2.2.2. X-ray Micro-CT

The instrument used for tailings sand sample testing is a nano Voxel-3000 micro-CT provided by Sanying Precision, Tianjin, China, as shown in Figure 3. The instrument has advanced nondestructive 3D imaging and image analysis capabilities, with secondary optical magnification and a spatial resolution of up to 0.5 μm. Using this equipment, the 3D visualization of the internal microstructure of materials can be nondestructively investigated.

**Figure 3.** nanoVoxel-3000 micro-CT.

In this test, the test voltage of the micro-CT equipment is 140 kV, the test current is 70 μA, the exposure time is 0.42 s, the standoff distance (SoD) from the source to the sample is 13.6 mm, and the distance from the source to the flat panel detector (SDD) is 279.6 mm. The resolution is 6.16 μm, which can distinguish the characteristics of pores with pore sizes greater than 6.16 μm and tailings sand particles. This experiment adopts the continuous scanning method of a cone-beam with a scanning rate of 0.25◦/frame and a total of 1440 projections. Numerous layers (1536) of the sample are cut longitudinally, and each layer thickness is 6.16 μm. Finally, 1536 two-dimensional (2D) slice images of 1800 × 1800 pixels are obtained. The parameters of the CT scan of the tailings sample are shown in Table 1.


**Table 1.** CT scan parameters.

### 2.2.3. Test Method and Process

To obtain the process of tailings sand infiltration and failure, the test was simulated by gradually increasing the head difference.

According to the critical head penetration failure heads of three different fine-grained tailings sand samples, the heads under test loading are divided into four grades. The top three grades are as follows: level 2 is the head value without seepage damage, level 3 is the critical head damage head value of the sample, and level 4 is the head value after seepage failure. After each stage is loaded stably, a CT scan is performed on the sample pressure relief, as shown in Figure 4. The meso-infiltration characterization inside the sample is obtained. After scanning, the head pressure of the seepage is increased to the next head value. According to the recorded experimental phenomena, the distribution of the head value for each stage of the sample is shown in Table 2.

**Figure 4.** CT scan of the tailings sand sample.


**Table 2.** Head loading values at various levels.

During the test, attention is paid to whether the test surface has water turbidity, bubbling, or fine particle bounce, whether there is a bulge or a rise on the sample surface, and whether there are faults in the sample [15]. When the average velocity of seepage suddenly increases and is unstable or when concentrated seepage occurs on the pipe wall, the test is stopped.

During the test, the values of the seepage velocity and permeability coe fficient are recorded as the head di fference gradually increases (see Table 3).


**Table 3.** Seepage velocity and permeability coe fficient during the test.

When the 30% fine-grained sample is loaded with a third-level head, a slight floating soil phenomenon appears on the top of the sample. When a fourth-level head is loaded, seepage failure occurs. The seepage at the top of the sample is very turbid, and the soil is completely jacked up. This phenomenon is a piping-type infiltration failure. When the 50% fine-grained sample is loaded with a third-level head, a slight floating soil phenomenon appears on the top of the sample, and microcracks appear in the middle and lower parts of the sample. When a fourth-level head is loaded, complete infiltration failure occurs. The seepage flow at the top of the sample is very turbid, the flow velocity increases sharply, and the lower cracks further expand. Fine particle bounces appear at the top of the sample, which suggests that fluid-type infiltration failure occurs. When the 70% fine particle sample is loaded with a second-level head, the flow rate increases sharply. The surface of the tailings sample exhibits a slight support phenomenon, and there are visible cracks in the lower part of the sample. When the third-level head is loaded, complete percolation failure occurs. The seepage at the top of the sample becomes very turbid, the flow velocity increases sharply, and the lower cracks further expand. Fine-grained beating appeared on the top of the sample, and it was judged that a fluid-type infiltration failure occurred.

