Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


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

Numerical investigation of the deformation properties of rock materials T


subjected to cyclic compression by the finite element method
Yongqiang Zhoua, Qian Shenga,b, Nana Lic, Xiaodong Fua,*
a
State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan, 430071, China
b
School of Engineering Science, University of Chinese Academy of Sciences, Beijing, 100049, China
c
Wuhan Branch National Science Library Chinese Academy of Sciences, Wuhan, 430071, China

ARTICLE INFO ABSTRACT

Keywords: Rock materials, which are involved in a myriad of engineering projects, are often subjected to cyclic loading, such as
Cyclic loading earthquakes. Understanding the dynamic deformation properties of rock materials under cyclic loading is necessary
Sub-loading surface theory for evaluating the stability of rock engineering structures. This study numerically investigated the influence of cyclic
Dynamic property loading conditions (e.g., waveform, frequency, maximum loading stress and amplitude) on the deformation prop-
Rock materials
erties of rock samples using the finite element method (FEM). On the basis of the Drucker–Prager (D-P) yield criterion
FEM
and sub-loading surface theory, a dynamic constitutive model for reproducing the hysteresis loops and the accu-
mulative plastic deformation of rock materials subjected to cyclic loading was established, numerically implemented
using FEM, and validated through comparison with experimental results on basalt and granite model materials. Then,
cyclic uniaxial loading tests were simulated with a range of loading parameters, and the deformation properties of
rock samples were analyzed. The results indicated that the hysteresis loops and accumulated plastic deformation of
rock materials during the cyclic process can be numerically reproduced in an effective manner. The loading para-
meters significantly affected dynamic deformation properties such as the maximum strain, irreversible strain and
hysteresis loop. The simulation results were almost consistent with the experimental results in the published lit-
erature. Therefore, through the proposed dynamic constitutive model, FEM can be used to numerically simulate the
influence of cyclic loading conditions on the deformation properties of rock materials.

1. Introduction laboratory experiments. In cyclic loading tests, the loading and un-
loading curves do not coincide to form a closed plastic hysteresis loop.
Rock engineering, such as underground caverns, mountain tunnel Meanwhile, as the number of cycles increases, the hysteresis loops
engineering and mining engineering, may be damaged during strong move in the direction of increasing strain, and the irreversible plastic
earthquakes, which directly affect the projects and may even be life deformation of rock materials increases with the strain [3–7]. Many
threatening. Therefore, it is necessary to study the stability of rock researchers have conducted studies on the influence of cyclic loading
engineering materials under earthquakes. Seismic action is very com- parameters on the strength and deformation of rock materials. For ex-
plex, but it is usually equivalent to cyclic loading [1]. It is well known ample, Tao and Mo [3], Bagde and Petroš [8] and Xiao et al. [2] studied
that the mechanical behavior of brittle material under dynamic loads is the behavior of rock under cyclic loading with different waveforms. It
extremely different from that under static loads. For example, cyclic was found that the loading waveform is of great significance and affects
loading often causes brittle material to fail prematurely at a stress level rock behavior. Three levels of confining pressure and three sets of fre-
lower than its determined strength under monotonic conditions [2]. quencies were applied for the axial cyclic loading tests by Liu et al. [7].
Rocks in underground engineering are under high in situ stress field. The results from the cyclic loading tests indicate that frequency has a
Thus, exploring the mechanical behavior of rock subjected to cyclic strong influence on the dynamic deformation, dynamic stiffness, and
compression loading is of great importance for rock engineering design failure mode at the same confining pressure. Cyclic loading can reduce
and construction [3]. the material strength and elasticity, depending on the loading ampli-
Over the past few decades, considerable efforts have been made to tude and the maximum applied load in each cycle [8–11]. Fuenkajorn
evaluate the mechanical response of rocks under cyclic loading via and Phueakphum [12] and Momeni et al. [13] further indicated that the

*
Corresponding author.
E-mail address: xdfu@whrsm.ac.cn (X. Fu).

https://doi.org/10.1016/j.soildyn.2019.105795
Received 22 January 2019; Received in revised form 3 July 2019; Accepted 6 August 2019
0267-7261/ © 2019 Elsevier Ltd. All rights reserved.
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

List of symbols surface (MPa)


P , A1, f1 Current load (MPa), amplitude (MPa) and frequency (Hz)
, Geometric center of the normal-yield surface and the sub- R, Rs Ratio of the size of the sub-loading surface and similarity-
loading surface (MPa) center surface to that of the normal-yield surface (−)
a, r Material parameters for kinematic hardening (−) Re Minimum set point of R (−)
B, Ve Strain matrix and volume of element e (−) s Similarity-center (MPa)
, F Parameters related to the cohesion and internal friction s , ŝ Similarity-center of the sub-loading surface and normal-
angle (−) yield surface considering (MPa)
c, Cohesion (MPa) and internal friction angle (°) sy Point on the normal-yield surface and conjugate to the
C Material parameter of s (−) point on the similarity-center surface with respect to
Dep, Del Elasto-plastic matrix and elastic matrix (GPa) (MPa)
d Plasticity multiplier (−) ToL Permissible minimum (−)
d p Plastic strain increment (−) u, U Material parameter and function for similarity ratio (−)
E, Young's modulus (GPa) and Poisson's ratio (−) Maximum value of Rs (−)
F(H) Function of hardening variable H (−) y Dual stress on the normal yield surface of the current
f Yield stress function (−) stress (MPa)
(f nk ++11) e (int), (fn + 1)e (ext) Internal force and external force of element e Current stress on the sub-loading surface (MPa)
(MPa) ˆy Stress on the normal loading surface and is conjugate to
H Hardening variable (−) considering (MPa)
I1, I1 First invariant of the stress on normal yield surface and the ˜y, ˜ Vector difference between y , and s (MPa)
current stress on sub-loading surface (MPa) ( ˜ )m , (ˆ)
s m , m Mean normal stress of ˜ , and ŝ (MPa)
J2, J2 Second invariant of the deviator stress on normal yield , K0 Average strain rate (/s) and stress rate (MPa/s)
surface and the deviator current stress on sub-loading

