Introduction

Sandy loess is a loose sediment formed by aeolian deposition since the Quaternary under arid climatic conditions1,2,3,4, covering an area of 640,000 km2, accounting for about 6.6% of China country’s total area. Liu Dongsheng, Academician of the CAS Member, through systematic investigation of the Loess Plateau, found that the grain size of sandy loess becomes finer from the northwest to the southeast of China5, with less sand content and greater cohesion. Based on this, he divided the sandy loess region, general loess, and clayey loess areas. Due to the different stratigraphic depositional time sequences, sandy loess often underlies the tertiary silt-fine sand stratum, and the two exhibit significantly different mechanical properties. Fine sand has low shear strength and is prone to disintegration under construction disturbance, resulting in poor stability of surrounding rocks and potential hazards such as sand leakage and tunnel face collapse during tunnel construction6,7,8. Therefore, a rational construction method is key to ensuring tunnel safety.

Li et al.9 proposed a comparative study between the optimal tunnel method and the CD method for large section tunnel engineering on loose loess foundation, and showed that the three step large wall foot temporary arch method has a good settlement control effect. Based on the characteristics of sandy loess tunnel construction, Shao et al.10 conducted simulation research on the strength variation law of loess tunnel construction from the aspects of loess structure evolution, excavation face instability, and shear deformation zone distribution changes, revealing the formation mechanism of different types of damage during loess tunnel construction. Li et al.11 conducted a series of in-situ experiments and numerical simulations on the displacement characteristics of the excavation section of a high-speed railway (HSR) loess tunnel using the three step excavation method. The displacement characteristics of a large-span loess tunnel were explained, and a metastable structure of loess was proposed. It was proved that the three step excavation method can effectively control the deformation of the loess tunnel face and is an excellent construction plan. Luo et al.12 conducted numerical simulation research on the disturbance law of large-span tunnel excavation in the loess surrounding rock of northern Shaanxi. The spatiotemporal relationship curve of arch deformation shows three stages of “rapid deformation stage continuous deformation stage slow deformation stage” in the construction of reserved core soil ring excavation, verifying the different stress laws of support structures at different positions. Wang et al.13 proposed that the efficiency of commonly used support methods in the current engineering on site is very limited, leading to serious support system problems in loess tunnel construction, mainly manifested as arch deformation, cracking of sprayed concrete, and surface settlement. This proves that the new construction support scheme has excellent practical value for optimizing the structural safety and bearing capacity of traditional support systems. Xiong et al.14 considers the asymmetry of tunnel loads and the weakness of surrounding rock, and takes the influence and deformation characteristics of temporary supports during the excavation of tunnels using the two-step center diaphragm method (CDM) as the research object. He found that there is a significant “step like” phenomenon in the change of principal stress of temporary steel supports.

Currently, the study of tunnel construction method of under loess and fine sand composite strata is still insufficient. In this study, based on the Panxianshan Tunnel of the Zhongwei to Lanzhou Railway, obtains the parameters of sandy loess and fine sand strata through on site investigation and indoor experiments, proposing a construction method of TTA. The engineering effect is validated by combining numerical simulation with on-site monitoring. The research results hold significant reference value for selecting construction methods for large-section tunnels in similar stratum conditions.

Project overview

The Panxianshan Tunnel is located in Zhongchuan Town, Lanzhou New Area, Gansu Province. The tunnel spans from DK236 + 376 to DK239 + 205, with a total length of 2,829 m, making it the longest sandy loess tunnel on the Zhongwei to Lanzhou Railway. The tunnel passes through strata primarily consisting of Quaternary (Q4) and Upper Pleistocene (Q3) sandy loess, with underlying bedrock of Upper Tertiary (N1) sandstone interbedded with mudstone15,16,17. The tunnel is designed for a speed of 250 km/h, with an excavation width of 14.52 m, an excavation height of 12.38 m, and an excavation area of 153.12 m2. According to the “Technical Specifications for Loess Tunnels on Railways” (Q/CR 9511–2014), it is classified as an extra-large span, super large cross-section loess tunnel. During excavation at the tunnel entrance (up to DK236 + 495), a silt layer approximately 2–3 m thick appeared on the left side of the upper step and at the crown. By DK236 + 570, Fine sand was exposed on the upper and middle steps, about 8–10 m thick, with sandy loess present on the lower step. Issues such as crown collapse and flowing sand were frequently encountered during construction. The tunnel location and site conditions are shown in Fig. 1 (The satellite images are produced by Ovi Interactive Maps Windows client v10.2.1). It is worth mentioning that groundwater is not developed in the tunnel area, and the sandy loess layer is very dry. Therefore, this article does not consider the influence of groundwater on the mechanical behavior of tunnel excavation, as shown in Fig. 2

Fig. 1
Fig. 1
Full size image

Panxianshan Tunnel’s location and the site construction.

