Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: A new poroelastic damage model is developed in the paper to describe the macroscopic failure of rock materials
Multiscale fracturing due to microcrack nucleation and propagation based on a multiscale framework. The model is deduced from
Heterogeneous rock locally periodic microstructure with dynamically evolved microcracks in heterogeneous rock body. The ho
Microcrack
mogenization method based on asymptotic expansions gives rise to the damage evolution law coupled with the
Damage law
Homogenization
poroelastic fracture system, which includes the fracture opening induced permeability change. The obtained
model takes into account the complex coupling between fluid pressure-deformation and hydro-mechanical (HM)
properties at the microscale, leading to the nonlinear anisotropic mechanical behavior, degradation of both
elastic stiffness and poroelastic properties at the macroscale, which is fundamental to describe the complex
fracturing behavior influenced by microcrack distribution. The homogenized coefficients are illustrated in detail
for a given set of initial material parameters, with dependence on the normalized damage variable. Results of
numerical simulations well reproduce specific experimental observations where fracturing in heterogeneous
rocks is shown to be a multiscale phenomenon that initiates from the microcrack nucleation and propagation,
while the fracture propagation direction is shown to be influenced by both external loadings and microcrack
distribution. The easy implementation in finite element framework and revealed micro-mechanism for macro
scopic failure under strict mathematical formulations make the wide application of model in rock mechanics
problems.
* Corresponding author. College of Water Resource & Hydropowerm Sichuan University, NO. 24, South Section 1, Yihuan Road, Chengdu, Sichuan, 610042, China.
E-mail address: liujf@scu.edu.cn (J. Liu).
https://doi.org/10.1016/j.ijrmms.2024.105676
Received 8 July 2023; Received in revised form 31 January 2024; Accepted 15 February 2024
Available online 24 February 2024
1365-1609/© 2024 Elsevier Ltd. All rights reserved.
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
stiffness and strength of each part comprise jointly the overall properties 2. Multiscale poroelastic model
of the rock, while the local failure between discrete parts represents the
microcracking process and as a result the macroscopic damage is grad 2.1. Problem statement
ually formed due to multiple failure events.14,15 In the significant efforts
made during the past decades, the discrete approach has proved its We consider a porous medium consisting of a large number of peri
capability by reproducing the complex fracture propagation behavior, e. odically distributed microcracks, in a 2D plane strain framework.
g., the wing crack evolution by interacting with pre-existed microcracks Infinitesimal deformation and isothermal conditions are adopted. Ten
in rocks,16 hydraulic fracturing (HF) in natural fractured rock mass.17 sile strain is positively counted.
Since the mechanical behavior and fluid flow are coupled at the The porous macrostructure and the set of microcracks are denoted by
microscale, the approach is able to describe pore fluid induced micro B and C , respectively, as seen in Fig. 1(a). Each crack is straight of
cracking behavior. However, the adopted microstructure and scale may length l and has a constant spacing lc , which is much smaller than the
not represent the real rock properties so that the model parameters may macro-structural size Lc . The microstructure containing one straight
not correspond to the measurable physical properties, which need to be crack can be scaled to a reference cell Y, which is considered as a
determined by curve fitting.18 In addition, the computational cost be representative elementary volume (REV). The single crack (denoted by
comes expensive when focusing on the macroscale problems. C Y ) contained in cell Y is oriented in an angle θ with respect to hori
For the continuum approach, the heterogeneous rock is generally zontal direction. The considered REV is made of compressible solid
idealized to be continuous body. A simple treatment is incorporating the phase and saturated by incompressible fluid.
elastic model together with damage criterion and post-failure behavior, We begin with the Biot’s theory for the deformable porous
as to represent the microcracking process, some widely-adopted elasto- macrostructure35,36:
damage models are referred to.19,20 This kind of elasto-damage models
may become more sophisticated by considering the effect of micro
∂σ ij
= 0, with σ ij = Cijkl exkl (u) − αpf δij (1)
cracks, the porosity change and the structural damage as internal vari ∂xj
ables, detailed works can be seen in.21–24 As another enrichment, the
∂qi k ∂pf
microcrack-based continuum model generally adopts the + φ̇ = 0, with qi = − (2)
∂xi μf ∂xi
semi-analytical method that evaluates the microcrack evolution and the
corresponding permeability change through fracture mechanics.6,25,26
φ̇ = αėxii (u) + βṗf (3)
To overcome the model limitations that can represent microcrack
induced anisotropy and progressive damage behavior, anisotropic
where σij is total stress tensor, u is displacement tensor, Cijkl is elastic
damage variables, strength softening laws and heterogeneous HM
stiffness tensor, exij is strain tensor with respect to x-axis, α is Biot co
properties may be incorporated.27–30 However, as the above-mentioned
efficient, pf is fluid pressure in porous matrix, qi is Darcy’s velocity, φ is
models are partly or totally based on macroscopic phenomenological
considerations to reproduce experimental observations, the physical porosity, β is Biot modulus, δij is the Kronecker delta.
mechanisms describing microstructural damage are missing. Inside the saturated cracks, fluid yields free-flow condition in a di
To address the above limitations and represent the microcracking rection parallel to crack. To ease the difficulty in describing the coupling
process in a more physical manner, a multiscale damage model is regions between porous-flow and free-flow, the following Stokes re
introduced that considers poroelastic coupling and stiffness degradation lations are introduced:
anisotropically due to microcrack propagation. Compared to our pre ∇x ⋅ σb = 0, with σb = 2μf ex (vb ) − pb I; ∇x ⋅ vb = 0 (4)
vious work in,31 the current contribution is improved from the following
aspects: (1) degradation of poroelastic properties is taken into account at where σb , pb and vb are stress tensor, fluid pressure and velocity inside
the microscale, which is fundamental to represent the real physical microcracks, respectively; ex (vb ) is rate of deformation tensor. Noted
process; (2) microcracks can have a certain arbitrary orientation, which that the crack geometry is a thin straight line which induces significant
is more aligned with the microcrack nature; (3) self-balanced unit cell anisotropy of fluid flow, where flow only takes place along the crack
problems under complex conditions are enhanced. direction.
