Sei sulla pagina 1di 20

EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2017; 46:159–178


Published online 11 July 2016 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2780

Numerical modelling of reinforced-concrete structures under seismic


loading based on the finite element method with discrete inter-
element cracks

Željana Nikolić*,†, Nikolina Živaljić, Hrvoje Smoljanović and Ivan Balić


Faculty of Civil Engineering, Architecture and Geodesy, University of Split, Split, Croatia

SUMMARY
The behaviour of the earthquake excitated reinforced concrete (RC) structures up to the collapse, previously
designed according to the prescriptions of Eurocode 8 cannot be fully predetermined because of the diversity
of the nature of the earthquake. Thus, one needs to make a judicious choice between the two ductility classes
balancing between the desired response of the structure, dimension of structural elements, costs and com-
plexity of the design and construction. Powerful numerical models stand as significant tools to obtain more
insight into the peculiarity of each structure. This paper presents a numerical model for dynamic analysis and
prediction of the collapse of RC structures under the seismic loading based on finite discrete element
method. The model includes several numerical algorithms which are essential for realistic description of
highly non-linear effects of these structures: cyclic crack opening-closing mechanism based on discrete
cracks, interaction between the reinforcement and concrete taken by steel strain-slip relation, the influence
of adjacent cracks to the slip of reinforcing bar, local slip of reinforcing bar because of a high plastic defor-
mation under reversed cyclic loading, the influence of the curvature of reinforcing bar to yield stress reduc-
tion of the steel and cyclic behaviour of concrete and steel. Performance of the model was demonstrated
through incremental dynamic analysis of RC wall, previously designed according to the Eurocode 8 for me-
dium and high ductility classes. The presented results demonstrated the ability of the model for simulation of
the crack patterns and collapse of these structures. Copyright © 2016 John Wiley & Sons, Ltd.

Received 14 July 2015; Revised 3 June 2016; Accepted 6 June 2016

KEY WORDS: reinforced concrete structures; finite discrete element method; seismic loading; discrete
cracks; incremental dynamic analysis

1. INTRODUCTION

Modelling of the behaviour of engineering structures due to hazardous loading conditions such as
intensive seismic excitation represents a great challenge for researchers. Apart from knowing the
value of the failure load, it is very important to know the behaviour of the structure up to failure,
especially with respect to crack initiation and plastification of particular zones, behaviour of the
structure after reaching failure load including failure mechanisms and residual load capacity after the
end of earthquake.
Each country regulates safety and rational design of structures subjected to earthquake excitation
with its own technical regulations. The regulations of most countries allow the design of structures
using linear calculations, whereas the influence of nonlinearity is taken into account by applying the
behaviour factor which represents a very rough estimation of the real behaviour of the structure.
Moreover, the European standard for the earthquake resistance design of the structures (Eurocode 8,

*Correspondence to: Željana Nikolić, University of Split, Faculty of Civil Engineering, Architecture and Geodesy, Split,
Croatia.

E-mail: zeljana.nikolic@gradst.hr

Copyright © 2016 John Wiley & Sons, Ltd.


160 Ž. NIKOLIĆ ET AL.

further referred as EC8) [1] defines the classification of reinforced concrete (RC) structures in
seismically active areas into two ductility classes [medium ductility class (DCM) and high ductility
class (DCH)] and the appropriate calculation which corresponds to respective terms of the
regulations. Nonlinear analysis of such dimensioned structures sometimes shows that, despite the
strict observance of the rules, the behaviour of such structures subjected to earthquake does not
always correspond to the anticipated expectations. Instead of linear methods which use the behaviour
factors, the regulations leave the possibility of using non-linear methods. Nonlinear static (pushover)
analysis [2–4] represents one of the simplified non-linear methods, but it cannot include in detail all
the influences caused by the cyclic nature of earthquake loading. For these reasons time-dependent
analysis, which can consider material nonlinearity, cyclic loading and unloading and the loss of
energy during an earthquake, is the most insightful approach for a detailed analysis of the behaviour
of structures subjected to earthquake. This method is the basis for conducting the incremental
dynamic analysis, in which the ordinate of applied ground acceleration increases up to the failure of
the structure and allows the determination of the failure load as well as a monitoring of damage
increasing as a result of fracturing and plastification of the parts of the structure.
The only way, besides the costly experiments, to verify the behaviour of certain engineering
structure subjected to more different earthquakes and driven up to final failure is with time
dependent analysis conducted with advanced numerical models. The total collapse occurs because of
progressive localized failure of structural elements, where discontinuities make a significant role in
such process. When it comes to the RC structures, this is mostly revealed through the localized
zones where cracking of the concrete occurs together with buckling of the reinforcement and its
pulling out from concrete. Therefore, modelling of crack initiation and propagation is one of the key
factors that affect the reliability of the model for analysing structures subjected to earthquake loads.
The majority of the available non-linear numerical models for time-dependent analysis of RC
structures under seismic loading are based on finite element method and smeared crack approach,
where the cracked material is represented as a continuum and local displacement discontinuities are
smeared over some tributary areas. This approach is usually combined with classical constitutive
laws, such us plasticity or damage models. As the dynamic calculation of RC structures is extremely
demanding, modelling of the concrete fracturing through smeared crack approach can be used in
order to reduce computational cost, leading to the relatively good approximation of the global
structural response, like shown in [5]. However, the main failure mechanism in structures under the
earthquake’s excitation concerns the representation of the cyclic opening and closing of cracks,
when represented by discrete approach, leads to more accurate description of real cracks and
computation of energy dissipation.
Modelling of crack propagation and localized failure based on discrete cracks is possible through the
framework of finite element method using the cohesive elements across element edges [6, 7] or by
considering the adaptive refinement techniques [8]. Because of the complexity of these approaches,
the modelling of the discontinuity inside the finite elements became great challenge for researchers,
which resulted in finite elements with embedded strong discontinuities (ED-FEM method) [9, 10] or
extended finite elements (X-FEM method) [11, 12]. These approaches have been exploited for
modelling of cracking in RC structures [13] and for solving the dynamic problems as well [14, 15].
Representation of damage planes in the microstructure, such as those that exist at the contact
surfaces between aggregate and cement, can be successfully performed with microplane models [16,
17]. The constitutive framework of these models, based on the monitoring of stresses in different
predefined directions, is similar to discrete models with the difference that the model is formulated
through the continuum approach. The microplane model was recently applied in predicting the
phenomena related to dynamic fracture [18, 19] as well as nonlinear dynamic analysis of RC frame
structures under seismic loading [20].
The other possibility is discontinuum based modelling where discontinuum state of the structure is
assumed a priori, leading to discrete element method (DEM), firstly developed for granular assemblies
with the common idea of idealization of the material as assembly of particles with joints between
them. The latter has been extended to other heterogeneous brittle materials like concrete [21].
Discrete models in the form of lattice or particle models have been successfully applied in
modelling of the heterogeneity, damage localization and fracture at micro, meso and macro scales

