Shoin Zhi ,Guosho Su ,Shunde Yin ,Bin Zho ,Liuin Yn

a Key Laboratory of Disaster Prevention and Structural Safety of Ministry of Education,School of Civil and Architecture Engineering,Guangxi University,Nanning,530004,China

b Department of Civil and Environmental Engineering,University of Waterloo,Waterloo,ON N2L 3G1,Canada

c State Key Laboratory of Geohazard Prevention and Geoenvironment Protection,Chengdu University of Technology,Chengdu,610059,China

Keywords:Rockburst Strainburst Hard brittle rocks True triaxial test Acoustic emission(AE)

A B S T R A C T Rockburst is a typical rock failure which frequently threatens both human life and construction equipment during highly stressed underground excavation.Rock lithology is a control factor of rockburst.In this paper,rockburst tests were conducted on rectangular prismatic specimens of six types of intact hard brittle rocks,i.e.granodiorite,granite,marble,basalt,sandstone and limestone,under one-free-face true triaxial loading conditions.With the use of high-speed cameras,an acoustic emission(AE)system and a scanning electron microscope(SEM),rockburst of different rocks was investigated.The results show that the strainbursts of granodiorite,granite and marble were accompanied by tensile splitting near the free face,and consequently were relatively strong with a large amount of fragment ejection and kinetic energy release.For basalt,sandstone and limestone,failure was primarily dominated by shear rupture.The strainbursts of basalt and sandstone were relatively small with minor fragment ejection and kinetic energy release;while no burst failure occurred on limestone due to its relatively low peak strength.Rock strength,fracturing and fragmentation characteristics,and failure modes of different rocks can significantly affect rockburst proneness and magnitude.The AE evolution coupled with SEM analysis reveals that the differences in the inherent microstructures and fracture evolution under loading are the primary factors accounting for different rockbursts in various rock types.

1. Introduction

Rockburst is a typical rock failure,and it could eject large quantities of rock fragments with high kinetic energy(Cook,1965a;Hedley,1992;Ortlepp and Stacey,1994;Kaiser et al.,1996;Li et al.,2007;Feng,2017;Cai and Kaiser,2018).Strainburst,which occurs in relatively intact hard brittle rocks around underground openings,is the most common rockburst type;the other two are pillar burst and fault-slip burst,referred to as rockbursts occurring in pillar and in the vicinity of certain geological structures(such as dykes,joints and faults),respectively(Ortlepp and Stacey,1994).As excavation and mining proceed to greater depths,rockbursts may occur under different geologic conditions(Zhou et al.,2012),which threatens both human life and construction equipment(Zhang et al.,2013).A good understanding of rockburst mechanisms and characteristics with respect to different rock lithologies is therefore critically important for safe underground construction and operation.

Over the years,various efforts have been made to address the rockburst problems.Based on a wide range of field observations and experimental/theoretical investigations,various theories and indices have been proposed to evaluate the rockburst proneness and severity(Cook,1965a;Kidybi'nski,1981;Vardoulakis,1984;Linkov,1996;Heal,2010;Qian and Zhou,2011;Li et al.,2017).Numerical modeling and in situ monitoring have also been used to study rockburst evolution and identify potential hazardous areas(Zubelewicz and Mróz,1983;Srinivasan et al.,1999;Jiang et al.,2010;Tang et al.,2010;Feng et al.,2015;Manouchehrian and Cai,2017).Additionally,field control and support strategies for mitigation of rockburst hazards have also been reported(Hedley,1992;Kaiser et al.,1996;Cai,2013;Li,2017).However,complex geological and geometric conditions as well as uncertainty or variability of triggering conditions make the rockburst mechanism rather complicated,which has not been fully understood yet.

Underground rocks are generally under triaxial stress conditions prior to excavation.Excavation leads to redistribution of in situ stress,i.e.tangential stress concentration and radial stress release.Many laboratory tests,including conventional uniaxial,biaxial,triaxial and true triaxial compression(or unloading),have been conducted to investigate the failure characteristics of rocks and to reveal the rockburst mechanism near the excavation boundary(Cook,1965b;Lei et al.,2000;Hua and You,2001;Thompson et al.,2006;Yun et al.,2010;Huang and Li,2014;Jiang et al.,2015;Feng et al.,2016).Nevertheless,the triggering conditions of rockbursts have not been well physically simulated.The pioneering work on rockburst laboratory tests was performed by He et al.(2007,2010)using a modified true triaxial apparatus that could suddenly unload the stress imposed on one surface of the specimen.This can simulate in situ stress and boundary state of rockbursts.The rockburst process has thereafter been successfully reproduced under true triaxial unloading conditions(He et al.,2012;Zhao and Cai,2014;Zhao et al.,2014;Sun et al.,2017).For example,Zhao and Cai(2014)and Zhao et al.(2014)conducted a set of strainburst tests on Beishan granite with different unloading rates and another set of strainburst tests on Tianhu granite with different width-toheight ratios(of the rectangular prismatic specimens)under true triaxial unloading conditions,in order to understand the influences of unloading rate and specimen width-to-height ratio on strainbursts. Furthermore, another improved true triaxial rockburst testing apparatus was developed by Su et al.(2017c,d).Rockbursts induced by continuously increased tangential stress in extremely highly stressed situations were successfully simulated using this apparatus,and the influences of tunnel axial stress,radial stress gradient,temperature,loading rate and dynamic disturbance were investigated(Su et al.,2017a,b,c,d,2018).However,to the best of our knowledge, there have been few systematic experimental studies that can investigate the rockburst failure with respect to different rock lithologies.

It has long been recognized that rock lithology is a control factor of rockbursts,because the mechanical behaviors of rocks or rock masses primarily depend on their basic composition and structure(Jaeger et al.,2007).Many experimental studies focused on the mechanical behaviors of different rock types. For example,Wawersik and Fairhurst(1970)conducted a series of controlled uniaxial compression tests on a range of rock types and obtained the complete stress-strain curves,which revealed two distinct classes of post-failure behavior(class I,stable;class II,unstable or self-sustaining).Moreover, Wawersik and Fairhurst (1970)and Wawersik and Brace (1971) also investigated the post-failure behavior of different rocks(Tennessee marble II,Westerly granite and Frederick diabase)under triaxial compression with different confining pressures.Their work indicated that Tennessee marble can present an obvious brittle-ductile transition with increasing confining pressure,whereas Westerly granite and Frederick diabase remain brittle,which is supported by Martin(1997),Haimson and Chang(2000),and Feng et al.(2016).Lockner et al.(1992)investigated the temporo-spatial distributions of impulsive microcracking events through acoustic emissions(AEs)during fault formation of granite and sandstone.Similarly,they found that in contrast to sandstone,microcrack growth was distributed widely prior to fault nucleation in granite and that fracture localization occurred during the late stages of loading.Haimson and Song(1993),Lee and Haimson(1993)and Haimson(2007)revealed different breakout failure(micro-scale)mechanisms around the borehole drilled in various lithologic cubical specimens(e.g.granites,limestones and sandstones)subjected to three-dimensional(3D)far-field stresses using optical microscope and scanning electron microscope(SEM).However,experimental studies on different rock types have primarily focused on the gentle,stable failure or on the dynamic impact failure(Zhang et al.,1999,2000;Zhang and Zhao,2014),rather than on violent rockburst failure.

