You are on page 1of 21

Computers and Geotechnics 157 (2023) 105329

Contents lists available at ScienceDirect

Computers and Geotechnics


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

Research paper

Fracture initiation and propagation in the lined underground caverns for


compressed air energy storage: Coupled thermo-mechanical phase-field
modeling
Caichu Xia a , Yingjun Xu b , Shuwei Zhou b ,∗, Shikang Qin b , Xiuhan He b
a
Institute of Rock Mechanics, Ningbo University, Ningbo 315211, PR China
b Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092, PR China

ARTICLE INFO ABSTRACT

Keywords: In this study, to investigate the mechanism of fracture initiation and propagation in the CAES caverns, a
Compressed air energy storage coupled thermo-mechanical phase-field model (PFM) considering cavern excavation is proposed. The proposed
Phase-field model PFM is coupled with the thermodynamic process of CAES and solved in COMSOL Multiphysics. To verify
Tensile fracture
the capability and accuracy of the numerical model for predicating the tensile cracks of the CAES caverns,
Cavern excavation
the numerical results obtained by using the proposed model are compared with the existing analytical and
Thermo-mechanical coupling
numerical results. The influence of the lateral pressure coefficient, critical energy release rate, elastic modulus
of rock mass and buried depth on the fracture patterns and critical internal pressure is investigated. The results
indicate that the fracture patterns are mainly related to the lateral pressure coefficient. As the lateral pressure
coefficient of rock mass increases, the tensile fractures gradually deflect to the horizontal direction. With the
increase in the lateral pressure coefficient, critical energy release rate, elastic modulus and buried depth, the
critical internal pressure inducing the tensile fractures of the CAES caverns increases.

1. Introduction lined underground caverns for CAES. The study on the mechanisms
of fracture initiation and propagation of lined underground caverns
Compressed air energy storage (CAES) has been increasingly inves- for CAES is of great theoretical and practical significance for CAES
tigated compared with conventional large-scale energy storage tech- engineering.
niques (Zhou et al., 2017; Kim et al., 2016). This technique uses excess To the authors’ best knowledge, there have not been studies on
electric energy to store compressed air and generate electricity when fracture initiation and propagation of lined underground caverns for
needed, which is an effective way to solve intermittency and instability CAES. Some most relevant studies can be seen in Kim et al. (2012a),
of renewable energy. Lined underground caverns can be a good choice Jongpradist et al. (2015), Tunsakul et al. (2013, 2014, 2018), Thon-
for storing high-pressure compressed air for CAES (Kim et al., 2013; graksa et al. (2023) and Perazzelli and Anagnostou (2018). Kim et al.
Zhou et al., 2015; Perazzelli and Anagnostou, 2016; Jiang et al., 2020), (2012a) proposed a concept model for uplift failure to investigate the
because the large underground space can be perfectly utilized while the safety factor of the pressured gas storage caverns under high internal
special geological formations such as the rock salt adopted for the two
pressure. Jongpradist et al. (2015) and Tunsakul et al. (2013) imple-
commercial CAES power stations can be well avoided (Guo et al., 2016;
mented a physical model test to investigate the rock failure mechanism
Raju and Khaitan, 2012). However, cavern stability is still a challenging
induced by high internal air pressure. In addition, they also proposed
issue for construction and operation of CAES in lined underground
a numerical model based on the element-free Galerkin (EFG) method
caverns.
to simulate the fracture propagation pattern of the rock mass around
For underground CAES, a large amount air is injected into the
the cavern under high internal pressure (Tunsakul et al., 2014, 2018;
cavern, which induces high tensile stress in the concrete lining and
rock mass. The frequently injected air also brings great temperature Thongraksa et al., 2023). These studies are of great significance in
variation around the cavern. The high tensile stress and varying tem- reveling the failure mechanism of underground cavern subjected to
perature field can cause macro fracture initiation and propagation high internal pressure, and in their studies, the in-situ stress ratio has
in the surrounding materials, which severely affects the stability of a great influence on rock failure patterns. Perazzelli and Anagnostou

∗ Corresponding author.
E-mail address: shuwei_zhou@tongji.edu.cn (S. Zhou).

https://doi.org/10.1016/j.compgeo.2023.105329
Received 3 December 2022; Received in revised form 8 February 2023; Accepted 12 February 2023
Available online 24 February 2023
0266-352X/© 2023 Elsevier Ltd. All rights reserved.
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

2015; Zhou et al., 2018b; Zhuang and Zhou, 2020; Zhou et al., 2021;
Nomenclature Yu et al., 2021; Zhu et al., 2022b). Being firstly proposed from the
𝑉 Volume of the CAES cavern variational formulation (Francfort and Marigo, 1998) and regularized
𝜌 Air density at any time by Bourdin et al. (2000), the phase field method has been applied to
many complex fracture issues such as brittle fractures (Zhang et al.,
𝑚̇ 𝑐 Air flow rate in the stage of charging
2017b; Zhou et al., 2020a), ductile fractures (You et al., 2021, 2020),
𝑇𝑖 Temperature of the injected air
dynamic fractures (Borden et al., 2012; Zhou et al., 2018a), hydraulic
𝑇𝑅𝐿 Temperature of cavern surface fractures (Mikelic et al., 2015; Shiozawa et al., 2019; Zhou et al.,
𝑇 Temperature of concrete lining or host rock 2020b; Zhuang et al., 2022), thermal fractures (Miehe et al., 2015;
𝑍 Compressibility factor of the air Liu et al., 2020; Mandal et al., 2021; Yu et al., 2021; Cheng et al.,
𝑐𝑣0 Specific heat at constant volume of air 2022b; Zhou et al., 2022b) and rock fractures (Zhang et al., 2017a;
𝑐𝑝 Specific heat at constant pressure of con- Zhou et al., 2019; Fei and Choo, 2021; Liu et al., 2022; Zhu et al.,
crete lining or host rock 2022a). The phase field method uses an extra scalar field to smear the
𝑘 Thermal conductivity of concrete lining or sharp fracture surface and transforms the fracture problem into a multi-
host rock field coupling problem without extra work on fracture surface tracking,
𝒒 Heat flux through the cavern surface which is a huge advantage on fracture modeling in underground CAES
engineering. In addition, the phase field method can be easily extended
𝑝𝑐 Air pressure at critical state
to a problem with more physical fields. However, to the authors’ best
𝐹𝑖 State function of charging
knowledge, no study has adopted the phase field method to model
𝜓 Phase field scalar fractures of CAES caverns. Therefore, the suitability of PFM in CAES
𝑙0 Length scale parameter caverns is also the focus of this study.
𝐴𝑐 Cavern surface area The purpose of this study is to propose a thermo-mechanical phase-
𝜌0 Air density at initial state field model considering the effect of excavation and apply this model
𝑇𝑎𝑖𝑟 Air temperature inside the cavern to investigate the fracture initiation and propagation mechanism of
𝑇𝑐 Air temperature at critical state lined CAES caverns. The thermodynamic process of the CAES caverns,
𝑇0 Initial temperature of concrete lining or displacement field, temperature field and phase field are coupled and
host rock solved in COMSOL Multiphysics. Some numerical cases are presented
to verify the used phase field model. Finally, the initiation and propa-
𝑅 Gas constant of air
gation mechanisms of macro-fractures in the CAES caverns are revealed
𝑍𝑇 0 Derivative of 𝑍 with respect to 𝑇 at 𝑇0
while the effects of some influencing factors are investigated.
𝑐𝑝0 Specific heat at constant pressure of air The paper is organized as follows. Section 2 establishes the govern
ℎ𝑐𝑣 Coefficient of heat exchange equations of all physical fields involved. Section 3 shows the couple
𝑄̇ Heat transfer rate across the cavern surface relationship of all physical fields and the numerical implementation
𝑝 Air pressure inside the cavern algorithm before Section 4 verifies our model by comparing with previ-
𝛼𝑇 Thermal expansion coefficient of concrete ous numerical, analytical, and experimental studies. Section 5 discusses
lining or host rock the thermodynamic responses and fracture mechanisms of the CAES
𝐹𝑒 State function of discharging cavern. Section 6 gives some important conclusions based on the model
𝐻 History strain field presented in this study.
𝐺𝑐 Critical energy release rate
2. Governing equations for fracture modeling in CAES caverns

2.1. Governing equations for air state


(2018) investigated the uplift failure mechanism in pressured rock
The operation cycle of CAES has four typical stages, namely, charg-
tunnel by using the method of upper bound limit analysis. However,
ing, first storage, discharge and second storage. The high-pressure air
these studies are not focused on fracture mechanisms and not on the
inside the cavern satisfies the mass conservation and energy conserva-
CAES caverns while the effect of temperature is not considered.
tion regardless of the negligible air leakage, as follows (Zhou et al.,
Numerical methods are effective in predicting fracture mechanisms
in materials and can be used to study fracture initiation and prop- 2015, 2017):
agation in lined underground caverns for CAES. The fracture mod- d𝜌 ( )
𝑉 = 𝐹𝑖 + 𝐹𝑒 𝑚̇ 𝑐 (1)
eling methods can be classified into continuous and discontinuous d𝑡 ( )
2
approaches. For example, widely used rock failure progress analysis d𝑇 ( ) 𝑅𝑇 𝑍𝑇 0
(RFPA) (Wang et al., 2012; Zhao et al., 2019; Zhang et al., 2021), 𝑉 𝜌𝑐𝑣0 𝑎𝑖𝑟 = 𝐹𝑖 𝑚̇ 𝑐 𝑐𝑝0 𝑇𝑖 − 𝑇𝑎𝑖𝑟 + 𝑍0 𝑅0 𝑇𝑎𝑖𝑟 + 𝜌 0
d𝑡 𝜌0
peridynamics (Ren et al., 2017; Gao et al., 2020; Song and Menon, ( )
2019; Zhou et al., 2022a) are continuous approaches. The discrete 𝑅𝑇02 𝑍𝑇 0
+ 𝐹𝑒 𝑚̇ 𝑐 𝑍0 𝑅𝑇𝑎𝑖𝑟 + 𝜌 + 𝑄̇ (2)
element method (DEM) (Yang et al., 2014; Hu et al., 2021; Zhu et al., 𝜌0
2022c), extend finite method (XFEM) (Jin et al., 2017; Shi and Liu,
2021; Cheng et al., 2022a), cracking particle method(CPM) (Rabczuk where 𝑉 is the total volume of the cavern; 𝜌 and 𝜌0 are air density
and Belytschko, 2004, 2007; Rabczuk et al., 2010) and discontinuous and initial air density, respectively; 𝑚̇ 𝑐 denotes air mass rate in the
deformation analysis (DDA) (Chen and Wu, 2018; Zhang and Shi, stage of charging; 𝑐𝑣0 and 𝑐𝑝0 are specific heat at constant volume and
2021) belong to the discontinuous approaches. However, the above- specific heat at constant pressure of the air, respectively; 𝑍0 denotes
mentioned fracture methods have some unavoidable difficulties in compressibility factor of air at initial state; 𝑍𝑇 0 is the derivative of the
modeling fractures in underground CAES engineering because of the air compression function 𝑍 with respect to temperature at 𝑇0 ; 𝑇𝑎𝑖𝑟 and
features of thermo-mechanical coupling, high air pressure and stress, 𝑇𝑖 represent the temperature of the air in the cavern and the injected
and complex fracture surface tracking. air, respectively; 𝑇0 represents the initial temperature of host rock and
The recently widely concerned continuous approach–phase field concrete lining; 𝑅 is the air constant; 𝑄̇ represents the heat exchange
model (PFM) seems a potential numerical tool for modeling frac- rate between the air in the cavern and the concrete lining; 𝐹𝑖 and 𝐹𝑒
ture problems in lined underground caverns for CAES (Msekh et al., define the states of air injection and air extraction in the cavern.

