*2.2. Instrumentation*

The experiment mainly includes rainfall device and sensor monitoring equipment. The rainfall device is mainly composed of water pipeline, water tank, water pump, rainfall sprinkler, portable control center and rainfall meter. In order to ensure the uniformity of rainfall, the rainfall height was designed to be 6 m according to the top area and height of the model. In the control system interface, the continuous change of rainfall intensity can be realized by adjusting the opening degree of 0–150 mm/h, and the specific rainfall intensity value can be obtained by connecting the rain gauge with the control system. The sensors are pore water pressure sensor and moisture content sensor. The range of pore water pressure sensor is −20–20 kpa, the output voltage range is 0–5 v, and the acquisition frequency can reach 1 Hz, that is, 1 s can collect a datum. The moisture content sensor can collect one datum in one minute, and the mass moisture content can be obtained by dividing the corresponding dry density.

#### *2.3. Materials and Methods*

The test samples were taken from a Loess Fill in Qilipu community, Yan'an, Shaanxi, China (Figure 3), and the retrieved soil samples were screened for 5 mm to remove large particles and impurities. According to the actual situation, the dry density of the original slope in the model test is 1.63 g/cm3, the water content is 13.5%, and the natural gravity (the natural gravity of soil refers to the weight of soil under the condition of natural moisture content, which is equal to the ratio of the total weight of soil to the total volume of soil) is 18.53 kg/m3. The dry density of the fill slope is 1.58 g/cm3, the moisture content is 10%, and the natural gravity is 17.38 kg/m3. Before the beginning of the model test, the moisture content of the packaged soil was measured by the drying method, and the moisture content was about 16.3%. Therefore, the soil needs to be turned and aired. The measured water content after airing was about 14.2%. Considering the water loss in the filling process, the original slope was filled directly with the water content. Test slope model length × wide × Height = 3.2 m × 1.4 m × 1.2 m, the filling slope height is 0.9 m, and the interface between the original slope and the filling body and the filling slope angle are taken as 30◦ (Figure 2).

The modeling idea is to use the dry density and moisture content to jointly control the manual compaction. Each 10 cm is divided into one layer, and the required mass of each layer is calculated. Taking into account the slope shape to be formed in the later stage, the soil that is appropriate greater than the calculated mass is weighed. The filling of the layer is carried out, and the density of the compacted soil layer is measured by the ring knife method. After meeting the requirements, the next layer of filling is carried out. Firstly, the original slope is filled on the left side, and it is placed for a week. Then, the dry density and moisture content of the filling area are controlled on the right side. The slope on the right side is filled as the filling slope. When filling the filling slope, the penetration test ring knife sample is taken to fill the filling area. The sensor part is buried using Luoyang shovel. According to the meteorological data of Yan 'an, the maximum rainfall in an hour was 62 mm (1979) and the maximum daily rainfall was 139.9 mm (1981). According to the similarity ratio of rainfall intensity, the model rainfall intensity is 12.4 mm/h. Since the uniformity of rainfall device is poor at low rainfall intensity, and considering the uniformity of rainfall time, by controlling the hourly rainfall to be 12.4 mm, 9 h–18 h per day, the rain intensity was 24.8 mm/h. Observation and data collection are carried out during and after rainfall. In the evening, the model was covered with plastic cloth to keep moisture until the slope no longer changed significantly, and the experiment lasted for eight days. Three layers of sensors were arranged on the left and right interface of the original slope and the filled slope, with elevations of 1.0 m, 0.8 m and 0.4 m, respectively (Figure 2c,e). There were six moisture measuring points and nine pore pressure measuring points in total, and the collection interval of the two sensors was 1 min, which was mainly used to analyze the interface effect on the same horizontal line.

**Figure 3.** Location map of sampling points. (**a**) Sampling site, Shanxi Province; (**b**) loess fill in Qilipu community, Yan'an; (**c**) soil samples for test.

In order to simulate the interface effect between loess fill and original slope, plastic mesh is used as sliding zone in this test [22] (Figure 2d). After the completion of slope filling, marked nails are placed on the slope and the top of the fill slope in order to observe the deformation characteristics of the slope and the top of the fill slope and analyze the deformation and failure process of the fill slope (Figure 2f). Specifically, set for the top of the slope inserts two rows, slope inserts four rows and each row inserts five marked nails.