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 161

of brittle materials [22]. An integral part of discretization is Delaunay/Voronoi tessellations. The main
approaches are lattice discrete models [23–29] and rigid body spring models [30, 31]. The last two
papers [30, 31] present crack propagation analysis of RC structures under monotonic and cyclic
loading using rigid body spring model. Although these models successfully simulate brittle material
fracture, only few works are dealing with the modelling of real RC structures under dynamic and
seismic loading.
In the last two decades an increasing number of models attempted to combine the advantages of
continuum-based and discontinuum-based modelling [32–34]. With respect to these works, this
paper brings one such approach, based on combined finite discrete element method [35], which is
taken as a basis for numerical model capable of simulating the static and dynamic analysis of RC
structures [36, 37]. Although the main strength of the presented model is in discrete
representation of cracks and discontinuities, the model is capable of simulating the deformation of
uncracked structural parts in continuum sense, which is the main characteristic of the finite
discrete element method. This is in contrast to referred discrete models that assume rigid
behaviour of particles.
The purpose of this paper is to present the application of the numerical model developed by Živaljic
[36, 37], combined with findings obtained from sensitivity analyses [38] related to spatial
discretization, slip model parameters and value of penalty coefficient in the numerical modelling of
the earthquake excitated realistic RC structures previously designed according to EC8 prescriptions
[1]. More precisely, the time-dependent analyses of few recorded characteristic earthquakes are
conducted for two ductility classes (DCM and DCH), where structural behaviour can be monitored
until the complete failure.
The model presented in this work tends to consider some of the most important effects which are
dominant in the behaviour of RC structures under the seismic loading until collapse. Among these is
discrete representation of cracks together with their initiation and propagation depending on position
of reinforcement. More precisely, the interaction between the reinforcement and concrete is
considered by steel strain-slip relation with the influence of adjacent cracks to the slip of reinforcing
bar, local slip of reinforcing bar because of a high plastic deformation under reversed cyclic loading
and the influence of the curvature of reinforcing bar to yield stress reduction of the steel. All these
phenomena occur under the earthquake’s cyclic excitation and, together with inertial effects due to
motion and contact between separated structural parts, dominantly influence the failure process and
energy dissipation. Previously mentioned phenomena are based on approximation of the
experimental curves for behaviour of the concrete and steel in the crack interface and are taken from
literature; thus they are used to develop numerical algorithms suitable for incremental dynamic
analysis of these structures in the finite-discrete element framework.
The paper consists of a description of the numerical model, performance of the model for cyclic
loading validated on two examples and full incremental dynamic analysis of real RC wall structure
previously designed according to the prescriptions of EC8 for two ductility classes (DCM and DCH).

2. MODELLING OF THE REINFORCED CONCRETE STRUCTURE

In the presented model, the concrete structure is discretized with constant strain triangular finite
elements, while the reinforcing bars are modelled with one-dimensional elements, which can be
placed inside the concrete finite elements. The material model in triangular finite elements, which
takes into account finite displacement and rotations, is linear elastic [35], while the material non-
linearity, fracture and fragmentation are considered through joint elements which are implemented
within the finite element mesh. The interaction between discrete elements is considered through the
contact interaction algorithm based on the principle of potential contact forces with penalty function
method [39] and the Coulomb-type law for friction [40]. The method uses an explicit numerical
integration of the equation of motion.
The model of the RC structure with an embedded reinforcing bar is shown in Figure 1. The
reinforcing bar is defined by its first and end points. The intersection between the sides of triangular

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
162 Ž. NIKOLIĆ ET AL.

Figure 1. Discretization of reinforced concrete structure.

concrete finite elements and reinforcing bars gives the reinforcing finite elements and reinforcing joint
elements (Figure 1).
The structure is assumed to behave as a linear elastic continuum until the initiation of the cracks and
discontinuities, which are allowed to propagate through the joint elements of concrete, leading to the
deformation in the reinforcing bar joint elements [37].

2.1. Concrete modelling


The concrete model assumes the discontinuities in the joint elements, where latter is used for
simulation of crack initiation and propagation in tension and shear [37]. Nonlinear behaviour of
concrete in tension and shear can be observed through the two major regimes: strain-hardening prior
to reaching the peak stress ft, which is implemented like shown in [41], and strain-softening
(Figure 2), reflected through joint elements behaviour, which is based on an approximation of the
experimental stress-displacement curves taken according to Hordijk [42].
Outside the cracks, the material bulk is linear elastic. Moreover, the structure under the compression
load usually fails because of Mode I and Mode II (tensile and shear failure). The reason for this
phenomenon is that the tensile crack opening and shear sliding due to reaching the tensile and shear
strength occur before the significant nonlinear deformation in compression. Observed macroscopic
nonlinearity in compression is a consequence of tensile and shear failure which is modelled inside
the cracks.
The cracks are assumed to coincide with the finite element edges, which have been achieved in
advance through the topology of adjacent elements being described by different nodes. The

