Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Geophysical Journal International

Geophys. J. Int. (2013) 194, 1023–1041 doi: 10.1093/gji/ggt143


Advance Access publication 2013 May 16

Numerical simulation of stick-slip behaviours of typical faults


in biaxial compression based on a frictional-hardening
and frictional-softening model

X. B. Wang,1,2 J. Ma1 and Y. S. Pan2


1 State Key Laboratory of Earthquake Dynamics, Institute of Geology, China Earthquake Administration, Beijing 100029, China. E-mail: wxbbb@263.net
2 College of Mechanics and Engineering, Liaoning Technical University, Fuxin, Liaoning 123000, China

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Accepted 2013 April 8. Received 2013 March 31; in original form 2012 November 12

SUMMARY
Stick-slip behaviours of typical faults and interactions among faults are numerically modelled
using a proposed frictional-hardening and frictional-softening elastoplastic continuum model.
Forty numerical tests in biaxial compression are conducted quasi-statically or dynamically in
plane strain and in small or large strain mode using FLAC-3D. Faults are modelled by square
or quadrilateral elements from a viewing angle perpendicular to the maximum surface of a
specimen. An incremental plastic shear strain in a stick-slip cycle is involved in the model,
which is calculated at the beginning of slip and then is set to be zero upon reaching its maximum
at the end of stick. Thus, the repeated stick-slip behaviour can be modelled using the same set
of equations, and only the evolution of an internal frictional angle is required to be different at
different stages. At the slip stage, a decrease of the angle leads to an increase of the incremental
plastic shear strain, while at the following stick stage, it is updated according to the present
incremental plastic shear strain. Nodal velocities change at the two stages because of the use
of dynamic equations even though a rate- and state-dependent law is not introduced. Effects
of loading rate, fault width and maximum incremental plastic shear strain are investigated.
To obtain size-independent stress-deformation curves, a slower loading is required for a finer

GJI Seismology
mesh to ensure the same propagating distance of stress wave. For two intersecting faults or
an echelon fault, a few small events are observed at the stick stage because of interactions
among faults, whereas only one large event is observed at the slip stage. For a specimen with a
bending fault, as the angle between two fault segments is large, the asynchronic softening and
hardening of fault elements lead to a small stress drop at loading ends (stable sliding) rather
than a saw-tooth-like behaviour (stick-slip). To validate the proposed model, a laboratory test
is modelled quasi-statically that was performed on a specimen including an inclined fault
oriented an angle of 60◦ with the horizontal direction in biaxial compression. The physical
and numerical results agree well. Numerical results show that the stick-slip period is doubled
if the loading velocity is halved. Compared with the rate- and state-dependent law, the present
model is simple and can be implemented in FLAC-3D, capable of modelling the effects of
creep, seepage and temperature. For complex faults under complex loading conditions, the
proposed model can be used to identify faults prone to sliding or those with large stress
drops, and to investigate active sequences of faults and their interactions. Advantages and
disadvantages of square or quadrilateral meshes are also addressed, especially on the aspect
of stick-slip modelling. For a bending or a single straight fault, a long stick-slip period and a
low load-carrying capacity are obtained for quadrilateral meshes, but not for square meshes.
Key words: Numerical solutions; Seismic cycle; Friction; Rheology and friction of fault
zones; Continental tectonics: strike-slip and transform; Dynamics and mechanics of faulting.

gists. Many theoretical and numerical studies have been conducted


1 I N T RO D U C T I O N
to investigate various seismic problems using different constitutive
As a mechanism of earthquakes, the frictional stick-slip phe- models and numerical methods. Amontons’ law requires a constant
nomenon of faults has attracted much attention of many seismolo- frictional coefficient, and yet cannot consider the static friction and


C The Authors 2013. Published by Oxford University Press on behalf of The Royal Astronomical Society. 1023
1024 X. B. Wang, J. Ma and Y. S. Pan

the dependence upon the velocity of sliding (Lorig & Hobbs 1990). 2001; Hillers et al. 2006) results in complex seismic behaviours.
The static–kinetic friction law requires that the friction is static be- Beroza & Mikumo (1996) found that the short rise times observed
fore slip initiates, followed by a low kinetic friction once motion in an earthquake can be fully explained by a strongly heterogeneous
commences. However, the law always predicts an unstable motion spatial distribution of stress drop and strength excess, and that a self-
once slip starts. healing mechanism, such as a strongly velocity-dependent friction,
One constitutive relation, commonly used to describe slip pos- is not required.
sibly leading to an instability, is the slip-weakening law (Ida 1973; In this study, to model stick-slip phenomena of some typical
Day 1982; Andrews 1985; Shibazaki & Matsu’ura 1992; Harris faults, a frictional-hardening and frictional-softening model is pro-
& Day 1993, 1997; He 1995; Ide & Takeo 1997; Guattei & Spu- posed and implemented in a commercial continuum software devel-
dich 2000; Wolf et al. 2006; Ampuero & Ben-Zion 2008; Tinti oped by Itasca Corporation, Fast Lagrangian Analysis of Continua
et al. 2009). The law, in which the friction is only a function of in 3 Dimensions (FLAC-3D). The model can reproduce the peri-
slip distance, has been used to model various seismic phenomena. odic stress change, that is saw-tooth-like behaviour of stress, and
Because it inherently neglects the time and rate effects, it cannot consider interactions among faults. As a key ingredient, an incre-

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


reproduce more than one seismic cycle without resetting the fric- mental plastic shear strain in a stick-slip cycle is introduced, which
tion level when slip stops (Rice 1983). The slip-weakening law is increased from zero at the beginning of softening, and then is set
with initial slip-hardening was also proposed (Ohnaka & Yamashita to be zero upon reaching its maximum, to control the evolution of
1989; Ohnaka 1993). Bizzarri et al. (2001) discussed implications an internal frictional angle. Though no velocity is introduced into
of initial slip-hardening. the present model that is simpler than the rate- and state-dependent
Another constitutive relation, commonly used, is the rate- and law, a change in velocity can be described at the slip and stick stages
state-dependent law where the slip velocity and state variables en- because equations of motion are solved. At slip and stick stages, the
ter the constitutive relation. Different versions have been proposed same set of equations is used, and only the evolution of the internal
so far (Dieterich 1979, 1992, 1994; Ruina 1983; Kato & Tullis frictional angle is required to be different. Forty numerical tests are
2003; Kato et al. 2007). The law is believed to be a more com- performed in plane strain biaxial compression. Effects of loading
plete constitutive description of fault behaviours (Okubo 1989). To conditions (confining pressure and loading rate), fault width and
avoid different singularities for a locked fault (i.e. zero slip ve- fault geometry on stick-slip behaviours are investigated. The vali-
locity) as well as for very high slip velocities, numerous different dation of the proposed model is checked through a comparison with
regularizations have been suggested (Beeler et al. 1994; Marone a physical test in biaxial compression, and size-independent stress-
1998; Bizzarri et al. 2001). Even though, no existing friction law deformation curves are obtained through decreasing the loading
can describe all experimental data, and it is not known which law velocity for fine meshes.
is appropriate for describing seismic cycles (Kato & Tullis 2003).
Many modifications of the friction law lead to redundant analyt-
ical formulations (Bizzarri et al. 2001), and numerical solutions 2 I N T RO D U C T I O N T O F L AC - 3 D
are not easy to obtain (Miyatake 1992). In some models, a quasi-
static method during slow deformation is used, and then a dynamic FLAC-3D can be used to model mechanical behaviours, failure
method is triggered once an instability starts (Okubo 1989). The processes, and plastic flows of geological structures (Strayer &
abrupt switching from one scheme to another disrupts the natural Hudleston 1997; Mckinnon & De La Barra 1998; Erickson et al.
development of the instability, and the effects of this disruption on 2001). In FLAC-3D, properties of elements in different regions can
the further model response cannot be determined easily. To avoid be assigned to be different. Thus, joints and imperfections within
this difficulty, Lapusta et al. (2000) proposed an efficient and rigor- rock specimens can be modelled (Wang 2005; Wang 2007a,b, 2008;
ous numerical procedure for calculating the elastodynamic response Wang et al. 2009). Also, using a provided command, a normal
of a fault subjected to slow loading processes. Detailed comparisons (Gaussian) distribution for parameters of elements can be assigned,
between the slip-weakening law and the rate- and state-dependent or an arbitrary random distribution can be accomplished through
friction law have been made by Okubo (1989) and Bizzarri et al. users’ development (Cundall 1989; Fang & Harrison 2002; Wang
(2001). Bizzarri et al. (2001) emphasized that the rate- and state- & Pan 2008; Wang & Zhang 2009; Wang et al. 2010, 2012).
dependent friction law includes the most prominent features of the In FLAC-3D, equations of motion are solved for each node.
slip-weakening law. Okubo (1989) demonstrated that when a fric- Forces acting at a node i include two parts: one is the out-of-balance
tion law with high-speed cut-offs to the rate- and state-dependent force that is a resultant force of gravity, external force and elastic
term is used, the fault behaviour closely resembles slip weakening. force of surrounding elements; another is the local non-viscous
Some other models, such as various spring-block models, gran- damping force (Cundall 1989). The use of local damping is simpler
ular models (Mora & Place 1994; Abe et al. 2006) and cellular than Rayleigh damping, because it is unnecessary to specific a
automata (Bak & Tang 1989; Ito & Matsuzaki 1990), with simple frequency. Thus, damped dynamic equations can be written as
friction law or without, have been used to model complex seismic Mi ai = Fi + f i , (1)
phenomena. In the granular model of Mora & Place (1994), no
any frictional property is included. Doz & Riera (2000) used a lat- where Mi is the nodal mass, ai is the acceleration, Fi is the out-of-
tice model without friction. Abe et al. (2006) treated the fault as balance force and f i is the damping force.
a discrete arrangement of particles with the same frictional coeffi- In a time step or cycle of FLAC-3D, constitutive equations for
cient. The application of these models is difficult at a large scale. elements and equations of motion at grid points or nodes are exe-
Alternatively, continuum models are attractive and appropriate be- cuted one time. Using the Gauss divergence theorem, the velocity
cause many factors can be taken into account, such as complex at a node i at time t is related to the strain rate or strain increment
boundary conditions, loading conditions, fault geometries and ma- of elements. Moreover, the strain increment depends on the stress
terial models including heterogeneities. Inclusion of spatial het- increment in terms of a constitutive relation. Adding the stress in-
erogeneities in continuum models (Miyatake 1992; Bizzarri et al. crement to the old stress leads to a new stress. Application of the
Stick-slip behaviours of typical faults 1025