Herein, the asymptotic homogenization upscaling scheme by32 is On the boundary of crack face, the continuity of total stress and
adopted to describe the adjacent microcrack interaction under the local pressure is reached, while the fluid pressure in matrix and crack is
periodicity assumption. Starting with a fully coupled poroelastic frac equal37:
ture problem at the microscopic scale, we use a micromechanics-based
approach to obtain the poroelastic damage model, which considers the σ ⋅ N = σb ⋅N, and pf = pb (5)
complex coupling between fluid pressure, deformation and HM prop
where N is the unit vector normal to crack face.
erties. As a result, the nonlinear anisotropic mechanical behavior,
For fluid flow condition on crack face, since parallel flow is assumed
degradation of both elastic stiffness and poroelastic properties are con
along crack direction, the fluid velocity with respect to solid deforma
structed at the macroscale, and the change of effective HM coefficients
tion is in an equilibrium state between matrix and cracks.37,38 While at
depends on the introduced damage variable. Apart from the change of
the direction normal to crack face, there is no flux taking place. There
poro-mechanical behaviors due to microcracking, the change of fluid
fore, we have the following relations along and normal to crack face:
flow behavior is implemented through damage induced permeability
change. q ⋅ e = (vb − u̇)⋅e (6)
In the remainder of the paper, we formulate the multiscale poroe
lastic model for solid body containing periodically distributed micro qi Ni = 0 (7)
cracks in Section 2. Then the damage evolution law and the
corresponding permeability change are proposed in Section 3. The in which e is unit vector along crack direction, as seen in Fig. 2.
dependence of homogenized coefficients on damage variable is analyzed
in Section 4. Numerical examples in view of demonstrating the robust 2.2. Separation of scales
ness of the multiscale damage model are presented in the last section.
In order to derive the macroscopic effective equations for the above
local problem, it is assumed that the introduced crack length and crack
spacing in Fig. 1 are significantly smaller than the macroscopic
2
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 1. (a) The macroscopic porous medium with length Lc consists of periodically distributed microcracks of length l, lc is length of micro-period. (b) Basic principles
of field f along x-axis using asymptotic expansions (up to second order). (c) Reference cell with rescaled crack length d × Lc (adapted from Dascalu et al.33;
Orgeas et al.34).
characteristic length. Such condition of scale separation is directly In the microstructure containing one straight crack that evolves sym
correlated to a scale parameter ε, which represents the difference be metrically with respect to the center point, the normalized crack length
tween the microscale and macroscale, expressed as: can be defined between the crack length and microstructural length:
lc
ε= ≪1 (8) d=
l
(9)
Lc lc
3
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
where d is the normalized crack length at microscale, also the damage Introducing such expansions (18)–(20) in problem (13)–(17) yields
variable at macroscale. to a set of boundary value problems corresponding to different orders of
Since the heterogeneous medium is assumed to consist of periodi ε over the period Y. After the analysis of each problem (more details are
cally distributed microstructure, the material response also varies peri referred to Appendix A), we can obtain the macroscopic governing
odically over the macroscale domain. The damage evolution process at equations in the following form:
macroscale can be understood in a way that the microcrack length ∂ hm ( hm )
gradually propagates from small length to large length until completely Σ = 0, with Σhm ∗
ij = Cijkl exkl u − α∗ij phm (21)
∂xj ij f
Following the methodology developed in,34 we introduce in Eqs. (1)– where the superscript ’ hm’ denotes the homogenized medium, Σhm hm
ij , Qi ,
(5) the following microscopic coordinates: uhm , phm
f are the macroscopic total stress tensor, Darcy’s velocity vector,
yi =
xi
(10) displacement vector, and fluid pressure in the homogenized medium,
ε respectively. The homogenized coefficients, e.g., stiffness tensor C∗ijkl ,
where xi and yi are the macroscopic and microscopic coordinates, Biot coefficients α∗ij and γ∗ij , Biot modulus β∗ and intrinsic permeability k∗ij
respectively. Introducing different coordinates in the domain allows us are defined by:
to scale the micro-period to a reference cell, which contains the rescaled ( )
∗
Cijkl = 〈Cijkl + Cijmn eymn ξkl 〉 (23)
crack, as seen in Fig. 2.
By means of the above coordinate transformation and the adopted
α∗ij = αδij + 〈(1 − α)Cijkl eykl (κ)〉 (24)
local periodicity condition, all the involved fields in Eqs. (1)–(4), i.e., σ ij ,
u, pf , qi , φ, σ bij , pb and vb depend on ε. To underline this dependence, they ( )
γ ∗ij = αδij − (1 − α)〈eykk ξij 〉 (25)
are denoted by σεij , uε , pεf , qεj , φε , pεb and vεb . Hence, the material properties
and responses are Y-periodic and the field f on microscale can be
β∗ = β + (1 − α)2 〈eyii (κ)〉 (26)
expressed as:
( ) ∂ζ j
f ε (xi ) ≅ f (xi , yi ) = f xi , yj + nYj n = 1, 2, … (11) kij∗ = kA∗ij with A∗ij = 〈δij + 〉 (27)
∂yi
where f ε (xi ) is the material property or response, the superscript ε de
notes that the function f ε (xi ) depends on ε, which can be defined with where the symbol 〈 ⋅〉 stands for volume average. The proof of γ∗ij = α∗ij
respect to both macroscopic and microscopic coordinates in the can be found in.37
following form32: The Y-periodic vector ξpq is the solution of the cell problem:
df ε ∂f 1 ∂f ∂ ( )
= + (12) C e (ξpq ) = 0 in Y (28)
dxi ∂xi ε ∂yi ∂yj ijkl ykl
4
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
5
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
increases very slowly when damage initiates from a small value. When 2000 0.3 0.6 0.146
6
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 4. Evolution of homogenized mechanical coefficients with respect to normalized damage variable d∗ and crack orientation θ for elastic parameters E = 2 GPa
and υ = 0.3.
