Tunnelling and Underground Space Technology 43 (2014) 315–326

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

2D numerical simulation on excavation damaged zone induced by

dynamic stress redistribution
W.C. Zhu a,⇑, J. Wei a, J. Zhao b, L.L. Niu a
a
Key Laboratory of Ministry of Education on Safe Mining of Deep Metal Mines, Northeastern University, Shenyang 110819, China
b
Ecole Polytechnique Fédérale de Lausanne (EPFL), Laboratory of Rock Mechanics (LMR), CH-1015 Lausanne, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: The formation of an excavation damaged zone (EDZ) around an opening in a deep rock mass is associated
Received 28 June 2013 with the dynamic stress redistribution that starts from transient release of high in situ stress to the final
Received in revised form 8 May 2014 quasi-static stress state after the excavation. This study applies a theoretical analysis of stress redistribu-
Accepted 29 May 2014
tion due to transient unloading in surrounding rock under hydrostatic stress field, and develops a numer-
Available online 26 June 2014
ical elastodynamics model for finite element analysis. Coupling the theoretical and the numerical
solutions, a general damage model for heterogeneous rock mass is proposed by taking the dynamic stress
Keywords:
redistribution due to excavation into account. Finally, the dynamic stress redistribution, as well as the
Excavation damaged zone (EDZ)
Transient unloading
induced damage zone around the excavation under different lateral pressure coefficients is numerically
Dynamic stress redistribution simulated. The numerical result indicates that, the stress wave induced by the transient unloading will
Quasi-static far-field stress initially cause the damage only in the 1/3 radius vicinity of excavation perimeter. The damage zone
Lateral pressure coefficient may then develop further under the constant quasi-static far-field stress. Therefore, the EDZ development
Numerical simulation during deep excavation is closely dependent on in situ stress, rock strength and excavation method.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction released stress is rather significant in a relatively short duration.

It is unclear whether the quasi-static assumption is acceptable
The formation of an excavation damaged zone (EDZ) is expected (Lu et al., 2012). Some studies have revealed the necessity of study-
around excavated openings at depth in civil (e.g., tunnels and cav- ing the transient process of stress redistribution and related
erns), mining (e.g., shafts, tunnels, drifts and stopes), and petro- dynamic response.
leum engineering (e.g., borehole) (Zhu and Bruhns, 2008). As early as 1966, Cook et al. (1966) indicated that impulsive
Extensive studies have been performed to understand and predict release of the applied load could lead to over-relaxation of the dis-
the extent of EDZ, and in recent years advances have been made in placing rock, generating tensile stresses in the medium. Abuov and
the understanding of the formation mechanism of EDZ (Falls and Aitaliev (1988) pointed out that with the formation of a new open
Young, 1998; Backblom and Martin, 1999; Cai et al., 2001; surface in the rock mass, load-relief waves are formed, which leads
Diederichs et al., 2004; Read, 2004; Martino and Chandler, 2004). to the transfer of potential energy of bulk compression into kinetic
It is generally accepted that in high in situ stress conditions the energy. During this process, particles of the rock move toward the
excavation induced stress redistribution is the main cause for the surface of rock exposure, and a rockburst may occur when the
formation of EDZ, which plays a more important role on the extent potential energy of compression reaches a specific level. Carter
of EDZ than that of the excavation method (Falls and Young, 1998). and Booker (1990), as well as Wang and Huang (1998) deemed that
Many researchers deem that the stress redistribution during the the dynamic disturbance due to transient release rate of high
formation and development of EDZ is a quasi-static process. This in situ stress has great influence on the extent of EDZ and resultant
approximation is generally acceptable if the level of the in situ rockburst occurrence. Cai (2008) thought that, in addition to
stress is low. In fact, excavations trigger sudden release of in situ dynamic stress wave and blasting-induced gas pressure, there is
stress which leads to a strong transient disturbance to the sur- another mechanism that is dynamic unloading that may contribute
rounding rock mass. Under high in situ stress condition, the to the blasting-induced rock damage. For blasting induced damage
in excavation walls, the loss of confinement (excavation) and
dynamic loading from wave propagation cause both intended
⇑ Corresponding author. Tel.: +86 24 83687705; fax: +86 24 83681186. and unintended damage. Zhou and Qian (2007) and Li et al.
E-mail address: zhuwancheng@mail.neu.edu.cn (W.C. Zhu). (2009) took the stress redistribution as a dynamic process to

http://dx.doi.org/10.1016/j.tust.2014.05.023
0886-7798/Ó 2014 Elsevier Ltd. All rights reserved.
316 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

