Journal of Constructional Steel Research 125 (2016) 142163

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Plastic analysis-based seismic design method to control the weak storey

behaviour of concentrically braced steel frames
D.B. Merczel a,b, J.-M. Aribert a, H. Somja a, M. Hjiaj a,
a
LGCGM/Structural Engineering Research Group, Institut National des Sciences Appliques de Rennes, 20 Av. des Buttes de Cosmes 35708, Rennes Cedex 7, France
b
Department of Structural Mechanics, Budapest University of Technology and Economics, Megyetem rkp. 3-9, Budapest 1111, Hungary

a r t i c l e i n f o a b s t r a c t

Article history: This article focuses on the likelihood of the weak storey behaviour and the weak storey collapse of diagonally con-
Received 6 May 2015 centrically braced frames designed according to Eurocode 8 provisions. The emphasis is primarily put on the na-
Received in revised form 1 May 2016 ture and development of the weak storey behaviour in order to designate the effects that shall be taken into
Accepted 15 May 2016
account in an effective design procedure. In a second stage, the focus is on developing supplementary conditions
Available online 22 June 2016
to Eurocode 8 based on plastic analysis that can enhance the designs by preventing the occurrence of weak
Keywords:
storeys.
Concentrically braced steel frame 2016 Elsevier Ltd. All rights reserved.
Seismic design
Eurocode 8
Weak story mechanism
Plasticity-based design method

1. Introduction Conversely, CBFs are susceptible to exhibit weak storey collapse by

Steel concentrically braced frame (hereinafter CBF) systems are eco-
nomic forms of providing lateral resistance to high-rise buildings.
Owing to the geometry, the lateral forces are resisted by truss action
resulting mainly axial effects in the participating members. The truss be-
haviour incorporates large stiffness, which limits the lateral drifts and
low-frequency vibrations in the braced buildings, and therefore pro-
vides occupancy comfort and impedes damages in the non-structural
parts. Therefore, CBFs may be favoured over moment resisting frames
in non-seismic design states or for low return period seismic actions.
However, in the case of severe earthquakes, CBFs are known to have a
bad performance [13].
According to the principles of capacity design, structures are to be
provided with large plastic ductility capacity as this allows the re-
duction of the design horizontal forces. Although the steel material
may assure considerable structural ductility, it is also necessary to
distribute the inelastic deformations along the height of the building.
In design, the expected behaviour of a CBF corresponds to the global
plastic collapse mechanism depicted in Fig. 1a. In this case, the com-
pression braces are buckled, and all the tensile ones are undergoing
plastic deformation. This plastic mechanism provides the largest dis-
sipation and lateral drift capacity as all the storeys of the building
equally contribute.
Corresponding author.
E-mail address: mohammed.hjiaj@insa-rennes.fr (M. Hjiaj).
developing a localised storey mechanism as it is shown in Fig. 1b. In
the weak storey phenomenon, the plastic deformations and drifts are
localised on one or a limited number of storeys. Evidently, the dissipa-
tive and drift capacity of the weak storeys is by far inferior to the capac-
ity of the whole CBF. Furthermore, the weak storey behaviour
incorporates a signicant bending of the columns, which may result in
plastic hinges at the top and the bottom of the bent columns. The
weak storey behaviour is unfavourable as it results in low seismic per-
formance and may even lead to early collapse; therefore, it needs to
be prevented.
In the following sections, the corresponding parts of EN 1998-1[4]
(Eurocode 8 or EC8) will be summarised, various CBFs designed accord-
ing to this standard will be presented, and their performance evaluation
will be carried out via incremental dynamic analysis (IDA). A deeper in-
sight into the observed weak storey behaviour will provide the basis for
the determination of the most inuential effects on the weak storey be-
haviour. Finally, new design criteria, based on plastic analysis, will be
dened, which will be utilised to enhance the performance of the CBFs
investigated in previous sections.
2. Eurocode 8 design of CBFs
In the design of building structures, Eurocode 8 provides two basic
concepts for the seismic analysis. Earthquake-resistant buildings can ei-
ther have a low-dissipative or a dissipative behaviour. In a dissipative
structure, controlled inelastic deformations are expected to dissipate
signicant energy and damp the response of the structure, and in the

http://dx.doi.org/10.1016/j.jcsr.2016.05.008
0143-974X/ 2016 Elsevier Ltd. All rights reserved.
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
a)
Fig. 1. Geometry model of collapse mechanism. (a) Global mechanism, (b) local mechanism.
meantime to provide adequate ductility to meet displacement de-
mands. The behaviour factor, q, that accounts for this effect is 4 for
CBFs, which is the lower boundary of high ductility class (DCH) struc-
tures. The seismic action is divided by the behaviour factor and the
members are designed to resist only this reduced-intensity design
action.
In Eurocode 8, the criteria for the verication of dissipative struc-
tures follow the capacity design philosophy. For the design analysis,
only a reduced-intensity seismic action is considered, as a substantial
part of the seismic input energy is expected to be dissipated by plastic
deformations. Plastic deformations are strictly required to occur in dis-
sipative members that are to be designed with sufcient ductility,
whereas the yield or collapse of non-dissipative zones is to be evaded.
To ensure the elastic behaviour of non-dissipative zones, their seismic
design is required with a sufcient overstrength. In CBFs, the braces
are meant to be the dissipative members. Therefore, CBFs are designed
so that the yield of the diagonals takes place before the failure of the col-
umns, the beams and the connections.
For the sake of homogeneous dissipative behaviour, a simultaneous
yield of the braces on every storey has to be ensured. Eurocode 8 aims to
promote this by a condition given for the overstrength factors, i,
realised on the different storeys, making them closely uniform. It
needs to be veried that the maximum overstrength does not differ
from the minimum by more than 25%.
max
1:25
where
mini
max maxi
Npl;Rd;i
i 4
N br;Ed;i
For the design of columns and beams, Eurocode 8 requires the full-
ment of the following condition:
NRd MEd NEd;G 1:1  ov   NEd;E
where NRd(MEd) is the axial resistance in accordance with Eurocode 3,
taking into consideration the interaction with the design bending effect
MEd. NEd,G and NEd,E are the axial forces due to the non-seismic and the
seismic actions, respectively. The material overstrength factor, ov, ac-
counts for the random variability of the material properties. Its recom-
mended value is 1.25, but it may be varied in National Annexes. The
amplication coefcient 1.1 represents the increase of the yield stress
143
b)
of the dissipative members due to strain hardening. The total
overstrength, 1.1ov , is a way to account for the resistance reserve
of the diagonal members. In principle, it ensures that the resistance of
the non-dissipative members is adequate until the plastic yield of the
diagonals.
In addition, Eurocode 8 denes limitations to the relative slender-
ness:
2:0 6
and for X-braced congurations also:
1:3 7
The upper bound dened in Eq. (6) is imposed to prevent the rapid
degradation of the resistance of the braces. Furthermore, this limits the
plastic out-of-plane deformation of gusset plates, which are prone to
low-cycle fatigue fracture. The lower bound Eq. (7) assures the suf-
cient exibility of the diagonals in compression.
With its requirements, Eurocode 8 assumes and attempts to pro-
mote the development of the global plastic mechanism. However,
Eurocode 8 criteria are based on the elastic response of the structure
subjected to the reduced-intensity seismic action via the q-factor.
These may not be well adapted to control the inelastic response, partic-
ularly if the CBF exhibits a weak storey behaviour. That is to say, in an
elastic analysis, due to the truss action of the bracing, only negligible
1
bending can be expected. On the contrary, the columns in a weak storey
response are subjected to signicant bending, so the requirement im-
posed on the axial resistance of the columns, Eq. (5), may not be satis-
2 factory to provide adequate column sections. In a weak storey
response, the axial forces of the braces may also be different from the
3 ones obtained by elastic analysis; therefore, the uniformity condition
of the storey overstrength factors, Eq. (1), may be violated.
The drawbacks related to the Eurocode 8 design of CBF have been
addressed before by various authors. Elghazouli [1] deals with the
main behavioural issues involved in the seismic design of typical
forms of concentrically braced frames. Tremblay [5] also gives a com-
prehensive description of the seismic behaviour of CBFs and suggests
novel additional bracing congurations that may favour the global
5 mechanism. Martinelli et al. [6] apply the tension only concept of EC8
while proposing a new strategy based on the denition of a set of static
equivalent seismic forces, computed from response spectrum analysis.
Having observed the relevance of the resistance reserve provided by
the columns, Elghazouli [7] proposes that the ratio of the sum of the
horizontal stiffness provided by the continuous columns and the axial
stiffness of the brace on every oor shall be kept sufciently high to pre-
vent the localisation of the ductility demand on a weak storey. In their
articles, Longo et al. [8,9] and Brandonisio et al. [2] highlight certain
144 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
drawbacks of Eurocode 8 and propose modications. In the referenced
articles, it is pointed out that dening the overstrength factor, , as
the minimum of the storey overstrength to penalise the seismic internal
forces of non-dissipative members does not satisfy the principles of ca-
pacity design sufciently. Also, it is emphasised that the non-
dimensional slenderness limits of the braces may result in unfeasible
or vastly oversized designs. Understanding that the brace buckling and
deterioration, both in terms of resistance and dissipation, are account-
able for the development of the weak storey behaviour led to the ap-
pearance of alternative solutions that exhibit superior response. One
innovative bracing member that has dynamically spread throughout
the world in building practice and also studied by many researchers
[10,11] is the buckling restrained brace or commonly BRB. Another al-
ternative of enhancing seismic braced frame behaviour is the applica-
tion of self-centering braced frames [12]. These braced frames are
xed to the base in a way that the uplift of the column base beyond a
predened overturning moment is enabled.
3. Design of various braced frames
In order to evaluate the performance of CBFs that comply with the
requirements of Eurocode 8, a series of concentrically braced steel
frame structures were designed. The designs, their features, and funda-
mental periods are presented in Table 1. There are 4, 6, 8, and 10 storey-
high buildings denoted by CBF4, CBF6, CBF8, and CBF10 respectively.
The buildings are identical in plan. They are built on an orthogonal
grid that consists of the same four 6-m bays in both plan directions.
The storey height is 3 m in every building and on every storey. In the
buildings where the lateral resistance in both directions is provided by
braces on the four external frames, see Fig. 2a; only the columns that
are connected to the diagonals participate in the lateral resistance.
These are continuous throughout the whole height of the building but
pinned at the base. All the other column splices are hinged and located
at the slab levels. Although such detailing does not reect current prac-
tice, it has been adopted to better demonstrate the efciency of the
RSBD method developed in Section 7. These structural types are denot-
ed by the number 1 following the storey number, e.g., CBF41 (4-storey
CBF with bracing in both directions). In the buildings where the lateral
resistance is provided by moment resisting frames in one direction
and by bracing in the other, see Fig. 2b; all the columns are continuous.
These buildings are denoted by the number 2 after the storey number,
e.g., CBF42. At the end of the designation, the acronym EC8 refers to
the design procedure.
Both the gravity and the lateral loads are transmitted to the steel
frame members by a concrete slab. However, for the steel I-section
beams placed under the slab as ribs, the design has been carried out as-
suming that there is no composite behaviour.
The change of the mass on some oors in some CBFs has also been
considered as special cases. In CBF81, on the upper 4 oors a 20%, and
in CBF61, on the upper 3 oors, a 50% weight increase is assumed. The
dimensioning according to Eurocode has been carried out with regard
Table 1
Various CBF designs
Building Features
CBF41-EC8 4-storey CBF in both directions
CBF42-EC8 4-storey MRFCBF
CBF61-EC8 6-storey CBF in both directions
CBF62-EC8 6-storey MRFCBF
CBF81-EC8 8-storey CBF in both directions
CBF82-EC8 8-storey MRFCBF
CBF101-EC8 10-storey CBF in both directions
CBF102-EC8 10-storey MRFCBF
CBF61M-EC8 6-storey CBF, 50% mass irregularity, upper half
CBF81M-EC8 8-storey CBF, 20% mass irregularity, upper half
to the augmented gravity and seismic actions. In these cases, the letter
M is indicated in the designation.
For the elastic analysis, 3D nite element models have been built,
and the dimensioning of the sections has been done in compliance
with the corresponding parts of Eurocode 1, 3, and 8. The structural
self-weight is 6.77 kN/m2 distributed load. It is assumed that the build-
ing is of B occupation category; therefore, the live loading is 3 kN/m2
and the 2 combination factor is 0.3; therefore, 30% of the live loading
is concurrent with the seismic action. The earthquake acceleration re-
sponse spectrum is considered to be the type 1 in EC8 assuming B
type soil conditions. The design ground acceleration is taken to be
0.25g, where g is the acceleration of gravity. The behaviour factor, q, is
4 in agreement with the structure-specic regulations of the standard.
The seismic design has been carried out with the modal response spec-
trum analysis method, considering a sufcient number of modes to ex-
ceed 90% mass participation in the lateral directions. To account for
buckling of the braces in compression, one brace in a pair has been re-
moved. For the combination of the modal responses, the complete qua-
dratic combination method has been used. All the members were
designed to respect the corresponding parts of both Eurocode 3 (EN
1993.1.1, in particular clause 6.3.3) and 8. In Table 2, the cross sections
of the internal and external (faade) columns and the square hollow
section (SHS) braces, as well as the slenderness of the braces and the
storey overstrength results, are presented for selected, representative
examples. The steel grade of every member is S235.
4. Model for nonlinear dynamic analysis
To observe the actual behaviour and to evaluate the performance of
the CBF designs, nonlinear dynamic analyses have been carried out. For
the IDA, seven articial accelerograms were selected to respect the min-
imum requirement of Eurocode 8 (clause 4.3.3.4.3) so that the average
of the results shall be considered. Selecting articial accelerograms re-
duces the uncertainty of the seismic action and is therefore less
favourable than real motion (natural accelerograms) to detect the struc-
tural susceptibility to develop a weak storey mechanism. Large discrep-
ancies between accelerograms increase the probability to observe a
singular structural behaviour such as weak storey mechanism. On the
other hand, the limited uncertainty of the seismic action allows to eval-
uate the efciency of the adopted design method and to make better
comparison of design approaches.
The length of each excitation is 20 s and the spectra of the ground
motions were tted to the design spectrum, see Fig. 3.
The used accelerograms are benchmarks in the RFCS OPUS research
program [13]. The acceleration records were scaled by multiplying the
peak ground acceleration (PGA) by a scale factor ranging from 0.1 up
to 2.0. The analysis was conducted with FinelG [14] nonlinear nite ele-
ment software. The model consists of 2D bre beam elements with a co-
rotational formulation that can adequately describe large displace-
ments. Due to the planar modelling, the parallel columns were
projected on the plane of the braced frame on the facade and connected
at the oor levels by horizontal links, see Fig. 4.
The members were divided into multiple beam elements to account
for the large deformations (of the braces primarily). The beam elements
have 3 nodes and 7 degrees of freedom. The total number of DOFs cor-
responds to two rotational at end nodes and ve translational, see
Period (T1 s) Fig. 5. As usual for bre element, the section forces at the element
1.13 nodes are computed using both a longitudinal and transversal integra-
1.19 tion scheme. The integration along the beam length is performed
1.65 using 3 integration points. For each longitudinal integration point, a
1.61
transversal integration scheme is performed. On the webs of the thin-
2.10
2.04 walled section, 5 integration points are located and the anges are di-
2.54 vided by 7 points along the plate length and 3 along the thickness. In
2.41 the llets additional integration points are considered.
1.78 The cyclic nonlinear response of the material was modelled by the
2.17
application of the GiuffrMenegottoPinto cyclic law [15] (with
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 145

