Study on microscopic mechanisms of slurry infiltration in calcareous sand based on CT scanning

A novel detachable microslurry infiltration device tailored for computed tomography (CT) scanning was developed. Using this device, a series of slurry infiltration tests was conducted on calcareous and Fujian sand columns with various bentonite slurry concentrations. CT scanning technology nondestructively captured the cross-sectional image slices of each specimen for analysis. A comprehensive image processing methodology was deployed to precisely differentiate the three phases—calcareous sand, slurry, and air—enabling 3D reconstruction of the infiltrated sand column. This approach combines three techniques: threshold segmentation, the watershed algorithm, and deep learning. Distinctive particle morphologies inherent to the calcareous and Fujian sands were observed. Notably, CT scans revealed a markedly greater angularity and irregularity of the calcareous sand compared with the Fujian sand, leading to different pore characteristics. Given an identical permeability coefficient, the calcareous sand porosity exceeded that of the Fujian sands. Furthermore, image analyses revealed distinct features of filter cake formation in these two soil columns. In calcareous sands with finer grain sizes, the slurry particles were more prone to clogging because of constricted and irregular seepage pathways. Conversely, with increasing particle size, the internal pores of the calcareous sand particles augmented channels for slurry infiltration, hindering filter cake formation.


Introduction
The slurry pressure-balanced shield tunnelling technique has been widely used for constructing underwater tunnels because of its adaptation to different soil conditions.As the development of islands and reefs continues worldwide, shield tunnelling in calcareous sand strata has gradually gained prominence.
In recent years, new microscopic techniques have become increasingly popular in engineering.Notably, X-ray computed tomography (CT) [1] and scanning electron microscopy [2] have emerged as valuable nondestructive testing tools.These techniques offer numerous advantages for quantifying and analyzing internal and external features and structures without compromising the integrity of the tested specimens.Numerous studies have consistently demonstrated that calcareous sand particles exhibit distinct characteristics compared to sand derived from land sources.These differentiating features include significantly higher angularity, irregular shape, increased internal pore volume, and greater heterogeneity.[1][3][4] 2 In this study, the mechanism of filter cake formation under different strata conditions was investigated through a series of microslurry infiltration tests on columns composed of calcareous and Fujian sand.The analyses focused on the particle morphology, filter cake distribution, and pore distribution to gain insights into the filter cake formation process.These findings contribute to a better understanding of the formation of filter cakes under varying strata conditions.

Test apparatus and materials
The penetration column size was reduced from that used in previous studies [5][6] and made into a detachable unit for the CT scanning of specimens.Miniature water manometers and electronic scales were used to record the pore water pressure and water flow at a frequency of 1 Hz, as shown in figure 1.

Figure 1. Test apparatus
The strata were prepared using Fujian and calcareous sand with grain sizes of 0.315-0.63mm and 0.63-1.25 mm, respectively.Sodium bentonite was used to prepare the slurries.

Test procedure.
Owing to the effect of the internal porosity of calcareous sands, the porosity of calcareous sands under the same grading and natural accumulation states was much higher than that of Fujian sands.Because of the difficulty in controlling the porosity of the two sands, the same permeability coefficient was determined after conducting constant head tests.
During the slurry infiltration phase, the test was completed by controlling 50 ml quantities of slurry with different concentrations at pressures of 50 and 150 kPa.

Test result
After observing filter cake generation, the Peclet number [7] was calculated according to the relevant formulae to determine the filter cake status, as shown in table 1.
is the presence of a visible filter cake forming on the formation surface.○ is the presence of holes within the filter cake.× is the near absence of filter cake formation.The tests highlighted in yellow were used for CT scans.

Principle of CT imaging
Recently, CT has become popular in several fields.CT scanning is a nondestructive testing method that can visualize internal features and structures in great detail.
During scanning, the test object was exposed to a continuous stream of X-rays.As X-rays pass through different materials, they undergo various phenomena, such as photoelectric effects and scattering, leading to attenuation.X-rays of different densities are attenuated to different degrees, resulting in the formation of a grayscale image that conveys information on the varying densities of the materials within the object.
In this study, test samples were measured by the YXLON FF85 Computed Tomography Inspection System with a scanning accuracy of up to 18 µm .

Image processing methodology
Figure 3 shows the distribution of grey scale values obtained from the CT scan results.The grayscale values of the sample primarily range from 1800 to 7700.The distribution of the grayscale values reveal 4 the presence of four distinct peaks, corresponding to air, water, slurry, and sand in increasing order of density.

Processing flowchart
The conventional method of CT image processing involves the separation of different phases using threshold segmentation, which is more effective for two-phase materials [8].However, the effectiveness of this method decreases when dealing with multiphase materials.Because an overlap existed between the grayscale values of neighboring phases, it was challenging to differentiate these four phases using threshold-segmentation-based binarization alone.Therefore, a comprehensive image-processing methodology was employed.
Figure 4 shows the detailed processing of the CT images.First, threshold segmentation was used to select the peak region of each phase.Then, a gradient calculation was performed on the gray value of the image to identify the boundaries between the phases.Finally, a watershed algorithm [9] was utilized to distinguish the four phases based on the identified boundaries.Logical operations and deep learning techniques can be utilized to further optimize the image and complete 3D reconstruction based on the detailed information obtained.

