Instability Mechanism of Shield Tunnel Face Induced by Seepage and Soil Softening in Water-Rich Silty Sand: Case Study of Jingu-Haihe Tunnel

Abstract

The coupling mechanism involving high-pressure seepage and soil degradation regarding the face stability in water-rich silty sand environment remains to be comprehensively elucidated. This paper employs 3D fluid–solid coupling simulations to investigate these interactions taking the Jingu-Haihe Tunnel as a case study, and the dry and saturated hydraulic environments alongside three softening scenarios are set. Results indicate that hydro-mechanical coupling significantly compromises face stability, elevating the limit support pressure from 140 kPa in dry mechanical state to 231 kPa. The failure mechanism transitions from localized “horn-like” shear bands in dry states to global quasi-symmetric “bulb-like” visco-plastic diffusion in saturated seepage field scenarios. Softening effects cause stress-dependent stiffness degradation, increasing the deformation rate by 53.8% under low support pressure, and inducing uneven deformation where the crown displacement increases by 32.8 times, exceeding the 11.8-fold increase at the center as the support pressure drops from 600 kPa to 100 kPa. Moreover, the fluid–solid coupling effect amplifies the stratum’s sensitivity to shear strength parameters by up to 26 times at the face center compared to the dry condition. These findings may offer theoretical insights for optimizing support pressure determination in deep-buried saturated excavations.

1. Introduction

Driven by rapid global urbanization and the urgent need to alleviate traffic congestion, the development of underground space—encompassing metro systems, subsea, and river-crossing tunnels—has become a cornerstone of modern infrastructure [1,2]. While closed-face technologies like slurry shields offer high efficiency and minimal ecological footprints, the current trend toward greater burial depths and larger diameters poses formidable challenges to maintaining the stability of the tunnel face [3,4]. In deeply buried environments, especially when traversing water-rich weak zones or complex fault-impacted strata, the inevitable redistribution of stress fields and the accumulation of excess pore water pressure significantly compromise the integrity of the excavation face [5]. The complex interaction between high groundwater pressure, seepage forces, and the soil-arching effect increases the risk of face instability, ground loss, and catastrophic mud inrushes [6,7]. Therefore, it is essential to conduct a reliable analysis of tunnel face stability under high water pressure conditions and accurately predict the limit support pressure required during excavation.

The evaluation of tunnel face stability has traditionally relied on three primary methodologies: theoretical analysis, numerical simulation, and experimental testing [8,9,10,11]. Foundational theoretical frameworks, notably the wedge-silo model developed by Horn [12] and the plastic limit analysis proposed by Leca and Dormieux [13], provided the initial physical basis for calculating limit support pressures. Over the past decades, these models have been iteratively refined, through the development of multi-piece truncated cone mechanisms and discrete point-generated methods, to better account for diverse soil layers and non-uniform pressure distributions [14,15]. Building upon these theoretical foundations, investigations into fluid–solid coupling have advanced significantly, with notable applications in major global projects [16,17]. These studies have improved the assessment of tunnel face stability in saturated soils by incorporating seepage flow fields derived from numerical simulations [18,19]. While previous studies have provided a robust foundation for analyzing seepage forces, theoretical models often adopt the simplifying assumption of a constant soil skeleton. This idealization may not fully capture the simultaneous degradation of mechanical properties induced by pore water pressure. Consequently, further investigation is warranted to precisely quantify the instability mechanism in deeply buried slurry shield tunnels traversing water-rich strata. In this context, numerical simulation has emerged as the most robust approach for investigating fluid–solid coupling effects. Unlike theoretical solutions constrained by simplified assumptions or physical experiments limited by scale effects, numerical methods offer superior capability in rigorously implementing full-scale hydro-mechanical interactions [20,21].

