Study on the mechanism of water-induced degradation of slip zone soils and FDEM coupled simulation of slopes based on multi-scale characteristic

The evolution of the mechanical properties of slip zone soils had a greater impact on the structure of soil slopes with large upper porosity and excellent infiltration conditions. Especially under the rainfall effect, it was easy to be affected by the infiltration effect of internal injury, which will lay a ‘hidden danger’ for the slope body to trigger sliding in the secondary transformation process. In this paper, based on Haijiao ping landslide in Guizhou, SEM image recognition techniques were used to reveal the water-induced degradation mechanism of the sliding zone soil from multiple scales and the damage parameters were obtained. Meanwhile, the FDEM numerical model was established to simulate the stability of the slope coupled with indoor TCT test results. The results proved that the angle of internal friction decreased linearly with the increase of water content on the macroscopic scale, but the cohesive force showed an increase and then a significant decrease. The effect of matrix suction in the microscopic scale was significant at lower water content, the internal cohesion formed a large number of agglomerate structures to resist external deformation, but the microstructure was loose and porous after sufficient water immersion. The pore space spreads directionally and the area increases by 2.66 times. The cross-scale discrete-finite element coupled simulation method based on image recognition can visually respond to the macroscopic mechanical properties and stability change response of the slope body caused by microscopic damage. The water-induced degradation effect of rainfall on slip zone soils was the inherent factors for the initiation deformation of landslide. The artificial excavation was the external factor that triggered the slope to slide. This type of landslide was more concealed in its natural state and prone to deformation when excavated after long-term rainfall.