## **3. Results**

#### *3.1. Variation of Slope Water Content*

3.1.1. Variation of Water Content at Top of Slope (H1 = 1.0 m)

Figure 4 shows the variation curve of water content increment with time at different measuring points at the top of the slope (H1 = 1.0 m). In the LFSI and the filling slope, the variation of water content with time under the action of rainfall is similar: the water content first increases gradually, and then gradually decreases after a period of rainfall when it reaches a certain stable value or maximum value. Before the end of the daily rainfall, the increment of volumetric water content at the interface W2 is greater than that in the filling body W3, indicating that LFSI is an easy channel for rainfall seepage. The time of water content decline at W2 measurement point was later than that at W3 after the end of rainfall, indicating that part of the water will accumulate at LFSI and continue to permeate along LFSI after rainfall, and the water at W2 at the interface is supplemented, resulting in the lag of water content decline at W2.

**Figure 4.** Variation of water content in the following days, from the first to the eighth (H1 = 1.0 m).

3.1.2. Variation of Water Content in the Middle and Upper Slope (H2 = 0.8 m)

Figure 5 shows the variation of water content in the middle and upper slope (H2 = 0.8 m) (W5 is located at LFSI and W6 is located in the filling body). It can be found that the water content in the middle and upper part of the slope increases gradually with time under the action of rainfall. When it increases to a certain stable value or maximum value, it gradually decreases after a period of rainfall, which is basically consistent with the variation trend of W2 and W3 measuring points at the top of the slope. The increment of volume moisture content before the end of rainfall shows that water content W5 is greater than W6, indicating that LFSI has infiltration advantage. The reason for the abnormality on the third day may be that the water content at W6 increased for a period after the rainfall on the second day. It is speculated that cracks occur near W6, and the rainwater is easy to gather at the measuring point, resulting in the increment of moisture content of W6. At 17 h 26 min on the third day, a shallow slope slip occurred, and the stress inside the slope was redistributed, resulting in no obvious seepage advantage at W6. Therefore, the advantageous seepage at the interface is reflected again after the fourth day. The time of water content decline in W5 after the end of rainfall was later than that in W6, indicating that water distribution at the end of rainfall was collected at the interface and continued to infiltrate along the interface, supplementing water content in W5, resulting in lagging water content decline in W5.

#### 3.1.3. Variation of Water Content in Slope Toe (H3 = 0.4 m)

Figure 6 shows the variation of water content increment at different measuring points at the foot of slope (H3 = 0.4 m) with time (W8 is located at LFSI and W9 is located in the filling body). From the fourth day of rainfall, with the increase of rainfall time, the

increment of moisture content of W8 began to be greater than that of W9, which also indicates that there is a lag phenomenon of rainfall infiltration along LFSI.

**Figure 5.** Variation of water content in the following days, from the first to the eighth (H2 = 0.8 m).

According to the analysis of the above test results, the essence of the "interface effect" of the loess fill slope is that the interface is a rainfall dominant seepage channel, and there is a lag phenomenon of rainfall infiltration along the interface (Figures 4–6), that is, after the end of rainfall, part of the water will continue to be collected at LFSI and penetrate along LFSI.

#### *3.2. Variation of Pore Water Pressure of Slope*

Pore water pressure increment (the pore water pressure measured at each point at each time was subtracted from the pore water pressure before the test (9 h per day)) is adopted for analysis.

#### 3.2.1. Variation of Pore Water Pressure at Top of Slope (H1 = 1.0 m)

Figure 7 shows the curve of pore water pressure increment versus time at the top of slope (H1 = 1.0 m) (pp1 is located in the original slope, pp2 is located in the LFSI, and pp3 is located in the fill slope, where pp represents pore water pressure, the number is the corresponding measuring point position number). It can be found that the pore pressure at each measuring point increases sharply first and then decreases in the process of rainfall. The change of pore water pressure mainly comes from seepage. In the process of rainfall, rainwater infiltrates from the top of the slope and the slope surface, and seepage occurs inside the slope, which makes the pore water pressure gradually increase. As the infiltration of rainwater gradually forms a transient saturated zone on the surface of the slope body, the infiltration rate of rainwater decreases, resulting in a slow or declining increase of pore

pressure. Pore water pressure decreases gradually after rainfall due to the lack of water replenishment and evaporation of slope.

**Figure 6.** Variation of water content in the following days, from the first to the eighth (H3 = 0.4 m).

3.2.2. Variation of Pore Water Pressure in Middle and Upper Slope (H2 = 0.8 m)

Figure 8 shows the curve of pore water pressure increment versus time in middle and upper slope (H2 = 0.8 m) (pp5 is located in the original slope, pp6 is located in the LFSI, and pp6 is located in the fill slope). It can be found that the pore pressure at each measuring point gradually increases sharply, then increases slowly or decreases, and decreases gradually after the end of rainfall.