In the tunnel excavation process, parameter sensitivity analysis has emerged as a fundamental instrument for risk mitigation [22,23]. Extensive research has been dedicated to evaluating the influence of geotechnical parameters on face stability, utilizing rigorous methodologies such as orthogonal experimental design, stochastic response surface methods, and global variance-based techniques [24,25,26,27,28]. These studies have successfully established a hierarchy of parameter importance, demonstrating that the shear strength indices, specifically cohesion and internal friction angle, play a dominant role in determining the limit support pressure and safety factors [29,30]. Concurrently, the characterization of failure mechanisms has also become well-established, with classical wedge models and logarithmic spiral surfaces widely adopted to describe the geometry of the collapse zone under standard loading conditions [31,32]. Building upon these comprehensive evaluations, it is valuable to further explore the comparative nuances between different hydraulic environments. Specifically, while static conditions are well documented, a quantitative analysis regarding the disparity in parameter sensitivity between dry and saturated mechanical environments could be further extended. Consequently, the non-linear amplification effect of fluid–solid coupling, whereby seepage forces magnify the stratum’s sensitivity to parameter degradation, invites more precise quantification. Moreover, it would be beneficial to delve deeper into the evolutionary mechanism of the support pressure-displacement response, thereby providing a more detailed insight into the spatial deformation modes and the emergence of bottom rheological characteristics induced by the softening process. Extensive studies have addressed tunnel face stability and parameter sensitivity under either dry or saturated conditions, a quantitative comparison of parameter sensitivity between dry and saturated mechanical environments could be further extended. In particular, the non-linear amplification effect induced by fluid–solid coupling on parameter degradation has yet to be fully quantified.

Above all, relying on the Jingu-Haihe Tunnel project of the Jinwei High-speed Railway, this study establishes a full-scale 3D fluid-solid coupling model on tunnel face stability in deeply buried silty sand strata based on Flac3D. The evolution laws of limit support pressure in dry soil (dry mechanical state), saturated soil (saturated seepage field scenarios), and under different degrees of softening are quantitatively assessed. The mechanism of the support pressure-displacement response triggers by the softening effect is analyzed to reveal the bottom rheological features caused by material softening. In addition, this study conducts a quantitative analysis on the difference in parameter sensitivity under dry and saturation mechanical environments, quantifying the non-linear amplification effect of the fluid–solid coupling effect on parameter sensitivity. The research results may offer theoretical references and technical support for the accurate determination of tunnel face support pressure and safety control in deeply buried water-rich weak strata.

2. Methods

2.1. Engineering Background

The Jingu-Haihe Tunnel of the New Tianjin-Weifang Railway is selected in this study, with a total length of approximately 3258 m, the tunnel serves as the key controlling project of the entire line (Figure 1). Its maximum burial depth reaches 49.7 m. The shield tunneling section (Chainage DK17 + 200~DK22 + 765) is constructed using a large-diameter slurry balance shield TBM with an excavation diameter of 13.8 m. The segment dimensions are 13.3 m (outer diameter), 12.2 m (inner diameter), and 2 m (ring width).

A representative deeply buried crossing section of the tunnel was selected for analysis. The burial depth of the tunnel axis in this section is approximately 30 m, located in a geological environment with high-pressure confined water. The strata crossed by the tunnel mainly consist of Quaternary alluvial-proluvial deposits. The overlying strata, from top to bottom, are silty clay, silt, silty sand and weak mudstone. The tunnel body and the tunnel face are located within the silty sand layer. The groundwater in this area is classified into phreatic water and shallow confined water. The phreatic water level is relatively shallow (0.9~1.9 m depth) and has a negligible impact on deep construction. The confined water hosted in the silt and silty sand layers constitutes the primary risk source, with a water level depth of approximately 13.6 m. Due to the moderate permeability of the silty sand layer and the high hydraulic head difference (approximately 30 m water column pressure), the excavation unloading process is highly prone to inducing water inrush and instability or collapse of the tunnel face.

2.2. Governing Equations for Fluid–Solid Coupling

This study is based on Biot’s consolidation theory, assuming the soil skeleton as a deformable porous medium and pore water as a compressible fluid, with fluid flow in the pores governed by Darcy’s law. Under quasi-static conditions, the mechanical equilibrium equation for saturated soil can be expressed in terms of the total stress tensor ${\sigma }_{ij}$ as ${\sigma }_{ij,j}+\rho g_i=0$, where $\rho$ is the density of saturated soil and $g_i$ is the gravitational acceleration vector. According to Terzaghi’s effective stress principle, the coupling relationship between total stress, effective stress ${{\sigma }^{\prime }}_{ij}$, and pore water pressure $p$ is defined as