Figure 2. Material model in concrete joint element.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 163

separation of these edges induces a bonding stress which is taken to be a function of the size of
separation δ (Figure 2). The area under stress-displacement curve represents the energy release rate
Gf =2γ, where γ is the surface energy, that is, the energy needed to extend the crack surface by a
unit area.
Before reaching tensile strength ft, there is no separation of element edges which is enforced through
the penalty method [41]. For separation δ ≤ δt the bonding stress is given by
"  2 #
2δ δ
σ cj ¼  ft (1)
δt δt

where δt = 2hft/p0 represents the separation corresponding to the bonding stress being equal to the tensile
strength ft, h is the size of particular element and p0 is penalty term. In that way, the separation of
adjacent elements before the reaching of the tensile strength is normalized by the element size [42].
The error in the displacements is controlled through setting penalty p0 as function of modulus of
elasticity Ec. Increase of penalty parameter produces more accurate solutions, but because of the
numerical stability, it causes smaller time steps and longer calculation time. The analysis of the
influence of penalty to the error in displacements presented in [38] has shown that the relative error
for p0 = 100Ec is less than 1%. The numerical stability is also ensured for chosen value of penalty
coefficient with appropriate time step.
After reaching a tensile strength ft stress decreases with an increasing separation δ and at δ = δc the
bonding stress tends to zero. For separation δt < δ < δc bonding stress is given by

σ cj ¼ z f t (2)

where z is a heuristic scaling function representing an approximation of the experimental stress-


displacement curves taken according to Hordijk [43]:
h i  
z ¼ 1 þ ðc1 Dt Þ3 ec2 Dt  Dt 1 þ c31 ec2 (3)

where c1 = 3 and c2 = 6.93, while the damage parameter Dt is determined according to the following
expression:

ðδ  δt Þ=ðδc  δt Þ; if δt < δ < δc ;
Dt ¼ (4)
1; if δ > δc

The same model for describing shear stress τ s and shear displacement ts relation is adopted for
concrete behaviour in shear, also shown in [37].
This paper presents the formulation for the behaviour in a concrete joint element, adjusted to capture
the main mechanisms related to cyclic loading in tension. For this purpose the material model shown in
Figure 2 is adopted according to Reinhardt [44] where the ratio of δ1/δmax was obtained experimentally
from uniaxial cyclic tests being equal to 0.73.

2.2. Reinforcement modelling


Behaviour of the reinforcement is modelled separately inside finite elements and joint reinforcement
elements. Namely, intersection between the sides of triangular concrete finite elements and
reinforcing bars gives the reinforcement finite elements and reinforcement joint elements (Figure 1).
The stress–strain relationship in the finite element of the reinforcing bar is linear elastic which means
that the concrete tensile strength will be exceeded before attaining the yield stress of steel. The
influence of the redistribution of the loads along the bar is not directly considered in the finite
element; however, bonding between the reinforcing bar and concrete was taken into account by
experimental curves which define relation between local slip and deformation of the bar in crack
[45] as it will be shown in this section.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
164 Ž. NIKOLIĆ ET AL.

Figure 3. Reinforcement finite element in initial and current configuration.

Strain and stress in reinforcing bars are calculated considering current lc and initial li length of the
reinforcing bar finite element (Figure 3). From the determined stress σ se, force in the reinforcing bar
is obtained according to

f 0se ¼ f 1se ¼ σ se As (5)

where As is cross-sectional area of the reinforcing bar. Forces f0se and f1se acting in points P0 and P1 are
distributed into the nodes of the parent concrete triangular element in the form of equivalent nodal
forces.
The model of the reinforcing bar in the joint elements produces two different regimes as well, before
and after the occurrence of cracks in concrete.
Prior to opening of cracks in concrete, the continuity between the reinforcing bar finite elements is
ensured through the penalty function method, such that the proportion of stress in the joint element of
the reinforcing bar σ sj and the joint element of concrete σ cj in the direction of the bar is the same as the
proportion of stress in the finite element of the reinforcing bar σ se and the finite element of concrete σ ce.
The model of the reinforcing bar in the joint element, with crack propagation and corresponding
post-peak softening behaviour in concrete, is based on a path-dependent mechanical model for a
deformed reinforcing bar at the RC interface developed by Soltani and Maekawa [45]. The influence
of the redistribution of the loads along the bar to the local slip and deformation of the bar in crack is
considered through the experimental curve [46].
An approximation of the experimental crack opening-strain curves is used to describe the behaviour
of the reinforcing bar at cracks faces. The presented model considers the bond deterioration in the
reinforcement near the crack plane, providing accurate expression of the behaviour of a reinforcing
bar that undergoes a high plastic deformation under reversed cyclic loading. After the crack
opening, the axial tension force developed in the reinforcing bar is further partially transferred to the
concrete between adjacent cracks through bonding between the reinforcing bar and concrete.
Consequently, the local stress along the bar differs from that at the interface, reflecting no uniform
distribution of strains along the bar, which among other factors depends on the bar pull out S from
the crack interface (Figure 4) [45].

