You are on page 1of 19

Engineering Structures 212 (2020) 110514

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Seismic analysis of older and contemporary reinforced concrete frames with T


the improved fish-bone model
Aleš Jamšek, Matjaž Dolšek

Faculty of Civil and Geodetic Engineering, University of Ljubljana, Jamova 2, Ljubljana, Slovenia

ARTICLE INFO ABSTRACT

Keywords: The improved fish-bone (IFB) model for seismic analysis of older and contemporary reinforced concrete frames is
Reinforced concrete frame proposed by introducing a new procedure for the estimation of parameters of structural elements of the fish-bone
Fish-bone model model. The procedure makes it possible to approximately account for the importance of structural elements on
MDOF model the seismic response of a frame building and the effect of potential redistribution of demands between structural
Pseudo-dynamic test
elements of the frame building. Firstly, the IFB model is described. Follows demonstration of its capability by
Seismic analysis
Pushover analysis
simulating the response history of storey shears, storey drifts and storey accelerations of selected pseudo-dy-
Fragility analysis namically tested frames. Finally, it is shown that the IFB models can provide pushover curves, the corresponding
damage states and fragility functions based on response history analysis which are very similar to those obtained
by the conventional multi-degree-of-freedom models. The capability of the IFB model with respect to conven-
tional fish-bone model is also demonstrated by the pushover analysis. It can be observed that the IFB model
significantly improves the simulation of the older reinforced concrete frames, which do not fulfil the strong
column – weak beam concept, and the frames for which the columns or beams of one storey of a frame building
differ significantly from each other. Additional research is needed to extend the capabilities of simplified
structural models to plan irregular frame building.

1. Introduction parameters of reinforced concrete moment frames [11]. The energy-


consistent approach for reducing the number of degrees-of-freedom was
Seismic response of building structures is often evaluated by multi- used by Araki et al. [12] in order to extend the generic frame models for
degree-of-freedom (MDOF) models (Fig. 1a), where each column, beam estimating the response of tall steel frames with non-regular bay
or wall in a storey is modelled by one finite element. However, many lengths. Li and Kurata [13] presented a probabilistic method of up-
simplified MDOF models [1,2] and single-degree-of-freedom (SDOF) dating the fish-bone (FB) models for assessment of damage on the
models [3–6] have been developed. Simplified MDOF models can be beam-column connection of steel frames. Continuous models consist of
classified into generic frame models, and continuous models called stick a flexural and shear cantilever beam representing flexural and shear
models. Different variants of the simplified frame models: the generic deformations of multi-storey buildings that are connected with axially
frame (GF) (Fig. 1c) and fish-bone (FB) (Fig. 1b) models have been used rigid links [2,14,15].
to study the response of frame buildings. The generic frame (GF) pro- Simplified models are computationally efficient, but their capacity
posed by Nakashima et al. [1] was used for estimating the seismic de- for simulating phenomena observed during the seismic response of
mand of steel moment-resisting frames [7], while the initial idea of the buildings is not yet well understood. The reduced degrees of freedom
fish-bone model was presented by Ogawa et al. [8]. Later on, Khaloo represent the basic uncertainty in the case of simplified models, while
and Khosravi [9] modified the fish-bone model to more accurately si- the estimation of the input parameters of simplified models represents
mulate storey drift demands of analysed moment-resisting frames [9]. the second-level of uncertainty. In the case of the SDOF models, which
The modified fish-bone (MFB) model [9] was used as a reference model are the most simplified models, the direct estimation of the input
to study inter-storey drift demand from low-rise to high-rise moment parameters is particularly challenging. The parameters of the SDOF
frames [10]. Further on, the MFB model was adapted to extend its models are, therefore, often evaluated via pushover analysis by utilising
capability for the estimation of several different engineering demand detailed building models. Such an approach provides reliable estimates

Abbreviations: FB, fish-bone; IFB, improved fish-bone; GF, generic frame; PsD, pseudo-dynamic

Corresponding author.
E-mail address: mdolsek@fgg.uni-lj.si (M. Dolšek).

https://doi.org/10.1016/j.engstruct.2020.110514
Received 27 September 2019; Received in revised form 3 March 2020; Accepted 9 March 2020
Available online 25 March 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

results of seismic analysis obtained the fish-bone model are not sig-
nificantly affected if frames have more than 1-bay, and their height is
less than three times its length. All the frames analysed in this study
fulfil these conditions.
The FB model was modified in order to eliminate some short-
comings of the initial FB model [9]. The authors performed extensive
studies investigating moment resisting frames with an emphasis on the
consideration of the effect of column elongation and contraction in
order to take into account overall flexural deformation of the structure.
Based on the analysis of cyclic lateral loading Khaloo and Khosravi [9]
proposed a bi-linear moment-rotation relationship for beams rather
than using the tri-linear relationship proposed by Nakashima et al. [1].
This modification was made because it was found that the simplified
Fig. 1. Example of 2D (a) MDOF model of a frame, (b) equivalent fish-bone (FB) model is more appropriate for seismic analysis [9]. The assumption of
model, (c) equivalent generic-frame (GF) model. equal stiffness of joints in a given storey was considered by Nakashima
et al. [1], while Khaloo and Khosravi [9] proposed to account for re-
duced rotational stiffness of exterior joints compared to that of interior
of input parameters of the SDOF models, although the use of the SDOF
joints. Therefore, the equivalent rotational stiffness of the modified fish-
models is still limited because the effect of higher failure modes is
bone model is slightly lower than that of the conventional fish-bone
disregarded [16]. This, however, can be approximately taken into ac-
model [9]. For analysis of reinforced concrete frame modified fish-bone
count by different approaches (e.g. [2,17,18]).
model was further adapted with consideration of deteriorating peak-
The detailed MDOF models can provide a reliable estimation of
oriented hysteretic behaviour of the moment-rotation relationship of
seismic demands in structural elements of building structures, but there
the plastic hinges and taking into account the migration of the point of
are several potential applications where analyst still cannot afford to
contra-flexure from the mid-span due to the considered peak-oriented
use the detailed MDOF models. For example, for the seismic risk as-
hysteretic behaviour of reinforced concrete beams [11].
sessment of a building portfolio, the analyst has to simplify building
models of building portfolio either due to the lack of data or limited
2.2. Description of the IFB model
resources available for seismic risk estimation [14,15].
The paper first introduces the improved fish-bone (IFB) model as a
2.2.1. Required information about the building, input data for the definition
variant of the fish-bone (FB) model (Fig. 1b) [8]. The configuration of
of the IFB model and the basic assumptions of the IFB model
elements of the presented IFB model is equivalent to that of the FB
Although the IFB model is a simplified model, it is preferred that the
model. Novelty is presented by the procedure for the estimation of the
information about the building is the same as that required for the
parameters of the beams and columns of the IFB model, which is
definition of the MDOF model. In the optimal case, the required in-
especially important if beams are stronger than columns and if columns
formation about the building is the building geometry in plan and
or beams of one storey of a frame building differ significantly from each
elevation, reinforcement drawings, material characteristics of designed
other. In the second part of the paper, the capabilities of the IFB model
or built-in material, the storey masses and the gravity loads. The in-
are demonstrated by simulating the seismic response of three buildings,
formation about the building can then be used to prepare the input data
which were pseudo-dynamically tested in full scale. For these buildings
for the definition of the IFB model. Some steps are the same as in the
and three additional buildings, the capabilities of the IFB model for
process of preparation of input data of the MDOF model. The analyst
performing pushover and fragility analysis are then discussed.
has to define the geometrical constants of the structural elements (e.g.
length L , cross-section area A , the moment of inertia I ), gravity loads
2. Overview of simplified frame models and description of and calculate the corresponding loads on the beams and columns. Then,
improved fish-bone model the moment-rotation relationship in plastic hinges of columns and
beams has to be calculated.
2.1. Overview of simplified frame models The above-described data refer to the investigated building, while
the following process is then related to the definition of the input data
Several simplified MDOF models of buildings have been developed. of the IFB model. Firstly, the geometrical constants of the columns and
The generic frame (GF) model [1] consists of one column and one beam beams of the IFB model, from now on called IFB columns and beams,
per each storey (Fig. 1c). The beams are condensed into one rotational are calculated from the geometrical constants of the columns and beams
spring with the assumption of the point of contra-flexure of all beams at of a frame building. Next, the moment-rotational relationships of the
the middle of the span. This implies that rotations at joints (beam-to- IFB columns and beams are calculated. The adopted hysteretic beha-
column connections) at a given storey are assumed identical. The viour in the plastic hinges of the IFB columns and beams is usually
masses are assumed lumped at floor levels, and material nonlinearity is similar to the one that would be considered in the MDOF model. The
modelled by plastic hinges at member ends. The P-delta effects can be same applies to the model of viscous damping.
considered as a first-order approximation where the storey shear re- When defining the IFB model of a building, there are two levels of
sistance is subtracted by the effect of the axial force and storey dis- assumptions. The first level of assumption is already used in the case of
placements [1]. The stiffness of the column or the beam of the simpli- the MDOF models, which are considered sufficiently accurate for the
fied MDOF model is usually defined by summing stiffnesses of columns simulation of the seismic response of a building in the range up to the
or the beams of the detailed MDOF model [1,8]. Fish-bone model (FB) near collapse (e.g. [19,20]). For MDOF models, it is usually assumed
[7,8] was used for the analysis of steel moment-resisting frame build- that the mass can be modelled by lumped storey mass in the centre of
ings. This model (Fig. 1b) is similar to the GF model, but with beams on gravity and that floor is rigid in its plane. The strain-stress relationship
both sides of the column. The beams of the fish-bone model are half the for the moment-curvature analysis of the cross-section of the columns
length of the beams from the frame and restrained at one end to prevent and beams can be calculated according to the code requirements (e.g.
vertical displacements (Fig. 1b) [8,9]. However, the use of the simpli- [21]). However, several assumptions are needed to evaluate the mo-
fied models is limited, but it was shown before [1] that by neglecting ment-rotation relationship in the plastic hinges of the columns and
the effect of the column and beam elongation and contraction, the beams. In this study, the moment-rotation relationship at the ends of