In this study,experimental investigations into rockburst behaviors are conducted on six types of intact hard brittle rocks under true triaxial loading conditions.The rockburst herein primarily refers to as strainburst(Ortlepp and Stacey,1994;Kaiser and Cai,2012).With the use of high-speed cameras and an AE system,the rockburst characteristics,including ejection failure process,fracture pattern,fragments and kinetic energy,and fracturing evolution are studied.Finally,rockburst failures are discussed in terms of fracturing and fragmentation characteristics and failure modes,as well as evaluation of rockburst proneness and severity.

2. Rock samples

Rectangular prismatic specimens, with dimensions of 100 mm×100 mm×200 mm(length×width×height),were cut from the same blocks of six types of hard brittle rocks:granodiorite from a quarry in Guangdong Province,China;granite from a quarry in Guangxi Zhuang Autonomous Region,China;basalt from the underground powerhouse of Baihetan hydropower station,China;marble from the headrace tunnel of Jinping II hydropower station,China;sandstone from a quarry in Shandong Province,China;and limestone from a quarry in Iran,as shown in Fig.1.The physicomechanical properties of the rocks tested are presented in Table 1.

Fig.1.Rectangular prismatic specimens of the rocks tested.

Fig.2 shows the naked-eye observations,the 3D hyper-focal distance microscopic images and the optical cross-polarized micrographs of the rocks tested.The mineralogical compositions and grain sizes evaluated from the optical cross-polarized micrographs are listed in Table 2.The following can be observed:

(1)In Fig.2a,the rock used is medium-grained biotite granodiorite,which displays a granitic texture and a massive structure.The predominant mineral constituents are plagioclase,quartz,K-feldspar,and a small percentage of biotite and hornblende.These mineral particles are within a general size range of 0.6-5 mm,unevenly embedded and distributed in the rock.

(2)In Fig.2b,the rock used is coarse-grained biotite granite,which has a granitic texture and a massive structure and is primarily composed of K-feldspar,quartz,and a small percentage of biotite and plagioclase.The mineral grains are generally within a size range of 2-22 mm that are unevenly embedded and distributed in the rock.

(3)In Fig.2c,the prepared rock is fine-grained weak alteration basalt with a blastodiabasic and blastopoecilitic texture and a massive structure.The major minerals are plagioclase,pyroxene,chlorite,actinolite,and a small amount of epidote and glass,with sizes generally ranging from 0.1 mm to 1 mm.The pyroxene and plagioclase are unevenly embedded and distributed,with the other minerals non-uniformly distributed in the rock.

(4)In Fig.2d,the prepared rock is micro-fine-grained calcitedolomite marble with a granoblastic texture and a massive structure.The primary mineral constituents are calcite,dolomite and a small percentage of kaolinite and sericite,with sizes ranging between 0.02 mm and 0.16 mm.The calcite is unevenly embedded and distributed in the dolomite.

(5)In Fig.2e,the rock tested is fine-medium-grained inequigranular dirty arkose sandstone with an arenaceous texture.The predominant minerals are feldspar,quartz and a small percentage of sericite,kaolinite and limonite,which are within a size range of 0.06-0.25 mm,followed by 0.25-0.5 mm and 0.5-1.2 mm.These minerals are mixed in coarse and fine sizes within the rock.

(6)In Fig.2f,the rock tested is microlite bioclastic limestone,which has a bioclastic texture and a massive structure,and is primarily composed of calcite and a small percentage of sericite,kaolinite and quartz.The size of the particles varies between 1 mm and 12 mm.The particles of different sizes are unevenly distributed in the rock.

3. Experimental set-up and testing procedure

3.1. True triaxial rockburst testing system

A true triaxial rockburst testing system developed by Su et al.(2017c,d)was used in this study.This testing system is primarilycomprised of a true triaxial rockburst testing machine,an AE system,and a high-speed camera system(Fig.3).An improved true triaxial electrohydraulic servo-controlled compression apparatus is specially equipped with two independent loading systems in one horizontal direction(y-direction).It can therefore rapidly unload and expose only one loading face(perpendicular to the y-direction)of the rectangular prismatic specimen under triaxial compression.It can also load five faces of the specimen continuously while keeping one face free with the help of friction or shear stresses between the rigid plate and the specimen.It is crucial that an apparent stress gradient,which is analogous to the radial stress gradient near the boundary of underground excavations,can be simulated on the specimen in y-direction during the aforementioned actions(Su et al.,2017d).Thus,using this testing machine,the triggering conditions of rockbursts in underground excavations can be well reproduced on the rectangular prismatic rock specimen,i.e.a representative rock element.This rockburst testing machine has relatively large loading capacities,i.e.5000 kN and 3000 kN in vertical direction and two horizontal directions,respectively,and it can thus accommodate relatively large representative rock elements(e.g.100 mm×100 mm×200 mm or larger).

Table 1 Basic physico-mechanical properties of the rocks tested.

3.2. Experimental procedure

In the field,most rockbursts occurred in a certain time period after excavation,usually several hours or a few days,rather than immediately despite the fact that some of them may be affected by the remote seismic disturbance(Feng, 2017). Thus,rockbursts induced by the gradually increased tangential stress,i.e.loading rockbursts,rather than by the sudden unloading of the radial stress immediately after excavation,were investigated.

Accordingly,the loading path illustrated in Fig.4 was used in the rockburst tests,which simplified the initial loading/unloading actions of the radial stress for higher efficiency.For more details,readers are referred to Su et al.(2017c,d).The loading steps were described as follows:

(1)Load σxand σzindependently at a low rate,until σxreaches the target value.

(2)Load σy2soon after the beginning of step(1)at a low rate to the target value while keeping the opposite specimen surface free(i.e.σy1=0 MPa).