2
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

where 𝛹𝑒 represents the stored elastic energy; 𝛹𝑓 represents the fracture


surface energy; 𝛹𝑒𝑥𝑡 is the work by external forces; 𝛹0 is the elastic
energy induced by initial geostress field; 𝒃 represents the body force;
𝝈 0 represents the initial geostress field; 𝜺 represent the strain field.
Using the strain spectrum decomposition method, the total elastic
strain is decomposed into tensile strain and compressive strain in the
following form:

𝑑
⟨ 𝑒⟩
𝜺𝑒± = 𝜀𝑎 ± 𝒏𝑎 ⊗ 𝒏𝑎 (7)
𝑎=1

Considering the thermo-mechanical model, the elastic strain is equal


to the difference between total strain and thermal strain:
1[ ]
𝜺𝒆 = ∇𝒖 + (∇𝒖)𝑇 − 𝜺𝑇 (8)
2
where 𝜺𝑇 denotes the thermal strain.
The stored elastic energy 𝛹𝑒 in body is decomposed into two parts of
𝛹𝑒+ and 𝛹𝑒− which are induced by tensile strain and compressive strain
respectively. Now, considering that the positive elastic energy 𝛹𝑒+ is
controlled by the phase variable, we can obtain:
Fig. 1. A typical compressed air energy storage cycle.
𝛹𝑒 (𝜀) = 𝑔 (𝜓) 𝛹𝑒+ + 𝛹𝑒− (9)
[ 2
]
where 𝑔(𝜓) is the degradation function and 𝑔(𝜓) = (1 − 𝑘)(1 − 𝜓) + 𝑘 .
As shown in Fig. 1, in a CAES cycle, 0∼ 𝑡1 denotes the charging 𝑘 = 10−9 is used as a stability parameter to prevent numerical singular-
stage; 𝑡1 ∼ 𝑡2 denotes the storage stage after charging (the first storage ity.
stage); 𝑡2 ∼ 𝑡3 denotes the discharging stage; 𝑡3 ∼ 𝑡4 denotes the storage The positive elastic energy density and negative elastic energy
stage after discharging (the second storage stage). 𝜉𝑐𝑑 is the ratio of density can be expressed as:
charging mass rate to discharging mass rate. 𝜆 ( )
Air pressure in the cavern can be determined by using the gas state 𝛹𝑒± = ⟨𝑡𝑟 (𝜀)⟩2± + 𝜇𝑡𝑟 𝜀2± (10)
2
equation: where 𝜆 and 𝜇 are Lamé constants.
𝑝 = 𝜌𝑍𝑅𝑇𝑎𝑖𝑟 (3) The elastic energy 𝛹0 induced by initial geostress field is expressed
as:
where 𝑍 is the compressibility factor of high-pressure air (Zhou et al.,
𝝈 0 ∶ 𝜺dΩ = 𝑔 (𝜓) 𝝈 0 ∶ 𝜺dΩ (11)
2015): ∫𝛺∕𝛤 ∫𝛺
( )
9𝑝 𝑇𝑐 6𝑇𝑐2 where 𝑔(𝜓) is the degradation function, referred to as Eq. (9); 𝝈 0 is the
𝑍 =1− −1 (4) initial geostress tensor.
128𝑝𝑐 𝑇𝑎𝑖𝑟 𝑇 2
𝑎𝑖𝑟
The variational method is applied to the total potential energy
with 𝑝𝑐 and 𝑇𝑐 being the pressure and temperature of air at critical shown in Eq. (6). The first variation of the energy functional reads:
state. The heat exchange rate between the air inside the cavern and
the cavern inner surface can be written as:
( ) 𝛿𝐿 (𝑢, 𝑇 , 𝜓, ∇𝜓) = 0 (12)
𝑄̇ = ℎ𝑐𝑣 𝐴𝑐 𝑇𝑅𝐿 − 𝑇𝑎𝑖𝑟 (5)
According to Euler–Lagrange equation, and considering the irre-
where ℎ𝑐𝑣 represents the heat exchange coefficient between the air and versibility of phase field evolution, Eq. (12) achieves governing equa-
concrete lining; 𝐴𝑐 represents the heat transfer area between air and the tions of strong form:
cavern inner surface; 𝑇𝑅𝐿 represents the temperature of cavern inner ( )
⎧ 𝐃𝐢𝐯 𝝈 𝑡𝑜𝑙 + 𝒃 = 0
surface. ⎪[ ]
⎨ 2𝑙0 (1 − 𝑘) 𝐻 2𝑙 (1 − 𝑘) 𝐻 (13)
⎪ + 1 𝜓 − 𝑙02 ∇2 𝜓 = 0
⎩ 𝐺𝑐 𝐺𝑐
2.2. Governing equations for thermo-mechanical coupled phase field model
where 𝐻 represents the maximum positive strain energy during the
loading history; 𝝈 𝑡𝑜𝑙 is the total stress tensor. They are given as follows:
Fig. 2 considers an elastic body where 𝜕𝛺 and 𝛤 denote the ex-
ternal boundary and fracture surface respectively. The external bound- { }

ary includes two parts, 𝜕𝛺𝑡 and 𝜕𝛺𝑢 , while 𝜕𝛺 = 𝜕𝛺𝑡 𝜕𝛺𝑢 and 𝐻 (𝑥, 𝑡) = max 𝛹𝜀+ (𝑥, 𝑠) + 𝝈 0 ∶ 𝜺𝑒 (𝑥, 𝑠) (14)
⋂ 𝑠∈[0,𝑡]
𝜕𝛺𝑡 𝜕𝛺𝑢 = ∅. The external force 𝒇 is applied on 𝜕𝛺𝑡 and the
and
prescribed displacement vector 𝒖 is applied on 𝜕𝛺𝑢 .
We establish a new phase field model for thermo-elastic problems. 𝝈 𝑡𝑜𝑙 = 𝑔 (𝜓) 𝝈 𝟎 + 𝐃 ∶ 𝜺𝒆 (15)
The total potential energy 𝐿 in the elastic body 𝛺 is written as:
( ) where 𝐃 is the elastic matrix; 𝜺𝑒 is the elastic strain, as shown in Eq. (8).
1 2 𝑙0 The constitutive equation of the displacement field can be expressed
𝐿= 𝛹𝑒 (𝜺) dΩ + 𝐺𝑐 𝜓 + |∇𝜓|2 dΩ
∫𝛺 ∫𝛺 2𝑙0 2 as:
⏟⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏟ ⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
𝛹𝑒 𝛹𝑓 [ ] [ ⟨ ( )⟩ ] ⟨ ( )⟩
𝝈 = (1 − 𝑘) (1 − 𝜓)2 + 𝑘 𝜆 𝑡𝑟 𝜀𝑒 + 𝑰 + 𝜀𝑒+ + 𝜆 𝑡𝑟 𝜀𝑒 − 𝑰 + 2𝜇𝜀𝑒−
− 𝒃 ⋅ 𝒖dΩ − 𝒇 ⋅ 𝒖𝑑𝜕𝛺 − 𝝈 ∶ 𝜺dΩ (6)
∫𝛺 ∫𝜕𝛺 ∫𝛺∕𝛤 0 (16)
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟ ⏟⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏟
𝛹𝑒𝑥𝑡 𝛹0 where 𝑰 is a unit tensor.

3
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 2. Phase field approximation of the fracture surface.

The elasticity tensor 𝐃 is given by: value 𝑘0 . In the fully broken zone (𝑐2 < 𝜓 < 𝑐1 ), the thermal conduc-
𝜕𝝈 {[ ] ( ( )) ( ( ))} tivity 𝑘𝑓 of calculated domain is defined as a value close to zero. In
𝐃 = = 𝜆 (1 − 𝑘) (1 − 𝜓)2 + 𝑘 𝐻𝜀 𝑡𝑟 𝜺𝒆 + 𝐻𝜀 −𝑡𝑟 𝜺𝒆 𝑱 the transition zone, the thermal conductivity 𝑘𝑟𝑓 of calculated domain
𝜕𝜺𝒆
{ } is defined by Eq. (20). We adopt the linear interpolation method to
[ ] 𝜕𝜺𝑒+ 𝜕𝜺𝑒−
+ 2𝜇 (1 − 𝑘) (1 − 𝜓)2 + 𝑘 + (17) obtain a continuous distribution of thermal conductivity in the whole
𝜕𝜺𝑒 𝜕𝜺𝑒
calculation domain. The interpolation functions 𝜒𝑟 and 𝜒𝑓 are defined
where 𝐻𝜀 (𝑥) is the Heaviside function with 𝐻𝜀 (𝑥) = 1 for 𝑥 > 0 while as:
𝐻𝜀 (𝑥) = 0 for 𝑥 ≤ 0; 𝑱 is a fourth order tensor.
The Dirichlet boundary conditions and Neumann boundary condi- ⎧ 1, 𝜓 ≤ 𝑐1
⎪ 𝑐 −𝜓
tions of the governing equations for displacement field and phase field 𝜒𝑟 (⋅, 𝜓) = ⎨ 𝑐 2−𝑐 𝑐1 < 𝜓 < 𝑐 2 (21)
2 1
are as follows: ⎪ 0, 𝜓 ≥ 𝑐2