4.2. Poroelastic coefficients normal to crack face. This result is consistent with the numerical anal
ysis of Biot’s coefficient with respect to the rock microstructure in.65
The homogenized Biot coefficients α∗ij and Biot modulus β∗ are pre
sented in Fig. 5 with respect to normalized damage variable d∗ and crack
4.3. Hydraulic coefficients
orientation θ. The usual symmetries hold for the coefficients: α∗ij = α∗ji .
With an increase of damage value, the upper and lower parts of the cell The homogenized hydraulic coefficients A∗ij are presented in Fig. 6
are progressively partitioned by a traction free and impermeable crack
with respect to normalized damage variable d∗ and crack orientation θ.
face. The deterioration degree of material with respect to the virgin state
The usual symmetries hold for the coefficients: A∗ij = A∗ji . The presence of
is becoming stronger and as a result the values of Biot coefficients in
microcracks leads to anisotropic hydraulic response and the hydraulic
crease gradually.
coefficients have a nonlinear dependence on the normalized damage
By comparing Figs. 4 and 5, it is noted that the principal axes of Biot
variable. When the porous matrix is totally partitioned by the crack,
coefficients tensor correspond to the direction of extreme material ri
fluid flow only occurs along the crack face and flow within the porous
gidity degradation. The associated principal values of α∗ij represent the
matrix perpendicular to the crack plane is interrupted.
extreme deterioration degree of the elastic stiffness. Therefore, the fast
increase in Biot coefficients correspond to the abrupt decrease in elastic
5. Numerical examples
coefficients, as seen from the evolution of α∗22 and C∗2222 for horizontal
crack.
To get a deeper understanding of the multiscale model, the local
As concerns the Biot modulus, the virgin material (d∗ = 0) has the
response analysis is generally needed to be conducted. This part of ex
minimum value β∗ (0) = (α − φ0 ) /Ks , where Ks is the bulk modulus of
amples has been investigated in detail in our previous works, see e.
solid grain. While the homogenized Biot modulus is increasing for a
g.,31,46 where the rate effect and size effect of the damage model are
gradually damaged porous matrix. In addition, since we focus on fluid
discussed. Thus it will not be explored in the current study, our objective
pressure acting to open crack face, it is obvious from Fig. 5 that the
is to evaluate the model performance for fracturing in heterogeneous
obtained Biot’s coefficient tensor shows significant anisotropy along and
rocks due to microcrack nucleation and propagation. In the following
7
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 5. Evolution of homogenized poroelastic coefficients with respect to normalized damage variable d∗ and crack orientation θ.
Fig. 6. Evolution of homogenized hydraulic coefficients with respect to normalized damage variable d∗ and crack orientation θ.
8
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
9
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Table 2
Parameters for simulation of direct tension test31.
E (MPa) υ (− ) α (− ) φ0 (− ) θ (◦ ) K0 [Pa•m0.5] n [− ] v0 [m/s] lc [m]
4 − 5 4
2000 0.3 0.6 0.146 0 2.31×10 3 1×10 1×10−
10
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 11. Damage band in refined mesh zone with respect to different meshing schemes.
The lower water pressure value in the center part of domain is closely
associated with fracture propagation, where a slight stress concentration Table 4
is formed ahead of fracture front and moving with hydraulic fracture Parameters used in the simulation of borehole failure.
correspondingly.
K0 [Pa•m0.5] n [− ] v0 [m/s] lc [m] k0 [m− 2] at [− ] θ [◦ ]
To validate the relation between water pressure change and fracture
4 − 5 − 4 − 18
propagation, we present the damage distribution in the domain at 2.31×10 3 1×10 1×10 5×10 3.5 90
Fig. 13. Failure characteristics in the simulation and experimental images reproduced from Hashiba and Fukui.68 Note: Red line represents the failure plane. (For
interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
11
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 15. Example of borehole failure along preferred stress direction: (a) evolution of borehole water pressure, (b) water pressure along the horizontal line at the
height of 50 mm.
penetrates the rock formation. It is clearly seen from Fig. 16 that the respect to the initial stress, see the stress profile of t = 20 s in Fig. 17(c).
hydraulic fracture has formed in a way parallel to the direction of the As the hydraulic fracture front almost reaches the cutting line, see
maximum in-situ stress. The fracture area acts as preferential pathways Fig. 17(a), the applied water pressure acts to increase the effective stress
for fluid flow directly from borehole into the formation. in both directions. Under the condition of stress difference, the hydraulic
It should be clarified here that the multiple fractures formed around fracture preferably propagates along the maximum principal stress di
the borehole may be closely associated with the effect of injection rate, rection, which induces the anisotropic distribution of Biot coefficients
the viscosity of fracturing fluid, HM properties of rock, etc. For the around the crack front. This is validated by the different Biot coefficients
multiscale model predicted single crack propagation under other con of cracked elements along horizontal and vertical directions, see the
ditions, readers are referred to our previous works in.46 Current re curve of t = 195.5 s in Fig. 17(d). As a result, the effective stress in
searches mainly relate multiple crack propagation with fracture horizontal direction increases more rapidly than that in the vertical di
spacings,75 sequencing effect,76 natural fracture distribution,77 etc. The rection, see the stress profile of t = 100 s in Fig. 17(c). This result is
effect of injection rate on multiple crack propagation needs to be further consistent with the experimentally observed anisotropic damage in the
investigated in the future work. fractured rock material.6
To study the fundamental hydraulic fracturing mechanism along When the hydraulic fracture passes the cutting line and gets to the
preferred stress direction, we present water pressure distribution in the impermeable boundary, as seen in Fig. 17(b), the no flow condition leads
fracturing area (defined by the area of d > 0.9) and effective stress to water flow back. Due to the microcracking mechanism along vertical
around the fracturing area in Fig. 17(a)-(b). When the hydraulic fracture direction, the fracturing area presents a higher permeability with sig
front is far from arriving the cutting line (t = 20 s), the effective stresses nificant anisotropy in different directions, see Fig. 17(e). Thus, the
in horizontal and vertical direction only have a minor variation with flowback water is more likely to be transported along the existed
12
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 17. Simulation results of hydraulic fracture propagates along preferred stress direction: (a) water pressure in the fracturing area (d > 0.9) and effective stress
distribution at t = 100 s, (b) water pressure and effective stress distribution at t = 195.5 s, (c) effective stress σ‘11 and σ‘22 along the line at different time, (d) effective
Biot coefficients along the cutting line at different time, (e) components of effective permeability tensor along the line at different time.
