所用的两相流体分别为：以水基墨水染色的水作为非润湿相，用于实现流动可视化的对比度；以及氟化液Fluorinert FC-43作为润湿相。为调控非润湿相的粘度，我们使用99.5%的甘油来达到目标粘度。向水中添加墨水，以及在需要时添加甘油，均不会改变其物理属性，即最终混合液的密度与粘度与20℃下的纯水基本一致。 每组实验中，非润湿相的体积流量固定，对应固定的毛细管数。由于孔隙空间内预先充满了润湿相，因此实验实现了驱替场景。非润湿相通过一台neMESYS mid pressure syringe pump 1000N注射泵注入孔隙空间。注射泵的控制通过QmixElements©软件与个人计算机完成。在驱替过程中，使用StreamPix 8.0©软件记录孔隙空间的图像，采集帧率介于1至15帧每秒之间，具体取决于过程的速度。 本次提供的数据文件为每组实验过程中拍摄的图像，其存储与命名结构基于孔隙网络类型、毛细管数、粘度比以及采集帧率进行组织。