$${\sigma }_{ij}={{\sigma }^{\prime }}_{ij}+\alpha p{\delta }_{ij}$$

(1)

where ${\delta }_{ij}$ denotes the Kronecker delta; $\alpha$ represents the Biot effective stress coefficient; for loose silty sand, $\alpha \approx 1.0$. Fluid transport within porous media is governed by the law of conservation of mass. Considering the compressibility of both the fluid and the solid skeleton, the fluid continuity equation is expressed as

$${\displaystyle \frac{𝜕\zeta }{𝜕t}}+𝛻\cdot q=0$$

(2)

where $\zeta$ is the variation in fluid content, and $q$ is the seepage velocity vector. According to Darcy’s law, the seepage velocity is linearly related to the pore water pressure gradient, i.e., $q_i=-k_{ij}{\displaystyle \frac{𝜕}{𝜕x_j}}(p-{\rho }_w{g}_k{x}_k)$, where $k_{ij}$ is the permeability tensor. Substituting Darcy’s law into the continuity equation and introducing the Biot modulus $M$ yields the diffusion equation governing the evolution of pore water pressure:

$${\displaystyle \frac{1}{M}}{\displaystyle \frac{𝜕p}{𝜕t}}=𝛻\cdot (k𝛻p)-\alpha {\displaystyle \frac{𝜕{\epsilon }_{vol}}{𝜕t}}$$

(3)

2.3. Soil Constitutive Model and Yield Criterion

The Mohr–Coulomb elastic-perfectly-plastic constitutive model is adopted. This model assumes that the material follows Hooke’s law prior to yielding and obeys a composite criterion of shear failure and tensile failure after yielding. The shear yield function $f^s$ is defined as:

$$f^s={\sigma }_1-{\sigma }_3{N}_{\varphi }+2c\sqrt{N_{\varphi }}$$

(4)

where ${\sigma }_1$ and ${\sigma }_3$ represent the maximum and minimum principal stresses, respectively; $c$ is the cohesion; and $N_{\varphi }$ is a material constant dependent on the internal friction angle $\phi$, defined as $N_{\varphi }=(1+\mathrm{s}\mathrm{i}\mathrm{n}\phi )/(1-\mathrm{s}\mathrm{i}\mathrm{n}\phi )$. When the stress state satisfies $f^s\ge 0$, the material undergoes shear plastic flow. Additionally, considering the potential tensile failure in shallow soils or disturbed zones, a tension cutoff is introduced. Its yield function $f^t$ is expressed as:

$$f^t={\sigma }_3-{\sigma }^t$$

(5)

where ${\sigma }^t$ denotes the tensile strength of the soil (typically taken as 0). During numerical calculations, a non-associated flow rule is employed to accurately reflect the dilatancy behavior of the silty sand. The plastic potential function $g^s$ follows a similar form to the yield function but uses the dilation angle $\psi$ instead of the internal friction angle $\phi$.

2.4. Numerical Model Establishment and Boundary Conditions

Based on the aforementioned theoretical framework, a full-scale three-dimensional numerical model is established using FLAC3D Version 9.6 (Itasca Consulting Group, Inc., Minneapolis, MN, USA) software to simulate the excavation process of the deep-buried slurry shield TBM. The model geometry is set as 120 × 120 × 120 m, considering an influence range of approximately 9D (D = 13 m, tunnel diameter) ahead of the tunnel axis and 9D laterally (Figure 2). This spatial extent is selected to eliminate the interference of boundary effects on the stress redistribution in the core area of the tunnel face. To accurately capture the localized shear zone ahead of the tunnel face under high hydraulic head conditions, mesh refinement is employed near the face, with the minimum element size controlled within 0.05 m. The total number of model elements is approximately 1.07 million.