⎧ 𝝈 𝑡𝑜𝑙 ⋅ 𝒏 = 𝒇 𝜕𝛺𝑡 ⎧ 0, 𝜓 ≤ 𝑐1
⎪ ⎪ 𝜓−𝑐
⎨𝒖 (𝑥, 𝑡) = 𝒖0 𝜕𝛺𝑢 (18) 𝜒𝑓 (⋅, 𝜓) = ⎨ 𝑐 −𝑐1 𝑐1 < 𝜓 < 𝑐 2 (22)
⎪ 2 1
⎪ 1,
⎩ ∇𝜓 ⋅ 𝒏 = 0 𝜕𝛺 ⎩ 𝜓 ≥ 𝑐2
where 𝒏 is the outward normal vector of 𝜕𝛺. where 𝑐1 and 𝑐2 are two phase field threshold values that define the
transition domain for the temperature field.
2.3. Governing equations for temperature field Heat exchange occurs on the surface between the air inside cavern
and concrete lining during typical CAES cycles. Therefore, the heat
For the underground caverns for CAES, heat transfer in the rock transfer equation obeys the following boundary condition:
mass and concrete lining are governed by: ( )
−𝑘𝑟𝑓 ∇𝑇 = ℎ𝑐𝑣 𝑇 − 𝑇𝑎𝑖𝑟 , 𝑆 = 𝑆𝑛 (23)
𝜕𝑇
𝜌𝑐𝑝 − 𝑘𝑟𝑓 ∇2 𝑇 = 0 (19)
𝜕𝑡 where 𝑆𝑛 represents the cavern inner surface.
where 𝜌 represents the density of concrete lining or host rock; 𝑐𝑝 is the Dirichlet boundary 𝑆𝑑 for solving the governing equation is as
specific heat at constant pressure of concrete lining or host rock; 𝑘𝑟𝑓 follows:
is thermal conductivity of concrete lining or host rock; 𝑇 denotes the
temperature of concrete lining or host rock. 𝑇 = 𝑇0 , 𝑆 = 𝑆𝑑 (24)
In the proposed thermo-mechanical phase field model for the CAES where 𝑇0 is the initial temperature of the concrete lining and rock mass
caverns, the thermal conductivity of concrete lining or host rock can before CAES operation.
be approximated by an interpolation function with respect to the phase At the initial state, it is assumed that the temperature of the concrete
field evolution variable 𝜓, which is determined from the intact and fully lining and rock mass is a constant:
broken states (Zhou et al., 2022b):
𝑇 = 𝑇0 , 𝑡=0 (25)
𝑘𝑟𝑓 = 𝑘0 𝜒𝑟 + 𝑘𝑓 𝜒𝑓 (20)

where 𝑘𝑟𝑓 represents the thermal conductivity of concrete lining or host 3. Numerical implementation
rock; 𝑘0 and 𝑘𝑓 represent the thermal conductivity of concrete lining
or host rock in an intact state and fully broken state, respectively; 𝜒𝑟 3.1. Finite element discretization
and 𝜒𝑓 are the interpolation functions.
We divide the conductive medium into three zones, namely, the in- For numerical implementation, the equations of stress equilibrium,
tact zone 𝛺𝑖 , the transition zone 𝛺𝑡 and the fracture zone 𝛺𝑓 according phase field evolution and heat conduction are solved by using the finite
to the value of phase field variable 𝜓. In the intact zone (𝜓 < 𝑐1 ), we element method, while the mass conservation and energy conservation
consider the thermal conductivity of calculated domain as an initial equations are solved by using the finite difference method because they

4
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

are ODEs. The weak forms of the governing equations for displacement in which
field 𝒖, phase field 𝜓 and temperature field 𝑇 can be written as ∑
𝑲𝑢 = 𝑩 T 𝑫𝑩 𝑢 d𝛺
∫𝛺 𝑢
𝑒
𝝈 𝑡𝑜𝑙 ∶ 𝛿𝜺d𝛺 − 𝒃𝛿𝒖d𝛺 − 𝒇 𝛿𝒖d𝛺 = 0 (26) ( ) (37)
∫𝛺 ∫𝛺 ∫𝜕𝛺𝑡 ∑
T
𝑷𝑢 = − 𝑩 𝑫𝜀𝑇 𝑰 0 d𝛺 + 𝑵 𝒃d𝛺 + 𝑵 𝒇 d𝑆
( ) 𝑒
∫𝛺 𝑢 ∫𝛺 𝑢 ∫𝜕𝛺𝑡 𝑢
1
−2 (1 − 𝑘) 𝐻 (1 − 𝜓) 𝛿𝜓d𝛺 + 𝐺𝑐 𝑙0 ∇𝜓 ⋅ ∇𝛿𝜓 + 𝜓𝛿𝜓 d𝛺 = 0 and
∫𝛺 ∫𝛺 𝑙0
{ [ ] }
∑ 𝐺𝑐
(27) 𝑲𝜓 = 𝑩 T𝜓 𝐺𝑐 𝑙0 𝑩 𝜓 + 𝑵 T𝜓 + 2(1 − 𝑘)𝐻 𝑵 𝜓 d𝛺
𝑒
∫𝛺 𝑙0
∑ (38)
[ ( ) ] ( )
𝜕𝑇 𝑷𝜓 = − 2(1 − 𝑘)𝐻𝑵 T𝜓 d𝛺
𝛿𝑇 𝜌𝑐𝑝 − 𝑘𝑟𝑓 ∇𝛿𝑇 ⋅ ∇𝑇 d𝛺 − 𝛿𝑇 ℎ𝑐𝑣 𝑇𝑎𝑖𝑟 − 𝑇 d𝑆 = 0 (28) ∫𝛺
∫𝛺 𝜕𝑡 ∫𝑆𝑛 𝑒

For a two-dimensional plane strain problem, the displacement field and



𝒖, phase field scalar 𝜓 and temperature field 𝑇 in an element are 𝑪= 𝜌𝑐𝑝 (𝑵 𝑇 )T (𝑵 𝑇 )d𝛺
discretized as follows: 𝑒
∫𝛺
∑ ∑

𝑚 ∑
𝑚 ∑
𝑚
𝑲𝑇 = 𝑘𝑟𝑓 (𝑩 𝑇 )T (𝑩 𝑇 )d𝛺 + ℎ𝑐𝑣 (𝑵 𝑇 )T (𝑵 𝑇 )d𝑆 (39)
𝒖= 𝑵 𝑢𝑖 𝒖𝑖 𝜓= 𝑁𝑖𝜓 𝜓𝑖 𝑇 = 𝑁𝑖𝑇 𝑇𝑖 (29) 𝑒
∫𝛺 𝑒
∫𝑆𝑛
𝑖=1 𝑖=1 𝑖=1 ∑
𝑷𝑇 = ℎ𝑐𝑣 𝑇𝑎𝑖𝑟 𝑵 𝑇 d𝑆
where 𝑵 𝑢𝑖 , 𝑁𝑖𝜓 and 𝑁𝑖𝑇 are the shape function matrices at the node 𝑖; 𝑚 𝑒
∫𝑆𝑛
denotes the node number for an element. The linear triangle elements ∑
where 𝑒 represents the assembly of element stiffness matrix.
or quadrilateral elements are usually chosen in the two-dimensional The generalize-𝛼 method is adopted for time discretization when
problems, corresponding to 𝑚 = 3 or 𝑚 = 4, respectively. solving Eq. (36). In time step 𝑛, it is assumed that the displacement
The derivatives of Eq. (29) are given by field 𝒖𝑛 , temperature field 𝑻 𝑛 and phase field 𝝍 𝑛 have been given. We

𝑚 ∑
𝑚 ∑
𝑚 define that 𝒗𝑛 = 𝒖̇ 𝑛 , and 𝒂𝑛 = 𝒖̈ 𝑛 . For the displacement field, based on
𝜺𝑒 = 𝑩 𝑢𝑖 𝒖𝑖 − 𝜀𝑇 𝑰 0 , ∇𝜓 = 𝐵𝑖𝜓 𝜓𝑖 , ∇𝑇 = 𝐵𝑖𝑇 𝑇𝑖 (30) generalize-𝛼 method, we can obtained 𝒖𝑛+1 , 𝒗𝑛+1 , 𝒂𝑛+1 , 𝒖𝑛+𝛼𝑓 , 𝒗𝑛+𝛼𝑓 and
𝑖=1 𝑖=1 𝑖=1 𝒂𝑛+𝛼𝑓 when 𝒖𝑛 , 𝒗𝑛 and 𝒂𝑛 are given, as follows:
According to Eqs. (29) and (30), we can obtain: ( )
𝒖𝑛+𝛼𝑓 = 𝒖𝑛 + 𝛼𝑓 𝒖𝑛+1 − 𝒖𝑛

