Research of Transitional Failure Mode as Damage Evolution in Rock Wall

+e stress condition of tunnel surrounding rock mass is complex. +e stress concentration of in situ brittle rock mass caused by excavation results in localized damage evolution parallel to the free face, which is called surface instability.+e rock wall shows the transition characteristics of the failure mode with the distance from the surface to the depth. Low strength surface instability and transition failure modes of the tunnel’s rock wall are common in deep condition but cylindrical specimens cannot simulate stress state of rock wall surface well in conventional rock mechanics tests. +is paper conducted the indoor experimental study of the biaxial stress state and studied the surface instability of samples. An indoor test device for the simulation of transitional surface failure of the rock wall was developed.+rough a biaxial stress loading test on the rectangular rock sample, the damage process and crack development of rock samples were analyzed, and the law of stress and strain related to the failure mode transition was characterized as well. Based on test results and strength analysis, an explanation of the failure theory and its corresponding model are proposed based on the maximum strain strength theory. Furthermore, this paper concludes that the failure mode of surface instability presents transition feature from brittle to ductile with the increase of distance from the surface to depth.


Introduction
With the construction of tunnel engineering, petroleum industry, and disposal tunnel of nuclear wastes in the past 40 years, the deformation and failure characteristics of deep brittle hard rocks have been extensively investigated and studied. Excavation in deep hard rock causes a sudden release of high intensity energy stored in the rock mass, which plays an important role in inducing surface spalling and rock burst [1][2][3]. It has been widely accepted that the spalling and rock burst near the surface of rock wall is a typical brittle failure [4,5]. W. f. Brace et al. studied the entire process of brittle-shear fracture under compression and proposed that the failure process of brittle rock can be divided into the following stages [6,7]: fracture closure, linear elastic deformation, fracture initiation, stable fracture propagation, fracture coalescence, development of unstable fractures, failure, and fracture mode after peak strength.
e initial stress (σ ci ) and fracture damaging stress (σ cd ) can be determined according to the stressstrain relationship of uniaxial compression test. e failure of deep-buried brittle hard rock is an unconventional brittle failure that has the dual mechanism of brittle failure and shear failure. Recent research has mainly focused on the brittle-shear properties of high-stress rock mass.
Hoek-Brown proposed that the brittle rock mass will transform into the ductility when enough confining pressure is applied [8]. Hoek-Brown failure criterion provides an acceptable estimate of the peak strength for shear failure but a cutoff has been added for tensile conditions. However, the criterion does not adequately explain the progressive coalition of tensile cracks and the final shearing of the specimens at higher confining stress [9]. Parterson conducted experiments on the marble at room temperature and proved that the rock deformation behavior changed from brittleness to ductility with the increase of pressure [7]. Mogi has published similar experimental results and pointed out that the brittle-toductile transformation is usually related to rock strength [10]. On the other hand, some studies have shown that the overall strength of the rock increases with the increase of depth, and the rock becomes more brittle along with the increasing depth, which has been verified in the field monitoring of the Mine by experimental tunnel and the deep underground laboratory at Auspar [11]. Rock mass damage at great depths near underground openings is often of shear-tensile failure character [12]. To further study the failure mode transformation of the surrounding rock along the vertical inward direction of the chamber wall and the tensile failure of the rock wall and to simulate the bidirectional stress state of the chamber, a chamber wall failure simulation instrument was developed to study the transitional failure mode of the rock wall [13]. Maximum principal strain theory and maximum shear stress theory are two failure strength theories which are often used in the failure analysis of rock and rock mass. e difference between them is failure mode which is controlled by extensional strain or shear stress. Barton and Shen [14] mentioned that many scholars pointed out that the importance of axial splitting in response to the compressive stress fields, when neighbouring surfaces were available to allow the "lateral" extensional strain to cause extensional fracturing. Stacey [15] showed that, for a material that showed linear deformation behaviour, the onset of failure and the depth of failure could be related to a consideration of extensional strain. Based on the failure features analysis, the maximum principal strain formula is proposed for the initiation of surface instability calculation in this paper.

