Engineering Structures 30 (2008) 1802–1819

www.elsevier.com/locate/engstruct

Seismic performance of composite moment-resisting frames

A.Y. Elghazouli ∗ , J.M. Castro, B.A. Izzuddin
Department of Civil and Environmental Engineering, Imperial College London, UK

Received 29 April 2007; received in revised form 10 September 2007; accepted 10 December 2007
Available online 11 January 2008

Abstract

This paper examines the seismic performance of composite steel–concrete moment-resisting frames. A number of studies are carried out in
order to assess the influence of key parameters, related to the structural configuration as well as design and loading considerations, on the inelastic
seismic behaviour. Several sensitivity and parametric investigations are undertaken using an advanced analysis program that accounts for material
and geometric nonlinearities. Particular emphasis is given to composite frames designed according to the provisions of the European seismic code,
Eurocode 8. The validity of employing simplified nonlinear-static loading approaches is evaluated by comparison against the results of incremental
dynamic time-history analysis. Natural earthquake acceleration records, which are specifically selected and adjusted for compatibility with the
adopted design spectrum, are utilised for this purpose. In terms of frame configuration, it is shown that the span of the composite beam, the extent
of gravity loading and the number of stories can all have significant implications on the actual inelastic response characteristics of the structure.
Moreover, several design assumptions and decisions, such as those related to the behaviour factors, drift considerations, moment redistribution
and panel zone contribution, can also have a direct impact on the resulting performance. The studies presented in this paper highlight important
behavioural observations and trends, some of which point towards the need for further consideration and refinement of current design procedures.
c 2007 Elsevier Ltd. All rights reserved.

Keywords: Seismic response; Moment frames; Composite steel/concrete; Nonlinear analysis; Eurocode 8

1. Introduction This is attributed, on the one hand, to the assumption

The use of moment frames incorporating composite
steel–concrete floor systems offers several behavioural and
practical advantages over bare steel and other alternatives. The
increase of stiffness and capacity due to composite action
enables the use of larger beam spans under the same loading
conditions. Accordingly, the demand for larger and more
flexible usable space, coupled with the need for faster optimised
construction processes, has led to an increased utilisation of
composite frames in recent years.
In terms of seismic performance, composite frames
exhibit favourable behaviour due to the enhanced response
characteristics including ductility properties. However, the
number of detailed studies carried out on the seismic response
of composite frames, in comparison with bare steel or
reinforced concrete counter-parts, has been relatively limited.
∗ Corresponding author.
E-mail address: a.elghazouli@imperial.ac.uk (A.Y. Elghazouli).
that the behaviour could be largely inferred from studies
on steel or reinforced concrete frames and, on the other
hand, to the complexity of some of the analysis and
design issues involved, as reported in previous investigations
(e.g. [1–3]). Earlier studies have dealt with several modelling
and design considerations including connections [4,5], shear
interaction [6], slab effects [7] and behaviour factors [8,9].
Although previous studies have addressed and resolved
important behavioural aspects, there is a need for assessing
the key parameters influencing the seismic performance of
composite moment frames with typical geometric and loading
configurations. In particular, with the recent introduction of
Eurocode 8 [10], it is important to examine the performance
of composite frames designed according to the new European
seismic code. Similar to other seismic codes, the design of
composite frames in Eurocode 8 (EC8) largely stems from
provisions for bare steel frames with some modifications.
Consequently, it is also necessary to explore specific features of
composite frames and their implication on the inelastic seismic
behaviour.

c 2007 Elsevier Ltd. All rights reserved.