a) b)
Fig. 2. Plan and structural model of CBF buildings. (a) CBF in both plan directions (type 1). (b) MRFCBF building (type 2).

parameters E0 = 210 GPa, Eh = 2.1 GPa, fy = 235 MPa, = 0.3, R1 = 20,
A1 = 18.5, A2 = 0.15, see Appendix A). The masses and simultaneous
loads were lumped into the intersections of the beams and columns.
For the diagonals, their self-weight and an initial sinusoidal imperfec-
tion with an amplitude of L/250 (in accordance with EC3 [16]) were
both considered. The nonlinear time history analysis has been carried
out by the Newmark direct integration method (with parameters
= 0.25, = 0.5). The equivalent structural damping has been
implemented by Rayleigh damping [17], the two parameters of which
have been determined assuming 4% damping ratio [18] and considering
the rst and second mode of the structural model.
In order to validate the appropriateness of the nite element models,
the results of an experimental test, carried out by Black et al. [19],
has been adopted and modelled with FinelG. A bar with a
4 in. 4 in. 0.25 in. square hollow section and 10 feet length is loaded
with an alternating and monotonically growing-amplitude imposed

Table 2
Details of the selected CBF designs

st. int.col. ext.col. brace st. int.col. ext.col. brace

CBF41-EC8 4 HEA 160 HEA 140 100 4 1.83 1.05 CBF61-EC8 6 HEA 160 HEA 160 90 5 2.08 1.00
3 HEB 180 HEA 200 100 6.3 1.87 0.99 5 HEB 180 HEA 180 100 6 1.89 0.98
2 HEB 220 HEB 200 100 8 1.91 0.97 4 HEB 220 HEB 200 100 8 1.95 1.04
1 HEB 260 HEB 240 100 10 1.95 0.99 3 HEB 240 HEB 220 100 10 1.95 1.04
CBF81-EC8 8 HEA 160 HEA 160 90 5 2.06 0.99 2 HEB 280 HEB 240 100 10 1.95 1.01
7 HEB 180 HEA 180 100 6.3 1.87 1.05 1 HEB 300 HEB 280 120 10 1.63 1.00
6 HEB 220 HEB 200 100 8 1.95 1.01 CBF82-EC8 8 HEA 180 HEA 200 90 5 2.06 1.02
5 HEB 240 HEB 220 100 10 2.01 1.10 7 HEB 180 HEB 180 100 6.3 1.87 0.99
4 HEB 280 HEB 240 120 10 1.63 1.01 6 HEB 220 HEB 200 100 8 1.95 0.99
3 HEB 300 HEB 280 120 10 1.63 0.98 5 HEB 260 HEB 240 100 10 2.01 1.03
2 HEM 240 HEB 320 140 10 1.38 0.98 4 HEB 280 HEB 260 120 10 1.63 1.13
1 HEM 240 HEM 240 140 10 1.38 1.00 3 HEB 300 HEB 300 120 10 1.63 1.03
CBF101-EC8 10 HEA 160 HEA 160 100 5 1.86 0.98 2 HD320 158 HD320 158 140 10 1.38 1.08
9 HEB 180 HEA 180 100 8 1.95 1.00 1 HD360 179 HD320 158 140 10 1.38 1.01
8 HEB 220 HEB 200 100 10 2.01 1.00 CBF81M-EC8 8 HEA 180 HEA 160 100 5 1.85 1.00
7 HEB 240 HEB 220 120 10 1.60 1.11 7 HEB 200 HEA 200 100 8 1.91 1.08
6 HEB 280 HEB 260 120 10 1.60 1.02 6 HEB 240 HEB 200 100 10 2.01 1.03
5 HEB 300 HEB 280 140 10 1.36 1.07 5 HEB 260 HEB 240 140 7.1 1.32 1.02
4 HEM 240 HEB 320 140 10 1.36 1.01 4 HEB 300 HEB 280 120 10 1.63 1.02
3 HEM 240 HEM 240 140 12.5 1.39 1.09 3 HEB 320 HEB 320 140 10 1.38 1.09
2 HEM 260 HEM 260 140 12.5 1.39 1.03 2 HEM 240 HEM 240 140 10 1.38 1.00
1 HEM 280 HD320198 150 12.5 1.39 0.98 1 HEM 260 HEM 260 150 10 1.27 1.00
146 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
Fig. 3. Design and articial spectra
displacement sequence. In Fig. 6, the measured and the computed be-
haviour of the brace is presented. As it can be seen, the curves are mostly
overlapping one another. Some differences can be seen in the compres-
sion behaviour as the local buckling of plates at the mid-section [20]
that decreases the compressive resistance is not represented by the
beam element model applied in the analysis.
5. Evaluation of the performance of EC8 CBF designs
5.1. Representation of the responses
In Fig. 7, the performance of the CBF designs is depicted by diagrams
derived from the IDA. The diagrams do not have the usual appearance
[21] where the damage against the seismic intensity is represented by
one curve for each acceleration record. The likelihood of the weak storey
behaviour is better visualised by taking the average (of seven accelera-
tion records) relative interstorey drift results divided by the storey
height, i.e., the interstorey drift ratio (IDR), and plotting its evolution
for every oor separately in the considered seismic intensity range. In
the diagrams, the emphasised scale factor 1.0 is the design seismic in-
tensity, above which unfavourable response or failure is of less signi-
cance. In this evaluation, failure is considered above the 2% IDR limit.
Fig. 4. 3D to 2D projection of CBF structure.
On one hand, this limit correlates with the EC8 limitation of interstorey
drift (clause 4.4.3.2) assuming no damage sensitive secondary members
and a reduction factor of the seismic action = 0.5. On the other hand,
this value can be substantiated by the experimental results of various
authors.
In order to examine the fracture life of rectangular and square hol-
low section braces, several authors conducted comprehensive experi-
mental surveys [1923]. Upon the assessment of the cyclic test results,
various methods and formulas were established that aim to predict
the fracture life [2427]. Although the formulas may differ, the different
authors agree that the most important factors inuencing the rupture of
hollow braces are the global and the local slenderness (i.e., and
breadth-to-thickness ratio) and the material grade. Assuming S235
grade material and a judicious choice of both local and global slender-
ness from the EC8 permitted range, the failure of SHS braces may be ex-
pected in the 23% percent interval. Available literature on the
resistance of H-section columns [28,29] and the gusset plates [30,31]
also indicate the viability of the designated failure limit.
The diagrams in Fig. 7 are representative examples of the behaviour
obtained for every CBF considered in the research. It is worth to mention
that below 1.5 times the design seismic intensity, numerical conver-
gence problems were not observed in any building.
5.2. Interpretation of IDA diagrams
In order to characterise the responses in terms of weak storey sus-
ceptibility, two hypothetical 4-storey IDA curves are presented in
Fig. 8. The Fig. 8a shows the ideal behaviour of a CBF assuming the devel-
opment of the global plastic yield mechanism where the interstorey
drifts are equal on every storey and the IDA curves of the different oors
are the same. The real curves of Fig. 7 are more alike the ones presented
in Fig. 8b.
In the real responses, it can be observed, that the families of the
curves are rather broad. The upper oors deviate from the rest at a
small scale factor, approximately 0.25, where the occurrence of rst
yield is anticipated, as the behaviour factor is 4. Beyond 0.25 scale factor,
in the inelastic response range, the largest and the smallest maximum
IDR results show great differences. The differences are large even at
scale factors below the design level and especially in the taller buildings
the discrepancy is agrant. The buildings therefore do not exhibit a
well-distributed dissipation; there is at least one, or in the taller build-
ings there are several weak storeys. It is also important to note that
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
Fig. 5. DOF and integrations points of nite element.
the rise of the weak storey drifts on the upper oors decreases and at
higher scale factors other oors may even catch up.
Regarding the differences that the moment resisting frames (type 2
buildings) or the mass irregularity cause, we may see that the behaviour
fundamentally does not change, see Fig. 7/c, d, e diagrams. The numer-
ous continuous columns in CBF82-EC8 mitigate the development of the
weak storey on the top compared to CBF81, but the weak storeys are
still there, and the failure limit is attained at closely the same seismic in-
tensity. The mass irregularity also modies the shape of the IDA dia-
gram, but due to the reinforcement to carry the excess loading, the
performance of CBF81M is not worse than of CBF81.
As it has been pointed out, the weak storey behaviour may occur al-
ready at low seismic intensity but may also get strongly mitigated as the
intensity increases. This shows that the likelihood of the weak storey be-
haviour is not related solely to the intensity of the loading. Thus, the
weak storey phenomenon is not simply due to inadequacy to the im-
posed loading but the weakness of one oor, in comparison with the
other oors. The weak storey behaviour is an inherent susceptibility of
the structure coming from the structural arrangement, the cross-
sectional properties and connections. If so, the possibility of the occur-
rence of a weak storey may be controlled taking into consideration
only the characteristics of the structure, without particular consider-
ation to the design seismic intensity [32].
As the results indicate a unanimous adverse behaviour, we can con-
clude that the CBFs designed according to Eurocode 8 are susceptible to
develop weak storeys even when subjected to seismic actions that are
signicantly smaller than the design level.
Looking at the IDA curves of CBF41-EC8 and with a bit of allowance
on the failure limit CBF61-EC8, one can see that the lateral drift
disproportionally increases on one oor at a low scale factor, but early
collapse is not realised. The collapse limit is attained at a scale factor
higher than the design level; therefore, the building is considered ade-
quate. Yet, the behaviour of the building is not optimal as the lateral dis-
placements and dissipation are concentrated on some oors, the other
oors are not efcient.
Fig. 6. Experimental and numerical axial forcedisplacement curves of SHS brace
147
The diagram in Fig. 8b depicts a soft storey phenomenon-induced
early collapse. In this case, the building is considered collapsed at a
scale factor smaller than the design level. The failure, however, is due
to unacceptably large deformations, not asymptotic increase of the dis-
placements. Although the building may have been designed to have ad-
equate lateral resistance, so that (asymptotic) failure is reached beyond
the design level, still the disproportional plastic deformations lead to an
early collapse. CBF81-EC8, 82, 101, and 81M exhibit such inadequate
response.
It is clear that the weak storey phenomenon has to be avoided as it
facilitates collapse or just entails an adverse plastic behaviour. In order
to obtain the optimal global plastic or, more precisely, a distributed dis-
sipative behaviour, the differences between the storey drifts have to be
reduced. However, even a building with a weak storey likelihood can be
adequate as it has been discussed above. Therefore, the development of
a weak storey behaviour and the actual collapse can be uncoupled as
one does not necessarily mean the other. The weak storey phenomenon
does facilitate and therefore may result in collapse due to extensive
localised plastic deformations, but it may also result only uneven
dissipation.
5.3. Tentative strengthening of the weak storey braces
Considering CBF41-EC8 and CBF61-EC8, it can be seen that there are
moderately developed weak storeys on the top oors in both cases. Let
us make an attempt to enhance the behaviour of the two buildings by
providing additional resistance on these oors. Replacing the braces ob-
tained by the Eurocode 8 design on the top oors by SHS 100 6 diag-
onals, we get the IDA curves presented in Fig. 9. First, we can see that the
top oors became over-reinforced. This is not surprising as the selection
of the new cross sections is not based on a rened design analysis. Com-
paring to the originals in Fig. 7a and b, it can also be noticed that the sto-
reys below (3rd and 5th) exhibit a worse behaviour after reinforcement.
In CBF61 the presence of the weak storey behaviour can clearly be seen
on the 5th oor. The additional resistance does not enhance the perfor-
mance of the buildings but modies the storey of the localisation. The
reinforcing of the storeys that were found to be inadequate does not
necessarily yield satisfactory designs and the disappearance of the
weak storey phenomenon. Therefore, it is clear that strengthening the
structure on the basis of the experienced effects may not provide better
designs and that the likelihood of the weak storey behaviour is indeed
not simply a storey resistance issue.
The observations so far indicate, that it is necessary to investigate the
underlying reasons of the development of the weak storey behaviour
and that a design method shall be elaborated to prevent its occurrence.
6. Development of the weak storey behaviour
6.1. Permanent deformations due to seismic actions
When a CBF is subjected to seismic action, the braces can undergo
several cycles of alternating deformations. Fig. 10 depicts the axial
forcedeformation diagram of the hollow section brace on the 4th
oor of CBF41-EC8 obtained with one of the nonlinear time history
148 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

