Geomechanical model test for analysis of surrounding rock behaviours in composite strata

Linken Shi, Hui Zhou,*, Ming Song, Jingjing Lu, Zhenjiang Liu

a State Key Laboratory of Geomechanics and Geotechnical Engineering,Institute of Rock and Soil Mechanics,Chinese Academy of Sciences, Wuhan,430071,China

b University of Chinese Academy of Sciences, Beijing,100049, China

c CCCC Second Highway Consultants Co., Ltd., Wuhan, 430056, China

Keywords:Model test Tunnel excavation Composite strata Deformation and failure mechanism Stability analysis

ABSTRACT Due to the large differences in physico-mechanical properties of composite strata,jamming,head sinking and other serious consequences occur frequently during tunnel boring machine (TBM) excavation. To analyse the stability of surrounding rocks in composite strata under the disturbance of TBM excavation,a geomechanical model test was carried out based on the Lanzhou water supply project. The evolution patterns and distribution characteristics of the strain,stress,and tunnel deformation and fracturing were analysed. The results showed that during TBM excavation in the horizontal composite formations (with upper soft and lower hard layers and with upper hard and lower soft layers), a significant difference in response to the surrounding rocks can be observed. As the strength ratio of the surrounding rocks decreases, the ratio of the maximum strain of the hard rock mass to that of the relatively soft rock mass gradually decreases.The radial stress of the relatively soft rock mass is smaller than that of the hard rock mass in both types of composite strata, indicating that the weak rock mass in the composite formation results in the difference in the mechanical behaviours of the surrounding rocks.The displacement field of the surrounding rocks obtained by the digital speckle correlation method(DSCM)and the macro-fracture morphology after tunnel excavation visually reflected the deformation difference of the composite rock mass. Finally, some suggestions and measures were provided for TBM excavation in composite strata,such as advance geological forecasting and effective monitoring of weak rock masses.

1. Introduction

The sustainable development and effective use of underground space has become a hot issue in the field of geotechnical engineering (Li et al., 2021). To meet different requirements, a large number of tunnelling projects have to be carried out in complex environments. However, the layout of a tunnel is not always determined as per site-specific geological conditions, and sometimes driving a tunnel boring machine (TBM) through composite strata is inevitable (Tóth et al., 2013; Ma et al., 2015). As shown in Fig.1,the composite strata are characterised by sudden changes in physico-mechanical properties of adjacent rock layers, which is considered as one of the most unfavourable ground conditions.TBM excavation mainly involves the interaction between the machine and the ground(Villeneuve,2017;Hasanpour et al.,2020),in this case adverse ground conditions can significantly influence TBM performances and stability of the surrounding rocks (Laughton,2005). Thus, a good understanding of the mechanical behaviours of surrounding rocks during excavation in complex geological environments is necessary to ensure safe construction and subsequent operations of such engineering projects.

Composite strata or mixed grounds,defined as a unit consisting of two or more geomaterials with significant differences in hydrogeomechanical characteristics,may appear on the tunnel excavation surface or in a longitudinal direction(Bosse Marc,2005;Tóth et al.,2013). Rock layers of the same type but with different degrees of weathering also belong to that category of composite strata (Gong et al.,2016).Compared with common homogeneous rock layers,the surrounding rocks behave quite differently in composite strata.These heterogeneous conditions have been reported in many tunnel projects,e.g.KargiTunnel inTurkey(Home,2014),Chengdu Metro Line in China(Gao et al.,2009),and Kranji Tunnel in Singapore(Zhao et al.,2007). Case studies have shown that accidents such as low TBM driving rates,tool wear,TBM subsidence,and jamming have occurred when crossing composite strata sections (Barton, 2000; Zhao et al.,2007). In view of TBM performances and prediction models, many studies have been intensively conducted (e.g. Steingrimsson et al.,2002;Zhang et al.,2010;Ren et al.,2018;Zhou et al.,2018;Agrawal et al., 2019). In contrast, studies on the evolution of deformation and mechanical behaviours of composite strata tunnels are limited.For instance, Chakeri et al. (2015), Li et al. (2019), and Zhang et al.(2020) numerically investigated deformations and ground surface settlements of subways or metros in composite strata.Xu et al.(2019)revealed the mechanical responses of lining with a non-persistent longitudinal crack in soft-hard layers using numerical simulations.Feng et al. (2012) established a three-dimensional (3D) numerical model to analyse the changes of stress and strain in composite rocks and concluded that stress transfers towards and concentrates on the hard rock stratum.Such studies focused on tunnels in shallow strata or composite strata encountered along tunnel axis, primarily using numerical analyses. However, other methods, e.g. model tests, are needed to understand the surrounding rock behaviours of deep tunnels in alternating soft and hard strata.