theorem of virtual work results in forces at the node i. Thus, a new


velocity at the next time t + t will be calculated using an explicit
central difference approach, where t is the length of a time step.
The procedure mentioned above is executed continuously until a
satisfactory result is reached.
For a dynamic analysis, the nodal mass, velocity and acceleration
are true or physical, and the dynamic time step t is extremely small
to provide a numerically stable solution. Thus, probably, a dynamic
problem will consume a very long computational time. The dynamic
time step in FLAC-3D is automatically calculated, and is related to
the P-wave speed and geometrical sizes of subzones (one element
or zone includes ten subzones).
In practice, a quasi-static analysis is appropriate for modelling a Figure 1. A frictional-hardening and frictional-softening model for the soft-

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


quasi-static experiment in laboratory, and a uniform unit time step ening and hardening behaviour of a failed element.
is adopted, that is t = 1. Thus, the nodal mass must be enlarged
enough to fulfil the numerical stability condition. The enlarged
nodal mass is usually called fictitious mass that is dependent on dergoes the strain-softening behaviour, ϕ approaches its initial value
some elastic constants and geometrical sizes of subzones. Corre- ϕm , the peak value. At the softening stage, as deformation proceeds,
spondingly, the nodal velocity and acceleration are not true, whose ϕ will decrease from ϕm to its residual value ϕ0 , causing an incre-
units are m time step−1 and m time step−2 , respectively. mental plastic shear strain γ p . When ϕ decreases to ϕ0 , softening
In FLAC-3D, some classical elastic and plastic constitutive mod- terminates, and then hardening begins, describing the phenomenon
els are available, among which two are selected in this study. One that an increase of γ p or slip distance causes a recovery in strength.
is the isotropic linear elastic model controlling the relation between At the hardening stage, γ p continuously increases because of the
incremental stress and elastic strain: further plastic deformation under displacement-controlled loading,
  and ϕ is assumed to be linearly increased with γ p :
2
σi j = 2Gεi j + K − G εkk δi j , (2)
3 γ p
ϕ = ϕ0 + p · (ϕm − ϕ0 ) . (4)
where σi j is the stress tensor, εi j is the strain tensor, G is the γm
shear modulus, K is the bulk modulus and δi j is the Kronecker sign. When γ p reaches its permitted maximum value γmp , ϕ will
G and K are related to two elastic parameters: elastic modulus E reach its maximum ϕm and the Mohr–Coulomb criterion will be
and Poisson’s ratio υ. satisfied soon or immediately under displacement-controlled load-
Another is the strain-softening or strain-hardening model for ing. Therefore, the next slip will occur. Before the next slip, γ p
failed elements, which comprises a failure criterion and a post- is enforced to return to zero. The implication of the present model
failure description that specifies the relation between incremental suggests that in a cycle the maximum accumulated plastic shear
stress and plastic strain. The Mohr–Coulomb criterion is selected strain γmp or the corresponding critical slip distance is constant.
as the failure criterion: Once γmp is reached, then the present cycle terminates, and the
 next will start. The zero γ p corresponds to the onset of softening,
σ1 = σ3 Nϕ + 2c Nϕ , (3)
 whereas the maximum γ p corresponds to the end of hardening.
where Nϕ = (1 + sin ϕ) (1 − sin ϕ), ϕ is the internal frictional an- Thus, the repeated cycles can be modelled. If the slip distance in a
gle, c is the cohesion, σ1 and σ3 are minor and major principal cycle is divided by the fault width, then γ p can be obtained. A too
stresses, respectively. The Mohr–Coulomb criterion comprises two low γmp can cause an advanced termination of slip, namely, ϕ has
shear strength parameters ϕ and c. These two parameters are permit- not decreased to ϕ0 . Apparently, stress drops and stick-slip periods
ted to change with plastic shear strain after an element fails in FLAC- will be shortened correspondingly under this condition.
3D. Thus, the strain-softening or strain-hardening behaviour can be Note that γ p does not enter the constitutive relation. γ p
modelled straightly without introducing special solving strategies. is set to be zero when ϕ reaches its maximum, whereas the ac-
cumulated plastic shear strain ε p = ε p of a failed element al-
ways increases as plastic deformation proceeds under displacement-
3 T H E F R I C T I O NA L - H A R D E N I N G A N D
controlled loading, as shown later. In each time step, the incremental
F R I C T I O NA L - S O F T E N I N G M O D E L
plastic shear strain ε p in the constitutive relation is very small,
If ϕ and c are assigned to be linearly or non-linearly decreased with corresponding to a small increment ϕ (Fig. 1). In a time step of
plastic shear strain before the residual stage, then the behaviour FLAC-3D, ε p of a zone is calculated as a volumetric average of
resembles slip-weakening because the slip distance along a fault is the second invariant of plastic shear strain increment vectors over
proportional to the plastic shear strain of the fault provided that the all subzones in the zone. These plastic shear strain increment vec-
fault is modelled as a continuum. Nevertheless, the slip distance in tors are determined by a non-associated flow rule specifying the
the slip-weakening law needs to be converted to the plastic shear direction of the plastic shear strain increment vector as that normal
strain given that fault width. However, such model cannot reproduce to the potential surface. In a stick-slip cycle, γ p ranging from 0
the seismic phenomena more than one stick-slip cycle. To avoid to γmp can be divided into several small ε p . γ p is increased
this difficulty, herein, a frictional-hardening and frictional-softening from zero at the beginning of softening, corresponding to a specific
model is proposed (Fig. 1), including two parts: linear softening and ϕ. At the entire softening or hardening stage, calculations need to
hardening parts, describing slip and stick processes, respectively. consume many time steps. In each time step, ε p will be created.
For the sake of simplicity, only one shear strength parameter ϕ is Summing all of ε p from the onset of softening in a cycle leads to
permitted to change after an element fails. When an element just un- the present γ p .
1026 X. B. Wang, J. Ma and Y. S. Pan

Both similarities and differences exist between the proposed crease of the size of elements leads to a relatively smooth fault
model and the rate- and state-dependent friction law. No veloc- surface together with a decrease of computational efficiency. Also,
ity is introduced in the present model, and yet a change in velocity formation of new failure zones or faults can be modelled effec-
can be described during the stick or slip process owing to the use tively (Fang & Harrison 2002; Wang et al. 2010, 2012). Quadri-
of equations of motion in FLAC-3D. That is to say, the velocity is lateral meshes can be only applicable for relatively simple fault
treated as a response, not as a constitutive parameter. systems with smooth surfaces, such as an inclined straight fault or
At the slip or softening stage, a decrease of ϕ of elements on a bending fault, whereas new faults are difficult to model because
a fault leads to a decrease of the load-carrying capacity of the of different mesh sizes at different positions. If a model is divided
fault, whereas the stress in the elastic body outside the fault cannot into quadrilateral meshes, then the number of elements is possibly
decrease immediately. Thus, a non-equilibrium in stress between fewer, improving computational efficiency. Accuracy increases as
fault and elastic body will be created, leading to an increase of element aspect ratios tend to unit. Therefore, when two or several
the nodal out-of-balance force on the fault. According to equations intersecting faults are included in a model, it is not easy to design
of motion, these nodes will move in a rapid manner until a stick the mesh or grid for accurate solutions. In addition, once the size of

