Intelligent Evaluation Method for Sandstone Pores Based on Multi-Source Data Fusion from SEM
摘要
砂岩储层孔隙类型多样,孔隙结构的精细表征与类型划分是理解成岩改造与物性演化的关键.大视域高分辨率拼接扫描电镜成像(MAPS)可识别纳米~毫米尺度孔隙,但难以区分原子序数接近的矿物;扫描电镜矿物定量分析(Quantitative Evaluation of Minerals by Scanning Electron Microscopy,QEMSCAN)可准确识别矿物类型及其空间分布,但分辨率相对较低,难以准确识别微孔隙.本文以砂岩样品为例,提出一种MAPS 与 QEMSCAN 跨尺度融合的孔隙智能评价方法.首先,获取样品端面低分辨率 QEMSCAN 矿物嵌布图像;随后,获取同一端面大视域高分辨率MAPS图像,采用跨尺度配准与引导滤波上采样实现两类图像的像素级对齐与分辨率统一;最后,采用机器学习方法从MAPS图像中识别孔隙空间,并融合配准后的矿物信息,根据孔隙邻域矿物组成与接触关系划分孔隙类型,统计特征参数.结果表明:该方法可有效区分粒间孔、长石溶蚀孔、石英铸模孔及黏土晶间孔等典型孔隙类型;不同矿物发育的孔隙在空间分布与形态上存在差异,体现了压实与成岩作用对孔隙发育的改造;对比分析显示,Berea 砂岩以粒间孔为主、连通性较好,而致密砂岩粒间孔占比降低、粒内孔贡献增加,孔隙更趋扁平,孔径显著减小.本方法评价结果与覆压孔隙度、压汞和核磁共振测试结果具有较好对应关系,为孔隙类型的精细识别与定量统计提供了可行途径,并为高精度三维数字岩心建模与多尺度岩石物理分析奠定数据基础.
Sandstone reservoirs exhibit diverse pore types,and detailed characterization and classification of pore structures are key to understanding diagenetic modification and petrophysical evolution.Large-area,high-resolution stitched images acquired by mosaic acquisition by scanning electron microscopy(MAPS)can identify pores over scales from nanometers to millimeters,but it cannot distinguish minerals with similar atomic numbers.In contrast,quantitative evaluation of minerals by scanning electron microscopy(QEMSCAN)can accurately identify mineral types and their spatial distributions,but its low-resolution limits reliable characterization of micropores.Using sandstone samples as case studies,we propose an intelligent pore evaluation method based on cross-scale fusion of MAPS and QEMSCAN images.First,low-resolution QEMSCAN mineral distribution images are acquired from the sample end face.Large-area,high-resolution MAPS images of the same end face are then obtained,and cross-scale registration together with guided-filter upsampling is used to achieve pixel-level alignment and resolution harmonization between the two datasets.Finally,machine learning is used to identify pore space from the MAPS images,and the registered mineral information is integrated to classify pore types and quantify characteristic parameters according to the mineral composition and contact relationships in the pore neighborhood.The results show that the proposed method can effectively distinguish typical pore types,including intergranular pores,feldspar dissolution pores,quartz moldic pores,and clay intercrystalline pores.Pores associated with different minerals differ in spatial distribution and morphology,reflecting the modification of pore development by compaction and diagenesis.Comparative analysis further indicates that Berea sandstone is dominated by intergranular pores and exhibits relatively good connectivity,whereas tight sandstone shows a reduced proportion of intergranular pores,an increased contribution from intragranular pores,flatter pore shapes,and significantly smaller pore sizes.The evaluation results obtained by this method are in good agreement with porosity under confining pressure,mercury intrusion measurements,and NMR measurements.This workflow provides a feasible route for refined pore-type identification and quantitative statistical analysis,and provides a data foundation for high-fidelity 3D digital rock modeling and multiscale rock physics analysis.