3.2.3. Variation of Pore Water Pressure in Slope Toe (H3 = 0.4 m)

Figure 9 shows the curve of pore water pressure increment versus time in slope toe (H3 = 0.4 m) (pp7 is located in the original slope, pp8 is located in the LFSI, and pp9 is located in the fill slope). As can be seen from Figure 9, pore water pressure at each measuring point first rises sharply, then rises slowly or decreases, and gradually decreases after the end of rainfall. The pore pressure data fluctuated significantly after the rainfall at pp9 measurement point on the fourth day, which may be caused by the mudflow destruction at the foot of slope on the third day and the rainfall on the fourth day, resulting in the development of cracks near pp9 measurement point, and the pore pressure oscillated significantly under the action of rainfall in the following days.

**Figure 7.** Variation of pore water pressure in the following days, from the first to the eighth (H1 = 1.0 m).

**Figure 8.** Variation of pore water pressure in the following days, from the first to the eighth (H2 = 0.8 m).

**Figure 9.** Variation of pore water pressure in the following days, from the first to the eighth (H3 = 0.4 m).

## **4. Discussion**

#### *4.1. Variation Characteristics of Water Content and Pore Pressure*

Figure 10a shows the daily maximum increment change curves of water content at W2 and W3 at the top of the slope (H1 = 1.0 m). It can be seen that the maximum daily increment of water content Δ*w* gradually decreases with the increase of rainfall days, and the cumulative increment value is 99.5% and 63.4%, respectively, which may be because with the increase of rainfall days, the soil at W2 and W3 is closer to the saturated water content [23], resulting in the decrease of the maximum increment, which is consistent with the model test conclusion of Crosta [24] and Li et al. [25]. In Figure 10b, W5 shows a similar law with the W2 and W3, and the cumulative increment value is 79.2%, while the daily maximum increment of water content of W6 in the filling slope increases first and then decreases gradually with the increase of rainfall days. It can be seen from Figure 10c that the daily maximum increment of water content at W8 and W9 increased first and then decreased gradually with the increase of rainfall days, and both reached the maximum value on the second day of rainfall, but the cumulative increment value is55.0% and 51.7%, respectively. It is speculated that this may be related to scouring damage in the middle of slope on the second day of rainfall.

Figure 11a shows the daily maximum increment change curves of pore water pressure at the top of the slope (H1 = 1.0 m). On the first to fourth day of rainfall, pp1, pp2 and pp3 near the top of the slope showed that the daily maximum pore water pressure increment Δ*μ* gradually increases and the increasing speed is slower with the increase of rainfall days. Under the action of rainfall, the daily maximum pore pressure increment at LFSI (pp2) near the top of the slope is greater than that of the fill slope (pp3), which once again shows that rainfall at LFSI is easy to infiltrate. As can be seen from Figure 11b, the daily maximum

pore pressure increments at pp4, pp5 and pp6 are characterized by a sharp increase at the beginning of the rainfall, then a steady increase, and finally an increase trend. The daily maximum increments at LFSI and within the filling slope are greater than those within the original slope (pp4 measuring point), while the daily maximum pore pressure increment at pp5 at LFSI is basically stable in the later stage of rainfall. It can be seen from Figure 11c that the daily maximum pore pressure increment of each measuring point pp7, pp8 and pp9 on the first to fourth day of rainfall gradually increases with the increase of rainfall days. Locally, the daily maximum pore pressure increment of pp9 at the foot of the fill slope first decreases, then increases and finally decreases, which is closely related to the process of water enrichment at the foot of the slope and water seepage after slump, and also consistent with the experimental results of Chueasamat et al. [26] and Orense et al. [27].

**Figure 10.** Daily maximum increment of moisture content at different measuring points (**a**) H1 = 1.0 m, (**b**) H2 = 0.8 m and (**c**) H3 = 0.4 m.

#### *4.2. Analysis of Slope Deformation Characteristics and Failure Process*

The increase of pore water pressure will cause slope failure during rainfall [28–30]. The foot of the filled slope is the most vulnerable place to damage under the action of rainfall [13,17,26,27], which is due to the low terrain at the toe of the slope, the rainwater is easy to collect, and the soil is soaked [17,27,31], resulting in the reduction of shear strength. At 14 h 25 min on the first day of rainfall, two mud flow failures occurred at the slope toe (Figure 12a). Figure 12b shows that there are many small-scale mud flow damages on the slope surface at 17 h 45 min after the end of rainfall on the first day, and local erosion damages occur in the middle and both sides. Figure 12c shows the picture of slope surface 18 h after the end of rainfall on the second day. The slope erosion continued to intensify on the second day of rainfall, and the erosion on the right side is the most serious. Obvious cracks appeared at 8 h 50 min before rainfall on the third day near the slope shoulder (Figure 12d), which may be related to the evaporation of water at night on the previous day. The crack L1 was about 18 cm from the slope shoulder, about 90 cm in length, and about 47 cm from the left side of the model box.

