Landslides are serious geohazards that occur under a variety of climatic conditions (Medina et al., 2021; Fang et al., 2023), especially in the Three Gorges reservoir area of China (Guo et al., 2020). The tested soil mass was taken from Xietan, Zigui County, Hubei Province. It is composed of small soil grains and gravel. The soil mass is purple-red and has a high content of gravel grains. The soil sample is shown in Figure 1.

Particles with a maximum diameter of 60 mm are suitable for the shear apparatus used in the tests. The mass of particles with a diameter greater than 60 mm in the undisturbed sample accounts for only 0.71% of the total. Therefore, particles with a diameter larger than 60 mm were excluded by the exclusion method to obtain the grading curve, as shown in Figure 2. According to the standard for engineering classification of soil (GB/T 50145-2007, 2007), the particle with a diameter larger than 2 mm was defined as stone, and the particle smaller than 2 mm was defined as soil. The test soil sample had a stone content of 65.88% and a soil content of 34.12%, belonging to coarse-grained soil consisting of silty clay and gravel.

In order to qualitatively analyze the possible engineering properties of the gravel soil, a systematic test of its basic physical and mechanical properties was carried out. The basic physical and mechanical indexes of the test soil sample are shown in Table 1.

Direct shear and simple shear tests are two important methods to study the mechanical properties of the interface. Simple shear apparatus has a ten-layer stacked ring, which corresponds to the upper shear box of direct shear apparatus, and to some extent, it overcomes the shortcoming caused by the restriction on shear failure surface in direct shear apparatus. In this research, the large-scale DHJ50 type shear apparatus developed by Sichuan Huaxi Geotechnical Research Institute was applied. It has a diameter of 504.6 mm and a height of 400 mm. The reconstituted gravel soil sample was used. Its dry density was 1.6 g/cm³, and the moisture content was 7.41%. Normal stresses were set to be 100 kPa, 200 kPa and 400 kPa. The horizontal shear force was applied at a constant shear rate of 0.88 mm/min. The horizontal shear displacement and normal deformation were measured by displacement transducers arranged horizontally and vertically, respectively. The test was performed according to the Specification of Soil Test (SL237-1999, 1999). The working principle of the apparatus is shown in Figure 3. The bottom concrete cushion block and the surface concrete block are shown in Figures 4, 5 separately.

The roughness of the structural surface has a significant impact on the mechanical properties of the interface. In order to better simulate the contact between soil and concrete, in this research, a new method for making lower box concrete test block was proposed on the basis of the conventional direct shear apparatus (Du et al., 2016). Specifically, two steel molds were prepared for pouring concrete cushion block and surface block; hoisting steel bars were pre-embedded into the concrete. In this way, the problems in concrete preparation, loading, and unloading, and replacement could be solved.

The test results show that the two test methods will affect the shear stress-shear displacement relationship curve. Since the direct shear test limits the location of the interface, with the increase in shear displacement, the shear failure surface grows, and then coalescence occurs. The shear stress decreases, and the curve shows strain softening. During the simple shear test, the upper shear box does not produce rigid constraints on the soil sample. The upper box soil deformation increases with the increase of the shear displacement, which is a gradual failure process. Since the shear strength of the soil-concrete interface is greater than that of the soil sample, shear failure often occurs in a certain range of the soil mass. As the shear displacement increases, the shear stress increases, and the curve shows strain hardening.

Under different normal stresses, the initial shear modulus, and peak shear stress of the direct shear test are significantly larger than those of the simple shear test. The peak shear stresses of the direct shear test and simple shear in Figure 6 were extracted, and the chart column of them is shown in Figure 7. It can be seen that the peak shear stress of the direct shear test is about 30% larger than that of the simple shear test.