𝑚 ∑
𝑚 ( )
𝛿𝒖 = 𝑵 𝑢𝑖 𝛿𝒖𝑖 , 𝛿𝜺 = 𝑩 𝑢𝑖 𝛿𝒖𝑖 𝒗𝑛+𝛼𝑓 = 𝒗𝑛 + 𝛼𝑓 𝒗𝑛+1 − 𝒗𝑛
𝑖=1 𝑖=1
( )
𝒂𝑛+𝛼𝑚 = 𝒂𝑛 + 𝛼𝑚 𝒂𝑛+1 − 𝒂𝑛 (40)
∑𝑚 ∑
𝑚
𝛿𝜓 = 𝑁𝑖𝜓 𝛿𝜓𝑖 , ∇𝛿𝜓 = 𝐵𝑖𝜓 𝛿𝜓𝑖 (31) 1 [ ] 1 − 2𝛽
𝒂𝑛+1 = 𝒖𝑛+1 − 𝒖𝑛 − (𝛥𝑡)𝒗𝑛 − 𝒂𝑛
𝑖=1 𝑖=1 𝛽(𝛥𝑡)2 2𝛽
∑𝑚 ∑
𝑚 [ ]
𝒗𝑛+1 = 𝒗𝑛 + (𝛥𝑡) (1 − 𝛾)𝒂𝑛 + 𝛾𝒂𝑛+1
𝛿𝑇 = 𝑁𝑖𝑇 𝛿𝑇𝑖 , ∇𝛿𝑇 = 𝐵𝑖𝑇 𝛿𝑇𝑖
𝑖=1 𝑖=1 where 𝛥𝑡 = 𝑡𝑛+1 − 𝑡𝑛 is the time step; 𝛼𝑓 , 𝛼𝑚 , 𝛽, 𝛾 are the free parameters
Substituting Eq. (31) into Eqs. (26) ∼ (28), we can obtain: defined in this method.
The relationship among the four free parameters is given as follows:
− (𝛿𝒖)T 𝑩 T 𝑫𝑩 𝑢 d𝛺𝒖 + (𝛿𝒖)T 𝑩 T 𝑫𝜀𝑇 𝑰 0 d𝛺
∫𝛺 𝑢 ∫𝛺 𝑢 ( )
[ ] 𝛾=
1
+ 𝛼𝑚 − 𝛼𝑓 , 𝛽 =
1
𝛾+
1
(41)
2 4 2
+ (𝛿𝒖)T 𝑵 𝑢 𝒃d𝛺 + 𝑵 𝑢 𝒇 d𝑆 =0 (32)
∫𝛺 ∫𝜕𝛺𝑡 where
( )
1 3 − 𝜌∞ 1
{ [ ] } 𝛼𝑚 = , 𝛼𝑓 = (42)
𝐺𝑐 2 2 + 𝜌∞ 1 + 𝜌∞
−(𝛿𝝍)T 𝑩 T𝜓 𝐺𝑐 𝑙0 𝑩 𝜓 + 𝑵 T𝜓 + 2(1 − 𝑘)𝐻 𝑵 𝜓 d𝛺𝝍
∫𝛺 𝑙0 in which 𝜌∞ is the spectral radius of the amplification matrix at 𝛥𝑡 = ∞.
(33)
T For the temperature field 𝑻 𝑛 and phase field 𝝍 𝑛 , we still use the
+(𝛿𝝍) 2(1 − 𝑘)𝐻𝑵 T𝜓 d𝛺 =0
∫𝛺 generalized-𝛼 method to discretize time domain. More details can be
referred to the discretization procedures of the displacement field, as
d𝑻 shown in Eqs. (40) ∼ (42).
(𝛿𝑻 )T 𝜌𝑐𝑝 (𝑵 𝑇 )T (𝑵 𝑇 )d𝛺 + (𝛿𝑻 )T 𝑘 (𝑩 )T (𝑩 𝑇 )d𝛺𝑻
∫𝛺 d𝑡 ∫𝛺 𝑟𝑓 𝑇 At each time step, Newton–Raphson method is adopted to solve the
nonlinear equations. Letting 𝑖 be the Newton iteration step, and at the
− (𝛿𝑻 )T 𝒉𝑐𝑣 (𝑵 𝑇 )T 𝑵 𝑇 d𝑆 = 0 (34) iteration step 𝑖, the solution residual of all physical fields are defined
∫𝑆𝑛
as 𝑹𝑢𝑖 , 𝑹𝜓𝑖 and 𝑹𝑇𝑖 , respectively. At each time step, the iteration will
in which continue until the convergence condition is met, which can be given as
𝑵 𝑢 = [𝑵 𝑢1 , 𝑵 𝑢2 ⋯ , 𝑵 𝑢𝑚 ], 𝑵 𝜓 = [𝑵 𝜓1 , 𝑵 𝜓2 ⋯ , 𝑵 𝜓𝑚 ], follows:
𝑵 𝑇 = [𝑵 𝑇1 , 𝑵 𝑇2 ⋯ , 𝑵 𝑇𝑚 ] ⎧ ‖ 𝜓‖ ‖ 𝑇‖⎫
⎪‖
𝑢‖ ‖𝑹 𝑖 ‖ ‖𝑹 𝑖 ‖ ⎪
(35) ‖𝑹𝑖 ‖ ‖ ‖, ‖ ‖ ⩽ 𝑡𝑜𝑙
𝑩 𝑢 = [𝑩 𝑢1 , 𝑩 𝑢2 ⋯ , 𝑩 𝑢𝑚 ], 𝑩 𝜓 = [𝑩 𝜓1 , 𝑩 𝜓2 ⋯ , 𝑩 𝜓𝑚 ], max ⎨
‖ ‖
,
‖ 𝜓‖ ‖ 𝑇‖⎬
(43)
⎪ ‖𝑹𝑢0 ‖ ‖𝑹 0 ‖ ‖𝑹 0 ‖ ⎪
𝑩 𝑇 = [𝑩 𝑇1 , 𝑩 𝑇2 ⋯ , 𝑩 𝑇𝑚 ] ⎩ ‖ ‖ ‖ ‖ ‖ ‖⎭
where ‖⋅‖ represents the Euclidean norm; 𝑡𝑜𝑙 is the tolerance error.
The equivalent discrete forms of Eqs. (32) ∼ (34) are as follows,
𝑲 𝑢𝒖 + 𝑷 𝑢 = 0 3.2. COMSOL implementation
𝑲𝜓 𝝍 + 𝑷 𝜓 = 0 (36)
In this study, the coupled governing equations of discrete form for
d𝑻
𝑪 + 𝑲𝑇 𝑻 + 𝑷 𝑇 = 0 temperature field, displacement field, history strain field and phase
d𝑡

5
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 3. Coupling relationship of established physic fields.

field are solved by utilizing COMSOL Multiphysics, a finite element soft-


ware which can easily handle complex multi-field coupling problems.
In accordance with the governing equations in Section 2, the following
assumptions are used in the simulation:

• The concrete lining and rock mass are homogeneous and


isotropic;
• The sealing layers inside the CAES caverns are not considered in
this study;
• The concrete lining and rock mass are thermoelastic with good
integrity, and the plastic deformation is not considered in this
study;
• The CAES caverns have good air tightness, and air leakage is not
considered in the operation stage of CAES;
• The lined rock caverns have been filled with a certain amount of
compressed air that has an initial temperature 𝑇0 and pressure 𝑝0
before operation.

Details on the coupling of the physical fields are shown in Fig. 3.


Firstly, for the Thermodynamic process of the cavern, the mass con-
servation and energy conservation equations of stored air are solved.
The air temperature 𝑇𝑎𝑖𝑟 , density 𝜌, pressure 𝑝 are given under an
unknown cavern surface temperature. Then, the temperature of rock
mass is given under an unknown air temperature in the Heat transfer Fig. 4. First verification model.
module. Therefore, the Thermodynamic process of the cavern and Heat
transfer in the surrounding media must be coupled to solve the corre-
sponding unknowns. Furthermore, in the Solid mechanics module I, the
discretize the variables. A segregated scheme is employed to solve the
geostress balance computation is conducted, and the initial geostress
𝜎0 is given. Subsequently, the displacement field are obtained in the coupled governing equations, which indicates that all variables are
Solid mechanics module II via the output of other modules, including solved independently. For each time step, the implicit Generalized-𝛼
Thermodynamic process of the cavern, Heat transfer module, Solid me- method is applied to achieve unconditional stability in solving dif-
chanics module I and Phase field module under an unknown elasticity ferential equations with respect to time. There are several iteration
matrix 𝐃. According to the results of Solid mechanics module II, the steps in each time step. After each iteration step, we will evaluate the
unknown principal strain and its direction are stored in the module of
relative error of the calculation, which will not enter into the next
Variable storage. The History strain field is obtained via the positive
time step until the relative error is less than the targeted tolerance
elastic energy density from the Variable storage module. Finally, for the
Phase field module, the phase field variable is given under an unknown error. The Anderson acceleration method is adopted to accelerate the
history strain field 𝐻. Therefore, the perfect coupling relationship must iteration progress. To guarantee the accuracy of results, the stiffness
be established between Solid mechanics module II and Phase field matrix and load vector of governing equations for displacement field,
module for solving the phase field variable 𝜓. temperature field, history strain field and phase field are updated in
Algorithm 1 presents the detailed solution procedures for solving every iteration step. Our original codes can be downloaded from https:
the coupled governing equations of phase field modeling in COMSOL //github.com/xuyingjun1997/CAES_Phase-field-model.git.
Multiphysics. For all physical fields, the linear element is adopted to

6
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 5. Fracture initiation and propagation of the square plate with a elliptical hole subjected to initial geo-stress and internal pressure.

4. Model validation 4.1. Quasi-static fractures induced by high internal pressure

In this section, some numerical examples are presented to verify This example is square plate with elliptical holes subjected to initial
the feasibility and accuracy of the proposed phase field model. The geo-stress and an increasing internal pressure. The process of fracture
numerical results obtained by using the proposed phase field model are initiation and propagation can be observed clearly in this simulation.
compared with the previous analytical methods and experiments. The geometry and boundary conditions are shown in Fig. 4.

7
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 6. Hoop stress of point 𝑃1 and 𝑃2 with an increasing internal pressure.

Fig. 7. Final fracture patterns under different lateral pressure coefficient.

Fig. 8. Geometry and boundary conditions of rock specimens.

8
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 9. Comparisons of temperature and stress along the section 𝐴𝐴1 among the presented numerical results, previous experimental results and theoretical results.