**Figure 11.** Daily maximum increment of pore pressure increment at different measuring points (**a**) H1 = 1.0 m, (**b**) H2 = 0.8 m and (**c**) H3 = 0.4 m.

At 17 h 26 min on the third day of rainfall, two large-scale shallow sliding failures occurred at about 0.85 m from the bottom of the slope (Figure 13a). The sliding width of the left side is about 74.2 cm and the right sliding width is about 65.8 cm, and the right sliding trailing edge was 22 cm more than the left sliding trailing edge. It can also be seen from slope monitoring landmarks 6-2 and 6-4 that there is also a small scale of sliding damage above shallow sliding. Under the action of long-term rainfall, shallow sliding failure is more likely to occur in the lower part of the slope, which is consistent with Wu et al. [32] and Kim et al. [33] model test. At 8 h 50 min before the rainfall on the fourth day, there were five cracks near the slope shoulder (Figure 13b), and four new cracks (L2, L3, L4 and L5) were generated on the slope surface and slope top near the slope shoulder. The length of L1 does not change, but the width increases; L2 and L3 are located on the slope, L2 is about 38 cm long, 22 cm away from the slope shoulder, and L3 is about 32 cm long, 35 cm away from the slope shoulder; L4 and L5 are located on the slope top, L4 is about 52 cm long, and the longest distance slope shoulder is about 14 cm and L5 is about 31 cm long, 12 cm away from the slope shoulder. The development of cracks indicates that the continuous action of rainfall produces the change of stress field in the slope [34,35], and the tensile stress is generated near the top and shoulder of the slope, which is easy to produce tensile failure [35,36].

**Figure 12.** The deformation characteristics and failure process of slope. (**<sup>a</sup>**,**b**) Mud flow at slope toe and slope erosion on the first day, (**c**) slope erosion on the second day and (**d**) slope cracks on the third day.

**Figure 13.** Large-scale shallow sliding failures. (**a**) Shallow sliding of slope toe, and (**b**) slope top and surface cracks.

Figure 14a shows the change of shoulder cracks in the first 8 h 50 min of daily rainfall from day four to day nine. Under the action of continuous rainfall, the cracks near the slope shoulder gradually increased, expanded and deepened, L4 and L5 cracks began to connect from the sixth day, and gradually developed in the later stage, but there was no significant change in the overall shape of the slope from the third day. Moreover, on the eighth day of rainfall, the slope body was still not damaged along the crack, which may be due to the blockage of the shear outlet of the sliding body on the slope shoulder by the mud flow accumulated at the slope foot (Figure 14b). Considering the gentle angle of the slope itself and no further change in the overall shape of the slope since the mud flow at the foot of the slope, the test was stopped after the rainfall on the eighth day.

**Figure 14.** Deformation and failure process of fill slope under rainfall. (**a**) Crack development process near slope shoulder and (**b**) shallow sliding process.

In summary, the deformation and failure mode of loess fill slope under rainfall is shallow slip, and the failure process can be summarized as follows: local mud flow failure at the toe of the slope → erosion in the middle of the slope → crack initiation on the shoulder of the slope → local slip on the slope → crack propagation on the shoulder of the slope → shallow slip on the shoulder of the slope. Under the long-term effect of rainfall, the loess fill slope of mountain excavation and city construction project will first appear small mud flow damage at the foot of the slope, and then cracks appear at the shoulder of the slope. Rainwater infiltrates along the interface between the original slope and the filling slope and cracks, resulting in a sharp increase in the internal moisture content and pore pressure of the filling slope, which reduces the strength of the slope soil. Especially, under the action of repeated rainfall, the shallow surface soil will suffer obvious damage by dry-wet cycle [36,37], rainwater is more easily infiltrated, and the shear strength of soil will be reduced [38,39], resulting in the shallow landslide disaster of the filling slope. Therefore, it is necessary to pay attention to the treatment of slope shoulder and interface, and strengthen drainage and ecological protection measures for loess fill slope of mountain excavation and city construction project [13].