Figure 8A, B show that both test methods produce a shear contraction effect, and the trends of normal displacement at different sites of the upper surface of the soil sample vary. Due to the combined effect of normal and horizontal loads, there is friction between the soil-concrete interface, which causes the soil grains to move with the lower shear box. In the direct shear test, due to the rigid constraint of the upper shear box, the movement of the soil grains is limited to a certain height above the interface. With the increase of horizontal displacement, the upper surface of the soil mass will eventually show a stable tilt state, with the left being lower than the right. The simple shear test does not restrict the failure surface. There is not only soil grain movement on the interface but also aggregate movement inside the soil sample, which effectively compresses the internal voids of the soil sample. At 40 mm of the shear displacement, the shear contraction effect of the simple shear test is 31% larger than that of the direct shear test, as shown in Figure 8C.

C1ough and Clough and Duncan, 1971 fitted the sand-concrete direct shear test results and concluded that the relationship between the shear stress and shear displacement of the interface could be described by a hyperbola. Since it has only a few parameters and they are easy to determine, and its numerical calculation is easy, it is still widely used. The specific expression is shown in Eq. 1. τ = ω 1 k I γ ω σ n / P a n + R f ω σ n ⁡ tan ⁡ φ + c (1) Where τ is the average shear stress; ω is the relative shear displacement of the interface; k _I is the stiffness coefficient; n is the stiffness index; γ _w , and P _a are the unit weight of water and standard atmospheric pressure, respectively; R_f is reduction factor.The direct shear test data were fitted with the hyperbola model to determine the parameters. The results are shown in Table 2. The fitting curves and test curves were compared, as shown in Figure 9. It can be seen that the hyperbolic model can well represent the strain hardening characteristics of the interface.

The soil mass is a cylinder with a diameter of 50 cm and a height of 20 cm. The concrete is a 70 cm × 70 cm × 20 cm rectangular body. The soil mass and concrete both adopted the C3D8R solid element. The number of soil and concrete grid cells was 2023 and 784, respectively. The three-dimensional model is shown in Figure 10. Constraints in X and Y directions were set on the sides of the soil mass; 100, 200, and 400 kPa normal stresses were applied to the upper surface of the soil mass, respectively; constraints in Y and Z directions were set on the bottom surface of the concrete; horizontal displacement load of 80 mm was applied to the sides of the concrete. Three analysis steps were set in ABAQUS: geo-stress balance, normal pressure application, and horizontal displacement load application.

The soil mass adopted the Mohr-Coulomb elastic-plastic model, and its parameters were determined according to the direct shear test. The Poisson’s ratio of gravel soil is generally 0.15–0.25. It can be found from the trial calculation that Poisson’s ratio has little influence on the calculation result of the interface (as shown in Figure 11). Therefore, Poisson’s ratio of soil was set as 0.23. According to the empirical elastic modulus E = 2.0–5.0 Es, and the comparison of trial calculation results (as shown in Figure 12), the elastic modulus was determined as 15 MPa. The mechanical parameters of the materials are shown in Table 3.

Figures 13, 14 show the nephogram of the stress and displacement before and after the initial geo-stress balance. It can be seen that the magnitude of the stress after geo-stress balance remains unchanged, and the magnitude of the soil displacement reaches E−10(m). Therefore, the results of the geo-stress balance are regarded as reasonable, and it has little impact on subsequent analysis.

Without considering the tangential and normal coupling effects, the interaction between interfaces includes normal, and tangential actions. ABAQUS applies hard contact in the normal direction; in other words, the normal pressure can only be transmitted when the two objects are in a compression state, while no normal pressure can be transmitted when there is a gap between the two objects. Only when there is contact pressure on the interface can the interface transmit tangential force. If the tangential force is less than a certain limit value, ABAQUS regards the interface to be in a bond state; if the tangential force is greater than the limit value, the interface begins to show relative sliding deformation.