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


stage begins. Therefore, a decrease of ϕ of elements can result in an square elements is determined, then the minimum fault width will
increase of nodal velocity. On the contrary, at the stick or hardening be specific. Thus, only wide faults can be modelled by use of square
stage, an increase of ϕ leads to a progressive decrease of nodal elements unless the size of meshes is further decreased, leading to
velocity. Therefore, the nodal velocity on a fault cannot remain a decrease of computational efficiency. However, if we use quadri-
constant in a cycle. A change in velocity stems from a change in ϕ. lateral meshes to model a straight fault, then its width can be small
In summary, the present model is based on a classical elastoplastic enough, and it can change continuously. In dynamic analyses of
continuum model where an internal frictional angle is used instead FLAC-3D, the length of time steps depends on the minimum size
of a slip distance. As a key ingredient, an incremental plastic shear of all elements in a model to ensure the numerical stability. Thus, if
strain in a stick-slip cycle is introduced, which is increased from some narrow quadrilateral meshes are included in the model, then
zero, and then is set to be zero upon reaching its maximum, to control the computational efficiency will be greatly decreased. Therefore,
the evolution of the internal frictional angle, analogous to the slip two kinds of meshes have advantages and disadvantages. If we want
distance in a cycle in the rate- and state-dependent friction law to to study the influence of fault width, quadrilateral meshes have
some extent. Even though no velocity is introduced into the present apparent advantages. If we want to model complex faults where
model, the nodal velocity changes at the slip and stick stages because new faults are possibly generated under quasi-statical and dynamic
equations of motion are solved. At the slip or stick stage, the same loading conditions, square meshes can be more appropriate.
set of equations, including constitutive, geometrical, and dynamic Displacement-controlled loading is exerted in σ1 direction
equations, is used, and only the evolution of one shear strength through a velocity on two loading ends of a specimen to shorten the
parameter, that is internal frictional angle, is required to be different. computational time, different from the usual experimental condi-
The present model is written by use of a built-in programming tion. Thus, the needed time or number of time steps will be halved
language, FISH. in comparison with the case of loading at one end. In numerical
calculations, quasi-static loading is adopted in schemes 1–33 and
35–40, to save the computational time, as will be mentioned later.
In scheme 34, dynamic loading is conducted. The average compres-
4 E X A M P L E S A N D R E S U LT S
sive stress at two ends can be calculated by use of a FISH function,
provided in the FLAC-3D’s user manual. At two lateral surfaces, a
4.1 Numerical specimens
confining pressure σ3 is applied.
Forty numerical tests (Table 1) are performed in plane strain biaxial A specimen is composed of two parts: fault and rock block.
compression and in small or large strain mode. The plane strain Correspondingly, elements include fault and rock elements. At the
condition is accomplished by application of roller boundaries on elastic stage, for two kinds of elements, the constitutive relations
the front and back surfaces of a 3-D specimen. These numerical and parameters are identical, and the isotropic linear elastic model
specimens are shown in Fig. 2 except for scheme 40. In schemes 1– (eq. 2) is used: E = 26.5 GPa and υ = 0.21. Two faults can form a
25, each specimen is divided into square elements (cubic elements bending or echelon fault and two intersecting faults. In this study,
from a spatial viewing angle), whereas in schemes 26–39, elements we assume that fault elements will fail once the Mohr–Coulomb
are quadrilateral (hexahedral from a spatial viewing angle). Thus, criterion (eq. 3) is satisfied, whereas rock elements remain elastic
faults are modelled by solid elements. Faults can also be modelled except for a part of rock elements in echelon faults. In schemes
by use of interface elements without thickness (Rutqvist et al. 2007). 1–34 and 35, the cohesion of a failed element remains unchanged,
In schemes 1–34 and 36–40, calculations are conducted in a small that is 0.2 and 2.8 MPa, respectively.
strain mode, whereas in scheme 35, a large strain mode is triggered. In some stick-slip tests, some faults are prescribed in rock spec-
In schemes 1–25, uniform square meshes are adopted. Inevitably, imens and fractures of rock blocks are seldom observed (Sobolev
zigzag fault surfaces will be generated. In the present numerical et al. 1996; Miao et al. 2010; Ma et al. 2012). However, for rock
specimens, interconnected elements, whose centres fall into be- specimens including echelon faults, the case is quite different. For
tween two parallel linear segments, are used to approximately model echelon faults, theoretical results (Segall & Pollard 1980; Zachari-
a fault or gouge zone with non-zero width (Marone & Kilgore 1993; asen & Sieh 1995), experimental results (Ma et al. 1986, 2007,
Kato et al. 1994), as accomplished through a written FISH function 2008, 2010; Jiang et al. 2002), numerical results (Wang et al. 2010,
(Wang 2005, 2007a). The fault width is determined by the spac- 2012) and field observations (Tchalenko & Ambraseys 1970; Sibson
ing of the two linear segments. In schemes 26–35, non-uniform 1985) show that new failed regions or cracks will form in the jog or
quadrilateral meshes are used to form smooth fault surfaces. outside during the loading process, providing necessary conditions
Square meshes can model complex fault systems including sev- for occurrence of stick-slip behaviours along faults. Therefore, in
eral inclined or straight faults whose surfaces are uneven. A de- schemes 24 and 25, some elements in rock blocks are permitted
Table 1. Numerical tests and related parameters.
p
Scheme Type Height (m) Width (m) Number of elements Loading velocity Fault width (m) γm σ3 (MPa) ϕ m (◦ ) ϕ 0 (◦ ) Other features
1 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 30◦
2 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 38◦
3 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 45◦
4 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 53◦
5 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 60◦
6 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 2.5 20 10 α = 45◦
7 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 5.5 20 10 α = 45◦
8 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 10.5 20 10 α = 45◦
9 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 2 × 10−4 0.5 20 10 α = 45◦
10 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 6 × 10−4 0.5 20 10 α = 45◦
11 Straight 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.0035 6 × 10−4 0.5 20 10 α = 45◦
12 Intersecting 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.0035 6 × 10−4 0.5 20 10 α 1 = α 2 = 45◦
13 Intersecting 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.0035 6 × 10−4 0.5 20 10 α 1 = 45◦ , α 2 = 30◦
14 Intersecting 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.0035 6 × 10−4 0.5 20 10 α 1 = 45◦ , α 2 = 60◦
15 Intersecting 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 6 × 10−4 0.5 20 10 α 1 = 45◦ , α 2 = 60◦
16 Intersecting 0.1 0.05 40 × 80 1 × 10−9 m time step−1 0.00175 6 × 10−4 0.5 20 10 α 1 = 45◦ , α 2 = 60◦
17 Intersecting 0.1 0.05 40 × 80 5 × 10−10 m time step−1 0.00175 6 × 10−4 0.5 20 10 α 1 = 45◦ , α 2 = 60◦
18 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = β 2 = 45◦
19 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 38◦
20 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 53◦
21 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 57◦
22 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 60◦
23 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 50◦
24 Echelon 0.1 0.1 40 × 40 1.25 × 10−10 m time step−1 0.0035 1.2 × 10−3 0.5 20 5 Compressive
25 Echelon 0.98 0.1 40 × 32 1.25 × 10−10 m time step−1 0.0035 1.2 × 10−3 0.5 20 5 Extensional
26 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.007 4 × 10−4 0.5 20 10 α = 45◦
27 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.0035 4 × 10−4 0.5 20 10 α = 45◦
28 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.00175 4 × 10−4 0.5 20 10 α = 45◦
29 Straight 0.1 0.05 10 × 20 1 × 10−9 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦
30 Straight 0.1 0.05 10 × 39 1 × 10−9 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦
31 Straight 0.1 0.05 10 × 39 5 × 10−10 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦
32 Straight 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦
33 Straight 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦
34 Straight 0.1 0.05 10 × 20 5.7 × 10−3 m s−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦ , Dynamic
35 Straight 0.1 0.05 20 × 40 1 × 10−9 m time step−1 0.00175 1.6 × 10−3 0.5 20 10 α = 45◦ , large strain
36 Straight 0.1 0.05 20 × 40 5 × 10−10 m time step−1 0.0035 6 × 10−4 0.5 20 10 β 1 = β 2 = 45◦
37 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 - 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 38◦
38 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 - 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 53◦
39 Bending 0.1 0.05 20 × 40 5 × 10−10 m time step−1 - 6 × 10−4 0.5 20 10 β 1 = 45◦ , β 2 = 60◦
40 Straight 0.3 0.15 10 × 20 (1–2) × 10−9 m time step−1 0.003 1.88 × 10−2 5 31 8 α = 60◦
Stick-slip behaviours of typical faults
1027

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


1028 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 2. Numerical specimens in plane strain biaxial compression. (a) scheme 1; (b) scheme 2; (c) schemes 3 and 6–10; (d) scheme 4; (e) scheme 5; (f)
schemes 11 and 18; (g) scheme 12; (h) scheme 13; (i) scheme 14; (j) scheme 15; (k) schemes 16 and 17; (l) scheme 19; (m) scheme 20; (n) scheme 21; (o)
scheme 22; (p) scheme 23; (q) scheme 24; (r) scheme 25; (s) scheme 26; (t) scheme 27; (u) schemes 28, 29 and 34; (v) schemes 30 and 31; (w) schemes 32,
33 and 35; (x) scheme 36; (y) scheme 37; (z) scheme 38; (z1) scheme 39.

to fail. For the compressive echelon fault (Fig. 2q), a column of of 45o with the horizontal direction, whereas the right fault angle
elements can fail, going through two fault tips, and extending to is from 38o to 60o . For echelon faults (Figs 2q–r), the orientation
two loading ends. For the extensional echelon fault (Fig. 2r), a few of each fault is 45o with the horizontal direction, and the fault
elements can fail to form a short failed region, linking two fault tips. overlapping is about 0.01225 m, equal to the spacing.
For failed rock elements and fault elements, there is no difference
in mechanical properties.
For the single straight fault (Figs 2a–f and s–w), the angle between
4.2 Numerical results of an inclined fault using
fault and horizontal direction is in the range of 30o –60o . For two
square elements
intersecting faults (Figs 2g–k), the angle between one fault and
horizontal direction is 45o , whereas another ranges from 30o to 60o . First, we investigate the influence of the angle α between fault
For the bending fault (Figs 2l–p), the left fault is oriented at an angle and horizontal direction on the evolution of stress with the number
Stick-slip behaviours of typical faults 1029

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 2. (Continued.)