fatigue strength and deformation properties of rocks are significantly 2. Dynamic constitutive model of rock materials
influenced by the characteristic parameters of cyclic loading, including
the input loading waveform, maximum applied stress, amplitude and A constitutive model of rock materials is a comprehensive description
frequency. To study the damage properties of jointed rock mass mate- of their strength and deformation and a macroscopic representation of
rials under dynamic cycling loads with different frequencies, ampli- their internal cracks. Therefore, the dynamic constitutive model of rock
tudes and durations, Li et al. [14] and Liu et al. [15] performed a model materials under cyclic loading should be a comprehensive reflection of
test with intermittently jointed rock samples. It was found that the fa- the hysteresis loop and plastic deformation accumulation. Regarding the
tigue strength and the deformation modulus of jointed rock samples mechanical properties of rock under cyclic loading, Liu et al. [34] pro-
decrease with increasing loading duration and loading frequency and posed a binary constitutive model based on the theory of failure me-
exhibit opposite variation with increasing loading amplitude. chanics for geological materials. Other researchers have proposed en-
According to relevant studies, the mechanical properties of rock dochronic plasticity model [35], P-M model [36], etc. These models have
materials under cyclic loading were studied mainly through experi- good practicability in simulating the dynamic response of rock under
ments and rarely through numerical simulation. Numerical simulation cyclic loading. However, they have too many parameters to determine,
is a complementary method of experimental methods, and also a useful and the physical meaning of many of the parameters in these models is
method for solving problems, such as those involving high-frequency not clear enough to determine their value. In addition, Wang et al. [37],
cyclic loads and time consumption, and the focus of numerical simu- Zhang et al. [38], Li et al. [39] and Xiao et al. [40] proposed the internal
lation is on the mechanical properties of jointed rocks under cyclic variable fatigue constitutive model based on cyclic loading tests. The
loading using the discrete element method (DEM). For example, model can reflect the evolution law of the deformation modulus and the
Nemcik et al. [16] simulated the behavior of rock joints under cyclic plastic strain of rock materials with respect to the number of cycles under
loading and analyzed the dynamic stability of an underground structure cyclic loading. However, the loading and unloading sections of each
using DEM. Liu et al. [1] numerically investigated the dynamic prop- hysteresis loop need to be considered separately. It is difficult to com-
erties of intermittent jointed rock models subjected to cyclic uniaxial pletely describe the stress-strain relationship of rock materials, and it
compression using DEM. However, the influence of the cyclic loading cannot reflect the hysteresis loop phenomenon well. Contrary to rock
parameters (i.e., loading waveform, loading frequency, maximum materials, there are many different models, such as the bounding surface
loading stress and amplitude) on the deformation response of intact model [41,42], the focal points model [43], and the sub-loading surface
rocks has been rarely evaluated systematically through numerical si- model [44], that exist for the cyclic behavior of concrete, which is similar
mulation by FEM. Due to its versatility and flexibility, FEM has been to that of rock materials. The formulation proposed hereinafter is based
successfully applied to various rock dynamics problems [17–27] and to on sub-loading surface theory [45]. Sub-loading surface theory is ad-
simulate the behavior of concrete under cyclic loading, which is similar vantageous for describing the hysteresis loop and plastic deformation
to that of rock materials, and has obtained successful results [28–33]. In accumulation of a material and satisfies the continuity and smoothness of
this work, we mainly focus on the influence of the cyclic loading a cyclic plasticity model. Hashiguchi and Tsutsumi [45,46] used the sub-
parameters, the hysteresis loop and the accumulation of the plastic loading surface model to simulate the deformation of metals and soils
deformation of rocks subjected to cyclic compression using FEM. This under cyclic loading, and other researchers applied the sub-loading
paper is organized as follows. First, a dynamic constitutive model for surface model to concrete [44] and soft rock [47] and obtained good
rock materials that combines the D-P yield criterion and sub-loading results. Sub-loading surface theory is mainly chosen due to its ability to
surface theory is established and numerically implemented through produce plastic deformation accumulation upon constant amplitude cy-
FEM. Then, using this model, the dynamic responses of rock samples are cles under shear strength and its physically based concept and relative
systematically analyzed via numerical uniaxial cyclic loading tests with simplicity. Therefore, sub-loading surface theory can be used to describe
different cyclic loading parameters. Finally, the entire study is sum- the deformation properties of rock materials under cyclic loading.
marized.

2
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

2.1. Dynamic constitutive model principles f= J2 + I1 F (H ) = 0 (9)

As mentioned above, a dynamic constitutive model for rock mate- F (H ) = F + FH d p


(10)
rials should describe the hysteresis loop and cumulative plastic strain in where I1 and J2 represent the first invariant of the stress and the second
rock materials under cyclic loading. Therefore, based on the theory of a invariant of the deviator stress, respectively, and F are parameters
sub-loading surface, this article chooses the D-P yield criterion as the related to the cohesion and internal friction angle of rock, and d p re-
normal yield surface to construct a dynamic constitutive model for rock presents the plastic strain.
materials. Based on the D-P criterion, the sub-loading surface of the dynamic
constitutive model can be established by inserting Eq. (9) into (5), and
2.1.1. Sub-loading surface theory the expression is written as follows:
The basic theory of the sub-loading surface model assumes that there
is a sub-loading surface within the normal yielding plane and that the f( )= J2 + I1 RF (H ) = 0 (11)
current stress point is always located on the sub-loading surface. where I1 and J2 represent the first invariant of the current stress and the
Compared with those of the elastoplasticity model, the transitions of the second invariant of the deviator current stress, respectively.
elasto-plastic phase of the sub-loading surface model can be smooth. The
sub-loading surface is shown in Fig. 1, and the normal yield surface is 2.2. Derivation of the constitutive matrix
expressed as
f ( y ) = F (H ) (4) According to the consistency condition, Eq. (6) can be expressed as
through derivation:
where y is the dual stress on the normal yield surface of the current
f
stress , which is on the sub-loading surface, F (H ) is a function of a d F (H ) dR RdF (H ) = 0
(12)
hardening variable H, and f is a yield stress function. Since the sub-
loading surface is geometrically similar to the normal yield surface, the From Eq. (6), the following expression can be written
sub-loading surface is
d =d d =d Rd (1 ˆ
R) ds + sdR (13)
f ( ) = RF (H ) (5)
According to the assumption that the similarity-center surface is
where R is the ratio of the size of the sub-loading surface to that of the within the normal yield surface, the following inequality must hold.
normal yield surface, and is the current stress on the sub-loading sur- 0 f (ˆ)
s F (H ) 0 Rs (14)
face, considering the back stress of the sub-loading surface. From Fig. 1,
the following geometric similarity relationships can be obtained: where Rs represents the ratio of the size of the similarity-center surface
to that of the normal yield surface. is the maximum value of similarity
= = R ˆy = R( y ) (6) ratio Rs . The time differentiation of the above inequality at the limit
s = F (H ) yields [49,50].
state that s lies on the limit surface f (ˆ)
s s
sˆ = s = =
R R (7) f (ˆ)
s dF (H )
• ds d sˆ 0 for R s =
sˆ F (H ) (15)
˜ = s = R ˜y = R ( y s) (8)

where and represent the geometric center of the normal yield surface f (ˆ)
s
•ˆs = f (ˆ)
s = F(H) for R s =
sˆ (16)
and that of the sub-loading surface, respectively, and also can be called
back stress, ˆ y is the stress on the normal loading surface and is conjugate The inequality (15) or (16) is called the enclosing condition of si-
to the current stress on the sub-loading surface considering the back milarity-center. The following relationship can be derived from the
stress, ˜ y is the vector difference between the normal yield stress and the enclosing condition of the similarity-center [50]:
center of similarity, ˜ is the vector difference in the stress between the
dF (H )
sub-loading surface and the center of similarity, s is the center of simi- ds d ˆ 0
larity of the normal yield surface and the sub-loading surface and is
F (H ) (17)
called the similarity-center, s is the similarity-center of the sub-loading
surface considering back stress, and ŝ is the similarity-center considering
back stress. When R = 0, = s = , and the sub-loading surface coin-
cides with the similarity-central surface. When R = 1, = y , re-
presenting the traditional elastic-plastic model. Hence, the traditional
elasto-plastic model is a special case of the sub-loading surface model.