Geomechanical model test represents an effective and important method of investigating the stability of geotechnical engineering projects, and is widely used in many other fields, such as transportation(Meguid et al.,2008),coal mining(Wang et al.,2015),and hydropower development (Zhu et al., 2010). With this method, the spatial relationship between the underground structure and rock mass can be visualized,and the responses of surrounding rocks due to underground excavation can be revealed intuitively. In the premise of the similarity principle, various scenarios in practical engineering cases can be truly reproduced and the results can be used for numerical modelling. However, for some engineering problems, different methods, such as the finite element method(FEM) and discrete element method (DEM), may yield different results(Lisjak and Grasselli,2014;Liu et al.,2019;Vazaios et al.,2019).Also,due to time and cost constraints(He,2011),it is relatively more difficult and even more dangerous to conduct an in situ test.

Various geomechanical model tests have been carried out to understand the instability mechanism of tunnels excavated in composite strata in recent years. Jeon et al. (2004) used reducedscale tests to investigate the influences of faults, weak planes, and grouting on tunnel stability. Numerical analyses were also conducted to compare and verify the results obtained in model tests.Meguid et al. (2008) reviewed physical model tests in soft ground tunnelling and discussed various approaches for recording soil deformations induced by tunnelling. Concerning underground openings,the physically finite elementary slab assemblage(PFESA)approach by He et al. (2009) was employed to simulate underground roadway excavation in horizontal strata based on infrared(IR) thermography. The test revealed that the IR radiation temperature can effectively characterise the mechanical responses of the roadway after external loading.Baziar et al.(2014)conducted a series of centrifuge tests in a dry sandy soil with a tunnel embedded in soil layer, and found that the tunnel and soil responses were dependent on the tunnel position,soil relative density,and tunnel rigidity. To study the water inrush mechanism during excavation,adverse geological conditions, e.g. faults and karst, were also considered in model tests of underwater tunnels (Li et al., 2016a,2018).In addition,Ghabraie et al.(2017)studied the characteristics of multi-seam subsidence using several sand-plaster physical models,and the results showed that the panel configurations of the two seams have a significant impact on the development of multiseam subsidence. Zhang et al. (2019) carried out a model test and numerical simulations for a deep tunnel in complex rock strata,and the results showed that the evolution and distribution characteristics of displacement and stress are significantly different in various rock strata.Previous studies have provided a new approach to solve various problems encountered in underground spaces;however, most of these works were focused on the deformation and failure mechanisms of the tunnel excavated in singlehomogeneous or isotropic rock stratum under complex geological conditions. Physical model tests are rarely reported on composite strata with different surrounding rock strength ratios,which refers to the ratio of the uniaxial compressive strength(UCS)of relatively soft rock to that of relatively hard rock.

In this paper, a large-scale geomechanical model test based on the engineering background of Lanzhou water supply project was carried out in horizontal composite strata with different surrounding rock strength ratios. Using the experimental results, the evolution and distribution of deformation and stress,as well as the failure behaviours of the surrounding rocks were analysed.Finally,some measures were also provided for TBM excavation.The results can serve as a reference for the design and construction of tunnelling projects in similar composite strata.

Fig.1. Sketch of the different composite strata encountered during TBM excavation.

2. Engineering background

The Lanzhou water supply project is located in the southwest of Gansu Province, China, to transport water from the Liujiaxia Reservoir in the upper reaches of the Yellow River to Lanzhou City. The daily water supply of the project is 2.273×106m3.The main line of the water conveyance tunnel has a total length of 31.29 km and a maximum burial depth of 918 m.It also serves as a link between the reservoir and the area being supplied with water (see Fig. 2a). The sections with double-shield TBM construction are 21.9 km length in total, and the excavated cross-sectional diameter is 5.46 m. After installing the single-layer segmented lining structure for support,the inner diameter of the tunnel is 4.6 m.The geological conditions along the water conveyance tunnel are complex (see Fig. 2b), as it passes through 3 regional faults and 7 small-scale faults.The exposed strata,in order from older to younger,mainly include the pre-Sinian Maxianshan group (AnZmx4), middle to upper Ordovician Wusushan group (O2-3wx2), lower Cretaceous Hekou group (K1hk1), Neogene Linxia formation(N2l1)and Quaternary(Q).