of time steps (Fig. 3a). In schemes 1–5, the angle α is 30◦ , 38◦ , A detailed examination of change in stress shows that for the
45◦ , 53◦ and 60◦ , respectively. The number of time steps is pro- former stick-slip events, rapid stress drops are found, leading to
portional to the strain ε1 in σ1 direction: ε1 = 2v 0 t0 /L, where v 0 acute local stress peaks, that is acute saw-tooth-like curves (Fig. 3b).
is the loading velocity, L is the specimen’s length in σ1 direction In scheme 3, the acutest stress peak and lowest stress corresponding
and t0 is the number of time steps. In schemes 1 and 2, stress to the onset of slip are found. In addition, the stress at the beginning
drops are seldom found. At the same number of time steps, stresses of slip or the stress drop decreases as deformation proceeds until a
in schemes 1 and 2 are higher than in schemes 3–5, suggesting constant stress level is reached (Fig. 3b). The reason for this special
a higher load-carrying capacity. This means that, slip is easy to phenomenon will be discussed later.
take place at a moderate α. However, slip is more difficult at a Specimens are subjected to different lateral pressures σ3 in
higher or a lower α, and occurs only at a higher stress. Moreover, schemes 3 and 6–8. Thus, the influence of confining pressure can be
in schemes 1 and 2, the duration between two adjacent stress drops investigated (Fig. 4). It is found that the stick-slip period becomes
is longer, reflecting a longer stick-slip period, especially in scheme longer at a higher σ3 , especially for the former stick-slip events.
1; and interestingly, the stress for the occurrence of slip is ele- Moreover, the stress drop becomes larger, and the first slip is later.
vated continuously. These characteristics cannot be observed for In scheme 8 (σ 3 = 10.5 MPa), for the former three events, the stress
schemes 3–5. at the beginning of slip increases as deformation proceeds, similar
1030 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 5. Evolution of stress with the number of time steps for a specimen
p
with a single straight fault. Maximum incremental plastic shear strains γm
are different in different specimens.

stick-slip phenomenon at a later stage of deformation. A too small


γmp that is smaller than γ p corresponding to ϕ0 (Fig. 1), can
cause the slip behaviour to end beforehand (ϕ has not decreased
to ϕ0 before the stick stage), and then ϕ reaches ϕm immediately.
Therefore, a small stress drop will be expected, and the stick-slip
period will be shortened correspondingly.
We present the evolution of the accumulated plastic shear strain
ε p with the number of time steps in scheme 11 (Fig. 6). The deforma-
tional meshes with a 30 times exaggeration of nodal displacements
are also presented. It is found that elements on the fault have the
nearly same ε p , whereas ε p is zero outside the fault because the fail-
ure of rock blocks is not permitted. ε p is found to always increase
under displacement-controlled loading, as mentioned above.
Figure 3. Evolution of stress with the number of time steps for a specimen In FLAC-3D, the small strain mode assumes small displacements,
with a single straight fault. Angles α between fault and horizontal direction
displacement gradients, and rotations in each cycle or time step.
are different in different specimens.
Thus, nodal coordinates are not updated, and stress rotation cor-
rections are not taken into account in each time step. During the
stick-slip simulation, if the number of time steps is large enough,
elements on faults can reach relatively large strains (Fig. 6). Thus,
this founding is inconsistent with the assumption of small strain.
Seemingly, considering large strain mode is necessary. We will com-
pare results under different deformational modes using quadrilateral
meshes later.

4.3 Numerical results of two intersecting faults using


square elements
An increase of the number of faults alters the fault structure, and in-
teractions among faults could be apparent. For a single fault (scheme
11), the curve is relatively simple in a cycle: the stress first in-
creases and then decreases (Fig. 7). However, for two conjugate
faults (scheme 12), about three small stress drops in a cycle are ob-
Figure 4. Evolution of stress with the number of time steps for a specimen served at the stick stage, because of interactions among two faults
with a single straight fault. Confining pressures σ3 are different in different with a moderate angle α, and then only a large stress drop is found
specimens. at the following slip stage. Moreover, the stress at the beginning
of slip is low in scheme 12 because two faults greatly decrease the
to results in schemes 1 and 2 (Fig. 3a) and then the local stress peak load-carrying capacity of the specimen. The period in scheme 12 is
decreases continuously. obviously longer than in scheme 11.
The maximum incremental plastic shear strain γmp significantly We observe the effect of fault geometry (Fig. 8). In schemes
affects post-failure deformation characteristics. The strains γmp 12–14, the left fault angle is fixed as α 1 = 45o , whereas right fault
are prescribed to be different in schemes 3, 9 and 10 (Fig. 5). The angles α 2 are different. In scheme 13 (α 2 = 30o ), a regular saw-
stick-slip period is shorter, the stress drop is smaller, and the first tooth-like curve is found in a cycle, above other two curves. The
event is earlier at a lower γmp . A lower γmp leads to a less obvious curve in scheme 12 (α 2 = 45o ) is given again for a comparison, with
Stick-slip behaviours of typical faults 1031

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 6. Distribution of the accumulated plastic shear strain ε p in scheme 11. (a–c) the number of time steps is 10 000, 30 000 and 50 000, respectively.

Figure 7. Evolution of stress with the number of time steps for specimens Figure 9. Evolution of stress with the number of time steps for a specimen
with different numbers of faults. with two intersecting faults. Number of elements and fault width are different
in different specimens.

specimen. It is deduced that one fault is difficult to slip so that the


influence on another fault is negligible. To verify this, a detailed
comparison shows that only a few differences exist for the former
stick-slip events in schemes 13 and 11 (Figs 7–8). Furthermore, as
deformation proceeds the difference can be neglected. This suggests
that the fault oriented at an angle of α 2 = 30o with the horizontal
direction is difficult to slip, as also found in Fig. 3(a). A large
majority of slip events take place on a 45o fault.
Use of square elements leads to a wide fault unless the size of
square elements is further decreased. We focus on the effect of
minimum fault width controlled by the size of square elements
(Fig. 9). In schemes 14–16, the number of elements is 800, 200 and
3200, respectively; that is to say, the mesh size is 0.005, 0.0025 and
Figure 8. Evolution of stress with the number of time steps for a specimen 0.00125 m, respectively. Seemingly, results are mesh-dependent or
with two intersecting faults. The left fault angle is fixed, whereas right fault mesh sensitive (De Borst 1989). It is noted that fault widths are dif-
angles α 2 are different in different specimens. ferent in schemes 14–16: 0.0035, 0.007 and 0.00175 m, respectively.
Therefore, these results are not considered to be the sole influence
a larger stress drop and a longer stick-slip period. The lowest stress of mesh size, including the remarkable influence of fault width. The
corresponding to the onset of slip is found in scheme 12, whereas sole influence of fault width and size-independent results will be
the highest is found in scheme 13. In scheme 14 (α 2 = 60o ), a large presented later by use of quadrilateral meshes. A fine mesh and a
stress drop occurs after four small stress drops, and the stress for the narrow fault in scheme 16 lead to a complex curve, a short period
occurrence of each small slip increases as deformation proceeds, as and a high stress at the beginning of slip. For a narrower fault, the
also found in scheme 12. The complex curves in schemes 12 and accumulated plastic shear strain in a cycle will be higher at the
14 are because of interactions among two faults. same displacement increment in σ1 direction or the same critical
Different from schemes 12 and 14, scheme 13 has a relatively slip distance along a fault so that γmp is reached easily, leading to
regular stick-slip behaviour though two faults are included in the a smaller stress drop and a shorter period. However, a coarse mesh
1032 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 10. Evolution of stress with the number of time steps for a specimen
with two intersecting faults. Number of elements and loading velocity v 0
are different in different specimens.
and a wide fault in scheme 15 result in a simple and regular stick-
slip curve with a low stress corresponding to the onset of slip. It is
known that the load-carrying capacity of a specimen is affected by
fault width (Wang 2007a), supporting the present numerical result
that a low strength is expected for the specimen with a wide fault.
The reason for the ideal stick-slip phenomenon for coarse meshes
will be explained as follows. For large elements, the propagation
of stress wave from faults to loading ends of a specimen is rapid.
Otherwise, during the propagating process, stress waves stemming
from different positions of faults can superimpose each other. Thus,
a relatively uniform stress can be reached at loading ends of the
specimen, which cannot reflect the true state of stress on faults.
An increase of loading velocity exerted on one loading end of a
specimen leads to an increase of the axial displacement or strain of Figure 11. Evolution of stress with the number of time steps for a specimen
the specimen in a time step. We present the effect of loading ve- with a bending fault. The left fault angle is fixed, whereas right fault angles
locity (Fig. 10). Specimens in schemes 14, 15 and 17 are subjected β2 are different in different specimens.
to different loading velocities. It can be found that higher veloci-
ties are necessary for coarser meshes to model a regular stick-slip To find the reason for the smooth curve in scheme 22, the evolu-
phenomenon. If the mesh is refined, then v 0 must be decreased. A tion of ϕ of elements on the left and right faults is monitored during
high velocity leads to an increase of the stress at the beginning of the loading process. These elements are marked in Fig. 2(o) with the
slip. The use of a low v 0 increases the computational time on two black colour. For a comparison, ϕ of elements on a 5o bending fault
aspects. On one hand, the number of elements increases. On the (scheme 23), common in the field (Segall & Pollard 1980; Barka
other hand, slow loading requires a number of time steps to reach & Kadinsky-Cade 1988; Aydin & Du 1995; Hemendra 1997; Kato
the same strain ε1 . et al. 1999; Ma et al. 2012), is also monitored. These results together
with the evolution of stress are shown in Fig. 12, and results in the
range of (2–3) × 104 time steps are shown in Fig. 13. The evolution
4.4 Numerical results of the bending fault using
of ϕ is very chaotic in scheme 22, lacking synchronization (Figs 12a
square elements
and 13a). For the element at the bending point, a slow change in ϕ is
A change in the angle between two fault segments of a bending found, suggesting a high deformational resistance. Ma et al. (2012)
fault alters the fault geometry. The influence of fault geometry is found from physical tests that the temperature change is complex in
shown (Fig. 11). In schemes 18–22, the left fault angle is set to the vicinity of the bending point, indicating the stress redistribution
be 45o , whereas angles between two fault segments are different. and fault sliding. Asynchronic evolution of ϕ on the left and right
At β2 = 38◦ (scheme 19), the stress approximately increases as faults is found to be responsible for the smooth curve without appar-
deformation proceeds, and no apparent stick-slip phenomenon is ent stress drops. However, the softening and hardening behaviours
found, similar to the result in scheme 2 (Fig. 3a). As we know, indeed take place for each element.
for the single straight fault oriented at an angle of 45o with the A distinct phenomenon can be observed in scheme 23 (Figs 12b
horizontal direction, slip is prone to occur (Fig. 3). After adding a and 13b): nearly all elements behave similarly except for those at
gentle 38o fault, a very straight curve is observed. This means that the bending point. For different elements, a nearly simultaneous
slip on a 45o fault is prohibited by the 38o fault. increase or decrease of ϕ is found. Synchronic softening and hard-
Interestingly, at β2 = 60◦ (scheme 22), a smooth and horizontal ening behaviours create large slip events. The element at the bending
curve without a periodic change in stress, resembling stable sliding, point shows a long period, suggesting a difficult deformation, simi-
is observed. When β2 decreases from 60o to 45o , the stick-slip lar to results (Figs 12a and 13a). Apparently, the difficult extent of
behaviour becomes more apparent; the peak stress that a specimen deformation is different at the bending point and at other positions of
can reach decreases; and the stress drop increases (Fig. 11b). faults. Also, on a fault, the difficult extent of deformation is different
Stick-slip behaviours of typical faults 1033