4.5 4.5
St4 St6
CBF41-EC8 CBF61-EC8
Interstorey drift ratio [%]

Interstorey drift ratio [%]

4.0 St3 4.0 St5
St2 St4
3.5 St1 3.5 St3
St2
3.0 3.0 St1
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
a) b)
4.5 4.5
St8 St8
CBF81-EC8 CBF82-EC8

Interstorey drift ratio [%]

St7 St7
Interstorey drift ratio [%]

4.0 4.0
St6 St6
3.5 St5 3.5 St5
St4 St4
3.0 St3 3.0 St3
2.5 St2 2.5 St2
St1 St1
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
c) d)
4.5 4.5
St8 St10
CBF81M-EC8 CBF101-EC8
Interstorey drift ratio [%]

St9
Interstorey drift ratio [%]

4.0 St7 4.0

St6 St8
3.5 St5 3.5 St7
St4 St6
3.0 St3 3.0 St5
2.5 St2 2.5 St4
St1 St3
2.0 2.0 St2
St1
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
e) f)
Fig. 7. IDA diagrams of selected CBFs. (a) CBF41-EC8, (b) CBF61-EC8, (c) CBF81-EC8, (d) CBF82-EC8, (e) CBF81M-EC8, (f) CBF101-EC8.

analyses. In this diagram, most of the tensile peaks do not cause yield;
however, in three cycles, considerable plastic elongation is realised. As
the deformed length of the brace becomes larger than the original
length, buckling occurs even in the positive deformation range. Also
due to the increased length, for the development of the tensile resis-
tance of the brace larger axial, displacements are needed. Therefore, in
the case of small tensile displacements, the counteracting force in the
brace is diminished signicantly, i.e., the tangent stiffness decreases. In
the diagram, both phenomena are indicated.
When the brace is in compression, owing to the combined effect of
the axial force and the bending resulted by the buckling, a semi-
plastic hinge develops at mid length. This plastic hinge generally ap-
pears associated with local plate buckling. In the repeated cycles, the
buckled pattern folds and unfolds over and over again and the edges
in the buckled zone undergo repeated alternating plastic deformations.
The accumulation of plastic deformations decreases the moment
resistance of the mid-section, facilitates buckling, and may even lead
to a local low-cycle fatigue rupture.
Together, the axial elongations and the development of the plastic
hinge at the mid length result in having triangular imperfect-shape
braces, see Fig. 11. The triangularity of the deformed braces in a CBF
following an earthquake is not necessarily equal on every storey.
Large triangular deformation is a consequence of large interstorey
drifts during the earthquake. As in the weak storey behaviour, large
lateral drifts are localised on one oor; the prominent triangular
degradation of the braces is coherent with the development of a
weak storey. Since the dissipation of seismic input energy through
plastic deformation of all the braces is actually the desired
behaviour, the described degradation of one brace is not avoidable
in regular braced buildings. To prevent the degradation, buckling re-
strained braces may be used [10,11], but this otherwise more costly
solution is not in the scope of this article.
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 149

a) b)
Fig. 8. Global mechanism and realistic IDA curves of CBFs. (a) Global mechanism. (b) Weak storey with early failure.

As the degradation of the braces and the development of a weak sto-
rey are seemingly interrelated, the effect of the deformations on the CBF
response needs to be investigated.
6.2. A simplied model to simulate the effect of brace deformations on the
CBF's response
To evaluate the effect of the brace deterioration on the dynamic
characteristics of a CBF, a series of natural frequency analyses were con-
ducted on the elastic model of the four-storey braced frame (CBF41-
EC8). The perfect model with two straight bracing bars was altered. At
the midpoint of two braces on the same oor, different magnitudes of
imperfection were introduced. The midpoints of the diagonals were
pulled perpendicularly to the bar by 10 cm at each step up to 50 cm.
This resulted in triangular-shaped braces. The considered maximal al-
teration (50 cm) corresponds to a 2.76% relative storey drift ratio. The
triangular alteration of the diagonals has been done at every oor sepa-
rately to analyse the effect of developed weak storeys. Also the same im-
perfections have been applied on every storey simultaneously, which
corresponds to the development of the global plastic mechanism. Fur-
thermore, the natural frequencies of the perfect model with one diago-
nal on every oor have been computed, neglecting the buckled bar in
compression, accordingly to Eurocode 8 provisions. In Table 3, the fre-
quencies of the modes of vibration for the more relevant rst and sec-
ond modes are listed. The rows of the table correspond to the different
levels of the introduced imperfection while the columns differ in the
storey of the imperfection. For the sake of comparison, the modal
frequencies of the perfect model with one brace are 0.84 Hz and
2.26 Hz, respectively. One can see in the results of the table that with
the concurrent increase of the imperfection of all the braces the de-
crease of the frequencies in the two modes is roughly proportional.
However, this is not the case when the triangular modication is done
to the braces of one storey only. For example, the growing imperfection
on the rst oor halves the frequency of the rst mode but barely
changes the second. Consequently, it is not only the frequency of the
modes that changes.
The introduced imperfection of the braces modies the stiffness ma-
trix of the structure. It decreases the stiffness values corresponding to
the end nodes of the altered braces only. Therefore, the new stiffness
matrix of the structure is not proportional to the original while the
mass matrix remains the same. Thus, the mode shapes have to change
also. In Fig. 12, the modal frequencies and shapes and the modal mass
participation (MMP) are indicated. These results are presented for the
perfect four-storey building and two imperfect ones. Fig. 12.b corre-
sponds to a concurrent 40 cm imperfection of every brace, while
Fig. 12.c depicts the case where the same imperfection is introduced
only on the 4th storey. The mode shapes of Fig. 12a and b are seemingly
similar. Thus, the modal analysis of the perfect model well describes the
response of a CBF where every oor participates in the dissipation. In
Fig. 12.c, the mode shapes of the building with a developed weak storey
show that the rst mode is approximately the movement of the oor
above the weak one. Conversely, the second mode primarily consists
of the motion of the stories below the weak storey. Visibly, the existence
of a oor with a low lateral stiffness decouples the relevant modes of the

4.5 4.5
St4 St6
Top reinforcement!
Interstorey drift ratio [%]

Top reinforcement!
Interstorey drift ratio [%]

4.0 St3 4.0 St5

St2 St4
3.5 3.5
St1 St3
3.0 3.0 St2
St1
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
a) b)
Fig. 9. IDA diagrams of reinforced CBFs. (a) CBF41-EC8. (b) CBF61-EC8.
150 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

1.2
tension Table 3
1 Natural frequencies of imperfect 4-storey CBFs

0.8 Imperfection Natural frequency of mode 1 Natural frequency of mode 2

magnitude [cm] [Hz] [Hz]
0.6
decrease of
stiffness Storey of imperfection Storey of imperfection
0.4
N/Npl