2
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

n
1 1
LbF, i = · Lb, i, k
2 n k=1 (2)

where Lb, i, k is the length of the k -th beam in the i -th storey of the frame
building. One half of the length is used because the IFB beams are re-
strained at the mid-span because the configuration of structural ele-
ments of the IFB model is the same as that of the FB model (Fig. 1b).
However, the variation of the length of the beams within the i -th storey
of the building is taken into account in the calculation of the moment of
inertia of IFB beamIbF, i , which is calculated as follows:
n
Fig. 2. Bi-linear moment-rotation relationship with a linear post-capping be- 2LbF, i
haviour and three characteristic points needed for its definition. IbF, i = Ib, i, k ·
k=1
Lb, i, k (3)

the columns and beams was determined according to the procedure where Ib, i, k is the moment of inertia of the k -th beam in the i-th storey of
used in previous studies [22,23]. In addition, it was assumed acceptable the frame building.
that the moment-rotation relationship in the plastic hinges is modelled The vertical loads are applied to the IFB model as point loads in
by a bi-linear relationship with consideration of linear softening branch beam-column joints and uniformly distributed loads on the IFB beams
in the post-capping range (see Fig. 2). The moment-rotation relation- in such a manner that the storey axial forces of the MDOF model are
ship is thus defined by three characteristic points. They correspond to equal to the axial forces in the IFB column. The point load on the i -th
yield (MY ), maximum moment (MM ) and ultimate moment (MU ), which IFB column is defined as the sum of point loads on the columns of the
are from hereinafter called characteristic moments, while the corre- i -th storey of a building. For the IFB beams, the uniformly distributed
sponding rotations are termed as characteristic rotation ( Y , M , U ) line loads qbF, i are defined with the consideration of the redistribution of
(Fig. 2). Note that for the calculation of yield and maximum moments, vertical loads due to the support on one side and the plastic hinges on
the axial forces from the gravity analysis were considered. It was as- the other side of the beams:
sumed that the variation of axial forces does not have any significant qi ·Ai
impact on the seismic response of columns. The axial force in the beams qbF, i = cqF
LbF, i (4)
was assumed zero. In the process of estimation of the characteristic
rotations, it is usually assumed that the zero moments (i.e. point of where qi and Ai are, respectively, the uniformly distributed area load
contra-flexure) in the columns and the beams are at the mid-span of the and the area of the i -th storey, LbF, i is the IFB beam length and cqF is the
elements [24]. Such an assumption was also made in this study. Ad- coefficient of load redistribution, which accounts for the appropriate
ditionally, the effective beam widths were assumed according to redistribution of the vertical load in the IFB beam due to the effect of
Eurocode 2 [21] and the effective stiffness of the columns and beams as the initial stiffness in the plastic hinge of the IFB beam. If the initial
50% of the initial stiffness. The shear behaviour and the non-linear rotational stiffness of the plastic hinge of the IFB beam is practically
behaviour of beam-column joints was neglected, but it could be ap- zero, which is the case of the pinned connection of the IFB beam to the
proximately taken into account with modification of the moment-ro- IFB column, then cqF should be equal to 1.0 in order to get the consistent
tation relationship at the ends of the columns and beams (e.g. [25]). redistribution of vertical reactions and the corresponding axial forces in
The second level of assumptions refers to the definition of the IFB the IFB column. In the opposite case, if the IFB beam is considered fixed
model. These assumptions are related to the procedures used to con- at the side connected to the IFB column, then the vertical reaction
dense the input data of the MDOF model to the input data of the IFB transmitted to the IFB column increases and cqF has to be reduced in
model. Assumptions at this level are related to procedures for the es- order to guarantee that the resulting axial force in the IFB column is
timation of the stiffness of the IFB columns and beams and the esti- consistent with the storey axial forces from the MDOF model. The ap-
propriate value of cqF in this case is be equal to 0.8, because the vertical
mation of the moment-rotation relationship in the plastic hinges of the
reaction at the fixed support of the IFB beam is 1.25 times the vertical
IFB columns and beams. These procedures and the corresponding as-
reaction resulting from the uniform distribution of load to both sup-
sumptions are described in the following section.
ports of the IFB beam. However, the appropriate value of the cqF for all
IFB models used in this study is between the presented bounding values,
because of the initial rotational flexibility of the plastic hinge in the IFB
2.2.2. Definition of the stiffness and the moment-rotation relationship of the beam. If the initial rotational stiffness of the plastic hinge of the IFB
IFB columns and beams beam is equal to the rotational stiffness of the elastic IFB beam fixed at
The stiffness of the IFB model is controlled by the length of the the side connected to the IFB column, then the vertical reaction at the
structural elements of the IFB models and the corresponding moment of side of the IFB beam with plastic hinge is equal to 9/16 of the entire
inertia. The length of the IFB column is assumed to be equal to the uniform load on the IFB beam. Consequently, the coefficient cqF should
storey height, while the corresponding moment of inertia IcF , where F be equal to 8/9 in order to simulate that 50% of the load is transferred
denotes IFB model, is equal to the sum of moments of inertia of columns to the IFB column. Note that the distributed line loads on IFB beams
in the corresponding storey of a frame building: were defined because the line loads were also considered on beams of
m MDOF model [23]. The bending moments in plastic hinges of IFB beams
IcF, i = Ic, i, j due to gravity loads were thus similar to those observed in the beams of
j =1 (1) the MDOF model.
Simplifications are also needed in the estimation of the properties
where Ic, i, j denotes a moment of inertia of the j -th column of a frame affecting the capabilities of the IFB models to simulate the nonlinear
building in the i -th storey. This approach is equal to that used in the response of buildings. For this study, the moment-rotation relationship
case of the GF model [1] or in the case of modified FB model [9]. The in plastic hinges of the IFB columns and beams was assumed bi-linear
cross-section area of beams and columns of the IFB model is determined with an additional linear post-capping behaviour as presented in Fig. 2.
by analogy to Eq. (1). However, since the length of the beams in a It is assumed that the p -th characteristic moment of the h -th plastic
storey of a frame building can vary significantly, the IFB beam length hinge of the i -th storey column McF, i, h, p is defined as the sum of the
LbF, i is defined by one half of the average length of the beams in a storey: characteristic moments of plastic hinges of the columns in a given

3
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Fig. 3. Presentation of indices for plastic hinges of columns and beams of (a) MDOF and (b) IFB model.