The sectionsusing TBM constructioninthetunnelrange from T5+850m toT30+500 m.Sections T10+700-T13+800 m have a burial depth of 284-431 m and lie in the contact zone between Cretaceous sandstone and Caledonian granite.Asshownin Fig.3a and b,borehole ST07 crosses the boundary between sandstone and granite at a depth of 298.5 m. The former is light purple, with poor cementation and strong disintegration in presence of water,while the latter is greyishwhite with severe weathering and alteration. The magnetotelluric(MT)geophysical prospecting profile(see Fig.3c)clearly shows that the boundary between the two layers in the sections T11 +650 -T12+650 m is undulating and that the hole axis is located near the boundary. Marked differences in the mechanical properties, hydrological conditions,and degrees of weathering of these soft and hard formations can be observed. When a TBM is passing through the composite strata,it may encounter water inrush and collapse occurrences.Moreover,the head may sink locally and cause incidents such as jamming,which can seriously affect the safety and stability of engineering construction.

Fig. 2. (a) Geographical position of the Lanzhou water supply project; and (b) Longitudinal geological profile along the tunnel axis.

3. Physical model tests

3.1. Similarity principle

The premise of a successful model test is to prepare similar materials that meet the test requirements, and the model is required to meet a certain proportional relationship in terms of geometric dimensions, boundary conditions, loads, and material mechanical parameters.The relevant parameters in the model test are listed in Eq. (1). According to the similarity principle and dimensional analysis(Fumagalli,1973),the similarity coefficient Cidescribed in Eq. (2) represents the ratio of the physical quantity values with the same dimensionality between the prototype and the model. In this study, taking into account the effects of simulation objects, model frame size, and boundary effects, the reducedscale of dimension CLwas selected as 27, and the reduced-scale of volume-weight Cγ was selected as 1. A similarity relation in the model test is shown in Table 1.

Table 1 Similarity relationship of parameters adopted in the model test.

where L represents the length; γ is the volume-weight; E is the elastic modulus; ν is the Poisson’s ratio; σ is the stress; ε is the strain; Ciis the similarity coefficient of different parameters in the model test; ipand imrepresent the prototype and model parameters, respectively.

3.2. Model material

According to on-site investigation of the project, the argillaceous sandstone in contact with the hard rock layer along the tunnel has strong hydrophilicity and poor stability in contact with water. In response to this engineering problem, multiple sets of physico-mechanical property tests at different saturation times were carried out using argillaceous sandstone samples sampled from the study site. Fig. 4a shows that the water content of the argillaceous sandstone increases with saturation time,and its wave velocity shows an opposite trend. When the saturation time increased to 144 h, the water content w of the argillaceous sandstone was gradually stabilised at 1.79%-4.2%,and the wave velocity also decreased from 3600 m/s to 1670.3 m/s. Considering the significant water-softening effect of the argillaceous sandstone,argillaceous sandstone samples with water contents of 1.79%,1.97%,and 3.32% were selected as the original rock samples and were defined as medium-hard rock(MH rock),soft rock A(SA rock),and soft rock B (SB rock), respectively. These definitions only represented the different strength characteristics of rock samples under various water contents, and were used to distinguish the three relatively soft rock strata in the model tests.These rocks were used to form different types of composite strata with field weathered granite. The UCS values of the three original rocks are shown in Fig. 4b.

Fig.3. Composite strata revealed at the site:(a)Drill core and(b)the lithological column for on-site sampling of ST07,and(c)Electrical resistivity cross-section along the MT profile of sections T11 +650 - T12 + 650 m. Contact zone refers to the area where sandstone and granite are mixed.