Test Methods for the Transitional Failure
Simulator of the Rock Wall 2.1. Manufacture of the Sample Chamber. Referring to the design idea of the surface instability instrument developed by Kao [16], to consider the geometric factors (plane shape) and the biaxial stress state of the surrounding rock surface of tunnel, techniques such as endoscopy laser speckle technology, LVDT technology, digital image technology, and acoustic emission technology are utilized to measure and locate the internal and external deformation of the specimens. Compared with the traditional triaxial test, rock wall surface damage is progressive; this would be involved in the research on the transformation of failure mode and the geometric factors [17]. Meanwhile, the axis displacement, surface displacement, internal displacement, cracking signal, pressure data, and other relevant information require to be obtained. Furthermore, a servo loading test should be carried out to simulate the stable failure process of the surface rock of deep tunnel, which is in cooperation with a lateral displacement-control technique. e instrument is presented in Figure 1.
To ensure the measurement accuracy, the LVDT sensor was calibrated to obtain a linear range of around 5 mm, which was used as the measurement range in the test. Figure 2 shows the calibrating data as follows.

UCS and Triaxial Compression Test.
Uniaxial and triaxial compression tests were performed on 23 circular cylinders, including 14 sandstone specimens and 9 granite specimens. e specimens were prepared with diameters of 30 mm, 50 mm, and 100 mm to evaluate the size dependence of the strength. e loading program was in accordance with ISRM standards and tested within a servo-hydraulic system of TAW-2000. Failure specimens are shown in Figure 2, and the results are listed in Table 1. e number of images is defined as "rock type-size-confining pressure-ID." For example, S3060-0-1 means the first sandstone specimen with 30 mm × 60 mm size and 0 confining pressure.
It can be concluded that the strength of this kind of sandstone would show a deviation between different sizes. When the diameters of specimen change from 30 mm to 100 mm, the uniaxial compression strength increases from 40 MPa to 59.35 MPa. If the size is larger than 50 mm, the UCS changes a little and the trend is the same as the findings of Chen Yu et al. [18], Peng Chen and Zhiwei Zhou [19], and Chuangye Wang and Xiaoya Du [20].
Strength curves of cylindrical sandstone samples with different radiuses are shown in Figure 3. e transition of failure modes can be obtained, and the samples show more ductile feature with the increasing of radial stress. Similarly, a set of tests for chalk were conducted by Ali Tarokh et al. in 2016 [21], and the results showed the transition of failure modes (see Figure 4). Both in sandstone and granite tests, the orientation of the failure planes varies from vertical to horizontal with the increase of mean stress (see Figure 5) which is consistent with the announced results of Tarokh's research.

Preparation and Operation of the Surface Instability
Simulation Test

Surface Treatment of Rock
Samples. e spot image was made by the random seal method. In the test, the speckled surface of the sample needs to contact with the acrylic plate and generate intermediate principal stress, and the coating will increase the thickness of the sample [23]. erefore, through the steps of surface cleaning-background coating-speckle fabrication-surface lubrication, a relatively clear and distinguishable speckle image with a uniform coating is formed. 2 Advances in Civil Engineering

Test Operation Procedure
Step 1. Cyclic preloading and unloading were applied before the test to extrude excessive lubricant and form a uniform membrane, specifically, two cycles were conducted with a preloading start from 0 kN to 100 kN at the rate of 2 kN/s and then unloading to 1 kN.
Step 2. e experiment required to set up the imaging angle and focal length of the camera in VIC-3D system; turn on the VIC-3D system and set the frame rate of the    Advances in Civil Engineering image collector to 1 frame/second until the end of the experiment. e destruction simulation experiment is shown in Figure 7.
Step 3. e test loading was divided into two steps: (a) loading to 30% of the uniaxial compressive strength of the sample at 0.5 kN/s load-control rate. (b) e control mode was switched to the transverse LVDT displacement-control at the rate of 0.01 mm/min. e loading proceeded until the failure of the sample and stopped after the occurrence of multiple fractures.