Algorithm 1: Solution flow-chart for simulating thermo- initiate from the left and right orifice of the ellipse when 𝑝𝑖 = 14 MPa.
mechanical fracture in the CASE cavern As the internal pressure increases, the cracks gradually propagate along
Input: 𝜎0 , 𝑇𝑎𝑖𝑟 , 𝑇 , 𝜓, 𝐻, 𝑢 at 𝑡𝑛 the horizontal direction.
Output: 𝑇𝑎𝑖𝑟 , 𝑇 , 𝜓, 𝐻, 𝑢 at 𝑡𝑛+1 Fig. 6 presents the hoop stress of points 𝑃1 and 𝑃2 obtained from
Initialization: construct the initial guess the numerical simulation and analytical method. Points 𝑃1 and 𝑃2 are
(𝑇𝑎𝑖𝑟 )𝑚=0 , (𝑇𝑗 )𝑚=0 𝑚=0 , 𝐻 𝑚=0 , 𝑢𝑚=0 .
, 𝜓𝑛+1 located at the right orifice and the upper orifice of the ellipse as shown
𝑛+1 𝑛+1 𝑛+1 𝑛+1
in Fig. 4. For point 𝑃1 , Fig. 6a indicates that the hoop stress calculated
while 𝑡𝑜𝑙 ≤ 𝑒𝑟𝑟 do
1. Calculate the thermal strain 𝜀𝑇 = 𝛼𝑇 , elastic strain by the numerical simulation is consistent with that calculated by the
𝜺𝒆 = 𝜺 − 𝜀𝑇 𝑰. analytical method when 𝑝𝑖 is less than 6 MPa. However, with the
2. Calculate positive elastic energy increase in the internal pressure, the phase field 𝜓 at point 𝑃1 increases,
⟨ ⟩2 ⟨ ⟩2 which results in the reduction of hoop stress at point 𝑃1 . For point
𝛹𝑒+ = 𝜆2 𝑡𝑟(𝜺𝒆 )+ + 𝑢 𝑡𝑟(𝜺𝒆 )+ , the degradation function
[ 2
] 𝑃2 , the hoop stress calculated by the numerical simulation and the
𝑔(𝜓) = (1 − 𝑘)(1 − 𝜓) + 𝑘 .
analytical method are in good agreement because the phase field 𝜓 is
3. Calculate the elastic matrix 𝐃𝑚 by using elastic strain
𝑚 .
𝑛+1 considerably small.
𝜺𝒆 and 𝜓𝑛+1
Fig. 7 compares the final fracture patterns of the square plate with
4. Update 𝑢𝑚+1
𝑛+1
by using the elastic matrix 𝐃𝑚 𝑛+1
. an inclined elliptical hole subjected to geo-stress and internal pressure
𝑚+1
5. Update 𝐻𝑛+1 by using 𝑢𝑚+1
𝑛+1
and 𝐻 𝑚 .
𝑛+1 under different lateral pressure coefficients. It can be found that for 𝛾 =
𝑚+1 𝑚+1 2.5, 5, the final fracture gradually deflects towards the horizontal direc-
6. Update 𝜓𝑛+1 by using the new crack driving force 𝐻𝑛+1 .
7. Update (𝑇 )𝑚+1 by using the new 𝜓 𝑚+1
. tion with the increasing lateral pressure coefficient, which is consistent
𝑛+1 𝑛+1
with hydraulic fracture patterns observed in engineering.
8. Update (𝑇𝑎𝑖𝑟 )𝑚+1
𝑛+1
by using the new (𝑇 )𝑚+1
𝑛+1
.
9. Evaluate the relative error 𝑒𝑟𝑟 of (𝑢, 𝐻, 𝜓, 𝑇 , 𝑇𝑎𝑖𝑟 )
4.2. Fractures induced by thermal stress
end
In the second example, we mainly focus on fracture initiation and
propagation in a square rock specimen with a central bore under the
inner boundary heated by using the proposed phase field model. A
Analytical solution for stress distribution of this problem can be
similar thermal experiment of rock has been conducted by Ishida et al.
derived by using complex functions. Particularly, the radial and hoop
(2004). To verify the accuracy of the proposed numerical model, our
stresses around the elliptical orifice are given by:
results are compared with that obtained by using rock failure process
𝜎𝑟 = −𝑝𝑖 (44) analysis (RFPA) code (Huang et al., 2017).
Firstly, we consider an isotropic homogeneous disk with a central
[2𝑞(1 + 𝛾)𝑚 + 2𝑞(1 − 𝛾)](𝑚 − cos 2𝜃) borehole, as shown in Fig. 8a. The outer and inner radii of this disk
𝜎𝜃 = −𝑞(1 + 𝛾) +
𝑚2 − 2 𝑚 cos 2𝜃 + 1 are 150 mm and 14 mm, respectively. The initial temperature of the
2 𝑚 cos 2𝜃 − 3𝑚2 + 1 rock specimen is set as 𝑇0 = 28 ◦ C. On the inner boundary of borehole,
+ 𝑝𝑖 (45)
𝑚2 − 2 𝑚 cos 2𝜃 + 1 a heater is placed to heat the rock specimen and the inner boundary
where 𝑎 and 𝑏 are the transverse and longitudinal axes of the eclipse; is subjected to temperature of 𝑇 (𝑡) = 𝑇0 + 1.5 ◦ C/min. The outer
𝑚 is a constant 𝑚 = (𝑎 − 𝑏)∕(𝑎 + 𝑏); 𝑞 and 𝑝𝑖 are the vertical geo-stress and inner boundaries of this model are free. Other mechanical and
and internal pressure, respectively; 𝛾 is the lateral pressure coefficient; thermodynamic parameters of the rock specimen are: Young’s modulus
𝐿 is the length of the numerical model. 𝐸 = 47 GPa, Poisson’s ratio 𝜐 = 0.25, density 𝜌 = 2730 kg/m3 , specific
In the simulation and analytical computation, the parameters are heat capacity 𝐶𝑝 = 1025.64 J/(kg K), thermal conductivity 𝑘 = 2.3
determined as follows: 𝑎 = 0.4 m, 𝑏 = 0.05 m, initial vertical geo-stress W/(m K), thermal expansion coefficient 𝛼 = 0.85 × 10−5 1/K.
𝑞 = 0.5 MPa, lateral pressure coefficient 𝛾 = 1.5, Young’s modulus 𝐸 = The problem described in Fig. 8a has a theoretical solution and the
60 GPa, Poisson’s ratio 𝜐 = 0.3, critical energy release rate 𝐺𝑐 = 500 thermal stress can be calculated by:
[ 2 ]
N/m, length scale parameter 𝑙0 = 0.1 m, internal pressure loading rate 2
𝛼𝐸 𝑟 + 𝑟0
𝑟1 𝑟
2
𝛥𝑝𝑖 (𝑡) = 1 MPa/s. 𝜎𝜃 = 𝑟𝑇 𝑑𝑟 + 𝑟𝑇 𝑑𝑟 − 𝑇 𝑟 (46)
𝑟2 𝑟2 − 𝑟2 ∫𝑟0 ∫𝑟0
Fig. 5 shows the fracture initiation and propagation of the square [ 21 0
2
]
𝛼𝐸 𝑟 − 𝑟0
𝑟1 𝑟
plate with a central elliptical hole subjected to initial geo-stress and
𝜎𝑟 = 𝑟𝑇 𝑑𝑟 − 𝑟𝑇 𝑑𝑟 (47)
high internal pressure. It can be found that the fractures begin to 𝑟2 2 2 ∫
𝑟1 − 𝑟0 𝑟0 ∫ 𝑟0

9
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 10. Comparison of thermal fracture initiation, propagation and final pattern in the rock specimen with a center borehole between (a)∼(d) the presented numerical results
and (e) the previous numerical result (Huang et al., 2017).

where 𝜎𝜃 and 𝜎𝑟 are the hoop stress and radial stress, respectively; 𝑟0 and then increases as the distance from the borehole increases. It varies
and 𝑟1 are the inner and outer radius of the disk; 𝑟 is the distance to approximately between −13 and 0 MPa. The hoop stress increases
borehole center; as the distance from the borehole increases and gradually changes
Fig. 9 compares the temperature and stress of the rock specimen from compressive stress to tensile stress. On the outer boundary, the
along the section 𝐴𝐴1 at 94 min with previous experiments and the- maximum hoop stress reaches 6.9 MPa, which is satisfactorily close to
oretical method. Fig. 9a indicates that the temperature obtained by the theoretical result.
using the proposed method is in good agreement with that measured in Then, we consider a heterogeneous numerical model to investi-
Ishida’s experiment. The temperature has a maximum value of 165 ◦ C gate the initiation and propagation mechanism of thermal fractures
at the borehole wall and gradually decreases as the distance from the as shown in Fig. 8b. The rock specimen is a square plate with a
borehole increases. In addition, Fig. 9b shows that the radial and hoop length of 𝑎 = 300 mm. A borehole with a radius of 14 mm is located
stresses obtained by our simulation are also in good agreement with in the center of the rock specimen. The initial temperature of rock
that obtained by the analytical method. The radial stress first decreases specimen is set as 28 ◦ C. The temperature on the outer boundary of

10
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 11. Geometry and boundary conditions of the numerical model for the CAES cavern.

the rock sample is fixed at 28 ◦ C and the temperature on the inner Table 1
Calculation parameters for the displacement field.
boundary increases as a linear function of 𝑇 (𝑡) = 𝑇0 + 1.5 ◦ C/min.
The Young’s modulus of rock specimen follows the Weibull distribution Variables Definitions Values Units

where the homogeneous index 𝑁 is adopted as 5. Other mechanical 𝑉 Volume of the cavern 19.63 m3
𝑟 Radius of the cavern 3.0 m
and thermodynamic parameters of the rock specimen are consistent
𝛿𝑟 Thickness of concrete lining 0.5 m
with those of the homogeneous model. Particularly, for the phase field ℎ Buried depth of the cavern 100 m3
model, the critical energy release 𝐺𝑐 is set as 16.5 N/m and the length 𝜌1 Density of concrete lining 2,500 kg/m3
scale parameter 𝑙0 is defined as 5 mm from a calibration from the 𝜌2 Density of rock mass 2,500 kg/m3
experimental tests. 𝐸1 Elastic modulus of concrete lining 30 GPa
𝐸2 Elastic modulus of rock mass 35 GPa
Fig. 10 shows the initiation and propagation of thermal fractures.
𝜐1 Poisson’s ratio of concrete lining 0.2 –
As observed, a single macro-fracture initiates from the top boundary of 𝜐2 Poisson’s ratio of rock mass 0.3 –
the rock specimen at 78 min. As the heating time increases, the fracture
starts to propagate towards the borehole along the radial direction.
When 𝑡 = 80 min, the fracture extends to the borehole wall. Fig. 10e
gives the final fracture pattern obtained by RFPA code. Comparisons in applied to the whole calculation domain. For the sake of simplification,
Fig. 10 shows that the fracture pattern obtained by our simulation is we perform the cavern excavation and operation stage in a calculation
in good agreement with the previous numerical result. In conclusion, step. As shown in Fig. 11b, the excavation stage is considered as the
the verification results indicate that the proposed numerical simulation initiate state of the operation stage. Therefore, at the initial moment
method has the good capability and accuracy of capturing the thermal (corresponding to excavation stage), the air internal pressure 𝑝𝑖 is not
fractures. applied on the inner surface of the CAES cavern. When 𝑡 >0, the cavern
enters the operation stage and the air internal 𝑝𝑖 is applied on the inner
5. Fracture initiation and propagation of a CAES cavern surface of the CAES cavern. As for heat transfer, the initial temperature
is set to 𝑇0 . For simplicity, only half of the entire cavern is modeled.
In this section, a coupled thermo-mechanical numerical model for Therefore, the left boundary of the model is thermally insulated while
fracture initiation and propagation in a CAES cavern is established. On the other outer boundaries of the model have a constant temperature
this basis, we analyze the key factors influencing the fracture initiation 𝑇0 .
and propagation in the CAES cavern. Heat convection occurs between the air inside the cavern and inner
surface of concrete lining during the operation stage of the CAES cav-
5.1. Boundary conditions and calculation parameters ern. In a CAES cycle, compressed air is injected into the underground
cavern at a certain rate for 8 h firstly and stored for 6 h. Then, the
The geometry and boundary conditions of the numerical model are compressed air is withdrew at a certain rate for 4 h and stored for 6 h.
shown in Fig. 11. The 2D model has a domain of 25 m × 50 m. In this simulation, we set the charging rate 𝑚̇ 𝑐 for the CAES cavern
The excavation radius of the cavern is 3.0 m and the thickness of as 0.044 kg/s. The ratio of discharging rate to the charging rate 𝜉𝑐𝑑 is
the concrete lining is 0.5 m. The cavern size is mainly referred to a set as 2. The calculation parameters for the CAES cavern are listed in
specific CAES project of Yungang Mine in Shanxi province, China and Tables 1 and 2. Particularly, for the thermo-mechanical coupled phase
a CAES pilot cavern presented in the studies of Kim et al. (2016) and field model, the critical energy release rate 𝐺𝑐 and 𝐺𝑓 of rock mass and
Kim et al. (2012b). As shown in Fig. 11a, the geostress balance stage concrete lining are 150 N/m and 80 N/m, respectively. The length scale
requires obtaining the initial geostress field by applying the overburden parameter 𝑙0 is 0.1 m. We divide the whole computation domain into
pressure 𝑞 and linearly varying pressure from 𝛾𝑞 to 𝛾(𝑞+𝜌𝑔𝐻) to the top two parts, namely 𝛺1 and 𝛺2 , respectively. To reduce the computation
and right boundaries of the numerical model, respectively. The normal cost, a more refined mesh is used in the region of 𝛺1 and the maximum
displacement constraint and fixed constraint are applied to the left and size of elements are less than 𝑙0 ∕2, while a relatively sparse mesh is used
bottom of the numerical model, respectively. In addition, gravity is in the region 𝛺2 .

