Figures 5a–c illustrate three-dimensional stress distributions and representative cross-sectional stress contour plots of the slope subsequent to tunnel excavation. In accordance with FLAC3D conventions, negative values correspond to compressive stresses, while positive values denote tensile stresses. The principal compressive stresses are primarily distributed throughout the slope.

Under gravitational loading, the magnitude of stress increases progressively with depth. The X-direction stress contour plot (Figure 5a) reveals a localized zone of tensile stress at the crest of the model, indicating a heightened susceptibility to tensile failure at the slope’s uppermost and surface regions. These results are consistent with field observations, where collapse deposits are predominantly concentrated at the hilltop, and crack development is more prevalent on the surface of the talus slope.

As depicted in Figure 5b, tunnel excavation disturbs the initial stress regime of the talus slope, resulting in pronounced stress concentrations within the vault and sidewalls of the surrounding rock mass (see Figure 5c). During excavation, the vault is prone to the development of tangential tensile stresses. If these tensile stresses surpass the tensile strength of the surrounding rock, structural damage may ensue, manifesting as tension cracks or even vault collapse.

Figure 6a depicts the total displacement of the slope subsequent to excavation. Comparative analysis of pre- and post-excavation displacements reveals an increase of approximately 1∼2 cm, with a notably pronounced increment at the tunnel portal. The deformation exhibits a radially attenuating pattern, demonstrating strong agreement with theoretical predictions.

Figures 6b,c display displacement contour plots at two representative cross-sections. The non-uniform displacement distribution elucidates distinct failure mechanisms: Section A-A reveals two potential slip surfaces, whereas Section B-B demonstrates pronounced deformation in proximity to the tunnel portal, indicating a heightened propensity for sliding along the delineated failure plane. Field observations further corroborate these numerical results; an arcuate surface crack (Figure 6d), approximately 12 m in length and 2∼3 cm in width, was observed to be nearly parallel to the slope inclination. The strong correspondence between field evidence and simulation outcomes affirms the predictive reliability of the numerical model.

Numerical simulations reveal a post-excavation displacement increment of approximately 1∼2 cm. Field measurements, obtained with a Leica TS30 robotic total station, indicate displacements reaching up to 5 cm at monitoring points 1# through 5# (Figure 3). Comparative analysis of numerical and in situ monitoring data (Figure 7) identifies two distinct displacement anomalies occurring at approximately 32 and 45 days after the commencement of construction. Notably, monitoring points 1# and 2#, located proximal to the tunnel portal, exhibited significantly greater displacement magnitudes than the more distal points 4# and 5#, in accordance with expected distance-dependent deformation patterns.

Among the monitored points, 1# and 3# exhibit optimal agreement between simulated and observed displacements, as evidenced by minimal fitting errors. During the 35∼45 days construction period, the fitting error for point 2# ranges from 5 to 8 mm, while points 4# and 5# demonstrate consistent errors of approximately 7 mm during days 45∼55. The strong correspondence between numerical simulation outcomes and field measurements substantiates the reliability of the model in predicting slope deformation trends under the combined influences of reservoir water level fluctuations and train-induced vibrations.

However, there is a difference between the local numerical simulation value and the actual monitoring,The possible reasons are as follows:

Parameter Variability of talus deposits: The parameters of a real talus deposits may vary at different locations on the slope, whereas the physical and mechanical parameters used in numerical simulation are constant values, which cannot fully and accurately reflect the characteristics of this heterogeneous material.

Under the influence of reservoir water level fluctuations, variations in Y-direction displacement are particularly pronounced in the vicinity of peak water stages, with displacements at the slope toe being appreciably greater in unreinforced sections than in reinforced counterparts (Figure 8a).

The majority of Y-direction displacements are negative, signifying downward deformation of the slope. Positive displacements observed in front of the anti-slide pile further corroborate the identification of this region as the deformed back edge of the slope. Z-direction displacements are negative in the central segment of the slope and positive at the front (Figure 8b), indicating subsidence in the middle and rear sections, and uplift at the slope front. Displacements within the reinforced zone are substantially lower than those observed in the unreinforced area. Critically, uplift deformation is mainly concentrated behind the anti-slide pile, demonstrating its effectiveness in mitigating the slope’s response to reservoir water level fluctuations.

The maximum total displacement occurs at the slope toe and progressively decreases with elevation (Figure 8c), suggesting that traction-induced deformation is primarily attributable to reservoir water level fluctuations.

Following reservoir drawdown, deformation is primarily concentrated at the slope toe, with total measured displacements remaining below 5 mm. In comparison to reservoir filling conditions, both the spatial extent of the deformation zone and the magnitude of displacement at the toe are substantially reduced. Reinforced zones exhibit markedly lower displacements relative to unreinforced sections, thereby substantiating the effectiveness of anti-slide pile reinforcement. Figure 9 demonstrates negligible deformation—approximately 1 mm—in the mid-slope and rear slope regions. Analysis of the displacement distribution reveals an arc-shaped pattern at the slope toe, with surface displacements exceeding those recorded within the slope mass. Collectively, these findings indicate that reservoir drawdown exerts a minimal influence on talus slope deformation.