storey: characteristic moment at the h -th hinge of the IFB beam. The first
m condition of Eq. (6) is defined for the full contribution of the beam
McF, i, h, p = Mc, i, j, h, p strength, while the second condition of Eq. (6) is defined for the re-
j=1 (5) duced contribution of the beam strength to the characteristic moments
of the IFB beam. Once Mb, i, k, h, p has been estimated for all the beams of a
where Mc , i, j, h, p is the p -th characteristic moment of the h -th plastic hinge
frame building, the characteristic moments of the IFB beam MbF, i, h, p can
of the j -th column of the i -th storey of a frame building. For a more be defined (Eq. (7)) by the same principle as used in the case of the
straightforward presentation of indices, see Fig. 3. columns:
The estimation of the above-introduced parameters of the model is
n
straightforward, but special attention is given to the estimation of the MbF, i . h, p = Mb, i . k . h, p
moment-rotation relationship of the IFB beams in order to improve the k=1 (7)
simulation of seismic response of frames. Especially, when frames were
The same approach that is used for the estimation of the char-
not designed with consideration of the strong column – weak beam
concept (e.g. without consideration of capacity design according to acteristic moments of the IFB beam (Eq. (6)) is also used for the esti-
mation of the first characteristic rotation (yield rotation) of the IFB
Eurocode 8 provisions [26]), and for which the columns or beams of
beam in order to guarantee the appropriate initial stiffness of the plastic
one storey of a frame building differ significantly from each other. For
hinge of the IFB beam. If such modification is not taken into account,
example, if a strong beam is connected to a weak column, then it is not
the initial stiffness of the plastic hinge of the IFB beam is under-
likely that the moment demand in the beam will exceed yield moment
MY or maximum moment MM (Fig. 2), because yielding of the columns estimated.
The characteristic rotations, as well as the characteristic moments in
occurs prior to the yielding of the beams. Thus, the characteristic mo-
ments of the IFB beams cannot be estimated by analogy to Eq. (5), the plastic hinges of the beams and columns of a frame building, can
vary significantly from beam-to-beam and column-to-column in a se-
because in such a case, the potential for the demand in the IFB column
could be significantly overestimated, which could result in the over- lected storey. This is another issue, which should be at least approxi-
mately considered in the definition of the characteristic rotations of the
estimation of strength or underestimation of deformation capacity of
the IFB model. This issue can be solved approximately if the redis- IFB columns and beams. It is thus proposed that the characteristic ro-
tations of the plastic hinge of the IFB model are defined as the weighted
tribution of the demands in structural elements due to nonlinear be-
haviour of the frame building is approximately taken into account in average of the characteristic rotations of the corresponding plastic
hinges of the MDOF model. Thus, the formulation for the characteristic
the process of condensation of the properties of the frame building
rotations of the IFB columns cF, i, h, p and IFB beams bF, i, h, p is defined
beams to the properties of the IFB beams. It is thus proposed that the
according to Eq. (8) and Eq. (9).
characteristic moments of the IFB beams are defined on the basis of
m
maximum moment equilibrium of the beam-column joints (MM , Fig. 2). j=1
(Mc, i, j, h, p· c , i , j, h, p )
F
If the sum of the maximum moments of beams in an analysed joint is c, i, h, p = m
j =1
Mc, i, j, h, p (8)
greater than the sum of the maximum moments of the corresponding
columns Mc , i, M , the characteristic moments of the beams have to be n
(Mb, i, k, h, p·
k=1 b, i, k, h, p )
reduced in the estimation of the characteristic moments of the corre-
F
b, i, h, p = n
sponding IFB beams. It is assumed that the reduction factor is based on k=1
Mb, i, k, h, p (9)
the ratio of the sum of maximum moments of the columns to the sum of where c , i, j, h, p and b, i, k, h, p are the p -th characteristic rotations of the
maximum moments of the corresponding beams. In the opposite case, h -th hinge, respectively, of the j -th column and the k -th beam in the i -th
when the frame is designed with consideration of the strong column – storey of the analysed building. Mc , i, j, h, p and Mb, i, k, h, p are the corre-
weak beam concept, such a reduction is not necessary. These rules can sponding characteristic moments of the columns and beams, respec-
be formulated as follows: tively. These moments are used to define the weights assigned to the
characteristic rotation of the columns and beams.
Mb, i, k, h, p. .. Mb, i, M Mc , i, M
The weighted average for the definition of characteristic rotations of
Mb, i . k . h, p = Mc, i, M
Mb, i, k, h, p . .. Mb, i, M > Mc, i, M the IFB beam and column is proposed because the characteristic mo-
Mb, i, M (6)
ments in plastic hinges of the columns and beams (MY , MM , MU ) cor-
where Mb, i, k, h, p is the p -th characteristic moment in the h -th plastic respond to different values of the corresponding characteristic rota-
hinge of the k -th beam in the i -th storey of a frame building and Mb, i . k, h, p tions. This cannot be directly accounted for by the definition of the
is a contribution of the k -th beam in the i -th storey to the p -th moment-rotation envelopes of the IFB beam and column. With the

4
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

proposed weights, it is at least approximately taken into account that from the 1st to the 4th storey were equal to 87 t, 86 t, 83 t and 83 t [23],
the elements with greater strength have a more significant influence on respectively. The accelerogram for the first pseudo-dynamic test was
the definition of the characteristic rotations in the IFB columns and scaled to PGA = 0.12 g and for the second test to PGA = 0.45 g. The
beams. accelerogram was matched to Eurocode’s spectrum. More data about
Note that the proposed IFB model was investigated by OpenSees the building and experiments can be found elsewhere [29,30].
[27]. Hysteretic behaviour in plastic hinges was modelled by a uniaxial The SPEAR building is a 3-storey plan asymmetrical building
material Hysteretic, which is a typical peak-oriented model, similar to (Fig. 4b). It is a typical frame building design according to an older
Takeda’s hysteretic rules. The presented IFB model was incorporated design practice from Greece. The building was designed to gravity loads
into Performance-based Earthquake Engineering (PBEE) Toolbox [23], only [31]. Pseudo-dynamic tests were performed at ELSA Laboratory,
which is based on OpenSees [27] and Matlab [28]. The implemented Ispra, by considering an accelerogram that matched Eurocode’s spec-
model accounts for unsymmetrical moment-rotation relationship for trum, and was scaled to PGA = 0.15 g and PGA = 0.20 g for the first
beams with consideration of the effective slab width, the effect of de- and second test, respectively. In the analysis, the concrete strength of
gradation of stiffnesses as well as for the effect of significantly different 25 MPa and reinforcement steel tensile strengths of 459 MPa and
sizes of columns and beams in each storey of the frames. 377 MPa were assumed, depending on the diameter of the reinforce-
ment bar. Masses in the 1st and 2nd storeys amounted to 67 t and in the
3. Description of the analysed buildings 3rd storey to 63 t [35].
The third specimen considered in this study was ICONS building
Two sets of buildings were analysed with the aim to investigate the (Fig. 4c), which was also tested at ELSA Laboratory in Ispra [32–34].
capability of the IFB models. The first set includes buildings which were This 4-storey frame was designed to gravity loads only. In the analysis,
pseudo-dynamically tested in full-scale. These buildings are: 4-storey the concrete strength of 16 MPa and the reinforcement steel tensile
frame building [29,30], 3-storey plan asymmetrical reinforced concrete strength of longitudinal reinforcement of 343 MPa were assumed. In the
frame building [31] and 4-storey reinforced concrete frame building model, masses for the 1st, 2nd and 3rd storeys amounted to 46 t, while
[32–34], from now on denoted as PREC8, SPEAR and ICONS building, the mass of the 4th storey was considered equal to 40 t [36]. In addi-
respectively. The second set of buildings consists of symmetrical 6-, 8- tion, in this case, the accelerogram matched the Eurocode’s spectrum
and 11-storey reinforced concrete frame buildings, which were de- and was scaled to PGA = 0.22 g for the first test and to PGA = 0.29 g
signed according to the Eurocode 8 provisions [26]. The basic design for the second test, which corresponded to return periods of 475 and
principles of the investigated building are presented in Table 1. Note 975 years, respectively. The second test was terminated after approxi-
that the fundamental periods presented were based on the MDOF mately 7 s, i.e. at the middle of the test, because spalling and crushing
models that take into account the effective stiffness of the columns and of the concrete were observed at the top of the “strong” column in the
beams. 3rd storey, i.e. at the location of inadequate rebar lap splice, which
induced a soft storey mechanism [32,33].
3.1. Pseudo-dynamically tested buildings
3.2. Multi-storey code-conforming frame buildings
The elevation and the plan view of buildings from the first set,
which were pseudo-dynamically tested at ELSA Laboratory, are pre- The second set of buildings consisted of code-conforming multi-
sented in Fig. 4. storey frame buildings (Fig. 5). The 6-storey (6F) and 11-storey (11F)
The PREC8 building is a 4-storey reinforced concrete frame building buildings are regular in elevation, while the 8-storey building (8F) is
(Fig. 4a), for which a series of pseudo-dynamic tests in Y-direction were irregular in elevation due to the increased height of the first and the
performed. The building was designed in accordance with the pre- second storeys. All buildings were designed for ductility class medium
standard of the current Eurocode 8 by assuming ductility class high (DCM), as prescribed by Eurocode 8 [26], with PGA = 0.25 g and soil
(DCH, behaviour factor 5) and the medium soil condition [29,30]. The type B or C [37]. Concrete class C35/C45 was considered for 6F and
design PGA was equal to 0.3 g. The building was constructed with 11F, respectively, while concrete class C30/37 was adopted in the case
concrete class C25/30 and Tempcore reinforcement steel class B500. of building 8F. For all three buildings, B500 steel was used in the de-
The measured minimum and maximum concrete strengths (32 and sign. The mean concrete compressive strength was assumed 8 MPa
56 MPa) varied significantly from the nominal concrete strengths for greater than the characteristic value of compressive cylinder strength
C25/30 according to the Eurocode 2 provisions [21]. In the definition (i.e., 38 MPa and 43 MPa for C30/37 and C35/45, respectively) [21].
of the building models of the investigated building, the concrete For reinforcement, approximate mean yield strength was assumed 15%
strength was assumed to be equal to 42 MPa [23]. This simplification greater than the characteristic yield strength (575 MPa) [38].
does not affect the results significantly, because the reinforcement steel
governs the strength of the cross-section of RC frames designed ac- 3.3. Description of the IFB and MDOF building models
cording to the capacity design approach. The mean yield strength dif-
fered depending on the diameter of the reinforcing bar, while the The IFB models were developed for all the investigated buildings
average value amounted to approximately 580 MPa. Storey masses using the procedure described in Section 2. Note that 5% critical
damping proportional to the mass matrix was assumed for all analysed
Table 1 buildings for the fragility analysis presented in Section 5 [39]. How-
Basic design principles, the regularity of building, reference design peak ground ever, in the response history analysis of pseudo-dynamically tested
acceleration (soil type A) of the investigated buildings and fundamental periods building the damping was disregarded in order to be consistent with the
of the MDOF building models. test assumptions [30,31,33]. The pinching effect and the cyclic strength
degradation were disregarded in the Hysteretic uniaxial material
Building Designed according to Regularity agR [g] T1 [s]
OpenSees [27]. The parameter that controls the degrading unloading
PREC8 preEC8, DCH Plan, elevation 0.30 0.80 stiffness based on ductility was set to 0.8. The P-delta effect was con-
SPEAR Old design practice Elevation / 0.87 sidered for the IFB and MDOF models. 3D MDOF models were devel-
ICONS Old design practice / / 0.85 oped for all analysed buildings except the seismic response of ICONS
6F EC8, DCM Plan, elevation 0.25 1.00
frame was simulated by a 2D MDOF model [36]. The MDOF models
8F EC8, DCM Plan 0.25 1.76
11F EC8, DCM Plan, elevation 0.25 1.55 were developed with the PBEE Toolbox [23]. All analyses were per-
formed by OpenSees [27]. Masses and floors in models were assumed