To construct a representative geological model that captures the dominant hydro-mechanical behaviors, several idealizations were made based on the statistical analysis of the geotechnical investigation report. The stratum thickness and initial hydraulic head are determined using the arithmetic mean values from the multi-borehole data, and the soil layers were assumed to be isotropic and homogeneous. On this basis, the boundary conditions follow the actual engineering conditions. Mechanically, the model base is fully constrained, while the lateral boundaries are restrained in the normal direction. The top is loaded with an equivalent self-weight stress to simulate the overburden pressure at a 30 m burial depth. For hydraulic boundaries, constant-head boundaries are defined around the model perimeter and top. The pore water pressure distribution follows the hydrostatic gradient:

$$p(z)={\rho }_{w}g(H_{wt}-z)$$

(6)

where the groundwater level $H_{wt}$ is set equal to the ground surface elevation. During face instability simulations, the working face and tunnel lining sections are defined as impervious boundaries. In each load step, the fluid calculation sub-cycle and mechanical calculation sub-cycle alternate until the model reaches a new equilibrium state (with the maximum unbalanced force ratio maintained $<1.0\times {10}^{-1}$ for fluid calculations and <$1.0\times {10}^{-5}$ for mechanical calculations). The instability criterion for the tunnel face is defined as either the non-convergence of the model or the magnitude displacement exceeding 3% of the tunnel diameter [33,34]. The support pressure at this threshold is determined as the limit support pressure. Parameter values are selected from measured data (

Table 1. Relevant soil parameters.

Soil Type	$\mathit{\gamma }$/($\mathbf{k}\mathbf{N}\mathbf{·}{\mathbf{m}}^{\mathbf{-}\mathbf{3}}$)	$\mathit{\mu }$/(−)	${\mathit{K}}_{\mathbf{0}}$/(−)	E/(MPa)	c/(kPa)	$\mathit{\phi }$/(°)
Silty clay	20.3	0.39	0.65	2.53	22.9	10.8
Silt	20.3	0.35	0.55	4.73	15.4	27.1
Silty sand	19.8	0.31	0.45	8.63	5.1	27.2
Weak mudstone	20	0.3	0.42	50	40	35
Grout	21	\	\	3000	1000	36

). The soil porosity is obtained through in situ field testing, while the Poisson’s ratio, elastic modulus, cohesion, and internal friction angle are determined via laboratory saturated triaxial tests. Seepage-related parameters, including permeability and pore pressure distribution, are determined according to measured hydrogeological conditions and assumed to follow Darcy’s law under steady-state seepage.

Numerical results indicate that the displacement of the tunnel face exhibits a non-linear increasing trend as the support pressure decreases. The critical loading condition is identified by progressively reducing the support pressure derived from the actual in situ parameters until the instability criterion was met at approximately 231 kPa (Figure 3,

Table 2. Limit support pressure of dry mechanical state and saturated seepage field scenarios.

Various Scenarios	Dry Mechanical State			Saturated Seepage Field
	Intact Strength	Mean-Softening	Worst-Softening	Intact Strength	Mean-Softening	Worst-Softening
Limit support pressure/(kPa)	129	135	140	178	202	231

). While this value is lower than the operational service pressure used in engineering practice (approx. 276 kPa), this ultimate state is specifically selected to visualize the maximum potential plastic flow and the failure mechanism. At this critical state, the maximum displacement magnitude reaches 38.7 cm, demonstrating pronounced plastic flow characteristics. Specifically, the axial displacement is dominant, peaking at 38.6 cm, whereas the vertical and horizontal displacements perpendicular to the tunnel axis remain relatively small, with maximum values of 4 cm and 0.1 cm, respectively.

2.5. Configuration of the Softening Scenarios

The shear strength parameters adopted in this study are derived from test results under in situ saturated conditions. The hydro-mechanical coupled damage model established herein aims to dynamically simulate the strength degradation process induced by the increase in pore water pressure. To accurately reproduce the full process of damage evolution under hydro-mechanical coupling in the numerical simulation, it is necessary to back-calculate the initial shear strength parameters of the material in its intact state. By introducing a damage factor $\omega (p)$ a non-linear strength criterion based on the effective pore water pressure $p$ is established. Assuming the initial shear strength parameters of the soil in its natural low-moisture state are ${c^{\prime }}_0$ and ${{\phi }^{\prime }}_0$, the effective cohesion $c^{\prime }$ and effective internal friction angle ${\phi }^{\prime }$ undergo degradation as the pore water pressure increases:

$$c^{\prime }(p)={c^{\prime }}_0[1-{\omega }_c(p)]$$

(7)

$${\phi }^{\prime }(p)=\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{t}\mathrm{a}\mathrm{n}[\mathrm{t}\mathrm{a}\mathrm{n}{{\phi }^{\prime }}_0(1-{\omega }_{\phi }(p))]$$

(8)

where ${\omega }_c(p)$ and ${\omega }_{\phi }(p)$ represent the damage evolution functions varying with pore water pressure, with values ranging from 0 to 1.