2.1.2. The main idea of the dynamic constitutive model


The sub-loading surface model is highly adaptable and accurately
reflects the deformation characteristics of materials under cyclic
loading [44–46]; however, these materials are generally confined to soil
materials, metal materials and concrete and rarely include rock mate-
rials. Hence, the normal yield surface of the dynamic constitutive model
should choose the yield criterion that is applicable to rock materials.
The D-P criterion considers the hydrostatic pressure and intermediate
principal stress effects. The form of this criterion is simple and has a
clear physical meaning. Moreover, the yield curve on the plane is
round, i.e., smooth and nonsingular, and thus easily programmed.
Therefore, the D-P criterion is an appropriate choice for the normal
yield surface of the dynamic constitutive model [48]. The expression of Fig. 1. Diagram of a subloading surface [50] (note: dots indicate stress and
the D-P criterion is arrows indicate stress differences).

3
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

To meet the requirements of Eq. (17), the following is assumed: 2.3. Calculation of the similarity ratio

dF (H ) Rs
ds d sˆ = C d p
y (s y ) The similarity ratio R is an essential parameter for obtaining the
F (H ) expression of the sub-loading surface, and it can be calculated directly
p sˆ by R = f ( )/F (H ) . However, ( = ˜ + Rsˆ ) contains R. Therefore, it
=C d
R (18) needs to substitute the expression of into Eq. (11) to get the following
expression for the sub-loading surface:
where C is a material parameter, and sy is the point on the normal yield
surface and conjugate to the point on the similarity-center surface with f ( ) = f ( ˜ + Rsˆ) = tr ( ˜ + Rsˆ) + 1/2 ˜ + Rsˆˆ = RF (H ) (28)
respect to . Then, Eq. (18) can be rewritten as By transforming the above equation, the similarity ratio R is ob-
sˆ dF (H ) tained as follows:
ds = C d p +d + sˆ
R F (H ) (19) B + B2 4Ac
R=
2A (29)
For geotechnical materials, cyclic loading and dynamic problems
should adopt hardening or mixed hardening conditions [51]. To de- where
scribe the dynamic characteristics of rock materials subjected to cyclic
1 ^
loading, the nonlinear kinetic hardening criterion is used [50], and A= s 2 9 2 (ŝ)2
m (F (H ))2 + 6 F (H )(ŝ)m
(30)
2
kinematic hardening is expressed by
f f B = (~ ^s ) 18 2 (~)
m (ŝ)m + 6 F (H )(~)m (31)
d = a rF (H ) / d p
(20) 1 2 2 (˜)2
c = ˜ 9
2
m
(32)
where a and r are material parameters.
During the loading process, the sub-loading surface gradually ap- where (˜)m and (ŝ)m are the mean normal stress of ˜ and ŝ , respectively.
proaches the normal yield surface but will not pass it. When the sub-
loading surface reaches the normal yield surface, R is equal to 1. During 2.4. Numerical implementation
the unloading process, when the sub-loading surface reduces to a si-
milarity-center surface, it becomes a point, i.e., R is 0. Then, the sub- Based on the elastic prediction-plasticity correction, the numerical
loading surface gradually increases; this process is called reverse implementation of the dynamic constitutive model is created through
loading. Therefore, regarding the plastic strain, R must meet the fol- FEM (the integration is explicit). The detailed flow of the model is seen
lowing requirements: as follows:
Re R 0, dR = +
(1) Selecting the initial internal variables: when the time step is n +1
1 > R > Re , dR > 0
and the iterative step is 1, and the corresponding internal variables
R = 1, dR = 0
are (note that the initial value of R is set to a very small value, such
R > 1, dR < 0 (21) as 1 × 10−8, so that sub-loading surface theory is suited to dealing
where Re is the minimum set point of the size ratio of the sub-loading with rock materials that yield immediately upon loading):
surface to that of the normal yield surface. In this paper, Re is 0. Based snk + 1 = s n, k
= Rnk+ 1 = Rn , Fnk+ 1 = Fn (33)
n+1 n,
on the above requirements, R can be expressed as
dR = U d p = u ln R d p (22) (2) Performing the elastic prediction: when the time step is n +1, the
iterative step is k, assuming that the stress is
where u is a material parameter and U represents a function.
Substituting Eqs. (13), (19), (20) and (22) into Eq. (12), we can k+1
n+1 = k
n+1 + Deld k+1
n+1 (34)
obtain the following expression of the constitutive matrix for simulating
the behavior of rock materials subjected to cyclic loading (the plastic (3) Determining the yield judgments: the corresponding stress in-
flow rule adopts the associated flow rule): variants I1 and J2 are calculated using nk ++11 ; then, the following
Dep = Del inequality is evaluated:

f T
f f T
f f T
dF (H ) ˜ f( k+1
n + 1) Rnk+ 1 Fnk+ 1 (H ) 0 (35)
Del Del / Del + ˆ +d +U
F (H ) R
If inequality (35) is satisfied, then the stress at this time is the stress
sˆ in step (2) and ③ in step (4) should be performed; otherwise, the cor-
+ C (1 R) responding plasticity correction should be performed.
R (23)

f I1 1 J2 (4) Performing the plasticity correction: the purpose of the plasticity


= + correction is to obtain the correct stress so that the yield function
2 J2 (24)
can coincide with the yield surface.
I1 T
= [ 1 1 1 0 0 0] ① Solving for the plasticity multiplier:
(25)
k+1
f( n + 1) Rnk+ 1 Fnk+ 1 (H )
J2 T d k+1
=
= m[ 1 1 1 0 0 0] n+1 k + 1)
(26)
f( n +1
(Mnk+ 1 + Nnk + 1 Del Nkn+ 1)
k+1
n+ 1 (36)
dF (H ) H
= k + 1)
f( n
(27)
p +1
F (H ) 1+H d k+ 1
n+ 1
ep el
Nnk + 1 =
where D and D are the elastic-plastic matrix and the elastic matrix, f( nk + 1)
+1

respectively, and m is the mean normal stress of .