Fig. 2
Fig. 2
Full size image

tunnel excavation.

Indoor tests

To obtain the mechanical characteristics of the surrounding rock of the Panxianshan Tunnel, sampling was conducted at section DK236 + 510 (buried depth approximately is 50 m), with fine sand taken from the arch top and sandy loess from the lower step. To preserve the properties of the original soil, the samples were wrapped with plastic film, sealed in plastic bags, and taped to prevent any pressure during transportation.Four samples were made for each soil layer, particle analysis, density determination, moisture content assessment, and direct shear tests were conducted indoors18,19, as shown in Fig. 3. Figure 4 shows the particle size distribution curves of two types of soil layers. It can be observed that the sandy loess contains a high proportion of fine particles (0.05–0.005 mm) and sand particles (> 0.05 mm), with the particle size primarily distributed in the range of 0.01–0.1 mm. Particles smaller than 0.1 mm account for 88.4% of the total mass. The fine sand’s particle size is mainly distributed in the 0.1–1 mm range, with particles in the 0.1–0.25 mm range constituting 51.4% of the total mass. Table 1 presents the indoor test results for the two soil layers, the data in the table is the average value of four samples. It can be seen that the fine sand has a moisture content of 2.41%, a high content of coarse particles, and poor particle cohesion. As a result, its cohesion and friction angle are much lower than those of the sandy loess. Therefore, sand leakage and collapse are common hazards after tunnel excavation.

Fig. 3
Fig. 3
Full size image

Indoor test process.

Fig. 4
Fig. 4
Full size image

Grain diameter graduation cumulative curves of two types of soil.

Table 1 Laboratory soil test results.
Full size table

Numerical simulation

Finite element model

Finite element analysis was conducted using Midas GTS NX. Based on design specifications and on-site construction conditions, a three-dimensional finite element model was developed for the section from DK236 + 500 to DK236 + 60020,21. The tunnel excavation influence zone extends approximately 3–5 times the tunnel diameter. Hence, the model dimensions were set to 100 m × 100 m × 100 m. The y-axis represents the tunnel excavation direction, while the x and z axes denote the horizontal and vertical coordinates, respectively. Within the model, the fine sand layer has a thickness of 14 m, the sandy loess layer spans 56 m, and the mudstone layer measures 30 m. A monitoring section is positioned at y = 50 m at the model’s center. The initial support system comprises C25 shotcrete (30 cm), I22a steel arch (spacing 0.5 m per segment), and φ42 anchor pipes (L = 3.5 m). The surrounding rock is represented using 3D solid elements, with the Mohr–Coulomb failure criterion applied. Both the initial support and secondary lining are modeled as elastic elements. Plate elements represent Shotcrete, while the steel arch and anchor pipes are modeled using embedded beam elements22,23,24. The analysis considers two types of advanced support structures: pipe umbrellas and horizontal jet grouting piles. The parameters for these structures are as follows:

(1) The advanced support system employs φ108 large-diameter pipe umbrellas, each 15 m long, with a 3 m overlap, an insertion angle of 10°, and a circumferential spacing of 0.4 m. The pipe umbrella is simulated using embedded beam elements in the finite element model. Upon filling with mortar, the equivalent elastic modulus (EP) is calculated using the following formula:

$$E_{p} = \frac{{E_{s} I_{s} + E_{c} I_{c} K_{w} }}{{I_{p} }}$$
(1)

where, Es and Is represent the elastic modulus and moment of inertia of the steel pipe, respectively, Ec is the elastic modulus of the mortar inside the pipe, Ic is the moment of inertia of the mortar within the pipe, Ep denotes the equivalent elastic modulus of the grouting pipe umbrella, Ip represents the equivalent moment of inertia of the grouting pipe umbrella, and Kw is the stiffness reduction factor for the mortar resulting from the cracking of the pipe umbrella, which is assumed to be 0.6.