Based on the test results of the argillaceous sandstone samples and similarity principle, the basic physico-mechanical parameters of the above three original rocks and similar materials are listed in Table 2. To ensure the easy maintenance of the model and guarantee the integrity of the material, the material to mimic the surrounding rocks should be composed of iron powder, quartz sand,barite powder, and cement, in addition to water. Due to the difficulty that the simulated material and the prototype do not exhibit the same similarity coefficients for all parameters, the key parameters of the model material (e.g. E,ν, and UCS) should be in accordance with the characteristics of the prototype.The above materials were also used to prepare similar hard rock(H rock)materials as a common rock stratum to form three types of composite strata.The basic parameters are shown in Table 3. Additionally, the strength ratio η of the surrounding rocks was defined to better describe the difference in rock mass of different composite strata and can be expressed by

Fig.4. Laboratory test results for the argillaceous sandstone:(a)Moisture content and velocity tests, and (b) UCS values at different moisture contents.

where σrsand σrhare the UCS values of relatively soft and hard rock layers in the same excavated tunnel, respectively.

The ratios of similar materials in the surrounding rocks are listed in Table 4. As shown in Fig. 5, the surrounding rocks of the three simulated tunnels in this model test were all horizontal composite strata, and the area of different lithological strata in the excavated face is 50%each.The basic information of the simulated tunnels and surrounding rocks is listed in Table 5.

3.3. Model test system

A large-scale geomechanical model test system was developed,as shown in Fig. 6a, to perform the test in view of TBM excavation disturbance in composite strata. The length, width, and height of the steel frame of this test system are 3.95 m, 2.3 m, and 0.3 m,respectively.This system can be used to test a model with a size of 2.2 m(length)×1.78 m(width)×0.3 m(height),as shown in Fig.5.The system is mainly composed of the following parts: stress loading system, excavation simulation device, and information monitoring system.

Five hydraulic jacks were set at the top of the frame to simulate vertical loads, and four on the left and right sides were used to simulate horizontal loads(see Fig.6a).The test equipment applied pressure through a series of 30 mm thick loading plates and can be used to carry out uniaxial static test and biaxial plane strain test.At the same time, nine pairs of steel plates installed at the front and back of the equipment were also used to form displacement constraints during the laying of the model materials. The diameter of each jack was 220 mm,with a maximum load of 200 kN.As shown in Fig. 6b, normal stress was applied at the vertical and horizontal boundaries of the model, and the displacement was fixed at the bottom of the model.

3.4. Model measurement

As shown in Fig. 7, to reveal the dynamic responses of the surrounding rocks of the composite strata under TBM excavation disturbance, this experiment used fibre Bragg grating (FBG) strain sensors and the digital speckle correlation method (DSCM) to obtain information of the strain, stress, and displacement of the surrounding rocks in tunnels. FBG sensor is small in size, resistant to electromagnetic interference, and highly accurate (Li et al.,2009). Regardless of the influence of external temperature, a 16-channel demodulator (see Fig. 7a) was used to collect the phase change of the reflected wavelength of the grating during excavation. After the excavation of the rock body near the cave wall, the surrounding rock stress was readjusted. Using the elastic mechanics formula in Eq.(4),the radial stress of each monitoring point in the surrounding rocks can be obtained. Since the DSCM (see Fig.7b)does not damage the model material during the monitoring process, this method has also been used effectively in model tests(Li et al., 2016b).

where σi, Δσ, and σ represent the radial stress, stress change amount, and initial in situ stress of the measuring point,respectively.

In this model test, the arrangement of monitoring elements is consistent in the surrounding rocks of the three excavated tunnels(see Fig. 8a). The monitoring sections of Tunnels 1, 2, and 3 were marked as I,II,and III,respectively.As shown in Fig.8b,during the laying of the model materials, six fibre gratings were buried in plane ABCD, which is at the middle position along the excavation direction of the model. The first measuring point on each measuring line is 1 cm away from the periphery,and the distances between the marked measuring points#1-5 are 3 cm,5 cm,7 cm,and 14 cm, respectively. The front surface (A1B1C1D1) of the excavated tunnel was also sprayed with speckles to capture the overall deformation of the surrounding rocks.

3.5. Model test procedures

The main procedures are summarized as follows:

(1) Similar material preparation. According to the ratios of similar materials listed in Table 4, the raw materials of the similar materials were weighed separately and then mixed uniformly by a mixer (Fig. 9a) to generate a model material that met the test requirements.