interpret the tension–compression alternation and zonal disinte- under a change of applied stresses (positive for tension) is
gration phenomena around a deep tunnel. Sun et al. (2011) and expressed as (Timoshenko and Goodier, 1951)
Lu et al. (2012) considered that the dynamic unloading wave
induced by transient release of in situ stress plays an important G @ 2 ui
Gui;jj þ uj;ji þ F i ¼ q 2 ; ð1Þ
role in vibration of surrounding rock. Chen et al. (2011) suggested 1  2m @t
a method to determine the radius of the broken and plastic soften- where ui (i = x, y, z) is displacement (m), t is time (s), q is rock den-
ing zone when considering the dynamic response of surrounding sity (kg/m3), G is shear modulus (Pa), v is the Poisson’s ratio, and Fi
rock. Yin et al. (2012) found the fractal dimension of sandstone is the components of the net body force in the i-direction (N/m3).
gradually increases with the unloading rate increasing. Wei et al. This equation expresses the mechanical equilibrium in rock sub-
(2014) explained the failure mechanism of transient unloading in jected to dynamic loading. It could be used for quasi-static analysis
surrounding rock. In conclusion, the dynamic stress redistribution when the acceleration term in right-hand term is set to zero. This
induced damage zone results from two factors, i.e., a dynamic is general 3D equation, as for numerical simulation in Section 4, it
unloading stress induced by transient release of in situ stress and is simplified for 2D plain strain problem.
a quasi-static secondary stress due to the in situ stress. Therefore, As illustrated in Fig. 1, the damage in tension or shear mode of
the stress redistribution under high in situ stress condition is a rock is initiated when its state of stress satisfies the maximum ten-
dynamic process and rock dynamics approach could be applied sile stress criterion or the Mohr–Coulomb criterion, respectively, as
(Zhao et al., 1999, 2011). expressed by:
The main mechanism in the development of EDZ is the initia-
tion and growth of cracks and fractures, owing to stress redistribu- F 1  r1  ft0 ¼ 0 or F 2
tion. Because of the anisotropy and heterogeneity of rock, which ¼ r3 þ r1 ½ð1 þ sin /Þ=ð1  sin /Þ  fc0 ¼ 0 ð2Þ
may be also altered with damage evolution, it is difficult to theo-
retically characterize the EDZ. The field instrumentation records where ft0 and fc0 are uniaxial tensile and compressive strength (Pa),
contain unique deformation signatures that provide insight into respectively, / is internal frictional angle, and F1 and F2 are two
the mechanical response of rock mass to stress redistribution and damage threshold functions used to link the tensile and shear dam-
the formation of an EDZ. However, due to the limit of in situ data age, respectively.
obtained, it is usually difficult to clarify the associated mechanism According to the principle of elastic damage, the elastic modu-
that is responsible for the formation of EDZ. Moreover, attempts to lus of an element degrades monotonically as damage evolves, and
generate fractures by impulsive unloading in laboratory tests have the elastic modulus of damaged material is expressed as:
been unsuccessful (Brady and Brown, 2004). Many researchers
E ¼ ð1  DÞE0 ð3Þ
(Lajtai, 1998; Hajiabdolmajida et al., 2002; Suknev, 2008; Feng
et al., 2012) tend to believe that the rock failure is dominated by where D represents the damage variable, which lies between 0 and
tensile fracture at the beginning, especially under unloading condi- 1, and E and E0 are the elastic moduli of the damaged and the
tion, but arguments still exist. Therefore, it is significant to develop undamaged material (Pa), respectively. In this kind of numerical
effective numerical models that can capture the damage evolution simulation, the element as well as its damage is assumed isotropic,
during the stress redistribution caused by both dynamic unloading so the E, E0 and D are all scalar. According to Fig. 1, the damage var-
and quasi-static in situ stresses, in order to fully characterize the iable can be calculated as:
spatial and temporal development of EDZ in rock mass. 8
To this end, it is the dynamic and quasi-static response induced > 0 F 1 < 0 and F 2 < 0
>
>  n
by excavation of rock mass that defines the objective of this work.
< et0 
D¼ 1   e1  F 1 ¼ 0 and dF 1 > 0 ð4Þ
In this respect, an elastodynamic analysis on the dynamic unload- >
>  n
>
: 1  ec0 
ing response of surrounding rock under hydrostatic stress field is e3 F 2 ¼ 0 and dF 2 > 0
firstly given. Then, when the stress redistribution resulted from
transient release of in situ stress and quasi-static far-field stress where et0 and ec0 are maximum principal strain in tension and max-
is taken into account, a general damage model for simulating imum principal strain in compression when damage occurs, respec-
EDZ in heterogeneous rock is proposed and programmed into tively, and n is a constitutive coefficient and it is 2.0. In this respect,
COMSOL Multiphysics, a partial differential equation (PDE)-based the damage variable calculated with Eq. (4) is always from 0 to 1.0
multiphysics modeling environment (COMSOL, 2008). In addition, regardless of what kind of damage it may suffer. However in the
the numerical model is validated by simulating the elastodynamic damage zone figure, in order to distinctly display the two kinds of
response during the excavation in homogeneous rock under hydro- damage modes (i.e. tensile damage and shear damage), the tensile
static stress field. Finally, the 2D numerical simulations on the
dynamic stress redistribution and resultant damage zone under
different lateral pressure coefficients (i.e., j = 0.2, 1.0, 2.0, it is a
ratio of horizontal far-field stress to vertical far-field stress) are
comprehensively conducted. Although this 2D analyses is some-
what different from the real tunneling practices, for example,
blasting in the mining face and the existed fractures are not prop-
erly addressed, it is really important for clarifying the associated
mechanism responsible for the EDZ development due to the
dynamic stress redistribution.

2. Damage mechanics – based model

Initially the porous medium is assumed elastic, with constitu-

tive relationship defined by a generalized Hooke’s law. In this
regard, a modified Navier’s equation, in terms of displacement Fig. 1. The elastic damage-based constitutive law under uniaxial stress condition.
 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326 317

damage is represented as negative numbers, while the shear dam-

age is represented as positive ones.
It should be noted that, in the numerical implementation of Eq.
(4), the tensile damage is always preferable to shear one, that is to
say, the maximum tensile stress criterion is firstly used to judge
whether the elements damage in tension or not. Only the elements
that do not damage in tensile mode will be checked its shear dam-
age with the Mohr–Coulomb criterion.

3. Problem definition and its analytical solution Fig. 3. Time-dependent boundary loads for model II for analytical analysis of stress
redistribution in consideration of transient unloading.
3.1. Analytical solution under hydrostatic stress field

Assuming that there is an infinite geologic body, under the Under the hydrostatic stress field, namely j = 1, the above prob-
in situ horizontal stress px (equals to jp0) and vertical stress py lem can be solved analytically. The stresses in model I yield the fol-
(equalling to p0), respectively. A circular tunnel with a radius a is lowing expressions:
excavated, resulting in the release of normal and shear stresses
which unload simultaneously linearly with duration of t0 on the
rr1 ¼ p0 rh1 ¼ p0 srh1 ¼ 0 ð6Þ
excavation surface. Extensive studies reveal that the duration t0 In model II, axisymmetric cylindrical stress wave is spread in
depends on the different excavation methods (such as tunnel bor- the medium. Radial displacement is irrelevant to h and z, but it is
ing machine (TBM) or drilling-and-blasting) (Li et al., 2009; Yan the functions of r and t due to the axial symmetry (Yang and
et al., 2009). The dynamic stress induced by transient release of Zhang, 1988). The following statements for the unloading-wave
in situ stress would spread outwards into the surrounding rock problem can be made (Miklowitz, 1960):
mass.
This problem can be resolved into a superposition of two @2U 1 @U 1 @2U
þ  ¼ 0 ðr P a; t P 0Þ ð7Þ
mechanical models (Miklowitz, 1960), as shown in Fig. 2. In model @r 2 r @r c2d @t2
I, the in situ stresses px and py are applied at external boundary in
X- and Y-directions, respectively, and the radial compressive stress @ Uðr; tÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ur2 ¼ ; cd ¼ ðk þ 2GÞ=q;
rr1 and shear stress srh1 are applied at the inner boundary, which @r
both vary with the polar angle h and can be gained by the coordi- Em E
k¼ ; G¼ ð8Þ
nate transformation formulas like Eq. (5) as (Xu, 2006) ð1 þ mÞð1  2mÞ 2ð1 þ mÞ

rr1 ¼ rx cos2 h þ ry sin2 h þ 2sxy sin h cos h Uðr; 0Þ ¼

@ Uðr; 0Þ
¼ 0ðr P aÞ; limUðr; tÞ ¼ 0ðt P 0Þ; rr2 ða; tÞ ¼ pðtÞ
2 @t r!1
¼ p0 ðj cos2 h þ sin hÞ ð5aÞ
ð9Þ
8
srh1 ¼ ðry  rx Þ sin h cos h þ sxy ðcos2 h  sin2 hÞ <0
> ðt < 0Þ
p
¼ p0 ð1  jÞ sin h cos h: ð5bÞ pðtÞ ¼  t00 t ð0 6 t < t0 Þ ð10Þ
>
:
p0 ðt P t0 Þ
This is a quasi-static model representing the strain and dis-
placement caused by quasi-static in situ stresses, in which the Eqs. (7), (9), and (10) are the displacement equations of motion, ini-
stresses of each point are not time-dependent. tial and boundary conditions, respectively, where U is potential
In model II, there is no in situ stress applied. At the excavation function of displacement, ur2 is the displacement in the r-direction,
moment t = 0, the radial tensile stress rr2 and shear stress srh2, r is the distance from excavation centre, a is the excavation radius,
which are both time-dependent loads, as shown in Fig. 3 (positive t0 is unloading duration, cd is longitudinal wave velocity, E is elastic
for tension), are applied at the inner boundary suddenly. Finally modulus, and k and G are lame constants.
the surrounding rock (considered as elastic) remains stable under The tangential stress expressions can be obtained by making
the constant quasi-static far-field stress, which is the denoted with use of Laplace transform, inversion integral and contour integra-
‘t ? 1’, as shown in Fig. 2(c). tion (Miklowitz, 1960):

(a) Model I (static) (b) Model II (dynamic) (c) Final stress state
Fig. 2. Partition of mechanical model for analytical solution of stress redistribution in consideration of transient unloading.
318 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

Table 1
Basic parameters used in theoretical analysis under hydrostatic stress field.

Elastic modulus E (GPa) Poisson’s ratio m Density q (kg m3) Excavation radius a (m) In situ vertical stress p0 (MPa)
47.2 0.23 2700 5 25

 
rh2 ðr; tÞ ra are specified, in order to examine the effect of this unloading dura-
¼0 06t< ð11aÞ
p0 cd tion on the resultant dynamic response during the excavation.
Fig. 4(a and b) represent the radial stress rr/p0 and tangential
  Z 1
rh2 ðr; tÞ a2 ra 1 stress rh/p0 of particular unloading durations at r = 5.1 m (i.e.
¼ t 
p0 2
2r t 0 cd t0 p 0 0.1 m away from the excavation boundary), respectively. In this
8 h  i h   i9
<ðA þ BÞ cos xt  cos x ra þ ðC  DÞ sin xt  sin x ra = figure, including the subsequent figures, the stress is normalized
cd cd
 dx with p0. As seen from this illustration, radial stress firstly decreases
: ðF þ KÞx 2 ;
  substantially and then rebounds. Hereafter it undergoes several
ra ra
6t< þ t0 ð11bÞ smaller fluctuations and finally stabilizes until the quasi-static
cd cd
far-field stress conditions (Fig. 4(a)). Tangential stress increases
Z 1 rapidly and then decreases. Hereafter it also undergoes several
rh2 ðr; tÞ a2 1
¼  smaller fluctuations and remains stable finally (Fig. 4(b)). The
p0 2r 2 t 0 p 0
 dynamic effect of stress redistribution is more substantial for the
ðA þ BÞ½cos xt  cos xðt  t 0 Þ þ ðC  DÞ½sin xt  sin xðt  t 0 Þ
 dx shorter unloading duration. In particular, when t0 equals to 2 ms,
ðF þ KÞx2
  the tensile stress is induced at about t = 4 ms. In contrast, the
ra
tP þ t0 ð11cÞ
cd dynamic effect for t0 = 24 ms could be negligible, and thus tensile
stress is never found. However, dynamic effect does indeed exist,
A ¼ a1 a2 ; B ¼ a3 a4 ; C ¼ a3 a2 ; D ¼ a1 a4 ; F ¼ a2 a2 ; K ¼ a4 a4 which can be seen from the magnified curve of tangential stress
history (t0 = 24 ms) as shown in Fig. 4(b).
    Fig. 5 shows the radial stress rr/p0 and tangential stress rh/p0, at
2G x G x x the specific locations away from the excavation face. The dynamic
a1 ¼ J1 r þ J0 r ;
ðk þ 2GÞr cd ðk þ 2GÞ cd cd effect could be observed obviously. In particular, radial stress
   
2G x G x x shows tension–compression fluctuation adjacent to the excavation
a2 ¼ J1 a  J0 a ;
ðk þ 2GÞa cd ðk þ 2GÞ cd cd perimeter (r = 5.1 m), where the induced tensile stress may poten-
tially contribute to tensile damage around the excavation, because
    the rock mass is usually prone to tensile damage. Moreover, as
2G x G x x
a3 ¼ Y1 r þ Y0 r ; shown in Fig. 5(b), it may also incur the maximum dynamic tan-
ðk þ 2GÞr cd ðk þ 2GÞ cd cd
    gential stress that is considerably larger than the static stress there,
2G x G x x
a4 ¼ Y1 a  Y0 a ; which may possibly induce the compressive or shear damage
ðk þ 2GÞa cd ðk þ 2GÞ cd cd
around the excavation. In general, the transient unloading stress
where J and Y are ordinary Bessel functions. Similarly, the radial during the underground excavation may induce damage prior to
stress expressions can be obtained as a function of time. the final quasi-static stress redistribution resulted from excava-
The superposition of the solutions of the two models is the com- tion, which may be detrimental to the stability of underground
plete solution in the present problem. The basic constants specified opening.
in the theoretical analysis are listed in Table 1.
Extensive studies reveal that the transient stress redistribution 3.2. Comparison between numerical and analytical solutions
can finish within a few milliseconds during excavation either by
tunnel boring machine (TBM) or drilling-and-blasting methods When the lateral pressure coefficient is not equal to one (j – 1),
(Li et al., 2009; Yan et al., 2009). According to Yan et al. (2009), it is difficult to obtain the analytical expressions of the problem.
the durations of 2 ms and 24 ms are used to represent unloading However, it can be solved with resort to the numerical methods
process during the TBM and drilling-and-blasting excavation, such as finite element method (FEM). In this section, the theoreti-
respectively. The unloading durations of 2 ms, 10 ms and 24 ms cal model presented in Section 3.1 is implemented into COMSOL

2
1.9

1.2 2.2 1.8

20 25 30 35 40
1 2
t0=2ms
0.8 1.8
t0=10ms
σr /p0

σθ /p0

0.6 1.6 t0=2ms

0.4 t0=24ms 1.4 t0=10ms
0.2 1.2 t0=24ms
0 1
-0.2 0.8
0 10 20 30 40 50 0 10 20 30 40 50
t (ms) t (ms)
(a) Radial stress (b) Tangential stress
Fig. 4. Analytical solution to stress time history for different unloading durations (r = 5.1 m, j = 1.0, p0 = 25 MPa).
 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326 319

1 2.1
0.8 1.9
r=5.1m
0.6 r=5.1m 1.7

σθ /p0
σr /p0
r=10m
0.4 r=10m 1.5
r=25m
0.2 r=25m 1.3
0 1.1
-0.2 0.9
0 10 20 30 40 50 0 10 20 30 40 50
t (ms) t (ms)
(a) Radial stress (b) Tangential stress
Fig. 5. Analytical solution to stress time history at different locations (t0 = 2 ms, j = 1.0, p0 = 25 MPa).

1.2
t0=0.2ms
1
t0=2ms
0.8
t0=24ms

σr /p0
0.6
0.4
0.2
0
-0.2
0 5 10 15 20 25 30
t (ms)

Fig. 8. Radial stress time history of numerical solutions under different unloading
durations (r = 5.1 m, j = 1.0, p0 = 25 MPa).

Table 2
Weibull distribution parameters of heterogeneous rock.

Homogeneity Mean of Mean of Ratio of uniaxial Internal

index Young’s uniaxial compressive frictional
Fig. 6. Finite element mesh. modulus compressive and tensile angle (°)
(GPa) strength (MPa) strength
5 47.2 80 80/6 35
Multiphysics (COMSOL, 2008). In the numerical model, the 2D
domain is 600 m  600 m in dimension and is discretized into
27,408 triangular elements, and it has been proved that the stress mesh around the excavation boundary is refined, as shown in
wave cannot arrive at the external boundary and being reflected Fig. 6. The termination time of the transient analysis is 45 ms.
back throughout the duration of transient unloading. In order to Due to the material heterogeneity set to every element, the diffrac-
guarantee the convergence of finite element analysis (FEA), the tion may occur at every element within the numerical model,

1.2 2.2
1 2
0.8 Theoretical 1.8
σθ /p0
σr /p0

0.6 1.6 Theoretical

Numerical
0.4 1.4
Numerical
0.2 1.2
0 1
-0.2 0.8
0 10 20 30 40 50 0 10 20 30 40 50
t (ms) t (ms)
(a) Radial stress, t0 = 2 ms (b) Tangential stress, t0 = 2 ms
1.2 2.2
1 2
0.8 1.8
Theoretical
σr /p0

σθ /p0

1.6 Theoretical
0.6
Numerical 1.4
0.4 Numerical
1.2
0.2 1
0 0.8
0 10 20 30 40 50 0 10 20 30 40 50
t (ms) t (ms)
(c) Radial stress, t0 = 24 ms (d) Tangential stress, t0 = 24 ms
Fig. 7. Comparison between theoretical and numerical solutions to the stress distribution under different unloading durations (r = 5.1 m, j = 1.0, p0 = 25 MPa).
320 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

therefore the time step must be shorter enough to represent this a logical way to validate the numerical model is to compare the
effect. In this regard, during the numerical simulation, the initial numerical solutions with the analytical ones. As shown in Fig. 7,
time step is 0.2 ls and the maximum time step is 0.5 ls, namely it is found that the numerical solutions align favorably with the
the time step varying between 0.2 ls and 0.5 ls is adopted. The theoretical solutions, even though the induced tensile radial stress
relative and absolute tolerances for the iteration during the FEA at about t = 4 ms is not captured. The reason for this may be the
are 1e5 and 1e6, respectively. The time interval for outputting precision of FEM in solving this elastodynamic problem. However,
the numerical simulation is set as 0.3 ms. It has been checked that as shown Fig. 7(a), the tensile stress is only about 0.04 p0 when
the time step is short enough to guarantee the convergence and t0 = 2 ms, compared to the in situ stress, it is too low to induce
high resolution for this FEA. the tensile damage of the rock.
Prior to the FEA of transient stress redistribution around the As shown in Fig. 8, the existence of tensile stress around the
underground opening under different lateral pressure coefficients, boundary can also be found with our numerical simulation when

Fig. 9. Stress and damage zone distribution in heterogeneous surrounding rock around the excavation when lateral pressure coefficient j equals to 1.0 (t0 = 24 ms).
 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326 321

t0 is 0.2 ms, thus denoting the correlation of this induced tensile it is deemed that this damage-based model is feasible to simulate
stress with the unloading duration. The induced tensile stress is the excavation damaged zone induced by dynamic stress redistri-
prone to occurring under the shorter unloading duration. In addi- bution with certain accuracy.
tion, the existence of tensile stress, together with resultant tensile
damage zone, is also dependent on the rock heterogeneity, because
the rock heterogeneity may trigger transmission and reflection of 4. Numerical simulation on damage zone development
the stress wave. Therefore the rock heterogeneity is a crucial factor
dominating the tensile stress and damage zone distribution around In this section, the damage model considering the stress wave
the excavation, which will be discussed in Section 4. induced by the transient release of in situ stress and quasi-static
Based on the comparison between numerical and analytical stress due to the in situ stress field is used to assess the damage
solutions, together with the validation of this model in predicting zone development around the opening when the rock heterogene-
the rock blasting in the previous publications (Zhu et al., 2013), ity is taken into account. According to the description of rock

Fig. 10. Stress and damage zone distribution in heterogeneous surrounding rock around the excavation when lateral pressure coefficient j equals to 0.2 (t0 = 24 ms).
322 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

heterogeneity used in previous publications (Zhu and Tang, 2004; Weibull distribution with the parameters as listed in Table 2, and
Zhu et al., 2005, 2010), the rock is assumed to be heterogeneous no heterogeneity is introduced in ratio of uniaxial compressive/
with its Young’s modulus and strength specified according to a tensile strength and internal frictional angle. The values of other

Fig. 11. Stress and damage zone distribution in heterogeneous surrounding rock around the excavation when lateral pressure coefficient j equals to 2.0 (t0 = 24 ms).
 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326 323

parameters are specified identical to the above-mentioned elastic sistency between the homogeneous model (Fig. 7) and
analysis in previous sections. Due to the heterogeneity of material heterogeneous model (Fig. 9) suggests that the effect of rock heter-
parameters within the numerical model, the wave diffraction may ogeneity on the stress wave propagation is a very important factor
occur at different elements, in this regard, the accuracy of FEA for the induced tensile stress even though the unloading duration
could be guaranteed when the time step shorter than 0.5 ls is t0 is as long as 24 ms.
adopted. By comparing the damage zone at t = 25.2 ms and that at t ? 1,
The 2D domain for the numerical simulation is 600 m  600 m it is found that the quasi-static far-field stress hardly contributes to
in dimensions. However, in the following figures, the display is the further development of damage zone under this lateral pres-
focused only around the excavation with a size of 10 m  10 m sure coefficient. The lateral pressure coefficient of j = 1.0 is favor-
or 30 m  30 m in order to observe the damage zone clearly. The able to the stability of the excavation with circular cross section;
dynamic analysis under transient unloading will not be stopped therefore, the damage caused by the quasi-static far-field stress
until there is no further increase in damaged area, and it is fol- is very limited. Therefore, the tensile stress induced by transient
lowed by the quasi-static analysis under far-field stress that is unloading is the predominant mechanism responsible for the dam-
denoted by ‘t ? 1’. age around the excavation. Moreover, due to the rock heterogene-
ity incorporated in the numerical model, the stress distribution,
4.1. Effect of lateral pressure coefficient tensile stress zone and tensile damage zone are not symmetrically
distributed in the sidewall and roof although the lateral pressure
4.1.1. Case I (j = 1.0) coefficient is 1.0.
When lateral pressure coefficient j is 1.0, the stress and damage
zone distribution are shown in Fig. 9, in which the elastic solution 4.1.2. Case II (j = 0.2)
for homogeneous rock specimen is also shown, in order to high- Fig. 10(a) presents the distribution of the major principal stres-
light the effect of rock heterogeneity on the damage zone develop- ses along the horizontal line (h = 0°) and the vertical line (h = 90°)
ment. As shown in Fig. 9(a), at t = 21.3 ms, the stress along the under the lateral pressure coefficient j = 0.2, in which the numer-
radial direction for the heterogeneous rock model with a homoge- ical result for homogeneous rock specimen is also shown. Fig. 10(b)
neity index of 5.0 fluctuates near that for the homogeneous elastic highlights this tensile stress zone induced by transient unloading,
model, and no tensile stress is induced. However, along with the which may be high enough to induce tensile damage, as shown
propagation of unloading wave, at t = 25.2 ms the tensile stress in Fig. 10(c). At t = 21.3 ms, the tensile stress is found in the roof
(Fig. 9(b)) and the associated tensile damage (Fig. 9(c)) with a even for the homogeneous elastic model, so is the heterogeneous
depth of 0.6 m emerge around the excavation, which is not charac- model. Therefore, as shown in Fig. 10(c), very shallow damage zone
terized for the homogeneous model as shown in Fig. 7. The incon- is induced in the roof and floor for the heterogeneous model. At

Fig. 12. Final damage zone under different lateral pressure coefficients when the contribution of both transient unloading stress and constant quasi-static far-field stress are
taken into account.
324 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

Fig. 13. Final damage zone distribution under different unloading durations and that without considering transient unloading (j = 2).

t = 25.2 ms, the tensile damage zone extends further in the roof excavation perimeter. Thereafter, independent of the lateral pres-
until a depth of about 1.5 m, because of the propagation of tran- sure coefficient j of the in situ stress, the tensile damage zone
sient unloading stress wave. At the same time, there is odd distri- induced by transient unloading is very shallow, approximating
bution of shear damage found in the east and west sidewalls. As 0.6 m, 1.5 m and 1.3 m for j = 1.0, 0.2 and 2.0, respectively. Most
show in Fig. 10(a), the stress concentration in the damaged zone importantly, the shear damage with a depth of 1.6 is induced
in the roof is released because the load-bearing capacity in the around the excavation under the lateral pressure coefficient of
damage zone declines significantly. 2.0. Anyway, under these 3 lateral pressure coefficients, the dam-
After the contribution of transient unloading, the damage zone age zones induced by transient unloading are not deeper than
may develop a little under the action of quasi-static far-field stress 1/3 radius around the excavation. As suggested by Brady and
(at t ? 1), indicating that the quasi-static far-field stress only Brown (2004), the impulsive release of the geo-stress due to exca-
plays a little role in the damage zone development. Due to the rock vation may lead to over-relaxation of the rock, based on which the
heterogeneity, once damage occurs in an element, it may induce over-break of surrounding rock during blasting can be explained
the stress concentration around it, thus enhancing rock heteroge- most satisfactorily. In this regard, the damage zone induced by
neity further. Therefore the damage zone around the excavation transient unloading could satisfactorily reveal the mechanism of
is asymmetrical even although the loading conditions and geome- over-break when the dynamic stress redistribution due to excava-
try of the domain are symmetrical. tion is taken into account.

4.1.3. Case III (j = 2) 4.2. Effect of unloading duration

Fig. 11 gives the distributions of the major principal stress,
minor principal stress and damage zone under the lateral pressure As shown in Fig. 12(a), under the increasing lateral pressure
coefficient j = 2.0. At t = 21.3 ms, no tensile stress appears around coefficients from 0.2 to 2.0, the shear damage zone in the roof
the opening, but the compressive stress concentration is found, and floor becomes emerged, because the higher in situ stress at
resulting in a small amount of shear damage in the roof and floor. the horizontal direction is applied. The lateral pressure coefficient
As time elapses, at t = 25.2 ms, the tensile stress can be seen all is a key factor that not only affects the damage zone induced by
around the excavation but the tensile damage zone is mainly transient unloading, but also governs the damage zone induced
induced in the sidewall (with a depth of 1.3 m), while the shear by quasi-static far-field stress.
damage zone (with a depth of 1.6 m) is dominated in the roof When the unloading duration is as short as 2.0 ms, as shown in
and floor because of the highly concentrated compressive stress. Fig. 12(b), the larger extent of damage zone is induced by transient
In general, the damage zone induced by transient unloading is still unloading even the damage pattern is similar, thus confirming that
shallow around the excavation with a maximum depth of 1.6 m. the effect of dynamic effect caused by transient unloading is clo-
After the transient unloading, the damage zone in the roof and sely related to the unloading duration. For example, when
floor develops further under the contribution of quasi-static far- j = 1.0, the damage zone at the roof induced by transient unload-
field stress. At this stage (t ? 1), as shown in Fig. 11(d), the high ing is 1.7 m in depth under unloading duration of 2 ms. In contrast,
compressive stress is induced in the roof and floor of the excava- it is 0.6 m in depth when the unloading duration is 24 ms.
tion, thus extending the shear damage at this area until a depth Fig. 13 shows the comparison of final damage zone under differ-
of 12.1 m. The stress concentration in the roof and floor is released ent unloading durations when the lateral pressure coefficient is
because of the local damage zone development, which can be 2.0. In contrast, the incorporation of transient unloading during
found from the curve of major principal stress and minor principal the numerical simulation can increase the extent of damage zone
stress as shown in Fig. 11(a and b), respectively. Under this lateral adjacent to the excavation perimeter, which may become larger
pressure coefficient, the damage zone development under the with the shorter unloading duration. In this respect, the stress
quasi-static far-field stress is predominant to develop the damage wave induced by transient unloading will firstly cause the damage
zone around the excavation. at the vicinity of excavation perimeter; however, it may also con-
The dynamic tensile stress is induced when the transient tribute to the further expansion of the final damage zone under
unloading during the excavation is taken into account, it may exist the quasi-static far-field stress. For instance, when the lateral pres-
around the deep underground excavation regardless of the lateral sure coefficient is 2.0, the damage zone depth at the roof is 7.6 m if
pressure coefficient j. However, under these three lateral pressure transient unloading is not taken into account. By contrast, it may
coefficients as discussed above, the damage zone induced by the increase to 12.1 m and 13.4 m when the transient unloading with
transient unloading is only in a small range in the vicinity of the the duration of 24 ms and 2 ms are considered.
 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326 325

5. Conclusion South Africa Joint Research Programme (Grant No. CS06-L01/

2012DFG71060), the Key Project of Chinese Ministry of Education
The excavation-induced stress redistribution and resultant (No. 113019A), the Fundamental Research Funds for the Central
damage zone are dominated by a dynamic process starting from Universities of China (Grant Nos. N110201001 and N120101001)
transient release of high in situ stress to the final quasi-static and the Sino-Swiss Science and Technology Cooperation (SSSTC)
far-field stress after the excavation. In this study, the excavation program (Grant No. EG31-092011).
damage zone induced by transient unloading and by quasi-static
far-filed stress, as well as their relevance to the in situ stress con-
dition and unloading duration, are studied mainly based on the 2D
numerical simulation. The following conclusions are drawn: References

Abuov, M.G., Aitaliev, S.M., 1988. Studies of the effect of dynamic processes during
(1) Excavation can trigger strong disturbance in surrounding explosive break-out upon the roof of mining excavations. J. Min. Sci. 24 (6),
rock mass due to the transient release of in situ stress. It 581–590.
Backblom, G., Martin, C.D., 1999. Recent experiments in hard rocks to study the
could lead to tension compression alternation of stress adja-
excavation response: implications for the performance of a nuclear waste
cent to the excavation perimeter. The dynamic effect caused geological repository. Tunn. Undergr. Space Technol. 14 (3), 377–394.
by transient unloading during underground excavation, Brady, B.H.G., Brown, E.T., 2004. Rock Mechanics for Underground Mining, third ed.
becoming more obvious under the shorter unloading dura- Springer, The Netherlands.
Cai, M., 2008. Influence of stress path on tunnel excavation response – numerical
tion, may induce foregoing tensile stress prior to the final tool selection and modeling strategy. Tunn. Undergr. Space Technol. 23 (6),
quasi-static stress redistribution. This dynamic tensile stress 618–628.
induced by transient unloading can be captured with our Cai, M., Kaiser, P.K., Martin, C.D., 2001. Quantification of rock mass damage in
underground excavations from microseismic event monitoring. Int. J. Rock
numerical simulations when the unloading duration is as Mech. Min. Sci. 38 (7), 1135–1145.
short as 0.2 ms, which may also be dominated by heteroge- Carter, J.P., Booker, J.R., 1990. Sudden excavation of a long circular tunnel in elastic
neity of surrounding rock even though the unloading dura- ground. Int. J. Rock Mech. Min. Sci. 27 (2), 129–132.
Chen, J.G., He, H., Zhang, Y.X., 2011. Dynamic and static analysis of mechanism of
tion t0 is as long as 24 ms. loosen zone in surrounding rock (in Chinese). Chinese J. Geotech. Eng. 33 (12),
(2) The dynamic tensile stress, together with the resultant dam- 1964–1968.
age zone, induced by transient unloading during the excava- COMSOL AB. 2008. COMSOL Multiphysics Version 3.5, User’s Guide and Reference
Guide <http://www.comsol.com>.
tion may exist around the deep excavation regardless of the
Cook, M.A., Cook, U.D., Clay, R.B., 1966. Behavior of rock during blasting. Trans. Soc.
lateral pressure coefficient. However, independent of the lat- Min. Eng. 10 (2), 17–25.
eral pressure coefficient j of the in situ stress, such as the Diederichs, M.S., Kaiser, P.K., Eberhardt, E., 2004. Damage initiation and propagation
in hard rock during tunnelling and the influence of near-face stress rotation. Int.
in situ stress condition specified in this study, the tensile
J. Rock Mech. Min. Sci. 41 (5), 785–812.
damage zone induced by transient unloading is very shal- Falls, S.D., Young, R.P., 1998. Acoustic emission and ultrasonic-velocity methods
low, approximating 0.6 m, 1.5 m and 1.3 m for j = 1.0, 0.2 used to characterize the excavation disturbance associated with deep tunnels in
and 2.0, respectively. Most importantly, the shear damage hard rock. Tectonophysics 289 (1–3), 1–15.
Feng, X.T., Chen, B.R., Ming, H.J., 2012. Evolution law and mechanism of rockburst in
with a depth of 1.6 is induced around the excavation under deep tunnels: immediate rockburst (in Chinese). Chinese J. Rock Mech. Eng. 31
the lateral pressure coefficient of 2.0. Moreover, transient (3), 433–444.
unloading-induced damage at the vicinity of excavation Hajiabdolmajida, V., Kaiser, P.K., Martin, C.D., 2002. Modelling brittle failure of rock.
Int. J. Rock Mech. Min. Sci. 39 (6), 731–741.
perimeter, may also contribute to the expansion of final Lajtai, E.Z., 1998. Microscopic fracture processes in a granite. Rock Mech. Rock Eng.
damage zone under the constant quasi-static field-field 31 (4), 237–250.
stress. It reveals the mechanism of over-break satisfactorily Li, S.C., Qian, Q.H., Zhang, D.F., 2009. Analysis of dynamic and fractured phenomena
for excavation process of deep tunnel (in Chinese). Chinese J. Rock Mech. Eng.
which cannot be predicted without considering the transient 28 (10), 2104–2112.
unloading. Lu, W.B., Yang, J.H., Yan, P., 2012. Dynamic response of rock mass induced by the
transient release of in-situ stress. Int. J. Rock Mech. Min. Sci. 53, 129–141.
Martino, J.B., Chandler, N.A., 2004. Excavation-induced damage studies at the
Based on the 2D numerical simulation in this work, the mecha-
underground research laboratory. Int. J. Rock Mech. Min. Sci. 41 (8), 1413–1426.
nism for formation of EDZ induced by dynamic stress distribution Miklowitz, J., 1960. Plane-stress unloading waves emanating from a suddenly
due to underground excavation is illustrated. However, there is punched hole in a stretched elastic plate. J. Appl. Mech. 27 (4), 165–171.
Read, R.S., 2004. 20 years of excavation response studies at AECL’s underground
still a gap between the numerical results discussed in this paper
research laboratory. Int. J. Rock Mech. Min. Sci. 41 (8), 1251–1275.
and the tunneling engineering practices. In reality, the geometry Suknev, S.V., 2008. Formation of tensile fractures in the stress concentration zone in
and proximity of the tunnel face have a significant effect on the gypsum. J. Min. Sci. 44 (1), 43–51.
distribution and magnitude of stresses in the walls of tunnel. In Sun, J.S., Jin, L., Jiang, Q.H., 2011. Loosing mechanism of jointed rock mass induced
by transient adjustment of in-situ stress (in Chinese). Chinese J. Vib. Shock 30
addition, there may be fractures into the walls and ahead of the (12), 28–34.
tunnel face before the dynamic excavation takes place, which Timoshenko, S., Goodier, J.N., 1951. Theory of Elasticity. McGraw-Hill Book
could not been considered in 2D modeling. A rigorous 3D numeri- Company, Inc.
Wang, X.N., Huang, R.Q., 1998. Analysis of deformation and failure features
cal analysis is really necessary to simulate this excavation process characteristics of rock under release conditions and rock burst effect.
more realistically, so as to capture the real excavation response due Mountain Res. 16 (4), 281–285 (in Chinese).
to transient unloading. In this respect, this 2D numerical simula- Wei, J., Zhu, W.C., Niu, L.L., Wei, C.H., 2014. Finite element analysis of transient
unloading and plastic zone distribution in surrounding rock (in Chinese). J.
tion is very helpful for clarifying the rock damage mechanics Northeastern Univ. 35 (1), 117–121.
induced dynamic stress redistribution due to excavation, which Xu, Z.L., 2006. Theory of Elasticity (in Chinese). China Higher Education Press,
may lay a solid basis for the future 3D numerical analysis. Beijing, pp. 59–61.
Yan, P., Lu, W.B., Chen, M., 2009. Study of the damage characteristics of surrounding
rocks for tunnels constructed using TBM and drill-and-blast (in Chinese). China
Civil Eng. J. 42 (11), 121–128.
Acknowledgements Yang, G.T., Zhang, S.Y., 1988. Elastodynamics (in Chinese). China Railway Press,
Beijing, pp. 199–203.
Yin, Z.Q., Li, X.B., Jin, J.F., 2012. Failure characteristics of high stress rock induced by
The present work is funded by National Science Foundation of impact disturbance under confining pressure unloading. Trans. Nonferrous
China (Grant Nos. 51222401, 51374049 and 51304037), China- Metals Soc. China 22 (1), 175–184.
326 W.C. Zhu et al. / Tunnelling and Underground Space Technology 43 (2014) 315–326

Zhao, J., Zhou, Y.X., Hefny, A.M., Cai, J.G., Chen, S.G., Li, H.B., Liu, J.F., Jain, M., Foo, S.T., Zhu, W.C., Tang, C.A., 2004. Micromechanical model for simulating the fracture
Seah, C.C., 1999. Rock dynamics research related to cavern development for process of rock. Rock Mech. Rock Eng. 37 (1), 25–56.
ammunition storage. Tunn. Undergr. Space Technol. 14, 513–526. Zhu, W.C., Liu, J., Tang, C.A., Zhao, X.D., Brady, B.H., 2005. Simulation of progressive
Zhao, J., Zhou, Y.X., Xia, K.W., 2011. Keynote: advances in rock dynamics modelling, fracturing processes around underground excavations under biaxial
testing and engineering. In: 12th International Congress on Rock Mechanics, compression. Tunn. Undergr. Space Technol. 20 (3), 231–247.
Beijing. Zhu, W.C., Li, Z.H., Zhu, L., Tang, C.A., 2010. Numerical simulation on rockburst of
Zhou, X.P., Qian, Q.H., 2007. Zonal fracturing mechanism in deep tunnel (in underground opening triggered by dynamic disturbance. Tunn. Undergr. Space
Chinese). Chinese J. Rock Mech. Eng. 26 (5), 877–885. Technol. 25 (5), 587–599.
Zhu, W.C., Bruhns, O.T., 2008. Simulating excavation damaged zone around a Zhu, W.C., Wei, C.H., Li, S., Wei, J., Zhang, M.S., 2013. Numerical modeling on
circular opening under hydromechanical conditions. Int. J. Rock Mech. Min. Sci. destress blasting in coal seam for enhancing gas drainage. Int. J. Rock Mech.
45 (5), 815–830. Min. Sci. 59, 179–190.