1 2 3 4 All 1 2 3 4 All
0.2 10 0.95 0.99 1.03 1.09 0.69 2.75 3.12 2.95 2.36 1.90
20 0.67 0.87 0.78 0.82 0.41 2.52 3.07 2.71 1.90 1.13
0
30 0.54 0.76 0.65 0.64 0.29 2.46 3.03 2.62 1.79 0.82
-0.2 40 0.47 0.59 0.58 0.53 0.22 2.43 2.99 2.58 1.75 0.65
decrease of buckling load
compression 50 0.43 0.57 0.54 0.46 0.18 2.42 2.98 2.56 1.73 0.56
-0.4
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10
Axial deformation []
of the diagonals. The initial localisation of plasticity on one or very few
Fig. 10. Characteristic normalised axial forcedeformation diagram of brace. oors can be caused by the overstrength relations of the oors [1], higher
mode effects [33], or random effects of the seismic excitation [34], but in
the cluttered response of a building, it is more likely than the simulta-
building to modes that describe the motion of the oors above or below neous plastic deformation taking place everywhere.
the weak storey and these modes substantially differ from the originals The triangular deformation devolves the lateral loads on the col-
of the perfect model. It needs to be pointed out, that the frequency of the umns. The columns counteract the loading by bending, insofar as they
rst imperfect mode is signicantly smaller than of the second and are continuous on at least one end. The bending develops even as a re-
upper modes; therefore, the displacements associated with this mode sult of the brace deterioration, not only in a storey plastic collapse mech-
are dominating in the response. anism. Thus, the braced frame stops being a perfect truss, developing
As it can be seen, the degradation of the braces results in the abrupt only axial forces; therefore, continuous columns are indispensable in a
loss of lateral stiffness. If the loss has a similar amplitude on every sto- CBF. The brace deformation also causes a crucial drop of the storey
rey, then the CBF just becomes more exible; the general appearance shear stiffness. The resulting irregularity of the horizontal stiffness
of the response to excitation does not change signicantly. However, if along the building height modies the global dynamic behaviour of
the stiffness decrease is localised on a weak storey, the response of the the building and facilitates the localisation of large relative drifts on
CBF becomes as it is depicted by Fig. 13 (for a hypothetical case). This the storey under consideration. With the large drifts of the storey, plas-
displacement pattern demonstrates the localisation of the displace- tic deformations cumulate in the hinges at the middle of the braces, the
ments on the weak storey. This localisation may lead to even further bars deteriorate, the resistance of the braces decrease, and further plas-
degradation of the braces, decrease of the stiffness, and consequently, tic yield is facilitated.
to the gradually amplied development of the weak storey phenome- This process may be a recurring loop gradually increasing the dam-
non in repeated loops as explained in the next section. age of the weak storey. Hence, the development of the weak storey
mechanism can be regarded as auto amplifying phenomenon in case
6.3. Repeated process resulting in the weak storey behaviour of cyclic seismic action. The loop is broken by the collapse of the weak
storey. This can happen either by not having adequate columns on,
As the displacement response of a CBF with deteriorated braces on a above, and below the weak storey or by the fatigue failure of the braces
weak storey may lead to further plastic deformations, the process of the and gussets. The presented loop evidently leads to failure if a sufciently
development of the weak storey behaviour can be ascertained. Fig. 14 large number of cycles are exerted. Yet as the number of cycles during a
depicts an organigram of this process. The occurrence of plastic yield of seismic action is limited, by preventing excess brace deformations and
the braces on one oor results in the permanent triangular deformation providing exural column resistance and stiffness by choosing

crack
initialization
plate
buckling
fatigue
fracture

Fig. 11. Inelastic deformations of a CBF due to seismic action.

 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 151

M1 f=0,84 Hz M2 f=2,26 Hz M1 f=0,22 Hz M2 f=0,65 Hz M1 f=0,53 Hz M2 f=1,75 Hz

MMP=0,83 MMP=0,12 MMP=0,85 MMP=0,10 MMP=0,40 MMP=0,50
1,000 -1,000 1,000 -1,000 1,000 -0,229

0,755 0,364 0,787 0,442 0,152 1,000

0,510 0,955 0,532 0,907 0,087 0,743

0,245 0,705 0,277 0,628 0,043 0,371

a) b) c)
Fig. 12. Mode shapes of perfect and imperfect 4-storey models. (a) Perfect model, (b) concurrent 40 cm imperfection, (c) 40 cm imperfection on 4th oor.

adequately large cross sections, the gradual development of the weak
storey may be restrained.
Upon the description of the behaviour and the response of CBFs to
seismic actions presented above, an effective design method that will
be elaborated in the following should
- take into consideration the change of the modal behaviour due to
the brace deterioration and the resulting response of the CBF;
- quantify the participation of the columns and impose requirements
to their performance in the lateral resistance of a storey;
This method should prevent both the weak storey behaviour and the
weak storey collapse.
7. The robust seismic brace design method (RSBD)
The RSBD method proposed in this paper is not a stand-alone de-
sign method exclusively based on plastic analysis, which is in sharp
contrast with other methods proposed in the literature [3537].
The RSBD method, also based on the kinematic theorem of limit anal-
ysis, is an effective way to address the shortcomings of the elastic
analysis given in EC8 which, as shown in Section 5, fails to eliminate
weak storey mechanisms. Therefore, the RSBD method is rather a
complement to EC8 than an alternative design method as those pro-
posed in [3537]. In addition, the RSBD method always leads to a
strengthening of the seismic-resistant structure so that buckling of
columns is not anymore an issue since they have been checked in
the EC8 procedure.
The RSBD method imposes two requirements that are both based
on the kinematic theorem of plasticity. The rst requirement allows
to ensure that if a weak storey behaviour starts to develop at any sto-
rey, it will not fail before subsequently involving further storeys in
the dissipation. The second requirement tries to enforce equal
interstorey drifts with the aim to obtain a proportionally distributed
plastic dissipation, i.e., the global mechanism. It is worth mentioning
that the second requirement prevents also the occurrence of partial
plastic mechanisms involving several storeys. This is in contrast (sig-
nicant difference) with the TPMC method developed in [3537]. In
principle, these requirements should be formulated considering

Model Mode 1 Mode 2 Combination of modes

m4
k4
m3
k3
m2
k2
m1
k1

Fig. 13. Modal response of a CBF with a weak storey (modelled by triangular braces).
152 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
Fig. 14. Organigram of the weak storey behaviour of CBFs.
second-order geometrical effects. For the sake of clarity, these re-
quirements are developed below without second-order geometrical
effects. The second-order RSBD formulation is given in appendix B.
7.1. Criteria imposed on CBFs by the RSBD method
As Eurocode 8 criteria are based on the elastic response of the struc-
ture subjected to the reduced-intensity seismic action, they may not be
adapted in a random but probable inelastic response, such as the ap-
pearance of a weak storey. First, in an elastic analysis, due to the truss
action of the bracing, only negligible bending can be expected. Con-
versely, the columns in a weak storey response are subjected to signi-
cant bending due to the degradation of the braces, so the requirement
imposed on the axial resistance of the columns, Eq. (5), may not be sat-
isfactory to provide adequate column sections. In a weak storey re-
sponse, the axial forces of the braces may also be different from the
ones obtained by elastic analysis; therefore, the uniformity condition
of the storey overstrength factors, Eq. (1), may be violated.
Let us consider several kinematically admissible collapse mecha-
nisms of a CBF and compare them. According to theory, the mechanism
that can be realised is the one that corresponds to the smallest resis-
tance to lateral loading. The considered plastic mechanisms have to be
obviously the favourable global mechanism and the unfavourable
weak storey mechanisms. Partial mechanisms that involve several, but
not every storey in the plastic deformation may be regarded also. Con-
sequently, the lateral collapse load of the global mechanism and n-
storey mechanisms (in a n-storey CBF) have to be computed. In order
to take into account the change of the modal behaviour of a CBF, a lateral
loading that corresponds to the inelastic response shall be applied. It has
to be noted though, that if we consider the inelastic response of a CBF,
various load arrangements are possible. The deterioration of the braces
on the ith storey results in a response, where only the oors above the
weak storey sway with large amplitudes. The equivalent seismic load
distribution corresponding to this response involves concentrated
Fig. 15. Equivalent lateral loading of CBF with developed weak storey.
forces of equal magnitude (if the slab masses are equal) applied to the
slabs from the ith storey up to the top, see Fig. 15. In an n-storey CBF, n
different load distributions should be analysed each ranging from the
top down to a particular storey. For each load pattern, the load multiplier
corresponding to the formation of the global, glob, and the local, loc, plas-
tic mechanism can be computed, i.e., n multiplier pairs can be calculated.
If the local load multiplier is larger than the global, then according to the
kinematic principle, the occurrence of a storey collapse is prevented.
Thus, in order to secure that all the storeys have enough reserve so that
no localised collapse shall occur, the following criterion has to be met:
loc;i
1:0 where i 1::n 8
glob;i
Although the condition of Eq. (8) is necessary to prevent weak storey
collapse, it is not sufcient to keep the differences of the interstorey
drifts on all the oors within an acceptable range. Sufciently large sto-
rey resistance can be provided by enlarging either the brace or the col-
umn cross sections as they are both involved in the plastic resistance. If
in an extreme case the resistance is only given by the brace and the col-
umns are negligible (or hinged at both ends), the load corresponding to
the yield of the brace is equal to the ultimate plastic resistance of the
storey. As the method under consideration aims to prevent the attain-
ment of the storey resistance, the brace yield will not be reached either
and the storey remains elastic throughout the seismic action. Converse-
ly, if all the resistance is provided by the columns and the braces are
neglected, the lateral displacement of this certain oor will be a lot larg-
er than of other oors. The bending of the columns permits substantially
larger displacements than the stiff bracing. Such a oor has the tenden-
cy to exhibit the weak storey behaviour. The anticipated responses of
these two extreme cases are depicted in Fig. 16 along with the theoret-
ical behaviour of a regular well-designed oor. The two cases may be
satisfactory, satisfying all resistance design criteria, but certainly they
are both unfavourable because they impair the dissipative behaviour
of the CBF. In the former case, the oor is not exploited in the dissipa-
tion, whereas the latter case is too much involved by developing weak
storey behaviour.
Let us divide the local multiplier into two parts, i.e., the participation of
the brace(s) and the participation of the columns. Let us denote the for-
mer by br and the later by col. The diagrams of Fig. 17 depict the two ex-
treme cases where the load multiplier is plotted against the anticipated
lateral interstorey drift response of a storey. The thin lines depict the ini-
tial response of the storey, the thick lines are anticipated when the trian-
gular brace deterioration is present. When the resistance provided only
by the brace is high, i.e., br is close to glob (Fig. 17a), large inelastic defor-
mations are not attained, the brace deterioration is relatively small. On the
contrary, in the case of the diagram on Fig. 17b when signicant involve-
ment of the columns is needed to respect the rst criterion of Eq.(8), the
inelastic displacements and the deterioration are larger. The objective of a
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 153

obvious that if the BPR is close to 1.0, then the bracing is overdesigned,
column resistance is barely needed, and the storey is not involved in the
plastic dissipation. Therefore, it is recommended that the maximum BPR
shall not exceed 0.9. It has to be noted that the 0.1 and 0.9 constants
have been determined empirically, using numerous examples referred
to in this article:

BPR min 0:1BPR max 12

The two criteria of the RSBD method, Eqs. (8) and (12), will be used
to redesign the CBFs obtained by the Eurocode 8 procedure. The original
Fig. 16. Improper brace or column participation in storey resistance. cross sections obtained by EC8 should not be decreased.
The rst criterion does not require the denition of the amplitudes of
the equivalent seismic forces. This is a signicant advantage of the
second criterion is to prevent such inelastic displacement differences be-
method in comparison with other plasticity-based methods given in
tween storeys that otherwise satisfy the rst criterion.
the literature where the denition of these loads may be questionable.
Regarding the drift corresponding to the yield of the brace, dbr, if the
The second criterion tends to impose uniform interstorey drifts
yield stress is the same at every storey, then this drift also is closely the
without resorting to their explicit determination. This is accomplished
same constant on every storey. In order to promote a uniform response
by the Brace Performance Ratio parameter determined again by the
on every oor, it needs to be ensured, that the inelastic drift correspond-
use of the kinematical theorem. This was made possible by considering
ing to the attainment of the global mechanism, dglob, is also closely the
the gradual change of the dynamic behaviour of the structure due to the
same on every oor. Therefore, the following ratio has to be quasi con-
deterioration of the braces.
stant for all the storeys:

dbr;i 7.2. Determination of the parameters used in the RSBD method

const 9
dglob;i
The evaluation of the RSBD criteria requires the determination of the
load multipliers. This necessitates the computation of the internal and
Let us consider the dashed lines dening the secant stiffness of the in-
external works on the plastic yield mechanisms. In the following, the
elastic response. The ratio of the stiffness values, denoted by , describes
necessary formulae will be derived.
the relative stiffness loss resulted by the deterioration of the brace.
The parameters describing the geometry of a braced bay are present-
br;i glob;i ed in Fig. 18. The connections of the parameters related to the lateral
= i 10 displacement are approximated as follows:
dbr ; i dglob;i

As it has been described in Section 6.2, a signicant localised decrease B

sin 13
of the stiffness modies the modal behaviour in a way that promotes the H
development of the weak storey. Consequently, in order to prevent the
occurrence of the weak storey, it is required that shall also be the L
B 14
same on every storey. By combining Eqs. (9) and (10), the so-called cos
Brace Performance Ratio (BPR) can be expressed, which can be well ap-
proximated by the ratio of load multipliers, given that the left side of where is the angle between the brace and the horizontal direction. In a
the equation below is considered to be quasi constant on every storey. rst-order analysis, the vertical drift is considered to be equal to zero. In
case of large lateral displacement, the vertical drift cannot be neglected
br;i dbr;i anymore and a second-order analysis is required (see Appendix B).
BPRi i const: 11
glob;i dglob;i

As a second design criterion, additional to Eq. (8), it is required that B2

H 0 15
the maximum and minimum BPR do not differ by more than 0.1. It is 2H

a) b)
Fig. 17. Theoretical elasto-plastic load multiplier displacement curves. (a) High resistance provided by braces. (b) High resistance provided by columns.
154 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

where Ab is the area of the section, b is the breadth, and tf is the thickness
of the ange. If the criteria of Eq. (18) or Eq. (19) are not met, the mo-
ment resistance is to be decreased with a linear expression in case of
the y axis and a quadratic in case of the z:

1n
My;c Nc Mpl;y;c  20
10:5  a

na2 
Mz;c Nc Mpl;z;c  1 21

1a

For the sake of simplicity, in this investigation both interaction

curves are considered to be bilinear. In order to do so, an appropriately
chosen nlim needs to be determined, which is the limit axial force as be-
fore in Eq. (19). Let us compute the difference of the nonlinear interac-

tion curve, Eq. (21), and its bilinear approximation as the integral of the
square of the differences:
Fig. 18. Geometry model.

Z1 
na2  2
npl;c n
diff 1 dn 22
7.2.1. Internal work of the braces and the columns 1a npl;c n lim
0
In the calculation of the internal work of the braces, the compression
brace is neglected supposing that the plastic rotation of the hinge at the
mid length of the brace dissipates signicantly less energy in a plastic Dening nlim to make the above difference minimal gives an optimal
mechanism than the tensile yield of the brace in tension. Therefore, denition of the bilinear interaction curve. For most H-section columns
the internal work of the brace on the ith oor is used in the CBFs of the present research, the limit has been found to be
between 0.45 and 0.5. In Fig. 20, the interaction and bilinearised interac-
W t;br;i L Npl;br;i  Li Abr;i  f y  Li 16 tion curves are presented.
For bending about both axes, the interaction can be expressed in the
following form:
By substituting Eqs. (13) and (14) into Eq. (16), we get
8
W t;br;i N pl;br;i  cos i  H i  17 < M pl;c if Nc N lim
Mc Nc Npl;c Nc 23
: M if Nc NN lim
To determine the plastic work of the columns, the momentaxial Npl;c N lim pl;c
force interaction has to be considered as the deformation of the cross
section can be described by the axial shortening, N, and the rotation
(see Fig. 19): where Nlim is located at the intersection of the two solid lines
In Eurocode 3, the interaction of the axial force and the bending mo- (Fig. 20).
ment for both axes of an H-section is dened in Clause 6.2.9.1. For bend- The eccentricity of the neutral axis, e, indicated in Fig. 19, can be cal-
ing about the rst principal axis, y, the bending resistance is not culated as follows:
decreased by the axial force if:
8
0:5  hw  t w  f y < 0 if Nc N lim
N c 0:25 Npl;c and N c 18 eNc M pl;c 24
M0 : if Nc NN lim
Npl;c N lim
where Nc is the column axial force, tw is the thickness of the web, fy is the
yield stress (235 MPa), Npl,c is the axial plastic resistance, hw is the
height of the web of the section (h-2 tf), and M0 is a (safety) partial
factor taken equal to 1.0. For bending about the z axis, the resistance is 1
not decreased if 0.9
0.8
Nc Ac 2b  t f
n a with n and a 0:5 19 0.7
Npl;c Ac
Mc(Nc)/Mpl,c

0.6
0.5
0.4
0.3
My-N EC interaction
0.2 Mz-N EC interaction
0.1 bilinear Mz-N interaction
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Nc,/Npl,c

Fig. 19. Normal stress distributions and deformation diagram of column. Fig. 20. Interaction curves.
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 155

Nc,i+1,j My,c,i+1,j floor i

Nc,i,j My,c,i,j

column j
N M

My,c,i,j
Nc,i-1,j floor i-1
My,c,i-1,j
- + - +
Fig. 21. Internal forces and plastic hinges of columns in storey mechanism.

If we express the plastic work of the bent and axially loaded column
cross section as
W c Mc Nc Nc  eN c 25
and we combine it with Eq. (23) and Eq. (24) it is found that, as a result
of the bi-linearization, the work is not dependent on the axial force:
8 to be considered in the computation. The pointer vector, J, has only
< M pl;c  if N c N lim
W c Npl;c 26
: M  if Nc NN lim
Npl;c N lim pl;c
As it can be seen, the bilinear interaction curve simplies the strong
dependence of the plastic work on the axial force into two intervals. The
choice of the suitable expression from Eq. (26) still depends on the axial
force, but this may only lead to a two-step iterative calculation.
The internal work of the columns is considered only in storey mech-
anisms and obviously only on the storey of the mechanism itself. In a
building, all the columns, a total number of m, may participate in the lat-
eral resistance depending on the moment resistance of the column
splices. The peak values of the bending moment diagram, see Fig. 21,
are located at the oor slabs.
As the beams are usually connected to the columns, in a convention-
al CBF, by ordinary hinges, the moment peaks may form a plastic hinge
right above or under the oor slabs, where the cross sections may be dif-
ferent (although such conguration does not reect usual practice).
Furthermore, for the determination of the plastic moment resistance
the signicant axial forces in the columns have to be considered also.
As these increase at every oor slab, the resistance of the column
below a certain slab is not necessarily larger than the one above. Conse-
quently, attention is to be paid to appropriately select the smallest resis-
tance at both oor slabs of the analysed storey. For each column, a
pointer vector can be dened that identies which column's work has
zero or one values. The vector for the jth column on the ith oor, assum-
ing bending about the y axis, is
2 3
if Mc;i1; j bMc;i; j then 1; otherwise 0
6 if Mc;i1; j NMc;i; j then 1; otherwise 0 7
J i; j 6
4 if Mc;i1; j NM c;i; j then 1; otherwise 0 5
7 27
if Mc;i1; j bM c;i; j then 1; otherwise 0
It is important, that in the case of hinged column splices, the corre-
sponding entry are considered to be zero in the vector above. With
the pointer vector, the total internal work of a column j assigned to
the storey i is computed as follows:

W col;i; j W c;i1; j W c;i; j W c;i; j W c;i1; j
 J i; j
28

mpn. glob,i mpn. loc,i

mpi. glob,i mpi. loc,i

Fig. 22. External work of lateral loading and gravity forces.

156 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

This equation can be rewritten by factoring the rotation: Table 4b

CBF41-RSBD
h i Storey Int.col. Ext.col. Brace loc glob loc BPR
W col;i; j M sub
c;i1; j Msub
c;i; j M sub
c;i; j M sub
c;i1; j  J i; j 29
glob

4 HEA 160 HEA 200 90 6.3 1.43 578.7 568.3 1.02 0.86
where Msub is a moment resistance substituting the one given by 3 HEB 180 HEB 200 100 6.3 0.99 395.6 324.8 1.22 0.84
Eq. (23): 2 HEB 220 HEB 200 100 8 0.97 343.7 252.6 1.36 0.90
1 HEB 260 HEB 240 100 10 0.99 240.2 227.3 1.06 0.91
8
< M pl;c if N c N lim
Msub N pl;c 30
c
: M if Nc NN lim Fig. 22 depicts the relevant external forces of a CBF noting that mpi
Npl;c N lim pl;c
glob,i or mpi loc,i are the horizontal forces and Gi,j is the lumped gravity
loads. In the local mechanism, the lateral forces are equally displaced,
their work is
For every column, the product of the substituting moments and the
pointer vector can be interpreted as a total moment that has to be used X
n
W ;loc;i  H i  loc;i  mpk 34
for the determination of the plastic work:
ki

h i
Mcol;i; j M sub Msub Msub M sub In the global mechanism, the lateral displacement increases with the
c;i1; j c;i; j c;i; j c;i1; j  J i; j 31
height of the building. Therefore, for every concentrated loading in-
volved in the mechanism, the lateral displacement has to be calculated
individually. The external work can be computed as follows:
By substituting Eq. (31) into Eq. (29) and by summing up all the m
!
columns on a storey, the internal work of the columns in the storey X
n X
k

mechanism is expressed as a function of the lateral drift as follows: W ;glob;i  glob;i  mpk  Hl 35
ki l1

X
m
W col;i Mcol;i; j 
j1
7.2.2. External work of the lateral forces and gravity loads
In the RSBD method, the seismic loading is dened as horizontal
concentrated forces applied at the height of the oor slabs. The forces
are dened by an unknown parameter, denoted by , and to account
for the variability of the masses, it is multiplied by a reduced mass vec-
tor, mp. This vector is dened as the sum of the masses on a storey divid-
ed by the smallest of all the storey masses. If the mass is constant along
the height, the vector is an identity vector; otherwise, it shows the ratio
of the overweight of the particular oors.
32 The vertical shortening has been considered to be negligible in
Eq. (15) as the vertical displacement is innitesimally small for the
sake of the plastic limit analysis. Consequently, the second-order work
of the gravity loads is zero with this approximation, but in a storey
mechanism, the work on the plastic axial deformation has to be taken
into consideration. The axial deformation is (see Fig. 19):
N  eN c 36
The location of the plastic axial deformation is the same as the loca-
tion of the plastic hinge, which has been treated already by the use of a
pointer vector, see Eq. (27). Using this vector the axial deformation of
the column j on storey i is

X 
N;i; j ey;i1; j Nc ey;i; j Nc ey;i; j Nc ey;i1; j Nc  J i; j  37
mi; j
j
mpi 0 1 33
X By collecting the eccentricities as it has been done for the moments
min @ mi; j A in Eq. (31) and noting
i
j

ecol;i; j ey;i1; j Nc ey;i; j Nc ey;i; j Nc ey;i1; j Nc  J i; j 38

The loading is distributed from the top down to the oor under in- the work of the gravity forces above the considered storey on the
vestigation. In an n-storey-tall building n, different load patterns can axial deformation of the columns in a storey mechanism is
be dened. The i index appended to the load parameter refers to a dis-
Xn X
m 
tribution ranging from the ith storey up. According to the basic idea of W G;loc;i Gk;l  ecol;l 39
the RSBD method, the works are computed for global and local plastic ki l1
mechanisms, so for the n load arrangements, n load multiplier pairs,
glob,i and loc,i, have to be calculated.

Table 5a
CBF61-EC8
Table 4a
CBF41-EC8 Storey Int.col. Ext.col. Brace loc glob loc BPR

glob
Storey Int.col. Ext.col. Brace loc glob loc BPR
6 HEA 160 HEA 160 90 5 1.00 364.4 513.6 0.71 0.70
glob
5 HEB 180 HEA 180 100 6 0.98 265.0 280.1 0.95 0.81
4 HEA 160 HEA 140 100 4 1.05 342.5 471.6 0.73 0.70 4 HEB 220 HEB 200 100 8 1.04 245.7 205.4 1.20 0.91
3 HEB 180 HEA 200 100 6.3 0.99 291.8 269.5 1.08 0.91 3 HEB 240 HEB 220 100 10 1.04 230.7 171.2 1.35 0.96
2 HEB 220 HEB 200 100 8 0.97 255.5 209.6 1.22 0.95 2 HEB 280 HEB 240 100 10 1.01 197.7 154.1 1.28 0.90
1 HEB 260 HEB 240 100 10 0.99 193.1 188.7 1.02 0.95 1 HEB 300 HEB 280 120 10 1.00 148.8 146.7 1.01 0.90
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 157

Table 5b Table 6b
CBF61-RSBD CBF81-RSBD

Storey Int.col. Ext.col. Brace loc glob loc BPR Storey Int.col. Ext.col. Brace loc glob loc BPR

glob glob

6 HEA 160 HEA 200 100 6.3 1.42 632.3 652.1 0.97 0.84 8 HEA 160 HEA 220 100 8 1.74 800.2 798.7 1.00 0.76
5 HEB 180 HEB 200 100 8 1.17 439.7 355.7 1.24 0.90 7 HEB 180 HEB 200 100 8 1.21 479.7 426.0 1.13 0.71
4 HEB 220 HEB 200 100 8 1.00 300.2 260.8 1.15 0.82 6 HEB 220 HEB 200 100 10 1.22 336.3 304.3 1.11 0.75
3 HEB 240 HEB 220 100 10 1.02 257.8 217.4 1.19 0.88 5 HEB 240 HEB 220 100 10 1.11 269.7 245.7 1.10 0.75
2 HEB 280 HEB 260 100 10 0.96 239.3 195.6 1.22 0.84 4 HEB 280 HEB 240 120 10 1.17 244.5 213.0 1.15 0.80
1 HEB 300 HEB 280 120 10 0.97 188.0 186.3 1.01 0.85 3 HEB 300 HEB 280 120 10 1.07 219.3 193.6 1.13 0.73
2 HEM 240 HEB 320 140 10 1.08 226.3 182.6 1.24 0.80
1 HEM 240 HEM 240 140 10 0.98 182.4 177.5 1.03 0.72

7.2.3. Determination of the load multipliers

In order to express the load multipliers needed to evaluate the two
X
m Xn X
m 
criteria of the RSBD method we consider the equivalence of the external Npl;br;i  H i  cos i M col;i; j Gk;l  ecol;l
and the internal works in both the local and global mechanisms. In case j1 ki l1
loc;i 45
of the global mechanism the equation is as follows: X
n
Hi  mpk
X ki
W ;glob;i W t;br;i 40
i
The formula for the calculation of br, which is necessary for the de-
termination of the BPR, is deduced from Eq. (45), neglecting the partic-
Substituting Eqs. (17) and (35) into Eq. (40) we get: ipation of the columns and the gravity forces as those are also related to
the plastic hinges of the columns:
!
X
n X
k X
 glob;i  mpk  Hl Npl;br;i  H i  cos i 41 Npl;br;i  H i  cos i
br;i 46
ki l1 i X n
Hi  mpk
ki
In case of the local mechanism the energy equivalence is the follow-
ing: In classical plastic limit analysis, the parameter of the displacements,
hereby , is innitesimally small. In the seismic response of building
W ;loc;i W G;loc;i W t;br;i W col;i 42 structures, the drifts are moderate but hardly innitesimal. Consequent-
ly, the second-order work of the large gravity forces may substantially
affect the results that we could obtain with the limit state analysis. To
Substituting Eqs. (17), (32), (34), (39) into Eq. (42), we get take into account this effect, Eqs. (44)(46) were elaborated including
the second-order work of the gravity forces also (see Appendix B).
X
n Xn X
m  With this approach, the RSBD redesign yields slightly more economic
 H i  loc;i  mpk Gk;l  ecol;l
and better performing CBFs, but the difference is mostly not prominent.
ki ki l1
Xm 43 For further details see [38] and the appendix.
Npl;br;i   H i  cos i Mcol;i; j 
j1
7.3. Outline of the RSBD redesign procedure

From Eqs. (41) and (43), both the global and the local multipliers can The RSBD redesign method has to be conducted in the following
be expressed explicitly. However, it has to be noted, that for the deter- steps:
mination of the local multiplier a two-step iterative calculation may - Initial design of the structure according to related codes,
be needed as the column work depends on the axial force, which is af- i.e., Eurocode 3 and Eurocode 8 considering the response spectra, be-
fected by the lateral loading for certain columns. haviour factor, conducting an elastic analysis, and dimensioning the
X members according to its internal force results. In the seismic design,
N pl;br;i  H i  cos i the uniformity condition (Eq. (1)) may be neglected.
i
glob;i ! 44
X
n X
k
mpk  Hl
ki l1
Table 7a
CBF82-EC8
Table 6a
Storey Int.col. Ext.col. Brace loc glob loc BPR
CBF81-EC8
glob
Storey Int.col. Ext.col. Brace loc glob loc BPR
8 HEA 180 HEA 200 90 5 1.02 458.5 644.2 0.71 0.52
glob 7 HEB 180 HEB 180 100 6.3 0.99 370.3 343.6 1.08 0.66
8 HEA 160 HEA 160 90 5 0.99 402.0 609.5 0.66 0.53 6 HEB 220 HEB 200 100 8 0.99 310.5 245.4 1.27 0.70
7 HEB 180 HEA 180 100 6.3 1.05 296.3 325.1 0.91 0.67 5 HEB 260 HEB 240 100 10 1.03 306.0 198.2 1.54 0.77
6 HEB 220 HEB 200 100 8 1.01 245.0 232.2 1.05 0.71 4 HEB 280 HEB 260 120 10 1.13 308.6 171.8 1.80 0.88
5 HEB 240 HEB 220 100 10 1.10 229.0 187.5 1.22 0.77 3 HEB 300 HEB 300 120 10 1.03 282.0 156.2 1.81 0.79
4 HEB 280 HEB 240 120 10 1.01 217.5 162.5 1.34 0.88 2 HD320 HD320 140 10 1.08 306.0 147.2 2.08 0.86
3 HEB 300 HEB 280 120 10 0.98 192.3 147.8 1.30 0.78 158 158
2 HEM 240 HEB 320 140 10 0.98 210.8 139.3 1.51 0.85 1 HD360 HD320 140 10 1.01 198.7 143.2 1.39 0.76
1 HEM 240 HEM 240 140 10 1.00 147.2 135.5 1.09 0.74 179 158
158 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

Table 7b Table 8b
CBF82-RSBD CBF81M-RSBD

Storey Int.col. Ext.col. Brace loc glob loc BPR Storey Int.col. Ext.col. Brace loc glob loc BPR

glob glob

8 HEB 180 HEB 180 110 6.3 1.55 691.6 714.9 0.97 0.72 8 HEA 180 HEA 240 100 8 1.44 629.3 591.9 1.06 0.76
7 HEB 180 HEB 180 100 8 1.21 443.0 386.6 1.15 0.74 7 HEB 200 HEA 240 110 8 1.20 402.4 315.7 1.27 0.81
6 HEB 220 HEB 200 100 10 1.24 365.9 276.1 1.33 0.82 6 HEB 240 HEB 200 100 10 1.11 256.7 225.5 1.14 0.79
5 HEB 260 HEB 240 100 10 1.12 319.6 223.0 1.43 0.74 5 HEB 260 HEB 240 140 7.1 1.02 209.9 182.1 1.15 0.74
4 HEB 280 HEB 260 120 10 1.18 316.6 193.3 1.64 0.82 4 HEB 300 HEB 280 120 10 1.02 209.7 161.4 1.30 0.75
3 HEB 300 HEB 300 120 10 1.07 288.7 175.7 1.64 0.74 3 HEB 320 HEB 320 140 10 1.09 227.1 148.7 1.53 0.83
2 HD 320 HD 320 140 10 1.10 306.0 165.7 1.85 0.76 2 HEM 240 HEM 240 140 10 1.00 211.5 141.3 1.50 0.74
158 158 1 HEM 260 HEM 260 150 10 1.00 150.0 137.9 1.09 0.73
1 HD 360 HD 320 140 10 1.00 203.7 161.1 1.26 0.71
179 158

- For n different loading distributions, considering the global and n
local plastic mechanisms computation of n glob,i and loc,i parameter
pairs. Computation of the br,i load multipliers of the braces for n
local mechanisms.
- Strengthening of the braces to respect the second criterion of the
RSBD method. Attention is to be paid that the modication of one
brace does not only change the corresponding local load multiplier
and the Brace Participation Ratio of that oor, but also all the global
multipliers and all other BPRi in the same time.
- Strengthening of the columns or introduction of more continuous
columns in the building to satisfy the rst condition.
The procedure is by nature iterative and requires usually 23 or a
few iterations to converge to an efcient seismic-resistant design. In
the case of very different BPRi, signicant strengthening of the braces
is needed. Due to the explicit and linear nature of the equations in the
rst-order analysis, the redesign can be carried out by hand calculations.
7.4. Performance of reinforced CBFs
In the following, both the original Eurocode 8 designs and the rede-
signs will be presented in tables in conjunction with the IDA diagrams
that describe their seismic performance. In the tables, the cross section
of the internal and the external columns and the braces are listed. This
is followed by the storey overstrength factors, , the load multipliers
of the local and the global mechanisms, the ratio of this two and the
BPR results. In the Tables 4 to 9, the results that substantially differ
from the corresponding requirements are highlighted in bold letters.
Also, the reinforced cross sections and the maximum and minimum
BPR are with bold letters in the tables.
Looking at the tables, it can be seen that both the relation of the load
multipliers and the BPR results remarkably indicate the existence of the
weak storeys right where they can be observed in the diagrams of Fig. 7.
The reinforcement is mostly necessitated on the upper oors. As the re-
inforcement of the braces of any oor increases the global load multipli-
er, during the redesign process, the need of reinforcing oors below the
Table 8a
CBF81M-EC8
ones that were found to be weak in the rst (a) tables may arise. This
shows that with the RSBD method, cases where the reinforcement
causes the weak storey occurrence to shift to another storey can be suc-
cessfully prevented. The nal reinforced designs comply (or very closely
comply) with the RSBD requirements of Eqs. (8) and (12). On the other
hand, they strongly violate the uniformity condition given in Eq. (1).
Nevertheless, the response of every building provides lot slimmer fam-
ilies of curves than of the original designs, see Figs. 23 and 7. This indi-
cates that a favourable, well-distributed dissipative behaviour is
realised with the redesign. These RSBD designs are also adequate,
early failure is not observed; in fact, the buildings tend to be a lot better
performing than necessary, reaching failure at elevated scale factors.
The failure typically occurs on the rst and second oors in the form
of rapid increase of the storey drifts. This, however, is not weak storey
collapse in the terminology of the present article as it is not preceded
by weak storey behaviour. It is a regular load bearing capacity issue.
It can also be observed that the RSBD method equally enhances the
performance of type 1 (CBF-CBF) and type 2 (CBF-MRF) CBFs and of
the building with mass irregularity too. Apart from these detailed exam-
ples, similar results were obtained for every other building listed in
Table 1. In the following, a comprehensive summarised evaluation on
the efciency of the RSBD method, considering every CBF, will be given.
7.5. Evaluation of the efciency of the RSBD method
So far the IDA diagrams of the original EC8 designs have been found
to be signicantly broader than the ones of the redesigns. This broad-
ness can be described numerically for the sake of a better comparison.
It is appealing to consider the difference between the maxima and the
minima of the IDR results, but it may also be necessary to distinguish
cases where the curves scatter out like a fan and when only one agrant
curve results in the large difference. The IDR results of the oors for a
given scale factor can be considered as values of a discrete random var-
iable, the scatter of which can be described by the standard deviation
(hereinafter ). As in a low scale factor range, the broadness may not
be developed, and in a high range, the results may be distorted by the
Table 9a
CBF101-EC8
Storey Int.col. Ext.col. Brace loc glob loc BPR