Table 2 Physico-mechanical parameters of relatively soft rocks in the composite strata.

Table 3 Basic properties of similar materials of hard rock in the composite strata.

Table 4 Component ratios of similar model materials.

Table 5 Basic information on the surrounding rocks in the excavation tunnel.

(2) Model ground construction. The model body was produced by pouring layers from the bottom to the top. Each layer of the material was filled and rammed until the designed height was reached. When laying SA and SB rocks, partitions were erected at the boundaries to complete the filling of the two materials in sequence.

(3) Embedding of the monitoring element.When the model was poured to the monitoring position, the position of the monitoring point was measured with a ruler. The fib

re grating was embedded in the dense material and slotted(Fig. 9b), and then the wire was passed through the circular hole reserved on constraining steel plate.

(4) Loading and tunnel excavation. After the model body was cured for 20 d, the simulated load was applied using the hydraulic loading system and an excavation process was carried out.A horizontal load of σx=0.45 MPa was applied to the left and right sides, and a vertical load of σy= 0.4 MPa was applied to the upper surface.The TBM excavation of the project site was simulated by the stepwise excavation with a set of self-developed excavation equipment (Fig. 9c), which mainly includes a drilling rig that provides torque,a slide rail that guides the TBM forward, and a cutter head with a diameter of 20 mm that used to simulate rock breaking. A total of 12 excavation steps were set.Tunnels 1,2,and 3 were excavated in sequence. To avoid mutual influence between the tunnels during TBM excavation, each tunnel was separated at a safety distance greater than three times the diameter of the tunnel.

(5) Physical information measuring. After each excavation step was completed, a demodulator and a high-speed camera(Fig. 9d) were used to collect monitoring information of the surrounding rocks in composite strata.

Fig. 5. Layout drawing of the composite strata in the geological model (unit: cm).

Fig. 6. (a) Geo-mechanical model test system, and (b) boundary conditions of the model. The numbers in (a): 1 - Steel plate; 2 - vertical loading block; 3 - horizontal loading block; 4 - high pressure pipe; 5 - vertical jacks; 6 - four jacks; 7 - loading plate; and 8 - horizontal jack (four in total).

4. Results and analysis

4.1. Feature analysis of strain

By analysing the data recorded by the FBG sensors, the strain evolution processes at the monitoring points can be obtained in terms of microstrain. Taking sections I and II as examples, Fig.10 shows the strain states of the monitoring sections in two typical composite formations, including horizontal and vertical radial strains. Taking I-L1-12 in the legend as an example, I indicates the monitoring section, L1 the line number, and 12 the number of excavation steps.

Fig. 7. Measuring method used in the physical model: (a) FBG measurement system,and (b) DSCM.

Fig. 8. Design of the monitoring elements in the tunnel of the model test: (a) Spatial distribution of monitoring elements; and (b) Distribution of FBG strain sensors and digital speckle area (DSA).

The horizontal and vertical radial strains of the rock mass in the composite strata show a similar trend,as shown in Fig.10.The change trend of internal strains at different measuring points is characterised by"L"shape.Along the radial direction of the tunnel,the strain of the deep surrounding rocks is much smaller than that of the shallow surrounding rocks.All strain curves intersect at measuring point#4 and tend to be small, indicating that the surrounding rocks are significantly affected by excavation disturbance in the range of 1.5R(R is the tunnel radius). Moreover, the strain in MH and SA rocks decreases more rapidly, which indicates that tunnel excavation has a greater impact on soft rock mass.Due to the differences in mechanical strength and deformation of the composite strata, the strain of soft rock is obviously greater than that of hard rock. Specifically, in the vicinity of the interface,the maximum horizontal radial strains in the harder and softer rock masses in Tunnel 1 are 2726 με and 956 με,respectively(Fig.10a),and those in Tunnel 2 are 1541 με and 1066 με,respectively(Fig.10b).Interestingly,by comparing the horizontal and vertical strains,it is concluded that the strain at the base of the arch in thesectionI(upperhardandlower soft formation)isgreater than that in the vault,and the hance is somewhere in between.In contrast,in the section II(upper soft and lower hard formation),the strain is the highest at the vault,followed by those at the hance and at the base of the arch.

Fig.10. Distributions of radial strain in the rock surrounding the tunnel: (a)Section I;and (b) Section II.