Figure 4. Discrete crack and steel strain-slip relation under monotonic loading.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 165

The monotonic slip–strain relations are defined according to the non-dimensional slip s given by
   2=3
S fc
s¼ Kf c; Kf c ¼ (6)
D 20

where D is the diameter of the bar and fc is the compressive strength of concrete (MPa).
Non-dimensional slip–strain relationship before and after yielding of steel is used according to the
expressions shown in Figure 4 or in [37], where εs represents the strain at the reinforcing bar in the
crack, εy is the yielding strain of the bar, fu and fy are the tensile strength and yield stress of steel
(MPa), respectively, while εsh is the strain at the onset of hardening.
After yielding of the reinforcing bar, the normalized steel slip s is expressed as the sum of the slip spl
in the yield region and se in the elastic region [45] as follows:

s ¼ spl þ se (7)

Calculation of the slip in yield region is based on several assumptions: (1) linear distribution of the
strain in yield region; (2) the steel yielding zone ly expands only when new loading cycle begins; (3)
the steel yielding zone ly does not change during unloading and reloading [45]. The normalized steel
slip spl, based on previously mentioned assumptions, is expressed as follows:

ð1 þ βÞεs þ εsh  βεmax  


spl ¼ smax  sy (8)
εmax þ εsh
where εmax and smax represent steel strain and non-dimensional slip immediately after the transition
from loading to unloading, β is a factor obtained from experiments and is taken as 1.0. By
substituting Eqn (8) in Eqn (7), the strain in the reinforcing bar at the crack can be obtained from
the known non-dimensional slip s (Figure 5).
The calculation of the shear force carried by the bar is based on experimental curves [46] which
describe the curvature of the reinforcing bar in the vicinity of crack faces as a function of deflection
of the reinforcing bar at the interface ts. The shear force is thus given by

34:9091 t s
V s ¼ Es I s (9)
Lc 3
where Es is the Young modulus of the bar, Is is the moment of inertia of the bar and Lc is the length of
the curvature-influencing zone [37].
The influence of adjacent cracks is approximately taken into account through a reduction factor α
[45] which depends on distance lcr (Figure 6). The steel slip scr, which considers the influence of
adjacent cracks, is expressed for monotonic loading as follows:

scr ¼ α s (10)
where s is the non-dimensional slip defined by expression (6), while the reduction factor α is given by

Figure 5. Distribution of strain along the reinforcing bar after yielding.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
166 Ž. NIKOLIĆ ET AL.

Figure 6. Relation of the reduction factor α and crack spacing lcr in tension. RC, reinforced concrete.

3
α ¼ 1  eð0:065lcr =Dþ0:5Þ ; α ≤ 0:087 l cr =D (11)
The position of the crack is defined by a finite element edge, so the lcr is adopted as an input
parameter, which is equal to h/2 where h is the concrete finite element length.
2.2.1. Steel material model. The stress–strain relationship for a monotonically increasing load for
steel is shown in Figure 7a. The hysteresis behaviour of a steel bar is enforced through Kato’s
stress–strain model [47] shown in Figure 7b. From the given strain, stress is calculated by the
following expressions:
• during unloading (Figure 7b, curve (1))

σ s ¼ f y  E s ðεsh  εs Þ (12)

• during negative loading (Figure 7b, curve (2))


" ( ! )
σ s ¼ f y a  faða  1Þg = 
EB 
fy

εs  εsh þ εy þ a  1  (13)

where EB =  (Es/6)log 10(εsh  εy), a = Es/(Es  EB);


• during reloading-unloading (Figure 7b, curve (3))
 
σ s ¼ σ pm þ E s εs  εpm (14)

where σ pm is the minimum value of σ s in its loading history;

Figure 7. Stress-strain model of steel: (a) monotonic loading; (b) cyclic loading [47].

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 167

• during reloading (Figure 7b, curve (4))


" ( ! )
σ s ¼ f y þ σ pm þ f y a  faða  1Þg= 
EB 
fy

εy  εs þ εpm þ a  1  (15)

The flow chart which explains the presented numerical procedure is shown in Figure 8.

3. VALIDATION OF THE MODEL UNDER CYCLIC LOAD

The examples of a RC beam and RC wall under cyclic load have been used to validate the ability of the
presented model against experimental results, incorporating all the previously mentioned effects.

3.1. Reinforced concrete beam under cyclic load


An example of a RC beam under cyclic load with the known experimental results [45] was used to
validate the model under large inelastic cyclic load conditions, specifically stress–strain relation in
the reinforcing bars at the crack interface. The geometry of the beam and reinforcing bars are shown in

Figure 8. Flow chart of the overall numerical procedure.

Figure 9. Reinforced concrete beam.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
168 Ž. NIKOLIĆ ET AL.

Figure 9. Tensile external forces were applied according to the horizontal cyclic loading programme
shown in Figure 10.
Material characteristics are shown in Table I.
The crack opening in the experiment [45] was pre-determined by making notches at 300 mm
intervals. Numerical analysis was performed for two discretizations (Figure 11). In the coarse mesh
(Figure 11a), the cracks were enforced through the mesh, where the only possible cracks patterns
correspond to the experimental crack patterns. In the fine mesh (Figure 11a), the initiation of the
first cracks was enforced by reduced tensile strength in the joint elements at the position of the
notches. The crack patterns obtained numerically correspond to the experimental ones in the both cases
(Figure 12).
Figure 13 shows force-average axial strain relation under the cyclic loading of the beam obtained by
experiment and by presented model (finite discrete element method) for coarse and fine mesh. These
results reveal that force-average axial strain ε curve follows the experimental curve very accurately
during the cyclic loading for both discretizations, which is very important to describe the loss of
energy in the structure under cyclic load. In addition, the force-average axial strain relation under
the cyclic loading provides mesh-independent response.

3.2. Reinforced concrete wall under cyclic load


An example of a RC wall under cyclic load with the experimental results [48] was used for validation
of the model. The geometry and reinforcement of the wall are shown in Figure 14a. Flexural

Figure 10. Cyclic loading history.

Table I. Material characteristics of the beam.


Steel
Concrete
Young’s
Young’s modulus Compressive modulus Es/ Yield stress Ultimate Strain at the onset Reinforcement
Ec/MPa strength fc/MPa MPa fy/MPa stress fu/MPa of hardening εsh diameter D/m

29000 29 190000 350 540 0.0165 0.019

Figure 11. Finite element mesh: (a) coarse mesh; (b) fine mesh

Figure 12. Crack pattern for coarse and fine mesh.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 169

Figure 13. Force – average axial strain relation. FDEM, finite discrete element method.

Figure 14. Wall characteristics: (a) Geometry and reinforcement; (b) Mesh discretization.

reinforcement ratio in boundary elements is 3,3%. Ratio of vertical and horizontal reinforcement are
1.5% and 0.75%, respectively.
The concrete compressive strength was 35.2 MN/m2. The material characteristics of the reinforcing
steel are shown in Table II.
The wall was discretized into 986 triangular elements. The vertical reinforcing bars were modelled
with 56 two-node elements and the horizontal reinforcing bars with 28 two-node elements. Finite
element mesh is shown in Figure 14b.

Table II. Material characteristics of the reinforcing steel.


Bar type Yield stress fy/MPa Ultimate stress fu/MPa Bar diameter D/mm

Vertical 470 565 8.0


Horizontal 520 610 6.0
Stirrups 420 490 4.0

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
170 Ž. NIKOLIĆ ET AL.

The wall was subjected to the horizontal cyclic loading programme shown in Figure 15. After the
fourth cycle, a monotonically increasing load up to failure of the structure was applied.
The load-displacement curves of the experimental and numerical results are compared in Figure 16.
The envelope of the load-displacement curve and failure load show a very good agreement of the
results obtained by finite discrete element method with experimental ones. The failure load obtained
by this model was 103 kN, while the failure load in the experiment was 115 kN. The numerical
prediction of the shape of hysteresis loops shows slightly different behaviour. Namely, in this
model, the response of the concrete under cyclic loading is described by linear unloading and
reloading curves, while in reality some energy dissipation exists in reloading cycles. This is a reason
for differences in cyclic hysteretic response in relation to the experiment.

Figure 15. Cyclic loading history.

Figure 16. Horizontal load – displacement curve of the wall. FDEM, finite discrete element method.

Figure 17. Crack pattern for the failure: (a) numerical; (b) experimental.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 171

Figure 17 shows the crack patterns for the failure obtained numerically and experimentally. The
numerical prediction of the crack patterns reasonably agree with the experimental ones. Certain
differences between the numerical and experimental crack patterns occur because of the chosen
mesh discretization. Flexural cracks appeared near the bottom part of tensile zone of the wall. When
the horizontal load achieved about 60% of the failure load diagonal cracks appears. The collapse of
the wall occurred because of the failure of compressive zone accompanied with diagonal cracking.

4. SEISMIC ANALYSIS OF REINFORCED CONCRETE WALL STRUCTURES

The application of the presented numerical model for simulation of the dynamic response of the
structure was performed on a five-storey RC building with uncoupled wall system, which is shown in
Figure 18. The vertical load of the building consists of the own weight of structural elements, an
additional dead load of 2.5 kN/m2 and impose load of 4.0 kN/m2 at floor slabs.
The building was previously designed according to the prescriptions of EC8 [1] for importance
factor II (γI = 1), type 1 response spectrum, damping ξ = 5%, ground type B, design ground
acceleration ag = 0.3 g and ductility classes DCM and DCH. The behaviour factors equal to q = 3,0
for DCM and q = 4,4 for DCH are adopted.
Material characteristics of concrete and steel used in a numerical analysis are shown in Table III.
The application of the presented model in incremental dynamic analysis [49] of RC structure was
performed for the left boundary wall with geometry, discretization and reinforcement shown in
Figure 19.
Seismic loading is represented by a time function of the horizontal ground acceleration recorded on
the soil class B during real earthquakes. The set of seven ground motion records are chosen from the
European Strong-Motion Database [50] according to [51]. Selected earthquakes are listed in Table IV.
Figure 20 shows the acceleration response spectra of applied earthquakes together with the elastic
response spectrum for peak ground acceleration of 0.3 g and soil class B with a 5% viscous damping
ratio [1] and its 90% value.

Figure 18. Geometry of RC building with uncoupled wall system.

Table III. Material characteristics of the wall.


Concrete

Young’s Poisson ratio Tensile strength Compressive Fracture energy Spatial weight3
modulus strength
Ec/MPa ν ft/MPa fc/MPa Gf/N/m ρ/kN/m3
32800 0.2 3.80 38 150 25
Steel

Young’s Yield stress Ultimate stress Strain at the onset Ultimate strain
modulus of hardening
Es/MPa fy/MPa fu/MPa εsh εu
210000 500 600 0.02 0.1

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
172 Ž. NIKOLIĆ ET AL.

Figure 19. Boundary RC wall: (a) geometry, (b) discretisation, (c) reinforcement for DCM, (d) reinforce-
ment for DCH.

Table IV. Applied ground motion records [50].


Event name Country Station name Code Date amax (g)

South Iceland Iceland Selsund 004677xa 17/06/2000 0.33


Montenegro Montenegro Bar-Skupstina Opstine 000199ya 15/04/1979 0.44
Montenegro (aftershock) Montenegro Petrovac-Hotel Rivijera 000229ya 24/05/1979 0.33
Montenegro (aftershock) Montenegro Budva-PTT 000230ya 24/05/1979 0.32
Erzincan Turkey Erzincan-Meteorologij Mudurlugu 000535ya 13/03/1992 0.61
South Iceland Iceland Hella 004673ya 17/06/2000 0.57
South Iceland (aftershock) Iceland Kaldarholt 006328ya 21/06/2000 0.47

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 173

Figure 20. The acceleration response spectrum of seven real earthquakes together with the response spec-
trum for peak ground acceleration of 0.3 g and soil class B with a 5% viscous damping ratio [1] and its
90% value.