5
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Fig. 4. The elevation and plan views of the pseudo-dynamically tested buildings: (a) PREC8, (b) SPEAR, (c) ICONS buildings.

lumped in the centre of masses at the floor level and rigid, respectively. that the coefficient of load redistribution cqF (see Eq. (4)) for all IFB
Characteristic points of the moment-rotation relationships of the models was considered equal to 8/9, as described in Section 2.
structural elements of the MDOF models were used to estimate the The PREC8 building was tested and analysed in Y-direction only.
characteristic points of the moment-rotation relationship of the IFB Beams in Y-direction for the MDOF model were modelled as 4.0 m and
columns and beams. The effective rotational stiffness of the beams and 6.0 m long. The IFB beam length LbF, i was calculated according to Eq. (2)
column were assumed equal to about 50% of the corresponding stiffness and it amounted to 2.5 m. Full contribution to the characteristic mo-
of the uncracked elements according to the Eurocode 8 provisions [26]. ments of the IFB beams according to the first condition of Eq. (6) was
The rotational stiffness of a column or beam was defined as proposed by considered because the building was designed according to the provi-
Ibarra and Krawinkler [40] with the stiffness of the beam-column ele- sions of the Eurocode 8 pre-standard.
ment (i.e. elastic beam-column element) and the stiffness of the plastic The SPEAR building is a frame building, for which the detailed 3D
hinge (i.e. rotational spring) [40]. For the MDOF models, the beam- MDOF model was already developed by several authors [24,35,41]. The
column joints were not modelled, which was also the case in previous MDOF model used in this study was based on the modelling assump-
studies [23–24,35–37,39]. The same applied to the IFB models. Note tions considered in previous studies [35], but it was restrained

Fig. 5. The elevation and plan views of the code-conforming frame buildings: (a) 6-storey, (b) 8-storey, (c) 11-storey buildings.

6
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

perpendicular to the XZ plane and subjected to ground motions in X accounted for the effect of concrete cracking, while practically no da-
direction only. This ensured consistency with the IFB model, which is mage was observed in the pseudo-dynamic test. Consequently, such
capable of simulating 2D response, but the results of the MDOF model models are most likely too flexible to simulate low-level demands [23],
are not consistent with that of the pseudo-dynamic test, since the but they are more appropriate for the simulation of high-level demand
SPEAR building is asymmetrical in the plan, and some torsional re- test, as it can be observed for the second test from Fig. 6.
sponse was observed during the pseudo-dynamic test [41]. However, by The relative error between the maximum values of EDPs is pre-
numerical simulations of the 3D MDOF model, it was observed that the sented in Table 2 for the 2nd storey. In the case of the second test, the
engineering demand parameters (EDPs) at the centre of mass are only relative errors between the results of the models ( F / M ) as well as be-
slightly affected by consideration of torsional response and the ground tween the results of the IFB model and the results of the PsD test ( F / PT )
motion in the Y direction. Thus, the MDOF model used in this study are low. The F / PT for the storey drift and the storey shear, are −3% and
provided a good approximation for the X component EDPs at the centre +1%, respectively. The largest relative error was observed for the
of mass, which have been obtained during the pseudo-dynamic test. The maximum values of absolute accelerations in the 2nd storey. This EDP
IFB model of the SPEAR building was developed according to defini- was overestimated by the IFB model by 14% if compared to the values
tions presented in Section 2.2. The IFB beam length LbF, i was calculated calculated based on the second PsD test. The relative errors for the EDPs
to be 2.2 m. The IFB model was developed to simulate the seismic re- in other storeys are for the second test similar to those presented for the
sponse of the SPEAR building in X-direction, although the building was 2nd storey. Slightly larger relative errors were observed for the 4th
simultaneously tested in both horizontal directions (X and Y). storey during the second test, where the lateral stiffness of both models
For the ICONS building a 2D MDOF model was developed by ana- was slightly overestimated. During the first test, relative errors F / PT
logy to the model proposed by Dolšek and Fajfar [36]. The IFB beam presented for the 2nd storey (Table 2) were similar for the EDPs in the
length was assumed 2.1 m according to Eq. (2). 1st storey and lower for those in the 3rd and 4th storey. The reason is
The MDOF models for the 6F, 8F and 11F buildings were obtained that damages concentrate in the bottom two storeys.
from previous studies [39,42]. The buildings were analysed only in X- It can be concluded that the IFB model of the PREC8 building
direction (Fig. 5). Since all three frame buildings were designed ac- provides EDPs that are very similar to those observed in the case of the
cording to the Eurocode 8 provisions and ductility class medium [26], MDOF model. Both models are also capable of simulating the EDPs with
the columns were stronger than the beams, which implies that there sufficient accuracy for the second high-demand PsD test, while the in-
was no need to consider the reduction of the beam strength for the itial stiffness of the models is too low for simulating the response of the
definition of the IFB model. However, this was not the case for the PREC8 building in the case when the structure is practically un-
SPEAR and ICONS buildings. More information about the analysed damaged. The simulation of the first low-level demand PsD test could
buildings is presented in Appendix A. be improved by considering a tri-linear moment-rotation relationship
for plastic hinges as it was shown before [23]. In this case, the ap-
4. Capability of the IFB models to estimate engineering demand proximate effective stiffness should be assumed equal to uncracked
parameters observed in pseudo-dynamic tests initial stiffness. However, such models are numerically less robust.

The capability of the IFB models to simulate the response history of 4.2. SPEAR building
the most important engineering demand parameters (EDPs) during
seismic excitations was investigated for buildings that were pseudo- In the case of the SPEAR building, the maximum storey drift was
dynamically (PsD) tested in full scale. The maximum storey shears, the also observed in the 2nd storey. The response histories of the observed
maximum storey drifts and the maximum values of absolute storey EDPs in this storey are presented in Fig. 7. The results of the IFB and
accelerations were the observed EDPs for the two tests of each in- MDOF models are very similar for the first and second test. The relative
vestigated building. The values of storey accelerations from the pseudo- error between the observed EDPs in the case of the MDOF model and
dynamic tests were calculated from the relative displacement response the IFB model ( F / M ) were less than 11% for both PsD tests (Table 3). In
history with the consideration of accelerograms used in pseudo-dy- other storeys, the observed relative errors for the EDPs are similar to
namic tests. In addition, the results of the IFB models were also com- those presented for the 2nd storey. Thus, it can be concluded that the
pared to those obtained by the MDOF models. The observed EDPs of the IFB model of the SPEAR building provides the observed EDPs with
SPEAR model corresponded to the X component of the EDPs at the accuracy similar to that of the MDOF model.
centre of the mass. The accuracy of the prediction of EDPs simulated by Both models were capable of simulating the maximum storey shear
the IFB models was measured with the relative error by taking into as well as the absolute storey accelerations, while the maximum storey
account the EDPs observed in the case of the MDOF model F / M or those drifts were overestimated for both PsD tests. However, the prediction of
observed in the case of pseudo-dynamic tests F / PT : the SPEAR building storey drifts is very sensitive to the level of accel-
eration, since the drift demand in the case of the second test was al-
EDP F EDP M
= ·100% ready slightly in the softening range of the structure (Fig. 7a). Conse-
F /M
EDP M (10)
quently, this means that small increment of peak ground acceleration of
EDP F EDP PT the considered accelerogram can cause a substantial increment of drift
= ·100% demand.
F / PT
EDP PT (11)