A marked feature of the strain distribution in composite strata is that there are significant differences in the strains of the rock masses on both sides of the interface. Fig.11 shows the maximum strain difference curves between two adjacent rock masses after tunnel excavation. The strain difference at different points on the same measurement line varies greatly.Measuring points#1 and#2 near the periphery were significantly disturbed by TBM excavation;therefore, the strain difference is much larger than that at other points.In addition,for different types of composite strata,the strain differences of sections II,III,and I decrease in turn.The relationship between the surrounding rock strength ratio η and the maximum horizontal strain of the rock mass after final excavation is shown in Fig.12.The maximum horizontal radial strain difference is inversely proportional to η. We also observed that with decrease in η, the proportion of hard rock’s maximum strain in the relatively soft rock mass gradually decreases. Specifically, in Tunnels 1, 2, and 3, the proportions are approximately 69.1%, 39.6%, and 50.3%, respectively. Thus, the physico-mechanical properties and the degree of rock mass difference are important factors to guide TBM construction in composite strata.

Fig.12. Horizontal strain ratio,strain difference,and strength ratio in composite strata after the final excavation.

Strain monitoring data show that both excavated and unexcavated rock masses are affected by TBM excavation disturbance. As shown in Fig.13,one measuring point is selected in each of the three monitoring sections to analyse the strain behaviours of the surrounding rocks during excavation. Taking I-P52as an example, I represents the monitoring section,and P52represents the measuring point #2 on measuring line L5. The strain evolution curve mainly undergoes three phases during excavation: (1) Pre-development phase: As the working face advances, the balance of forces in the rock mass is disrupted;(2)Sharp increase phase:When the working face advances to the range of 12.5-17.5 cm, corresponding to excavation steps 6-7, the strain increases rapidly; and (3) Stabilization phase:The strain gradually increases and stabilises at a certain level.In general,the impact of TBM excavation on the deformation of the surrounding rocks is mainly concentrated within a certain range ahead of or behind the monitoring sections, but the surrounding rocks without excavation is also affected.

Fig. 9. Model test procedures: (a) Stirring materials with a mixer; (b) Embedding the FBG sensor; (c) Excavation with a TBM; and (d) Model measurement.

4.2. Variation analysis of stress

During TBM advancement, the stress in the rock mass near the tunnel is released due to excavation disturbance. The dynamic adjustment of the surrounding rock stress can cause deformation in the rock mass. According to the results of the strain analysis in Section 4.1,the change in the radial strain at measuring point#4 is really small while the strain at measuring point #5 is approximately zero. The radial stress value here can be regarded as the initial in situ stress. Thus, we can use Eq. (4) to approximate the stress variation and radial stress of the surrounding rocks.

As shown in Fig. 14, because of spatial limitations, a typical upper soft and lower hard composite stratum (section III) was selected to analyse the evolution of the radial stress in the surrounding rocks during tunnel excavation, and the two rock mass types are H and SA rocks, respectively.

(1) During TBM excavation, the radial stress in the soft rock above the interface is less than that of the hard rock below.Different from a generally homogeneous stratum, for the same excavation step, a significant difference in the radial stress of the two rock masses was observed in composite stratum. The process of rock mass excavation was accompanied by stress redistribution of surrounding rocks. In contrast, soft rock is more susceptible to excavation disturbances; therefore, the stress release of the soft rock is also much greater than that of hard rock.

(2) The radial stress of the surrounding rocks changes stepwise with the distance between the measuring point and the periphery.The stress value at the measuring points near the periphery is much smaller than that at the distant measuring points. Meanwhile, as the excavation progresses, the radial stress gradually decreases while the stress release rate gradually increases.

(3) The radial stresses of the two rock masses near the interface in the composite strata show the same change trend. The stress value at the soft rock is smaller than that at the hard rock, while the change in stress release rate follows the opposite trend.