(2) The advanced support system incorporates horizontal jet grouting piles, each with a diameter of 600 mm, circumferential spacing of 500 mm, and a 100 mm overlap between adjacent piles. The horizontal jet grouting piles are modeled using solid elements, and the elastic modulus of the piles is converted to that of the reinforced soil layer using the stiffness equivalence method.

$$E_{c} = \frac{{E_{a} S_{a} + E_{g} S_{g} }}{{S_{c} }}$$
(2)

where, Ec represents the elastic modulus of the equivalent reinforced soil; Ea is the elastic modulus of the fine sand; Sg is the cross-sectional area of the jet grouting pile; Eg is the elastic modulus of the jet grouting pile, assumed to be 4600 MPa; Sc is the cross-sectional area of the reinforced zone; Sa is the area of fine sand within the reinforced zone, where Sa = Sc—Sg.

Currently, for large-section sandy loess tunnel excavation in high-speed railways, commonly used methods include the CRD, double side heading method, and three steps and seven parts excavation method. The CRD and double side heading method offer effective deformation control but are complex and slow in progress. Therefore, the three-step temporary arching method is employed on-site, with temporary steel vertical brace installed during the excavation of the upper step to mitigate crown settlement. The temporary arching system consists of I22a steel I-beams and C25 shotcrete (25 cm) and is constructed concurrently with the initial support. After tunnel completion, secondary lining is applied. The surrounding rock is modeled using the Mohr–Coulomb criterion, while the initial support and secondary lining use elastic elements. The calculation parameters are listed in Table 2. In the analysis, each excavation step is 2 m, with the initial support lagging one step behind the excavation. The model consists of 77,138 nodes and 77,913 elements, as illustrated in the finite element model in Fig. 5, The model was modeled using Midas GTS NX v2021 R1 software.

Table 2 Calculation parameters.
Full size table
Fig. 5
Fig. 5
Full size image

Finite element model.

Assuming the length of the advanced suport is L1 and the longitudinal overlap length is L2, the length of the excavation step, L3, must satisfy the condition L3 ≤ L1—L2. Under the assumption of unchanged initial support, three working conditions are defined based on the combination of excavation step length and advanced support, as shown in Fig. 6, The model was modeled using Midas GTS NX v2021 R1 software:

  1. 1)

    Case 1 The upper, middle, and lower steps length are all 6 m, and the advanced support consists of φ108 large-diameter pipe umbrellas, each 15 m long with a 3 m overlap. The construction steps for this case in Midas GTS can refer to Table 3.

  2. 2)

    Case 2 The upper step length is 12 m, while the middle and lower step lengths are 6 m. The advanced support employs φ600 horizontal jet grouting piles, each 15 m long with a 3 m overlap. The construction steps for this case in Midas GTS can refer to the Table 4.

  3. 3)

    Case 3 The upper, middle, and lower steps length are all 6 m, and the advanced support consists of φ600 horizontal jet grouting piles, each 9 m long and with a 3 m overlap. The construction steps for this case in Midas GTS can refer to Table 5.

Fig. 6
Fig. 6
Full size image

Calculation condition.

Table 3 Case 1 pipe unbrella.
Full size table
Table 4 Case 2 horizontal jet grout pile 15 m.
Full size table
Table 5 Case 3 horizontal jet grout pile 9 m.
Full size table

Calculation results analysis

Tunnel deformation

Figure 7 illustrates the crown settlement curves as a function of excavation steps under different calculation conditions. The crown exhibits an S-shaped profile, indicative of a rapid increase in settlement following excavation disturbance. Once the excavation of the lower step is completed and the initial support forms a closed ring, the settlement rate significantly diminishes. In Case 1 (pipe umbrella advanced support), the maximum arch crown displacement reaches 71.21 mm. In Case 2 (jet grout pile advanced support with a 12 m upper step length), the arch crown displacement settlement is 62.9 mm. In Case 3 (jet grout pile advanced support with a 6 m upper step length), the arch crown displacement settlement is the smallest at 59.57 mm, approximately 5% less than in Case 2.