Fracture Expansion
Image of the Specimen. e 80 × 90 × 110 mm cubic sandstone and 80 × 90 × 60 mm cubic granite specimen were loaded, respectively, and the sample results are shown in Figure 8. In Figure 8, the right side of the sample is the free surface. It can be seen from the figure that the failure of the sample in the bidirectional stress state first occurs at the free surface end, mainly near-vertical splitting, and then the cracks extend to the depth. e crack spacing increases as the crack angle increases.
According to the definition of x, y, and z directions as shown in Figure 9, the imaging results (positive as tensile strain and negative as compressive strain) of S8090110-3, S8090110-4, and S8090110-5 in the critical state are shown in Figure 10, and the tensile strains at failure are 2157 με, 2423 με, and 2401 με, respectively.
According to the DIC images, with the increase of the pressure, the deformation of the sample below side of the free face reaches the limit value firstly, and a local vertical crack appears. Afterward, the cracks will not stop developing and expanding along a similar vertical direction until the entire sample is destroyed. By comparing the global strain measured by the LVDT sensor with the local strain of the DIC image, the DIC image would be more suitable for the monitor of local cracks development.

Sandstone.
e test results of the vertical stressvertical strain curve and the test results of the vertical stress-transverse strain curve of cubic sandstone under test loading are shown in Figure 11. It can be found that  Table 1) of the same stone sample, indicating that the shape of the sample does not influence its strength.

Granite.
e results of the vertical stress-vertical strain curve and the vertical stress-transverse strain curve of cubic granite are shown in Figure 12. It can be found that the failure of granite samples occurred when the transverse strain reached 4500. e strength will increase again after the first cracking, even exceeding the first cracking strength.

Analysis of Transitional Failure Mode.
Failure of rock is a complex process and failure mechanisms which would show a transition feature according to stress level and loading condition [24,25]. Previous research proposed that an increase in effective pressure shifts the macroscopic failure mode from brittle to ductile [26]. According to the difference   Advances in Civil Engineering of microcracks propagation, Charles Derek Martin [27] divided the stress-strain curves of brittle rock into five regions, which represent the process of microcracks propagation, and indicated that the initial step in the failure process is sliding along grain boundaries, with the wing or axial cracks developing much later in the failure process. e transition of failure modes will cause the change of the orientation of the failure plane, which decreases from 90°to 0°with the increase in mean stress [21]. From the analysis of fracture angle, cracks spacing, failure location, and stress-strain relationship curves shown in Figures 8-12, it can be seen that the first fracture of the samples occurs close to the free surface, which could be thought as a type of tensile failure. As the pressure continually rises, shear failure appears and a sliding surface can be found in deeper location of samples. After the fractures are fully developed, it can be found that the transitional failure mode from tensile failure to shear failure occurs from the free face to the depth, which simulates the stress and strain characteristics of the rock in the biaxial stress state and reproduces the stress state of the tunnel rock wall. In the uniaxial and triaxial tests in the laboratory, the failure mode of the sample is a combination of tensile and shear failure, while the first failure of the rock in the surface instability chamber is in tensile failure mechanism; however, the failure mode has gradually transformed to shear mode as the increase of the failure depth.