Introduction
In recent years, as people pay more attention to the stability of mountain slopes, the use of 'air-sky-ground' integration and the use of remote sensing satellites, drones, other multi-source data observation and numerical simulation of landslide interpretation research began to gradually promote [1,2]. The former is the application of new technologies and methods in the field of geological hazards, but the applicability of the scenario and the accuracy of the results need to be further considered; the latter is also a major obstacle to the accuracy of the model and parameter selection in the process and the authenticity of the results [3,4]. Therefore, for this type of hazards, it is also necessary to focus the research on the geological factor of the physical and mechanical parameters of the geotechnical body. This is also the key for the implementation of the two aforementioned methods. Slip zone soils are less strong and prone to plasticity due to the specific physicochemical conditions within the slip zone. The spatial and temporal variation of slip zone soil parameters is an intrinsic controlling factor for the variability of landslide formation development and damage. The probability of slope instability and the degree of instability hazard are closely related to the evolution of slip zone soil properties [5]. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Currently, many studies have been conducted on the effects of rainfall on slope stability. including indoor physical experiments, dynamic monitoring of the relationship between rainfall and landslide deformation, simulation analysis of rainfall-induced landslides [6,7]. Montgomery [8] and Jakob [9] proposed a more reasonable rainfall threshold based on the hydraulics and landslide stability model, considering rainfall and pore water pressure, permeability, matrix suction, landslide morphology, and bedrock structure. Frydman [10] conducted an indoor experimental study on the remodeled slip zone soil. The evolution of peak and residual shear strength was obtained by multi-case and different forms of interface shear tests. Kimura [11] concluded from tests and analysis that when the slip zone soil took the residual strength, the slip surface was partially in a completely softened state, while the result of rupture of the slip surface can be derived. Chen [12] studied the evolution of mechanical properties of slip zone soils in loess areas under rainfall conditions. The results show that the angle of internal friction decreased significantly with the increase of water content, while the cohesion also showed a decreasing trend. He [13] deeply explored the correlation between landslide triggering and rainfall by carrying out long-term displacement monitoring of the Zhangmu landslide in Tibet. Using the sustained rainfall increment as the loading dynamics parameter of the landslide, the evaluation parameters and prediction model of the rainfall dynamics increment displacement response ratio were established and determined. Fa [14] had deeply explored the correlation between landslide triggering and rainfall by monitoring landslide displacement in Zhangmu, Tibet. Tohari [15] concluded that landslide was a dynamic evolutionary system, once the surface cracks appear, it would form a dominant flow, which was more conducive to surface water infiltration to the deep soil body to trigger the deformation or landslide damage. Wu [16] concluded the characteristics of the strength parameters from the field calibration and indoor shear tests, and found that its shear strength was generally lower than that of the neighboring rock layers by comparison. Dong [17] used three kinds of indoor shear tests for the characteristics of slip zone soils, and concluded that the mechanical properties of slip zone soils differed greatly when the landslide was at different stages of development. Cao [18] conducted a sensitivity analysis study on the factors influencing slope stability and concluded that the mechanical property attenuation of slip zone soil has more influence on the slope with large angle inclination. Liu [19] analyzed how rainfall infiltration changed the physical and mechanical properties of slide soil and then studied the changes in slope stability under different rainfall conditions through numerical simulation. Duan [20] conducted an experimental study on the decay law of strength parameters under longterm immersion conditions of slip zone soils, and carried out a simulation analysis with actual slope projects. He concluded that rainfall increased the saturation zone of immersion of slip zone soils, causing slope instability. Gao [21] concluded that heavy rainfall was the key factor leading to the remote movement of landslides and used numerical simulation to show the progressive damage process of cascading landslides. B Legowo [22] detected the potential sliding surface of the slope using physical sounding and SW matrix algorithm. He carried out a study on the difference in characteristics and damage pattern of the sliding surface caused by the change of soil properties in the slip zone. Based on the research results, a more accurate prediction analysis of the stability of the landslide was made. Gao [23] considered the complex landslide body as an equivalent fluid and determined the frictional resistance of the substrate by different rheological models, which can better simulate the movement velocity and accumulation thickness of the landslide. Nguyen [24] found that the weakening of the support force at the foot of the slope led to sliding resistance of the slope. Under the coupling effect of strong rainfall and the cutting of engineering slope, the dynamic water pressure and sliding force increased, further leading to slope instability.
As can be seen from the above research results, a large number of research findings have been obtained by scholars for the quantitative evaluation of mechanical properties of slip zone soils and qualitative analysis of slope stability [25][26][27]. However, most of the results were based on indoor parametric tests or finite element simulations to evaluate the stability of slopes [28]. Few studies were conducted on the mechanism of hydrodynamic deterioration and mechanical response of landslide soils from multi-scale joint analysis, especially the correlation mechanism between the evolution of the mechanical properties of landslides and the change of slope stability across scales [29]. The existing research methods had certain limitations for a deep understanding of the causal mechanism of such landslides. Therefore, taking a reservoir landslide in Guizhou as an example, this paper used indoor scanning test and triaxial compression test (TCT) to quantify the mesoscopic damage degree of soil in the sliding zone by pore image identification technology for the first time. On this basis, a coupled finite-element-discrete-element model was established across scales to analyze the landslide instability mechanism. Finally, it was verified by combining with on-site geological investigation and monitoring data. The research results can provide reference for the disaster prevention and control of road and hydropower construction in the reservoir area.

Project overview
The study area was located in the south side of the reservoir bank erosion low mountainous terrain, the slope body was a single slope structure, narrow at the top and wide at the bottom. The top and bottom end was steeper and the middle was wider and gentler, the slope on the longitudinal gully development (see figure 1). The geological structure of the site was relatively simple, and no folded faults pass through. The stratigraphic lithology from top to bottom was cultivated soil, residual slope soil of the Fourth Series, old landslide accumulation, and gray to dark gray thin to medium-thick laminated sandy mudstone of the Permian Upper Xuanwei Group. The hydrogeological conditions in the area were generally simple, surface water was mainly seasonal water flow, and the borehole measurement showed that the buried depth of groundwater level was basically 0-20.4 m. The groundwater level in the study area of this paper is about 12 m. The on-site borehole and inclinometer had identified that the landslide contains a powder layer at a depth of about 10-11 m and the deformation was obvious. On-site seepage test showed that the pore connectivity of the mound and its upper part was permeable, and it is discharged in the form of infiltration or spring at the front edge of the landslide body. The survey results show that due to the good permeability of the landslide mound, a large amount of rainfall infiltrates and replenishes to the slide zone of the landslide body.
The field test results were shown in figure 2. The most obvious changes in the data of the borehole are located at 11-12 m and 17-23 m below ground, according to the drilling data, the thickness of the subsoil layer in this area is 13.4 m, so the potential sliding surface of the borehole is located at 11-12 m below ground, and the largest displacement is located at 11 m below ground. To obtain the permeability coefficients of the clay containing gravelly powder, the double-ring percolation test method was used. A test pit of 1 m * 1 m and 0.5 m deep was excavated on the ground surface. Two iron barrel rings (50 cm high, 0.5 m outer diameter, 0.25 m inner diameter) were placed concentrically and inserted into the soil layer for 8 cm, a steel straightedge was placed inside the 2 barrels, and the inner and outer water rings were kept at the same depth of 10 cm. When recording, control the length of water release. Initially 5 min record once, record the time interval and the amount of water seepage after becoming 10 min once. Flow observation accuracy to 0.1 L, stable (twice in a row) the difference between the observed flow is not more than 5%, and then measured 2-3 h end, take the last injection of water as the calculated value. Finally, the permeability coefficient is calculated by Darcy's law.