Three parameter selection scenarios are designed: the intact strength scenario, the mean-softening scenario, and the worst-softening scenario. Given that the geotechnical parameters are determined via in situ tests under saturated conditions, the measured values are considered to represent the worst-softening scenario. The mean-softening scenario is derived from statistical data of conventional geotechnical immersion tests, adopting the mean softening coefficients for each soil layer. This scenario aims to simulate the progressive damage behavior of the strata under conventional hydrogeological conditions. The shear strength parameters for the mean-softening and intact strength scenarios are back-calculated based on Equations (7) and (8), and the range of the parameters refers to the studies of Erguler and Ulusay [35] and Vasarhelyi and Ván [36]. The results are presented in

Table 3. Back-calculated soil shear strength parameters.

Softening Scenario	Soil Type	c/(kPa)	$\mathit{\phi }$/(°)
Mean-softening scenario	Silty clay	24.3	11
	Silt	16.9	27.7
	Silty sand	6.9	29.8
	Weak mudstone	41.8	30.5
Intact strength scenario	Silty clay	25.7	11.3
	Silt	22.8	30.5
	Silty sand	11.6	34
	Weak mudstone	47.4	31.7

.

3. Results

3.1. Comparison of Limit Support Pressure Under Dry and Saturated Scenarios

The full-process support pressure evolution (600~100 kPa) under the dry mechanical state and saturated seepage field are compared (Figure 4). The fluid–solid coupling effect forces the limit support pressure required to maintain critical stability to ascend from 140 kPa to 231 kPa, with approximately 65% pressure increment, which quantitatively characterizes the additional support resistance required to counteract the seepage body force and compensate for the attenuation of effective stress. The sensitivity of the stratum’s response to support pressure variation is also altered by hydraulic conditions. The stratum under fluid–solid coupling method exhibits more severe stiffness degradation compared to dry mechanical state as support pressure decreases. Introducing the deformation amplification factor ($\eta =u_{wet}/u_{dry}$), $\eta$ is approximately 1.98 in the high support pressure stage (600 kPa) but rises to 3.63 in the low support pressure stage (100 kPa). In the low-pressure phase, seepage forces simultaneously impose external loads and accelerate soil degradation, causing resistance to decay significantly faster than in dry mechanical state [37,38].

Figure 5 compares the spatial disparities and symmetry characteristics of the face displacement field under dry and saturated states. In the dry mechanical state, as the support pressure decreases the displacement contours bifurcate at the upper and lower peripheries, forming sharp “horn-like” displacement zones with distinct vertical symmetry. This morphology emerges because the low support force fails to counteract the active earth pressure, allowing narrow and distinct shear bands to propagate from the crown and invert. In contrast, the deformation field maintains a quasi-symmetric, rounded, expansive “bulb-like” distribution throughout the entire pressure range under saturated seepage field scenarios. The failure in dry mechanical state is governed by the formation of brittle shear bands, whereas fluid–solid coupling transforms the failure into a global visco-plastic diffusion, with lower support pressures intensifying this morphological divergence.