The average values of the spectra are also shown in Figure 20. The average elastic spectrum of seven
selected earthquakes should not drop below the value of 90% of elastic spectrum prescribed by EC8 in
any period of the structure [1].
The observed boundary RC wall is exposed to horizontal ground acceleration of selected seven
earthquakes. The amplitudes were gradually increased until the collapse of the wall. The relation
between the ratio of peak ground acceleration and gravity constant (a/g) and the ratio of maximum
top displacement u and the building height H (u/H) obtained by incremental dynamic analysis for
both ductility classes is shown in Figures 21 and 22.
Analysis of the average dynamic response of the wall for a series of seven records of real
earthquakes (Figures 21 and 22) shows that the behaviour of the wall designed for DCM class is
linear up to the ground acceleration a = 0,30 g. Significant non-linearity starts for a = 0,48 g, while
the collapse of the wall happens for a = 0,70 g. Wall designed according to DCH is in linear elastic
region up to the a = 0,40 g, followed by the occurrence of nonlinear behaviour. Average collapse
acceleration a = 0,84 g was observed for DCH class.
The most destructive earthquake for presented wall and both ductility classes is Bar, with ultimate
collapse accelerations a = 0,40 g and a = 0,50 g for DCM and DCH classes, respectively. The least
destructive earthquakes, considered through the aspect of collapse accelerations, are Petrovac and
Selsund, which produced the collapse of the wall for a = 0,90 g (DCM) and a = 1,0 g (DCH).
It can be observed that the wall reinforced according to the DCH possess average seismic resistance
20% higher with respect to DCM. It is interesting to emphasize that the longitudinal flexural
reinforcement in the boundary elements is same for both ductility classes in order to satisfy
condition of the minimum percentage of reinforcement. The differences pertain to confining

Figure 21. The ratio between the maximum roof displacement u and the wall height H for RC wall rein-
forced according to the DCM in dependence to the peak ground acceleration.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
174 Ž. NIKOLIĆ ET AL.

Figure 22. The ratio between the maximum roof displacement u and the wall height H for RC wall rein-
forced according to the DCH in dependence to the peak ground acceleration.

reinforcement in boundary elements and web shear reinforcement causing higher seismic resistance of
the high ductility wall. The average responses for both ductility classes still remain linear before they
reached design ground acceleration.
Figures 23 and 24 show crack pattern of the wall, reinforced corresponding to DCM and DCH
classes, exposed to earthquake excitation of Montenegro (aftershock) – Budva, until the collapse.
The reason for choosing the Budva earthquake is in a fact that this earthquake has response which is
near to the average value for considering seven earthquakes. The collapse of the structure for DCM
and DCH is achieved reaching accelerations of a = 0,60 g and a = 0,80 g.
The first cracks for DCM occur for acceleration of a = 0,30 g in the bottom part of the wall. The
reason for opening of the cracks here is relatively low percent of flexural reinforcement in the
boundary element (0,5% of the boundary element area) leading to the increase of the stress in
reinforcement and crack initiation in the concrete. With additional increase of acceleration, initial
horizontal cracks are elongated and connected with each another for a = 0,50 g. In this moment the
horizontal crack at the wall base is completely opened. Furthermore, the stress in the reinforcing
bars continues to propagate followed by the additional crack growth. Thus, the main cause of the
wall collapse is the failure of flexural reinforcement in the base of the wall, followed by the yielding
of the web vertical reinforcement. Figure 23d shows post-failure behaviour of the wall (a = 0,60 g).
The wall reinforced according to the DCH has also developed two cracks at the bottom part of the
wall for the same acceleration as in DCM wall (a = 0,30 g). The similar behaviour can be observed for
DCH wall with the difference in coalescence of initial horizontal cracks in the bottom part of the wall

Figure 23. Crack pattern for the wall reinforced according to the DCM, earthquake Montenegro (aftershock)
– Budva: (a) a = 0.3 g; (b) a = 0.4 g; (c) a = 0.5 g; (d) a = 0.6 g.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 175

Figure 24. Crack pattern for the wall reinforced according to the DCH, earthquake Montenegro (aftershock)
– Budva: (a) a = 0.3 g; (b) a = 0.4 g; (c) a = 0.5 g; (d) a = 0.6 g; (e) a = 0.7 g; (f) a = 0.8 g.

when ground acceleration reaches a = 0,70 g. The failure of the structure occurred again because of the
fracture of flexural reinforcement accompanied by pronounced crack in the base of the wall. Figure 24f
presents post-failure behaviour of the wall for acceleration a = 0,80 g characterized with rotating of the
part of the wall above the crack and significant cracking. Despite the similar failure pattern, smaller
number of larger cracks occurred in DCM wall, while finer crack pattern, followed by concrete
crushing, came as a result of greater amount of confining and shears web reinforcement in DCH
wall. As expected, DCH shows higher resistance to large ground accelerations. This is due to
vertical web reinforcement contribution to the flexural resistance of the wall, while the smaller
distance between the hops in confining elements ensures slower development of the cracks and
higher failure load.
Figure 25 shows the stress in the vertical edge reinforcement bar (φ14). The stress is monitored in
the joint element of the vertical bar, in the crack which occurred at the bottom of the wall.
It can be seen that the stress in the vertical edge reinforcement bar is almost the same up to
acceleration a = 0,30 g. Following that, the stress in the wall increases for both ductility classes.
Ultimate steel stress for DCM and DCH is reached for the collapse acceleration of 0,6 g and 0,8 g,
respectively.
Figure 26 shows the ratio of the maximum interstorey drift Δu and storey height h (Δu/h), as a
function of the ground acceleration.
This earthquake causes small value of interstorey drifts (less than 0.2%) up to the value of
approximately 80% of ultimate acceleration. All interstorey drifts suddenly increase after the failure
of flexural reinforcement. Maximum interstorey drift obtained for DCM wall is 1.33%.

Figure 25. Stress in the vertical edge reinforcement bar (φ14) for the wall reinforced according to the DCM
and DCH.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
176 Ž. NIKOLIĆ ET AL.

Figure 26. Maximum interstorey drifts for all storeys of the wall reinforced according to the DCM and DCH,
exposed to peak ground acceleration of earthquake Montenegro (aftershock) – Budva.

