2.2.1 HMPI testing

HMIP is one of the most commonly used experimental techniques to measure pore throat size distribution and effective porosity. Based on the mercury injection method, the lower limit of fluid flow, throat, and its connected pore volume can be quantitatively evaluated, which is helpful for understanding the law of fluid seepage [47]. An Autopore IV 9500 pore analyzer was used in this HPMI experiment. The test pressure ranges from 0.1 Psia to 33000 Psia, and the measured pore throat diameter ranges from 360 μm to 5 nm.

The pore throats are assumed to be connected capillary tubes of different diameters, and the permeability of the pore throat of radius r_j is [48, 49]

where k_j represents the permeability of the pore throat with radius r_j, and ∅_j is the ratio of the volume of the pore throat with radius r_j to the total pore volume.

The core permeability contribution rate corresponding to a certain radius of the pore throat is characterized by

2.2.2 NMR testing

NMR experiments transform the T₂ spectrum of nuclear magnetic resonance into the distribution curve of the pore radius, which can be used to evaluate the pore structure of cores and characterize the pore size distribution at multiple scales [50]. NMR experiments were carried out on the MacroMR12-150H-I core nondestructive and microcosmic testing imaging and analysis system.

2.2.3 XRD testing

XRD can obtain information on the mineral composition, crystal structure, or morphology, through the diffraction patterns of samples, so as to quickly and accurately identify the minerals of ore rocks. A representative portion of the samples (approximately 10 g) was crushed and ground into powder (particle size < 0.2 mm), and the main minerals in the sample were analyzed using a Bruker D8 Advance X-ray diffractometer. The Jade6.0 software was used to analyze the phases of the XRD patterns. After peak division, background value deduction, and smoothing treatment, combined with the geological background of the study area, the phase compositions were retrieved using standard PDF reference cards.

2.2.4 WD-XRF testing

Wavelength dispersive X-ray fluorescence spectrometer is a type of spectrometer that uses crystal spectrum. It is widely used in the determination of various elements and oxides in geological studies. The WD-XRF spectrometer Axios MAX was used for the quantitative analysis of mineral elements in solid disk samples with a diameter of 38 mm and a thickness of 10 mm.

2.2.5 SEM-EDS experiment

SEM-EDS can automatically identify the mineralogy of each image pixel according to the detected element spectrum with the known chemical compositions and generate a mineral distribution map [31, 40]. The surfaces of the samples were gently polished with 1000, 2000, and 4000 grade polishing paper, and then polished to 0.5 μm in a silicon solution. The samples were ultrasonically cleaned in water to remove any silica-like material during polishing and drying. In order to improve the conductivity of the samples under an electron beam, several nanometers of amorphous carbon was sputtered on the surface of the samples.

SEM-EDS experiments were carried out on an FEI Scios™ system, which is equipped with secondary and backscatter electron detectors and silicon drift detectors of the Oxford instrument for EDS. By adjusting the scanning voltage, it was found that the contrast of the SEM images was good at 10 kV, so 10 kV was selected to scan the end face of the sandstone sample in this study. The spot size of the scanning electron microscope image was 10 nm and the working distance was 6.8 mm. EDS mapping images of the same surface were also obtained using AZtec software.

2.2.6 Multi-resolution CT scans

Micro-CT scans are based on the different X-ray absorption coefficients of different density components in the rock, and the computer image reconstruction is used to make the material information of different densities on the designated layer displayed in high-resolution digital images. The low permeability uranium-bearing sandstone used in this study is easily damaged when sampling with a drill bit because of its low mechanical strength, so the rock sample is embedded in epoxy resin to ensure the strength. Then, samples of 15 mm × 15 mm × 20 mm and 1.5 mm - 1 mm microplugs were drilled by bits of different sizes. Micro-CT scans were performed on the nanoVoxel-2000 X-ray CT scanner system produced by Sanying Precision Instrument Co., Ltd. (Tianjin, China), which has a maximum resolution of 0.5 μm. The basic sample size used for the entire study were rectangular samples of 15 mm × 15 mm × 20 mm. First, the basic samples were scanned by micro-CT with medium resolution (12.58 μm) to observe the overall structure and internal porosity. According to the medium-resolution CT scan results, the volume of interest (VOI) was then selected for further imaging. The locations of the subsamples (1.5 mm - 1 mm) were marked and extracted from the basic cores and scanned at a higher resolution (1 μm) to study the finer pore structure. Table 2 shows the settings of the scanning parameters of the micro-CT equipment at different resolutions.