### **3. Test Results and Analysis**

#### *3.1. Three-Dimensional Reconstruction of the CT Scan of Fine-Grained Tailings Sand*

Cone shadows are generated in the upper and lower slices of the data generated by the CT scan, which a ffects the subsequent 3D reconstruction and quantitative characterization of the pore structure. Therefore, the CT layers need to be cropped in the VGStudio max 3.0 visualization software provided by Sanying Precision, Tianjin, China to obtain 1000 layers of 1800 × 1800 pixels. The 2D slice map was imported into the Avizo software for the 3D reconstruction for further analysis. The 2D slice image of the tailings sand sample obtained by cone-beam scanning is a ffected by various types of system noise and artifacts. The gray image obtained by the local mean filter needs to be filtered and denoised [16]. Then, the artifacts are e ffectively removed by the Avizo software provided by Sanying Precision, Tianjin, China, and the scanned data are reconstructed by the software's iterative reconstruction algorithm to obtain a 3D map of the tailings sand sample under the head pressures of 1–4, as shown in Figure 5. Afterward, the image segmentation processing is performed according to the commonly used Otsu algorithm [17], and the grayscale image is binarized. Finally, the 3D pore structure and tailings sand particle structure in the data volume are extracted by the Avizo software, and a 3D pore and tailings sand particle network model of the pores and tailings sand particle spatial arrangemen<sup>t</sup> is generated from the micro-CT image.

**Figure 5.** Three-dimensional reconstruction of the three fine-grained tailings samples. (**a**) First-level head of 30% fine content; (**b**) Second-level head of 30% fine content; (**c**) Third-level head of 30% fine content; (**d**) Fourth-level head of 30% fine content; (**e**) First-level head of 50% fine content; (**f**) Second-level head of 50% fine content; (**g**) Third-level head of 50% fine content; (**h**) Fourth-level head of 50% fine content; (**i**) First-level head of 70% fine content; (**j**) Second-level head of 70% fine content; (**k**) Third-level head of 70% fine content; (**l**) Fourth-level head of 70% fine content.

#### *3.2. Characterization of the Microstructure of the Pores during Infiltration Failure of Tailings Sand*

Based on the previous digital image processing, to obtain the pore structure model of the tailings sand samples, the grayscale image is binarized using image threshold segmentation, as shown in Figure 6a–x. The volume rendering module of the Avizo 3D visualization software is used to directly perform the 3D reconstruction of the pore structure of the sample, and the overall 3D pore microstructure of the three fine-grained tailings sand samples is obtained, as shown in Figure 7.

**Figure 6.** Pore threshold segmentation of the three fine-grained tailings sands. (**<sup>a</sup>**,**b**) First-level head of 30% fine content; (**<sup>c</sup>**,**d**) Second-level head of 30% fine content; (**<sup>e</sup>**,**f**) Third-level head of 30% fine content; (**g**,**h**) Fourth-level head of 30% fine content; (**i**,**j**) First-level head of 50% fine content; (**k**,**l**) Second-level head of 50% fine content; (**<sup>m</sup>**,**<sup>n</sup>**) Third-level head of 50% fine content; (**<sup>o</sup>**,**p**) Fourth-level head of 50% fine content; (**q**,**<sup>r</sup>**) First-level head of 70% fine content; (**<sup>s</sup>**,**<sup>t</sup>**) Second-level head of 70% fine content; (**<sup>u</sup>**,**<sup>v</sup>**) Third-level head of 70% fine content; (**<sup>w</sup>**,**<sup>x</sup>**) Fourth-level head of 70% fine content.

It can be seen from Figure 7 that the overall pore structures of the three fine-grained tailings sand samples are fully developed, the pores are unevenly distributed throughout the sample, and the pore shape changes are diverse at the surface of the pore structure. Small pores gradually penetrate into large pores, and the pore structure under the head pressure of level 4 is the most obvious.

**Figure 7.** The overall pore structure of the three fine-grained tailings sand samples. (**a**) First-level head of 30% fine content; (**b**) Second-level head of 30% fine content; (**c**) Third-level head of 30% fine content; (**d**) Fourth-level head of 30% fine content; (**e**) First-level head of 50% fine content; (**f**) Second-level head of 50% fine content; (**g**) Third-level head of 50% fine content; (**h**) Fourth-level head of 50% fine content; (**i**) First-level head of 70% fine content; (**j**) Second-level head of 70% fine content; (**k**) Third-level head of 70% fine content; (**l**) Fourth-level head of 70% fine content.

### 3.2.1. Characterization of the Pore Number and Volume

 81,410

4

 140,923  44,661

Based on the 3D model of the tailings sand pore data obtained by Avizo 3D visualization software, the label analysis command in Avizo software is used to statistically analyze the sizes of the pores in the tailings sand sample, and the target pore size and number are filtered by the analysis filter command.