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 12. Evolution of internal frictional angle and stress of elements on Figure 13. Evolution of internal frictional angle and stress of elements
the bending fault (a) scheme 22 (b) scheme 23. on the bending fault [from (2–3) × 104 time steps] (a) scheme 22 (b)
scheme 23. In scheme 22, curves with red, pink, orange, green, brown, cyan,
at different segments. A detailed examination of the evolution of ϕ and blue colors are for elements 243# , 306# , 369# , 411# , 494# , 577# , and
reveals that elements far away from the bending point facilitate slip. 619# , respectively (Fig. 2o). In scheme 23, curves with the same colours
mentioned above are for elements 243# , 306# , 369# , 431# , 513# , 595# , and
698# , respectively (Fig. 2p).
4.5 Numerical results of echelon faults using
square elements
We compare stick-slip behaviours of two kinds of echelon faults
(Fig. 14). For the compressive one, the stick-slip occurs at a high
stress, and the evolution of stress in different cycles is relatively
complex and irregular. Experimental tests (Ma et al. 2008) show
that the stick-slip behaviour along an extensional echelon fault oc-
curs easily, whereas it is difficult along a compressive one because
of the prohibition of the compressive jog. Based on a heteroge-
neous strain-softening model, numerical results show that the load-
carrying capacity of the specimen with a compressive echelon fault
is higher than with the extensional one (Wang et al. 2012). These
evidences support the present numerical results.
We compare the compressive stress evolution on three faults, Figure 14. Evolution of stress with the number of time steps for specimens
including the left, right and middle faults, and at loading ends of with echelon faults.
specimens with two kinds of echelon faults (Figs 15a and 16a).
Failed rock elements between the left and right faults form a middle the specimen with an extensional echelon fault is lower than on the
fault. Close-up figures in a few cycles are also shown (Figs 15b and left or right fault; and that on the middle fault, an ideal saw-tooth-
16b). As shown in Fig. 15, the evolution of stress on the middle like curve is expected, different from that in Fig. 15(a). The simple
fault is complex during a stick-slip process, and the stress is the or regular evolution of stress and a low level of stress mean that slip
highest. The average stress on the left or right fault is lower than on is easy on the middle fault.
loading ends of the specimen with a compressive echelon fault. In Several small events take place at the stick stage in a cycle,
addition, the stress on the right fault is always higher than on the left whereas only one large event occurs at the slip stage (Figs 14b
fault, which is related to the slightly different numbers of elements and 15b), consistent with some experimental observations (Sobolev
on both faults. No apparent difference exists in stick-slip periods of et al. 1996; Miao et al. 2010; Ma et al. 2012). This phenomenon
both faults. Fig. 16 shows that the average stress at loading ends of can also be found for two intersecting faults (Figs 7 and 8).
1034 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 15. Evolution of stress with the number of time steps for a specimen
with a compressive echelon fault in scheme 24.

4.6 Sole influence of fault width using Figure 16. Evolution of stress with the number of time steps for a specimen
quadrilateral elements with an extensional echelon fault in scheme 25.

As mentioned above, use of square elements cannot ensure a con-


tinuous change in fault width, and the fault width depends on the
size of square elements. Herein, we demonstrate the sole influence
of fault width w by use of quadrilateral elements (Fig. 17). Also, the
result in scheme 3 is also provided for a comparison. Fault surfaces
are even in schemes 26–28 because of the use of quadrilateral ele-
ments. In schemes 3 and 26, the fault width is identical. In scheme
3, owing to the use of square elements, the fault surface is uneven.
A saw-tooth-like fault surface leads to high peak stress and stress
corresponding to the end of slip. However, two kinds of fault sur-
faces result in the nearly same stick-slip period when the fault width
is identical. Moreover, the influence of fault width on the stick-slip
period and stress drop is remarkable. A narrower fault leads to a
shorter stick-slip period and a smaller stress drop, similar to results
in Fig. 9. The result in scheme 28 is like a stable sliding. When the Figure 17. Evolution of stress with the number of time steps for a specimen
fault width is halved, the stick-slip period is also halved. In schemes with a single straight fault. Fault widths are different in different specimens.
26–28, the maximum incremental plastic shear strain γmp remains
constant. Thus, a wider fault means a higher critical slip distance, the number of time steps. In schemes 28 and 29 (Figs 17 and 18),
and the stick-slip period will be longer. a higher γmp leads to a longer stick-slip period, consistent with
results (Fig. 5). When γmp is increased by a factor of 4, the period
is quadrupled.
4.7 Size-independent results using quadrilateral elements
Compared with scheme 29, schemes 30 and 32 have finer meshes.
To obtain size-independent stress-deformation results, we focus If the number of elements outside the fault is doubled in scheme
on the evolution of stress with displacement on one loading end 29, then the mesh in scheme 30 will be generated. If the mesh in
in schemes 29–33 with different mesh sizes or loading velocities scheme 29 is refined one time in the horizontal direction, then the
(Fig. 18). The displacement is a product of the loading velocity and mesh in scheme 32 will be expected. Indeed, different meshes lead
Stick-slip behaviours of typical faults 1035

the required true loading velocity in scheme 34 can be calculated


as 5.7 × 10−3 m s−1 .
Schemes 29 and 34 have the same stick-slip period despite dif-
ferent loading modes (Figs 18 and 19a). It is found that the nodal
velocity for 221# node (Fig. 2u) rapidly rises during the slip process.
The reached maximum velocity (approaching 0.1 m s−1 ) is greatly
higher than at two loading ends of the specimen. At the follow-
ing stick stage, the velocity decreases in an oscillating manner until
this node again gains an extraordinarily high velocity at the next slip
stage. Interestingly, the evolution of the velocity is similar to famous
rate- and state-dependent friction laws to some extent. Therefore,
the present model can describe the change in velocity during stick
and slip processes although no rate- and state-dependent friction