13
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
hydraulic fractures in a preferred manner. As water is continuously multiple hydraulic fractures evolve towards the wall, as shown in
injected in the borehole, the flowback water is interacting with the Fig. 20.
injected water, which leads to the significant decrease of effective To study the fundamental hydraulic fracturing mechanism along
stresses in vertical direction that is even lower than the initial stress preferred microcrack direction, we present the simulation results in
value, see the stress profile of σ‘22 (t = 195.5 s) in Fig. 17(c). Fig. 21. When the hydraulic fracture front is far from arriving the cutting
The fluid pressure at failure and dominant flow streamlines in the line (t = 20 s), the effective normal stress and shear stress only have a
domain are shown in Fig. 18. Comparison of damage distribution and minor variation with respect to the initial stress, see the stress profile of
flow streamline reflects that the developed fractures provide clear paths t = 20 s in Fig. 21(c). As the hydraulic fracture front almost reaches the
for fluid directly flow from the borehole into the rock formation. The cutting line at t = 90 s, see Fig. 21(a), the applied water pressure acts to
modelling results on fracture pattern and fluid flow qualitatively agree increase the effective stresses in normal and shear directions to the crack
with the experimental description in.57 It is noted that a negative pore face, as seen from the stress profile of t = 90 s in Fig. 21(c). Due to the
pressure is occurring ahead of the fracture tip, which is caused by the cracking mechanism of microstructure degraded by the fluid filled
water flowing back towards the borehole due to impermeable boundary cracks, the hydraulic fracture preferably propagates along the micro
conditions. Similar results have also been reported by other researchers, crack direction.
see e.g., Eghbalian et al..11 When the hydraulic fracture passes the cutting line and moves close
to the impermeable boundary in Fig. 21(b), the redistribution of pore
5.2.2. Fracturing along preferred microcrack direction pressure due to boundary effect acts to preventing the further propa
To evaluate the model performance to reproduce the borehole failure gation of hydraulic fracture as well as leading to water flowback, and as
along preferred microcrack direction, an example of similar hydraulic a result injected water is hardly transported along the cracking path,
fracturing test is conducted in the following part. In this example, the which can be validated by the negative pore water pressure in Fig. 22.
geometry and boundary conditions are the same as that in Fig. 14, Besides, the interaction between flowback water and continuously
except that σh = σH = 6 MPa and the set of microcrack is oriented in an injected water gives rise to the significant decrease in the crack normal
angle of 45∘ with respect to the horizontal direction. An initial crack of 2 stress in the hydraulic fractures, which is much lower than the initial
mm length is inserted into the geometry around the borehole at a di stress value, see the stress profile of σ‘n (t = 160 s) in Fig. 17(c).
rection of 45◦ with respect to x-axis, and the crack permeability is To illustrate the poroelastic behavior of the multiscale damage
1.0×10− 14 m2. The modelling parameters used in the test are referred to model and the dominant flow path, we present the evolution of Biot
Table 4 except that θ = 45∘◦ . coefficients along the cutting line and the distribution of water pressure
In the numerical example of borehole failure along preferred with fluid velocity in Fig. 22. By comparing Fig. 22(a) with damage
microcrack direction, the evolution of borehole water pressure and distribution in Fig. 20, the abrupt increase of Biot coefficients occurs in
water pressure along the horizontal line at the height of 50 mm are the damaged area, while the values in the non-damaged area are kept
shown in Fig. 19. The drop in borehole water pressure is due to the water constant, which shows the highly localized HM behavior. For the rock
flowing into the porous medium through induced fracture pathways, as material with low porosity, the increase of Biot coefficients in the HF is
validated by the damage distribution in Fig. 20. Water pressure along accompanied by the local drop in pore water pressure, as validated by
the horizontal line increases gradually with the process of water injec the decreased trend of pore water pressure distribution along the
tion. The pressure value around the damaged area presents an opposite cracking direction in Fig. 22(b), which is caused by the pore dilation
trend, where the positive pressure value is existed in the area away from around the fracture tip. As a consequence, the maximum water pressure
the hydraulic fractures, while the negative value is found in the area may be lower than the confining pressure, see Fig. 19(a), but the HF can
close to the hydraulic fractures, as seen from Fig. 19(b). still propagate along the preferred microcrack direction. Under this
The damage distribution in the domain is shown in Fig. 20 with circumstance, the distribution of water pressure shows a significant
respect to different time, in which the damage propagates along the anisotropic behavior, and negative pressure value occurs ahead of the
preferred microcrack direction. Due to the existence of initial crack, fracture tip due to the impermeable boundary effect, as seen in Fig. 22
more fluid penetrates this area that causes the tensile deformation of (b).
rock material. When the principal effective stress reaches the critical This example shows the dominant effect of microcrack distribution
stress that provides a driving force for fracture evolution, damage on fracture propagation behavior obviously, which is able to explain the
propagates along the initial crack area, as seen from damage distribution complex behavior of hydraulic propagation in naturally fractured rock
at t = 20 s. As constant water injection rate is applied continuously, materials, where the fracture propagation is influenced by both the
maximum stress and the microcrack distribution.
14
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 19. Example of borehole failure along preferred microcrack direction: (a) Evolution of borehole water pressure. (b) Water pressure along the horizontal line at
the height of 50 mm.