5. CONCLUSIONS

Crack opening is one of the dominant causes of nonlinearity in RC structures which leads to localized
failure and stands out as a serious challenge in numerical modelling. This becomes even greater
challenge when it comes to dynamic effects induced by the earthquakes, leading to cyclic crack
opening-closing mechanisms. Fracturing, contact detection and contact interaction between the parts
of the concrete separated by the cracks, cyclic behaviour of the material and interaction between
concrete and steel play a key role in modelling of energy dissipation mechanisms of RC structures
subjected to earthquake. Interaction between concrete and steel is represented by an embedded
reinforcing bars implemented in concrete elements, accompanied with steel strain-slip relation, the
influence of adjacent cracks to the slip of reinforcing bar, local slip of reinforcing bar near the crack
plane when the bar undergoes a high plastic deformation under reversed cyclic loading and the
influence of the curvature of reinforcing bar to yield stress reduction of the steel. All these effects
are successfully provided in the numerical model based on finite discrete element method, presented
in this paper.
The performance of the model was demonstrated by incremental dynamic analysis of the RC wall,
previously designed according to the prescriptions of EC8 for DCM and DCH. Presented results
demonstrated the ability of the model for simulation of the behaviour of earthquake excitated
structure until the collapse. In this example, the higher values of ultimate acceleration were obtained
for the DCH wall as a consequence of reinforcing the wall with greater amount of confining and
shear web reinforcement.
The presented model has a great potential for estimation of failure mechanism and collapse load of
both new and existing RC structures.

ACKNOWLEDGEMENTS
This work has been fully supported by Croatian Science Foundation under the project Development of nu-
merical models for reinforced-concrete and stone masonry structures under seismic loading based on dis-
crete cracks (IP-2014-09-2319).

REFERENCES
1. Eurocode 8: Design of structures for Earthquake resistance – part 1: general rules, seismic actions and rules for build-
ings. EN 1998-1, CEN, Brussels, 2005.
2. Fajfar P, Gašperšič P. The N2 method for the seismic damage analysis of RC buildings. Earthquake Engineering and
Structural Dynamics 1996; 25(1):11–46.
3. Fajfar P, Eeri M. A nonlinear analysis method for performance based seismic design. Earthquake Spectra 2000; 16
(3):573–592.
4. Kreslin M, Fajfar P. The extended N2 method considering higher mode effects in both plan and elevation. Bulletin of
Earthquake Engineering 2012; 10(2):695–715.
5. Kotronis P, Mazars J, Nguyen XH, Ile N, Reynouard JM, Bish P, Coin A. The seismic behaviour of reinforced con-
crete structural walls: experiments and modelling. In Geotechnical, Geological and Earthquake Engineering. Ansal
A. (ed.).Springer: Netherlands, 2004; 7: 363–376.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
NUMERICAL MODELLING OF RC STRUCTURES BY FINITE-DISCRETE ELEMENT METHOD 177

6. Xu XP, Needleman A. Numerical simulation of fast crack growth in brittle solids. Journal of the Mechanics and
Physics of Solids 1994; 42:1397–1434.
7. Ortiz M, Pandolfi A. Finite-deformation irreversible cohesive elements for three-dimensional crack-propagation anal-
ysis. International Journal for Numerical Methods in Engineering 1999; 44:1267–1282.
8. Ortiz M, Quigley JJ. Adaptive mesh refinement in strain localization problem. Computer Methods in Applied
Mechanics and Engineering 1991; 90:781–804.
9. Simo JC, Oliver J, Armero F. An analysis of strong discontinuities induced by strain-softening in rate-independent
inelastic solids. Computational Mechanics 1993; 12:277–296.
10. Armero F, Garikipati K. An analysis of strong discontinuities in multiplicative finite strain plasticity and their relation
with the numerical simulation of strain localization in solids. International Journal of Solids and Structures 1996;
33:2863–2885.
11. Moës N, Dolbow J, Belytschko T. A finite element method for crack growth without remeshing. International
Journal for Numerical Methods in Engineering 1999; 46:131–150.
12. Wells GN, Sluys LJ, Borst R. Simulating the propagation of displacement discontinuities in a regularized strain-
softening medium. International Journal for Numerical Methods in Engineering 2002; 53:1235–1256.
13. Ibrahimbegovic A, Boulkertous A, Davenne L, Brancherie D. Modelling of reinforced-concrete structures providing
crack-spacing based on X-FEM, ED-FEM and novel operator split solution procedure. International Journal for
Numerical Methods in Engineering 2010; 83:452–481.
14. Armero F, Linder C. Numerical simulation of dynamic fracture using finite elements with embedded discontinuities.
International Journal of Fracture 2009; 160:119–141.
15. Rethore J, Gravouil A, Combescure A. An energy-conserving scheme for dynamic crack growth using the eXtended
finite element method. International Journal for Numerical Methods in Engineering 2005; 63(5):631–659.
16. Bažant ZP, Caner F, Adley M, Akers S. Fracturing rate effect and creep in microplane model for dynamics. Journal
of Engineering Mechanics 2000; 126(9):962–970.
17. Ožbolt J, Li Y, Kožar I. Microplane model for concrete with relaxed kinematic constraint. International Journal of
Solids and Structures 2001; 38(16):2683–2711.
18. Ožbolt J, Bošnjak J, Sola E. Dynamic fracture of concrete compact tension specimen: Experimental and numerical
study. International Journal of Solids and Structures 2013; 50(25-26):4270–4278.
19. Ožbolt J, Sharma A, Irhan B, Sola E. Tensile behaviour of concrete under high loading rates. International Journal
for Impact Engineering 2014; 69(1):55–68.
20. Mitrović S, Ožbolt J, Travaš L. Three-dimensional finite element formulation for nonlinear dynamic analysis of
seismic site and structure response. European Journal of Environmental and Civil Engineering 2014; 19
(7):789–804.
21. D’Addetta GA, Kuhl E, Kun F, Ramm E. Micromechanical modelling of concrete cracking. Proceedings of
European Conference on Computational Mechanics, Munich, Germany, 1999, paper No. 307.
22. Cusatis G, Nakamura H. Discrete modelling of concrete materials and structures. Cement and Concrete Composites
2011; 33(9):865–866.
23. Schlangen E, Garboczi EJ. Fracture simulation of concrete using lattice models: computational aspects. Engineering
Fracture Mechanics 1997; 57(2/3):319–332.
24. Cusatis G, Bažant ZP, Cedolin L. Confinement-shear lattice model for concrete damage in tension and compression: I
Theory. Journal Engineering Mechanics 2003; 129(12):1439–1448.
25. Lilliu G, van Mier JGM. 3D lattice type fracture model for concrete. Engineering Fracture Mechanics 2003;
70:927–941.
26. Cusatis G, Bažant ZP, Cedolin L. Confinement-shear lattice CSL model for fracture propagation in concrete.
Computer Methods in Applied Mechanics and Engineering 2006; 195(52):7154–7171.
27. Benkemoun N, Hautefeuille M, Colliat JB, Ibrahimbegovic A. Failure of heterogeneous materials: 3D meso-scale FE
models with embedded discontinuities. International Journal for Numerical Methods in Engineering 2010; 82
(13):1671–1688.
28. Nikolic M, Ibrahimbegovic A, Miscevic P. Brittle and ductile failure of rocks: Embedded discontinuity approach for
representing mode I and mode II failure mechanisms. International Journal for Numerical Methods in Engineering
2015; 102(8):1507–1526.
29. Nikolic M, Ibrahimbegovic A. Rock mechanics model capable of representing initial heterogeneities and full set of
3D failure mechanisms. Computer Methods in Applied Mechanics and Engineering 2015; 290:209–229.
30. Gedik YH, Nakamura H, Yamamoto Y, Kunieda M. Evaluating of three-dimensional effect in short beams using a
rigid-body-spring model. Cement & Concrete Composites 2011; 33:978–991.
31. Yamamoto Y, Nakamura H, Kuroda I, Furuya N. Crack propagation analysis of reinforced concrete wall under cyclic
loading using RBSM. European Journal of Environmental and Civil Engineering 2014; 18(7):780–792.
32. Ghaboussi J. Fully deformable discrete element analysis using a finite element approach. International Journal of
Computers and Geotechnics 1997; 5:175–195.
33. Pearce CJ, Thavalingam A, Liao Z, Bićanić N. Computational aspects of the discontinuous deformation analysis
framework for modelling concrete fracture. Engineering Fracture Mechanics 2000; 65:283–298.
34. Munjiza A, Owen DRJ, Bićanić N. A combined finite-discrete element method in transient dynamics of fracturing
solids. Engineering Computations 1995; 12:145–174.
35. Munjiza A. The combined finite-discrete element method. John Wiley & Sons: UK, 2004.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe
178 Ž. NIKOLIĆ ET AL.