glob
Storey Int.col. Ext.col. Brace loc glob loc BPR
10 HEA 160 HEA 160 100 5 0.98 425.2 793.0 0.54 0.45
glob
9 HEB 180 HEA 180 100 8 1.00 332.7 417.4 0.80 0.62
8 HEA 180 HEA 160 100 5 1.00 365.9 560.7 0.65 0.54 8 HEB 220 HEB 200 100 10 1.00 281.2 293.7 0.96 0.68
7 HEB 200 HEA 200 100 8 1.08 301.9 299.0 1.01 0.76 7 HEB 240 HEB 220 120 10 1.11 279.1 233.2 1.20 0.84
6 HEB 240 HEB 200 100 10 1.03 231.7 213.6 1.08 0.77 6 HEB 280 HEB 260 120 10 1.02 235.5 198.2 1.19 0.76
5 HEB 260 HEB 240 140 7.1 1.02 209.9 172.5 1.22 0.78 5 HEB 300 HEB 280 140 10 1.07 233.1 176.2 1.32 0.85
4 HEB 300 HEB 280 120 10 1.02 209.7 152.9 1.37 0.79 4 HEM 240 HEB 320 140 10 1.01 209.5 161.8 1.29 0.77
3 HEB 320 HEB 320 140 10 1.09 227.1 140.9 1.61 0.88 3 HEM 240 HEM 240 140 12.5 1.09 224.4 152.5 1.47 0.88
2 HEM 240 HEM 240 140 10 1.00 211.5 133.9 1.58 0.78 2 HEM 260 HEM 260 140 12.5 1.03 203.3 146.9 1.38 0.79
1 HEM 260 HEM 260 150 10 1.00 150.0 130.6 1.15 0.77 1 HEM 280 HD320 198 150 12.5 0.98 142.3 144.2 0.99 0.70
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 159

Table 9b
CBF101-RSBD

Storey Int.col. Ext.col. Brace loc glob loc BPR

glob

10 HEA 160 HEB 220 100 10 1.94 1000.5 986.5 1.01 0.75
9 HEB 180 HEB 220 100 10 1.27 580.1 519.2 1.12 0.71
8 HEB 220 HEB 200 110 10 1.19 380.9 365.4 1.04 0.75
7 HEB 240 HEB 220 120 10 1.06 296.1 290.2 1.02 0.73
6 HEB 280 HEB 260 120 10 1.02 262.6 246.6 1.06 0.72
5 HEB 300 HEB 280 140 10 1.03 253.4 219.2 1.16 0.78
4
3
HEM 240
HEM 240
HEB 320
HEM 240
140 10
150 10
0.99
0.98
230.9
231.0
201.3
189.7
1.15
1.22
0.72
0.74

2 HEM 260 HEM 260 140 12.5 1.03 223.2 182.7 1.22 0.78
1 HEM 280 HD320 198 150 12.5 1.00 184.0 179.4 1.03 0.76

Fig. 24. Determination of the broadness and robustness coefcients.

4.5 4.5
St4 St6
CBF41-RSBD CBF61-RSBD
Interstorey drift ratio [%]

Interstorey drift ratio [%]

4.0 St3 4.0 St5
St2 St4
3.5 3.5 St3
St1
St2
3.0 3.0 St1
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
a) b)
4.5 4.5
St8 St8
CBF81-RSBD CBF82-RSBD
Interstorey drift ratio [%]

St7
Interstorey drift ratio [%]

4.0 St7 4.0

St6 St6
3.5 St5 3.5 St5
St4 St4
3.0 3.0 St3
St3
2.5 St2 2.5 St2
St1 St1
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
c) d)
4.5 St8 4.5
St10
CBF81M-RSBD CBF101-RSBD
Interstorey drift ratio [%]

St7
Interstorey drift ratio [%]

4.0 4.0 St9

St6 St8
3.5 St5 3.5 St7
St4 St6
3.0 St3 3.0 St5
2.5 St2 2.5 St4
St1 St3
2.0 2.0 St2
St1
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Scale factor of PGA Scale factor of PGA
e) f)
Fig. 23. IDA diagrams of reinforced CBFs (a) CBF41-RSBD, (b) CBF61-RSBD, (c) CBF81-RSBD, (d) CBF82-RSBD, (e) CBF81M-RSBD, (f) CBF101-RSBD.
160 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
Broadness Coefficient
1.40%
EC8
1.20% RSBD
1.00%
0.80%
0.60%
0.40%
0.20%
0.00%
41 61 81 101 42 62 82 102 61M 81M
a)
Fig. 25. Broadness and robustness coefcients of various CBFs. (a) Broadness coefcient. (b) Robustness coefcient.
asymptotic curves corresponding to attained collapse; this analysis is
conned to the 0.751.25 scale factor interval. To represent the broad-
ness of the numerous IDA curves, the mean of all the standard devia-
tions in this interval is computed. This way the broadness is expressed
by one number, hereinafter referred to as the broadness coefcient
that is in IDR unit, see Fig. 24.
An important characteristic of the RSBD method is that it provides
CBF designs that reliably resist earthquakes within their design intensity
range without sensitivity to any particularity of the acceleration record.
This can be demonstrated by quantifying the sensitivity, or in other
words, the robustness of the designs. To describe the robustness of the
obtained results, for a given scale factor, the scatter of the broadness
for every accelerogram shall be taken. If this scatter is small, so the broad-
ness is the same for every earthquake, then the response of the examined
CBF is not sensitive to variations of the excitation, and it is robust. The
broadness has been calculated with the standard deviation before and
to compute the scatter of these results; again the of these standard de-
viations is taken for every considered scale factor. The mean of the latter
is another number in IDR unit describing the robustness of the CBF
seismic response; therefore, it is called robustness coefcient.
Fig. A.1. GiuffrMenegottoPinto constitutive law.
Robustness Coefficient
0.50%
EC8
0.40% RSBD
0.30%
0.20%
0.10%
0.00%
41 61 81 101 42 62 82 102 61M 81M
b)
In Fig. 25, the broadness and robustness coefcients obtained both
for EC8 and RSBD design are presented in two bar charts. As it can be
seen, the broadness of the Eurocode 8 designs increases with the build-
ing height, whereas with the RSBD method the broadness is kept in an
interval of 0.30.5% IDR for every CBF. Also, the EC 8 designs can barely
show such a good performance; the broadness of the EC8 designs most-
ly exceed 0.6% and may go beyond substantial 1.2% IDR. The robustness
chart shows that the EC8 designs do not only have broader IDA curves
but they are also more sensitive to the particularities of the excitation.
The robustness of the EC8 originals is at least 0.20%, but it goes a lot
higher up in some cases, while all the RSBD redesigns stay in the 0.1
0.16% range. This demonstrates that the RSBD method indeed reliably
provides seismic-resistant CBFs. This reliability compelled the authors
to give the developed method the Robust Seismic Brace Design name.
8. Conclusions
The objective of this article was to investigate the susceptibility of
Eurocode 8 CBF designs to exhibit weak storey behaviour when subject-
ed to seismic action, to determine the reasons of this behaviour, and to
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163
provide a solution in design that can prevent the occurrence of storey
and partial mechanisms. The introduction of the structure-specic
rules of EC8 has been followed by the detailed description of CBF designs
and their models for nonlinear dynamic analyses.
The performance of the designs has been evaluated with the use of
IDA curves. While investigating the reasons of the weak storey behav-
iour, its development, along with the fundamental change of the
modal behaviour and the involvement in the lateral resistance of the
columns, has been directly related to the plastic deterioration of the
braces. The experiences of this rst half of the article led to the recogni-
tion of the factors that contribute to the development of weak storeys
and some weaknesses of the EC8 requirements. These are summarised
briey below:
- The EC8 CBFs are mostly prone to exhibit weak storey behaviour.
- The likelihood of the weak storey behaviour increases with the sto-
rey number.
- The susceptibility to exhibit weak storey behaviour can be analysed
independently of the design seismic intensity as the magnitude of
the weak storey phenomenon is not necessarily proportional to the
intensity.
- Attempted reinforcement of the buildings without accurate analysis
may not yield better seismic performance. This shows also that the
weak storey behaviour is not simply the lack of resistance on some
oors.
- The brace deterioration causes a signicant drop of the lateral stiff-
ness of the storey that it is located on, and this fundamentally affects
the response of the structure. The modied modal behaviour con-
tributes to the development of the weak storey; such effects are
not considered in any way in Eurocode 8.
- The interconnected plastic deformations, brace deterioration, and
change of dynamic response together create a loop that may amplify
the weak storey behaviour.
- The incipient elastic behaviour that the Eurocode 8 design procedure
refers to is signicantly different from the inelastic behaviour.
- The uniformity condition, Eq. (1), is not adequate to promote the
global dissipation as it cannot be in any inelastic state that the struc-
ture may undergo during the seismic excitation.
- The analysis with a perfect structural model leads to the underesti-
mation of the bending of the columns that develops with the deteri-
oration of the braces.
The observations above have been considered in the development of
the new Robust Seismic Brace Design method. The method gives two
criteria that are intended to replace the existing inadequate require-
ments in Eurocode 8. As the method is independent of the seismic inten-
sity, it needs to be conducted along with the basic EC8 provisions. The
rst requirement of the redesign method, imposed on the global and
local load multipliers, aims to avoid the storey collapse before realising
plastic deformations in all designated dissipative members, i.e., the
braces. The requirements concerning the Brace Performance Ratios pro-
vide a well-balanced distribution of the dissipation so that prominent
interstorey drifts do not impair the seismic performance of the CBF.
The presented buildings, that satisfy the two requirements of the
RSBD method but violate the uniformity condition (1) of Eurocode 8,
are adequate and exhibit a favourable dissipation without exception.
In the redesigned CBFs, weak storeys are not observed, and early col-
lapse is not attained; in fact, most CBFs show a signicantly higher resis-
tance than the demand. The reinforced buildings are also robust, not
showing sensitivity to the differences between acceleration records of
the same spectrum. The RSBD method is also capable of providing ro-
bust CBF designs independently of the examined storey numbers, struc-
tural models, and mass irregularities. Therefore, the use of the RSBD
method in seismic design is recommended to reinforce and enhance
concentrically braced steel frames.
161
Although the RSBD method is suitable for simple hand calculation,
its formulae, especially in the second-order analysis form presented in
the appendix, may be considered a bit complex. Therefore, in order to
introduce the method in a seismic design code utilised by practitioners,
it needs further simplication. Such simplication is possible and to
some extent already elaborated in [38].
Appendix A. GiuffrMenegottoPinto constitutive law
For the appropriate description of the real dissipative behaviour,
a constitutive model that accurately represents the cyclic dissipation
shall be used. For steel members in a one-dimensional stressstrain
state, the use of the GiuffrMenegottoPinto law is in common
practice. This model provides a smooth, curved transition between
the elastic and plastic branches. Furthermore, kinematic hardening
and the Bauschinger effect are taken into consideration. The rela-
tionship between the stress and the deformation is written as
follows:
1b
b
1=R A:1
1 R
r
A:2
0 r
r
A:3
0 r
0 and 0 dene the point which is the intersection of the tangent of
the initial elastic branch and the tangent of the asymptotic branch de-
scribing the yield, see Fig. A.1. r and r dene the point of the last
load reversal. The inclination of the hardening branch is dened by
the ratio of the initial (E0) and the hardening (Eh) modulus of elasticity:
Eh
b A:4
E0
The curved transition between the two branches is dened by R in
Eq. (A.1). In case of cyclic response, after each reversal the curvature re-
duces with the previous plastic excursion, , according to the following
expressions:
A1 n
Rn R0 A:5
A2 n
r;n r;n1
r;n r;n1
E0
n A:6
0;n r;n
where R0 is the value of the curvature in the rst loading, A1 and A2
are material dependent parameters. The applied set of material param-
eters is [15]: E0 = 210 GPa, Eh = 2.1 GPa, fy = 235 MPa, = 0.3, R1 = 20,
A1 = 18.5, A2 = 0.15, where fy and denote the yield stress and the
Poisson's ratio, respectively.
Appendix B. Second-order RSBD formulation
In order to take into account the second-order work of the gravity
forces on the vertical shortening, assuming that the lateral drifts are
not innitesimal, Eq. (15) is modied as follows:
B2 2
H H A:7
2H 2
It is also assumed that in the considered moderate, but not innites-
imal range of lateral displacements the principle of constant stresses is
applicable. Consequently, the internal forces are constant. To include
162 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163