Experimental design of slip band soil
The soil for the indoor test was taken from the slope slip zone, and the specimens were prepared by the dry soil method (technical lines of research was shown in figure 3). The specimen size was a standard cylinder with a height of 10 cm and a radius of 2.5 cm. Three valid specimens were prepared for compression in each group (according to specification of soil test 'SL237-017-199'. The moisture content of the natural soil sample was measured to be 10% in the previous foundation geotechnical test. In order to fully consider the water deterioration effect of the slip zone soil, the water content (WC) of the indoor test was set at 12%, 14% and 16% (as shown in table 1). Considering that the slope body was prone to infiltration and accumulation of rainwater in the area near the slip zone soil after rainfall, which may cause slow deformation and damage. The triaxial compression tester, which can strictly control the drainage conditions of the specimens and had a clear state of stress, was used for the triaxial compression test. The stress-strain data were obtained by the data acquisition system in the instrument, and the shear strength parameters of the specimens at four moisture contents were obtained by the Moore Coulomb model. Finally, the soil samples in natural state and at 16% water content were selected for SEM microscopic testing and analysis. The microcosmic tests were conducted on S-3000N scanning electron microscope. The scanning angle is between −20°and 90°and the magnification ranges from 5 to 300,000. The accelerating voltage and probe current are 0.5~30 kv and 10 −12~1 0 −8 A, respectively. The supporting pre-treatment instrument for specimen was the E-1010 carbon coater. The tests were followed the principle of high magnification focus and low magnification photography.

Features of shear curve
As seen in figure 4, the peak shear stress of the specimens maintained a linear growth relationship with the surrounding pressure regardless of the variation of the moisture content. When the surrounding pressure increased to 300 kPa, the peak shear force also reached its maximum value. The increase of the surrounding pressure can further compress the internal pore space of the geotechnical body by affecting the arrangement relationship of the soil particles. The enhancement of skeletal properties brought by the particle crowding can effectively resist the external shear action. It is obvious from figure 4 that the stress-strain curve of the soil exhibited the development characteristics of strain hardening when the water content exceeded 14%. When the water content was lower, the curve trend showed a rise in the early stage and a slight decline in the later stage. The increase of the surrounding pressure can effectively mitigate the reduction of the mechanical properties of the soil under the effect of water deterioration.

Parameters of shear strength
The shear strength of the specimens obtained from the triaxial compression test were analyzed in figure 5. For strain-reinforced soils, in order to effectively control the soil deformation in practical engineering, the strength  corresponding to no more than 0.2% strain was usually taken as the yield strength, which is the damage stress difference in the triaxial test. The peak variation law of the principal stress difference under different surrounding pressure and moisture content was shown in the following figure. When the confining pressure was greater than 200 kPa, the change was small although it increased under the condition of increasing surrounding pressure. When the water content of soil fluctuates in the range of natural water content, it increased slowly with the increase of water content. But its increase was obviously limited by the surrounding pressure. When the water content exceeds 14%, it decreases significantly with the increase of water content. It was also shown in figure 6 that the variation of soil moisture content was closely related to the evolution of shear strength parameters of slip zone soils. The angle of internal friction showed a steady decrease with the increase of water content, but the maximum attenuation was less than 10%. The cohesive force increases and then decreases with the increase of water content, and the two were quadratic functions. When the moisture content was between 10% and 12.6%, the cohesive force increased slowly. When the water content exceeded 12.6%, the cohesive force decreases, and the variation is more significant than that of the rising section. The equations were fitted to the results based on the Moore-Coulomb criterion and the experimental data as follows.