Research of Tunnel Wall Strength Based on the Maximum Principal Strain Theory
Under uniaxial compression or biaxial loading, the failure mode of the sample is mainly tensile failure rather than shear failure at first, so the strength is different from the value predicted by the Mohr-Coulomb criterion. In explaining these destruction phenomena, many scholars have put forward different viewpoints. Martin proposed the CWFS model [28] and emphasized that damage development caused that as friction was mobilized in the sample and led to the reduction of cohesion [29]. Martino and Chandler thought that the damage zone where the stress redistribution happened was too small to permanently change rock properties [30]. Brace and Gramberg have first provided explanations on the tensile crack under uniaxial compression condition [31,32] that is because the tensile strain is caused by tensile damage. Stacey attempted to use the theory of maximum principal strain to forecast the brittle failure mechanism of deep massive quartzite in South Africa; he suggested that for those materials with obvious linear deformation behaviors, the failure could be considered to relate to the tensile strain [15].
Since the failure mode of the surrounding rock is tensile failure, it cannot be explained by the Mohr-Coulomb criterion. Based on the maximum principal strain theory proposed by Stacey and Nick Barton and generalized Hooke's law, the following failure conditions can be obtained: where E is the elastic modulus, μ is Poisson's ratio value, σ 1,2,3 stands for principal stresses, and ε 1,2,3 means principal strains. If the calculated extensional strain was greater than the critical strain, then spalling would occur. So, if μ (σ 1 + σ 2 ) > σ 3 , then (negative) extensional strain will occur. According to the generalized Hooke's law, By iterative substitution of equation (2) in (1), the following can be obtained: erefore, a simplified maximum principal strain theory equation can be obtained: In the test, the minimum principal stress is 0 for free face, and equation (4) yields Elastic modulus, E, and Poisson's ratio, μ, can be calculated according to the UCS test. e parameters of sandstone samples are 7.07 GPa and 0.24, respectively. From equation (5), the critical principal strain ε c can be given as −2.526 × 10 −3 (negative value means tensile strain). Initial fracture creates strain values measured from LVDT sensor and DIC image (the measured value is 2300 and 2400, respectively) which are consistent with ε c , so it can be considered that the initial failure mode of sandstone samples is a tensile failure and the maximum principal strain theory has good applicability to explain the surface failure of the rock wall. Since the failure of the sample is not only shear failure in the uniaxial case, the strength value of the rock is higher in the uniaxial case when the Mohr-Coulomb criterion is used to predict it (see Figure 13). However, considering the plane strain problem under biaxial loading, especially the strength feature of the rock wall because the specimen failure is caused by tensile failure, the maximum principal strain theory is more in line with the actual situation (see Figure 13). e maximum principal strain theory can also be suitable for the calculation of uniaxial compression strength when the stretching lineation strain failure takes place (see Figures 3 and 8). Li Yuan et al. proposed a brittle-shear mixing strength criterion and considered rock failure as brittle fracture and shear failure [33]. e coefficient b was proposed to describe the transition from brittle failure to a shear failure caused by constant confining pressure; see the following: b � e mσ 3 + n , where (C 1 σ 3 + C 2 ) represents the brittle fracture strength;  indicates the transition coefficient from brittle failure to shear failure ranging from 0 to 1; m and n are parameters fitted out by data.
If equation (6) is used to explain the transitioning of failure mode with the stress state changing, the results are shown in Figure 13. Mohr-Coulomb strength line could be fitted by the higher stresses data and the maximum principal strain strength line could be calculated by E, μ, and UCS of 30 mm size sample mentioned in Table 2. But, according to the test condition of the uniaxial and triaxial compression test, equation (2) should be changed to e deductive program is the same as equations (3) to (5); it can be given that Because the average UCS value is 39.92 MPa, the related critical strain can be calculated by equation (9) as −1355 × 10 −3 . e strength line of ε c is illustrated in Figure 13 as the red line. Equation (6) is fitted by the test data shown in Figure 13. C1 and C2 are confirmed by slope and intercept of equation (9) and D1 and D2 by the parameters of Mohr-Coulomb strength theory which is fitted in Figure 13. Fitted and calculated parameters are given in Table 2, and the correlation coefficient arrives at 0.9961. e test data and transition trend of sandstone samples can further illustrate the reason why the in situ strength of tunnel wall is lower than the predicted strength value by Mohr-Coulomb theory; that is, the tensile failure is not considered, which leads to a higher predicted strength value.

Conclusion
Experimental study of surface instability of samples was carried out. An indoor test device for the simulation of transitional surface failure of the rock wall was developed. rough a biaxial stress loading test on the rectangular rock sample, the damage process and crack development of rock samples were analyzed and the law of stress and strain related to the failure mode transition was characterized as well.
e following conclusions can be deducted: (1) e surface fracture of the rock wall is a transitional failure mode. It is found in the simulation test that, under the condition of biaxial loading, with the continuous rise of pressure, the crack caused by tensile failure first appears at the end of the free surface and followed by shear failure. e experiment simulates the stress-strain characteristics under the in situ stress state of the rock wall. is phenomenon is also applicable to granite, and it is concluded that tensile failure is the main reason why the wall strength of both shallow and deep tunnel is lower than indoor strength. (2) e maximum principal strain theory is applied to explain the tensile failure of samples. Under the condition of biaxial stress, the maximum principal strain theory was deduced. Additionally, test data were used

Data Availability
e data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper.