Professional Documents
Culture Documents
Effect of Rock Joints On Lined Rock Caverns - 2023 - Journal of Rock Mechanics A
Effect of Rock Joints On Lined Rock Caverns - 2023 - Journal of Rock Mechanics A
Effect of Rock Joints On Lined Rock Caverns - 2023 - Journal of Rock Mechanics A
Effect of rock joints on lined rock caverns subjected to high internal gas
pressure
Davi Rodrigues Damasceno*, Johan Spross, Fredrik Johansson
Division of Soil and Rock Mechanics, KTH Royal Institute of Technology, SE-100 44, Stockholm, Sweden
a r t i c l e i n f o a b s t r a c t
Article history: The storage of hydrogen gas in lined rock caverns (LRCs) may enable the implementation of the first
Received 23 June 2022 large-scale fossil-free steelmaking process in Sweden, but filling such storage causes joints in the rock
Received in revised form mass to open, concentrating strains in the lining. The structural interaction between the LRC components
12 September 2022
must be able to reduce the strain concentration in the sealing steel lining; however, this interaction is
Accepted 14 November 2022
Available online 16 December 2022
complex and difficult to predict with analytical methods. In this paper, the strain concentration in LRCs
from the opening of rock joints is studied using finite element (FE) analyses, where the large- and small-
scale deformation behaviors of the LRC are coupled. The model also includes concrete crack initiation and
Keywords:
Lined rock cavern (LRC)
development with increasing gas pressure and rock joint width. The interaction between the jointed rock
Rock joints mass and the reinforced concrete, the sliding layer, and the steel lining is demonstrated. The results show
Strain concentrations that the rock mass quality and the spacing of the rock joints have the greatest influence on the strain
Lining interaction distributions in the steel lining. The largest effect of rock joints on the maximum strains in the steel
Finite element (FE) analysis lining was observed for geological conditions of “good” quality rock masses.
High gas pressure Ó 2023 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).
https://doi.org/10.1016/j.jrmge.2022.11.011
1674-7755 Ó 2023 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1626 D.R. Damasceno et al. / Journal of Rock Mechanics and Geotechnical Engineering 15 (2023) 1625e1635
Fig. 1. Conceptual model of (a) the main LRC components and (b) the opening of rock joints and cracking of the concrete layer from gas pressurization of the LRC.
lining. However, large strain concentrations can increase the risk or sharp edges in the 3D geometry. Zhou et al. (2020) used a
for gas leakage because of a rupture of the steel lining. The local thermo-mechanical damage model to study the long-term stability
effect of strain concentrations in the steel lining due to the opening of LRCs for CAES. They analyzed the influence of intersecting weak
of rock joints therefore needs to be considered in the design of LRCs. planes in the rock mass and rock mass heterogeneity on the LRC
For the LRC design, Johansson et al. (1995) presented an behavior. Other numerical models are available for similar under-
analytical model to calculate the wall displacement of rock caverns ground gas storages (e.g. Kim et al., 2012, 2013; Carranza-Torres
given an internal gas pressure. The derivations and a more concise et al., 2017; Zhou et al., 2017); however, they consider internal
form of these equations were presented by Damasceno et al. gas pressures less than 10 MPa, and small-scale effects are dis-
(2022). This model assumed a long rock cavern and isotropic in regarded. Thus, there is a lack of numerical models that concisely
situ stress conditions. Furthermore, Johansson et al. (1995) and capture the complex interaction between all components in an LRC
Johansson (2003) presented another analytical model to predict the and take into account the local strains induced by the opening of
strain concentration in the steel lining caused by the opening of rock joints because of the pressurization of the cavern.
concrete cracks. This latter solution included the effect of the In this paper, 2D FE analyses are used to investigate the con-
sliding layer as a frictional interaction between the steel lining and centration of strains in the steel lining because of the opening of
the concrete layer, and assumed equally spaced concrete cracks. rock joints in LRCs subjected to high internal gas pressure. The 2D
These analytical models contribute to the conceptual understand- geometry represents the horizontal cross-section at the LRC mid-
ing of the LRC deformation behavior and are easy to implement; height. The FE model includes the interaction between the main
however, their applicability is limited to fundamental problems LRC components and coupling between large- and small-scale LRC
with simplified LRC geometry, geological conditions, and lining deformation behaviors. Concrete crack initiation and development
interactions. Therefore, numerical modeling is needed to analyze because of the opening of rock joints with increasing gas pressure
the effect of opening rock joints and the complex lining interactions are considered. This concrete cracking induces strain concentra-
for the LRC design. tions in the steel lining. The internal radius, material properties,
The available numerical models for LRCs discussed in the liter- anisotropic in situ stresses, and internal gas pressure of the FE
ature usually only account for the large-scale deformation behavior, model are chosen based on the LRC in Skallen (Glamheden and
i.e. there is no consideration of the opening of rock joints or the Curtis, 2006). The interaction between individual LRC compo-
lining interactions. For example, Lu (1998) used a three- nents is demonstrated. Strain distributions in the steel lining are
dimensional (3D) finite element (FE) model to simulate field gas studied where the largest rock joint width increase occurs, i.e. at
pressure tests performed in a pilot LRC in Grängesberg, Sweden, the location of the maximum tangential strain in the rock cavern
and Glamheden and Curtis (2006) presented two-dimensional (2D) wall. The effect of geological conditions is investigated with respect
finite difference models simulating the wall convergence of the to the rock mass quality and the spacing of the rock joints, and the
rock cavern during the excavation in Skallen, Sweden. Damasceno influence of lining interactions is analyzed for different behaviors of
et al. (2022), on the other hand, used analytical 2D and 3D FE the reinforcement and sliding layer. Furthermore, the maximum
models to investigate the effect of horizontal in situ stress anisot- strains in the steel lining are studied, in a separate analysis, with
ropy and cavern height on the rock cavern response to an internal respect to the discrete rock joints and considering the effect of
gas pressure. For application in compressed air energy storage tangential strains in the rock cavern wall.
(CAES), analyses of LRC design issues using computational models
were performed by Perazzelli and Anagnostou (2016). They inves- 2. Modeling of the LRC
tigated failure related to rock mass uplifting, rock mass deforma-
tion, steel lining buckling, steel lining fatigue, and concrete plug 2.1. Model assumptions
stability. Park et al. (2013) proposed a probabilistic-based structural
design for LRCs against the ultimate failure of the steel lining using If the rock cavern is excavated deep enough and the cupola and
analytical and numerical methods, but disregarded the concen- invert of the LRC are spherical, the maximum strain in the steel
tration of strains due to the opening of rock joints in the rock mass lining is located close to the LRC mid-height. In this case, the critical
D.R. Damasceno et al. / Journal of Rock Mechanics and Geotechnical Engineering 15 (2023) 1625e1635 1627
Fig. 2. FE setups for horizontal cross-sections at the LRC mid-height: (a) Large-scale model (one quarter circumference); and (b) Small-scale model (one rock joint spacing arc
length).
failure mode is fatigue of the steel lining due to cyclic fillinge paper with applied loads and boundary conditions are presented in
emptying of the LRC. Considering that the LRC has a large height/ Fig. 2. These two FE model geometries are used to account for both
diameter ratio (cylindrical shape), a 2D plane-strain simplification the large-scale (more general) deformation behavior and the small-
at the horizontal cross-section of the LRC mid-height can be used scale (more detailed) deformation behaviors of the LRC. The dis-
for the analysis. The LRC design concept presumes that the built placements at the LRC wall from the large-scale model are im-
facility is completely gas tight, thus gas migration in fractures ported to the boundaries of the small-scale model. Static analysis is
would take place only if the steel lining is ruptured. Therefore, the used to simulate an “excavation step” (large-scale model), and
model does not consider gas leakage. In the case of a real structural quasi-static analysis is used for better convergence of the concrete
failure, gas leakage would be detected by the gas detection system. cracking in a “pressurization step” (large- and small-scale models).
However, post-failure behavior belongs to the general risk man- In accordance with the field tests presented by Mansson et al.
agement for the LRC, and is not within the scope of this paper. (2006), an LRC internal gas pressure of pi ¼ 25 MPa is applied
The FE simulations are analyzed for the maximum internal gas over 20 d, which represents a quasi-static process.
pressure in the LRC, which gives the largest width increase of rock Symmetry for the large-scale model geometry is taken at one
joints. Most of the plastic strains in the rock mass take place during quarter of the LRC horizontal cross-section (Fig. 2a), where the rock
the first storage filling, while the following storage fillinge cavern radius is 19 m. The whole model has a dimension size of
emptying cycles have a smaller strain range. For reference, typical 1000 m 1000 m. The “excavation step” includes only the rock
ultimate strains for ductile steels are in the order of 15%, and the cavern with the minimum and maximum horizontal in situ
strain range for the fatigue failure of the steel lining corresponds to stresses, sh and sH, respectively, applied in the rock mass. The
6& (for a monthly storage fillingeemptying cycle as considered in resulting stress field from the “excavation step” is imported to the
Damasceno et al. (2022)). Positive sign notation is adopted for rock mass of the “pressurization step”, which includes an 1-m thick
extension since tensile strains acting tangentially on the LRC are the concrete lining. Thus, the stresses in the lining are zero before the
critical strains for the analyses. gas pressure (pi ¼ 25 MPa) is applied to the lining internal surface.
Mansson et al. (2006) presented field tests at the LRC in Skallen Structured quadrilateral plane-strain elements (CPE4R) are used in
where fillingeemptying storage cycles were performed. Gas the whole model.
compression during filling increases the storage temperature, For the small-scale model, a portion of the large-scale lining
expanding the LRC wall, while gas expansion during emptying with arc length equal to one rock joint spacing, sj , is used (Fig. 2b).
decreases the storage temperature, contracting the LRC wall. In addition to the concrete layer, this model includes one rock joint,
Therefore, compared to constant temperature cases, the gas tem- the concrete reinforcement, and the steel lining. The intact rock
perature variation results in slightly smaller radial displacements at blocks are modeled with5-cm thickness. The thickness of the intact
maximum gas pressure and slightly larger wall radial displace- rock blocks has no major influence on the cracking of the concrete
ments at minimum gas pressure. The thermal influence is limited to layer, since the rock mass deformation behavior is governed by the
the surface of the LRC wall, and compared to the steel lining strain opening of rock joints rather than internal deformations in the
concentrations at concrete cracks, the magnitude of thermal strains intact rock blocks. The rock joint is modeled as a discrete contact
is negligible. Therefore, constant temperature is considered in the between the rock blocks. The orientation of the rock joints is
FE analyses in this paper. modeled orthogonally to the cavern walls because this results in
maximum opening of the rock joints and represents a worst case
scenario. As the LRC is a very large facility, it is likely that orthog-
2.2. FE model setup
onal or near-orthogonal rock joints would be present somewhere at
the rock cavern wall. Non-orthogonally oriented rock joints would
In this paper, the commercial software package Abaqus v2020
have larger resisting frictional forces at the contact because of
(Dassault Systèmes Simulia Corp, 2019) is used for all the FE ana-
higher normal stresses acting on the joint, resulting in smaller rock
lyses, i.e. geometry, materials, boundary conditions, loads, meshing,
joint opening. Since tensile opening of the rock joints, which have
simulation, and visualization. The FE model geometries used in this
1628 D.R. Damasceno et al. / Journal of Rock Mechanics and Geotechnical Engineering 15 (2023) 1625e1635
negligible tensile strength, occurs in the analyses, the strength The input parameters for the geological conditions considered
parameters of the contact do not have any influence on the inter- in the simulations are summarized in Table 1. A value of GSI ¼ 70 is
face behavior and are therefore not considered. used, which is similar to the values observed at the LRC in Skallen.
The reinforcement is embedded into the concrete layer. Dis- Based on the rock grade classification by Brown (1981), typical rock
placements from the large-scale model are transmitted to the types for hard rock mass conditions, e.g. marble, sandstone, granite,
boundaries of the steel lining using a tie constraint between the granodiorite, and gneiss, are within the “very strong” rock grade.
line edges and the laterals of the concrete layer. A cohesivee Based on Hoek (1983) and Hoek and Brown (1997), the values of
frictional interface interaction is used for the shotcrete at the rock fci ¼ 175 MPa and mi ¼ 25 are used as input parameters in the FE
blockseconcrete layer contact, and a frictional interface interaction model since they give the same average tangential strain at the LRC
is used for the sliding layer at the concrete layeresteel lining con- wall as observed in Skallen (Mansson et al., 2006). At the LRC mid-
tact. Movements at the boundaries of the small-scale model with height depth of 141 m, the horizontal in situ stresses are sH ¼ 8:3
respect to the displacements in the large-scale model are allowed MPa and sh ¼ 4:1 MPa (Glamheden and Curtis, 2006). The
by including a frictionless line interface at the outer boundaries of magnitude of sco ¼ 6:2 MPa is used as the confinement pressure to
the concrete layer and rock blocks. The small-scale model considers calculate the rock mass properties (Hoek et al., 2002) to be used
only one “pressurization step”, where the magnitude of pi ¼ 25 MPa during the rock cavern excavation. For a maximal internal gas
is applied to the internal surface of the steel lining, and the dis- pressure of pi ¼ 25 MPa, the shear yield limit of the rock mass is
placements from the large-scale “pressurization step” are applied not exceeded; therefore, no residual parameters are needed.
to the other boundaries of the small-scale model. Unstructured Following Hoek and Brown (1997)’s suggestions for “very good”
triangular plane-strain elements (CPE3) are used in the concrete quality hard rock masses, the Poisson’s ratio is set to 0.2. The
layer to allow greater flexibility for the cracking, and truss (T2D2) dilation angles for the intact rock and rock mass are estimated from
and beam (B21) elements are used for the reinforcement and steel Alejano and Alonso (2005). For rock joints oriented orthogonally to
lining, respectively. the rock cavern walls and assumed average values for “good”
The simulation of concrete crack initiation and propagation quality rock mass, a value of sj ¼ 1 m is assumed based on
follows the damaged plasticity material model (Lubliner et al., Bieniawski (1989)’s rock mass rating definition.
1989; Lee and Fenves, 1998). The concrete material is brittle, thus
the thickness of the concrete layer has no major influence on the
concrete cracking. An FE element size of le ¼ 1:9 cm yields well-
developed concrete cracks in the small-scale model and is there- 2.3.2. Concrete layer
fore used. This le is smaller than the maximum allowed FE element The guidelines from Malm (2016) regarding FE analysis using
size calculated from Rijkswaterstaat (2012). Based on Cervenka the concrete damaged plasticity model are followed. The damaged
et al. (2020), the expression for the concrete crack width is plasticity model was developed by Lubliner et al. (1989) and Lee
derived as wc ¼ 1:16le εcr , where εcr is the crack opening strain, i.e. and Fenves (1998) and gives a good description of the concrete
the plastic strain in tension. cracking behavior, as verified by other authors, e.g. Malm and
Holmgren (2008a, b), Genikomsou and Polak (2015), and Xiao
2.3. Characterization of LRC components et al. (2017). This material model allows for the simulation of
crack initiation and development by introducing a reduction in the
2.3.1. Jointed rock mass element stiffness. The default values of eccentricity (0:1), biaxial
Tensile and shear failures are allowed for both rock mass and
intact rock, following the Rankine and MohreCoulomb failure
Table 1
criteria, respectively. It is assumed that the blasting disturbance Properties of the jointed rock mass used in the analyses for the large-scale model
caused by the excavation is negligible. A “very strong” intact rock (“excavation step” and “pressurization step”) and small-scale model (“pressurization
and “good” quality rock mass are chosen to represent typical step”).
geological conditions of a hard-jointed rock mass, similarly to the Parameter Large-scale model (rock Small-scale model (intact rock
geological conditions at Skallen. The depth and horizontal in situ mass properties) properties) (pressurization step)
stresses of the LRC in Skallen (Glamheden and Curtis, 2006) are Excavation Pressurization
used in the model simulations. step step
The rock mass properties including Young’s modulus, cohesion, Rock cavern 19 19 19
friction angle, and tensile strength are estimated based on Hoek radius (m)
et al. (2002). The input parameters to estimate these rock mass Joint spacing, sj e e 1
properties are the geological strength index (GSI), the unconfined (m)
Density (kg/ 2700 2700 2700
compressive strength of the intact rock, fci , the intact rock constant,
m3)
mi , and the confining pressure in the rock mass, sco . When simu- Young’s 41.8 41.8 57.1
lating the excavation of the rock cavern, the average horizontal in modulus
situ stress is considered as the confining pressure, i.e. sco ¼ ðsh þ (GPa)
sH Þ=2, while sco ¼ pi is used during gas pressurization of the LRC Poisson’s ratio 0.2 0.2 0.2
Cohesion 3.9 7.1 22.6
because pi > ðsh þ sH Þ=2. (MPa)
The rock blocks have stiffer/stronger properties compared to the Friction angle 62.2 54.1 58.7
rock mass, favoring the opening of the rock joint instead of internal ( )
deformation of the rock blocks. The intact rock properties are Compressive 31.5 43.8 161.3
strength
chosen to represent the rock blocks in the small-scale model, which
(MPa)
yields a more conservative analysis (larger width increase of the Tensile 0.7 0.7 7
rock joint). The Young’s modulus of the intact rock is approximated strength
as proposed by Hoek and Diederichs (2006). The equivalent intact (MPa)
rock properties for cohesion, friction angle, and tensile strength are Dilation angle 9.8 4.9 5.4
( )
obtained from Hoek et al. (2002) using GSI ¼ 100.
D.R. Damasceno et al. / Journal of Rock Mechanics and Geotechnical Engineering 15 (2023) 1625e1635 1629
failure ratio (1:16), and shape of yield surface (0:67) suggested in Table 3
Abaqus v2020 (Dassault Systèmes Simulia Corp, 2019) are used. Properties of reinforcement used in the analyses.
The stressestrain relationship for the concrete under compres- Parameter Value
sion depends on the Young’s modulus of the concrete, the Diameter of rebar (cm) 1.6
compressive strength of the concrete at 28 d, and the concrete Spacing of rebar, sr (cm) 15
strain at maximum stress (EN 1992-1-1:2005, 2005a). For the Distance from steel lining (cm) 10
tensile behavior, the stress is expressed in terms of the crack Density of steel (kg/m3) 8000
Young’s modulus of steel (GPa) 200
displacement, as defined by Cornelissen et al. (1986). For normal-
Poisson’s ratio of steel 0.3
density concrete, the first and second material constants are 3 Yield strength of reinforcement (MPa) 500
and 6:93, respectively, and the stress-free crack width is calculated Ultimate strength of reinforcement (MPa) 540
based on the fracture energy, as estimated from the Fib (2010) Elongation at maximum stress of steel (%) 5
model code and the tensile strength of concrete.
The concrete properties are listed in Table 2. A concrete layer
with the thickness of 1 m and class C30/37 is chosen for the sim- Table 4
ulations, corresponding to a design used in the development of the Properties of the steel lining used in the analyses.
LRC concept (Johansson, 2003). Based on the studies by Malm Parameter Value
(2006), a suitable value of the concrete dilation angle for normal- Thickness of steel plates (cm) 1.2
grade concrete with a high degree of confining pressure is chosen Density of steel (kg/m3) 8000
to be 30 . In compression, the behavior of the concrete is assumed Young’s modulus of steel (GPa) 200
to be elastic up to 40% of the compressive strength of concrete at 28 Poisson’s ratio of steel 0.3
Yield strength of steel plates (MPa) 355
d.
Ultimate strength of steel plates (MPa) 490
Strain at hardening initiation (%) 1.2
2.3.3. Reinforcement Elongation at maximum stress of steel (%) 5
The properties of the reinforcement are listed in Table 3. In the
small-scale model, the concrete reinforcement of type B500B (EN
1993-1-9:2005, 2005b) is used. Johansson (2003)’s observations Table 5
from LRC experiments showed that a good concrete crack distri- Interface properties of the shotcrete and sliding layer used in the analyses.
bution close to the steel lining was achieved for a reinforcement Parameter Value
placed at a clearance of 10 cm from the steel lining, with a rebar
Adhesion of the shotcrete interface (MPa) 0.81
diameter of 1.6 cm, and a rebar spacing of sr ¼ 15 cm, which are Cohesion of the shotcrete interface (MPa) 1.62
therefore used in the simulations. The rebar spacing for the con- Friction coefficient of the shotcrete interface 1
crete reinforcement is modeled in 2D as an equivalent cross-section Friction coefficient of the sliding layer interface, ms 0.35
area.
2.3.4. Steel lining to the rock surface is about 0.81 MPa. Using the Griffith criterion,
The properties of the steel lining are listed in Table 4. For the the cohesion of the rock blockseconcrete layer interface is esti-
steel plates, the steel of type S355 with a plate thickness of 1.2 cm mated to be 1.62 MPa, i.e. twice the adhesion. Based on the labo-
was tested in the development of the LRC design (Johansson, 2003), ratory work by Krounis et al. (2016), an average friction coefficient
and this steel lining is therefore adopted in this paper. It is assumed value of 1 is assumed for the rock blockseconcrete layer interface.
that, after yielding initiation, the steel elongates 1:2% before strain The sliding layer for the LRC is modeled as an interface inter-
hardening occurs. Degradation of the steel properties due to, for action between the concrete layer and the steel lining. Baltay and
example, corrosion and hydrogen embrittlement is associated with Gjelsvik (1990) showed that, under high pressures, the static co-
the long-term behavior of the LRC and is not included. efficient of friction, ms , between concrete and steel can reach
magnitudes close to 0.6. If an intermediate thin layer of sliding
material, e.g. bitumen, is added between the concrete and steel, the
2.3.5. Shotcrete, sliding layer, and drainage system
value of ms can be decreased to as low as 0.1 (Johansson, 2003).
The interface properties used for the shotcrete and sliding layer
Because it is difficult to ensure that the sliding material homoge-
are listed in Table 5 and are explained in the next paragraphs. The
neously covers the whole concrete layeresteel lining interface, an
drainage system should not have any significant influence on the
average value of ms ¼ 0:35 is assumed for the sliding layer.
mechanical behavior of the LRC and is not included in the FE model.
The shotcrete for the LRC is modeled as a cohesive interface
between the rock blocks and the concrete layer. Bjureland et al. 2.4. List of analyses
(2019) showed that the mean value for the adhesion of shotcrete
A summary of 42 FE simulations analyzed in this paper is shown
in Table 6. In Section 3, 19 FE simulations are performed to study the
Table 2
Properties of the concrete layer used in the analyses. interaction between LRC components. Additional 12 FE simulations
are done in Section 4 to analyze strain distributions in the steel
Parameter Value
lining. To investigate the influence of rock joints on the maximum
Thickness of the concrete layer (m) 1 strains in the steel lining, additional 11 FE simulations are run in
Density of concrete (kg/m3) 2400
Section 5.
Young’s modulus of concrete (GPa) 33
Poisson’s ratio of concrete 0.2 The interactions between LRC components, which contribute to
Compressive strength of concrete at 28 d (MPa) 38 the final strain distribution in the steel lining, are studied in Section
Tensile strength of concrete (MPa) 2.9 3, considering the conditions of hard-jointed rock mass and lining
Concrete strain at maximum stress (%) 0.22 setup described in Sections 2.2e2.4. The influence of the in situ
Dilation angle of concrete ( ) 30
stress anisotropy on the width increase of rock joints is investigated
1630 D.R. Damasceno et al. / Journal of Rock Mechanics and Geotechnical Engineering 15 (2023) 1625e1635
Fig. 4. Simulation results of (a) concrete cracking patterns (tensile damage contour) and (b) crack width distribution along the radial thickness of the concrete layer. The definition
of xs is shown in (a) at qh ¼ 88:5 .
3.3. Response of lining components to the cracking of the concrete suitable sc value for the analytical model, and the maximum εs can
layer be overestimated if the steel yields without strain hardening. The
presented FE model is more robust and can better capture the in-
In Fig. 5, the interaction of the concrete cracks with the main teractions between the LRC components; therefore, this FE model is
lining components, i.e. the sliding layer, reinforcement, and steel
lining, is demonstrated. Fig. 5a shows the values of wc and the steel
lining slip length, ls , with respect to the tangential distance from
the rock joint, xj . Fig. 5b shows the corresponding values of strain in
the steel lining, εs , and the reinforcement strain, εr , along xj . Con-
crete cracks that cause strain concentrations in the steel lining have
wc > 0:05 mm; however, the two largest εs are observed for wc >
0:2 mm at xj ¼ 30 cm and xj ¼ 20 cm. The largest slip of
maxðls Þ ¼ 0:04 mm and largest strains of maxðεr Þ ¼ 12& and
maxðεs Þ ¼ 13& occur close to the location of the two largest
concrete cracks. Locations of peak strains for the reinforcement are
slightly shifted due to the diagonal development of the concrete
cracks from the rock joint. The steel lining slip and strain of the
reinforcement take place at the concrete crack locations, reducing
strain concentrations in the steel lining.
further analyzed in the next section for strain distributions in the Table 8
steel lining. Rock mass quality scenarios for the analysis.
Rock mass quality GSI εqw (&) min(sj) (m) mode(sj) (m) max(sj) (m)
Fig. 7. Analyses of the influence of the reinforcement and steel lining on the redis-
tribution of concrete cracks and strain concentrations in the steel lining. Distributions
of εs are shown for reference model and cases with ms ¼ 0 and ms / N considering (a)
the unreinforced concrete layer and (b) the reinforced concrete layer. “*” denotes that
the reference model from Sections 2.2e2.4 corresponds to sr ¼ 15 cm and ms ¼ 0:35.
layer, and can better simulate the expected LRC behavior ls Steel lining slip length relative to the concrete layer
than the analytical model from Johansson (2003) (derived in mi Intact rock constant
Appendix A). This FE model included the complex interaction pi Pressure inside the lined rock cavern
between LRC components and can be used to calculate sc Average spacing of concrete cracks
maximum strains in the steel lining in the design of the LRC. sj Average spacing of rock joints
(2) Attained cavern wall tangential strain and the rock joint sr Spacing of reinforcement rebar
spacing were in combination with the most critical factors wc Width of concrete crack
for the concentration of strains in the steel lining. Thus, a wj Width of rock joint
discrete rock joint has the largest influence on the LRC design xs Radial distance from the steel lining
for geological conditions of “good” quality rock masses xj Tangential distance from the rock joint
(GSI ¼ 70), because such geological conditions combine εcr Crack opening strain
moderate values for both cavern wall tangential strain and εr Strain in the reinforcement
rock joint spacing. For better (GSI ¼ 80) or poorer (GSI ¼ εs Strain in the steel lining
60) rock mass qualities, the combined effect was less, even εqw Tangential strain in the rock cavern wall
though individual factors were larger. qh Angle from the direction of minimum horizontal in situ
(3) The smallest and most evenly distributed strains in the steel stress
lining were achieved for the case of a reinforced concrete ms Friction coefficient of the concrete layeresteel lining
layer with a rough contact interaction at the concrete layere contact
steel lining interface. For this case, the steel lining acted as an sco Confining pressure in the rock mass
additional reinforcement and further reduced the size of sh Minimum horizontal in situ stress
concrete cracks. sH Maximum horizontal in situ stress
Prestudy for Work Package 2.3 in HYBRIT Research Program 1. KTH Royal Rijkswaterstaat, 2012. Guidelines for Nonlinear Finite Element Analysis of Concrete
Institute of Technology, Stockholm, Sweden. Technical Report TRITA-ABE-RPT- Structures, Scope: Girder Members. Ministry of Infrastructure and Environ-
182. ment, Delft, Netherlands.
Kim, H.M., Rutqvist, J., Ryu, D.W., Choi, B.H., Sunwoo, C., Song, W.K., 2012. Exploring Tengborg, P., Johansson, J., Durup, J.G., 2014. Storage of highly compressed gases in
the concept of compressed air energy storage (CAES) in lined rock caverns at underground lined rock caverns e more than 10 years of experience. In: Tun-
shallow depth: a modeling study of air tightness and energy balance. Appl. nels for a Better Life: Proceedings of the World Tunnel Congress 2014. Inter-
Energy 92, 653e667. national Tunnelling and Underground Space Association (ITA), Foz do Iguaçu,
Kim, H.M., Rutqvist, J., Jeong, J.H., Choi, B.H., Ryu, D.W., Song, W.K., 2013. Charac- Brazil.
terizing excavation damaged zone and stability of pressurized lined rock cav- World Steel Association, 2021. Climate Change and the Production of Iron and Steel.
erns for underground compressed air energy storage. Rock Mech. Rock Eng. 46, https://worldsteel.org/publications/policy-papers/climate-change-and-the-
1113e1124. production-of-iron-and-steel/.