3.1.1 Mineral composition and chemical constituents

XRD combined with XRF analysis can improve the accuracy of mineral composition identification. XRD analysis of the samples (Fig. 4) shows that the main minerals in the samples are quartz, ankerite, feldspar, and calcite, followed by clay minerals such as kaolin, illite, and chlorite, and a small amount of metal minerals (owing to the small content, it is not shown in the figure), such as pyrite, rutile, and sphalerite. Clay minerals mainly include kaolinite, illite, and chlorite. Kaolinite is the most developed, followed by chlorite and illite, which reflects the dissolution of feldspar during diagenesis and mineralization.

Fig. 4.Full-screen View

XRD analysis of low permeability uranium-bearing sandstone. (a) Histogram of minerals relative contents; (b) Histogram of clay minerals relative contents.

The results of petrochemistry analysis of low-permeability uranium-bearing sandstone are shown in Table 3. The K₂O+Na₂O content of the ore bearing rock samples ranges from 2.88 % to 3.30 %, and the K₂O/Na₂O value is greater than 1, which indicates that the feldspar clasts in the sandstone are mainly K-feldspar.

3.1.2 Mineral distribution characteristics

The mineral and pore distribution images of the polished faces were obtained using EDS-SEM analysis. EDS mapping analysis showed the distribution of Si, Al, Ca, Mg, O, C, K, and S (Fig. 5). It can be seen that the main intergranular pores in the rock sample are mainly filled with clays (Al+Si+Mg), while quartz (Si+O), dolomite (Ca+Mg+C), K-feldspar (Si+Al+K), and other particles that constitute the rock sample skeleton are covered with many clays. According to the mineralogical composition and mass fraction in Fig. 5, it can be seen that the relative content of ankerite and dolomite is high, while the relative content of calcite is low, which indicates that there is significant dissolution and metasomatism in the diagenesis of low permeability uranium-bearing sandstone reservoir [42].

Fig. 5Full-screen View

(Color online) EDS spectrum and map captured from low permeability uranium-bearing sandstone

3.1.3 Occurrence state of uranium

Under the SEM, uranium minerals are mostly bright white, and the particles are small and scattered in the rock minerals. In addition, coffinite often coexists with pyrite and disperses in sandstone cements in clumps (Fig. 6a). The BSE image shows that the irregular pyrite grains are mainly distributed around clastic quartz and rock debris, while star point pyrite is mainly distributed in cement and clay minerals (Fig. 6a). Due to the strong adsorption of clay minerals, coffinite is absorbed by clay minerals and then precipitated around the clay membrane or in clay minerals (Fig. 6b). Coffinite also exists on the surface of pyrite and in intergranular pores (Fig. 6c). Moreover, some intact coffinite occurs in the middle of clastic grains such as feldspar, which should be carried from the provenance area (Fig. 6d).

Fig. 6.Full-screen View

(Color online) BSE images of uranium in low permeability uranium-bearing sandstone. (a) Coffinite and pyrite in carbonate cements; (b) Coffinite precipitate in residual intergranular pores of clay minerals and sandstone interstitial materials; (c) Coffinite adsorption on pyrite surface; (d) Coffinite occurs in the interstitial materials of quartz and feldspar. Qtz = Quartz; Cly = Clay mineral; CC = Carbonate cement; Fsp = Feldspar; Cal = Calcite; Py = Pyrite; Cof = Coffinite; Org = organic.