(3)Maintain σxand σy2at the target values and continually load σzat a relatively fast rate until the rock specimen fails.

Rockbursts under this loading path are also known as“one-facefree true triaxial loading rockbursts”.In the studies on granite(Su et al.,2017c,d),rockbursts were relatively slight or not triggered under an extremely low axial stress σxor radial stress gradient that corresponds to σy2,whereas an extremely high σxor σy2is unusually concentrated around underground openings. Furthermore, an extremely high σy2may cause the specimen to slip along y-direction during loading,threatening the safety of the testing apparatus.Thus,in this study,σxand σy2were specified as 30 MPa and 5 MPa,respectively.σz(and σx)was initially loaded at 0.5 MPa/s and then at 1 MPa/s,a slightly higher rate to increase the likelihood of reproducing the rockburst process on different rocks(Su et al.,2018).σy2was loaded at 0.25 MPa/s.Following this loading scheme,rockburst tests on different rock specimens with relatively large dimensions(100 mm×100 mm×200 mm)(Fig.1),as the representative rock elements around the underground openings,were conducted.

Fig.2.Naked-eye observations(top-left),3D hyper-focal distance microscopic images(top-right)and the optical cross-polarized micrographs(bottom)of(a)granodiorite,(b)granite,(c)basalt,(d)marble,(e)sandstone,and(f)limestone(Pl-plagioclase,Qtz-quartz,Kfs-K-feldspar,Bt-biotite,Chl-chlorite,Ep-epidote,Aug-augite,Act-actinolite,Doldolomite,Cal-calcite,Fsp-feldspar,Bas-basalt debris,Tur-tourmaline,Lm-limonite,Biocl-bioclast).

During the tests,the ejection process upon failure and the AE activity were captured and monitored in real time by the high-speed cameras and AE sensors,respectively,as illustrated in Fig.5.A multichannel AE measurement system,SENOR HIGIHWAY II from Physical Acoustics Corporation and three(or more in a few tests for AE location)Nano30 sensors,with operating frequency range of 125-750 kHz were used.The trigger threshold was set to 40 dB,and the sampling rate was set to 1 million samples per second(MSPS).

4. Results

The results of the rockburst tests on different rock types are summarized in Table 3. One can see that violent rockbursts occurred as σzincreased except for the limestone(LS)specimens.The rockbursts of granodiorite(GD),granite(GR)and marble(MA)specimens were relatively strong with a large amount of fragment ejection and kinetic energy release,accompanied by a series of splitting slabs near the free face and notable failure precursors.In contrast,the rockbursts of basalt(BA)and sandstone(SS)specimens were relatively slight with little fragment ejection and kinetic energy release,which exhibited no or few splitting slabs near the free face.

4.1. Ejection failure process

Based on the ejection failure process of different rocks(or stable failure of limestone)recorded by the high-speed cameras,it was found that the ejection failure processes of the granodiorite,granite and marble specimens were similar,but those of the basalt and sandstone specimens differed significantly.Photographs during the failure processes of typical specimens of different rocks are presented in Fig.6(i.e.GD2,GR2 and MA1 in Fig.6a;BA3 and SS1 in Fig.6b;and LS2 in Fig.6c).

Table 2 Mineralogical composition and grain size distribution of the rocks tested.

Fig.3.True triaxial rockburst testing system:(a)True triaxial rockburst testing machine,(b)AE system,(c)High-speed cameras,(d)Zoomed-in photograph,and(e)Illustration of stress states of the rock specimen-rigid plate assembly and the interior rock,in which σx,σy2,σz and τ denote the axial stress σa,radial stress σr,tangential stress σθ and shear stress σrθ(τθr)of a rock mass element in the vicinity of the excavation boundary,respectively.

Fig.4.Schematic illustration of(a)stress states of a rock mess element in the vicinity of the excavation boundary,and(b)loading path used in the tests.

In Fig.6a,small particles were first ejected out from the free face with slight cracking sounds at approximately 3.141 s,21.343 s and 0.821 s before the rockbursts of GD2,GR2 and MA1,respectively.Afterwards,approximately 3.023 s,11.343 s and 0.535 s later,tensile splitting occurred in GD2,GR2 and MA1,respectively,which split the rocks into slabs.Subsequently,the splitting slabs expanded and bent outward.Approximately 50 ms,3589 ms and 160 ms later in GD2,GR2 and MA1,respectively,the slabs bucked and broke off,accompanied by a large amount of violent ejection and extremely loud sounds.Despite some differences between different rock types and even between different specimens of the same lithology,the ejection failure of these three specimens or corresponding rock types generally underwent four stages:grain ejection,splitting of rocks into slabs,bending of rock slabs and fragment ejection at the moment when the rock slabs broke off(Su et al.,2017c,d).In particular,the evolution of the rockburst process of granite(e.g.GR2)lasted much longer,and the corresponding rockburst ejection was thus relatively gentle,involving a smaller failure scale compared with those of granodiorite(e.g.GD2)and marble(e.g.MA1).

In Fig.6b,the specimens showed no significant macroscopic fracture on the free face until small particle ejection occurred shortly,i.e.at approximately 156 ms and 352 ms for specimens BA3 and SS1,respectively,before the rockbursts.Subsequently,approximately 146 ms and 322 ms later for BA3 and SS1,respectively, local rupture occurred on the free face, and immediately,approximately 10 ms and 30 ms later,the local broken rocks and a few fragments behind them were ejected with loud sounds.Different from GD2,GR2 and MA1,the ejection failure processes of these two specimens or corresponding rock types were seldom associated with the splitting of rocks into slabs and buckling of splitting slabs.The corresponding nucleation and evolution process were more sudden and no clear precursors were captured.Compared with basalt(e.g.BA3),the local ejection failure of sandstone (e.g. SS1) occurred more gradually and gently.

With respect to specimen LS2 shown in Fig.6c,gentle failure rather than violent rockburst failure occurred.

4.2. Fracture pattern

The observation of the ejection failure processes reveals the failure characteristics near the free face.However,the rockburst failure is also closely related to the fracture/rupture occurring away from the free face.Fig.7 shows the failed specimens after the rockburst tests in view of the fracture patterns throughout the tested specimens.Many experimental investigations revealed that in the majority of rocks,brittle fracture,splitting,shear rupture or faulting occurred in planes oriented parallel to the σ2direction(striking in the σ1-σ2plane,where σ1and σ2indicate the maximum and intermediate principal stresses,respectively)under the true triaxial loading(Haimson and Chang,2000;Nasseri et al.,2014;Feng et al.,2016).This is particularly the case for our one-facefree true triaxial rockburst tests,i.e.most of the fracture planes were oriented parallel to the σxdirection,and thus,in Fig.7,the photographs of the surfaces on which σxwas applied were taken.