Fig. 7
Fig. 7
Full size image

Variation curve of crown settlement with excavation step.

Figure 8 presents the horizontal convergence curves as a function of excavation steps under different calculation conditions. These curves also follow an S-shaped pattern, with the most significant variation occurring in the excavation-disturbed zone. Case 1 exhibits the highest convergence value, reaching 43.18 mm. The final convergence value in Case 2 is 27.95 mm, while in Case 3, it is slightly lower at 26.77 mm, showing only a minor difference compared to Case 2.

Fig. 8
Fig. 8
Full size image

Variation curve of horizontal convergence with excavation step.

Primary lining stress

Figure 9 presents the steel arch frame’s final first principal stress distribution at the monitoring section under the three conditions. The steel arch frame is predominantly in a state of compression, with higher compressive stresses extending from the crown to the haunch. The maximum first principal stress is observed at the shoulder of the arch. In terms of numerical values, the stresses follow the order: Case 1 (− 62.13 MPa) < Case 2 (− 50.4 MPa) < Case3 (− 46.09 MPa). This pattern is attributed to the high stiffness of the horizontal jet grouted pile support, which allows it to withstand more significant surrounding rock pressure, thereby reducing the pressure on the primary lining steel frame and decreasing the compressive stress.

Fig. 9
Fig. 9
Full size image

Cloud of the first principal stress the monitoring steel arch frame.

Plastic zone

Figure 10 shows the distribution of the plastic zone at the monitoring section after tunnel completion under the three conditions. In Case 1, the plastic zone is primarily distributed around the tunnel periphery, with the largest plastic strain occurring at the haunch. In Case 2 and Case 3, the plastic zone distributions exhibit a highly similar distinctive crescent-shaped pattern, with the major axis extending approximately 4.2 m along the tunnel’s axial direction and the minor axis measuring 1.8 m radially. Notably, compared to Case 1—where a plastic zone with a maximum radius of 2.5 m was observed at the crown—both Case 2 and Case 3 show extremely small plastic deformation at the crown, with radii below the measurement resolution (≤ 0.2 m) and thus virtually indistinguishable in field observations. This significant improvement is attributed to the increased support stiffness provided by the horizontal jet grouting piles. This stiffness enhancement reduced tensile stress concentrations at the crown by 45%, effectively suppressing plastic zone development to a magnitude nearly below detectability.

Fig. 10
Fig. 10
Full size image

Cloud of plastic zone under different calculation conditions.

Additionally, a comparative analysis between Case 3 and Case 2 reveals quantitative differences in plastic zone extent at the tunnel springing. Specifically, the plastic zone radius at the springing in Case 2 measures 1.3 m, which is approximately 15% larger than the 1.1 m radius observed in Case 3. This discrepancy highlights the nuanced impact of varying grouting parameters on local stress redistribution and plastic deformation control.

A comprehensive analysis of surrounding rock deformation, primary lining stress, and plastic zone development indicates that when employing the three-step temporary arch method for excavation, the pipe umbrella advanced support (Case 1) offers the least effective control over surrounding rock deformation. In contrast, applying horizontal jet grouted pile advanced support significantly reduces surrounding rock deformation and plastic zone development, thereby providing superior support performance compared to the pipe umbrella method. The support effects of Case 2 and 3 are mainly similar. However, increasing the jet grouted pile support length allows for a longer upper-step excavation, thus providing additional workspace for construction. This, however, increases both the difficulty and cost of construction. Therefore, considering the balance between support effectiveness and construction cost, the optimal construction parameters are determined to be a jet grouted pile length of 9 m and a step excavation length of 6 m.

On-site monitoring

Monitoring plan

To evaluate the efficacy of the TTA construction method, on site monitoring was conducted on the DK236 + 550 section within the study area. The on site construction setup is depicted in Fig. 11. The monitoring parameters included crown settlement, horizontal convergence, surrounding pressure, and steel arch strain. A total of 12 monitoring points were distributed across the tunnel crown, shoulder, haunch, springing, and temporary arch. Each monitoring point was equipped with a soil pressure gauge (Vibrating wire and double membrane type with a range of 0–2MPa: YT-200B) and a surface strain gauge(Vibrating wire type with a range of − 1800 to 1200με: YT-500B) mounted on both sides of the steel arch, with wiring led out to prevent damage during construction burial25,26,27,28. In total, 24 instrumen·ts were installed for this section. The monitoring schedule was as follows: two measurements per day for the first 15 days post-initial installation, one measurement per day after the 15th day, and two measurements per week after 30 days. The arrangement of the monitoring points is shown in Fig. 12.