0141-0296/$ - see front matter

doi:10.1016/j.engstruct.2007.12.004
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Fig. 1. Structural arrangement for the reference structure: (a) Plan configuration, (b) Elevation of moment-resisting frame.
In this paper, the seismic performance of composite moment
frames is examined. Although emphasis is given to frames
designed according to the provisions of EC8, most of the
discussions are of more general nature irrespective of the code
employed. After introducing the structural configuration as well
as design procedures and assumptions adopted, the numerical
models and response criteria utilised in this investigation
are described. This is followed by an assessment of the
nonlinear-static response of a typical composite moment frame
with due account of the sensitivity of the response to the
loading pattern and the presence of gravity loading. The
validity of using nonlinear-static approaches for this type of
frame is evaluated by comparison with dynamic time-history
analysis, using earthquake records which are carefully selected
and adjusted for compatibility with the design spectrum.
A number of numerical studies are then carried out in
order to assess the influence of key loading, geometric and
design parameters on the response of composite frames. The
main behavioural observations and trends are discussed and
their implications on seismic performance and design are
highlighted.
1803
2. Structural configuration and design procedures
For the studies described in this paper, a reference structure
is selected such that it can be used as a basis for parametric
variations. The structure consists of a five-storey composite
steel–concrete office building as shown in Fig. 1. In plan,
columns are spaced at 9 m in both directions. Unlike ‘tube’
systems, in which perimeter moment frames may exhibit
relatively short spans, the choice of an internal composite frame
arrangement would typically be associated with comparatively
large beam spans. With due account of column inertia, lateral
resistance is provided by composite moment frames in one
direction whilst the other plan direction is assumed to have a
braced system. The floors consist of 120 mm thick composite
slabs of the re-entrant (dovetail) profile with 1 mm thick steel
deck supported by secondary beams spaced at 3 m.
The internal composite steel/concrete moment-resisting
frame, depicted in Fig. 1(b), was adopted in this study as the
reference frame (CREF). The frame was designed according to
the provisions of Eurocodes 3, 4 and 8 [10–12]. European steel
profiles were used for the columns (HE) and the beams (IPE).
1804 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Full shear interaction was assumed in the composite beams to
avoid undesirable behaviour in connectors as indicated in recent
studies [13]. The structure was initially designed according to
Eurocode 4 for the gravity and wind loading scenarios. The
load values were obtained from Eurocode 1 [14] and comprise
the self-weight incorporating an allowance of 2 kN/m2 for
finishing/partitions, in addition to an imposed load of 3 kN/m2 ,
with appropriate load factors. To satisfy the gravity design
situation, IPE500, HEB300 and HEB450 cross-sections were
required for the beams, external columns and internal columns,
respectively.
The initial frame resulting from the vertical loading scenario
was then checked for seismic conditions according to EC8.
The design peak ground acceleration (PGA) was assumed as
0.30g, whilst a rock soil (Ground Type A) and Spectrum Type
1 were considered. Since the structure satisfies EC8 regularity
conditions in plan and elevations, equivalent lateral seismic
loading based on an idealised first mode of response was
adopted. The fundamental period of vibration (T1 ) based on the
simplified expression given by EC8 (T1 = 0.085H 3/4 , where
H is the overall height in metres) was 0.65 s. However, T1
was found to be nearly 1.0 s from modal eigenvalue analysis
performed in a simplified structural model assuming centreline
representation for the members. This value was deemed more
accurate and hence was used in the evaluation of the design base
shear. On the basis that use is made of Class 1 cross-sections,
which satisfy DCH (Ductility Class High) requirements, the
reference behaviour factor (q) in this case is 5αu /α1 . The
recommended code value for αu /α1 for multi-bay multi-storey
frames is 1.3 which results in a behaviour factor of 6.5. The
total mass considered for the seismic design situation was