11
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 12. Fracture initiation and propagation of the CAES cavern when the lateral pressure coefficient 𝛾 = 1.

Fig. 13. Historical Distribution of maximum principal stress and maximum positive strain energy on the cavern surface.

12
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 14. Thermal conductivity of rock mass and concrete lining from 𝑡 = 7.4 h to 𝑡 = 7.9 h.

Fig. 15. Thermodynamic responses of the CAES cavern before instability failure.

5.2. Result analysis tensile fracture initiates from the location of 𝛼 = 180◦ at 𝑡 = 7.6 h.
With continuous air charging, a tensile fracture propagates towards the
Fig. 12 shows the fracture initiation and propagation of the CAES direction of 45◦ from the vertical direction and another tensile fracture
cavern when the lateral pressure coefficient 𝛾 = 1. We define 𝛼 propagates towards the bottom of the model.
(0◦ ⩽ 𝛼 ⩽ 180◦ ) as the angle between a point of cavern periphery Fig. 13a shows the distribution of maximum principal stress on the
and the vertical direction of the cavern roof. As observed, a macro inner surface of the CAES cavern at 𝑡 = 1 h, 3 h, 5 h and 7 h. It
tensile fracture initiates from the location of 𝛼 = 45◦ and another can be observed that the maximum principal stress of each position

13
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 16. First principal stress at different positions of the CAES cavern.

Fig. 17. Fracture patterns of the CAES cavern under different lateral pressure coefficient.

on the inner surface of the cavern is relatively close because the lateral our proposed phase-field modeling. As observed, during the early stage
pressure coefficient is equal to 1. With the increase of time, the damage of air injection, there is no obvious difference for the fracture driving
of the inner surface of the cavern gradually increases, which results force on the inner surface of the CAES cavern. With the increase in
in the decrease of the maximum principal stress. Fig. 13b shows the time, the fracture driving force on the inner surface reaches the peak
distribution of the historical maximum positive strain energy on the at the locations of 𝛼 = 45◦ and 𝛼 = 180◦ . Overall, the results obtained
inner surface of the CAES cavern at 𝑡 = 1 h, 3 h, 5 h and 7 h. from Fig. 13 indicate that the locations of 𝛼 = 45◦ and 𝛼 = 180◦ are the
The historical maximum positive strain energy in the figure can be maximum positive strain energy concentration zones where the macro
regarded as the driving force for fracture initiation and propagation in tensile fractures are mostly likely to initiate. The figure can initially

14
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

first principal stress at 𝑃3 decreases during the charging, while the first
principal stress at 𝑃4 increases firstly and then decreases. The results
indicate that the damage on the inner surface of the concrete lining
is larger than that on the outer surface of the concrete lining, verified
from the results in Fig. 12.

5.3. Parameter sensitivity analysis and discussion

5.3.1. Effect of the lateral pressure coefficient 𝛾


Fig. 17 shows the fracture initiation locations and fracture patterns
of the CAES cavern during the charging stage under different lateral
pressure coefficient 𝛾 (𝛾 = 0.5, 0.7, 0.9, 1.1, 1.3, 1.5), while the
other parameters keep unchanged. In addition, the case for 𝛾 = 1
is presented in Fig. 12. The results indicate that the tensile fractures
initiate from the top and bottom of the cavern and propagate towards
the vertical direction when the lateral pressure coefficient 𝛾 = 0.5,
0.7 and 0.9. However, as afore-mentioned, when the lateral pressure
Fig. 18. Critical internal pressure inducing the tensile fractures of the CAES cavern 𝛾 = 1, there are two macro tensile fractures distributed around the
under different lateral pressure coefficient. CAES cavern. One of the fractures initiates from the location of 𝛼 =45◦
and the other initiates from the bottom of the cavern. As the lateral
Table 2 pressure coefficient increases, the fracture initiation locations gradually
Calculation parameters for the temperature field. deflect to the horizontal direction. When the lateral pressure coefficient
Variables Definitions Values Units 𝛾 = 1.5, the tensile fractures propagate nearly along the horizontal
𝑘1 Thermal conductivity of concrete lining 3.0 W/(m K) direction. The evolution of fracture patterns under different lateral
𝑘2 Thermal conductivity of rock mass 3.0 W/(m K) pressure coefficient is controlled by the coupled effects of gravity,
𝑐𝑝1 Specific heat of concrete lining at constant 900 J/(kg K) geo-stress, internal air pressure and temperature field.
pressure
Fig. 18 shows the effect of lateral pressure coefficient 𝛾 on the
𝑐𝑝2 Specific heat of rock mass at constant 900 J/(kg K)
pressure critical internal pressure inducing the tensile fractures of the CAES
𝛼1 Thermal expansion coefficient of concrete 1.0 × 10−5 1/K cavern. The figure indicates that the critical internal pressure increases
lining firstly and then decreases with the increase in the lateral pressure
𝛼2 Thermal expansion coefficient of rock mass 1.0 × 10−5 1/K
coefficient. The critical lateral pressure coefficient is 𝛾 = 1. When 𝛾
ℎ𝑐𝑣 Coefficient of heat exchange 50 W/(m2 K)
𝑅 Constant of air 287 J/(kg K)
is approximately equal to 1, the critical internal pressure can reach its
𝑝𝑐 Critical pressure of air 3.766 MPa maximum, which is approximately 10.77 MPa. However, as the lateral
𝑇𝑐 Critical temperature of air 132.65 K pressure coefficient increases continuously, the critical internal pres-
𝑐𝑣0 Specific heat of air at constant volume 717.86 J/(kg K) sure gradually decreases. When 𝛾 = 1.5, the critical internal pressure is
𝑐𝑝0 Specific heat of air at constant pressure 1,000.5 J/(kg K)
equal to 7.31 MPa.

5.3.2. Effect of the critical energy release rate 𝐺𝑐


explain the mechanism of fracture initiation and propagation of the Fig. 19 shows the fracture patterns of the CAES cavern under
CAES cavern presented in Fig. 12. different critical energy release rate of rock mass under a fixed 𝛾 = 1.2.
Fig. 14 shows the distribution of thermal conductivity of rock mass The simulation indicates that the critical energy release rate of rock
and concrete lining from 𝑡 = 7.4 h to 𝑡 = 7.9 h. It can be observed mass 𝐺𝑐 has little effect on the fracture pattern of the CAES cavern.
that the thermal conductivity of the concrete lining decreases firstly That is, a macro tensile fracture initiates from the right part of the CAES
and then the thermal conductivity of the rock mass decreases during cavern and then propagates horizontally towards the deep surrounding
the charging stage, which is consistent with the trend of phase-field rock. In addition, Fig. 20 shows the critical internal pressure inducing
evolution. The results also indicate that the thermal conduction capa- the tensile fractures of the CAES cavern under different critical energy
bility of the concrete lining and rock mass will reduce when the damage release rate 𝐺𝑐 . As observed, the critical internal pressure increases
of the concrete lining and rock mass increases. nearly linearly with the increase in the critical energy release rate 𝐺𝑐 .
Fig. 15 presents the temperature and pressure of compressed air The reason for this phenomenon is that a larger critical energy release
inside the CAES cavern before instability failure. It is observed that rate 𝐺𝑐 of the rock mass means a higher fracture resistance for tensile
the phase-field simulation are in good agreement with the analytical failure.
results obtain by Zhou et al. (2015). However, with the reduction
of the thermal conductivity of concrete lining and rock mass during 5.3.3. Effect of the elastic modulus of rock mass 𝐸
phase-field evolution, the air temperature in our simulation has a slight Fig. 21 shows the positions of fracture initiation points under differ-
deviation with the analytical result. The reason is that the heat of the ent elastic modulus of rock mass when the lateral pressure coefficient
air inside the cavern cannot be effectively transferred outward with the 𝛾 = 1.5, the buried depth 𝐻 = 100 m, and the critical energy release
decrease of the thermal conductivity of lining and surrounding rock. rate of rock mass 𝐺𝑐 = 150 N/m. The figure indicates that the positions
Fig. 16 shows the first principal stress at the fracture initiation of fracture initiation are influenced by the elastic modulus of rock
positions of the CAES cavern. Fig. 16a shows that the first principal mass. With the increase in 𝐸, the fracture initiation points gradually
stress at 𝑃1 decreases with air inflation. The reason is that the phase- moves from the inner surface to the outer surface of concrete lining.
field 𝜓 at 𝑃1 is relatively large than other positions, which can result in The reason is that a higher elastic modulus of rock mass commonly
the decrease of stress at 𝑃1 . As for Point 𝑃2 , firstly, the internal pressure has a higher tensile stress under high internal pressure, which can
on the inner surface of the CAES cavern increases with the air charging, result in stress and energy concentration near the host rock. Fig. 22
which can increase the first principal stress. However, if the air inflation presents the final fracture patterns of the CAES cavern under different
continues, the first principal stress at 𝑃2 tends to decrease because of an elastic modulus, which indicates that the fracture patterns are nearly
increasing phase field scalar 𝜓. In addition, Fig. 16b indicates that the not influenced by the elastic modulus of rock mass. In addition, the

15
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 19. Fracture patterns of the CAES cavern under different critical energy release rate.