Fig. 11
Fig. 11
Full size image

Site construction photos.

Fig. 12
Fig. 12
Full size image

Monitoring point layout.

Tunnel deformation

Figure 13 compares the on site monitoring data with the numerical simulation results for tunnel deformation. The on site data were corrected, and the “measured” and “simulated” data were aligned to the same time points in the construction process. The measured data also account for deformations before excavation, as represented by the numerical simulation values. As shown in Fig. 13a, it is evident that as the distance from the tunnel face increases, the crown settlement gradually rises and ultimately stabilizes. The trends observed in the on site and simulation data align closely. Between 2 and 20 m from the face, the crown settlement increases by 18.2 mm, accounting for approximately 85% of the total observed settlement. This range corresponds to the tunnel construction disturbance zone. Once the distance from the face exceeds 20 m, the closure of the initial support occurs, and the rate of crown settlement growth significantly slows. Subsequent construction disturbances have a reduced impact. The final measured crown settlement is 52.06 mm, while the simulated value is 58.07 mm, with the measured value slightly smaller than the simulated one. Figure 13b shows the horizontal convergence curve. The variation pattern follows that of the crown settlement. In the construction disturbance zone, horizontal convergence increases rapidly, with a marked deceleration following the closure of the initial support. The final measured horizontal convergence is 29.15 mm, while the simulated value is 27.95 mm, with the measured value slightly exceeding the simulated one. A comparison of the on site monitoring data with the numerical simulation results supports the reliability of the numerical simulations. The TTA construction method has proven effective in mitigating surrounding rock deformation.

Fig. 13
Fig. 13
Full size image

Deformation curve of monitoring and simulation data.

Surrounding rock pressure

According to the Technical Regulations for Monitoring and Measurement of Railway Tunnels (Q/CR9218-2015), using horizontal convergence and arch settlement as core monitoring indicators has significant scientific necessity and engineering practical value. Horizontal convergence directly characterizes the compressive effect of lateral stress release of surrounding rock on tunnel structure, while arch crown settlement reflects the vertical loosening of rock mass and the bearing response of the support system. The synergy between the two reveals the three-dimensional deformation characteristics of tunnel section and the dynamic mechanism of surrounding rock support interaction. By monitoring its spatiotemporal evolution in real-time, the stability of surrounding rock can be quantitatively evaluated.

Due to the asynchronous construction times of the inverted arch and tunnel arch walls and differing testing periods, these two aspects are discussed separately. Figure 14 shows the variation in surrounding rock pressure over time at the arch wall measurement points. The pressure values fluctuate significantly during the first 6 days and stabilize after 52 days. Notably, the maximum pressure values occur at different times for each point. The surrounding rock pressure at the left and right arch haunches (S4 and S5) is higher than at other measurement points. At the S4 point (left arch haunch), the pressure reached a maximum value of 149.47 kPa on day 4, fluctuated after that, and stabilized at 134 kPa. At the S5 point (right arch haunch), a second pressure increase occurred on day 8, reaching a maximum of 121.73 kPa on day 18 and stabilizing at 117.8 kPa.

Fig. 14
Fig. 14
Full size image

Time dependent curve of surrounding rock pressure of the arch wall measuring points.

Figure 15 presents the surrounding rock pressure variation over time at the inverted arch measurement points. The trend is similar across all points: the pressure increases over time, peaks around day 40, and stabilizes after day 60. The maximum pressure recorded at point Y2 is 150.82 kPa. Generally, the pressure variation at the inverted arch points is more consistent than at the arch wall points. This is because the pressure gauges at the inverted arch had better contact conditions. Moreover, the inverted arch was constructed after the completion of the initial support, minimizing the effects of tunnel excavation stages. After the initial support ring closure (22 days), the surrounding rock pressure continued to rise for longer. Disturbances from secondary lining construction and construction vehicle activity led to higher pressures at the inverted arch points compared to the arch wall points.