Furthermore,to analyse the stress variation patterns of different composite strata, the radial stress results of the three monitoring sections are plotted in Fig.15,showing the radial stress distribution curves before and after the TBM advances to the monitoring sections.When the tunnel is excavated to 17.5 cm(step 7)in the axial direction, the radial strain is significantly less than that at 12.5 cm(step 5). Since the tunnel has been excavated to the rear of the monitoring section, the stress release of the surrounding rocks is further accelerated.For the same excavation step,in theory,the SB rock with larger stress release actually exhibits a larger radial stress than the other two soft rocks do. This is because the calculated stress cannot accurately reflect the actual value at the measuring point to a certain extent,but it can facilitate a better understanding of the change trend of the composite strata stress. In general, the radial stress of the relatively soft rock is less than that of the hard rock represented by the black line for both the upper hard and lower soft formation(section I)and the upper soft and lower hard formation (sections II and III), which indicates that the weak rock mass in the composite strata results in the difference in the response of the surrounding rocks.

Fig.11. Maximum radial strain difference of the surrounding rocks after excavation.

Fig.13. Variations in the internal radial strain during tunnel excavation.

4.3. Analysis of deformation and failure characteristics

The strain data obtained by the FBG sensors represent the deformation of the monitoring section inside the surrounding rocks; however, they cannot reflect the overall deformation of the tunnel. Observing the surface displacements of the tunnel with naked eyes is difficult.The DSCM can be a good choice for obtaining the displacements of the surrounding rocks(Li et al.,2016b).During the model test,a series of excavation photographs of the tunnel in section II (upper soft and lower hard formation) was successfully captured. After processing by PhotoInfor software, the evolution process of the surrounding rock displacement field was obtained.This data set is taken as an example to analyse the characteristics of the overall deformation of the tunnel in the composite strata, as shown in Fig.16.

Fig.16. Displacement evolution of the surrounding rock with excavation in upper soft and lower hard strata: (a) Step 1; (b) Step 6; (c) Step 12; and (d) After tunnel excavation.

Fig.14. Radial stress distributions in the surrounding rocks near the interface in section III: (a) Step 3; (b) Step 6; and (c) Step 9.

Fig.15. Radial stress curves of the surrounding rocks in different composite strata.

Fig.16d shows a photograph of the tunnel captured during the test after excavation to the monitoring section. A part of the rock mass has been excavated, and the interface of the composite rock mass is clearly visible. The surrounding rocks were almost undisturbed at the early stage of excavation, indicating that the rock mass has good self-stability,as shown in Fig.16a.As TBM advanced,the surrounding rocks gradually lost the support of the rock mass.When TBM excavation reached the monitoring section (see Fig. 16b), the displacement of the surrounding rocks changed dramatically during the dynamic adjustment of the stress caused by the stepwise excavation. As a result, the surrounding rocks exhibited nonuniform deformation in the vertical direction.Compared with the small deformation of the hard rock, the overlying soft rock deformed significantly in most areas. Eventually,when entire rock mass in the tunnel was excavated (see Fig.16c),the deformation difference between the surrounding rocks of the composite strata increased further. The local maximum displacement was approximately 4.8 mm,which occurred at the vault of the tunnel.However,part of the rock mass at the base of the hard rock arch remained stable. In particular, the surrounding rocks underwent a significant deformation at the interface between the two rock masses. Because the overall deformation of the surrounding rocks of the tunnel is significantly greater than the relative deformation of the rock mass at different positions in the shallow part,the monitoring results show that the deformation is relatively uniform.

Fig.17 shows the macro-fracture morphology of the surrounding rocks recorded from the back of the model after tunnel excavation. The rock masses in the composite strata exhibit different macroscopic deformation and fracture characteristics. The deformation and destruction of the surrounding rocks are most intense in the soft rock masses.As shown in Fig.17,the overlying soft rock exhibits a movement towards the free surfaces at the spandrel and vault. Specifically, the inner surface (labelled B) and especially the outer surface (labelled A) of the tunnel are accompanied by the collapse of the rock mass. At the same time, tensile cracks at the base of the arch are observed in the hard rock (see Fig.17a) while multiple intersecting shear cracks appear in the soft rock (see Fig. 17b), indicating that the composite rock mass also shows different failure mechanisms after excavation.For the surrounding composite strata, an important feature is the difference in hydromechanical properties of the two rocks. Consequently, we observed another key feature at the interface of the composite strata,i.e.the uncoordinated deformation of the surrounding rocks(labelled C in Fig. 17a). As shown in Fig. 17b, the macroscopic characteristics of the tunnel recorded 2 h after excavation show that the soft rock mass collapses to a small extent and that the hard rock mass hardly changes.