It is observed that for granodiorite,graniteand marble specimens(Fig.7a-c),several macrocracks with smooth surfaces covered with a small amount of powder or grains occurred near the free face subparallel to the σzdirection.This indicates the fracture/rupture mechanism of tensile splitting,whereas macrocracks with a coarse surface mixed with white powder or small grains that dipped in the σydirection occurred away from the free face,suggesting fracture/rupture mechanism of shear rupture or faulting(Bobet and Einstein,1998;Haimson,2007).The rockburst pit with a relatively large volume occurred on the splitting slabs near the free face,resulting from fracturing and fragmentation of these slabs.The stair-step shape of the rockburst pit(e.g.GR3,MA1 and MA2 in Fig.7)can also reveal the tensile splitting mechanism near the free face.Compared with granodiorite and marble specimens,the splitting slabs of granite specimens were more irregular and crushed,corresponding to the relatively gradual and gentle rockbursts.

Table 3 Summary of key experimental results of the rockburst tests on different rocks.

Fig.6.Ejection failure processes of different rocks:(a)Granodiorite(GD2),granite(GR2)and marble(MA1);(b)Basalt(BA3)and sandstone(SS1);and(c)limestone(LS2).The numbers in the top left corners of the images indicate the frame number and the time in h:min:s.

Fig.7.Failed specimens after the rockburst tests on different rocks:(a)Granodiorite,(b)Granite,(c)Marble,(d)Basalt,(e)Sandstone,and(f)Limestone.The typical specimens of different rocks are specially presented.

Fig.8.Mass of spalling and ejection fragments of different rocks.

The fracture patterns of basalt,sandstone and limestone specimens(Fig.7d-f)were similar,but differed significantly from those of the granodiorite,granite and marble specimens.Macrocracks,characterized by an uneven surface and covered by white powder and dipped in σydirection,occurred throughout the specimen.This result indicates that the primary rupture mechanism/fracture type for these three rocks was of shear or tensile-shear rupture or faulting without splitting of rocks into slabs(Bobet and Einstein,1998;Haimson,2007).The rockburst pit(only for basalt and sandstone specimens)occurred mostly on the top close to the free face,i.e.in the top left or top right corner of the photographs,which resulted from the local compression-shear or tensile-shear rupture.The burst zone was relatively small and regular except for specimen SS2,which,unfortunately,was damaged when removed from the loading plates.In particular,the fracture of the limestone specimens was more concentrated and localized away from the free face and exhibited clear ductile characteristic.A relatively broad but shallow burst zone occurred on the basalt specimens,such as BA2 and BA3.In general,the fracture patterns of the tested specimens confirmed and further revealed the ejection failure processes and failure mechanism of rockbursts on different rocks.

4.3. Fragments and kinetic energy

In our tests,the rock fragments of spalling and ejection were collected and grouped using a standard sand sieve with square holes of 9.5 mm,4.75 mm,2.36 mm,1.18 mm,0.6 mm,0.3 mm,0.15 mm and 0.075 mm in side length and then weighed.For fragments with a maximum length greater than 9.5 mm,the maximum length,width and thickness as well as the weight were measured.

Fig.8 shows the mass of the spalling and ejection fragments,which was used to quantify the volume of the rockburst pits divided by the density of different rocks,as listed in Table 3.It can be seen that the fragment mass(or volume)of marble and granodiorite was relatively large,followed by that of granite and basalt.The fragment mass of sandstone and limestone that only generated spalling failure was relatively small.It is also found that the fragment mass as well as the proportion of spalling relative to ejection varied with different specimens of the same lithology,particularly for the granite followed by the granodiorite and marble.

To examine the degree of fragmentation induced by different lithologic rockbursts,the cumulative mass percentage-grain size(the maximum length of fragmentations) distributions of the fragments were investigated,as plotted in Fig.9.The figure shows that the mass distributions of specimens GD2,GR2 and MA1 are similar and more uniform,compared with those of specimens BA3,SS1 and LS2.The cumulative mass percentages of the fragments with a maximum length less than 9.5 mm and 4.75 mm are presented in Fig.9b and c,respectively.As shown in Fig.9b and c,the proportions of fine or medium fragments of specimens GD2,GR2 and MA1 are generally larger than those of specimens BA3,SS1 and LS2,despite the fact that the grain sizes of granodiorite and granite are much larger,which,coupled with the results shown in Fig.9a,indicates that a higher degree of fragmentation had occurred during the rockburst.

The kinetic energy of ejected fragments is estimated quantitatively,as presented in Fig.10(also in Table 3).For more details,please see the works of Su et al.(2017c,d).Fig.10 shows that the ejection kinetic energy of the granodiorite and marble was relatively large,which indicates strong rockbursts generated,followed by that of basalt and granite.The ejection kinetic energy of sandstone was relatively small,and no burst ejection occurred on limestone.The ejection kinetic energy essentially remained the same as that of the fragment mass.Similarly,the ejection kinetic energy of the granodiorite,granite and marble showed significant randomness and inconstancy.Further,it was observed that a relatively high ejection kinetic energy is more prone to occurring on a specimen with a relatively high strength despite some variability.On the other hand,it should also be noted that the kinetic energy differed significantly in lithologies which exhibited comparable strengths,such as granite,marble and basalt.

Fig.9.Plots of(a)cumulative mass percentage-fragment size distributions and cumulative mass percentage of fragments with a maximum length less than(b)9.5 mm and(c)4.75 mm of different rocks(only the typical specimens GD2,GR2,MA1,BA3,SS1 and LS2 are shown for clarity).

Fig.10.Kinetic energy of ejected fragments versus the peak stress of σz of different rocks.

4.4. AE evolution

AE techniques provide a tool to evaluate the fracturing evolution inside the rock volume(Scholz,1968;Lockner,1993;Lei et al.,2000;Thompson et al.,2006).Fig.11 shows the temporal changes of the AE hits,cumulative AE hits and cumulative AE energy detected by one of the AE sensors.Since the AE system malfunctioned during testing specimen MA1,the AE parameters and stress data of specimen MA3,on which similar failure occurred,are shown in Fig.11c.In Fig.11,AE records during the initial stage of loading(i.e.crack closure process)have been removed to avoid confusion in data analysis(Eberhardt et al.,1999).