3.2.1 Pore structure types

The pore types and genesis of low-permeability uranium-bearing sandstone were analyzed by micro-CT scanning and SEM. It can be seen that there are four main pore types in the cores of low permeability uranium-bearing sandstone, which include intergranular pores, intragranular dissolution pores, intercrystalline micropores, and micro fractures, as shown in Fig. 7. Among them, intergranular pores mainly include intergranular pores of detrital minerals and intergranular dissolution pores. The number of intergranular pores of clastic minerals is small, but the spatial scale is relatively large, and most of them are in the shape of narrow triangles or multilateral shapes with smooth straight edges (Fig. 7a). The intergranular pores of clastic minerals were altered by diagenesis, such as mechanical compaction or cemented filling of clay minerals, resulting in a large number of micro-nano scale intergranular dissolution pores (Fig. 7b). The dissolution pores are mainly secondary pores produced by carbonate particles, feldspar particles, and interstitials (kaolinite, chlorite) formed by dissolution. Dissolution pores within the grains are widely developed and have irregular pore surfaces with diameters ranging from a few nanometers to a few microns (Fig. 7c, d, and e). Intergranular micropores are mostly formed by authigenic minerals and clay minerals in rocks (Fig. 7f, g) with small pore diameters. Micro fractures mainly exist in carbonate minerals and are formed by dissolution as well as brittle mineral fracture and clay mineral shrinkage (Fig. 7h, i) with a width of less than ten nm and up to several hundred microns. The existence of micro fractures enhances the connectivity of reservoir pores, increases the permeability, and plays an important role in the drainage of the solution. In summary, it can be seen that in the samples, even the same mineral may contain multiple pore types. The pore structure of low-permeability uranium-bearing sandstone is complex and has multi-scale characteristics.

Fig. 7Full-screen View

(Color online) Representation of the typical pore types in low-permeability uranium-bearing sandstone: (a) intergranular pore of clastic minerals, CT; (b) intergranular dissolution pores between quartz and kaolinite, BSE; (c) intragranular pores formed by dolomite dissolution, BSE; (d) intragranular dissolution pores formed by feldspar dissolution, BSE; (e) intragranular dissolution pores formed by kaolinite dissolution, BSE; (f) intercrystalline pores in kaolinite, SEM; (g) intercrystalline pores in pyrite, BSE; (h) micro fractures formed by brittle mineral fracture, BSE; (i) microfractures formed by shrinkage of clay minerals, BSE.

3.2.2 Pore throat size distribution analysis by MIP

The MICP data show that the pore throat sizes of low-permeability uranium-bearing sandstone vary from millimeter to micrometer. According to the diameter, the term ''micro pore throat’’ is applied to pore throat sizes less than 1 μm, ''meso pore throat’’ is the term applied to pore throat sizes less than 10 μm, and ''macro pore throat’’ is the term applied to pore throat sizes greater than 10 μm [51, 52]. Based on the statistics of the influence of different pore-throat sizes on the permeability of the samples, it was found that the contribution of permeability of the samples mainly comes from the large throat (99.51 %), and it can be inferred that the large pore throat has good connectivity.

3.2.3 Pore size distribution analysis by NMR

The NMR test results of the samples were converted into pore diameter distributions, as shown in Fig. 9. The pore diameter of the samples is in the range of 50–300 μm. Although the seepage ability of micropores is small, the pore volumes of the nanopores and micropores are larger. The micropores are not easily wetted and permeated by the lixivium, which limits the free flow and permeability of the lixivium in its interior. Moreover, the influence of clay minerals on porosity and pore flow capacity cannot be ignored when clay minerals disperse into fine particles along with the flow of lixivium and block micropores.

Fig. 9Full-screen View

(Color online) NMR pore size distribution: (a) S1; (b) S2.

3.3.1 Image registration

Image registration is a necessary preliminary procedure for multi-image fusion. The feature-based image registration method uses the feature points of the image to achieve image registration. Firstly, the interested image features are detected by the feature extraction operator. Secondly, the feature matching algorithm is used to select the feature pairs with matching relationships in the corresponding feature set of two images. Then the transformation model between images is calculated by using the coordinate relationship between pairs feature parameters to achieve image registration [54]. The main goal of spatial registration of different resolution 3D CT images is to align and overlap the two-image data into the same coordinate system.

