In conventional Pore-network Modeling (PNM) approaches, fluid flow and transport are solved in a network of pore elements with idealized geometries. Such simplification can lead to inaccurate predictions when the original pore space features complex geometries. To overcome this limitation, this study introduces a novel workflow that integrates four key components: (i) an enhanced pore network extraction (PNE) platform capable of identifying and extracting pore cross-sections from high-resolution micro-computed tomography (micro-CT) images of the pore space, (ii) a computationally-efficient semi-analytical model that can faithfully predict capillary entry pressure and the corresponding fluid configuration of piston-like displacements using real two-dimensional cross-sections of pores, (iii) a PNM approach for two-phase flow modeling that utilizes the capillary entry pressure of pores predicted by the semi-analytical model, and (iv) an Artificial Intelligence (AI)-driven model that paves the way for future advancements in efficiently predicating fluid displacement properties in intricate pore structures. To validate this new workflow, we constructed various pore networks containing real pore and throat cross-sections over a diverse group of sandstone and carbonate rock samples. Subsequently, we simulate capillary pressure curves of Mercury Intrusion Capillary Pressure (MICP) and oil–water primary drainage displacements in Bentheimer and Berea sandstones, respectively, using both the conventional and enhanced PNM approaches. The latter demonstrated improved prediction accuracy compared to conventional methods. Next, primary drainage simulations are conducted for two carbonates, and the resulting capillary pressure curves from both PNM approaches are compared. In addition, we conduct an in-depth analysis of fourteen geometric features of the pore space, identifying key factors of hydraulic radius, circumradius, sphericity, and area, that significantly impact capillary entry pressure of pores. After that, we construct an Artificial Neural Network (ANN) to predict the capillary entry pressure of pores using their critical geometric features. This AI model, trained using data derived from the semi-analytical model, exhibits excellent predictive accuracy (with a R2 of 0.995 for the test data set) in estimating capillary entry pressure of pores. Overall, our newly proposed, integrated workflow represents a significant step forward in the field of digital rock technology (DRT), offering an accurate and efficient method for modeling fluid flow in rock samples with complex pore geometries.

中文翻译：

---

真实孔隙几何形状中的热力学一致界面曲率：对两相位移过程孔隙尺度建模的影响


在传统的孔隙网络建模 （PNM） 方法中，流体流动和传递是在具有理想化几何形状的孔隙单元网络中求解的。当原始孔隙空间具有复杂的几何形状时，这种简化可能会导致预测不准确。为了克服这一限制，本研究引入了一种新的工作流程，它集成了四个关键组件：（i） 一个增强的孔隙网络提取 （PNE） 平台，能够从孔隙空间的高分辨率微型计算机断层扫描 （micro-CT） 图像中识别和提取孔横截面，（ii） 一个计算高效的半分析模型，可以忠实地预测毛细管入口压力和相应的流体配置活塞状位移使用真实孔隙的二维横截面，（iii） 用于两相流建模的 PNM 方法，该方法利用半分析模型预测的孔隙的毛细管入口压力，以及 （iv） 人工智能 （AI） 驱动的模型，为未来在有效预测复杂孔隙结构中的流体位移特性的进步铺平了道路。为了验证这种新的工作流程，我们在一组不同的砂岩和碳酸盐岩样品上构建了包含真实孔隙和喉横截面的各种孔隙网络。随后，我们使用传统和增强型 PNM 方法分别模拟了 Bentheimer 和 Berea 砂岩中汞侵入毛细管压力 （MICP） 和油水初级排水位移的毛细管压力曲线。与传统方法相比，后者的预测准确性更高。接下来，对两种碳酸盐进行初级排水模拟，并比较两种 PNM 方法得到的毛细管压力曲线。 此外，我们对孔隙空间的 14 个几何特征进行了深入分析，确定了水力半径、圆周半径、球形度和面积等关键因素，这些因素对孔隙的毛细管入口压力有显着影响。之后，我们构建了一个人工神经网络 （ANN），利用孔隙的关键几何特征来预测孔隙的毛细管进入压力。该 AI 模型使用来自半分析模型的数据进行训练，在估计毛孔的毛细血管入口压力方面表现出出色的预测准确性（测试数据集的 R2 为 0.995）。总体而言，我们新提出的集成工作流程代表了数字岩石技术 （DRT） 领域向前迈出的重要一步，为模拟具有复杂孔隙几何形状的岩石样品中的流体流动提供了一种准确有效的方法。