Fig.11.Temporal changes of the AE hits,cumulative AE hits and cumulative AE energy together with the stress data of different rocks:(a)Granodiorite(GD2),(b)Granite(GR2),(c)Marble(MA3),(d)Basalt(BA3),(e)Sandstone(SS1),and(f)Limestone(LS2).

As shown in Fig.11,the AE temporal evolutions during the rockburst tests of different rocks basically undergo similar patterns.With increasing σzafter the initial stress state,a gradual increase of AE occurred for crack initiation and propagation,or low AE activity continued for insignificant crack development.As σzfurther increased shortly before failure,a rapid increase of AE or a continuously high AE activity could be observed due to crack coalescence.However,because several AE hits may be registered as one single hit instead of being detected individually when the rate of cracking is extremely high(Moradian et al.,2016),AE hits might show a drop.Checking the detected AE hits during the period when the AE hits exhibited a clear drop shortly before burst failure,it is found that the limited number of AE hits was cut based on the maximum duration(1 s)rather than on the hit definition time(HDT).This result indicates that severe overlapping might have occurred before burst failure,mostly attributed to violent macrocracking and friction between the crack surfaces.The corresponding burst increase of AE energy could also confirm this case.

The rise anglevalue(RA=therise time/the maximum amplitude)and the average frequency value(AF=AE ring-down count/the duration)are plotted versus time for specimens GR2,BA3 and SS1 in Fig.12 to understand the fracture mechanisms during rockburst processes.For more details about RA and AF,readers are referred to Grosse and Ohtsu(2008).In Fig.12,300,100 and 50 AE hits were chosen as the intervals for specimens GR2,BA3 and SS1,respectively(in terms of the total number of AE hits throughout the test),to carry out moving average calculation of the RA and AF values(Eberhardt et al.,1998;Ohno and Ohtsu,2010).One can see from Fig.12 that AF becomes lower and a higher RA value is obtained in the later or final stages,which reveals that during rockbursts,the fracture mode normally changes from tensile mode to shear(or mixed)mode(Grosse and Ohtsu,2008;Ohno and Ohtsu,2010).Additionally,partial waveforms(length of 8 K,lasting 8.192 ms with sampling rate of 1 MSPS)of example signals detected in the early loading stage and in the later/final stage,which correspond to a burst emission and a continuous emission,respectively,are illustrated in Fig.12.It is found that compared with the burst emission,the continuous emission has much higher amplitudes,longer duration and larger energy magnitude,resulting from consecutive macro-scale fracturing(or microcrack coalescence).

Fig.12.Plots of RA and AF versus time of granite(GR2),basalt(BA3)and sandstone(SS1).Partial waveforms(length of 8 K,lasting 8.192 ms)of example signals of burst emission(top left)and continuous emission(top right)detected in the tests are illustrated.

Except for these similar patterns in Figs.11 and 12,however,it is observed that the lithology has a significant influence on the AE evolution.For example,specimens GD2 and GR2(or other granodiorite and granite specimens)showed an active AE state almost throughout the loading procedure,characterized by the clearly increased AE hits and gradually cumulated AE hits and energy from the early loading stage(Fig.11a and b).Before the stress(σz)reached its yield point where the AE increased dramatically,a large number of AE hits appeared on specimens GD2 and GR2,accounting for over 60% of the total hits.However,the corresponding cumulative energy was much less(approximately 1-2 orders of magnitude)than the final cumulative energy,because these AE hits were mostly related to small(micro-scale)tensile fractures with relatively lower magnitudes.Similar to specimens GD2 and GR2,specimen MA3 (or other marble specimens) also underwent obvious cracking and exhibited notable AE activity before the peak stress was reached(Fig.11c).The difference is that a clear increase in AE has not occurred until σzincreases to approximately 158 MPa which is greater than 65% of the peak stress,and subsequently,the AE hits continually fluctuate strongly up to the occurrence of burst failure.This fluctuation may result from the alternative occurrence of growth and coalescence of microcracks and macrocracks.Contrary to specimens GD2,and GR2(or similar MA3),specimens BA3,SS1 and LS2(or other basalt sandstone and limestone specimens)did not exhibit distinct AE activity,except for individual bursts until shortly or immediately before the final burst failure(Fig.11d-f).Both the AE hits and cumulative AE hits were essentially constant or slightly increased.Therefore,it can be concluded that small local fractures were almost entirely suppressed.Another point in Fig.11 is that the cumulative graphs of the AE energy for specimens MA1,BA3 and SS1 all showed clear jumps before the stress(σz)increased to enough high levels.However,the jumps were not completely related to the sudden increase in the AE hits.A possible explanation for this is that because small local fractures were significantly or almost entirely suppressed in all the three specimens,the cracking energy could be preserved in the rocks until certain thresholds or conditions were met and then could be released rapidly.

It is likely that with increasing stress(σz),a large number of local tensile fractures would further develop and coalesce and eventually generate splitting slabs sub-parallel to σzdirection near the free face(Fig.7).As above mentioned,these splitting slabs have greatly influenced the rockburst mode on the tested specimens.On the other hand,a large number of small tensile fractures could provide a possibility for the occurrence of intense growth and coalescence of macrocracks before burst failure.The obvious drop in the AE hits,lasting about 50 s,and the burst increase of AE energy could be confirmed in Fig.11a and b.Consequently,the total AE hits and energy of specimens GD2 and GR2 were several times or even tens of times larger than those of specimens BA3,SS1 and LS2,which means that a higher degree of fracturing and fragmentation has occurred in these two specimens.Comparing the rockburst characteristics of different rocks,the importance of the small tensile fractures could be highlighted.

5. Discussion

5.1. Fracturing and fragmentation during rockburst

During the evolution of rockburst,the rock mass is fractured and fragmented,and on the other hand,excessive energy is rapidly transformed into the kinetic energy of fragments due to the rapid development of cracks(Kaiser et al.,1996;Cai,2013).These two processes are all essential for the occurrence of the final ejection failure of rockburst.For example,even though there is a large amount of excessive energy,the rock mass that is split into large pieces will not present strong ejection failure but tend to show obvious shake.In other words,the fragmentation characteristic of the rock mass can significantly affect the energy redistribution and transformation during rockburst development.

The experimental results reveal that the fracturing and fragmentation process and final degree of fragmentation generated in different rocks differed distinctly, roughly following two modes.