Based on the actual resolution of the tailings sand sample scanning, the pore size of the tailings sand sample is divided into five categories for statistical analysis: D ≥ 80 μm, 40 μm ≤ D < 80 μm, 20 μm ≤ D < 40 μm, 10 μm ≤ D < 20 μm, and 0 μm ≤ D < 10 μm. The statistical results are shown in Tables 4–6 and Figures 8–10.


 24,151  8.34 × 10<sup>10</sup>

**Table 4.** Statistics of the number of pores in the seepage process with the 30% fine content.

**Table 5.** Statistics of the number of pores in the seepage process with the 50% fine content.

 6170  297,315



**Table 6.** Statistics of the number of pores in the seepage process with the 70% fine content.

**Figure 8.** Variation in the pore number. (**a**) 30% fine content; (**b**) 50% fine content; (**c**) 70% fine content.

As seen from the above chart:


(3) During the seepage process, the sample with a 70% fine content has more fine content, and the connected pores are not developed. All pores are reduced during the early stage of seepage, and the force between soil particles is significantly increased. Fluid soil damage rapidly occurs, and the number of pores rapidly increases after the damage.

**Figure 9.** Change in the total pore quantity.

**Figure 10.** Total connected pore volume change.

Through a comparative analysis, we have determined the following:


(3) During the seepage process, the total number of pores in a 70% fine-grained sample is relatively small, the soil particles and the seepage force increase rapidly, and the sample quickly undergoes fluid soil failure. After breaking the ring, the number of pores increases rapidly. The connected pore volume also increases rapidly.

### 3.2.2. Evolution Characteristics of the Connected Pores

On the basis of obtaining the overall pore structure of three fine-grained tailings sand samples, the connected pore structure of the samples with 1–4 head pressures along the seepage direction is extracted by the Avizo 3D visualization software (see Figure 11). Then, the volume of the connected pores is counted.

**Figure 11.** Connected pore structure under three levels of the water content at the three levels of fine particle content. (**a**) First-level head of 30% fine content; (**b**) Second-level head of 30% fine content; (**c**) Third-level head of 30% fine content; (**d**) Fourth-level head of 30% fine content; (**e**) First-level head of 50% fine content; (**f**) Second-level head of 50% fine content; (**g**) Third-level head of 50% fine content; (**h**) Fourth-level head of 50% fine content; (**i**) Second-level head of 70% fine content; (**j**) Third-level head of 70% fine content; (**k**) Fourth-level head of 70% fine content.

With increasing head pressure, the connected pore volume of the 30% fine particle sample gradually increases. Fine particles migrate and flow under the action of the seepage, small pores easily penetrate into large pores, and the increase is obvious. The connected pore structure of the 50% fine-grained tailings sand sample runs through the entire tailings sand sample along the seepage direction because this sand sample is finer than the 30% fine-grained tailings sand sample. Increasing the particle content reduces the pores between the particles, and thus, the connected pores passing through the entire sample are relatively insignificant at the first head pressure. As the head pressure gradually increases, the connected pores increasingly develop around them. This behavior indicates that as the seepage force increases, the fine particles migrate and flow to expand the small pores into larger pores, increasing the connectivity of the sample. As the connectivity increases, the damage to the sample is more serious, and the development of the connected pores is more obvious.

The 70% fine-grained tailings sand sample did not pass through the entire sample through the pores under the first-level head. As the head pressure is increased, the fine particles migrate and flow. At this time, the connected pores penetrate the entire sample. The 70% fine-grained tailings sand sample has a smaller proportion of pores than the 30% and 50% fine-grained tailings sand samples due to the large proportion of fine content, and the connectivity of the pores is poor. From level 2 to level 4 heads, with increasing seepage force, the connected pores develop from the sample center to the surrounding area. As the head pressure increases, the connected pore volume tends to increase

first and then decrease. The 70% fine content tailings sand sample has a large proportion of fine content, and the pores between the particles are small. When the sample begins to infiltrate and break, the volume of the connected pores does not increase significantly. From the 2n<sup>d</sup> to the 3r<sup>d</sup> head, due to the continuous increase in the seepage force, a large number of fine particles migrate and flow, which makes the volume of the connected pores increase sharply to 4.80 × 10<sup>10</sup> μm3. From grade 3 to grade 4 heads, the volume of the connected pores slowly decreases, indicating that the tailings sand sample under the grade 4 head pressure is severely damaged. When a CT scan is carried out when the head pressure is unloaded, a large number of fine particles are uniform due to settlement.