In this paper, the friction model, and hyperbolic model were applied in the tangential direction to carry out a comparative analysis. The friction model is a conventional processing method and does not consider the cohesion of the interface. Its basic formula is τ _crit = μp (μ is the friction coefficient; p is the normal contact pressure). The coefficient of friction was 0.7455, obtained by fitting the shear strength relation curve of the direct shear test. The hyperbolic model can be easily realized by invoking the FRIC subroutine with ABAQUS. The basic formula is shown in Eq. 1, and the calculation parameters are determined according to Table 3.

Figure 15 shows that the numerical calculation results of both models are smaller than the results of the direct shear test. The friction model is an ideal elastoplastic model, and its calculation results deviate greatly from the experimental data. The hyperbolic model can better simulate the non-linear relationship between the shear stress and the shear displacement. Its calculation effect is better than that of the friction model. According to the Specification of Soil Test (SL237-1999, 1999), the peak stress or shear stress at the site where the shearing displacement reaches 1/15–1/10 of the sample diameter is taken as the shear strength. In this paper, the shear stress at 40 mm (1/12.5 of the sample diameter) was taken as the shear strength. The errors of different calculation models compared with the direct shear test are shown in Figure 16. It can be seen that the error of the hyperbolic model is smaller than that of the friction model. For the numerical analysis of actual engineering contact problems, the hyperbolic model is recommended since it can reflect the non-linear relationship between stress and strain. However, the soil-structure interaction force under high stress is about 16% smaller than the actual value and should be adjusted accordingly or locally strengthened.

According to the numerical calculation results, the distribution of the central node compressive stress along the horizontal path under the normal stress with different shear displacements w _s was extracted, as shown in Figure 17. The results show that the distribution of compressive stress on the interface is uneven during the shearing process. Figure 17A shows that the compressive stress of the interface is about 100 kPa when the shearing is not started; Figures 17B–D show that with the increase of the shear displacement, the compressive stress at the middle of the interface gradually increases while that at the sides gradually decreases and show a unimodal shape. The compressive stress in the middle reaches about 400 kPa. As the shear displacement further increases, the overall compressive stress decreases and presents a bimodal shape, as shown in Figure 17E. Due to the shear contraction and boundary effects during the shearing process, the compressive stress on the left is greater than that on the right, and the wave crest is finally located in the middle to the left, as shown in Figure 17F.

The critical shear stress is the critical value when the interface node transits from the bond state to the slip state. The critical shear stress is always greater than or equal to the actual shear stress. When the shear stress is smaller than the critical shear stress, the node is in a bond state and is subjected to static friction; when the shear stress increases to the critical shear stress, the node is in a slip state, and is subjected to sliding friction. Figure 18 reveals that shear failure of the interface is a progressive process from the edge to the interior (Yin et al., 1995). Figure 18C shows that the edge nodes on the sides first reach the critical shear stress and show slip failure. As the shear displacement increases, the slip failure gradually develops towards the middle, as shown in Figure 18D. When the critical shear stress curve and the actual shear stress curve completely coincide, as shown in Figure 18E, the interface shows integral slippage and coalescence of the failure; after that, the shear stress of the interface is only affected by the compressive stress of the interface.

During the shearing process, the soil grains on the interface will exhibit two motion states: translation and rotation. Under the action of shear stress, the left soil grains move to the right. The right soil grains are not only subjected to the shear stress but also pushed by the left soil grains. Due to boundary constraints, the soil grains accumulate at the right; with further increase in shear stress, the soil grains on the right side of the interface will squeeze and roll upwards, and the grains inside the soil sample will also rotate upwards. Eventually, the upper box soil mass will show a tilt state with the left lower than the right, as shown in Figure 19.

The structure drives the movement of soil grains on the interface through the shear stress, which affects the size, and distribution of the compressive stress and shear stress on the interface. According to the relationship between the actual shear stress and the critical shear stress, the interface will be in a bond state or slip state. The movement of soil grains and the size and distribution of the compressive stress on the interface directly affect the shear stress of the structure on the soil. The shear stress, in turn, drives the movement of the soil grains. In this way, the soil-structure interaction is gradually formed. The evolution process is shown in Figure 20.