where and
EDP F , are the observed parameters for the IFB
EDP M EDP PT 4.3. ICONS building
model, MDOF model and PsD test, respectively.
The simulation of the ICONS building seismic response is quite
4.1. PREC8 building challenging for the IFB model because one column is significantly
stronger than the others. This issue can be approximately solved by an
The response histories of the observed EDPs are presented for the appropriate definition of the properties of the IFB beams (Eq. (6), Eq.
2nd storey (Fig. 6), where the greatest storey drifts were obtained in the (7)) and by an appropriate definition of the characteristic rotations in
pseudo-dynamic tests. The presented response histories corresponding plastic hinges of the IFB beams and columns (Eq (8), Eq. (9)). This can
to the IFB and MDOF models are very similar. However, the storey be confirmed by the presented response history of the observed EDPs
drifts from the first test were overestimated by both models (see also (Fig. 8). It can be concluded that the IFB model was capable of simu-
Table 2). This was expected because the initial stiffness of the models lating the seismic response of ICONS building with similar accuracy as

7
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

(a)
(b)
(c)

First test Second test


Fig. 6. (a) 2nd storey shear versus 2nd storey drift, (b) 2nd storey drift history, (c) 2nd storey absolute acceleration history for the IFB model and the MDOF model of
the PREC8 building, and the corresponding results of the PsD tests.

Table 2 structure. Spalling and crushing of the concrete were observed at the
Maximum 2nd storey drifts, corresponding maximum storey shears and max- top of the “strong” column in the 3rd storey, i.e. at the location of rebar
imum storey accelerations for the IFB model and the MDOF model of the PREC8 lap splice. This reinforcement detail was not appropriate, which in-
building, and the corresponding results of the PsD tests with an indication of duced a soft storey mechanism [33,34]. The termination of the test is
relative errors F / M and F / PT , which were calculated without rounding decimals also indicated in Fig. 8b and 8c.
of the considered EDPs. The maximum relative error F / PT for the observed EDPs was only
IFB MDOF F/M PsD test F / PT 11% (Table 4). The IFB model provided slightly overestimated drift
demand for the first test and slightly underestimated drift demand for
First test
the second test (Fig. 8). Nevertheless, the seismic response simulated by
Maximum 2nd storey drift [/] 1.0% 1.0% −2% 0.37% +174%
Maximum 2nd storey shear [kN] 1015 968 +5% 525 +93% the IFB model was slightly less accurate than that obtained by the
2
Maximum 2nd storey accelerations [m/s ] 5.0 4.4 +13% 4.0 +26% MDOF model. The maximum storey drift was overestimated with the
Second test
MDOF model by about 10% with respect to that observed in the PsD
Maximum 2nd storey drift [/] 2.3% 2.4% −4% 2.4% −3% test, while the IFB model underestimated the maximum drift by about
Maximum 2nd storey shear [kN] 1195 1184 +1% 1183 +1% 11%. The magnitude of the observed errors F / PT and F / M in the 2nd
Maximum 2nd storey accelerations [m/s2] 13.2 12.4 +6% 11.6 +14% and 4th storeys are similar to those presented for the 3rd storey (Fig. 8,
Table 4). However, in the case of the second test, both models slightly
overestimated the observed damage in the 1st storey. The maximum
storey drift observed during the PsD test was 0.63%, while the esti-
it was demonstrated by the MDOF model. In this case, the highest mated drifts with the IFB and MDOF models were 0.93% and 0.91%,
seismic demand was observed in the 3rd storey, which is consistent respectively.
with the observations from the PsD test [32]. This is most likely because
the cross-section area of the “strong” column of the ICONS building was 5. Capability of the IFB models for pushover and fragility analysis
reduced in the 3rd storey. It should be noted that the second PsD test
was terminated after around 7.4 s due to the indicated softening of the The capability of the proposed IFB models for pushover and fragility

8
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

(a)
(b)
(c)

First test Second test


Fig. 7. (a) 2nd storey shear versus 2nd storey drift, (b) 2nd storey drift history, (c) 2nd storey absolute acceleration history for the IFB model and the MDOF model of
the SPEAR building, and the corresponding results of the PsD tests.

Table 3 assumed that one of the three limit states is attained when the rotation
Maximum 2nd storey drifts, corresponding maximum storey shears and max- in one of the plastic hinges of the IFB columns first exceeds one of the
imum storey accelerations for the IFB model and the MDOF model of the SPEAR corresponding characteristic rotations from the moment-rotation re-
building, and the corresponding results of the PsD tests with an indication of lationship of the IFB columns. In the case of the MDOF model, the limit-
relative errors F / M and F / PT , which were calculated without rounding decimals states were defined at the level of a storey. This makes it possible to
of the considered EDPs. introduce consistency between the definition of the limit-states in the
IFB MDOF F/M PsD test F / PT case of the IFB and MDOF models. Thus, in the case of the MDOF model
it was assumed that LS1, LS2 or LS3 are attained when the weighted
First test
average rotation demand of the lower (h = 1) or upper (h = 2) plastic
Maximum 2nd storey drift [/] 1.5% 1.6% −3% 1.2% +28%
Maximum 2nd storey shear [kN] 174 174 +0% 161 +8% hinges of columns of the MDOF model exceed the weighted average
Maximum 2nd storey accelerations [m/s2] 3.9 3.9 −0% 3.3 +19% characteristic rotations of the moment-rotation relationship of the
Second test
corresponding plastic hinges. The limit states defined in this way were
Maximum 2nd storey drift [/] 2.2% 2.4% −11% 1.9% +13% considered in the pushover and the response history analyses.
Maximum 2nd storey shear [kN] 165 168 −2% 166 −1% Note that the limit-states in the case of the MDOF models can be
Maximum 2nd storey accelerations [m/s2] 4.6 4.5 +2% 5.0 −8% defined as presented in previous studies (e.g. [43,44]) where the de-
signated limit state is attained when the rotation of the first column
plastic hinge exceed first, second or third characteristic rotation. Such a
definition can also be used in the case of the IFB model.
analysis was investigated for both sets of buildings. The capability of
the IFB models for pushover analysis was measured by the relative error
F / M for the maximum base shear, the limit-state drifts and limit-state
5.1. Pushover analysis
displacement. The limit-state drifts were also used for the definition of
limit-state spectral accelerations in the case of fragility analysis. Pushover analyses using the IFB or MDOF models were performed
Limit states LS1, LS2 and LS3 were defined with consideration of the for the modal distribution of lateral forces, which were defined as a
different levels of damage in the building. For the IFB model, it was product of the storey masses and the first modal shape of the in-
vestigated building. The lateral forces were applied at a storey level in

9
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

(a)
(b)
(c)

First test Second test


Fig. 8. (a) 3rd storey shear versus 3rd storey drift, (b) 3rd storey drift history, (c) 3rd storey absolute acceleration history for the IFB model and the MDOF model of
the ICONS building, and the corresponding results of the PsD tests.