36. Živaljić N. Finite-discrete element method for 2D seismic analysis of reinforced concrete structures. PhD thesis (in
Croatian), University of Split, Croatia, 2012.
37. Živaljić N, Smoljanović H, Nikolić Ž. A combined finite-discrete element model for RC structures under dynamic
loading. Engineering Computations 2013; 30(7):982–1010.
38. Živaljić N, Smoljanović H, Nikolić Ž. Sensitivity analysis of numerical parameters in FEM/DEM model for RC
structures. International Journal for Engineering Modelling 2012; 25(1-4):7–17.
39. Munjiza A, Andrews KRF, White JK. Penalty function method for combined finite-discrete element system compris-
ing large number of separate bodies. International Journal for Numerical Methods in Engineering 2000;
49:1377–1396.
40. Xiang J, Munjiza A, Latham JP, Guises R. On the validation of DEM and FEM/DEM models in 2D and 3D. Engi-
neering Computations 2009; 26:673–687.
41. Munjiza A, Andrews KRF, White JK. Combined single and smeared crack model in combined finite-discrete element
method. International Journal for Numerical Methods in Engineering 1999; 44:41–57.
42. Munjiza A, John NWM. Mesh size sensitivity of the combined FEM/DEM fracture and fragmentation algorithms.
Engineering Fracture Mechanics 2002; 69:281–295.
43. Hordijk DA. Tensile and tensile fatigue behaviour of concrete – experiments, modelling and analyses. Heron 1992;
37(1):3–79.
44. Reinhardt HV. Fracture mechanics of an elastic softening material like concrete. Heron 1984; 29(2):3–41.
45. Soltani M, Maekawa K. Path-dependent mechanical model for deformed reinforcing bars at RC interface under
coupled cyclic shear and pullout tension. Engineering Structures 2008; 30:1079–1091.
46. Qureshi J, Maekawa K. Computational model for steel embedded in concrete under combined axial pullout and trans-
verse shear displacement. Proceedings of the JCI 1993; 15(2):1249–1254.
47. Kato B. Mechanical properties of steel under load cycles idealizing seismic action. AICAP-CEB Symposium:
Structural concrete under seismic action, CEB 1979; 131:7–27.
48. Lefas ID, Kotsovos MD. Strength and deformation characteristics of reinforced concrete walls under load reversals.
ACI Structural Journal 1990; 87(6):716–726.
49. Vamvatsikos D, Cornell CA. Incremental Dynamic Analysis. Earthquake Engineering and Structural Dynamics
2002; 31(3):491–514.
50. European Strong-Motion Database, 2015. http://www.isesd.hi.is/ESD_Local/frameset.htm [Accessed 12 January
2015].
51. Iervolino I, Maddaloni G, Cosenza E. Eurocode 8 compliant real record sets for seismic analysis of structures.
Journal of Earthquake Engineering 2008; 12:54–90.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2017; 46:159–178
DOI: 10.1002/eqe

Potrebbero piacerti anche