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


law is included.
It is found that prominent rises in nodal velocities have occurred
Figure 18. Evolution of stress with displacement at one loading end of a before rapid stress drops, indicative of slip precursors and a few
specimen with a single straight fault. Mesh size and loading velocity are peaks can be observed (Figs 19b–d). There are some differences of
different in different specimens.
nodal velocities at different positions of the fault. For example, the
peak nodal velocity is high for the black curve (221# node is on
the left boundary of the specimen), whereas it is low for the purple
to different peak stresses and stress drops (Fig. 18). Seemingly, the curve (261# node is on the right boundary). Moreover, the rapid
present numerical results are size-dependent, lacking uniqueness. rise in velocity for 261# node is slightly delayed. Approximately,
However, in schemes 31 and 33, the size-dependency can be greatly velocities of these monitored nodes undergo the synchronic evolu-
decreased if the loading velocity exerted on two loading ends is tion, that is simultaneous increase and decay. This is related to the
halved. The reason for a finer mesh requiring a slower loading ve- relatively simple state of stress on the smooth fault surface, and no
locity to avoid non-uniqueness of numerical solutions is explained heterogeneity is introduced in the present model.
as follows. In a time step, the propagating distance of stress wave
will be the size of an element. When the number of elements in a
specimen is doubled, the propagating velocity of stress wave across 4.9 Results under different deformational modes using
elements or meshes will be halved if the exerted loading velocity quadrilateral elements
remains unchanged. Therefore, if we want to obtain a unique so-
lution for different meshes, the number of time steps needs to be As found before, with an increase of the number of stick-slip cycles,
increased for the specimen with fine meshes. If the size of elements the saw-tooth-like curve becomes less acute until its shape does not
is halved, then the number of time steps needs to be doubled. Cor- change (Figs 3b, 5 and 7–11). Note that these results are obtained
respondingly, the loading velocity needs to be halved to ensure the using square elements. For quadrilateral elements, the phenomenon
same propagating distance of stress wave for different mesh sizes, mentioned above still exists, whereas it is less obvious (Figs 17,
leading to an increase of the number of time steps or an increase of 18 and 19a). Therefore, changes in saw-tooth-like curves cannot be
the stick-slip period. only attributed to the influence of uneven fault surfaces.
One possible reason is that the inclusion of a small strain assump-
tion is not applicable for modeling stick-slip cycles at a late stage
of deformation. At the time, deformation of elements on a fault is
4.8 Results under dynamic loading using
large, as mentioned above. However, a detailed comparison of two
quadrilateral elements
results obtained using the small and large strain modes shows that
The velocity history of nodes at different positions of a fault can nearly no difference exists (Fig. 20). After introducing the large
demonstrate the stick-slip process of the fault. We present the evo- strain mode, changes in saw-tooth-like curves are still observed,
lution of stress and nodal velocity with time step in scheme 34 leading to a long computational time. Scheme 32, calculated under
(Fig. 19), focusing on three typical stick-slip cycles (Fig. 19b). Posi- a small strain mode, consumes about 8 min for 2 × 105 time steps,
tions of monitored nodes on the straight fault are shown in Fig. 2(u). whereas scheme 35, under a large strain mode, only consumes about
Scheme 34 is conducted under dynamic loading with a mass density 5 min.
of 2.7 × 103 kg m−3 and with an automatically calculated dynamic Though the shape of the saw-tooth-like curve is slightly differ-
time step of 1.763 × 10−7 s by FLAC-3D, different from schemes ent at different stages of stick-slip processes, the period does not
1–33 performed under quasi-static loading. In scheme 34, the ex- change approximately. Also, changes in stress drops and slopes
erted loading velocity at two ends of the specimen is true or physical of saw-tooth-like curves are not large at slip stages. Therefore,
velocity (m s−1 ). Whether specimens are loaded quasi-statically or changes in saw-tooth-like curves would not influence the qualita-
dynamically, the local damping is used. Nevertheless, damping co- tive understanding of stick-slip phenomenon. Perhaps, at a late stage
efficients are different. In quasi-static analyses above, the default of deformation, the synchronic change in level of stress for fault
value is 0.8, whereas in the present dynamic analysis, the coeffi- elements is worse than at an early stage of deformation. This leads
cient is set to be 0.1571 (0.05π , π is the critical damping), that is to a propagation of stress along a fault rather than a strictly simul-
5 per cent critical damping, which is a typical value for dynamic taneous increase or decrease of stress for different elements. Thus,
analyses. Schemes 34 and 29 have the same models and parameters, during strain-softening of a large majority of fault elements, a few
and the same number of time steps, that is 5 × 104 , is calculated to fault elements are undergoing strain-hardening, leading to a blunt
reach the same axial displacement or strain in σ1 direction. Thus, saw-tooth-like curve and a small stress drop.
1036 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 19. Evolution of stress and nodal velocity for a node (a), and that for some nodes during three typical cycles (b–d) with the number of time steps in
scheme 34.

form a unity. Three subgrids form the left part of the mesh, whereas
4.10 Numerical results of the bending fault using
the other three subgrids form the right part. These two parts have
quadrilateral elements
the same size. The left fault angle with the horizontal direction is
To generate a smooth fault surface, herein, the bending fault and fixed as 45o , and the width is 0.0035 m. When the right fault angle
elastic bodies outside the fault are modelled by quadrilateral ele- is larger or smaller than 45o , the width is smaller or larger than
ments. Thus, the effect of uneven fault surface on the evolution of 0.0035 m, respectively.
stress can be excluded. We focus on the influence of fault geometry Regular saw-tooth-like curves are found in scheme 36, but not
(Fig. 21). Schemes 36–39 have the same parameters and computa- in schemes 37–39. When the angle between two fault segments is
tional conditions as schemes 18–20 and 22, respectively, except for higher or lower, the stress corresponding to the beginning of slip
the fault width and mesh shape. In schemes 36–39, widths of left is higher, and the stress drop is smaller. These characteristics are
and right faults are not completely identical, which is caused by the basically identical to results in schemes 18–20 and 22 (Fig. 11).
present mesh generation method. For any mesh in schemes 36–39, Compared with results in schemes 18–20 and 22 (Fig. 11), results
six subgrids are first generated, and then are attached together to in schemes 37–39 are more complex, which is related to slightly
Stick-slip behaviours of typical faults 1037

different widths of two fault segments. In addition, the use of quadri-


lateral elements leads to a low load-carrying capacity and a long
stick-slip period, whereas square elements predict a high stress cor-
responding to the onset of slip and a short period, consistent with
results for a single straight fault (Fig. 17). Using provided com-
mands in FLAC-3D, at present, it is difficult to establish some
typical fault systems with smooth fault surfaces, such as a bending
fault with the same width at different positions, an echelon fault
and two intersecting faults. Alternatively, using square or cubic
elements, some typical or complex fault systems can be created
through writing FISH functions, providing a chance to investigate
stick-slip behaviours and interactions among faults. However, we
must acknowledge some disadvantages of square meshes, such as

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


inevitable uneven fault surfaces and a low computational efficiency.

Figure 20. Evolution of stress with the number of time steps for a specimen
with a single straight fault.
4.11 Experimental verification using quadrilateral
elements
Under biaxial compression, Miao et al. (2010) conducted stick-slip
experiments of cuboid rock specimens, each including an inclined
straight fault. The height, width and thickness of each specimen
(Fig. 22a) are 300, 200 and 50 mm, respectively. The fault with a
width of 3 mm is modelled with anhydrite, and the weight percent-
age of gypsum and water is 3:1. The fault surface is milled with
400# cumbersome. The applied confining pressures (σ3 ) are in the
range of 5–20 MPa on two lateral surfaces of specimens whose
area is 300 mm × 50 mm. Axially displacement-controlled load-
ing is quasi-statically conducted in σ1 direction at one end with a
low compressive velocity ν of 0.1–13.5 µm s−1 , whereas the axial
displacement of another end remains zero.
A part of a typical principal stress difference-time curve is shown
Figure 21. Evolution of stress with the number of time steps for a specimen in Fig. 22(b) with σ3 = 5 MPa and ν = 0.05–0.1 µm s−1 . It is found
with a bending fault. Angles between two fault segments are different in that the loading velocity influences the stick-slip period becoming
different specimens. longer with a decrease of loading velocity. The peak stress and
stress corresponding to the end of slip do not change greatly. At

Figure 22. Tested specimen with a single straight fault (a) and principal stress difference-time curves at different loading velocities v (b) from Miao et al.
(2010).
1038 X. B. Wang, J. Ma and Y. S. Pan

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


Figure 23. Numerical specimen with a single straight fault (a) and evolution of principal stress difference with the number of time steps at different loading
velocities v 0 (b) in scheme 40.

ν = 0.1 µm s−1 , six short slip-stick periods undergo about 2900 s, For the numerical specimen, the axial displacement increment
and an average period of 483 s is expected for each stick-slip. At u a is
ν = 0.05 µm s−1 , two long periods consume 2436 s, and an average
u a = 2v0 t0 , (6)
period is 1218 s.
In the following, the mentioned experiment above will be numer- where v 0 is the used velocity under quasi-static loading (m time
ically modelled. In tests, the fault is only 3 mm in width. Therefore, step−1 ), or called fictitious velocity, and t0 is the number of time
it is inconvenient to use square elements to model the narrow fault steps. To avoid oscillating numerical results, a low enough fictitious
unless the size of square elements is set to be small enough. Thus, a velocity is needed, such as in the range of 10−10 –10−8 m time
very long computational time will be expected because smaller ele- step−1 . To model experimental results for v = 0.1 and 0.05 µm s−1 ,
ments require lower loading velocities, as mentioned in section 4.7. fictitious velocities are selected as v 0 = 2 × 10−9 and 1 × 10−9 m
For the specimen with an inclined fault, that is simple fault sys- time step−1 , respectively.
tem, it is convenient to model the fault using quadrilateral elements. We let u a = u a . Thus, we can obtain the needed number of
Outside the fault, the elastic body is also divided into the same kind time steps under quasi-static loading. When v 0 = 2 × 10−9 m time
of elements. It is noted that four corners of the tested specimen step−1 , t0 = 7.3 × 104 time steps, corresponding to six short periods.
are cut-off to avoid intervention between loading platens (Fig. 22a). Thus, we can deduce that one time step corresponds to 0.04 s, or 1 s
Therefore, to accurately model the shape of the specimen, a few corresponds to 25 time steps.
quadrilateral elements must be deleted. Seemingly, it is unneces- Calculations include two stages in scheme 40: a hydrostatic stress
sary to do so. Alternatively, the width of the numerical specimen loading and then a displacement-controlled loading, similar to phys-
is decreased to one half of the height. That is to say, only a part ical tests above. It can be found that calculated and experimental
of the tested specimen is numerically modelled, which is in the results agree well (Figs 22b and 23b). For the same loading veloc-
dashed rectangle (Fig. 22a). Thus, the width of the numerical spec- ity, numerical saw-tooth-like curves are identical at different stick-
imen (Fig. 23a) is only 150 mm. The fault angle with the horizontal slip cycles because of few numbers of stick-slip cycles. Calculated
direction is 60o , consistent with the experimental condition. results, including the peak stress, minimum stress and stick-slip
The tested and numerical specimens have the same height. The period approach experimental observations. Strictly, the period is
axial displacement increment is mainly contributed from slip along doubled as the loading velocity is halved. This is slightly different
the fault because of the negligible deformation outside the fault. from experimental results shown in Fig. 22(b). Probably, anhydrite
Therefore, to ensure that each stick-slip causes the same slip dis- is subjected to certain damages during loading, leading to slightly
tance for two kinds of specimens, we only need to require that two different stick-slip periods even though the loading velocity is iden-
specimens undergo the same axial displacement increment during tical. The present calculation is ideal, and certain damages or other
each stick-slip process. factors, such as heat because of friction, have not been considered.
For the tested specimen, the axial displacement increment u a Therefore, if the loading velocity is decreased, then the stick-slip
is period will be increased because the accumulated process of the
plastic shear strain on the fault is delayed.
u a = vt, (5) In the following, the needed time or number of time steps will
be assessed if the dynamic loading mode is triggered to model the
where v is the actual or physical velocity (m s−1 ), and t is the time actual loading process in laboratory. Using the same computational
interval (s). At v = 0.1 µm s−1 , t = 2900 s, corresponding to six model and parameters as scheme 40 except for the loading mode,
short stick-slip periods. the automatically calculated dynamic time step is 1.267 × 10−7 s
Stick-slip behaviours of typical faults 1039