Fig. 20. Damage distribution along preferred microcrack direction in the domain.
holes. The fluid injection process continues for sufficient time so that damage model. The performed numerical analysis provides fundamental
fluid completely permeates the rock specimen. In this way, a insight into rock fracture behavior and may have a certain guidance to
non-homogeneous distribution of fluid pressure will be generated in the field application. In the shale gas exploitation with multiple fracturing
notch and through the specimen. In the second step, fluid is injected wells, the hydraulic fractures are not only influenced by the distribution
through the top hole at a constant rate q = 5 mL/min until fracture of natural cracks, but also influenced by the pore pressure gradient field
occurs and the fracture type is affected by the pore pressure distributed before fracture operation is started.
in the first step.
The parameters of rock specimen used in the simulation are listed in 6. Conclusions
Table 5. Water is chosen as the fracturing fluid and assumed to be
incompressible for simplicity. The phenomenon of macroscopic fracturing in rock materials has
The final pore fluid pressure distribution in two steps is presented in been numerically studied by incorporating the dynamically evolved
Fig. 24. The fracture path in the experiment and in the model are shown microcracks in the modelling framework through scale separation. The
in Fig. 25. It can be seen from the figures that the simulated pore fluid present approach constructing the coupled poroelastic damage model
pressure and damage distribution agree well with that in the experi appears to be the first one that proposes a multiscale description of fluid
mental observation. When the rock material is subjected to non-uniform flow-deformation due to distributed damage nucleation and propaga
pore pressure field and in the absence of stress difference conditions, the tion, which is fundamental to describe the complex fracturing behavior
generated fluid pressure gradient between the notch and left bottom influenced by microcrack distribution.
hole affects the fracture path, which is moving towards the region of Starting with a fully coupled poroelastic fracture problem at the
higher local pore pressure field. Two fracture propagation mechanisms microscale, a periodically distributed microstructure has been assumed
are observed during the simulation: one fracture path is affected by the and asymptotic homogenization method has been adopted to formulate
combined effect of microcrack distribution and local pore pressure dis the multiscale damage model. The model has considered the complex
tribution, in which the fracture initiates from the notch tip. The second coupling between fluid pressure-deformation and HM properties at the
fracture is initiating from the injection port towards the top boundary, microscale, giving rise to the nonlinear anisotropic mechanical
where the path is dominated by the microcrack distribution at vertical behavior, degradation of both elastic stiffness and poroelastic properties
direction. at the macroscale. The microstructural size has been included in the
This example clearly shows the effect of local pore fluid pressure field damage evolution law, which is able to describe the rate effect and size
on fracture propagation, which is well captured by the multiscale effect. In addition, a permeability model has been proposed accounting
15
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 21. Simulation results of hydraulic fracture propagates along preferred microcrack direction: (a) water pressure in the fracturing area (d > 0.9) and effective
stress distribution at t = 90 s, (b) water pressure and effective stress distribution at t = 160 s, (c) effective normal stress σ‘n and shear stress σ ‘s along the line at
different time.
Fig. 22. (a) Biot coefficients along the cutting line at t = 160 s, (b) water pressure with fluid velocity (represented by black arrow line) in the domain at t = 160 s.
for crack opening induced permeability change. performed for examples of direct tension test, borehole failure at the lab
The homogenized coefficients have been illustrated in detail for a scale and hydraulic fracturing in a three-hole rock specimen. The frac
given set of initial material parameters, with dependence on the turing process in heterogeneous rock materials has been shown to be a
normalized damage variable. Numerical simulations have been multiscale phenomenon that initiates from the microcrack nucleation
16
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
Fig. 23. The geometry and boundary conditions in the test of hydraulic frac Declaration of competing interest
turing in a three-hole specimen.
The authors declare that they have no known competing financial
Table 5
Parameters in the example of HF in a three-hole rock specimen.
E (MPa) υ α φ0 K0 [Pa•m0.5] n v0 [m/s] lc [m] k0 [m− 2] at θ [◦ ]
Fig. 24. The final pore fluid pressure distribution in (a) first step, (b) second step. Black arrow line denotes the major fluid flow path.
Fig. 25. (a) Fracture pattern in the experiment,56 (b) simulated fracture path.
17
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
In this appendix, the asymptotic homogenization method32 is adopted to perform the upscaling process.
Eq. (13) at the order ε− 2 and Eq. (16) at the order ε− 1 allow us to obtain the following boundary value problem:
∂ ( ( ))
C e u(0) = 0 in Y and (A.1)
∂yj ijkl ykl
( ( ))
Cijkl eykl u(0) Nj = 0 on C Y (A.2)
with periodicity condition applied on the opposite exterior boundary of cell Y. To solve problem (A.1)-(A.2), it is easily found that u = u (x,t). (0) (0)
To solve the balanced problem in the medium with fluid pressure in both pores and cracks, we first present the order ε− 1 term of Eq. (15) in the
following form:
∂p(0)
(A.3)
b (0) (0)
= 0, thus pb = pb (x, t)
∂xi
Use of the condition on crack Eq. (16), and comparing p(0)
f = p(0)
f (x, t) with Eq. (A.3), it is deduced that:
(A.4)
(0) (0)
pf = p b
Eq. (13) at the order ε− 1 and Eq. (16) at the order ε0 give a boundary value problem for the function u(1) considering that u(0) only depends on x-axis,
as follows:
∂ ( ( ))
Cijkl eykl u(1) = 0 in Y and (A.5)
∂yj
( ( ) ( ))
Cijkl exkl u(0) + eykl u(1) Nj = − (1 − α)p(0)
f Ni on C
Y
(A.6)
37
and with periodicity boundary conditions on the external boundary of the cell. The function u (1)
can be looked for in the form :
( )
(A.7)
(1) (0)
ui (x, y, t) = ξpq
i (y)expq u(0) − κi (y)(1 − α)pf
∂p(0)
(A.9)
f
N = 0 on C Y
∂yi i
with periodicity condition applied on the opposite exterior boundary of cell Y. We can easily deduce that p(0)
f = p(0)
f (x, t).