For the granodiorite and granite,long before the peak stress was reached,stress-induced initiation,propagation and coalescence of microcracks were generated largely and randomly distributed in the specimen but more concentrated near the free face.As the stress approached and exceeded the peak stress,a large number of macrocracks(fracture and rupture)associated with the coalescence of microcracks could thus develop, fragmenting the specimen severely,particularly near the free face.The resulting fragments or fine particles were easily ejected out from the free face of the specimen once excessive energy existed.Similarly,the above fragmentation process also occurred on the marble;however,a significant creation and accumulation of microcracks is initiated at a higher stress level,as shown in Fig.11.

For basalt and sandstone(or limestone),local microscopic or mesoscopic fracturing would not occur until the load was loaded up close to the peak stress.Shortly before the peak stress,microcracks were generated locally in the specimen,and then shear or tensileshear rupture nucleated.As the load approached the peak stress,cracks in front of the rupture tip were consecutively generated,rocks between these cracks were crushed or rotated for dominoblocks(Peng and Johnson,1972;Lei et al.,2000;Tarasov,2014),and shear rupture propagated,intersected or coalesced,which fragmented the specimen into pieces.Clearly,the resulting pieces were relatively large and not easily ejected during rockburst failure compared with the fragments or fine particles generated in the granodiorite and granite specimens,although these pieces might be further fragmented by a shock wave generated due to the‘instant’rupture(Tarasov and Stacey,2017).In summary,for basalt,sandstone and limestone, development of localized shear or tensile-shear rupture is neither attributed to nor accompanied by distributed microcracking, and accordingly, rocks outside the rupture maintains a good integrity.

The aforementioned two fracturing and fragmentation modes were initially reported by Wawersik and Fairhurst(1970).Subsequently,many experimental and theoretical studies investigated the micro-or meso-scale fracturing behaviors,which confirmed these fracturing and fragmentation phenomena and mechanisms(Fonseka et al.,1985;Horii and Nemat-Nasser,1985;Lockner et al.,1992;Kawakata et al.,1999;Lei et al.,2000;Thompson et al.,2006;Moradian et al.,2016).Tarasov and Stacey(2017)discussed features of the energy balance at failure and fragmentation of classes I and II rocks,and they proposed two different fragmentation mechanisms,i.e.‘compressive-force fragmentation’operating on the majority of rocks(including classes I and II)and‘shock-wave fragmentation’operating on the extreme class II rocks.They proposed that for the extreme class II rocks,catastrophic rupture propagation and sudden elastic energy release can excite shock waves,and these waves can cause a significantly high degree of fragmentation of rocks that still remain relatively intact.Although extremely brittle class II post-peak failure might not occur in the above rockburst tests,we consider that these two fragmentation mechanisms hold. The‘compressive-force fragmentation’was the dominant mechanism,especially for the granodiorite,granite or marble.It is noted that in Fig.7,some basalt specimens exhibited a clear‘domino-block’structure,which is known as a preferred failure mechanism of extremely violent ruptures(Tarasov,2014).The resulting shock waves might be one reason for the relatively broad but shallow burst zone that occurred on the basalt specimens,as mentioned in Section 4.2.Compared with basalt,the sandstone was softer and less brittle,and hence the efficiency of the‘shock-wave fragmentation’mechanism would decrease significantly.

Fig.13.SEM images of fragment surfaces of different rocks:(a)Granodiorite,(b)Granite,(c)Marble,(d)Basalt,(e)Sandstone,and(f)Limestone.

For different fracturing and fragmentation characteristics of the rocks, the inherent microstructure and corresponding microfracturing were investigated and considered as the primary factor(Wawersik and Fairhurst,1970;Tapponnier and Brace,1976;Kranz,1983;Fonseka et al.,1985;Kawakata et al.,1999;Rodríguez et al.,2016;Tarasov and Stacey,2017).Fig.13 shows the SEM images of fragment surfaces of different rocks.It can be observed that the medium-or coarse-grained polymineralic granodiorite and granite exhibit strong heterogeneity,which can cause significant stress concentration and microfracturing initiation at a relatively low stress level.On the other hand,mismatch of stiffness and other physical properties between different minerals can result in additional boundary tensile stress and cracking.For example,it was demonstrated by Kranz(1983)and Rodríguez et al.(2016)that many stress-induced microcracks are initiated in quartz crystals,which are stiffer in comparison with plagioclase,K-feldspar or microcline crystals.Relatively longer grain boundaries can also lead to(intergranular)microfracturing,while transgranular microcracks including cleavage are generally the major cracking form for most igneous rocks that exhibit a crystalline interlocking texture.Consequently,various types of microcracks can be generated on granodiorite and granite specimens.With regard to marble,basalt and sandstone,a significant decrease in the grain size of the minerals coupled with an increased textural homogeneity of the materials(e.g.marble and basalt)can change the internal stress environment and suppress the occurrence of significant local fracturing.Wawersik and Fairhurst(1970)proposed that the stress to induce local axial fracturing may exceed the mean resistance to faulting in very fine-grained rocks.Chang and Haimson(2005)demonstrated that no or few microcracks develop before the occurrence of through-going shear rupture on two fine-grained rocks,while the grain size in and by itself is not the major cause.Nevertheless,these three rocks exhibited different microfracturing features as they belong to metamorphic,igneous and sedimentary rocks,respectively,and have different microstructures.In finegrained marble, there were predominant intergranular microcracks with a degree of cleavage(transgranular microcracks)that occurred under a relatively high stress level.In basalt,the grains display a crystalline interlocking texture and are cemented together so firmly that when the rock fractures,the fractures are prone to passing through the grains rather than around them.In sandstone,the toughness of the grain boundary or the cement is generally weaker than that of the grain,and hence the majority of microcracks tend to be initiated along (cleaving or crushing) grain boundaries during the final loading stage.

Fig.14.Case examples of(a)tensile slabbing spalling or rockburst and(b)shear rockburst(Hou et al.,2011;Feng,2017).

Fragmentation is necessary for occurrence of ejection failure of rockburst;however,it should also be noted that significant preburst fracturing and fragmentation will decrease the mechanical properties(e.g.elastic modulus)of the rocks and associated energy accumulation(Wawersik and Fairhurst,1970;Tarasov and Stacey,2017).This may be one of the reasons for the relatively slight rockbursts that occurred on granite specimens.On the other hand,the decrease of the pre-burst fracturing and fragmentation intensity,indicating low energy dissipation,will facilitate more energy accumulated and thus can drive more violent or selfsustaining(class II)unstable failure.In the case of rockburst,if this decrease does not significantly influence the generation of fragments to be ejected potentially,it will intensify the rockburst failure.This may provide an explanation of the relatively large kinetic energy of rockbursts on marble specimens.