### 3.2.3. Change Characteristics of the Porosity Layer by Layer

Based on the Avizo 3D visualization software, the pores are extracted after the threshold segmentation of three kinds of fine-grained tailings sand samples, and statistical analyses of the layer-by-layer porosity of the 1000-layer 2D slice map of three kinds of fine-grained tailings sand samples at various head pressures are performed separately. The layer-by-layer porosity statistics of the three fine-grained samples are shown in Figure 12.

**Figure 12.** Layer-by-layer porosity statistics of the three fine-grained tailings sand samples.

The following conclusions can be drawn from Figure 12:

(1) With an increasing head pressure of the 30% fine-grained samples, the porosity of the samples in the same area shows an increasing trend from layer to layer. An increase in the seepage force affects the porosity of the tailings sample at each layer. The e ffect of the rate is obvious. With an increase in the number of slice layers in the z-axis direction, that is, an increase in the number of 2D slice layers along the seepage direction, the tailings sand samples at all the levels of the water head show an increasing trend in the porosity of each layer.


### *3.3. Microstructure Distribution Characteristics of the Particle Size and the Particle Size during Percolation Failure of the Tailings Sand*

Based on the Avizo 3D visualization software, the data obtained from CT scans of the three fine-grained tailings sand samples were binarized after threshold segmentation, and tailings sand particles with the highest gray values were extracted, as shown in Figure 13.

**Figure 13.** *Cont*.

**Figure 13.** Threshold segmentation of the tailings sand particles. (**<sup>a</sup>**,**b**) First-level head of the 30% fine content; (**<sup>c</sup>**,**d**) Second-level head of the 30% fine content; (**<sup>e</sup>**,**f**) Third-level head of the 30% fine content; (**g**,**h**) Fourth-level head of the 30% fine content; (**i**,**j**) First-level head of the 50% fine content; (**k**,**l**) Second-level head of the 50% fine content; (**<sup>m</sup>**,**<sup>n</sup>**) Third-level head of the 50% fine content; (**<sup>o</sup>**,**p**) Fourth-level head of the 50% fine content; (**q**,**<sup>r</sup>**) First-level head of the 70% fine content; (**<sup>s</sup>**,**<sup>t</sup>**) Second-level head of the 70% fine content; (**<sup>u</sup>**,**<sup>v</sup>**) Third-level head of the 70% fine content; (**<sup>w</sup>**,**<sup>x</sup>**) Fourth-level head of the 70% fine content.

Based on the data of tailings sand particles obtained after threshold segmentation, 200 layers of 2D slices were used as a region, and a total of 1000 layers were divided into five regions through the Extract Subvolume command in the 3D visualization software of Avizo. Particle size distributions of 10–20 μm, 20–40 μm, and 40–75 μm were counted.

3.3.1. Thirty Percent Fine Content Sample

> The statistical results are shown in Tables 7–10 and Figure 14.


**Table 7.** Variation in the number of fine particles percolated with the 30% fine content (0–10 μm).


**Table 8.** Variation in the number of fine particles percolated with the 30% fine content (10–20 μm).

**Table 9.** Variation in the number of fine particles percolated with the 30% fine content (20–40 μm).


**Table 10.** Variation in the number of fine particles percolated with the 30% fine content (40–75 μm).


**Figure 14.** *Cont*.

**Figure 14.** Variation in the number of fine particles percolated with the 30% fine content. (**a**) Particle size: 0–10 μm; (**b**) Particle size: 10–20 μm; (**c**) Particle size: 20–40 μm; (**d**) Particle size: 40–75 μm.

From the above chart, the following can be observed:


the obvious tailings particle sedimentation in the seepage channel, indicating that a penetrating infiltration failure channel is formed inside.

3.3.2. Fifty Percent Fine Content Sample

> The statistical results of the 50% fine content are shown in Tables 11–14 and Figure 15.

**Table 11.** Variation in the number of fine particles percolated with the 50% fine particle content (0–10 μm).


**Table 12.** Variation in the number of fine particles percolated with the 50% fine particle content (10–20 μm).


**Table 13.** Variation in the number of fine particles percolated with the 50% fine particle content (20–40 μm).


**Table 14.** Variation in the number of fine particles percolated with the 50% fine particle content (40–75 μm).


**Figure 15.** Variation in the number of fine particles percolated with the 50% fine content. (**a**) Particle size: 0–10 μm; (**b**) Particle size: 10–20 μm; (**c**) Particle size: 20–40 μm; (**d**) Particle size: 40–75 μm.

From the above chart, the following can be observed:


The statistical results of the 70% fine content are shown in Tables 15–18 and Figure 16.


**Table 15.** Variation in the number of fine particles percolated with the 70% fine particle content (0–10 μm).

**Table 16.** Variation in the number of fine particles percolated with the 70% fine particle content (10–20 μm).


**Table 17.** Variation in the number of fine particles percolated with the 70% fine particle content (20-40 μm).


**Table 18.** Variation in the number of fine particles percolated with the 70% fine particle content (40–75 μm).



**Figure 16.** Variation in the number of the fine particles percolated with the 70% fine content. (**a**) Particle size: 0–10 μm; (**b**) Particle size: 10–20 μm; (**c**) Particle size: 20–40 μm; (**d**) Particle size: 40–75 μm.

3.3.4. Changes in the Total Value of the Fine Content of the Three Samples

The statistics regarding the changes in the fine contents in three kinds of fine-grained tailings sand samples during percolation are shown in Tables 19–22 and Figure 17.


**Table 19.** Fine particle content change during seepage (0–10 μm).




**Table 21.** Fine particle content change during seepage (20–40 μm).

**Table 22.** Fine particle content change during seepage (40–75 μm).


**Figure 17.** The number of fine particles with di fferent particle sizes. (**a**) 30% fine content; (**b**) 50% fine content; (**c**) 70% fine content.


the fine particles do not have enough pores to flow, and the soil breaks down. After settlement, the fine particle content increases significantly.

(3) Through a comparative analysis, it can be observed that during the percolation of the sample, fine particles migrate along the percolation direction, and the smaller the particles are, the more obvious the migration. The di fference is that the number of fine particles will a ffect the type of osmotic failure of the tailings sand sample. The tailings sand with a fine content near 30% will undergo piping-type osmotic failure, and its fine particle content will continue to decrease. At the 30% fine content, the tailings sand will undergo osmotic failure, its fine particle content will decrease, and then all of the particles will flow.

Comparing the fine particle content migration with the di fferent fine particle contents and di fferent particle diameters, it can be seen that the occurrence of fine particle tailings sand permeation failure is closely related to the fine particle content migration and the formation of the seepage channels. The relationship between the changes in the fine particle content of di fferent particle sizes and the head pressure during the seepage process of tailing sand samples with di fferent fine particle contents is shown in Figure 18.

**Figure 18.** Variation in the particle number during infiltration with di fferent fine contents. (**a**) Particle size: 0–10 μm; (**b**) Particle size: 10–20 μm; (**c**) Particle size: 20–40 μm; (**d**) Particle size: 40–75 μm.

From Figure 18, the following can be observed:


pore volume grows less. As the seepage pressure increases, the flow-type soil infiltration is quickly damaged.

During the seepage process, the changes in the number of fine-grained particles under the heads at all levels are shown in Table 23 and Figure 19.


**Figure 19.** Fine particle content as a function of the number of infiltration processes.

From the above chart, the following can be observed:


Both the relationship between different proportions of fine-grained tailings sand samples and the initial number of fine particles and the relationship between different proportions of fine-grained tailings sand samples and the average number of fine-grained contents per unit volume are analyzed, as shown in Figures 20 and 21.

**Figure 20.** Relationship between the proportion of the fine content and the initial content of fine particles.

**Figure 21.** Relationship between the proportion of the fine content and the average content of fine particles per unit volume.

It can be seen that there is a clear linear relationship between the content of fine particles and the total number of fine particles. At the same time, the analysis of the fine particle content in the unit volume is the ratio of the fine particle content of the sample to the sample volume. Through data fitting, it can be found that the percentage of fine particle content and the fine particle content in the unit volume exhibit a very obvious linear correlation. Therefore, the fine particle content per unit volume can be used as the mesoparameter of samples with different fine particle contents.

### *3.4. Macroscopic and Mesofactor Analysis of the Seepage Process*

Based on the macrophysical and mechanical properties, a comparative analysis of the parameters of the microstructure characterization of the tailings sand sample is performed. The macro- and mesofactor comparison table is shown in Table 24. The relationship between the fitted macro- and mesofactors is shown in Figures 22–26.


**Table 24.** Macro- and mesofactor comparison table.

**\$YHUDJHQXPEHURIILQHVFRQWHQWSHUXQLWYROXPH**

**Figure 22.** Fitting curve of the fine particle content and permeability coe fficient per unit volume.

**\$YHUDJHQXPEHURIILQHVFRQWHQWSHUXQLWYROXPH**

**Figure 23.** Fitting curve of the fine particle content and internal friction angle per unit volume.

\$YHUDJHQXPEHURIILQHVFRQWHQWSHUXQLWYROXPH

**Figure 24.** Fitting curve of the fine particle content and cohesion per unit volume.

**Figure 25.** Fitting curve of the connected pore volume and permeability coe fficient.

**Figure 26.** The fitting curve of the average porosity and permeability coe fficient.

The following conclusions can be drawn from the above chart:

(1) Analyzing the data relationship between the content of fine particles per unit volume and the permeability coe fficient, internal friction angle, and cohesion that a ffect the seepage and failure of the tailings sand sample macroscopically, the following data characteristics can be found:

With increasing fine particle content per unit volume, the permeability coe fficient and the internal friction angle gradually decrease. The former shows a more obvious exponential function relationship, while the latter shows a more obvious positive correlation with a linear function. For the fitting formula, see Equations (1) and (2).

$$k = a\_1 \times e^{b\_1 n} \tag{1}$$

where *k* is the internal friction angle, cm·s<sup>−</sup>1; *n* is the fine particle content per unit volume; *a*1 is a material parameter, which can be obtained by fitting the experimental data; and *b*1 is a material parameter, which can be obtained by fitting the experimental data.

$$
\varphi = a\_2 + \mathbf{b}\_2 \times n \tag{2}
$$

where ϕ is the permeability coe fficient, ◦; *n* is the connected pore volume, μm3; *a*2 is a material parameter, which can be obtained by fitting the experimental data; and *b*2 is a material parameter, which can be obtained by fitting the experimental data.

As the content of fine particles in a unit volume increases, the cohesion gradually increases, and the two show a relationship in the form of an exponential function. The fitting formula is shown in Equation (3):

$$x = a \times e^{lm} \tag{3}$$

where *c* is the cohesion, kPa; *n* is the fine particle content per unit volume; *a* is a material parameter, which can be obtained by fitting the experimental data; and *b* is a material parameter, which can be obtained by fitting the experimental data.

(2) The connected pore volume and the average porosity of the sample can be used as the macroand micropore structure characteristics of the sample. The data relationship analysis reveals the following data characteristics:

The initial volume of connected pores has a positive linear correlation with the permeability coe fficient. The fitting formula is shown in Equation (4).

$$k = a + b \times V \tag{4}$$

where *k* is the permeability coe fficient, cm·s<sup>−</sup>1; *V* is the connected pore volume, μm3; *a* is a material parameter, which can be obtained by fitting the experimental data; and *b* is a material parameter, which can be obtained by fitting the experimental data.

The fitting curve of the average porosity and permeability coe fficient shows an exponential relationship, and it's fitting formula is shown in Equation (5).

$$k = a\_1 \times n^{b\_1} \tag{5}$$

where *k* is the permeability coe fficient, cm·s<sup>−</sup>1; *n* is the average porosity of the sample, %; *a*1 is a material parameter, which can be obtained by fitting the experimental data; and *b*1 is a material parameter, which can be obtained by fitting the experimental data.

Through the analysis of the above data, it can be found that the number of fine particles in a unit volume and the connected pore volume can be used as microstructure parameters to establish a relationship with the macroscopic physical and mechanical properties of the sample. The observation angle is combined with the macrophenomena for comparative analysis.