( )
Where c was the cohesion/kPa, j was the angle of internal friction/°, s was the normal stress/kPa, t was the shear stress/kPa, W was the moisture content/%, and a-e wass the fitting parameter to be determined.  Substituting equation (2) into equation (1) The analysis of the fitted expressions showed that the shear strength of slip zone soils decays after the water content reached a certain level, which was more consistent with the mechanical response law of the soil derived from the test, so the prediction formula had physical significance to some extent.
In summary, it can be seen that the shear resistance of the soil body appeared to be significantly reduced after the water content of the specimen exceeds 12.6%. The two mechanical strength indexes obtained from the test also further verified that higher water content of the slip zone soil had a large negative impact on the development of the bearing and deformation resistance of the soil. In order to further investigate the essential reasons for the decay of the mechanical properties of slip zone soils, it was necessary to select specimens under two extreme water content conditions for comparative microstructure analysis.

Microstructure and image recognition of slip band soil
Electron microscope scans of the samples under two conditions of 10% and 16% water content were analysed (see figure 7). In the natural state, the soil particles were fully cemented and the soil skeleton was intact, but micro-fractures still developed in local areas. As the water content increased, the bond between the soil particles weakened. The structural porosity increased and gradually showed the loose state, while the original interparticle superposition disappeared. Under the effect of prolonged water immersion, some of the cohesive particles were transported away and accumulated disorderly in the vicinity of the surrounding particles and pores. The surface of the sample appeared locally bumpy and wavy. In addition, the internal and surrounding media and the voids between them increased. This was mainly because the powder particle skeleton contained more hydrophilic minerals such as montmorillonite or illite, which changed the micro-metallic structure upon contact with water, making the soil appear softened by water immersion and dry cracked by water loss. In fact, the matrix suction effect within the soil body was often significant at low water content. At this time, soil saturation was low and the internal cohesive formation of the agglomerate structure was more able to resist compressive deformation. This was also one of the intrinsic reasons for the higher peak shear force of the soil body under the action of low water content.
To further quantify and analyse the microstructural evolution pattern of the slip zone soil under long-term water immersion, the images were subjected to a series of processing such as grey scale comparison, binarisation and noise reduction, and then the statistical analysis of the average pore size was carried out (see figure 8). It can be seen that a certain amount of pore distribution (about 494.82 pixels) existed inside the slip zone soil in the natural state. However, as the water content increased, the number of pores increased rapidly. When the water content reached 16%, the average pore area was about 1808.84 pixels, which was about 2.86 times that of the natural state. The variation pattern is consistent with the relationship between internal friction angle and water content, reflecting the macroscopic characteristic of decreasing shear strength parameter. In addition, the aspect ratio of pores at different water contents showed that the pore aspect ratio gradually decreased with increasing water content, with a maximum decrease of 8.8%. It indicated that the trend of fracture directional propagation became more obvious, and it was easier to develop and connect under loading. The maximum and minimum values of the pore space and its average pixel size were obtained by automatic identification of the statistical fineview parameters of the specimen.

Numerical simulation and analysis
According to the site investigation, the landslide started before local rainfall occurred. Rainwater penetrated downward into the slip zone soil, but the slope did not produce obvious sliding at this time. However, when the leading edge of the landslide was excavated, the slope produced significant and rapid deformation. So the rainfall condition was considered as the change of moisture content of the soil in the slide zone, and four conditions of 10% (condition A), 12% (condition B), 14% (condition C) and 16% (condition D) were set for simulation.