5.2. Failure modes of rockburst

Rockburst is the structurally unstable failure of rocks around underground openings. As mentioned previously, rockburst occurring on intact hard brittle rock specimens in this study primarily belongs to strainburst,the most common rockburst type.Slabbing/buckling burst characterized by splitting of rocks into slabs and buckling and ejection of slabs(slabbing structure)has been considered to be one of the most important modes of strainburst;however,strainburst should not be limited to this.Indeed,splitting/slabbing failures have been frequently reported in underground openings around the world(Fairhurst and Cook,1966;Martin,1997;Germanovich and Dyskin,2000;Diederichs,2007;Cai,2008;Jiang et al.,2017;Gong et al.,2018).Also,slabbing,buckling and breaking off of rocks at or near the excavation boundary can naturally satisfy the fragmentation demand for ejection failure.However,the excessive energy from this process must be limited because the buckling strength of rock slabs is much lower than the UCS of rocks,and tunnel closure can progressively occur during the separate failure of slabs.Thus,the slabbing/buckling burst is generally less violent,although it is more violent in comparison with the splitting/slabbing failure that has also been considered as a slight strainburst in several studies(e.g.Diederichs,2007;He et al.,2007;Du et al.,2016;Jiang et al.,2017).

As demonstrated in Section 4.2,for granodiorite,granite and marble specimens,the dominant fracture mechanism was tensile splitting near the free face,whereas shear rupture or faulting occurred farther from the free face(Fig.7).The strainburst on these specimens can thus be called slabbing/buckling-shear rupture burst.During the evolution of this burst,the rocks at or near the excavation boundary and those further away from the excavation boundary behave diversely as two different structural components.On one hand,the fragments can be generated mainly at or near the excavation boundary by slabbing,buckling and breaking off of rocks.On the other hand,the shear rupture of the rocks away from the excavation boundary can result in more excessive energy from the instantaneous compressive recovery of the surrounding rocks and the rapid tunnel closure(Kaiser and Cai,2012).Thus,the slabbing/buckling-shear rupture burst is generally relatively stronger.It should be noted that most of the excessive energy does not result from the fractured and fragmented rocks but from the relatively intact surrounding rocks,although some of this energy is ultimately dissipated and released(as kinetic energy)by the fractured and fragmented rocks.This is consistent with the statement“the violent fractures and rockbursts result from sudden increases in the effective size of an excavation caused by discontinuous enlargements of the fractured and failed region,as the size and geometry of the excavations change”(Cook,1965a).

In Fig.7,the shear rupture on granodiorite and granite specimens is moderate,unlike that on marble specimens.Su et al.(2017d)revealed that the failure mode of strainburst can be influenced by the radial stress gradient and a relatively high radial stress gradient will likely yield strong shear strainbursts around the excavation boundary.Thus,a possible explanation for this phenomenon is that a higher radial stress or stress gradient is required for granodiorite and granite to suppress local tensile fractures(see Section 5.1).

Fig.15.Schematic illustration of three strainburst modes:(a)Slabbing/buckling burst,(b)Slabbing/buckling-shear rupture burst,and(c)Shear rupture burst.

Shear rupture burst,which is characterized by shear rupture without obvious tensile splitting/slabbing near the free face,uniquely occurred on basalt and sandstone specimens in this study.It is clear that the fracturing and fragmentation mode of the two rocks is the key to the occurrence of this type of strainburst.When rocks near the free face or the excavation boundary are divided into large blocks without strong fragmentation,ejection failure will be limited.Nevertheless,the catastrophic rupture propagation and sudden elastic energy release may source intense shock waves,which,in extremely violent case,can cause a significantly high degree of fragmentation of rocks and then intense ejection failure occurs.These shock waves can also cause a remote dynamic stress increase or remote rockbursts.In this sense,shear rupture burst can be considered as a seismic event(Hedley,1992;Ortlepp and Stacey,1994). In some studies (e.g. Stacey, 2016), rupture rockbursts referred to as the events in which the seismicity(shear rupture)is tens or hundreds of meters away from the rockburst location.This is not exactly the case as discussed herein.

Ortlepp and Stacey(1994)and Ortlepp(1997)classified rockbursts into five types:strainburst,buckling,face crush/pillar burst,shear rupture and fault-slip burst. Buckling rockburst can be grouped into strainburst,and shear rupture rockburst into faultslip rockburst,and rockbursts can then be generally classified into three types:strainburst,pillar burst and fault-slip burst(Hedley,1992;Cai,2013).Field investigations have shown that in rocks near or at the excavation boundary,tensile slabbing spalling or rockburst and shear rockburst frequently occur,as shown in Fig.14.Experimental results,in addition to those in Du et al.(2016)and Su et al.(2017d),revealed that around underground openings of relatively intact rocks,slabbing/buckling burst,slabbing/bucklingshear rupture burst or shear rupture burst can occur,which are considered as specific modes of strainbursts herein,as shown in Fig.15.This has a close relationship with rock lithology,radial stress gradient and other field conditions.As discussed above,strainburst should not be simply considered as splitting of rocks into slabbing,or buckling and ejection of slabs at or near the excavation boundary.

As the fracturing mode and mechanism of granodiorite or granite are typical for the majority of rocks,most of the surrounding rocks will tend to undergo slabbing/buckling burst or slabbing/buckling-shear rupture burst, exhibiting obvious precursors such as grain ejection,splitting/slabbing or buckling of slabs and significant AE activity,as presented in Section 4.Strainbursts or other similar types of bursts are generally predictable(Srinivasan et al.,1999;Kaiser and Cai,2012;Feng et al.,2015).Nevertheless,for some fine-grained homogeneous or severely interfingering crystallized rocks,the stress to induce local axial fracturing may exceed the mean resistance to faulting and thus they will tend to generate shear rupture burst,exhibiting few precursors,as presented in Section 4.Thus,the primary concerns regarding prediction and support measures of these two types of strainbursts should be different.

5.3. Evaluation of rockburst proneness

It is acknowledged that strong brittle and elastic rocks in highly stressed underground openings generally have a potential for rockbursts,as they can accumulate high strain energy and abruptly release it during failure.Many rock indices,such as strength,brittleness and energy indices,have been proposed to estimate the rockburst proneness and severity of different rock types(Kidybi'nski,1981; Singh,1987; Wang and Park, 2001), and a continuing effort has been made(Tarasov and Randolph,2011).

Fig.16.The z-direction stress-strain curves of different rocks(only the typical specimens GD2,GR2,MA1,BA3,SS1 and LS2 are shown for clarity).

Table 4 Evaluation of rockburst proneness of different rocks.