The displacement evolution obtained by the DSCM(see Fig.16)and the macro-fracture morphology recorded by the camera (see Fig.17) intuitively reflect the deformation differences in the surrounding rocks composed of composite strata. The abovementioned analysis of the strain and stress in the surrounding rocks also reveal the pattern of dynamic responses inside a rock mass composed of composite strata after TBM excavation disturbance.

Fig. 17. Macro-fracture morphology of surrounding rocks with upper soft and lower hard formation: (a) At the end of excavation; and (b) 2 h after excavation.

5. Discussion

A significant difference in the behaviours of surrounding rocks can be observed during TBM excavation in composite strata. Specifically, the strain, stress, and displacement obtained from the model test (Figs.10,14 and 16) are significantly larger in the relatively hard rock strata than that in relatively soft rock strata.Similar phenomena have also been observed in other studies.Tien and Tsao(2000) and Tien et al. (2006) performed conventional triaxial compression tests on layered rock specimens and observed the phenomena of clear dislocation near the interface of two different materials.This experimental observation is in good agreement with the phenomena of “uncoordinated deformation” shown in Figs.16 and 17.By using FLAC3D,Feng et al.(2012)found that the stress and strain features are quite different near the contact zones of the softand-hard rock strata. The range of plastic deformation is larger in the soft rock zone than that in the hard rock zone. The process of TBM excavation in composite strata was reliably reproduced in the model test, which also reveals that the surrounding rock strength ratio is a key factor affecting the response of composite strata.

In normal operations,a TBM maintains an eccentric relationship with the surrounding rocks.However,in some extreme cases,TBM may be jammed or buried in composite strata (Gao et al., 2009;Hasanpour et al., 2020). This is mainly because the surrounding rock convergence deformation is greater than the gap between the shield and rock mass. The surrounding rocks squeezed the shield and caused the TBM jamming. Based on the analysis of the test results, two suggestions related to TBM construction in composite strata are proposed:

(1) Strengthen advanced geological prediction. Specifically, an advanced drill bit and forecast system can be installed on the TBM to assist in predicting the geological conditions ahead.

(2) Focus on monitoring the stress of the weak formations in the composite rock mass and the convergence rate of the surrounding rocks. In particular, grouting reinforcement measures should be adopted appropriately, or the diameter of the TBM excavation can be enlarged in upper soft and lower hard strata to increase the space for convergent deformations of the surrounding rocks.

In addition, along the axis direction of the tunnel, the alternating soft and hard rocks are also expected to be encountered with tunnel excavation (Ma et al., 2015), which also poses a threat to TBM performance and the stability of rock mass.

6. Conclusions

This paper analyses the dynamic responses of surrounding rocks related to TBM excavation,such as deformation characteristics and fracture mechanism, based on geomechanical model tests conducted on two typical types of horizontal composite strata. According to the test results,the conclusions of the study are drawn as follows:

(1) The model test shows that the change trend of the strain of the surrounding rocks at different measuring points is characterised by "L" shape. In composite strata, significant differences in rock strain on both sides of the interface can be observed, which is inversely related to the strength ratio of surrounding rocks. The strain evolution curves can be roughly divided into three phases during excavation,i.e.predevelopment deformation phase, sharp increase phase, and stabilisation phase.

(2) The process of surrounding rock excavation is accompanied by stress redistribution. The radial stresses of the two rock masses near the interface of the composite strata show the same change pattern.No matter it is an upper hard and lower soft formation or an upper soft and lower hard formation,the radial stress of the relatively softer rock mass is less than that of the hard rock.

(3) The evolution process and macro-fracture characteristics of the surface displacements of the rocks surrounding the tunnel obtained by the DSCM are analysed, and the results can help understand the different behaviours between composite strata.

(4) The test results reveal the dynamic responses of the surrounding rocks in the composite strata.For the two types of composite rock masses that may be encountered during TBM excavation, the associated risks are analysed and relevant suggestions are provided.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was supported by the National Natural Science Foundation of China (Grant No. 41941018) and the National Program on Key Basic Research of China (973 Program) (Grant No.2014CB046902). The partial support from the Hubei Province Natural Science Foundation Innovation Group (Grant No.2018CFA013) was also gratefully acknowledged.