Segmentation display
The grayscale value of the CT scan is determined by the substance density.The presence of significant quantities of bentonite flocs between particles, which form filter cakes, can increase the density of the surrounding liquid, which has a substantial impact on image segmentation.Figure 5 illustrates the contrast between the regions segmented using only the threshold method and those segmented using the watershed method.The red circle in figure 5(b) depicts the challenge of accurately discerning flocs between particles with densities similar to those of sand when utilizing threshold segmentation.This limitation results in a critical flaw in the 3D reconstruction, as shown in figure 6(a).However, after incorporating the watershed method, figures 5(c) and 6(b) indicate a significant improvement in the segmentation effectiveness.

Result analysis 3.3.1. Visualization
The results of the reconstructed slurry infiltration are depicted in figure 7.For 0.315-0.63mm grain size sand, a slurry condition with a mass concentration of 12% (figures 7(c) and (d)) can produce a high-quality filter cake on the surface of both sands, where the filter cake is comparatively thicker in the calcareous sand strata.In contrast, at a slurry concentration of 10%, the filter cake on the surface of the Fujian sand (figure 7(a)) was very thin and contained holes.After slurry infiltration, air broke through the strata, as indicated by the red area in figure 7(a).The calcareous sand layer was able to form a closed-air filter cake even at a slurry concentration of 10 %; although, the thickness of the film was lower than that formed at a slurry concentration of 12 %.
Under high slurry concentration conditions when the grain size is 0.63-1.25 mm, forming filter cakes is easy for the Fujian sand (figure 7(e)) and difficult for the calcareous sand (figure 7(f)).

Slurry analysis
For fine sands, the quality and thickness of the filter cake formed in calcareous sand were higher than those in Fujian sand under the same conditions, whereas the result converged in coarse sands.
Figure 8(a) shows that the volume fraction of the slurry in the sand column (10-45 mm in height) increased as the quality and thickness of the filter cake decreased.A high-quality filter cake typically results in lower filtration losses, indicating that less slurry enters the sand column interior.This phenomenon hinders slurry particle accumulation within the column.
As shown in figure 8(b), the volume fraction of the slurry is higher in the coarse sand column than that in the fine sand.This is due to the increased porosity associated with larger particle sizes, resulting in higher slurry retention in the pore space.

Pore porosity analysis
Figure 9 shows the pore distribution using the cross-sectional porosity.The porosities of Fujian and calcareous sand were approximately 35 and 45%, respectively.These values are similar to those obtained prior to the test.Notably, the differences between the two curves in figure 9(a) may be because the Fujian sand failed to form a closed air filter cake at a slurry concentration of 10%.Consequently, air penetrated the stratum, leading to disturbances within the stratum, as shown in Fig. 7(a).

Internal pore analysis
During slurry infiltration, the slurry gradually squeezed the water used in the saturated sandy soil out of the infiltration column, and the two liquids fused.Owing to the intermolecular forces and self-gravity, the slurry particles were adsorbed onto the sand surface and settled, as shown in figure 10.Figures 10 and 11 show two screws with different particle sizes.The larger screw had a greater volume occupied by air, whereas the interior of the smaller screw was filled with slurry and water.Air was present only at the opening of the screw mouth of the smaller screw.The presence of surface tension in the liquid resulted in a minimal flow inside the smaller screws, reducing the impact of the pores within the screws on the overall seepage.However, as the pore size inside the screw increased, the flow of liquid inside became significant and could not be overlooked.This phenomenon also accounts for the difficulty of filter cake generation between the calcareous and Fujian sands with different grain sizes.

Conclusion
In this study, CT scanning was performed after slurry filtration.The following conclusions were drawn from the test results and CT images.
Using a comprehensive image-processing methodology based on watershed segmentation, multiphase materials with different densities can be reliably distinguished.The porosity of the calcareous sand is higher than that of the Fujian sand under the same compacting conditions because of the internal pores and particle irregularity.
In fine calcareous sands, small pores inside the particles have less influence on the overall infiltration owing to the presence of surface tension in the liquid.Conversely, the large pores inside the coarse calcareous sand particles play a crucial role in infiltration, providing more seepage pathways that hinder filter cake formation.
In the engineering slurry concentration range, it is difficult for calcareous sand strata with particle sizes greater than 0.63 mm to form filter cakes, and subsequent consideration can be given to adding other solid-phase materials to the slurry or improving the bentonite gradation to meet the excavation requirements.

Figure 2 Figure 2 .
Figure2shows representative photographs of the three filter cake forms listed in table 1.

Figure 5 .
Figure 5. Selection regions of different image processing methods.(a) Original, (b) threshold, and (c) watershed images.

Figure 6 .
Figure 6.Reconstruction of different image processing methods.(a) Threshold and (b) watershed images.

Table 1 .
Filter cake formation situation.