1
Introducing Eq. (14) at the order ε− and Eq. (17) at the order ε0 with the consideration of pf = pf (x,t), we can derive the following boundary
(0) (0)
value problem:
∂2 p(1)
(A.10)
f
= 0 in Y
∂y2i
( )
∂p(0) ∂p(1)
(A.11)
f f
+ Ni = 0 on C Y
∂xi ∂yi
∂p(0)
f (x, t)
p(1)
f (x, y, t) = ζ i (y) (A.12)
∂xi
The displacement problem at the order ε0 is deduced from Eqs. (13) and (16):
∂ ( ( ( ) ( )) )
(0)
Cijkl exkl u(0) + eykl u(1) − αpf δij +
∂xj
in Y (A.13)
∂ ( ( ( (1) ) ( )) )
Cijkl exkl u + eykl u(2) − αp(1)
f δij = 0
∂yj
18
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
( ( ( ) ( )) )
(A.14)
(1) (1)
Cijkl exkl u(1) + eykl u(2) − αpf δij Nj = − pb Ni on C Y
In order to obtain the effective equations, we introduce the mean value operator as follows:
∫
1
〈⋅〉= ⋅dy (A.15)
|Y|
Y
where |Y| = L2c is the area of Y. Let us integrate Eq. (A.13) over the domain Y with the help of Eqs. (A.14)-(A.15), the following relation can be
obtained:
∫
( ( ) ( )) 1
〈Cijkl exkl u(0) + eykl u(1) − αp(0)
f δij 〉 = p(1)
b Ni dΓ (A.16)
|Y|
CY
To solve the right side member of Eq. (A.16), we look for the expression of fluid pressure pεb in microcracks at the order ε0 . Using Eq. (15) at the
order ε0 by taking into account of Eq. (A.3) is allowed to derive the following relation:
∂p(0) ∂p(1)
b
+ b =0 (A.17)
∂xi ∂yi
Considering the relation (16) on crack face at the order ε0 , it is known that p(1)
f = p(1)
b . As we consider small thinness crack that occupies no volume,
integrating the above equation (A.17) inside the crack with the help of divergence theorem enables us to derive the right side of Eq. (A.16):
∫
p(1)
b Ni dΓ = 0 (A.18)
CY
Lastly, making use of Eq. (A.7) and (A.18) in (A.16) gives the effective macroscopic equilibrium equation in the form:
∂ (0)
Σ =0 (A.19)
∂xj ij
where Σij = 〈σij 〉 = C∗ijkl exkl (u(0) ) − α∗ij pf is the macroscopic total stress tensor in the microporous medium with cracks. The homogenized stiffness
(0) (0) (0)
tensor C∗ijkl and Biot coefficient tensor α∗ij are defined by:
( )
∗
Cijkl = 〈Cijkl + Cijmn eymn ξkl 〉 (A.20)
The microscale mass balance (14) and Eq. (17) at the order ε0 writes:
( ( ) ( ))
∂q(0) ∂q(1) ∂exii u(0) ∂eyii u(1) ∂p(0)
(A.22)
i f
+ i +α + +β = 0 in Y
∂xi ∂yi ∂t ∂t ∂t
q(1)
i Ni = 0 on C
Y
(A.23)
Integrating Eq. (A.22) over the cell Y and applying the divergence theorem with the help of boundary condition (A.23), we deduce the macroscopic
mass conservation law:
( )
∂Q(0) ∂exij u(0) ∂p(0)
(A.24)
i f
+ γ∗ij + β∗ =0
∂xi ∂t ∂t
k∗ij ∂pf
(0)
where Qi is the macroscopic Darcy’s velocity, the homogenized coefficients γ∗ij , Biot modulus β∗ and intrinsic permeability k∗ij are
(0) (0)
= 〈qi 〉 = − μf ∂xj
given:
( )
γ ∗ij = αδij − (1 − α)〈eykk ξij 〉 (A.25)
∂ζj
kij∗ = kA∗ij with A∗ij = 〈δij + 〉 (A.27)
∂yi
Finally, we put the derived quantities together and substitute superscript ’ hm’ for superscript ’(0)’, thus obtaining the multiscale poroelastic model in
homogenized manner, as represented by Eqs. (21) and (22).
19
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
References 37. Lewandowska J, Auriault JL. Extension of Biot theory to the problem of saturated
microporous elastic media with isolated cracks or/and vugs. Int J Numer Anal
Methods GeoMech. 2013;37:2611–2628.
1. Li W, Frash LP, Lei Z, et al. Investigating poromechanical causes for hydraulic
38. Yan X, Huang ZQ, Yao J, et al. Numerical simulation of hydro-mechanical coupling
fracture complexity using a 3D coupled hydro-mechanical model. J Mech Phys Solid.
in fractured vuggy porous media using the equivalent continuum model and
2022;169, 105062.
embedded discrete fracture model. Adv Water Resour. 2019;126:137–154.
2. Lydzba D, Shao JF. Study of poroelasticity material coefficients as response of
39. Wrzesniak A, Dascalu C, Bésuelle P. A two-scale time-dependent model of damage:
microstructure. Mech Cohesive-Frict Mater. 2000;5:149–171.
influence of micro-cracks friction. Eur J Mech Solid. 2015;49:345–361.
3. Hu DW, Zhou H, Zhang F, Shao JF. Evolution of poroelastic properties and
40. Atkinson BK, Meredith PG. The Theory of Subcritical Crack Growth with Applications to
permeability in damaged sandstone. Int J Rock Mech Min Sci. 2010;47:962–973.
Minerals and Rocks. New York: Academic Press; 1987.
4. Liu J, Qiu X, Yang J, Liang C, Dai J, Bian Y. Failure transition of shear-to-dilation
41. Swanson PL. Subcritical crack growth and other time- and environment-dependent
band of rock salt under triaxial stresses. J Rock Mech Geotech Eng. 2023.
behavior in crustal rocks. J Geophys Res. 1984;89:4137–4152.