about 1440 tons based on a combination of the unfactored
dead load and 30% of the imposed load. The resulting design
base shear was about 700 kN which was distributed linearly
over the height. It should be noted that wind loading was also
considered, based on the assumption of a wind pressure of
0.65 kN/m2 , but this had no influence on the resulting member
sizes.
In addition to satisfaction of the seismic strength demands
in members, other seismic design checks include compliance
with stability and drift criteria as well as capacity design
considerations [10,15]. Second-order stability effects are
considered through the sensitivity coefficient (θ ) given as:
Ptot dr
θ= , (1)
Vtot h
where Ptot and Vtot are the total cumulative gravity load and
seismic shear applied at the storey under consideration; h is
the inter-storey height; and dr is the design inter-storey drift
(product of elastic inter-storey drift and ‘q’). For θ ≤ 0.1,
second-order effects may be ignored. If 0.1 < θ ≤ 0.2, the
multiplier 1/(1 − θ ) may be used to account for this effect and,
in any case, the value of θ should not exceed 0.3.
The other drift check is related to serviceability deforma-
tions, by limiting (dr ) in proportion to the storey height (h)
such that:
dr ν ≤ ψh (2)
where ψ depends on non-structural elements, given as
0.5%, 0.75% and 1.0% for brittle, ductile or non-interfering
components, respectively; and ν is a reduction factor in the
range of 0.4–0.5 to represent smaller serviceability events.
Apart from a number of checks to ensure ductile response
of beams, the main capacity design criterion is related to the
general concept of weak-beam/strong-column behaviour. For
steel and composite moment frames, a specific application
rule is stipulated whereby the design moments (M Ed ) for the
columns can be obtained from:
M Ed = M Ed,G + 1.1γov Ω M Ed,E (3)
where M Ed,G and M Ed,E are the moments due to the gravity
loads and lateral seismic forces, respectively; γov is the
material overstrength factor typically assumed as 1.25; Ω is a
beam overstrength factor determined as a minimum of Ωi =
M pl,Rd,i /M Ed,i of beams, where M Ed,i is the design moment
in beam ‘i’ and M pl,Rd,i is the corresponding plastic moment.
For the reference composite frame under consideration
(CREF), the member sizes selected on the basis of the gravity
combination were found to satisfy all the seismic design checks
as stipulated in EC8. However, the external columns had to be
increased to HEB 340 in order to limit (θ) to a value of 0.2 to
avoid nonlinear analysis at the design stage. Six reinforcement
bars of 20 mm diameter were also provided in the slab at
the beam-to-column joints, as recommended by the detailing
rules given in Annex C of EC8. The steel part of the beam-
to-column connections were considered to be fully welded in
order to achieve an idealised rigid response. Additionally, to
comply with EC8 provisions, the column panel zones had to
be strengthened through the addition of sets of two doubler
plates with thickness 21 mm and 29 mm to external and internal
columns respectively.
3. Analytical approaches
3.1. Numerical modelling
The nonlinear finite element program ADAPTIC [16] which
accounts for material and geometric nonlinearities is used for
the analysis of the composite frames. The accuracy of the
program in modelling the response of steel and composite
structures has been extensively validated in previous studies
(e.g. [17–22]).
For the steel columns, four cubic elasto-plastic elements
(cbp2) [17] are utilised. The cbp2 elements account for the
spread of plasticity across the section, using a fibre approach, as
well as along the length using Gaussian sections in conjunction
with a cubic shape function. The same elements are used for
the composite beams where two parallel lines of 18 cbp2
elements are considered at the centroids of the steel and
concrete constituent parts, as illustrated in Fig. 2.
Composite action is developed by introducing link elements
(lnk2) with rigid properties connecting the steel and concrete
line elements. A rigid connection is considered on the
assumption that a sufficiently high degree of shear connection
is present such that the influence of slip between the steel
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Fig. 2. Modelling of composite beams.
beam and the slab on the response is not significant. It should
be noted however that the degree of shear connection may
need to be notably higher than that implied by a theoretical
full-connection (i.e. 100% shear connection) in order to
achieve these idealised conditions. In a recent study [23] in
which composite beams were modelled with nonlinear springs
representing a range between 50% and 150% shear connection,
it was observed that only with 150% or more can the inelastic
properties be reasonably represented by a rigid connection
idealisation. Analytical simulations of experimental studies
carried out on a full-scale composite frame [24] designed
to Eurocode 8 also indicated a discrepancy of less than 3%
on the response when the composite beams were modelled
using the actual shear connection or with a rigid idealisation.
Whilst low and moderate levels of shear connection are not
recommended for seismic-resisting composite systems [13],
due to the susceptibility of the connectors to low-cycle fatigue,
it is important to emphasise that the slip between the slab and
the steel beam may need to be incorporated in numerical models
unless a very high degree of shear connection is adopted in
design.
For simplicity, the effective slab width was assumed as
1.25 m based on an estimate of the contra-flexure location and
the provisions of EC4 for the sagging moment region. Although
the provisions vary in EC4 and EC8 for sagging/hogging
regions and for design/analysis, a sensitivity study indicated
that these differences do not have a notable influence on the
overall frame response [23]. Further discussion dealing with
the issue of effective width in composite beams can be found
elsewhere [25].
Special attention is given to the modelling of the panel zone
regions, for which a modified frame approach [22] is adopted
by assembling link (lnk2) and joint (jel2) elements. Fig. 3
shows the model for a typical external connection. Distinction
is made between the main panel zone and the top part which
is in contact with the slab. The contact behaviour between
the composite slab and the column is also taken into account
through the introduction of joint elements. Asymmetric curves
are employed for panel zones located at external joints, whilst
symmetric relationships are considered for internal panels. For
the top panel, a linear elastic response is considered. On the
other hand, a plastic curve is considered for the slab-contact
joint. Under negative bending, the capacity at the contact region
is based on the maximum yield force that can develop in the
reinforcement bars whilst, under positive moment, the joint
capacity is based on the maximum axial force that can be
mobilised in the steel profile of the composite beam. The
inclusion of a joint element to model slab interaction also
1805
Fig. 3. Numerical models for composite joints [22].
permits the consideration of partial-strength connection effects
if necessary.
For steel materials, a bilinear elasto-plastic cyclic model
with strain-hardening (stl1) is adopted for both structural and
reinforcing steel (Fig. 4(a)). On the other hand, concrete
nonlinear behaviour is accounted for through a uniaxial cyclic
constitutive model (con1) featuring both compressive and
tensile softening (Fig. 4(b)). In all the analyses performed
in this study, the steel elastic modulus (E) and the strain-
hardening coefficient (µ) are assumed as 210 × 103 N/mm2
and 0.5%. The yield strength values ( f y ) for structural steel
and reinforcement bars are assumed as 275 and 500 N/mm2 ,
respectively. For concrete, the values of E c1 , E c2 , f c1 , and
f c2 are assumed as 15 × 103 , 8 × 103 , 30 and 18 N/mm2 ,
respectively. It is also assumed that the residual strength in
compression is reached at a concrete strain of 0.35%, whilst
concrete behaviour in tension is ignored.
For eigenvalue as well as time-history analysis, appropriate
representation of the mass and damping characteristics of the
structure is required. In the case of eigenvalue analysis, only
concentrated mass elements (cnm2) are incorporated in the
model and located at the thirds of the beam spans. For dynamic
analysis, viscous damping, assumed as 5% of the critical, is
also considered by inclusion of dashpot elements (cnd2) at
the same locations of mass elements. It is worth noting at
this point that discrete modelling of the panel zone has an
influence on the stiffness, and in turn on the natural periods
of vibration. As noted before, for the reference frame, the
fundamental period was found to be around 1.0 s, using a
conventional centreline modelling approach. When the panel
zone is realistically incorporated, with due account of the
1806 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819

Fig. 4. Constitutive models adopted for: (a) Steel and (b) Concrete.

Fig. 5. M–Φ curves for composite beam: (a) Sagging and (b) Hogging moment.

additional stiffness of the doubler plates, the period reduces to
0.88 s. On the other hand, if all the panel zone components
within the discrete model are assumed to be infinitely rigid, the
period reduces further to about 0.8 s.
3.2. Response criteria
In order to assess the inelastic seismic performance of
composite frames, it is necessary to define specific criteria
that can be associated with the yield and ultimate response
levels. Various approaches and assumptions can be adopted to
select comparative measures for this purpose. The notion of
‘significant yield’, which is typically associated with plastic
hinge formation in framed structures, is employed in this
investigation. Due to the more detailed strategy adopted in the
representation of the composite members, it was decided to
establish the local yield criteria based on fundamental cross-
section data that ADAPTIC is able to provide. Instead of
relating the plastic hinge formation to the achievement of
a theoretical plastic moment, plastic hinges are assumed to
form in the composite beam under negative moment when
both extreme fibres of the steel profile reach the yield strain.
On the other hand, a plastic hinge is assumed to form under
positive bending when the axial capacity of the steel beam is
reached; this criterion is valid for ductile sections in which
the plastic neutral axis is located in the slab. The two criteria
described above are highlighted in Fig. 5 which depicts the
moment-curvature relationship for the composite beam under
sagging (positive) and hogging (negative) bending. The two
curves clearly indicate that the selected criteria are closely
related to the plastic moment (M pl ) of the composite cross-
section, and hence can be used in identifying the formation
of a plastic hinge. For the steel columns, similar to beams in
negative bending, a plastic hinge is assumed to form when both
extreme cross-section fibres attain the yield strain as proposed
in previous studies [26].
In the case of ultimate response, global criteria related to
inter-storey drift are adopted, as in comparison with local
failure considerations (such as local buckling of lower flange
or concrete crushing in compression), they are found to govern
the behaviour of the frames dealt with in this investigation.
The study frames are designed and detailed such that all steel
members satisfy cross-section requirements for the highest
ductility class. Additionally, the amount of the additional slab
reinforcement over supports reduces the extent of possible
strain localisation in hogging moment. On the other hand, the
composite beams in sagging moment satisfy limitations on the
depth of the plastic neutral axis for achieving high ductility. It is
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
also important to note that the idealised uniaxial representation
of concrete behaviour makes the definition of a strain-
based failure criterion in the slab unrealistic. Moreover, this
investigation is concerned more with assessing the response in
terms of global capacity considerations and plastic mechanisms
rather than examining local failure conditions. It was therefore
decided to assess ultimate conditions, where necessary, based
on a limiting inter-storey drift which, in any case, is directly
related to the ductility demand imposed on structural members.
A limiting value in the range of 2.0%–4.0% [27,28] is normally
adopted for framed structures, with 3.0% being commonly used
(e.g. [9,29]).
The numerical model described above is utilised in the
following section to assess the nonlinear-static as well as
dynamic response of the reference composite frame presented
earlier. Subsequently, a number of parametric variations are
considered to examine key behavioural and design issues.
4. Lateral seismic loading
4.1. Nonlinear-static response
The use of nonlinear-static, or pushover, analysis for seismic
assessment and design has increased significantly in recent
years. It can be employed to assess the overall capacity and
stability, and to identify the likely plastic mechanisms and
associated dissipative regions. The attractiveness of pushover
analysis stems mainly from its relative simplicity, in terms of
modelling and computational demands as well as interpretation
of results, in comparison with nonlinear dynamic analysis [30].
The main purpose of this section is to examine the response
of the reference frame using ‘conventional’ forms of pushover
analysis, commonly referred to as the ‘triangular’ and ‘uniform’
distributions. The adequacy of the pushover idealisations in
representing the inelastic response of the frame is then assessed
in subsequent sections by comparison against the results of
incremental dynamic analysis.
The nonlinear-static response of the reference frame is
presented in Fig. 6. The analysis is conducted by increasing
the displacement at the top of the frame incrementally up to a
global lateral drift of about 3% of the overall height. The results
are shown for both the linear (triangular) and uniform patterns
of the lateral load. Due to the regularity and simplicity of the