the second-order effects, Eq. (39), the work of the gravity forces need to [2] G. Brandonisio, M. Toreno, E. Grande, E. Mele, A. De Luca, Seismic design of concen-
tric braced frames, J. Constr. Steel Res. 78 (2012) 2237, http://dx.doi.org/10.1016/j.
be modied as: jcsr.2012.06.003.
[3] G.A. MacRae, Y. Kimura, C. Roeder, Effect of column stiffness on braced frame seis-
X n X m    mic behaviour, J. Struct. Eng. 130 (No. 3) (2004)http://dx.doi.org/10.1061/

W G;loc;i Gk;l  ecol;l Hi A:8 (ASCE)0733-9445(2004)130:3(381) (pp. 381-39).
ki l1
2 [4] EN 1998-1:2004 Eurocode 8: Design of structures for earthquake resistancePart 1:
general rules, seismic actions and rules for buildings
[5] R. Tremblay, Achieving a stable inelastic seismic response for multi-storey concen-
In the global plastic mechanism, every gravity force has to be taken trically braced steel frames, Eng. J. (2003) 111129 (second quarter).
[6] Martinelli L., Perotti F., Bozzi A.: Seismic design and response of a 14-storey concen-
into consideration. The vertical displacements, coming only from the
trically braced steel building. STESSA 2000 Conference, Montreal, Canada, Mazzolani
second-order shortening this time, are the same on a oor, but with F., Tremblay R. eds., pp. 327-334.
the height they increase. The work of the gravity forces are computed [7] A.Y. Elghazouli, Seismic design of steel-framed structures to Eurocode 8, The 14th
World Conference on Earthquake Engineering, 2008 (Beijing, China).
with the formula below: [8] A. Longo, R. Montuori, V. Piluso, Failure mode control of X-braced frames under seis-
mic actions, J. Earthq. Eng. 12 (2008) 728759, http://dx.doi.org/10.1080/
" # " # 13632460701572955.
X
n Xm X
i X
n Xm X
i
2 [9] M.T. Giugliano, A. Longo, R. Montuori, V. Piluso, Plastic design of CB-frames with re-
W G;glob;i Gk;l  H k Gk;l  H k A:9 duced section solution for bracing members, J. Constr. Steel Res. 66 (2010) 611621,
k1 l1 k1 k1 l1 k1
2
http://dx.doi.org/10.1016/j.csr.2010.01.001.
[10] A. Watanabe, Y. Hitomi, E. Saeki, A. Wada, M. Fujimoto, Properties of brace encased
in buckling-restraining concrete and steel tube, Proc. 9th World Conf. Earthq. Eng. 4
To determine the parameters, Eqs. (40)(43) have to be amended (1988) 719724.
according to Eqs. (A.8), (A.9). [11] . Zsarnczay, Experimental and Numerical Investigation of Buckling Restrained
Braced Frames for Eurocode Conform Design Procedure Development(PhD Disser-
X tation) Department of Structural Engineering, Budapest University of Technology
W ;glob;i W G;glob;i W t;br;i A:10 and Economics, 2013.
i [12] R. Sause, J.M. Ricles, D. Roke, C.-Y. Seo, K.-S. Lee, Design of self-centering steel
concentrically-braced frames, Proc. 4th International Conference on Earthquake En-
! " # gineering, Taipei, Taiwan, paper no. 122, 2006.
X
n X
k X
n X
m X
i
2 [13] A. Braconi, et al., OPUS Project Final Report. European Union, RFCS Program, 2011.
 glob;i  mpk  Hl Gk;l  Hk [14] V. De Ville, FINELG: nonlinear nite element analysis program, User's Manual Ver-
2
X ki l1 k1 l1 k1 sion 9.0, Ulg et BEG, 2003.
Npl;br;i  H i  cos i A:11 [15] M. Menegotto, P. Pinto, Method of analysis for cyclically loaded reinforced concrete
i
plane frames including changes in geometry and nonelastic behaviour of elements
under combined normal force and bending, IABSE Symposium on Resistance and Ul-
timate Deformability of Structures Acted on by Well-Dened Repeated Loads,
W ;loc;i W G;loc;i W t;br;i W col;i A:12 Lisbon, 1973.
[16] EN 1993-1-1:2005 Eurocode 3: Design of steel structuresPart 1-1: General rules
X
n Xn X m    and rules for buildings
[17] Wilson E. L.: Three-Dimensional Static and Dynamic Analysis of Structures, Third
 H i  loc;i  mpk Gk;l  ecol;l Hi edition, Computers and Structures, ISBN 0-923907-00-9
ki ki l1
2
Xm A:13 [18] D.J. Dowrick, Earthquake Resistant Design for Engineers and Architects, second ed.
Wiley & Sons, Chichester, New York, 1987.
Npl;br;i   H i  cos i Mcol;i; j  [19] G.R. Black, B.A. Wenger, E.P. Popov, Inelastic buckling of steel struts under cyclic load
j1 reversals, Research report no. UCB/EERC-80/40, Earthquake Engineering Research
Center, Berkeley, CA, 1980.
[20] W.R. Walpole, Behaviour of cold-formed steel RHS members under cyclic loading,
By the rearrangement of the above formulae, the load parameters Research report no. 96-4, University of Canterbury, New Zealand, 1996.
can be expressed explicitly. The expressions below cannot be simplied [21] D. Vamvatsikos, C.A. Cornell, Incremental Dynamic Analysis, Earthq. Eng. Struct.
Dyn. 31 (3) (2002) 491514.
by the parameter of the deformation, . Therefore, this parameter has [22] V.A. Zayas, E.P. Popov, S.A. Mahin, Cyclic inelastic buckling of tubular steel braces,
to be chosen to obtain comparable values of the load parameters. In the Research report no. UCB/EERC-80/16, Earthquake Engineering Research Center,
conducted analyses, the rotation has been considered to be 0.02, which Berkeley, CA, 1980.
[23] T.Y. Yang, J.P. Moehle, B. Stojadinovic, Performance evaluation of innovative steel
is equivalent to a 2% relative drift. Such analysis provides somewhat bet- braced frames, Research report no. PEER 2009/103, Pacic Earthquake Research
ter dissipative behaviour of the reinforced CBFs, see [38]. Center, University of California, Berkeley, 2009.
[24] B. Shaback, T. Brown, Behaviour of square hollow structural steel braces with end
" # connections under reversed cyclic axial loading, Can. J. Civ. Eng. 30 (2003)
X X n X
m X
i
2 745754, http://dx.doi.org/10.1139/L03-028.
Npl;br;i  H i  cos i Gk;l  Hk [25] R. Tremblay, M.-H. Archambault, A. Filiatrault, Seismic response of concentrically
i k1 l1
2
glob;i ! k1 A:14 braced steel frames made with rectangular hollow bracing members, J. Struct.
Xn X
k Eng. 129 (No. 12) (2003) 16261636, http://dx.doi.org/10.1061/(ASCE)0733-
mpk  Hl 9445(2003)129:12(1626).
ki l1 [26] J.M. Goggins, B.M. Broderick, A.Y. Elghazouli, A.S. Lucas, Behaviour of tubular steel
members under cyclic axial loading, J. Constr. Steel Res. 62 (2005) 121131,
m    http://dx.doi.org/10.1016/j.jcsr.2005.04.012.
X
m X
n X
[27] K.H. Nip, L. Gardner, A.Y. Elghazouli, Cyclic testing and numerical modelling of car-
Npl;br;i   H i  cos i Mcol;i; j  Gk;l  ecol;l H
j1 ki l1
2 i bon steel and stainless steel tubular bracing members, Eng. Struct. 32 (2010)
loc;i 424441, http://dx.doi.org/10.1016/j.engstruct.2009.10.005.
X
n
[28] V. Gioncu, M. Mazzolani, Ductility of Seismic Resistant Steel Structures, Spon Press,
 Hi  mpk
London, 2002 (ISBN 0-419-22550-1).
ki
[29] V. Gioncu, D. Petcu, Available rotation capacity of wide-ange beams and beam-
A:15 columns, J. Constr. Steel Res. 43 (1997) 161217.
[30] D.E. Lehman, C.W. Roeder, D. Herman, S. Johnson, B. Kotulka, Improved seismic
X n X m   performance of gusset plate connections, J. Struct. Eng. 134 (No. 6) (2008)
2 890901.
Npl;br;i   H i  cos i Gk;l  Hi
2 [31] C.W. Roeder, E.J. Lumpkin, D.E. Lehman, A balanced design procedure for special
ki l1
br;i A:16 concentrically braced frame connections, J. Constr. Steel Res. 67 (No. 11) (2011)
X
n
17601772.
 H i  mpk [32] D.,.B. Merczel, H. Somja, J.-M. Aribert, J. Lg, On the behavior of concentrically
ki braced frames subjected to seismic loading Periodica Politechnica, Civ. Eng. 57/2
(2013) 113122.
[33] A.K. Chopra, R.K. Goel, C. Chintanapakdee, Evaluation of a modied MPA procedure
assuming higher modes as elastic to estimate seismic demands, Earthquake Spectra
References 20 (3) (2004) 757778, http://dx.doi.org/10.1193/1.1775237.
[34] J. Gyrgyi, Corrected ground motion functions in the case of near-fault earthquake,
[1] A.Y. Elghazouli, Seismic design procedures for concentrically braced frames, Proc. of The Tenth Int. Conf. on Computational Structures Technology, paper 326, pp.1-10,
the Inst. of Civil Engineers Struct. Build. 156 (2003) 381394. Valencia, Spain, 2010.
 D.B. Merczel et al. / Journal of Constructional Steel Research 125 (2016) 142163 163

[35] F.M. Mazzolani, V. Piluso, Plastic Design of Seismic Resistant Steel Frames, Earthq. [38] D.B. Merczel, Weak storey behaviour of concentrically braced steel frames subjected
Eng. Struct. Dyn. 26 (ssue 21) (1997) 167191. to seismic actions(PhD Dissertation) Department of Structural Mechanics, Budapest
[36] A. Longo, E. Nastri, V. Piluso, Theory of plastic mechanism control: state-of-the-art, University of Technology and Economics and Department of Civil Engineering, INSA
Open Constr. Building Technol. J. 8 (1) (2014) 262278 (2014). de Rennes. ID: 15ISAR 01/D15-01, 2015.
[37] A. Longo, R. Montuori, V. Piluso, Moment framesconcentrically braced frames dual
systems: analysis of different design criteria, Struct. Infrastruct. Eng. 12 (1) (2016)
122141 (2016).