Fig. 15
Fig. 15
Full size image

Time dependent curve of surrounding rock pressure of the inverted arch measuring points.

Figure 16 shows the distribution of surrounding rock pressures across various monitoring sections. The pressure at the inverted arch measurement points is generally higher than at the arch wall points. The pressure distribution forms a crescent shape, with lower pressure at the crown and higher pressures at the left and right haunches, exhibiting asymmetric stress distribution. According to the finite element model (Fig. 5), the thickness of the fine sand layer is 14 m, with approximately 11 m above the crown. The maximum loosening pressure at the initial support is calculated to be \(\gamma h\) = 15.3 × 11 = 168.3 kPa. However, as seen in Fig. 16, the measured surrounding rock pressure is lower than the maximum loosening pressure. This indicates that the horizontal jet grouting piles have effectively reinforced the fine sand layer, causing it to remain elastoplastic. As a result, the horizontal jet grouting piles bear most of the surrounding rock pressure, thereby reducing the pressure transferred to the initial support.

Fig. 16
Fig. 16
Full size image

Distribution map of surrounding rock pressure.

Steel arch stress

Figure 17 shows the variation in steel arch stress over time at the arch wall measurement points. It is evident that the stress variation patterns on both the inner and outer sides of each measurement point are consistent. Overall, the stress increases and stabilizes after approximately 50 days, with the steel arch stress being negative, indicating a compressive state. The arch crown (S1 point) and the left and right arch shoulders (S2 and S3 points) experience the longest period of disturbance, with the largest stress variation. The stress change exhibits a stepwise pattern, with the greatest disturbance occurring at the points above the shoulders. The maximum steel arch stress at the crown (S1-in) is 132.92 MPa. The stress is relatively small for the left and right arch feet (S6 and S7), and the fluctuations are less pronounced. Figure 18 illustrates the variation in steel arch stress over time at the inverted arch measurement points. The stress variation on the inner and outer sides is identical and negative, indicating that the inverted arch remains under compressive stress. The stress changes significantly during the first 20 days due to substantial disturbances from pouring the secondary lining and backfilling concrete. After day 20, the stress gradually decreases and stabilizes after 60 days.

Fig. 17
Fig. 17
Full size image

Time dependent curve of steel arch stress of the arch wall measuring points.

Fig. 18
Fig. 18
Full size image

Time dependent curve of steel arch stress of the inverted arch measuring points.

Based on the stress monitoring data from the inner and outer sides of the steel arch, the conversion formulas for stress and internal force can be derived from the combination of material mechanics and deformation Theory as follows:

$$\left\{ {\begin{array}{*{20}c} {N = (\sigma_{in} + \sigma_{out} ) \times A/2} \\ {M = (\sigma_{in} - \sigma_{out} ) \times W/2} \\ \end{array} } \right.$$
(3)

where N and M are the axial force (kN) and bending moment (kN·m) of the steel arch, respectively. \(\sigma_{in}\) ,\(\sigma_{out}\) are the inner and outer stress values (MPa) at the measurement points of the steel arch. A is the cross-sectional area of the steel arch (m2), W is the section modulus (m3) .

Figure 19 presents the final internal force envelope of the steel arch. It can be observed that the moment is asymmetrically distributed around the tunnel, with the moment on the right side being more significant than on the left. The values are relatively small, with the maximum moment of − 15.45 kN·m at the right arch shoulder (S3 point). In contrast, the axial force follows a pattern where the arch ring experiences a more significant axial force than the inverted arch. The maximum axial force of − 493.10 kN is observed at the left arch shoulder.

Fig. 19
Fig. 19
Full size image

Final internal force envelope diagram of steel arch frame.