3.2. Support Pressure-Displacement Response Induced by Softening Effects

The control exerted by different degrees of softening on the deformation modes of the tunnel face are analyzed (Figure 6). The softening effect induces a distinct stiffness degradation dependent on the stress state in the stratum, and the limit support pressures for the intact strength, mean-softening, and worst-softening scenarios are identified as 178 kPa, 202 kPa, and 231 kPa, respectively. In the high support pressure stage (>400 kPa), the three curves exhibit high convergence, demonstrating linear-elastic deformation characteristics. The displacement growth rate in this phase is approximately 6 × 10⁻² cm/kPa from 600 kPa to 400 kPa support pressure, indicating that under high confining pressure, the contribution of soil parameter degradation to macroscopic deformation is significantly suppressed. However, as the support pressure drops below 300 kPa, the curves display a distinct non-linear bifurcation. This bifurcation point corresponds to the critical yield stress threshold where the deviatoric stress induced by excavation unloading intersects the degraded Mohr–Coulomb yield surface. The softening effect shrinks the yield envelope, causing the stress path in the worst-softening scenario to breach the plastic limit earlier than in the intact scenario. Consequently, in the low support pressure stage (300~100 kPa), the softening effect fundamentally alters the displacement growth rate. The average growth rate for the intact strength scenario is 1.19 × 10⁻¹ cm/kPa, whereas it surges to 1.83 × 10⁻¹ cm/kPa in the worst-softening scenario, representing an increase of approximately 53.8% in the deformation rate. This surge indicates that once the support pressure falls below the critical pore pressure-dependent threshold, the failure mechanism transitions from stable deformation to accelerated plastic flow. This accelerated stiffness degradation causes the displacement difference to expand from 8.2 mm at 400 kPa to 6.4 cm at 250 kPa, and finally reaching 18.6 cm at 100 kPa. The softening effect significantly lowers the stratum’s yield threshold, rendering face deformation in the low-pressure stage extremely sensitive to support pressure variations.

The softening effect leads to a structural reconfiguration of the deformation field. The displacements from five characteristic points (a, b, c, d, e) across the tunnel face are selected (Figure 6). The softening effect significantly alters the relative deformation pattern of the tunnel face. In the worst-softening scenario, the center point displacement increases from 5.6 cm at 600 kPa to 66 cm at 100 kPa, an 11.8-fold increase. In contrast, the displacement at the crown surges from 1.2 cm to 39.3 cm, representing a 32.8-fold increase. The deformation growth rate in the peripheral region far exceeds that of the central region, leading to a significant elevation in the edge-to-center displacement ratio. Specifically, the normalized displacement ratio at the crown rises from 0.22 in the high support pressure stage to 0.63 in the low support pressure stage. Moreover, in the intact strength scenario, the gravitational stress field dictates a conventional displacement pattern where the crown displacement exceeds that of the invert. However, the magnitude displacement of invert reaches 39.3 cm, exceeding the displacement of crown under the worst-softening scenario. The bottom rheology dominant phenomenon is that the tunnel invert simultaneously bears the maximum pore water pressure and gravitational stress. Driven by high hydraulic gradients, the softened silty sand exhibits more pronounced rheological and plastic flow characteristics.

3.3. Shear Parameter Sensitivity Under Dry and Saturated Mechanical States

In the dry mechanical state, the stratum exhibits pronounced structural robustness, where face deformation displays an insensitive characteristic to the degradation of material parameters (Figure 7). When the support pressure is in the high support pressure stage (>400 kPa), the displacement curves for both the intact strength and softening scenarios almost perfectly coincide. Soil stiffness is dominated by skeletal contact stress, and the degradation of cohesion has negligible impact on macroscopic deformation under high effective support pressure. Even under the 100 kPa support pressure, the relative rate of change in face displacement remains limited to 0.3 cm. In the absence of pore water pressure, if provided the support pressure remains above the limit equilibrium threshold, the soil arching effect can automatically adjust the load transfer path through inter-particle frictional interlocking to accommodate minor parameter degradation. The limit support pressures also reflect the insensitivity to parameter variations, calculated as 129 kPa, 135 kPa, and 140 kPa for the intact strength, mean-softening, and worst-softening scenarios, respectively. The stratum’s response to parameter variations exhibits a hyper-sensitive explosive characteristic and distinct symmetry breaking when fluid-solid coupling effect is set. In the saturated environment, the same magnitude of parameter degradation triggers a severe chain reaction; particularly in the low support pressure stage (<300 kPa), the change in face displacement surges to 4.2~18.7 cm. The fluid–solid coupling effect amplifies the stratum’s sensitivity to parameter variations by up to 26 times at point_c than dry mechanical state.

4. Discussion

4.1. Comparison of Fluid–Solid Coupling Effects

Validation with classical theories regarding the hydro-mechanical interaction: Our finding that the limit support pressure increases by approximately 65% (from 140 kPa to 231 kPa) under seepage conditions is consistent with the classical theoretical solutions and recent numerical investigations [39,40,41]. These studies established the fundamental consensus that hydraulic gradients act as a significant destabilizing body force on the tunnel face, necessitating higher support pressures to maintain stability.