Table 4 observed relative errors F / M (Eq. (10)) for the maximum base shear and
Maximum 3rd storey drifts, corresponding maximum storey shears and max- fundamental period of the IFB model are practically negligible (Table 5)
imum storey accelerations for the IFB model and the MDOF model of the ICONS for all the investigated buildings. Relative error F / M in the prediction of
building, and the corresponding results of the PsD tests with an indication of the limit-state drifts and limit-state displacements did not exceed 5%
relative errors F / M and F / PT , which were calculated without rounding decimals (see Table 6). This is a good result provided that the observed limit-
of the considered EDPs.
state storey drifts varied from very low drifts of 0.30% to very high
IFB MDOF F/M PsD test F / PT drifts of 6.7%. For comparison reasons, the pushover curves were also
calculated by the conventional fish-bone (FB) models (Figs. 9 and 10). It
First test
can be observed that FB and IFB models provided very similar results in
Maximum 3rd storey drift [/] 0.88% 0.95% −7% 0.80% +10%
Maximum 3rd storey shear [kN] 150 151 −1% 139 +8% the elastic range of the response, while in the inelastic range, the IFB
2
Maximum 3rd storey accelerations [m/s ] 5.4 5.2 +4% 5.3 +2% models are more accurate than the FB models. This is because of the
Second test improved procedure for the definition of the moment-rotation re-
Maximum 3rd storey drift [/] 2.1% 2.6% −19% 2.4% −11% lationship of the plastic hinges of the IFB beams, which approximately
Maximum 3rd storey shear [kN] 148 150 −1% 154 −4% takes into account the effect of the different beam lengths, reduced
Maximum 3rd storey accelerations [m/s2] 5.6 5.7 −1% 5.3 +5% contribution of beam strength, bi-linear moment-rotation relationship
with softening branch and the unsymmetrical moment-rotation re-
lationship with consideration of effective slab widths. Note that the
pushover curves presented in Fig. 9, as well as the points representing
the centre of mass and the beam-column joints in the case of MDOF and the attained limit states (LS1, LS2 and LS3), were estimated at very
IFB models, respectively. The resulting pushover curves, which were similar values (Table 5) for the IFB and MDOF model. Therefore, the
obtained by utilising the IFB and MDOF models, are presented in Fig. 9 labels of the limit states of the MDOF model are basically behind the
and Fig. 10. It can be observed that the IFB models were capable of labels presenting the limit states of the IFB model.
predicting the pushover curves of the investigated buildings with high Please note that if the reduced contribution of the beam strength to
accuracy. The observation for pushover analyses in the negative di- the characteristic moments of the IFB beams were not taken into ac-
rection, which are not presented in Figs. 9 and 10, were similar. The count in the definition of the IFB model according to the second

10
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Fig. 9. Pushover curves with an indication of the three limit-state top displacements for the IFB, model and MDOF model using the modal distribution of lateral forces
for a) PREC8 building and b) 6F, 8F and 11F buildings. The pushover curves of the corresponding FB models are also presented.

condition of Eq. (6), the strength of the SPEAR and ICONS buildings Table 5
would be slightly overestimated, and the deformation capacity would The fundamental period of the IFB model, the maximum base shear – weight
be underestimated, especially in the case of the SPEAR building ratio and the corresponding relative errors between the IFB and MDOF models
(Fig. 10a). For example, for the SPEAR building, the base shear would of the investigated buildings.
be overestimated by 8% by the IFB model compared to the MDOF Building T1F [s] F/M FBS /W [%] F/M
model, whereas relative error F / M of top displacements for limit states
LS2 and LS3 would be −47% and −53%, respectively. As for the PREC8 0.79 −1% 36% −0%
ICONS building, the strength would be overestimated by 3% and re- SPEAR 0.85 −2% 10% +2%
ICONS 0.85 −1% 12% −2%
lative error F / M of top displacements for limit states, LS2 and LS3 6F 1.01 +1% 19% +0%
would be −12% and −7%, respectively (Fig. 10). It should be noted 8F 1.77 +1% 10% +1%
that the capabilities of the IFB model are not yet fully understood when 11F 1.56 +1% 12% −0%
used for the seismic performance assessment of buildings with weak
columns and strong beams. However, the proposed IFB model was
verified at least for the ICONS and SPEAR building, and the results of
the pseudo-dynamic tests were adequately simulated by the response IFB model with the full contribution of the beam strength under-
history analysis of the IFB models as presented in Sections 4.2 and 4.3. estimated the deformation capacity to the extent similar to that in the
An additional example was analysed to further check the capability of case of SPEAR building (Fig. 10a).
the IFB model. For this purpose, the columns of the 6F building were Although the configuration of structural elements of the IFB model
weakened intentionally to such extent that the beams became stronger was significantly different from that of the MDOF model, it was ob-
than the columns. Results of the pushover analysis proved that push- served that the distribution of the damage throughout the models is
over curves of the IFB and the MDOF model were very similar, while the similar. This can be observed in Fig. 11, where the damage in beam/

Fig. 10. Pushover curves with an indication of three limit-state top displacements for the IFB model, the IFB model with consideration of the full contribution of the
beam strength and MDOF model using the modal distribution of lateral forces for (a) SPEAR building and (b) ICONS building. The pushover curves the corresponding
FB models are also presented.

11
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Table 6 column plastic hinge rotations associated with the three limit states are
Limit-state storey drifts and limit-states top displacements for the IFB and presented for the IFB and MDOF models of the PREC8 building, and in
MDOF models of the investigated buildings and the corresponding relative error Fig. 12, where the damage associated with limit state LS3 is presented
F/M . for all the other analysed building models. It can be observed that the
Storey drift [/] Top displacement [m] damage pattern of the two models is similar also in the beams and not
only for the columns, which were taken into account in the definition of
Limit state IFB MDOF F/M IFB MDOF F/M the limit states.
PREC8 LS1 1.1% 1.1% +1% 0.10 0.10 0%
LS2 3.3% 3.2% +1% 0.28 0.28 0% 5.2. Fragility analysis
LS3 6.7% 6.7% +0% 0.50 0.50 0%

SPEAR LS1 0.32% 0.30% +5% 0.02 0.02 −4% The limit state drifts used for the definition of fragility functions
LS2 2.4% 2.4% −1% 0.14 0.13 +2% were determined by incremental dynamic analysis (IDA) [45], taking
LS3 3.8% 3.8% 0% 0.22 0.22 +1%
into account a set of 30 hazard-consistent ground motions (GM). The
ICONS LS1 0.42% 0.43% −3% 0.04 0.04 +3% ground motions for the fragility analysis of PREC8, 6F, 8F and 11F were
LS2 2.7% 2.6% +2% 0.11 0.11 −3%
the same as those used in the previous study [39]. The ground motions
LS3 4.3% 4.1% +1% 0.14 0.14 −1%
for the ICONS and SPEAR buildings were selected for this study using
6F LS1 0.52% 0.52% 0% 0.07 0.07 +1%
the ground-motion selection algorithm [46]. All sets of ground motions
LS2 3.8% 3.8% 0% 0.40 0.40 −1%
LS3 5.2% 5.2% 0% 0.51 0.51 +1% were selected from Strong ground motion database, which contains
database NGA [47] and RESORCE [48] that were recently combined
8F LS1 0.82% 0.82% 0% 0.12 0.12 0%
LS2 4.2% 4.2% 0% 0.48 0.48 0%
[49]. For the selection of ground motions, the target conditional spectra
LS3 5.6% 5.6% 0% 0.59 0.59 0% (CS) was defined with consideration of the spectral acceleration cor-
responding to the return period of 2475 years and location of Ljubljana,
11F LS1 0.94% 0.94% 0% 0.19 0.19 0%
LS2 4.1% 4.1% 0% 0.75 0.75 0% Slovenia. The conditional spectra (CS) was defined based on the official
LS3 5.2% 5.2% 0% 0.87 0.87 0% probabilistic seismic hazard maps for the region of Slovenia [50], and
the conditional mean spectrum was normalized to a target spectral
acceleration at the fundamental period T1 of the building model. The CS

Fig. 11. Distribution of damage associated with limit states LS1, LS2 and LS3 in the case of (a) the IFB model and (b) the MDOF model of the PREC8 building.

12
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Fig. 12. Distribution of damage associated with limit states LS3 in the case of (a) the IFB models and (b) the MDOF models of the SPEAR, ICONS, 6F, 8F and 11F
building.

PREC8 SPEAR ICONS

6F 8F 11F

Fig. 13. Target conditional spectra and corresponding spectra of the selected ground motions for the seismic fragility analysis of the PREC8, SPEAR, ICONS, 6F, 8F
and 11F buildings.

target spectra and the spectra of the selected ground motions are pre- observed for the median IDA curves of the SPEAR building. The points
sented in Fig. 13. in the IDA curves indicate the attainment of limit states LS1, LS2 and
The relationships between the story drift and the spectral accel- LS3 and define the limit-state spectral acceleration at the fundamental
eration at the fundamental period of the investigated buildings (IDA period of a building (Sae, LS ). It can be observed that the patterns of limit-
curves) are presented in Fig. 14. The IDA curves obtained by utilising state points based on the IFB and MDOF models are quite similar. With
the IFB and MDOF models are very similar even if they are compared the IFB models, the limit-state spectral accelerations are quite accu-
for a given ground motion. As a consequence, the median IDA curves of rately estimated compared to those obtained by the MDOF model.
the IFB and MDOF models match very well. Some differences can be However, a slightly more significant difference can be observed in the

13
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

PREC8 SPEAR

ICONS 6F

11F
8F

Fig. 14. IDA curves of each considered ground motion, the median, 16th and 84th percentile of IDA curves and points of the attained limit states for the IFB and
MDOF models of the PREC8, SPEAR, ICONS, 6F, 8F and 11F buildings.