k+ 1
n+ 1 (37)

4
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

F k
n+1 ˜ nk + 1 k
n +1 sˆ kn + 1 hysteresis loop. is the ratio of the maximum similar center surface to
Mnk+ 1 = Nkn + 1 ˆ nk + 1 + k
+ Unk+ 1 + C (1 Rnk+ 1)
k
Fn + 1
n +1
Rnk+ 1 Rnk+ 1 the normal yield surface, and the value will not exceed 1 [50]. Due to the
intersection between u and C, these two parameters can be determined
(38)
only by trial and error and need to be adjusted according to the triaxial
② Updating the internal variables: test results of the loading, unloading and reloading stress path until it can
p (k + 1 ) accurately simulate the stress-strain curve [52] (similar to the calibration
d n+1 =d k+1 k
n + 1 Nn + 1 (39)
of the material parameters of the boundary surface model [48]).
p (k + 1)
k+1
n+1 = k
n+1 + Depd k+1
n+1 = k
n+1 + Del (d k+1
n+1 d n+1 ) (40)
2.6. Verification
p (k + 1)
Fnk++11 = Fnk+ 1 + Fnk+ 1 H d n+1 (41)
A cylindrical compression test for basalt (diameter of 48.58 mm and
p (k + 1)
k+1
n+1 = k
n+1 + a (rFnk+ 1 Nnk + 1 k
n + 1) d n+1 (42) height of 99.92 mm) is carried out on an RDT-10000 rock high-pressure
dynamic triaxial test system developed by the Wuhan Institute of Rock and
k
n+1 sˆ nk + 1 F k
n+1 k p (k + 1) Soil Mechanics, Chinese Academy of Sciences. The test system is shown in
snk ++11 = s nk + 1 + C + k
n+1 + sˆ n + 1 d n+1
Rnk+ 1 k
Fn + 1 Fig. 2. The confining pressure is 20 MPa, and the loading waveform is
sinusoidal wave with an amplitude of 65 MPa and a frequency of 1 Hz. In
(43)
addition, taking the type II rock mass of the underground cavern of Da-
③ Solving for the similarity ratio R: gangshan Hydropower Station as the prototype material, each size of the
model material is 150 mm and the geometric scale is 100. The model
B nk++11 + B nk++112 4Ank++11 c nk++11
Rnk++11 = material is composed of iron powder, barite powder, quartz sand, gypsum
2Ank++11 (44) and water. Cyclic uniaxial compression is applied to the model material
1 ^ k+1 using a sinusoidal wave with an amplitude of 27 KN and a frequency of
Ank++11 = s n+1 2 9 2 (ŝ k+1 )2 (Fnk++11 )2 + 6 Fnk++11 (ŝ kn++11)m 1 Hz. The diagram of calculation model is shown in Fig. 3. The cyclic load
2
n+1 m
(45)
is applied at the top of the model, and vertical constraint is applied at the
B nk++11 = (~ nk ++11 ŝ nk ++11 ) 18 2 (~ k+1) (ŝ k+1 )
n+1 m n+1 m + 6 Fnk++11 (~ kn++11)m (46) bottom of the model. The total number of elements is 200. The parameters
of the basalt and model material are shown in Table 1, and the comparison
1 2 ( ˜ k+1)2 between the simulation results of the dynamic constitutive model and the
c nk++11 = ˜ k+1 2
9
2
n+1 n+1 m
(47)
experimental results are shown in Figs. 4 and 5. It can be found that plastic
deformation occurs during the unloading and reloading processes, which
(5) Determining whether the internal force and external force are ba- results in a hysteresis loop, and the cumulative plastic strain increases with
lanced: the number of loadings. A reasonable consistency is observed between the
m m numerical and experimental results for the cumulative plastic strain with
(f nk ++11) e (int) (fn + 1)e (ext) < ToL the number of cycles.
e=1 e=1 (48)

(f nk ++11) e (int) = BT k+1


n + 1 dV
3. Influence of the cyclic loading parameters on the dynamic
Ve (49) properties of rock

where (f nk ++11) e (int) and (fn + 1)e (ext) are the internal force and external force 3.1. Influence of the loading waveform
on element e of analysis object, respectively, and ToL is a permissible
minimum. m e=1
(f nk ++11) e (int) and m (f )
e = 1 n + 1 e (ext)
represent the total in- This paper aims to study the influence of the deformation properties
ternal force and external force of analysis object by integration from
first element to m element, respectively. B and Ve are the strain matrix
and volume of element e.
If Eq. (48) is satisfied, the next step (step (1)) is performed; other-
wise, step (2) is performed to continue the iteration.

2.5. Parameter calibration

The parameters of the dynamic constitutive model are E, ν, c, ϕ, H, a,


r, u, C and (E, ν, c, and ϕ represent Young‘s modulus, Poisson's ratio,
the cohesion and internal friction angle, respectively); notably, this
model has fewer parameters than the binary medium model [34], the
endochronic plasticity model [35], the P-M model [36] and the boundary
surface model [48]. The first four parameters can be determined by static
uniaxial and triaxial tests. Young's modulus E and Poisson's ratio can be
determined from static uniaxial tests. Triaxial compression experiments
are used to determine strength parameters such as cohesion and friction
angle (c, ϕ). H, a and r are hardening parameters (H is the isotropic
hardening parameter, and a and r are the kinematic hardening para-
meters, and they are assumed equal to zero in this case) and can be
obtained from the stress-strain curves in the initial and inverse loadings
[50,52]. The latter three parameters, i.e., u, C and , are parameters
specific to the dynamic constitutive model. u is a parameter that governs
the rate at which the stress point approaches the normal yielding state,
and it is initially determined from the slope of the stress-strain curve at
moderate deformation rates. The value of C affects the width of the Fig. 2. RDT-10000 rock high-pressure dynamic triaxial test system.

5
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

2A12 f1 arcsin(P / A1 ) sin(2 arcsin(P / A1 ))


K0 = +
P 2 4 (50)

where P , A1 and f1 are the current load, amplitude and frequency,


respectively. Therefore, the average strain rate is
K0
=
E (51)

where E is the Young's modulus.