Model establishment and parameter selection
In order to further characterize the changing characteristics of the slip zone soil after water immersion, the response law of the slip stability under the action of this factor was studied. Combined with the aforementioned electron microscope scanning results, the fourier reconstruction method was used in the discrete cell. The highstage coefficients were obtained by equation fitting, and the low-order coefficients were derived by separate statistics, thus realizing the reconstruction of the soil unit considering the pore characteristics. The spatial distribution location of the pores was established based on the identification of the scan of the slip zone soil, and the random reconstruction threshold of the pores was determined in combination with the geometric parameter distribution characteristics of the pores in figure 9. Finally, the image stitching was combined with the size of the study site and imported into the particle flow software for model reconstruction. The model size at a certain water content was cut out randomly from the assembled image. Using the SEM image processed in formal step  and different soil parts with a certain water content, the development procedure of this study was established. The generated PFC2D model was executed.
A two-dimensional FDEM model was used to evaluate the study area based on a typical engineering geological section. Considering the complexity of discrete-continuous coupling calculation, the model was somewhat simplified and the local discontinuous stratum was merged. The horizontal distance of the model was 368 m and the height was 135 m. The location of the sliding zone of the side slope body was initially determined through preliminary survey and borehole testing, and the thickness was 0.8-1.5 m. The X and Y direction constraints were set at the bottom of the model, the X direction constraints were set at the left boundary, and there were no constraints on the slope surface. The continuous cells were discretized by 2D mesh using hybrid mesh generator, and the model had 11242 cells and 11134 nodes. The discrete cell had 8872 particles and the parallel bond contact model was selected. The coupled model was shown in figure 10. The values of slip gravel soil were combined with the results of field survey and foundation geotechnical test (see table 2), and the permeability coefficient was taken as K = 7.6 × 10 −5 m s −1 according to the field experimental results, while the slip zone soil was assigned based on the calibration of fine parameters of triaxial test results with different water content (see table 3). The contact parameters are mainly divided into deformation parameters (young's model, poisson) and strength parameters (cohesion, frcition angle, tension). Before the calibration of individual mineral parameters, the deformation parameters are first calibrated. The target elastic modulus and Poisson's ratio are matched by adjusting the parameters of fj_emod and fj_kratio. After the calibration of the deformation parameters, the contact strength parameters are then adjusted. In general, the tensile strength of the specimen is more influenced by the parameter of fj_ten, and higher tensile strength is obtained by increasing the value of fj_ten. The final compressive strength is generally influenced by the parameters of fj_coh and fj_fa , but fj_fa mainly affects the triaxial test results and has little effect on the uniaxial test results. In the process of calibrating the strength parameters, the deformation parameters also need to be fine-tuned.

Microcrack evolution of slip band soil
The analysis of the simulation results after excavation at the foot of the slope showed that when the slip zone soil was not affected by infiltrating rainwater and a small amount of water, the slip zone area had very little crack  distribution and the overall stability was good. With the gradual increase of the water content of the slip zone soil, the slip zone area was subjected to large concentrated shear stress under long-term infiltration action in water and gravity-induced compression shear. This phenomenon became more and more obvious after the water content exceeded 12%. Combined with the crack monitoring data, it can be concluded from figure 11 that the number of cracks in the slip zone area increases only about 6 times when the water content was 12%. However, when the water content reached 16%, the number of cracks in the slip zone area increased 220 times compared with that in 10% water content condition. It can also be seen from the figure 12 that the cracks developed gradually and penetrated rapidly as the water content increased. It was noteworthy that the compression-shear cracks at the front edge of the slope and the shear cracks at the back edge were very dense. Finally, it was easy to trigger the sliding of whole slope under the action of gravity and external disturbance.

Change in slope displacement
Combining the horizontal displacement of the slope and the overall displacement vector map (with 16% water content as an example), it can be seen from figure 13 that the front edge moved upward diagonally as the downward misalignment of the back edge of the slope occurred. Under the effect of topographic influence, the overall backward sliding occurred. It can be seen from the horizontal displacement cloud map that the horizontal displacement of geotechnical body near the excavation surface was the largest, especially there was an  obvious trend of displacement toward the prograde at the excavation slope corner. When the locking section at the foot of the slope was destroyed, the pressure will induce large traction type slip. With the increase of water content of soil in slip zone, the horizontal displacement also increased significantly, and the maximum increase of horizontal displacement was 38% compared with that in natural state. Combined with the previous analysis, the back edge of the slope will be fully pulled and cracked down after the slope cracks are fully penetrated, and the front edge will be sheared out under the action of gravity and final damage will occur. This is more consistent with the crack division pattern found in the field investigation.

Stability analysis
A comprehensive analysis of the change in slope stability under the conditions of excavation and different submerging degrees of sliding zone soil showed that (see figure 14) when the soil samples were under natural moisture content (i.e., it was not affected by rainfall) and the leading edge was not excavated, the overall stability of the slope was relatively good. However, as the water content of the soil in the slip zone increased, the stability of the slope began to change. From the performance of stability coefficient, it showed a trend of slow decrease first and then rapid decrease. When the water content of slip zone soil exceeded 14.8%, the slope stability started to decrease significantly and the slope of the curve increased. When the slope body was affected by both rainfall and excavation at the foot of the slope, the safety coefficient decreased significantly. After the excavation at the foot of the slope, the slope tended to slide when the moisture content of the soil in the slip zone reached about 12.5%. As the water content continues to increase, the stability of the slope decreases more significantly than when it was affected by water deterioration alone. From the above analysis, it can be concluded that the stability of the landslide was controlled by both excavation at the foot of the slope and the deterioration of the mechanical properties of the sliding zone soil due to rainfall infiltration. Based on the analysis of the influence characteristics and degree of the two factors, it was concluded that the deterioration of soil properties in the slide zone caused by rainfall infiltration in the early stage was the basic internal factors for the initiation deformation of landslide, while the excavation of the leading edge of the slope is the trigger factor for the initiation of the landslide. Therefore, slope stability is more sensitive to excavation at the foot of the slope. The sensitivity is even greater for excavations at the foot of the slope where the softening effect of the slip zone soils has already occurred. Combined with the preliminary site investigation, before the landslide started to slide, it experienced a small amount of local rainfall, but at this time there was no obvious phenomenon such as pulling crack and slope displacement. After the excavation at the foot of the slope, the trend of the slope movement became more and more obvious, which was also more consistent with the results of the on-site borehole diagonal measurement and indoor analysis. Ultimately, it is worth noting that the method proposed in this paper is mainly applicable to the structure of slopes with large upper porosity and excellent infiltration conditions.

Conclusion
In this paper, the Haijiao Ping landslide was taken as the research object. Based on detailed field investigation data and a large number of indoor tests, a macro-mechanical calculation model based on the microstructural characteristics of the slip zone soil was established using the coupled finite-element-discrete-element method. The decrease regularity of the strength parameters of slip soil because of the influence of moisture was fully considered. By carrying out systematic analytical and numerical simulation research work on the Haijiao Ping landslide, the following results were obtained: 1. When the moisture content was in the range of 10% to 16%, the internal friction angle of powder soil is slowly decreasing with the increase of moisture content, and the decrease is within 10%. The cohesive force showed the change rule of rising first and then falling. When the water content was about 15%, the cohesive force reached the maximum value. The variation of main stress difference with water content was similar to that of cohesion.
2. The effect of matrix suction in the soil was significant at lower water content, and the internal cohesion formed a large number of agglomerate structures to resist external deformation. However, some of the cohesive particles are transported and disordered to accumulate in the vicinity of the surrounding particles and pores under the effect of long-term water immersion. Because the powder particle skeleton contained more hydrophilic minerals, it would weak the bond strength among the soil particles, and the microstructure will also become looser.
3. With the increase of water content, the damage of slip zone soil gradually intensified. The pore length to slenderness ratio gradually decreased, and the maximum decrease was 8.8%. The trend of fracture directional spreading was obvious. In addition, the number of pores increased rapidly and the pore size increased by about 2.66 times, which was consistent with the macroscopic mechanical property decay.
4. The results of cross-scale discrete-finite element coupled simulation based on image recognition showed that the number of cracks in the slip zone increased 6-220 times with the influence of rainfall infiltration, and the displacement in the prograde direction increased significantly. The development evolution and penetration of cracks in the slip zone triggered the sliding of the slope.

5.
Simple excavation or rainfall infiltration was not the main reason for the initiation of this type of landslide directly. On the one hand, excavation in the lower part of the slope will induce traction damage to the soil landslide; on the other hand, the mechanical properties of the slip zone soil will be decayed under the action of pre-infiltration. The deep slip zone of the slope will be damaged and the stability of the slope will be slowly reduced. When the moisture content of the slip zone soil exceeded 14.8%, the decline of slope stability after excavation was more significant than when it was affected by water degradation alone, and the coupling effect of the two will directly trigger the slope body to pull and deform or even slide.