5. Liu J, He X, Huang H, et al. Predicting gas flow rate in fractured shale reservoirs
42. Charles R. Dynamic fatigue of glass. J Appl Phys. 1958;29:1657–1662.
using discrete fracture model and GA-BP neural network method. Eng Anal Bound
43. Dascalu C. A two-scale damage model with material length. Compt Rendus Mec.
Elem. 2024;159:315–330.
2009;337:645–652.
6. Shao JF. Poroelastic behaviour of brittle rock materials with anisotropic damage.
44. Dascalu C, Bilbie G, Agiasofitou EK. Damage and size effects in elastic solids: a
Mech Mater. 1998;30:41–53.
homogenization approach. Int J Solid Struct. 2008;45:409–430.
7. Tan X, Konietzky H, Fruehwirt T. Experimental and numerical study on evolution of
45. Dascalu C, Gbetchi K. Dynamic evolution of damage by microcracking with heat
Biot’s coefficient during failure process for brittle rocks. Rock Mech Rock Eng. 2015;
dissipation. Int J Solid Struct. 2019;174:128–144.
48:1289–1296.
46. Yang J, Fall M. A two-scale hydro-mechanical-damage model for simulation of
8. Kranz RL. Microcracks in rocks - a review. Tectonophysics. 1983;100:449–480.
preferential gas flow in saturated clayey host rocks for nuclear repository. Comput
9. Ougier-Simonin A, Renard F, Boehm C, Vidal-Gilbert S. Microfracturing and
Geotech. 2021;138, 104365.
microporosity in shales. Earth Sci Rev. 2016;162:198–226.
47. Guo G, Fall M. A thermodynamically consistent phase field model for gas transport
10. Sprunt ES, Brace WF. Direct observation of microcavities in crystalline rocks. Int J
in saturated bentonite accounting for initial stress state. Transport Porous Media.
Rock Mech Min Sci. 1974;11:139–150.
2021;137:157–194.
11. Eghbalian M, Wan R, Pouragha M, Fung LS. Numerical modeling of fracturing in
48. You T, Zhu QZ, Li PF, Shao JF. Incorporation of tension-compression asymmetry into
permeable rocks via a micromechanical continuum model. Int J Numer Anal Methods
plastic damage phase-field modeling of quasi brittle geomaterials. Int J Plast. 2020;
GeoMech. 2019;43:1885–1915.
124:71–95.
12. Jain AK, Juanes R. Preferential Mode of gas invasion in sediments: grain-scale
49. Guo G, Fall M. Modelling of preferential gas flow in heterogeneous and saturated
mechanistic model of coupled multiphase fluid flow and sediment mechanics.
bentonite based on phase field method. Comput Geotech. 2019;116, 103206.
J Geophys Res Solid Earth. 2009;114(19), B08101.
50. François B, Dascalu C. A two-scale time-dependent damage model based on non-
13. Vassal JP, Orgeas L, Favier D, Auriault JL, Le Corre S. Upscaling the diffusion
planar growth of micro-cracks. J Mech Phys Solid. 2010;58:1928–1946.
equations in particulate media made of highly conductive particles. II. Application
51. Cuss R, Harrington J, Giot R, Auvray C. Experimental observations of mechanical
to fibrous materials. Phys Rev E. 2008;77, 011303.
dilation at the onset of gas flow in Callovo-Oxfordian claystone. Geol Soc Spec Publ.
14. Hazzard JF, Young RP, Maxwell SC. Micromechanical modeling of cracking and
2014;400:507–519.
failure in brittle rocks. J Geophys Res Solid Earth. 2000;105:16683–16697.
52. Yang J, Fall M. Coupled hydro-mechanical modelling of dilatancy controlled gas
15. Potyondy DO, Cundall PA. A bonded-particle model for rock. Int J Rock Mech Min
flow and gas induced fracturing in saturated claystone. Int J Rock Mech Min Sci.
Sci. 2004;41:1329–1364.
2021;138, 104584.
16. Duriez J, Scholtes L, Donze FV. Micromechanics of wing crack propagation for
53. Yang J, Fall M. A dual porosity poroelastic model for simulation of gas flow in
different flaw properties. Eng Fract Mech. 2016;153:378–398.
saturated claystone as a potential host rock for deep geological repositories. Tunn
17. Fu P, Johnson SM, Carrigan CR. An explicitly coupled hydro-geomechanical model
Undergr Space Technol. 2021;115, 104049.
for simulating hydraulic fracturing in arbitrary discrete fracture networks. Int J
54. Carman PC. Fluid flow through granular beds. Trans Inst Chem Eng. 1937;15:
Numer Anal Methods GeoMech. 2013;37:2278–2300.
150–166.
18. Yuan SC, Harrison JP. A review of the state of the art in modelling progressive
55. Witherspoon PA, Wang JSY, Iwai K, Gale JE. Validity of Cubic Law for fluid flow in a
mechanical breakdown and associated fluid flow in intact heterogeneous rocks. Int J
deformable rock fracture. Water Resour Res. 1980;16:1016–1024.
Rock Mech Min Sci. 2006;43:1001–1022.
56. Bruno MS, Nakagawa FM. Pore pressure influence on tensile fracture propagation in
19. Tang CA, Tham LG, Lee PKK, Yang TH, Li LC. Coupled analysis of flow, stress and
sedimentary-rock. Int J Rock Mech Min Sci Geomech Abstr. 1991;28:261–273.
damage (FSD) in rock failure. Int J Rock Mech Min Sci. 2002;39:477–489.
57. Zoback MD, Rummel F, Jung R, Raleigh CB. Laboratory hydraulic fracturing
20. Zhu WC, Tang CA. Micromechanical model for simulating the fracture process of
experiments in intact and pre-fractured rock. Int J Rock Mech Min Sci. 1977;14:
rock. Rock Mech Rock Eng. 2004;37:25–56.
49–58.
21. Basista M, Gross D. The sliding crack model of brittle deformation: an internal
58. Fall M, Nasir O, Nguyen TS. A coupled hydro-mechanical model for simulation of
variable approach. Int J Solid Struct. 1998;35:487–509.
gas migration in host sedimentary rocks for nuclear waste repositories. Eng Geol.
22. Krajcinovic D, Fonseka GU. The continuous damage theory of brittle materials. I.
2014;176:24–44.
General theory. Transactions of the ASME. J Appl Mech. 1981;48:809–815.
59. Olivella S, Alonso EE. Gas flow through clay barriers. Geotechnique. 2008;58:
23. Lyakhovsky V. Scaling of fracture length and distributed damage. Geophys J Int.
157–176.
2001;144:114–122.
60. Guo G, Fall M. Modelling of dilatancy-controlled gas flow in saturated bentonite
24. Tao J, Shi A-C, Li H-T, Zhou J-W, Yang X-G, Lu G-D. Thermal-mechanical modelling
with double porosity and double effective stress concepts. Eng Geol. 2018;243:
of rock response and damage evolution during excavation in prestressed geothermal
253–271.
deposits. Int J Rock Mech Min Sci. 2021;147, 104913.
61. Ma J. Review of permeability evolution model for fractured porous media. J Rock
25. Chiarelli AS, Shao JF, Hoteit N. Modeling of elastoplastic damage behavior of a
Mech Geotech Eng. 2015;7:351–357.
claystone. Int J Plast. 2003;19:23–45. Pii s0749-6419(01)00017-1.
62. Chen D, Pan Z, Shi JQ, Si G, Ye Z, Zhang J. A novel approach for modelling coal
26. Shao JF, Zhou H, Chau KT. Coupling between anisotropic damage and permeability
permeability during transition from elastic to post-failure state using a modified
variation in brittle rocks. Int J Numer Anal Methods GeoMech. 2005;29:1231–1247.
logistic growth function. Int J Coal Geol. 2016;163:132–139.
27. Lu YL, Elsworth D, Wang LG. Microcrack-based coupled damage and flow modeling
63. Mahjoub M, Rouabhi A, Tijani M, et al. Numerical study of callovo-oxfordian
of fracturing evolution in permeable brittle rocks. Comput Geotech. 2013;49:
argillite expansion due to gas injection. Int J GeoMech. 2018;18, 04017134.
226–244.
64. ANDRA. Référentiel du comportement THM des formations sur le site de Meuse/Haute-
28. Shen WQ, Shao JF. A micromechanical model of inherently anisotropic rocks.
Marne. France: Chatenay-Malabry; 2012.
Comput Geotech. 2015;65:73–79.
65. Tan X, Konietzky H. Numerical study of variation in Biot’s coefficient with respect to
29. Xie N, Zhu QZ, Shao JF, Xu LH. Micromechanical analysis of damage in saturated
microstructure of rocks. Tectonophysics. 2014;610:159–171.
quasi brittle materials. Int J Solid Struct. 2012;49:919–928.
66. Dai J, Liu J, Zhou L, He X. Crack pattern recognition based on acoustic emission
30. Zhu QZ, Kondo D, Shao JF. Micromechanical analysis of coupling between
waveform features. Rock Mech Rock Eng. 2022;56:1063–1076.
anisotropic damage and friction in quasi brittle materials: role of the
67. Gao Q, Tao JL, Hu JY, Yu X. Laboratory study on the mechanical behaviors of an
homogenization scheme. Int J Solid Struct. 2008;45:1385–1405.
anisotropic shale rock. J Rock Mech Geotech Eng. 2015;7:213–219.
31. Yang J, Fall M. A two-scale time dependent damage model for preferential gas flow
68. Hashiba K, Fukui K. Effect of water on the deformation and failure of rock in
in clayey rock materials. Mech Mater. 2021;158, 103853.
uniaxial tension. Rock Mech Rock Eng. 2015;48:1751–1761.
32. Sánchez-Palencia E. Non-homogeneous Media and Vibration Theory. Berlin: Springer;
69. Jin ZF, Li WX, Jin CR, Hambleton J, Cusatis G. Anisotropic elastic, strength, and
1980.
fracture properties of Marcellus shale. Int J Rock Mech Min Sci. 2018;109:124–137.
33. Dascalu C, François B, Keita O. A two-scale model for subcritical damage
70. Liao JJ, Yang MT, Hsieh HY. Direct tensile behavior of a transversely isotropic rock.
propagation. Int J Solid Struct. 2010;47:493–502.
Int J Rock Mech Min Sci. 1997;34:837–849.
34. Orgeas L, Geindreaua C, Auriault JL, Bloch JF. Upscaling the flow of generalised
71. Lee YK, Pietruszczak S. Tensile failure criterion for transversely isotropic rocks. Int J
Newtonian fluids through anisotropic porous media. J Non-Newtonian Fluid Mech.
Rock Mech Min Sci. 2015;79:205–215.
2007;145:15–29.
72. Jaeger JC. Shear failure of anistropic rocks. Geol Mag. 1960;97:65–72.
35. Biot MA. General theory of three-dimensional consolidation. J Appl Phys. 1941;12:
73. Bhat HS, Rosakis AJ, Sammis CG. A micromechanics based constitutive model for
155–164.
brittle failure at high strain rates. J Appl Mech-Trans ASME. 2012;79(12), 031016.
36. Coussy O. Poromechanics. Chichester, England: John Wiley & Sons; 2004.
74. Atiezo MK, Dascalu C. Antiplane two-scale model for dynamic failure. Int J Fract.
2017;206:195–214.
20
J. Yang et al. International Journal of Rock Mechanics and Mining Sciences 175 (2024) 105676
75. Gunaydin D, Peirce AP, Bunger AP. Laboratory experiments contrasting growth of 76. Wang H. Numerical investigation of fracture spacing and sequencing effects on
uniformly and nonuniformly spaced hydraulic fractures. J Geophys Res Solid Earth. multiple hydraulic fracture interference and coalescence in brittle and ductile
2021;126, e2020JB020107. reservoir rocks. Eng Fract Mech. 2016;157:107–124.
77. Wang H. Hydraulic fracture propagation in naturally fractured reservoirs: complex
fracture or fracture networks. J Nat Gas Sci Eng. 2019;68, 102911.
21