Talus slope deformation demonstrates a pronounced correlation with variations in reservoir water level, whereby deformation increments during the reservoir impoundment phase substantially exceed those recorded during the drawdown period.

An increase in reservoir water level disrupts the antecedent equilibrium of the slope, necessitating progressive self-adjustment via deformation until a new limit equilibrium state is established. During intervals of constant water level at a given elevation, slope stability is maintained. In contrast, reservoir drawdown disturbs the slope equilibrium, instigating renewed deformation that gradually stabilizes in an asymptotic fashion.

Prior to reservoir impoundment, slope stability is principally governed by gravitational forces and topographic conditions, which impart a pushing effect. Post-impoundment monitoring reveals substantially greater deformation in the frontal region of the slope compared to the rear, indicating a shift from a gravity-dominated to a traction-controlled deformation regime. Figure 10 presents the predicted deformation at monitoring points 1 through 4. Due to the close proximity of monitoring points 4# and 5# in both field observations and numerical simulation domains, these were consolidated as point 4# for analytical consistency. The monitoring data exhibit a progressive deformation gradient: point 1# records the highest deformation, followed sequentially by points 2# and 3#, with point 4# displaying the lowest values.

For computational efficiency, the train is idealized as a sequence of moving loads, P(t), systematically distributed along the track as follows: P t = P 0 + P 1 ⁡ sin ⁡ ω 1 t + P 2 ⁡ sin ⁡ ω 2 t + P 3 ⁡ sin ⁡ ω 3 t P i = m a i ω i 2 ω i = 2 π υ / L i

P ₀ denotes the static wheel load, whereas P ₁ to P ₃ represent the amplitudes of dynamic vibration loads arising from train ride comfort, dynamic supplementary loads, and rail surface wave wear, respectively. ω _i designates the angular frequency; m refers to the unsprung mass; ai represents the vibration amplitude associated with each respective condition; Li characterizes the representative wavelength of vibration induced by these factors; and v indicates the train velocity.

Drawing upon both domestic and international research (Zhu and Lin, 2015; Wei et al., 2024),

P 0 = 97.5 kN , m = 750 kg , v = 120 km / h , L 1 = 10 m , a 1 = 3.5 mm , L 2 = 2 m , a 2 = 0.4 mm , L 3 = 0.5 m , a 3 = 0.08 mm .

The temporal variation of the train-induced dynamic load is depicted in Figure 11. This sinusoidal excitation function is imposed upon the subgrade structure to simulate the dynamic response elicited by train-generated loading.

Owing to the negligible variations in stress cloud patterns following the imposition of train loads, the shear strain increment was adopted as the principal analytical parameter. Computational outcomes for Condition 1 reveal that shear strain increments are concentrated in the talus slope proximate to the tunnel portal, displaying an approximately elliptical spatial distribution (Figure 12a). After a period of reservoir water level fluctuation, results for Condition 4 illustrate a shift in strain concentration to the back-pressure zone of the anterior slope fill, while retaining elliptical spatial characteristics (Figure 12b). Additional zones of strain accumulation are evident within the talus mass and the tunnel vault preceding the portal (Figures 12c,d). The magnitude of shear strain increments at a water level of 425 m exceeds that observed at 406 m. Collectively, the measured shear strain increments remain relatively minor, signifying that train-induced vibrations impart a limited influence on both slope stability and the integrity of the tunnel’s surrounding rock mass.

During tunnel operation, deformation is most pronounced at the toe of the talus slope and within the tunnel entrance section, as compared to the central and posterior regions of the slope. Comparative analysis reveals a 5 mm increase in displacement under Condition 3 relative to Condition 1 (see Figures 13a,b), while Condition 4 demonstrates a further 3 mm increase in displacement compared to Condition 3 (see Figures 13c,d). These findings corroborate that train loads exert measurable effects on both slope stability and the integrity of the tunnel’s surrounding rock mass. However, the relatively modest displacement magnitudes indicate that train-induced vibrations have only a limited impact on the overall structural performance.

The results demonstrate that fluctuations in reservoir water level represent the principal factor governing talus slope stability and deformation throughout operational periods, with especially pronounced effects observed during phases of water level ascent.

Monitoring data reveal that, under train loading conditions, all measurement points display minimal displacement variations, with recorded values consistently remaining within 6 cm (see Figure 14). Slope deformation at an elevation of 425 m is approximately double that observed at 406 m. The slope exhibits substantially greater instability when subjected to train loading under elevated water level conditions as opposed to low water level scenarios. Nevertheless, the relatively modest magnitudes of displacement collectively indicate that train-induced loading exerts only a limited influence on the global stability of the talus slope. The corresponding deformation prediction curves are presented in Figures 15a–d.