Integration of softening mechanism building upon these established coupling frameworks: The current research further integrates the material softening mechanism to characterize the hydraulic deterioration of soil properties. Unlike standard coupling analyses that primarily focus on the seepage body force, our model shows the stress-dependent stiffness degradation induced by pore water pressure [42,43]. This approach allows for a more detailed quantification of the non-linear amplification effect (up to 26 times increase in parameter sensitivity), offering supplementary insights into the complex rheological failure modes observed in deep-buried water-rich strata.

4.2. Transition of Failure Modes Under Excessive Support Pressure

In the dry mechanical state, elevating the support pressure from 550 kPa to 600 kPa results in a deviation from the conventional convergence law, manifested as an increase in face displacement (Figure 4a). This phenomenon signals a fundamental transition in the interaction mechanics between the tunnel face and the support system-shifting from passive confinement to active compression. The stratum deformation diminishes as pressure increases within the range between active and at-rest earth pressures. When the pressure exceeds the at-rest threshold and approaches the passive earth pressure limit, the support pressure essentially transforms into an active load, performing work on the soil mass. The excessive thrust forces the frontal core soil into a passive compression state. Consequently, the soil skeleton undergoes extrusion under axial compressive stress, which is quantitatively reflected as an increase in absolute displacement values and a reversal of the displacement vector (Figure 8).

The high-pressure extrusion phenomenon revealed herein correlates well with predictions from the upper-bound solutions of classical limit analysis theory [15,44]. The stability of frictional faces has established that excessive support pressure shifts the failure mechanism from collapse to blow-out. In the deeply buried dry mechanical state, the immense confining pressure from the overburden prevents surface heave, where the excessive work done by the support system is dissipated through the axial compaction of the core soil ahead of the face. This aligns with the intrinsic consistency of findings by Mollon [14,45], which posit that failure modes under high pressure are governed by the synergy of stratum dilatancy and confinement levels.

5. Conclusions

Based on the full-process fluid-solid coupling numerical analysis concerning the face stability of large-diameter slurry shield tunnels in deeply buried silty sand strata, the main conclusions are derived:

• Seepage flow significantly elevates the limit support pressure required to maintain face stability. The limit support pressure increases from 140 kPa in the dry mechanical state to 231 kPa under fluid–solid coupling conditions. In the dry state, the range of displacement contours presents “horn-like”, reflecting the formation of localized shear bands, and a global “bulb-like” deformation pattern is exhibited under saturated seepage field scenarios. The seepage forces transform the failure mode from friction-controlled brittle shearing to visco-plastic rheological diffusion.
• The impact of soil softening exhibits a strong dependency on the support pressure. During the high support pressure stage, the effects of softening are inhibited by high effective confining pressure. However, the deformation rate in the worst-softening scenario increases by 53.8% compared to the intact strength scenario within the low support pressure stage. This indicates that saturated sand exhibits extremely high sensitivity to the softening effect at low stress levels.
• The water-rich environment alters the spatial distribution of deformation. During the unloading process in saturated state, the displacement growth multiplier at the crown reaches 32.8, far exceeding the 11.8 observed at the center. Driven by the superposition of maximum hydraulic head and gravitational stress, the ultimate displacement at the tunnel invert exceeds that of the crown.
• Fluid–solid coupling induces a non-linear amplification of parameter sensitivity. The fluid–solid coupling environment amplifies the stratum’s sensitivity to shear parameter variations by up to 26 times compared to dry mechanical state. This implies that in water-rich silty sand strata, minor errors in geological parameters can be hydraulically amplified into catastrophic mud inrushes.

In the future, the research intends to incorporate elastic-viscoplastic or strain-softening models, to more accurately characterize the rheological responses and long-term creep under high hydraulic gradients. Furthermore, the current focus on a single layer will be expanded to simulate multi-layer interactions due to the heterogeneous nature of the strata. Introducing interface elements and anisotropic permeability parameters will allow for a more rigorous evaluation of cross-layer seepage effects and differential deformation across geological boundaries, ultimately enhancing the optimization of support strategies for complex profiles.