Krounis, A., Johansson, F., Larsson, S., 2016. Shear strength of partially bonded Xiao, Y., Chen, Z., Zhou, J., Leng, Y., Xia, R., 2017. Concrete plastic-damage factor for
concreteerock interfaces for application in dam stability analyses. Rock Mech. finite element analysis: concept, simulation, and experiment. Adv. Mech. Eng. 9
Rock Eng. 49, 2711e2722. (9). https://doi.org/10.1177/1687814017719642.
Lee, J., Fenves, G.L., 1998. Plastic-damage model for cyclic loading of concrete Zhou, S.W., Xia, C.C., Zhao, H.B., Mei, S.H., Zhou, Y., 2017. Numerical simulation for
structures. J. Eng. Mech. 124 (8), 892e900. the coupled thermo-mechanical performance of a lined rock cavern for un-
Lu, M., 1998. Finite element analysis of a pilot gas storage in rock cavern under high derground compressed air energy storage. J. Geophys. Eng. 14 (6), 1382e1398.
pressure. Eng. Geol. 49, 353e361. Zhou, S., Xia, C., Zhou, Y., 2020. Long-term stability of a lined rock cavern for
Lubliner, J., Oliver, J., Oller, S., Oñate, E., 1989. A plastic-damage model for concrete. compressed air energy storage: thermo-mechanical damage modeling. Eur. J.
Int. J. Solid Struct. 25 (3), 299e326. Environ. Civ. Eng. 24 (10), 2070e2093.
Malm, R., 2006. Shear Cracks in Concrete Structures Subjected to In-Plane Stresses.
KTH Royal Institute of Technology, Stockholm, Sweden.
Malm, R., 2016. Guideline for FE Analyses of Concrete Dams. Energiforsk, Stockholm, Davi Rodrigues Damasceno obtained his BSc degree in
Sweden, p. 159. Report 2016:270. Petroleum Engineering from the Federal University of
Malm, R., Holmgren, J., 2008a. Cracking in deep beams owing to shear loading. Part Espírito Santo, Brazil in 2014 and MSc degree in Structural
1: experimental study and assessment. Mag. Concr. Res. 60 (5), 371e379. Geomechanics from the Missouri University of Science
Malm, R., Holmgren, J., 2008b. Cracking in deep beams owing to shear loading. Part and Technology, USA in 2016. He is currently a PhD stu-
2: non-linear analysis. Mag. Concr. Res. 60 (5), 381e388. dent at the Division of Soil and Rock Mechanics in the KTH
Mansson, L., Marion, P., Johansson, J., 2006. Demonstration of the LRC gas storage Royal Institute of Technology, Sweden. His research focus
concept in Sweden. In: Proceedings of the 23rd World Gas Conference. Inter- is on the improvement of the lined rock cavern design
national Gas Union, Amsterdam, Netherlands, pp. 404e420. methodology for the storage of hydrogen gas. This work is
Park, D., Kim, H.M., Ryu, D.W., Choi, B.H., Han, K.C., 2013. Probability-based struc- a contribution to the Hydrogen Breakthrough Ironmaking
tural design of lined rock caverns to resist high internal gas pressure. Eng. Geol. Technology (HYBRIT) initiative, which is a joint venture
153, 144e151. project between major companies in the mining, iron-
Perazzelli, P., Anagnostou, G., 2016. Design issues for compressed air energy storage making, and energy sectors, to implement a fossil-free
in sealed underground cavities. J. Rock Mech. Geotech. Eng. 8 (3), 314e328. steelmaking process in Sweden. His work involves
reliability-based design within geotechnical engineering using finite element analysis.