estimation of limit-state maximum storey drifts, since small increments 2


=
1 Ngm
(ln (Sae, LS (r )) ln (Sae, LS ))2
LS
in the spectral accelerations can cause a considerable increment in the Ngm r=1
(13)
maximum storey drift demands, especially in the near-collapse range. In
addition, the limit-state storey drifts can be attained at slightly different where Ngm is the size of the sample of Sae, LS (r ) , i.e., the number of
maximum storey drift demands of the IFB and MDOF models. ground motions used in the IDA analysis. The most significant differ-
The fragility functions were then defined in the form of lognormal ence in the estimation of fragility parameters Sae, LS and LS and the
cumulative distribution functions (Fig. 15). This requires the estimation corresponding fragility functions using the IFB and MDOF models can
of the median value of the limit-state spectral acceleration at the fun- be observed in the case of the SPEAR building (Fig. 15, Table 7). Error
damental period of the building model Sae, LS and the standard deviation F / M in the prediction of fragility parameter Sae, LS is less than 13%, 6%
of logarithmic values of limit-state spectral acceleration at the funda- and 7% for the LS1, LS2 and LS3 limit states, respectively. The highest
mental period of the building LS (e.g. [37,40]): error was calculated for the LS1 limit state, due to the small intensity
measure levels, where a small absolute difference can account to high
relative error (Table 7). Note that the fragility functions are also pre-
1 Ngm
ln (Sae, LS (r ))
Sae, LS = e Ngm r= 1
(12)
sented in the form of the empirical cumulative distribution functions,

14
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

PREC8 SPEAR

ICONS 6F

8F 11F

Fig. 15. Empirical and lognormal fragility functions for the IFB and MDOF models of the PREC8, SPEAR, ICONS, 6F, 8F and 11F buildings.

which were obtained directly from the limit-state spectral accelerations of numerical analyses are particularly necessary for carrying out the
resulting from the response history analysis either using IFB or MDOF seismic fragility analysis. For this reason, the computational efficiency
model (Fig. 15). of the IFB models of the analysed buildings is presented in Table 8. It
In order to provide more insight into the results associated with the can be observed that the computational time required for performing
seismic fragility analysis, the maximum storey drifts along the height of analyses by using IFB models of more complex frame buildings (i.e. 6F,
the building are presented (Fig. 16). They correspond to the response 8F and 11F building) is only a few percent of the time needed for
history analyses based on the IFB and MDOF models by taking into carrying out the analyses by the conventional MDOF models. The dif-
account the ground motions scaled to the median spectral accelerations ference in the computational time, as presented in Table 8, is to some
causing limit states LS2, as it was observed for the MDOF model. It can extent related to the ratio between the number of the elements of the
be observed that the median values of the maximum storey drifts de- IFB and MDOF model. In addition, IFB models are numerically very
termined by the IFB and MDOF model are very similar. robust because there were no problems related to numerical non-con-
In addition to accuracy, the computational efficiency and robustness vergence.

15
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Table 7 showed, that the relative error in predicting the maximum storey drifts,
Parameters of fragility functions obtained by using the IFB and MDOF models of storey shears and storey accelerations by the IFB models is less than
the PREC8, SPEAR, ICONS, 6F, 8F and 11F buildings and the corresponding 19%, 5% and 13%, respectively. The average of the absolute values of
relative error F / M . relative errors in predicting all three parameters for all three buildings
Sae,LS LS
was only 5%.
Furthermore, it was shown that the IFB model could be used to
Limit state IFB MDOF F/M IFB MDOF F/M estimate the capacity of the examined buildings either by pushover
curves or fragility functions. The limit-storey drifts and limit-state top
PREC8 LS1 0.48 0.42 +13% 0.14 0.10 +36%
displacements obtained by the pushover analyses were estimated by
LS2 1.55 1.56 −0% 0.28 0.29 −2%
LS3 2.79 2.83 −2% 0.36 0.34 +7% less than 5% error compared to the results obtained by MDOF models.
The maximum error estimated for the median limit-state spectral ac-
SPEAR LS1 0.08 0.08 +5% 0.15 0.23 −35%
LS2 0.54 0.50 +6% 0.34 0.25 +37% celerations obtained by IDA was observed less than 13% considering all
LS3 0.75 0.70 +7% 0.43 0.42 +2% examined buildings.
ICONS LS1 0.13 0.13 +2% 0.15 0.12 +23%
The use of the proposed IFB model is currently limited to pre-
LS2 0.52 0.52 −1% 0.26 0.26 −1% dominantly plan-symmetrical buildings. In addition, the IFB model is
LS3 0.65 0.64 +2% 0.35 0.34 +4% not capable of simulating all the potential failure modes of the struc-
6F LS1 0.22 0.21 +7% 0.14 0.12 +17% tural elements of frame buildings, which may not significantly affect
LS2 1.51 1.55 −3% 0.36 0.35 +1% accuracy. However, further studies are needed to fully understand the
LS3 1.88 1.90 −1% 0.36 0.35 +4% capabilities and limitations of the IFB model for the seismic perfor-
8F LS1 0.12 0.11 +6% 0.14 0.12 +11% mance assessment of predominantly plan-symmetrical reinforced con-
LS2 1.07 1.09 −2% 0.44 0.43 +1% crete frame buildings.
LS3 1.26 1.28 −2% 0.37 0.36 +3% The computational robustness and the computational efficiency in
11F LS1 0.20 0.18 +11% 0.26 0.22 +18% the case of buildings with a large number of structural elements are the
LS2 1.43 1.43 −1% 0.40 0.41 +0% main advantages of the IFB model. Thus, it makes sense to explore the
LS3 1.65 1.69 −2% 0.41 0.41 +0%
possibilities of the IFB models for fragility analyses and loss estimations
of building portfolio. For this purpose, it will also be necessary to ex-
6. Conclusions tend the use of the IFB models for the estimation of seismic response of
asymmetric buildings.
Improved fish-bone (IFB) model was introduced, and its capability
for the seismic analysis was tested for six predominantly plan-symme-
trical reinforced concrete frame buildings. Results of the simulation of Declaration of Competing Interest
the seismic response of three pseudo-dynamically tested full-scale
structures proved that the IFB models are practically as accurate as The authors declare that they have no known competing financial
conventional MDOF models. The use of IFB and MDOF models for the interests or personal relationships that could have appeared to influ-
response history analyses of the PREC8, SPEAR and ICONS buildings ence the work reported in this paper.

PREC8 SPEAR ICONS

6F 8F 11F

Fig. 16. The maximum storey drifts resulting from the response history analysis using IFB or MDOF model and the ground motions scaled to the median spectral
acceleration causing LS2 in the case of the MDOF model. The maximum storey drifts and the corresponding median values are presented for the PREC8, SPEAR,
ICONS, 6F, 8F and 11F building.

16
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

Table 8
The computational times, the percentage of the computational time required to complete incremental dynamic analyses for the set of ground motions, the percentage
and number of elements of the IFB and MDOF models of the PREC8, SPEAR, ICONS, 6F, 8F and 11F buildings.
The computational time for IDA using The computational time for IDA using the The percent of the computational time for The percent and the number of
the IFB model [h] MDOF model [h] IDA (IFB/MDOF model) elements (IFB/MDOF)

PREC8 8.9 53.1 17% 11% (28/252)


SPEAR 4.4 17.4 25% 11% (21/189)
ICONS 5.7 9.1 63% 33% (28/84)
6F 10.1 627 1.6% 4% (42/1116)
8F 14.0 340 4.1% 6% (56/960)
11F 21.8 1261 1.7% 4% (77/2046)

Acknowledgements shared the MDOF models of code-conforming buildings, is highly ap-


preciated. Research presented in this paper is based on work financed
The help of Jure Žižmond, who provided ground motions for fra- by the Slovenian Research Agency. This support is gratefully ac-
gility analysis of the ICONS and SPEAR buildings, and Mirko Kosič, who knowledged.

Appendix A. Basic information about analysed buildings

The basic information about the design of analysed buildings, building masses, design base shear, mean values of yield strength of reinforcement
fsym and the mean values of concrete strength fcm are presented. A cross-section of typical column and beam in the 1st storey of each building along
with the moment-rotation relationship of plastic hinges are also shown. Please note, that all the dimensions are in centimetres [cm] and all the
diameters of the reinforcement are in millimetres [mm].