results show that a larger elastic modulus of rock mass results in a with the increase in 𝐻, it can be observed that the damage induced by
larger fracture width. excavation gradually increases. The damage zone induced by cavern
Fig. 23 shows the variation in the critical internal pressure inducing excavation is mainly distributed at the top and bottom of the CAES
the tensile fractures of the CAES cavern with the elastic modulus of rock cavern, as shown in Figs. 24c and 24d.
mass. The figure indicates that critical internal pressure and the elastic Fig. 25 shows the critical internal pressure inducing the tensile
modulus of rock mass are positively correlated. The elastic modulus of fractures of the CAES cavern under different buried depth 𝐻. The figure
rock mass has a similar effect to the critical energy release rate on the indicates that the buried depth has a great effect on the critical internal
critical internal pressure. That is, a larger elastic modulus of rock mass pressure. The critical internal pressure increases as the buried depth
corresponds to a higher critical internal pressure. increases. Moreover, with the increase in 𝐻, the influence of buried
depth on the critical internal pressure is increasing. When 𝐻 = 75 m,
5.3.4. Effect of the buried depth 𝐻 the critical internal pressure is approximately 5.00 MPa. In addition,
Fig. 24 presents the fracture patterns of the CAES cavern under when 𝐻 = 150 m, the critical internal pressure increases to 8.34 MPa.
different buried depth 𝐻 when the lateral pressure coefficient 𝛾 = 1.5,
elastic modulus of rock mass 𝐸 = 25 GPa and critical energy release rate 5.4. Discussions
𝐺𝑐 = 150 N/m. It can be observed that the buried depth has little effect
on the fracture initiation positions and fracture propagation patterns. It is worth noting that some previous studies have investigated the
Under a fixed 𝛾 = 1.5, only a macro tensile fracture initiates from failure issues of highly pressurized caverns by using the physical model
the cavern side and propagates along the horizontal direction. This test and the element-free Galerkin (EFG) method such as Tunsakul et al.
conclusion is consistent with that obtained by Tunsakul et al. (2014). (2013, 2014) and Tunsakul et al. (2018). It can be concluded from
Their results also indicate that the buried depth has little effect on these studies that the lateral pressure coefficient has a strong influence
the fracture initiation positions and final fracture patterns. However, on the position of fracture initiation and fracture propagation paths,

16
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

initiate from the cavern roof and bottom and propagate along the verti-
cal direction, which is slightly different from the results obtained from
Fig. 26. In the studies of Tunsakul et al. (2013, 2014) and Tunsakul
et al. (2018), only a macro tensile fracture initiates from the cavern
roof and propagates along the vertical direction. The reason for this
difference may be that the cavern size and geometry in our numerical
model are different from those in the previous studies, which have
significant influence on the stress distribution around the cavern. In our
study, because the cavern size is relatively small, the stress difference
between the cavern roof and bottom is not obvious. Therefore, the high
internal pressure is the main factor controlling the fracture initiation
and propagation.
Differently from the previous studies, in our proposed phase-field
modeling, the fracture initiation positions and propagation paths can
be automatically captured without the requirement of determining the
Fig. 20. Critical internal pressure inducing the tensile fractures of the CAES cavern locations of fracture initiation in advance. Nonetheless, in this study,
under different critical energy release rate.
we only investigate the tensile fractures induced by high internal pres-
sure and temperature in the CAES cavern, while the shear fractures are

as presented in Fig. 26. Similar findings can also be concluded by our not considered (Thongraksa et al., 2023). In addition, the air seepage
research. However, as previously mentioned, our study indicates that process in the surrounding rock is also not considered. These limitations
when the lateral pressure coefficient is less than 1, the tensile fractures are intended to be solved in our future research.

Fig. 21. Positions of fracture initiation points in the CAES cavern under different elastic modulus.

17
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 22. Fracture patterns of the CAES cavern under different elastic modulus.

6. Conclusions

This study proposes a coupled thermo-mechanical phase-field model


for simulating fracture initiation and propagation in the CAES cav-
erns. Detailed algorithm and implementation procedures in COMSOL
Multiphysics are further given. To verify the capability and accuracy
of our proposed model in simulating brittle quasi-static fractures and
thermal fractures, two numerical examples are presented in this study
for comparison. Finally, the influence of lateral pressure coefficient,
critical energy release rate of rock mass, elastic modulus of rock mass
and buried depth on the fracture patterns and critical internal pressure
when fracture initiates is investigated. From the numerical simulations,
some important conclusions can be summarized as follows:
(1) During excavation and operation, the first damage gradually
accumulates in the concrete lining and then brittle tensile fractures
initiate from the concrete lining and propagate towards the deep rock
mass with air charging. Fracture patterns and critical internal pressure
are mainly related to fracture and mechanical parameters and initial
geostress.
Fig. 23. Critical internal pressure inducing the tensile fractures of the CAES cavern
(2) The lateral pressure coefficient 𝛾 has a significant effect on
under different elastic modulus.
the fracture initiation locations, fracture patterns and critical internal

18
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Fig. 24. Fracture patterns of the CAES cavern under different buried depth.

pressure of the CAES cavern. As the lateral pressure coefficient in-


creases, the tensile fractures gradually deflect towards the horizontal
direction, with an increasing critical internal pressure. When the lateral
pressure coefficient is equal to 1, the critical internal pressure reaches
its maximum.
(3) The critical energy release rate 𝐺𝑐 , has little effect on the
fracture patterns of the CAES cavern. However, the effect on the critical
internal pressure is significant. A larger critical energy release rate
corresponds to a higher critical internal pressure inducing the tensile
fractures of the CAES cavern.
(4) The elastic modulus of rock mass 𝐸 has little effect on the
fracture patterns of the CASE cavern. However, as the elastic modulus
of rock mass increases, the positions where fracture initiates move from
the inner surface of the concrete lining to the outer surface of the
concrete lining while the critical internal pressure increases.
Fig. 25. Critical internal pressure inducing the tensile fractures of the CAES cavern (5) The buried depth 𝐻 has little effect on the fracture patterns
under different buried depth.
and critical internal pressure of the CAES cavern. As the buried depth
increases, the fracture patterns are basically unchanged and the critical
internal pressure gradually increases.

19
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Cheng, P., Zhu, H., Yan, Z., Shen, Y., Fish, J., 2022b. Multiscale modeling for fire
induced spalling in concrete tunnel linings based on the superposition-based phase
field fracture model. Comput. Geotech. 148, 104832.
Fei, F., Choo, J., 2021. Double-phase-field formulation for mixed-mode fracture in rocks.
Comput. Methods Appl. Mech. Engrg. 376, 113655.
Francfort, G.A., Marigo, J.J., 1998. Revisiting brittle fracture as an energy minimization
problem. J. Mech. Phys. Solids 46 (8), 1319–1342.
Gao, C., Zhou, Z., Li, Z., Li, L., Cheng, S., 2020. Peridynamics simulation of surrounding
rock damage characteristics during tunnel excavation. Tunn. Undergr. Space
Technol. 97, 103289.
Guo, C., Pan, L., Zhang, K., Oldenburg, C.M., Li, C., Li, Y., 2016. Comparison of
compressed air energy storage process in aquifers and caverns based on the Huntorf
CAES plant. Appl. Energy 181, 342–356.
Hu, X., Zhang, H., Boldini, D., Liu, C., He, C., Wu, S., 2021. 3D modelling of soil-
rock mixtures considering the morphology and fracture characteristics of breakable
blocks. Comput. Geotech. 132, 103985.
Huang, X., Tang, S., Tang, C., Xie, L., Tao, Z., 2017. Numerical simulation of cracking
behavior in artificially designed rock models subjected to heating from a central
borehole. Int. J. Rock Mech. Min. Sci. 98, 191–202.
Ishida, T., Kinoshita, N., Wakabayashi, N., 2004. Acoustic emission monitoring during
thermal cracking of a granite block heated in a center hole. In: Proceedings of the
3rd Asian Rock Mechanics Symposium. Kyoto, pp. 133–138.
Jiang, Z., Li, P., Tang, D., Zhao, H., Li, Y., 2020. Experimental and numerical
investigations of small-scale lined rock cavern at shallow depth for compressed
air energy storage. Rock Mech. Rock Eng. 1–13.
Jin, Y., González-Estrada, O., Pierard, O., Bordas, S., 2017. Error-controlled adaptive
extended finite element method for 3D linear elastic crack propagation. Comput.
Methods Appl. Mech. Engrg. 318, 319–348.
Jongpradist, P., Tunsakul, J., Kongkitkul, W., Fadsiri, N., Arangelovski, G., Youwai, S.,
Fig. 26. Fracture patterns of underground highly pressurized cavern obtained from the 2015. High internal pressure induced fracture patterns in rock masses surrounding
previous studies (Tunsakul et al., 2014). caverns: experimental study using physical model tests. Eng. Geol. 197, 158–171.
Kim, H.M., Park, D., Ryu, D.W., Song, W.K., 2012a. Parametric sensitivity analysis of
ground uplift above pressurized underground rock caverns. Eng. Geol. 135, 60–65.
Kim, H.M., Rutqvist, J., Jeong, J.H., Choi, B.H., Ryu, D.W., Song, W.K., 2013.
CRediT authorship contribution statement Characterizing excavation damaged zone and stability of pressurized lined rock
caverns for underground compressed air energy storage. Rock Mech. Rock Eng. 46
(5), 1113–1124.
Caichu Xia: Investigation, Software, Writing – original draft, Vi- Kim, H.M., Rutqvist, J., Kim, H., Park, D., Ryu, D.W., Park, E.S., 2016. Failure
sualization. Yingjun Xu: Formal analysis, Methodology, Supervision, monitoring and leakage detection for underground storage of compressed air energy
Writing – review & editing. Shuwei Zhou: Resources, Supervision, in lined rock caverns. Rock Mech. Rock Eng. 49 (2), 573–584.
Kim, H.M., Rutqvist, J., Ryu, D.W., Choi, B.H., Sunwoo, C., Song, W.K., 2012b.
Conceptualization, Writing – review & editing. Shikang Qin: Data cu-
Exploring the concept of compressed air energy storage (CAES) in lined rock
ration, Supervision, Writing – review & editing. Xiuhan He: Resources, caverns at shallow depth: a modeling study of air tightness and energy balance.
Supervision, Writing – review & editing. Appl. Energy 92, 653–667.
Liu, S., Wang, Y., Peng, C., Wu, W., 2022. A thermodynamically consistent phase
field model for mixed-mode fracture in rock-like materials. Comput. Methods Appl.
Declaration of competing interest Mech. Engrg. 392, 114642.
Liu, J., Xue, Y., Zhang, Q., Yao, K., Liang, X., Wang, S., 2020. Micro-cracking behavior
The authors declare that they have no known competing finan- of shale matrix during thermal recovery: Insights from phase-field modeling. Eng.
Fract. Mech. 239, 107301.
cial interests or personal relationships that could have appeared to
Mandal, T.K., Nguyen, V.P., Wu, J.Y., Nguyen-Thanh, C., de Vaucorbeil, A., 2021.
influence the work reported in this paper. Fracture of thermo-elastic solids: Phase-field modeling and new results with an
efficient monolithic solver. Comput. Methods Appl. Mech. Engrg. 376, 113648.
Data availability Miehe, C., Hofacker, M., Schänzel, L.M., Aldakheel, F., 2015. Phase field modeling
of fracture in multi-physics problems. Part II. Coupled brittle-to-ductile failure
criteria and crack propagation in thermo-elastic–plastic solids. Comput. Methods
Data will be made available on request. Appl. Mech. Engrg. 294, 486–522.
Mikelic, A., Wheeler, M.F., Wick, T., 2015. A phase-field method for propagating fluid-
filled fractures coupled to a surrounding porous medium. Multiscale Model. Simul.
Acknowledgments 13 (1), 367–398.
Msekh, M.A., Sargado, J.M., Jamshidian, M., Areias, P.M., Rabczuk, T., 2015. Abaqus
The author would like to acknowledge the financial support from implementation of phase-field model for brittle fracture. Comput. Mater. Sci. 96,
the National Nature Science Foundation of China (52278402, 472–484.
Perazzelli, P., Anagnostou, G., 2016. Design issues for compressed air energy storage
52278408) and Fundamental Research Funds for the Central Univer- in sealed underground cavities. J. Rock Mech. Geotech. Eng. 8 (3), 314–328.
sities of China (22120220117, 22120210056). Perazzelli, P., Anagnostou, G., 2018. Upper bound limit analysis of uplift failure in
pressurized sealed rock tunnels. Int. J. Numer. Anal. Methods Geomech. 42 (5),
719–735.
References
Rabczuk, T., Belytschko, T., 2004. Cracking particles: a simplified meshfree method for
arbitrary evolving cracks. Internat. J. Numer. Methods Engrg. 61 (13), 2316–2343.
Borden, M.J., Verhoosel, C.V., Scott, M.A., Hughes, T.J., Landis, C.M., 2012. A phase- Rabczuk, T., Belytschko, T., 2007. A three-dimensional large deformation meshfree
field description of dynamic brittle fracture. Comput. Methods Appl. Mech. Engrg. method for arbitrary evolving cracks. Comput. Methods Appl. Mech. Engrg. 196
217, 77–95. (29–30), 2777–2799.
Bourdin, B., Francfort, G.A., Marigo, J.J., 2000. Numerical experiments in revisited Rabczuk, T., Zi, G., Bordas, S., Nguyen-Xuan, H., 2010. A simple and robust three-
brittle fracture. J. Mech. Phys. Solids 48 (4), 797–826. dimensional cracking-particle method without enrichment. Comput. Methods Appl.
Chen, K.T., Wu, J.H., 2018. Simulating the failure process of the Xinmo landslide using Mech. Engrg. 199 (37–40), 2437–2455.
discontinuous deformation analysis. Eng. Geol. 239, 269–281. Raju, M., Khaitan, S.K., 2012. Modeling and simulation of compressed air storage in
Cheng, L., Luo, Z., Zhao, L., Xie, Y., 2022a. Numerical analysis of fracture deformation caverns: a case study of the Huntorf plant. Appl. Energy 89 (1), 474–481.
and instability during CO2 geological sequestration using a THM-XFEM coupled Ren, H., Zhuang, X., Rabczuk, T., 2017. Dual-horizon peridynamics: A stable solution
model. Comput. Geotech. 145, 104664. to varying horizons. Comput. Methods Appl. Mech. Engrg. 318, 762–782.