The average stain rate of the triangle wave can be expressed by

Fig. 3. Diagram of calculation model. 4A1 f1


=
E (52)
Table 1 For the trapezoidal wave, the average stain rate can be calculated as
Parameters. follows:
Type of rock E (GPa) v c (MPa) (0) H a r u C
A1
=
basalt 32 0.2 6.4 50 1 0 0 800 100 0.7 0.5(t 1 t 2) + 0.25/ f1 (53)
model material 1.2 0.25 0.38 30 1 0 0 30 2 1
The average strain rate of the different loading waveforms is shown
in Fig. 10(b). The order of the average strain rates of the different
of rock materials under cyclic loading through numerical simulation, and loading waveforms is as follows: square wave > sine wave >
the fatigue life of rock materials is not considered. To reflect the influ- triangular wave. Due to the strain rate effect, the modulus and strength
ence of different loading waveforms, three commonly used waveforms of the rock increase with the strain rate [53]. Therefore, the cracks
are designed, as shown in Fig. 6; these waveforms are the sine wave, the
triangle wave and the square wave. Each waveform has the same am-
plitude, frequency, and number of cycles. Because a square wave is more
difficult to obtain, it is mainly acquired by expanding the platform stage
of the trapezoidal wave [8]. The time points t1, t2, t3 and t4 in Fig. 7
describe the trapezoidal wave, where t2-t1 = t4-t3. Fig. 5 shows the
stress loading time history of the sine wave: first, the stress increases to
the average loading stress = 1/2 (maximum loading stress
(M) + minimum loading stress) = 10 MPa; then, the sine wave is loaded
8 times. A1 = 1/2 (maximum loading stress (M) - minimum loading
stress) = 10 MPa is the amplitude of the sine wave, and T is the period.
Then, the frequency is f1 = 1/T = 1 Hz. For a trapezoidal wave, t = t2-
t1 = t4-t3 is set to 0.49 s, which is very close to 1/2 T = 0.5 s, therefore,
the trapezoidal wave (0.49 s) can be used to approximate a square wave.
The parameters are shown in Table 2 and the diagram of calculation
model subjected to cyclic uniaxial compression is shown in Fig. 3.
Fig. 8 shows the stress-strain curves of the rock samples under each
loading waveform with the dynamic constitutive model and the D-P
criterion. It can be found that the dynamic constitutive model can ac-
curately reflect the deformation behavior of rocks under cyclic loading,
whereas the D-P criterion cannot reflect the hysteresis loop and cumu-
lative plastic strain. Fig. 9 depicts the stress-strain curves of the rocks
under cyclic loading during the first and last cycles. The hysteresis loop
areas and the initial modulus for different loading wave patterns are not
identical [2], indicating that the dynamic constitutive model can reflect
the size of the hysteresis loop under different loading waveforms. The
residual strain is different under the loading waveforms, and the order of
the strain magnitude of different loading waveforms is triangular
wave > sine wave > square wave, as shown in Fig. 10(a), which is in
line with the experimental results [2,3]. Under cyclic loading, the de-
formation of rock specimens is characterized by three stages: initial de-
formation stage, constant velocity development stage and accelerated
failure stage [1,40]. Fig. 10(a) shows that at the beginning, due to par-
ticle compacting and microcrack closure, the deformation of rock grows
rapidly and then keeps a slow constant growth process.
In nature, the influence of the loading waveform, amplitude and
frequency on the mechanical properties can be integrated into the in-
fluence of the loading rate [2]. Xiao et al. [2] suggested that different
loading waveforms have different average loading rates and that the
average loading rate of the sine wave can be expressed by
Fig. 4. Comparison between the simulation and testing results of basalt
(a) Stress-strain curves
(b) Residual strain vs number of cycles.

6
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 7. Stress loading time history of the sine wave.

Table 2
Parameters.
E (GPa) v c (MPa) (0) H a r u C

12 0.2 4 30 0 0 0 600 1 1

3.2. Influence of the frequency

The main frequency band of the seismic wave is generally between


0 Hz and 10 Hz; therefore, cyclic loading with a fixed maximum loading
stress of 20 MPa, a fixed amplitude of 10 MPa and 3 frequencies (i.e.,
0.1 Hz, 1 Hz and 10 Hz) is applied.
Fig. 11 shows the stress-strain curves under cyclic loading with
distinct frequencies. The hysteresis loop and cumulative plastic strain
are not the same for different frequencies, and the smaller the fre-
quency, the larger the hysteresis loop and the greater the cumulative
plastic strain. Fig. 12(a) and Fig. 12(b) demonstrate the change in the
maximum strain and residual strain, respectively; the maximum and
residual strains of the rock specimen decrease as the frequency in-
creases: the maximum strain decreases from 0.002311 at 0.1 Hz to
Fig. 5. Comparison between the simulation and testing results of model ma- 0.002199 at 1 Hz and to 0.002117 at 10 Hz, and the residual strain
terial decreases from 0.000613 at 0.1 Hz to 0.00053 at 1 Hz and to
(a) Stress-strain curves 0.000466 at 10 Hz. As shown in Fig. 13, a linear negative correlation
(b) Residual strain vs number of cycles.
was found between the maximum strain, the residual strain and loga-
rithm of the frequency. The deformation under different frequencies
can also be explained by the loading rate. For each load, when the
amplitude is fixed, as seen in Eq. (53), the average loading rate in-
creases with frequency, as displayed in. Fig. 14. The average strain rate
is greater than 10−2/s when the frequency of cyclic loading is 10 Hz.
According to the analysis of the different loading waveforms mentioned
in Sect. 3.1, the order of the magnitude of the strain rate effect under
different loading frequencies should be as follows: 0.1 Hz < 1 Hz <
10 Hz. Therefore, the order of the maximum strain and the residual
strain with different frequencies is as follows: 0.1 Hz > 1 Hz > 10 Hz.
A similar phenomenon was also observed by Xiao et al. [2] and Ma et al.
[56] for intact rocks. These results indicate that the dynamic con-
stitutive model can reflect the deformation of rock materials under
different loading frequencies.

Fig. 6. Schematic of the loading waveforms.


3.3. Influence of the maximum loading stress

within the rock will produce different responses under different loading Studies showed that the stress level is a primary factor that affects
waveforms. In other words, the strain decreases with an increase in the deformation of rock materials under cyclic loading [12,13,58]. The
loading rate, which can explain the above results. For the following stress-strain curves of the whole process and the last cycle for rock
simulations, the sine wave is a good choice as the loading waveform samples under different maximum loading stresses (i.e., 20 MPa,
[54–57], and the parameters are shown in Table 2. 22 MPa, 25 MPa and 30 MPa) at a fixed amplitude of 10 MPa and a

7
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 8. Stress-strain curves under cyclic loading with different loading wave- Fig. 9. Stress-strain curves of rocks under cyclic loading with different loading
forms waveforms
(a) Stress-strain curves of dynamic constitutive model (a) Stress-strain curves of the first cycle
(b) Stress-strain curves of D-P criterion. (b) Stress-strain curves of the last cycle.