17
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

References [11] Soleimani R, Khosravi H, Hamidi H. Substitute Frame and adapted Fish-Bone
model: Two simplified frames representative of RC moment resisting frames. Eng
Struc 2019;185:68–89.
[1] Nakashima M, Ogawa K, Inoue K. Generic frame model for simulation of earthquake [12] Araki Y, Ohno M, Mukai I, Hashimoto N. Consistent DOF reduction of tall steel
responses of steel moment frames. Earthq Eng Struct Dynam 2002;31:671–92. frames. Earthq Eng Struct Dynam 2017;46(10):1581–97.
[2] Miranda E, Akkar SD. Generalized interstory drift spectrum. J Struct Eng [13] Li X, Kurata M. Probabilistic updating of fishbone model for assessing seismic da-
2006;132:840–52. mage to beam–column connections in steel moment-resisting frames. Comp-Aid Civ
[3] Fajfar P. A nonlinear analysis method for performance based seismic design. Earthq and Infr Eng 2019;34(9):790–805.
Spectra 2000;16(3):573–92. [14] FEMA. 440. Improvement of Nonlinear Static Seismic Analysis Procedures.
[4] Gupta A, Krawinkler H. Estimation of seismic drift demands for frame structures. Washington (DC): Federal Emergency Management Agency; 2005.
Earthq Eng Struct Dynam 2000;29:1287–305. [15] Ramirez CM, Miranda E. Building-Specific Loss Estimation Methods & Tools for
[5] Chopra AK, Goel RK. A modal pushover analysis procedure for estimating seismic Simplified Performance-Based Earthquake Engineering. Technical Report No. 171,
demands for buildings. Earthq Eng Struct Dynam 2002;31:561–82. John A. Blume Earthquake Engineering Center: Stanford, California; 2009.
[6] Hatzigeorgiou GD, Beskos DE. Inelastic displacement ratios for SDOF structures [16] Abdelnaby AE, Elnashai AS. Numerical modelling and analysis of RC frames sub-
subjected to repeated earthquakes. Eng Struct 2009;31:2744–55. jected to multiple earthquakes. Earthq Struct 2015;9(5):957–81.
[7] Luco N, Mori Y, Funahashi Y, Cornell CA, Nakashima M. Evaluation of predictors of [17] Kreslin M, Fajfar P. The extended N2 method considering higher mode effects in
non-linear seismic demands using ‘fishbone’ models of SMRF buildings. Earthq Eng both plan and elevation. Bull Earthq Eng 2012;10:695–715.
Struct Dynam 2003;32:2267–88. [18] Brozovič M, Dolšek M. Envelope-based pushover analysis procedure for the ap-
[8] Ogawa K, Kamura H, Inoue K. Modeling of moment resisting frame to fishbone- proximate seismic response analysis of buildings. Earthq Eng Struct Dynam
shaped frame for response analysis. Arch Inst Japan, J Struct Constr Eng 2014;43:77–96.
1999;521:119–26. [19] Haselton CB, Goulet CA, Mitrani-Reiser J, Beck JL, Deierlein GG, Porter KA, Stewart
[9] Khaloo AR, Khosravi H. Modified fish-bone model: A simplified MDOF model for JP, Taciroglu E. An Assessment to Benchmark the Seismic Performance of a Code-
simulation of seismic responses of moment resisting frames. Soil Dyn Earthq Eng Conforming Reinforced Concrete Moment-Frame Building. PEER Report 2007/12.
2013;55:195–210. Berkeley (CA): Pacific Earthquake Engineering Research Center, University of
[10] Khaloo AR, Khosravi H, Jamnani HH. Nonlinear interstory drift contours for idea- California at Berkeley; 2008.
lized forward directivity pulses using “Modified Fish-Bone” models. Adv Struct Eng [20] Dolšek M. Incremental dynamic analysis with consideration of modeling un-
2015;18(5):603–27. certainties. Earthq Eng Struct Dynam 2009;38:805–25.

18
A. Jamšek and M. Dolšek Engineering Structures 212 (2020) 110514

[21] CEN. European standard EN 1992–1-1: 2004. Eurocode 2: Design of concrete [36] Dolšek M, Fajfar P. The effect of masonry infills on the seismic response of a four-
structures – Part 1-1: General rules and rules for buildings. European Committee for storey reinforced concrete frame – a deterministic assessment. Eng Struct
Standardization; 2004. 2008;30:1991–2001.
[22] Haselton CB, Deierlein GG. Assessing Seismic Collapse Safety of Modern Reinforced [37] Kosič M, Fajfar P, Dolšek M. Approximate seismic risk assessment of building
Concrete Moment Frame Buildings. Department of Civil and Environmental structures with explicit consideration of uncertainties. Earthq Eng Struct Dynam
Engineering, Stanford University, CA. Report No. 156; 2007. 2014;43:1483–502.
[23] Dolšek M. Development of computing environment for the seismic performance [38] CEN. European standard EN 1998–2: 2005. Eurocode 8: Design of structures for
assessment of reinforced concrete frames by using simplified nonlinear models. Bull earthquake resistance. Part 2: Bridges. European Committee for Standardization;
Earthq Eng 2010;8:1309–29. 2005.
[24] Fajfar P, Dolšek M, Marušić D, Stratan A. Pre- and post-test mathematical modelling [39] Lazar Sinković N, Dolšek M, Žižmond J. Impact of the type of the target response
of a plan-asymmetric reinforced concrete frame building. Earthq Eng Struct Dynam spectrum for ground motion selection and of the number of ground motions on the
2006;35:1359–79. pushover-based seismic performance assessment of buildings. Eng Struct
[25] Celarec D, Dolšek M. Practice-oriented probabilistic seismic performance assess- 2018;175:731–42.
ment of infilled frames with consideration of shear failure of columns. Earthq Eng [40] Ibarra LF, Krawinkler H. Global Collapse of Frame Structures under Seismic
Struct Dynam 2013;42:1339–60. Excitations, John A. Blume Earthquake Engineering Center, Stanford University,
[26] CEN. European standard EN 1998–1: 2004. Eurocode 8: Design of structures for CA. Report No. 152; 2005.
earthquake resistance. Part 1: General rules, seismic action and rules for buildings. [41] Negro P, Mola E, Molina FJ, Magonette GE. Full-scale PSD testing of a torsionally
European Committee for Standardization; 2004. unbalanced three-storey non-seismic RC frame. Proceedings 13WCEE, Vancouver,
[27] McKenna F. OpenSees: A framework for earthquake engineering simulation. Comp B.C., Canada, Paper No. 968; 2004.
Sci Eng 2011;13(4):58–66. [42] Lazar N, Dolšek M. Incorporating intensity bounds for assessing the seismic safety of
[28] MathWorks. MATLAB the Language of Technical Computing; 2016. http://www. structures: Does it matter? Earthq Eng Struct Dynam 2014;43:717–38.
mathworks.com/. [43] Dolšek, M. 2012. Simplified method for seismic risk assessment of buildings with
[29] Negro P, Colombo A. Irregularities induced by nonstructural masonry panels in consideration of aleatory and epistemic uncertainty. Struct Infrastruct Eng
framed buildings. Eng Struct 1996;19(7):576–85. 2012;8(10):939-952.
[30] Fardis M. editor. Experimental and numerical investigations on the seismic response [44] Rejec K, Fajfar P. On the relation between the near collapse limit states at the
of R.C. infilled frames and recommendations for code provisions. ECOEST/PREC8, element and structure level. Second European Conference on Earthquake
Rep. No. 6, LNEC, Lisbon; 1996. Engineering and Seismology, Istanbul August 25–29. 2014.
[31] Fardis M, Negro P. editors. Seismic performance assessment and rehabilitation of [45] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthq Eng Struct
existing buildings. Proceedings of the international workshop, EC JRC, European Dynam 2002;31:491–514.
Laboratory for Structural Assessment, Ispra 4-5; 2005. [46] Jayaram N, Lin T, Baker JW. A computationally efficient ground-motion selection
[32] Pinho R, Elnashai AS. Dynamic collapse testing of a full-scale four storey RC frame. algorithm for matching a target response spectrum mean and variance. Earthq
ISET J Earthq Tech 2000;406(37):143–63. Spectra 2011;27(3):797–815.
[33] Carvalho EC, Coelho E. editors. Seismic assessment, strengthening and repair of [47] Chiou B, Darragh R, Gregor N, Silva W. NGA project strong-motion database. Earthq
structures. ECOEST2-ICONS report no. 2, European Commission – Training and Spectra 2008;24(1):23–44.
Mobility of Researchers Programme; 2001. [48] Akkar S, Sandıkkaya MA, Şenyurt M, Azari Sisi A, Ay BÖ, Traversa P, et al.
[34] Pinto A, Verzeletti G, Molina J, Varum H, Pinho R, Coelho E. Pseudo-dynamic test Reference database for seismic ground-motion in Europe (RESORCE). Bull Earthq
on non-seismic resisting RC frames (bare and selective retrofit frames). Joint Eng 2014;12:311–39.
Research Centre – Ispra, European Laboratory for Structural Assessment, Institute [49] Šebenik Ž, Dolšek M. Strong ground motion database. University of Ljubljana,
for Protection and Security of the Citizen, Europe Commission; 2002. Faculty of Civil and Geodetic Engineering, IKPIR; 2016.
[35] Dolšek M, Fajfar P. Post-test Analyses of the SPEAR Test Building. University of [50] Lapajne J, Šket Motnikar B, Zupančič P. Probabilistic seismic hazard assessment
Ljubljana; 2005. p. 112. methodology for distributed seismicity. Bull Seismol Soc Am 2003;93:2502–15.

19

You might also like