In this study, registration with the landmarks algorithm provided by FEI Avizo was used to register CT images of low-permeability uranium-bearing sandstone with different resolutions. In registration, typically a low-resolution CT image dataset is taken as the reference, and the model of high-resolution CT image data sets is transformed, moved, and possibly rescaled, until both data sets match. According to the location of the pre-marked subsample, the converter of the Transform Editor was used to translate, rotate, and scale the reference image, so that the model image was roughly aligned with the reference image. Landmark sets were then artificially selected and created to calculate the transformation equation matching the landmark sets. Using the Tcl command in Avizo console, the numerical values of the registration transformation can be retrieved. Owing to the distortion of the image in the process of sampling, the simple rigid body transform algorithm cannot satisfy the registration of the images. Therefore, the Bookstein image distortion algorithm was used to register the images fine. The points in the transformed group (model) were forced to fit to the reference points, and all image points in between were shifted by using the correlation of the surrounding points to achieve the best image registration [14, 55]. After verification and confirmation of overlapping known features, the registration is successful, and features (texture, crack, void) with the same coordinates are found on the final registered image (Fig. 11 c, d).

Fig. 11Full-screen View

(Color online) 3D to 3D image registration: (a) registration surface of medium-resolution CT; (b) registration surface of high-resolution CT, yellow curves indicate possible points for landmarks (A, A ’and B, B ’); (c) 2D registration between core and cross section of sub–sample, red curve indicates Fig. 11 (b), and (d) 3D registration between sub-sample and core.

3.3.2 Pore Segmentation

After fusion, CT images have both high- and low-resolution image data. For medium-resolution CT images, owing to the serious marginal fuzzy effect of pores and skeletons, they cannot be simply divided by the binary method [56]. The K-means method is a purely stochastic technique that estimates G_min and G_max based only on the gray-level distribution of the 3D microtomography images [24, 57]. Using the two threshold levels G_min and G_max estimated by the K-means algorithm, the gray-level of the CT image can be divided into three segments: pores, skeleton, and unresolved pores. The linear function of the porosity and probability distribution function of the grayscale of CT images with different resolutions are shown in Fig. 12. By calculating the G_min and G_max of the high-resolution voxel corresponding to the low-resolution voxel, and changing the gray value of G_min of the low-resolution CT image, although the pores and skeletons cannot be completely accurately divided, the overlapped gray levels in the low-resolution CT image can be better segmented and the pore characteristics of diameter greater than 1 μm in the high-resolution CT image can be fused [16]. Indeed, the contribution of the macro throat (> 10 μm) to the permeability of the sample in this test is dominant, and that of the micro throat (<1 μm) can almost be ignored, so the digital core pore model established after integrating the pores of the high-resolution CT image meets the reliability of the next permeability simulation analysis.

Fig. 12Full-screen View

(Color online) Porosity linear functions and the probability distribution functions of grey levels of different resolution CT images.

Finally, the three-dimensional digital core pore model of the plunger sample is obtained, as shown in Fig. 13. It can be seen from Fig. 8b that the pore connectivity in the medium-resolution CT image has been significantly improved after the fusion of the pores identified by the high resolution.

Fig. 13Full-screen View

(Color online) Pore segmentation results of low-resolution image: (a) medium-resolution CT image; (b) the segmentation result of medium-resolution CT images combined with high-resolution CT segmentation threshold, black represents the pores obtained from conventional binary segmentation, and green represents the pores after fusing the segmentation threshold G_min of high-resolution CT image

Fig. 8Full-screen View

(Color online) MICP pore throat size distribution: (a) S1; (b) S2.