constant frequency of 1 Hz are shown in Fig. 15. The hysteresis loop original internal cracks in the rock are more easily expanded, and new
area and size decrease with an increase in the maximum loading stress, cracks are more likely to form, increasing the rock damage. Conse-
which agrees with the experimental results [59]. The variations in the quently, with an increase in the maximum loading stress, the maximum
maximum and residual strains under different maximum loading strain and residual strain of the rock increase.
stresses are illustrated in Fig. 16, which shows that the deformation of
rock grows rapidly at the first cycle and then keeps a slow constant
growth process under cyclic loading, and the rate of increase is de- 3.4. Influence of the loading amplitude
termined by the maximum loading stress. The influence of the max-
imum loading stress on the simulated maximum strain and residual Four loading amplitudes were evaluated to analyze the influence of
strain is illustrated in Fig. 17. Both the maximum and residual strains different loading amplitudes (i.e., 10 MPa, 8 MPa, 6 MPa and 4 MPa) on
are approximately linearly proportional to the maximum loading stress. the deformation characteristics of rocks. All cyclic loads are applied
As the maximum loading stress increases from 20 MPa to 30 MPa, the with a given maximum loading stress of 20 MPa and a fixed frequency
maximum strain at the seventh cycle increases from 0.0022 to 0.0039, of 1 Hz. The stress-strain curves of the rock samples under different
and the residual strain at the seventh cycle increases from 0.000507 to loading amplitudes are shown in Fig. 18. The variations in the max-
0.0023. Therefore, the maximum loading stress not only controls the imum and residual strains under different loading amplitudes are illu-
maximum strain but also dominates the residual strain. strated in Fig. 19, which shows that the deformation of rock specimens
The influence of the maximum loading stress on the deformation varies with loading amplitude. The influence of the amplitude on the
characteristics of the rock samples can be explained by the micro- simulated maximum strain and residual strain is illustrated in Fig. 20. It
fracturing mechanism of rock during the cyclic loading process [13]. can be found that the residual strain of the rock samples is inversely
Due to the increase in the maximum loading stress, both the average related to the increase in the loading amplitude, which is consistent
loading stress and the minimum loading stress increase. Thus, the with the results observed by Liu et al. [1]. As the amplitude varies from
4 MPa to 10 MPa, the residual strain at the seventh loading cycle

8
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 10. The variation in the residual strain and strain rate with number of
cycles under different loading waveforms Fig. 11. Stress-strain curves under cyclic loading with different loading fre-
(a) Residual strain under different loading waveforms quencies
(b) Strain rate under different loading waveforms. (a) Stress-strain curves of the whole process
(b) Stress-strain curves of the last cycle.

decreases from 0.000507 to 0.00165. For each load, when the fre-
quency is fixed, as seen in Eq. (53), the average loading rate increases maximum loading stress on the deformation of rock materials. It can be
with amplitude. Therefore, even though the maximum loading stress of found that the relationships between deformation and these two factors
these four cases is the same, the maximum strain of the rocks slightly are opposites. Therefore, to further reveal the relationship between rock
decreases with loading amplitude. deformation and these two factors, this section focuses on the common
The influence of the loading amplitude on the deformation char- influence of the maximum loading stress and loading amplitude on the
acteristics of the rock samples can also be explained by the micro- deformation characteristics of rock. Four conditions were devised: to
fracturing mechanism of rock [13]. At a constant maximum loading maintain the average loading stress, the maximum loading stress is
stress, the higher the loading amplitude, the lower the minimum evaluated at 20 MPa, 18 MPa, 16 MPa and 14 MPa, and the loading
loading stress. Therefore, preexisting cracks within the rock are not amplitudes and frequencies are the same as those of Sect. 3.5.
easily extended, and new cracks are less likely to be generated, corre- The stress-strain curves of the rock samples under different max-
sponding to a decrease in rock damage. In addition, the development of imum loading stresses and loading amplitudes are shown in Fig. 21. The
the plastic zone in the rock is mainly affected by the maximum loading hysteresis loop area increases with the maximum loading stress and
stress, and the residual deformation is mainly determined by the loading amplitude, which is consistent with the experimental results
minimum loading stress for the same maximum loading stress. There- [60]. This phenomenon can be attributed to the increased energy dis-
fore, with an increase in loading amplitude, the maximum strain of the sipation generated during the cyclic loading process. With increasing
rock slightly decreases and the residual strain decreases. amplitude and maximum loading stress, the area of the hysteresis loop
produced in each cycle increases, which implies that an increasing
amount of energy is dissipated during subsequent loading and un-
3.5. Combined influence of the maximum loading stress and loading
loading processes. The variation in the maximum and residual strains
amplitude
under different maximum loading stresses and loading amplitudes is
displayed in Fig. 22. Fig. 23 illustrates that the maximum strain
The above sections analyzed the influence of loading amplitude and

9
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 14. The variation in strain rate with load under different frequencies.

Fig. 12. The variation in strain with number of cycles under different loading
frequencies
(a) Maximum strain under different loading frequencies
(b) Residual strain under different loading frequencies.

Fig. 15. Stress-strain curves under cyclic loading with different maximum
stresses
(a) Stress-strain curves of the whole process
Fig. 13. The variation in residual strain with loading frequency (the “8-
(b) Stress-strain curves of the last cycle.
Residual Strain” represents the residual strain at the 8th cycle, and the fol-
lowing is similar).

10
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 18. Stress-strain curves under cyclic loading with different loading am-
plitudes.

Fig. 16. The variation in strain with number of cycles under different maximum
stresses
(a) Maximum strain under different maximum stresses
(b) Residual strain under different maximum stresses.

Fig. 19. The variation in strain with number of cycles under different loading
Fig. 17. The variation in maximum strain and residual strain with maximum amplitudes
stress. (a) Maximum strain under different loading amplitudes
(b) Residual strain under different loading amplitudes.

11
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 20. The variation in strain with loading amplitude.

Fig. 22. The variation in strain with number of cycles under different loading
amplitudes and maximum stresses
(a) Maximum strain under different loading amplitudes and maximum stresses
(b) Residual strain under different loading amplitudes and maximum stresses.

maximum loading stress in Fig. 17, the rate of the increase in the
maximum strain decreases when considering the common effect of the
maximum loading stress and loading amplitude. However, the residual
strain shows a linear decrease with an increase in the maximum loading
stress, which is also very different from the results in Fig. 17. Therefore,
when considering the combined influence of the maximum loading
stress and loading amplitude on the deformation of rock materials, the
increase in loading amplitude decreases the rate of increase in the
maximum and residual strains and changes the trend of the residual
strain from increasing to decreasing. Meanwhile, the increase in the
maximum loading stress increases the rate of increase of the maximum
and residual strains.

Fig. 21. Stress-strain curves under cyclic loading with different loading am- 4. Conclusions and discussion
plitudes and maximum stresses
(a) Stress-strain curves of the whole process Exploring the mechanical behavior of rocks subjected to cyclic
(b) Stress-strain curves of the last cycle. loading is important for rock engineering design and construction.
However, the dynamic responses of rock samples subjected to various
increases linearly with the loading amplitude, and the residual strain cyclic loading conditions have been investigated rarely via numerical
tends to decrease linearly with loading amplitude, which diverges from simulation. In this study, a dynamic constitutive model based on the
the results displayed in Fig. 19. Compared with the influence of the sub-loading surface theory and D-P criterion is proposed, and cyclic
loading tests are numerically simulated on rock samples. The dynamic

12
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

Fig. 23. The variation in strain with loading amplitude and maximum stress.

deformation properties of rocks subjected to different cyclic loading Montreal.


conditions, including different waveforms, frequencies, maximum [6] Peng SS, Podrieks ER, Cain PJ. Study of rock behavior in cyclic loading. Soc Petrol
Engrs 1973:181–92. Pre-print SPE-4249.
loading stresses and amplitudes, are systematically studied. The main [7] Liu E, Huang R, He S. Effects of frequency on the dynamic properties of intact rock
conclusions and discussions are reviewed here: samples subjected to cyclic loading under confining pressure conditions. Rock Mech
Rock Eng 2012;45(1):89–102.
[8] Bagde MN, Petroš V. Waveform effect on fatigue properties of intact sandstone in
(1) The proposed dynamic constitutive model can effectively reproduce uniaxial cyclical loading. Rock Mech Rock Eng 2005;38(3):169–96.
the hysteresis loops and the accumulated plastic deformation of the [9] Singh SK. Fatigue and strain hardening behavior of graywacke from the flagstaff
rock samples during both the loading and unloading stages. formation, New South Wales. Eng Geol 1989;26(2):171–9.
[10] Ray SK, Sarkar M, Singh TN. Effect of cyclic loading and strain rate on the me-
(2) The hysteresis characteristics and rock deformation under a sine
chanical behavior of sandstone. Int J Rock Mech Min Sci 1999;36(4):543–9.
waveform is worse than that under a square waveform but better [11] Jun-Ichi K, Ishizuka Y, Abe T, Ishijima Y, Goto T. Estimate of the fatigue strength of
than that under a triangular waveform when the loading frequency granite subjected to long-period cyclic loading. Shigen Sozai 2001;116:111–8.
[12] Fuenkajorn K, Phueakphum D. Effects of cyclic loading on mechanical properties of
and amplitude remain constant. With increasing frequency, the
Maha Sarakham salt. Eng Geol 2010;112(1–4):43–52.
maximum and residual strains of the rock specimens decrease. The [13] Momeni A, Karakus M, Khanlari GR, Heidari M. Effects of cyclic loading on the
results are almost in line with the experimental results, which in- mechanical properties of a granite. Int J Rock Mech Min Sci 2015;77:89–96.
dicates that the dynamic constitutive model can reflect the strain [14] Li N, Chen W, Zhang P, et al. The mechanical properties and a fatigue-damage
model for jointed rock masses subjected to dynamic cyclical loading. Int J Rock
rate effect to some extent. Mech Min Sci 2001;38(7):1071–9.
(3) Both the maximum strain and residual strain are approximately [15] Liu Y, Dai F, Feng P, et al. Mechanical behavior of intermittent jointed rocks under
linearly proportional to the maximum loading stress. The maximum random cyclic compression with different loading parameters. Soil Dyn Earthq Eng
2018;113(10):12–24.
strain of the rocks remains approximately unchanged with an in- [16] Nemcik J, Mirzaghorbanali A, Aziz N. An elasto-plastic constitutive model for rock
crease in the loading amplitude, while the residual strain is in- joints under cyclic loading and constant normal stiffness conditions. Geotech Geol
versely related to the loading amplitude. Considering the combined Eng 2014;32(2):321–35.
[17] Saksala T, Hokka M, Kuokkala VT. Numerical 3D modeling of the effects of strain
effect of the maximum loading stress and the loading amplitude, the rate and confining pressure on the compressive behavior of Kuru granite. Comput
evolutions of the size of the hysteresis loop, maximum strain and Geotech 2017;88(8):1–8.
residual strain are different from those under the influence of the [18] Zhu J, Liao Z, Tang C. Numerical SHPB tests of rocks under combined static and
dynamic loading conditions with application to dynamic behavior of rocks under in
loading amplitude or maximum loading stress. These simulation
situ stresses. Rock Mech Rock Eng 2016;49(10):1–12.
results are basically consistent with experimental results or pub- [19] Bendezu M, Romanel C, Roehl D. Finite element analysis of blast-induced fracture
lished results. propagation in hard rocks. Comput Struct 2017;182(4):1–13.
[20] Gao Q, Lu W, Hu Y, et al. An evaluation of numerical approaches for S-wave
component simulation in rock blasting. J Rock Mech Geotech Eng 2017(5):52–64.
Acknowledgments [21] Jayasinghe LB, Zhou HY, Goh ATC, et al. Pile response subjected to rock blasting
induced ground vibration near soil-rock interface. Comput Geotech
The work reported in this paper is financially supported by the 2017;82(2):1–15.
[22] Mitelman A, Elmo D. Analysis of tunnel support design to withstand spalling in-
National Key R&D Program of China (2018YFC0809400), the National duced by blasting. Tunn Undergr Space Technol Inc Trenchless Technol Res
Natural Science Foundation of China (No. 51679232, No. 51809258, 2016;51(10):354–61.
No. 51779250), and the International Partnership Program of Chinese [23] Vosoughifar H, Madadi F, Rabiefar A. Modified dynamic stress concentration factor
for twin tunnels using a novel approach of FEM-scattering. Tunn Undergr Space
Academy of Sciences Grant No. 131551KYSB20180042. The authors are Technol 2017;70(11):30–41.
thankful for these supports. [24] Kouretzis GP, Bouckovalas GD, Gantes CJ. 3-D shell analysis of cylindrical under-
ground structures under seismic shear (S) wave action. Soil Dyn Earthq Eng
2006;26(10):909–21.
References [25] Shimbo Taiki. Development and application of a dynamic XFEM for the seismic
residual displacement analysis of an embankment. Soils Found 2017;57(3):357–70.
[1] Liu Y, Dai F, Zhao T, Xu N. Numerical investigation of the dynamic properties of [26] Khaledi K, Mahmoudi E, Datcheva M, et al. Stability and serviceability of under-
intermittent jointed rock models subjected to cyclic uniaxial compression. Rock ground energy storage caverns in rock salt subjected to mechanical cyclic loading.
Mech Rock Eng 2017;50(01):89–112. Int J Rock Mech Min Sci 2016;86(7):115–31.
[2] Xiao J, Ding D, Xu G, Jiang F. Waveform effect on quasi-dynamic loading condition [27] Ghamgosar M, Erarslan N. Experimental and numerical studies on development of
and the mechanical properties of brittle materials. Int J Rock Mech Min Sci fracture process zone (fpz) in rocks under cyclic and static loadings. Rock Mech
2008;45(4):621–6. Rock Eng 2016;49(3):893–908.
[3] Tao ZY, Mo HH. An experimental study and analysis of the behavior of rock under [28] Tejchman J, Bobinski J. Continuous and discontinuous modeling of fracture in
cyclic loading. Int J Rock Mech Min Sci Geomech Abstr 1990;27(1):51–6. concrete using FEM. Springer; 2014. p. 161–2. 1996.
[4] Attewell PB, Farmer IW. Fatigue behavior of rock. Int J Rock Mech Min Sci [29] Han L, Wang W, Zhao X. Behavior of steel beam to concrete-filled SHS column
Geomech Abstr 1973;10(1):1–9. frames: finite element model and verifications. Eng Struct 2008;30(6):1647–58.
[5] Hardy HR, Chugh YP. Failure of geologic materials under low cycle fatigue. [30] Sena-Cruz José M, Barros JAO, Coelho Mário RF, et al. Efficiency of different
Proceedings of the sixth Canadian symposium on rock mechanics. 1970. p. 33–47. techniques in flexural strengthening of RC beams under monotonic and fatigue