According to the Railway Tunnel Design Code (TB10003—2016), the safety factor for the steel arch can be assessed as follows:

$$\left\{ {\begin{array}{*{20}c} {K = \frac{{\varphi \alpha R_{a} bh}}{N}} & {(e \le 0.2h)} \\ {K = \frac{{1.75\varphi R_{1} bh}}{{6\frac{{e_{0} }}{h} - 1}}} & {(e > 0.2h)} \\ \end{array} } \right.$$
(4)

where: K—Safety Factor, N—Axial Force; Ra—Compressive Strength of Steel Arch, is 235 MPa; R1—Tensile Strength of Steel Arch, is 235 MPa; b—Width of Section, is 0.11 m; h—Height of Section, is 0.22 m; \(\varphi\)—(Longitudinal Bending Coefficient of Component, is 1; \(\alpha\)—Eccentricity Factor of Axial Force; e—Eccentric Distance of Section. (para26).

Figure 20 shows the distribution of safety factors at various measuring points of the steel arch frame. It can be seen that the safety factor of the left arch shoulder is the smallest, at 11.53, and the safety factor of the left wall foot is the largest, at 71.04. The safety factors of each measuring point are greater than the required values in the specifications (2.0 for small eccentricity compression failure and 2.4 for large eccentricity tension failure), and the surplus is relatively large, indicating that the supporting structure is in a safe state.

Fig. 20
Fig. 20
Full size image

Safety factor distribution map of steel arch frame.

Conclusion

This study focuses on the Panxian Mountain Tunnel of the Zhonglan Railway, employing experimental tests, numerical simulations, and field monitoring to comprehensively evaluate optimal construction methods for large-section tunnels in sandy loess-fine sand composite strata. Key findings are as follows:

  1. 1)

    The fine sand exhibits predominant particle sizes of 0.1–1 mm (moisture content: 2.41%) with high coarse particle content and cohesion measuring merely 29% of that in sandy loess. Tunnel excavation-induced particle separation triggers sand gushing and face collapse. These indoor-tested physical parameters provided reliable input for numerical simulations, though the study employed a continuous medium model that may underrepresent post-disturbance loose media behavior in fine sand, as noted in the discussion.

  2. 2)

    Numerical simulations of the three-bench temporary inverted arch method reveal that maximum surrounding rock deformation, initial support stress, and plastic zone occur under pipe-roof advanced support (Case 1). Advanced horizontal jet grouting piles significantly reduce these parameters. While 15m and 9m piles demonstrate comparable reinforcement effectiveness, longer piles increase upper bench excavation length, elevating construction difficulty and cost. A 9m pile length is therefore recommended as the optimal parameter.

  3. 3)

    Field monitoring data indicate three-stage temporal evolution of surrounding rock pressure and steel frame stress: rapid growth, fluctuation, and stabilization. Significant deformation occurs within 20 m behind the tunnel face, with pressure displaying crescent-shaped distribution (lowest at vault, highest at arch haunches). Inverted arch measurement points exceed sidewall values, while measured pressures remain below theoretical relaxation pressures. Steel frames exhibit minor eccentric compression characteristics, with safety factors consistently below code-specified limits.

  4. 4)

    Integrated simulation and monitoring validate horizontal jet grouting’s effectiveness in fine sand reinforcement. The three-bench method accelerates initial support closure through temporary inverted arches while employing temporary vertical supports to control vault settlement—a balance between deformation control and workspace efficiency that could benefit from future numerical-discrete element coupling simulations, as discussed, to optimize support configurations without compromising construction practicality. This methodology represents a preferred approach for large-section tunnel construction in similar strata.

Discussion

This article uses a combination of numerical simulation and on-site monitoring to study the construction method of high-speed railway tunnel in sandy loess and fine sand composite weak strata. The effectiveness of the proposed construction method is verified through on-site construction monitoring. However, this study does have certain limitations.

  1. 1)

    Through indoor soil physical and mechanical parameter testing, reliable input data has been provided for numerical simulation. However, the calculation model in this article adopts a continuous medium constitutive model. For fine sand, the mechanical behavior of loose media is exhibited after tunnel excavation disturbance. Subsequent research used a numerical method coupling finite elements and discrete elements for simulation to explore the instability mechanism of the palm face in such composite formations.

  2. 2)

    The proposed construction method employs temporary vertical supports to control crown settlement during the initial excavation phase. However, these temporary supports reduce the available workspace in actual construction and may interfere with subsequent horizontal rotary jet pile construction. Numerical simulations and on-site monitoring should be conducted to optimize the placement of vertical supports or potentially eliminate them, thereby enhancing construction efficiency.