3.3.3. Skeleton Segmentation

Owing to the blurring of edges caused by noise and volume effects in the process of image acquisition and reconstruction, the commonly used simple threshold segmentation method often produces inaccurate or even wrong segmentation results, especially in the phase transition region where more than two phases are segmented. Watershed segmentation can effectively solve these problems in many cases. The principle and operation of watershed segmentation are introduced in detail in [36, 46], which will not be repeated here. However, watershed segmentation is prone to undersegmentation or oversegmentation, and the segmentation results usually have some errors with the measured values.

BSE images have the advantages of high resolution, clear microscopic component morphology, and can be combined with energy spectrum analysis (EDS) to identify minerals. At the same time, the gray value of the SEM image pixels is related to the density of rock samples, so it can be used to verify the distribution of gray values of different mineral components in CT images [38, 58, 59]. In this study, SEM images, EDS spectra, and maps were used to calibrate the threshold segmentation of different components of the CT image. The SEM image obtained in this study is an 8-bit grayscale image. The histogram tool in ImageJ software was used to quantitatively count the gray range of each component in the sample, and the average value was obtained. The segmentation thresholds of pyrite and coffinite, feldspar, dolomite, quartz, dolomite, clay minerals, and pores are 242-255, 173-241, 172-146, 105-145, 104-85, 54-85, and 0-53, respectively.

Because the gray value in CT images depends on the density of each mineral phase, it is difficult to distinguish between two minerals with similar densities, such as calcite and dolomite. In this study, the CT images were divided into 7 phases, namely pore space, clay minerals, calcite/dolomite, quartz, ankerite, feldspar, pyrite, and coffinite for further analysis and volume performance simulation. Comparing the sample EDS spectrum and map with the segmentation threshold of watershed segmentation, reasonable segmentation thresholds of minerals are obtained. Multi-mineral SEM images after watershed segmentation are shown in Fig. 14b. Compared with the EDS image (Fig. 14a), it can be seen that good segmentation results can be obtained by using this method. The mineral skeleton of high-resolution CT images is also divided by watershed segmentation. The threshold value of each mineral in the CT image was determined by comparing the gray value and segmentation threshold of the BSE image. The multi-mineral images of CT after watershed segmentation are shown in Fig. 14d.

Fig. 14.Full-screen View

(Color online) 2-D image of (a) EDS image; (b) segmented micro BSE image; (c) CT image; (d). Segmented CT image. Qtz = quartz; Cly = clay mineral; Cal = calcite; Dol = dolomite; Fsp = feldspar; Py = pyrite; Cof = coffinite.

3.3.4 Digital core evaluation

A variety of morphological algorithms in Avizo Fire are applied to quantitatively characterize the micropore characteristics and mineral content proportion of low-permeability uranium-bearing sandstone. The fast watershed algorithm was used to mark the pores, and each pore was quantitatively analyzed. According to the volume of marked pores, the porosity of the digital core could be obtained [25, 60-62]:

where $V_{\text{p}}$ is the single pore volume, pix³; $V_{\text{v}}$ is the total volume, pix³.

On the basis of the 3D digital core pore model, isolated pores can be removed by using four connected region labeling algorithms, and connected pores can be obtained [63]. The porosity of open pores and close pores can be calculated according to formula (3). It can be seen that the CT images combined with the two resolutions can identify more pores, which is helpful to build an accurate digital core (Table 4).

According to the principle of equivalence, the equivalent pore diameter of each marked pore can be obtained [25]:

where d_eq is the equivalent pore diameter, pix.

The histogram of the equivalent pore diameter distribution is shown in Fig. 15. Due to the limitation of CT resolution, pore sizes with a diameter of less than 1μm were not counted, but the distribution of pore size with a diameter of more than 1 μm is in good agreement with the results of the NMR test (Fig. 9).

Fig. 15.Full-screen View

Equivalent diameter of pore size distributions based on micro-CT images: (a) S1; (b) S2.

Similar to the porosity, according to the volume of the marked minerals, the proportion of mineral components of the digital core can be obtained. The proportions of the six mineral phases are listed in Table 5. It can be seen that the digital core mineral volume fraction is consistent with the results of the XRD measurement, and the structural information of different components of the core is reconstructed.