20
C. Xia et al. Computers and Geotechnics 157 (2023) 105329

Shi, F., Liu, J., 2021. A fully coupled hydromechanical XFEM model for the simulation Zhao, Y., Cao, S., Shang, D., Yang, H., Yu, Y., Li, Y., Liu, J., Wang, H., Pan, R., Yang, H.,
of 3D non-planar fluid-driven fracture propagation. Comput. Geotech. 132, 103971. et al., 2019. Crack propagation and crack direction changes during the hydraulic
Shiozawa, S., Lee, S., Wheeler, M.F., 2019. The effect of stress boundary conditions fracturing of coalbed. Comput. Geotech. 111, 229–242.
on fluid-driven fracture propagation in porous media using a phase-field modeling Zhou, X.P., Du, E.B., Wang, Y.T., 2022a. Thermo-hydro-chemo-mechanical coupling
approach. Int. J. Numer. Anal. Methods Geomech. 43 (6), 1316–1340. peridynamic model of fractured rock mass and its application in geothermal
Song, X., Menon, S., 2019. Modeling of chemo-hydromechanical behavior of unsatu- extraction. Comput. Geotech. 148, 104837.
rated porous media: a nonlocal approach based on integral equations. Acta Geotech. Zhou, S., Rabczuk, T., Zhuang, X., 2018a. Phase field modeling of quasi-static and
14 (3), 727–747. dynamic crack propagation: COMSOL implementation and case studies. Adv. Eng.
Thongraksa, A., Jongpradist, P., Nanakorn, P., Tunsakul, J., 2023. Shear fracture Softw. 122, 31–49.
propagation in quasi-brittle materials by an element-free Galerkin method. Theor. Zhou, X., Wang, L., Shou, Y., 2020a. Understanding the fracture mechanism of ring
Appl. Fract. Mech. 123, 103713. Brazilian disc specimens by the phase field method. Int. J. Fract. 226 (1), 17–43.
Tunsakul, J., Jongpradist, P., Kim, H.M., Nanakorn, P., 2018. Evaluation of rock fracture Zhou, S.W., Xia, C.C., Du, S.G., Zhang, P.Y., Zhou, Y., 2015. An analytical solution
patterns based on the element-free Galerkin method for stability assessment of a for mechanical responses induced by temperature and air pressure in a lined rock
highly pressurized gas storage cavern. Acta Geotech. 13 (4), 817–832. cavern for underground compressed air energy storage. Rock Mech. Rock Eng. 48
Tunsakul, J., Jongpradist, P., Kongkitkul, W., Wonglert, A., Youwai, S., 2013. Investiga- (2), 749–770.
tion of failure behavior of continuous rock mass around cavern under high internal Zhou, S.W., Xia, C.C., Zhao, H.B., Mei, S.H., Zhou, Y., 2017. Numerical simulation for
pressure. Tunn. Undergr. Space Technol. 34, 110–123. the coupled thermo-mechanical performance of a lined rock cavern for underground
Tunsakul, J., Jongpradist, P., Soparat, P., Kongkitkul, W., Nanakorn, P., 2014. Analysis compressed air energy storage. J. Geophys. Eng. 14 (6), 1382–1398.
of fracture propagation in a rock mass surrounding a tunnel under high internal Zhou, S., Zhang, C., Xu, Y., Yuan, Y., 2022b. A hybrid phase field method for modeling
pressure by the element-free Galerkin method. Comput. Geotech. 55, 78–90. thermal fractures in brittle rocks: fracture diversity from a modified driving force.
Wang, S., Sloan, S., Tang, C., Zhu, W., 2012. Numerical simulation of the failure Int. J. Fract. 1–17.
mechanism of circular tunnels in transversely isotropic rock masses. Tunn. Undergr. Zhou, S., Zhuang, X., Rabczuk, T., 2019. Phase field modeling of brittle compressive-
Space Technol. 32, 231–244. shear fractures in rock-like materials: A new driving force and a hybrid formulation.
Yang, S.Q., Huang, Y.H., Jing, H.W., Liu, X.R., 2014. Discrete element modeling on Comput. Methods Appl. Mech. Engrg. 355, 729–752.
fracture coalescence behavior of red sandstone containing two unparallel fissures Zhou, S., Zhuang, X., Rabczuk, T., 2020b. Phase field method for quasi-static hydro-
under uniaxial compression. Eng. Geol. 178, 28–48. fracture in porous media under stress boundary condition considering the effect of
You, T., Waisman, H., Zhu, Q.Z., 2021. Brittle-ductile failure transition in geomaterials initial stress field. Theor. Appl. Fract. 107, 102523.
modeled by a modified phase-field method with a varying damage-driving energy Zhou, S., Zhuang, X., Zhou, J., Liu, F., 2021. Phase field characterization of rock
coefficient. Int. J. Plast. 136, 102836. fractures in Brazilian splitting test specimens containing voids and inclusions. Int.
You, T., Zhu, Q.Z., Li, P.F., Shao, J.F., 2020. Incorporation of tension-compression J. Geomech. 21 (3), 04021006.
asymmetry into plastic damage phase-field modeling of quasi brittle geomaterials. Zhou, S., Zhuang, X., Zhu, H., Rabczuk, T., 2018b. Phase field modelling of crack
Int. J. Plast. 124, 71–95. propagation, branching and coalescence in rocks. Theor. Appl. Fract. Mech. 96,
Yu, Z., Shao, J.F., Vu, M.N., Armand, G., 2021. Numerical study of thermo-hydro- 174–192.
mechanical responses of in situ heating test with phase-field model. Int. J. Rock Zhu, F., Tang, H., Liu, F., Zhang, X., 2022a. Adaptive fourth-order phase field method
Mech. Min. Sci. 138, 104542. for rock fractures using novel refinement criteria and improved data transfer
Zhang, K., Liu, X., Liu, W., Zhang, S., 2021. Influence of weak inclusions on the operators. Comput. Geotech. 151, 104987.
fracturing and fractal behavior of a jointed rock mass containing an opening: Zhu, F., Tang, H., Zhang, X., Papazafeiropoulos, G., 2022b. Fourth-order hybrid phase
Experimental and numerical studies. Comput. Geotech. 132, 104011. field analysis with non-equal order elements and dual meshes for simulating crack
Zhang, Q.H., Shi, G.H., 2021. Verification of a DDA-based hydro-mechanical model and propagation. Comput. Geotech. 142, 104587.
its application to dam foundation stability analysis. Int. J. Rock Mech. Min. Sci. Zhu, Z., Wang, J., Wu, M., 2022c. DEM simulation of particle crushing in a triaxial
138, 104627. test considering the influence of particle morphology and coordination number.
Zhang, X., Sloan, S.W., Vignes, C., Sheng, D., 2017a. A modification of the phase-field Comput. Geotech. 148, 104769.
model for mixed mode crack propagation in rock-like materials. Comput. Methods Zhuang, X., Li, X., Zhou, S., 2022. Three-dimensional phase field feature of longitudinal
Appl. Mech. Engrg. 322, 123–136. hydraulic fracture propagation in naturally layered rocks under stress boundaries.
Zhang, X., Vignes, C., Sloan, S.W., Sheng, D., 2017b. Numerical evaluation of the phase- Eng. Comput. 1–24.
field model for brittle fracture with emphasis on the length scale. Comput. Mech. Zhuang, X., Zhou, S., 2020. An experimental and numerical study on the influence
59 (5), 737–752. of filling materials on double-crack propagation. Rock Mech. Rock Eng. 53 (12),
5571–5591.

21

You might also like