by FLAC-3D with a true loading velocity of v = 0.1 µm s−1 . So for the compressive echelon fault, synchronic stick-slips are slightly
small time step is to ensure the numerical stability. Thus, six short bad, and more complex stress evolution in a cycle is found because
stick-slip periods (about 2900 s) in laboratory will require 2.3×1010 of the prohibition of the compressive jog.
time steps. However, under quasi-static loading, we only need 7.3 × FLAC-3D can consider the influence of creep, seepage, and tem-
104 time steps, and the needed time is less than 8 min using an or- perature on mechanical behaviours of geomaterials. These features
dinary personal computer. Thus, if we select the dynamic loading can be involved in the present model without any difficulties. In
mode, at least 7.4 × 104 min (i.e. 1.8 × 103 d) are needed to run, the future, the proposed model can be used to model stick-slip be-
an unacceptable task. As can be found from Fig. 23(b), numerical haviours and interactions among faults of actual sizes under com-
results for v 0 = 2 × 10−9 and 1 × 10−9 m time step−1 are smooth plex loading conditions. Among several faults, active sequences,
enough. Therefore, quasi-static loading is a feasible selection with- stick-slip periods, and stress drops can be determined numeri-
out affecting the accuracy and efficiency. cally. Even, faults prone to slip, and those with large stress drops
(stick-slips) or without (stable sliding), because of the synchronic
or asynchronic softening and hardening of elements, respectively,

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


can be identified numerically. Also, the calibration of the model
5 C O N C LU S I O N S
should be emphasized. These further investigations are worth con-
For modelling the periodic stress change at loading ends of rock ducting, especially important for fault mechanics and earthquake
specimens with one to three faults in plane strain biaxial compres- predictions.
sion, a frictional-hardening and frictional-softening elastoplastic
continuous model is proposed and implemented in FLAC-3D. The
model includes a linear softening part and then a linear hardening
AC K N OW L E D G E M E N T S
part, to model slip and stick behaviours, respectively. A key feature
of the model is the inclusion of an incremental plastic shear strain We are grateful to two anonymous reviewers for their helpful com-
in a stick-slip cycle, which is set to be zero upon reaching its maxi- ments and suggestions in improving the manuscript. We also thank
mum at the end of stick. This implies that the permitted maximum the Editor, Prof. Jörg Renner, for language polishing and sugges-
plastic shear strain or critical slip distance in a cycle is constant. At tions for correction. This work was supported by the National Natu-
the slip stage in a cycle, a decrease of an internal frictional angle ral Science Foundation of China (50974069), the open research fund
leads to an increase of the incremental plastic shear strain, whereas (LED2010B04) of State Key Laboratory of Earthquake Dynamics,
at the following stick stage, the angle is updated according to the Institute of Geology, China Earthquake Administration, and the Pro-
present incremental plastic shear strain. Although, no velocity is gram for Liaoning Excellent Talents in University (LJQ2011030).
introduced in the present model, the nodal velocities change in a
cycle because of the use of equations of motion. Using the pro-
posed model, stick-slip behaviours of 40 numerical specimens are
REFERENCES
modelled quasi-statically or dynamically, and the proposed model is
validated through a physical stick-slip test in biaxial compression. Abe, S., Latham, S. & Mora, P., 2006. Dynamic rupture in a 3-D particle-
For a single straight fault or two intersecting faults, the effects based simulation of a rough planar fault, Pure appl. Geophys., 163, 1881–
1892.
of parameters on the evolution of stress acting on loading ends are
Ampuero, J.P. & Ben-Zion, Y., 2008. Cracks, pulses and macroscopic
investigated, such as fault width, loading velocity and maximum in-
asymmetry of dynamic rupture on a bimaterial interface with velocity-
cremental plastic shear strain in a cycle. To obtain size-independent weakening friction, Geophys. J. Int., 173, 674–692.
stress-deformation curves, a low velocity is suggested for a fine Andrews, D.J., 1985. Dynamic plane-strain shear rupture with a slip-
mesh to ensure the same propagating distance of stress wave. Also, weakening friction law calculated by a boundary integral method, Bull.
the maximum incremental plastic shear strain in a cycle is found seism. Soc. Am., 75, 1–21.
to affect the stick-slip period and stress drop. For two intersecting Aydin, A. & Du, Y.J., 1995. Surface rupture at a fault bend: the 28 June
faults, the attention is paid on the effects of the fault geometry, and 1992 Landers, California earthquake, Bull. seism. Soc. Am., 85, 111–128
a few small slip events are observed at the stick stage because of Bak, P. & Tang, C., 1989. Earthquakes as self-organized critical phenomena,
interactions among faults, whereas only one large event is found at J. geophys. Res., 94, 15635–15637.
Barka, A.A. & Kadinsky-Cade, K., 1988. Strike-slip fault geometry in
the slip stage. A similar phenomenon is found for echelon faults.
Turkey and its influence on earthquake activity, Tectonics, 7, 663–684.
For bending faults, a small stress drop is observed at large an-
Beeler, N.M., Tullis, T.E. & Weeks, J.D., 1994. The roles of time and dis-
gles between two fault segments. The reason for this is attributed to placement in the evolution effect in rock friction, Geophys. Res. Lett., 21,
the asynchronic softening and hardening of fault elements, as con- 1987–1990.
firmed through monitoring evolution of the internal frictional angle. Beroza, G.C. & Mikumo, T., 1996. Short slip duration in dynamic rupture in
However, for a usual 5◦ bending fault, the synchronic evolution is the presence of heterogeneous fault properties, J. geophys. Res., 101(B10),
expected, creating a periodic change in stress. For a bending or a 22 449–22 460.
single straight fault, the use of quadrilateral elements leads to a low Bizzarri, A., Cocco, M., Andrews, D.J. & Boschi, E., 2001. Solving the
load-carrying capacity and a long stick-slip period, whereas square dynamic rupture problem with different numerical approaches and con-
elements predict a high stress corresponding to the onset of slip and stitutive laws, Geophys. J. Int., 144, 656–678.
Cundall, P.A., 1989. Numerical experiments on localization in frictional
a short period.
material, Ing.-Arch., 59, 148–159.
For two typical echelon faults, stick-slip behaviours of different
Day, S.M., 1982. Three-dimensional simulation of spontaneous rupture: the
faults, including the left, right, and middle faults, are compared. effect of nonuniform prestress, Bull. seism. Soc. Am., 72, 1881–1902.
New failed regions outside faults are permitted to appear to provide De Borst, R., 1989. Numerical methods for bifurcation analysis in geome-
necessary conditions for fault sliding. For the extensional echelon chanics, Ing.-Arch., 59, 160–174.
fault, stick-slips on the left and right faults are synchronic, suggest- Dieterich, J.H., 1979. Modeling of rock friction, 1: experimental results and
ing that the influence of the extensional jog is negligible. However, constitutive equations, J. geophys. Res., 84, 2161–2168.
1040 X. B. Wang, J. Ma and Y. S. Pan

Dieterich, J.H., 1992. Earthquake nucleation on faults with rate- and state- field during stick-slip instability on a 5o bending fault, Sci. China Earth
dependent strength, Tectonophysics, 211, 115–134. Sci., 55, 869–881.
Dieterich, J.H., 1994. A constitutive law for rate of earthquake production Ma, S.L., Chen, S.Y., Liu, P.X., Hu, X.Y., Wang, K.Y. & Huang, Y.M., 2008.
and its application to earthquake clustering, J. geophys. Res, 99(B2), Effect of fault jogs on frictional behavior: an experimental study, Sci.
2601–1618. China Earth Sci., 51, 1436–1445.
Doz, G.N. & Riera, J.D., 2000. Towards the numerical simulation of seismic Marone, C., 1998. Laboratory-derived friction laws and their application to
excitation, Nucl. Eng. Des., 196, 253–261. seismic faulting, Ann. Rev. Earth planet. Sci., 26, 643–696.
Erickson, S.G., Strayer, L.M. & Suppe, J., 2001. Initiation and reactivation Marone, C. & Kilgore, B., 1993. Scaling of the critical slip distance for
of faults during movement over a thrust-fault ramp: numerical mechanical seismic faulting with shear strain in fault zone, Nature, 362, 618–621.
models, J. Struct. Geol., 23, 11–23. Mckinnon, S.D. & De La Barra, I.G., 1998. Fracture initiation, growth and
Fang, Z. & Harrison, J.P., 2002. Development of a local degradation approach effect on stress field: a numerical investigation, J. Struct. Geol., 20, 1673–
to the modeling of brittle fracture in heterogeneous rocks, Int. J. Rock 1689.
Mech. Min. Sci., 39, 443–457. Miao, A.L., Ma, S.L. & Zhou, Y.S., 2010. Experimental study on frictional
Guattei, M. & Spudich, P., 2000. What can strong-motion data tell us about stability transition and micro-fracturing characteristics for anhydrite fault

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019


slip-weakening fault-friction? Bull. seism. Soc. Am., 90, 98–116. zones, Chin. J. Geophys., 53, 2671–2680 (in Chinese).
Harris, R.A. & Day, S.M., 1993. Dynamic of fault interaction: parallel Miyatake, T., 1992. Numerical simulation of the three-dimensional faulting
strike-slip faults, J. geophys. Res., 98(B3), 4461–4472. processes with heterogeneous rate- and state-dependent friction, Tectono-
Harris, R.A. & Day, S.M., 1997. Effects of a low-velocity zone on a dynamic physics, 211, 223–232.
rupture, Bull. seism. Soc. Am., 87, 1267–1280. Mora, P. & Place, D., 1994. Simulation of the frictional stick-slip instability,
He, C., 1995. Slip-weakening constitutive relation and the structure in the Pure appl. Geophys., 143, 61–87.
vicinity of a shear crack tip, Pure appl. Geophys., 145, 147–157. Okubo, P.G., 1989. Dynamic rupture modeling with laboratory-derived con-
Hemendra, K.A., 1997. Influence of fault bends on ruptures, Bull. seism. stitutive relations, J. geophys. Res., 94, 12 321–12 335.
Soc. Am., 87, 1691–1696. Ohnaka, M., 1993. Critical size of the nucleation zone of earthquake rupture
Hillers, G., Ben-Zion, Y. & Mai, P.M., 2006. Seismicity on a fault controlled inferred from immediate foreshock activity, J. Phys. Earth, 41, 45–56.
by rate- and state-dependent friction with spatial variations of the critical Ohnaka, M. & Yamashita, T., 1989. A cohesive zone model for dynamic
slip distance, J. geophys. Res., 111, 1–23. shear faulting based on experimentally inferred constitutive relation and
Ida, Y., 1973. The maximum acceleration of seismic ground motion, Bull. strong motion source parameters, J. geophys. Res., 94, 4089–4104.
seism. Soc. Am., 63, 959–968. Rice, J.R., 1983. Constitutive relations for fault slip and earthquake insta-
Ide, S. & Takeo, M., 1997. Determination of constitutive relations of fault bilities, Pure appl. Geophys., 121, 443–475.
slip based on seismic wave analysis, J. geophys. Res., 102, 27379–27391. Ruina, A.L., 1983. Slip instabilities and state variable friction laws, J. geo-
Ito, K. & Matsuzaki, M., 1990. Earthquakes as self-organized critical phe- phys. Res., 88, 10 359–10 370.
nomena, J. geophys. Res., 95, 6853–6860. Rutqvist, J., Birkholzer, J., Cappa, F. & Tsang, C.F., 2007. Estimating max-
Jiang, H.K., Ma, S.L., Zhang, L., Hou, H.F. & Cao, W.H., 2002. Spatio- imum sustainable injection pressure during geological sequestration of
temporal characteristics of acoustic emission during the deformation of CO2 using coupled fluid flow and geomechanical fault-slip analysis, En-
rock samples with compressional and extensional en-echelon fault pat- ergy Convers. Manage., 48, 1798–1807.
tern, Acta Seismol. Sin., 4, 402–413. Segall, P. & Pollard, D.D., 1980. Mechanics of discontinuous faults, J. geo-
Kato, N. & Tullis, T.E., 2001. A composite rate- and state-dependent law for phys. Res., 85(B8), 4337–4350.
rock friction, Geophys. Res. Lett., 28, 1103–1106. Shibazaki, B. & Matsu’ura, M., 1992. Spontaneous processes for nucleation,
Kato, N. & Tullis, T.E., 2003. Numerical simulation of seismic cycles with dynamic propagation, and stop of earthquake rapture, Geophys. Res. Lett.,
a composite rate- and state-dependent friction law, Bull. seism. Soc. Am., 19, 1189–1192.
93, 841–853. Sibson, R.H., 1985. Stopping of earthquake ruptures at dilatational jogs,
Kato, N., Yamamoto, K. & Hirasawa, T., 1994. Microfracture processes in Nature, 316, 248–251.
the breakdown zone during dynamic shear rupture inferred from labora- Sobolev, G.A., Ponomarev, A.V., Koltsov, A.V. & Smirnov, V.B., 1996. Sim-
tory observation of near-fault high-frequency strong motion, Pure appl. ulation of triggered earthquakes in the laboratory, Pure appl. Geophys.,
Geophys., 142, 713–734. 147, 345–355.
Kato, N., Satoh, T., Lei, X.L., Yamamoto, K. & Hirasawa, T., 1999. Effect Strayer, L.M. & Hudleston, P.J., 1997. Numerical modeling of fold initiation
of fault bend on the rupture propagation process of stick-slip, Tectono- at thrust ramps, J. Struct. Geol., 19, 551–566.
physics, 310, 81–99. Tchalenko, J.S. & Ambraseys, N.N., 1970. Structural analysis of Dasht-e
Kato, N., Lei, X.L. & Wen, X., 2007. A synthetic seismicity model for Bayaz (Iran) earthquake fractures, Geol. Soc. Am. Bull., 81, 41–60.
the Xianshuhe fault, Southweatern China: simulation using a rate- and Tinti, E., Cocco, M., Fukuyama, E. & Piatanesi, A., 2009. Dependence
state-dependent friction law, Geophys. J. Int., 169, 286–300. of slip weakening distance (Dc ) on final slip during dynamic rupture of
Lapusta, N., Rice, J.R., Ben-Zion, Y. & Zheng, G., 2000. Elastodynamic earthquakes, Geophys. J. Int., 177, 1205–1220.
analysis for slow tectonic loading with spontaneous rupture episodes on Wang, X.B., 2005. Joint inclination effect on strength, stress-strain curve and
faults with rate- and state-dependent friction, J. geophys. Res., 105(B10), strain localization of rock in plane strain compression, Mat. Sci. Forum,
23 765–23 789. 495–497, 69–74.
Lorig, L.J. & Hobbs, B.E., 1990. Numerical modeling of slip instability Wang, X.B., 2007a. Effects of joint width on strength, stress-strain curve
using the distinct element method with state variable friction laws, Int. J. and strain localization of rock mass in uniaxial plane strain compression,
Rock Mech. Min. Sci., 27, 525–534. Key Eng. Mat., 353–358, 1129–1132.
Ma, J., Du, Y. & Liu, L., 1986. The instability of en-echelon cracks and its Wang, X.B., 2007b. Prediction of height effect, plastic deformation and
precursors, J. Phys. Earth, 34, s141–s157. fracture energy for high-strength concrete by linear shear softening con-
Ma, J., Liu, L.Q., Liu, P.X. & Ma, S.L., 2007. Thermal precursory pattern of stitutive relation based on energy conservation method, Mag. Concr. Res.,
fault unstable sliding: an experimental study of en echelon faults, Chin. 59, 341–350.
J. Geophys., 50, 1141–1149 (in Chinese). Wang, X.B., 2008. Influence of imperfection number on acoustic emissions
Ma, J., Ma, S.P., Liu, L.Q. & Liu, P.X., 2010. Experimental study of thermal and elastic strain energy decrease of rock specimens with initially random
and strain fields during deformation of en echelon faults and its geological imperfections, Chin. J. Nonferr. Met., 18, 1541–1549 (in Chinese).
implications, Geodyn. Tectonophysics, 1, 24–35. Wang, X.B. & Pan, Y.S., 2008. Preliminary failure process modeling of
Ma, J., Sherman, S.I. & Guo, Y.S., 2012. Identification of meta-instable strain-softening rocks with heterogeneity and stiffness deterioration, in
stress state based on experimental study of evolution of the temperature the 3rd International Symposium on Modern Mining & Safety Technology
Stick-slip behaviours of typical faults 1041

Proceedings, pp. 767–771, Coal Industry Publishing House, Beijing, propagation process and anomalies related to the released energy during
P. R. China. a compressive jog intersection, J. Mech. Mat. Struct., 5, 1007–1022.
Wang, X.B. & Zhang, J., 2009. Numerical simulation of failure process of Wang, X.B., Ma, J. & Liu, L.Q., 2012. A comparison of mechanical behavior
three-point bending concrete beam considering heterogeneity of tensile and frequency-energy relations for two kinds of echelon fault structures
strength and post-peak softening curve, Eng. Mech., 26, 155–160 (in through numerical simulation, Pure appl. Geophys., 169, 1927–1945.
Chinese). Wolf, S., Manighetti, I., Campillo, M. & Ionescu, I.R., 2006. Mechanics of
Wang, X.B., Wu, D., Zhao, F.C. & Pan, Y.S., 2009. Simulation of failure normal fault networks subject to slip-weakening friction, Geophys. J. Int.,
processes, precursors and overall deformations of rock specimens with 165, 677–691.
different numbers of initially random imperfections, Prog. Geophys., 24, Zachariasen, J. & Sieh, K., 1995. The transfer of slip between two en echelon
1874–1881 (in Chinese). strike-slip faults: a case study from the 1992 Landers earthquake, southern
Wang, X.B., Ma, J. & Liu, L.Q., 2010. Numerical simulation of failed zone California, J. geophys. Res., 100(B8), 15 281–15 301.

Downloaded from https://academic.oup.com/gji/article-abstract/194/2/1023/599889 by guest on 04 September 2019

You might also like