Fig.16 shows the z-direction stress-strain curves of different rocks.Since violent,compressive failures in rockbursts require a sustained application of stress in the form of‘following load’except for the sufficiently high stress to initiate failure(Hedley,1992),it is necessary to follow the loading path characterized by monotonic loading of σz(as shown by the red line in Fig.4b)in the rockburst tests,without employing other controllable regimes near the peak stress(Wawersik and Fairhurst,1970;Lockner et al.,1992).The conventional unbiased post-peak stress-strain curve that can characterize the mechanical behavior of tested rocks had not been obtained.Besides,the instantaneous changes of stress and deformation upon rockburst failure make the recording of post-peak curve difficult.In this context,the majority of brittleness and energy indices that depend on the stress-strain curves could not be used directly.In fact,we note that the energy consumed and released in the specimen in the post-peak stage during rockbursts would be supplied and imposed additionally by the elastic energy stored in the test machine,and may be the input energy from an external energy source,i.e.the hydraulic system.They respectively correspond to the energy stemming from the instantaneous compressive recovery and the rapid tunnel closure(the circumferential shrinkage deformation)of the surrounding rocks in the field.This implies that a‘soft’loading system that generally exists in the rockburst case was reproduced(Kaiser and Cai,2012;Cai,2013;Manouchehrian and Cai,2017).For the extremely brittle class II rock or class II rock failure,the energy could be totally provided by the elastic energy stored in the specimen,but this may be rarely the case for rockburst failure,which still needs further study.Hence,it is likely that rockburst is not only related to the lithology but also to the loading condition.Further,as discussed above,the fracturing and fragmentation characteristics and failure modes can significantly influence the rockburst proneness and severity.The conventional indices potentially cannot cover these factors.

The parameters of mass(or volume),speed and more comprehensive kinetic energy of ejected fragments have close relationship with rockburst proneness and severity,and they have been widely employed in the field and laboratory research(Kaiser et al.,1996;Ortlepp, 2001; He et al., 2010; Jiang et al., 2015; Su et al.,2017c,d).In this study,with rockbursts of different rocks being successfully reproduced,rockburst proneness and severity of these rocks could be evaluated directly with the estimated ejection kinetic energy.

Table 4 shows the proneness of the six rocks to rockburst evaluated primarily based on the average kinetic energy of ejected fragments.It can be seen that granodiorite and marble are strong,basalt and marble moderate,sandstone slight,and limestone has no ejection.This confirms that igneous and metamorphic rocks are generally more prone to rockburst than sedimentary rocks(Singh et al.,2008).The fragmentation characteristics and failure modes of the six types of rocks during rockburst roughly exhibited two different regimes,and the former case could potentially permit the relatively strong burst.High rock strength could always contribute significantly to burst when overloaded,regardless of fragmentation characteristics and failure mode.In fact,rock strength is generally positively associated with most of the rockburst proneness indices,such as brittle index and elastic strain energy index,and thus highstrength rocks are generally associated with rockburst problems(Singh et al.,2008).On the other hand,rocks that exhibit comparable strengths,such as granite,marble and basalt,could generate different ejection kinetic energy or burst proneness for the differences in fragmentation characteristics and failure modes.Besides,in Fig.10,significant randomness of kinetic energy of ejected fragments is observed,resulting from the intrinsic variability of rocks and the structural uncertainty generated during evolution of rockburst.

It should also be pointed out that the stresses can cause failure of different rocks in this study.If the field stresses are not extremely large,the proneness of different rocks to rockburst will be different.For example,the sandstone will more likely produce burst than the granodiorite under a relatively low loading condition.A series of rockburst tests(He et al.,2007,2010;Su et al.,2017a,b,c,d)have indicated that loading/unloading rate,tunnel axial stress,radial stress gradient and dynamic disturbance can significantly influence the rockburst behaviors of rocks.To better evaluate the rockburst tendency of different rocks under various loading conditions,extensive rockburst experiments are required.

6. Conclusions

In the present study,rockburst tests on six types of hard brittle rocks(granodiorite,granite,basalt,marble,sandstone and limestone)were conducted,following the one-face-free true triaxial loading scheme.The rockburst behaviors were successfully reproduced and recorded with the use of high-speed cameras and an AE system.The rockburst failures were discussed with respect to the fracturing and fragmentation characteristics and failure modes,as well as the evaluation of rockburst proneness and severity.The main observations and conclusions are as follows:

(1)For granodiorite, granite and marble specimens, stressinduced local(micro-or meso-scale)cracks were largely generated during the earlier loading states,and then a large number of macroscopic cracks associated with coalescence of microcracks developed particularly near the free face.Accordingly,the rockbursts were characterized by obvious tensile splitting/slabbing,fragmentation and ejection near the free face and shear rupture away from the free face.

(2)For basalt,sandstone and limestone specimens,significant local fracturing had not occurred during the earlier loading states. Microcracks were locally generated, and shear or tensile-shear rupture nucleated shortly before the peak stress.The rockbursts of basalt and sandstone were characterized by local rupture near the free face and dominated by shear rupture or faulting throughout the specimens.For limestone,no burst failure occurred due to the relatively low strength.

(3)The differences in the inherent microstructure and corresponding microfracturing when loaded resulted in the two different rockburst regimes. Considering the necessary fragmentation and possible energy excess conditions,the former case,slabbing/buckling-shear rupture burst,could potentially permit the occurrence of relatively strong ejection failure.

(4)The proneness and severity of rockburst of different rocks were related to rock strength,fracturing and fragmentation characteristics and failure modes as well as the loading conditions.They could be evaluated primarily based on the kinetic energy of ejected fragments.Under the loading conditions simulated in this study, the relative rockburst proneness(severity)of granodiorite and marble was strong,that of basalt and marble was moderate,that of sandstone was light,and that of limestone was non-existent.

The test results provide insights into the rockburst characteristics and mechanisms on different rocks.However,there is limitation of specimen dimensions in the laboratory test.For example,the scale of fracturing and fragmentation in field rockbursts may be significantly different from that in laboratory.In addition,this paper merely focused on rockbursts of intact rocks,and thus study on rockbursts that initially involve structural features is needed in future.

Declaration of Competing Interest

The authors wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.

Acknowledgments

The authors would like to thank the financial support from the National Natural Science Foundation of China under Grant No.51869003.The work was also supported by the Opening Fund of State Key Laboratory of Geohazard Prevention and Geoenvironment Protection(Chengdu University of Technology)under Grant No.SKLGP2017K022 and Study Abroad Program for Excellent PhD Students of Guangxi University.