13
Y. Zhou, et al. Soil Dynamics and Earthquake Engineering 126 (2019) 105795

loading. Constr Build Mater 2012;29:175–82. Plast 2005;21:1941–69.


[31] Xiong Q, Chen Z, Kang J, Zhou T, Zhang W. Experimental and finite element study [47] Fu Y, Iwata M, Ding W, Yashima A. An elastoplastic model for soft sedimentary rock
on seismic performance of the LCFST-D columns. J Constr Steel Res considering inherent anisotropy and confining-stress dependency. Soils Found
2017;137:119–34. 2012;52(4):575–89.
[32] Rezapour M, Ghassemieh M. Macroscopic modelling of coupled concrete shear wall. [48] Cerfontaine B, Charlier R, Collin F, Taiebat M, et al. Validation of a new elasto-
Eng Struct 2018;169:37–54. plastic constitutive model dedicated to the cyclic behaviour of brittle rock mate-
[33] Toyama H, Kishida H, Yonezu A. Characterization of fatigue crack growth of con- rials. Rock Mech Rock Eng 2017;50(10):1–18.
crete mortar under cyclic indentation loading. Eng Fail Anal 2018;83:156–66. [49] Hashiguchi K, Ueno M, Ozaki T. Elastoplastic model of metals with smooth elas-
[34] Liu E, Zhang J, He S, et al. Binary medium model of rock subjected to cyclic loading. tic–plastic transition. Acta Mech 2012;223(5):985–1013.
J Chongqing Univ Technol 2013;27(9):6–12. [50] Hashiguchi K. Elastoplasticity theory. Springer Berlin Heidelberg; 2009.
[35] Chang K, Yang T. A constitutive model for the mechanical properties of rock. Int J [51] Zheng Y, Kong L. Generalized plastic mechanics and its application. Eng Sci
Rock Mech Min Sci Geomech Abstr 1982;19(3):123–33. 2006;4(1):21–36.
[36] Carmeliet J, Abeele KEVD. Application of the Preisach‐Mayergoyz space model to [52] Kong L, Zheng Y, Yao Y. Subloading surface cyclic plastic model for soil based on
analyze moisture effects on the nonlinear elastic response of rock. Geophys Res Lett Generalized plasticity(I): theory and model. Rock Soil Mech 2003;24(2):141–5.
2002;29(7). 48-41. [53] Lajtai EZ, Duncan EJS, Carter BJ. The effect of strain rate on rock strength. Rock
[37] Wang Z, Zhao J, Li S, Xue Y, et al. Fatigue mechanical behavior of granite subjected Mech Rock Eng 1991;24(2):99–109.
to cyclic load and its constitutive model. Chin J Rock Mech Eng [54] Tien YM, Lee DH, Juang CH. Strain, pore pressure and fatigue characteristics of
2012;31(9):1888–900. sandstone under various load conditions. Int J Rock Mech Min Sci Geomech Abstr
[38] Zhang PY, Xia CC, Zhou SW, Zhou Y, et al. A constitutive model for rock under 1990;27(4):283–9.
cyclic loading and unloading. Rock Soil Mech 2015;36(12):3354–9. [55] Bagde MN, Petroš V. Fatigue and dynamic energy behavior of rock subjected to
[39] Li N, Zhang P, Chen Y, et al. Fatigue properties of cracked, saturated and frozen cyclical loading. Int J Rock Mech Min Sci 2009;46(1):200–9.
sandstone samples under cyclic loading. Int J Rock Mech Min Sci [56] Ma L, Liu X, Wang M, Xu H, Hua R, et al. Experimental investigation of the me-
2003;40(1):145–50. chanical properties of rock salt under triaxial cyclic loading. Int J Rock Mech Min
[40] Xiao J, Ding D, Jiang F, Xu G. Fatigue damage variable and evolution of rock Sci 2013;62(9):34–41.
subjected to cyclic loading. Int J Rock Mech Min Sci 2010;47(3):461–8. [57] Vaneghi RG, Ferdosi B, Okoth AD, Kuek B. Strength degradation of sandstone and
[41] Yang B, Dafalias YF, Herrmann LR. A bounding surface plasticity model for con- granodiorite under uniaxial cyclic loading. J Rock Mech Geotech Eng
crete. J Eng Mech 1985;111(3):359–80. 2018;10(1):117–26.
[42] Fardis MN, Chen ES. A cyclic multiaxial model for concrete. Comput Mech [58] Bagde MN, Petroš V. Fatigue properties of intact sandstone samples subjected to
1986;1(4):301–15. dynamic uniaxial cyclical loading. Int J Rock Mech Min Sci 2005;42(2):237–50.
[43] Yankelevsky DZ. Model for cyclic compressive behavior of concrete. J Struct Eng [59] Jiang X, Yin G, Hong W, Tang Y. Experimental research on the evolution of hys-
1987;113(2):228–40. teresis curve of rock in different axial stress levels. J Chongqing Jianzhu Univ
[44] Khosroshahi AA, Sadrnejad SA. Substructure model for concrete behavior simula- 2006;28(2):40–2.
tion under cyclic multiaxial loading. Int J Eng Trans A Basic 2008;21(4):329–46. [60] Deng H, Hu Y, Li J, Wang Z, Zhang X, et al. Effects of frequency and amplitude of
[45] Hashiguchi K. Generalized plastic flow rule. Int J Plast 2005;21:321–51. cyclic loading on the dynamic characteristics of sandstone. Rock Soil Mech
[46] Tsutsumi S, Hashiguchi K. General non-proportional loading behavior of soils. Int J 2017;38(12):3402–9.

14

You might also like