structure, the triangular pattern represents closely a first mode-
dominated response. On the other hand, the uniform pattern is
expected to be important if higher modes contribute notably to
the response or when significant inelastic concentrations occur.
As shown in Fig. 6(a), the uniform loading pattern results in
higher stiffness and capacity in comparison with the triangular
distribution, as expected. The inter-storey drifts in the frame
at yield (formation of first plastic hinge) as well as at ultimate
(achievement of 3% inter-storey drift) are depicted in Fig. 6(b).
The inter-storey drift distribution is almost uniform at the
yield stage since the behaviour is largely elastic. At ultimate,
however, lower stories exhibit significant inter-storey drifts
compared to upper levels. The inelastic concentration is more
pronounced for the uniform load pattern which forces a
1807
more distinct soft-storey behaviour at the first storey. The
inelastic distributions shown in Fig. 6(b) are closely related to
Fig. 6(c) which illustrates the sequence and location of plastic
hinges occurring in the beam and column members up to the
attainment of the defined ultimate limit.
Apart from the difference in response for the linear and
uniform patterns, a number of important observations are
noteworthy with reference to Fig. 6. Firstly, the frame is shown
in Fig. 6(a) to possess an actual capacity which is considerably
higher than that assumed in design. A degree of overstrength
is generally expected for framed structures [15], particularly
when a relatively high behaviour factor is employed, since
the member sizes are likely to be governed by the gravity
loading scenario or by deformation-related limits (from inter-
storey or stability checks). This overstrength effect becomes
even more pronounced in the case of composite frames. One
of the reasons for this is the relatively large spans that may be
present in composite frames, hence increasing the dependence
of beam sizes on the gravity loading design scenario. Moreover,
composite frames exhibit significant levels of redistribution,
in terms of the ratio between the ultimate base shear and
that at yield (referred to as αu /α1 in EC8), which contributes
significantly to the global overstrength of the frame.
Another important observation is related to the formation of
plastic hinges predominately on one side of the beam spans,
as illustrated in Fig. 6(c). This is caused by a combination of
the influence of gravity moments (which can be significant in
relatively large spans enabled by composite action), coupled
with the difference between the plastic moment capacity of the
cross-section under positive and negative moments which is
characteristic of composite beams. Unless considerable levels
of overstrength are provided to the columns to preclude the
formation of column hinges at significant inelastic drifts, beam
hinges may not form in the other side of the spans at ultimate,
which is the case in the reference frame. This characteristic
behaviour occurs even when the gravity loading is disregarded
(as shown in Fig. 6(d)), thus indicating the significance of the
asymmetry in the moment capacity of the beams.
In relation to the above discussion, it is also important to
note that whilst codes aim to achieve a ‘weak-beam/strong-
column’ behaviour, a degree of column hinging is unavoidable
in many cases unless large overstrength factors, well beyond
those implied in codes, are adopted for column design. This is
illustrated by the hinge patterns shown in Fig. 6(c), and was
also reported in previous studies on steel frames (e.g. [31]).
In the inelastic range, the points of contra-flexure in members
change and consequently the distribution of moments vary
considerably from the idealised conditions assumed in design.
Therefore, in terms of actual response, the purpose of meeting
code requirements can perhaps be viewed as aiming to achieve
relatively strong columns such that beam rather than column
yielding predominates in several stories. Further discussion of
these capacity design issues is presented later on in this paper.
In order to examine the general validity of nonlinear-
static procedures for assessing the overall seismic response of
composite moment frames in further parametric and sensitivity
studies, the pushover results of the reference frame are
1808 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819

Fig. 6. Reference structure: (a) Pushover curve, (b) Inter-storey drifts, (c) Hinge formation for different load patterns and (d) Hinge formation for case without
gravity loads.

compared with those from dynamic analysis in subsequent
sections. To achieve a realistic basis for this comparison, careful
attention is given to the selection and scaling of the time-history
records in relation to the code-spectrum assumed in design.
4.2. Input for time-history analysis
A number of difficulties arise when choosing specific
records for dynamic analysis, particularly in cases where
seismic hazard studies are not available. In general, two
different criteria for record selection can be employed. One
consists of choosing the records according to strong-motion
parameters and the other is based on geophysical criteria.
The search based on strong-motion parameters consists of
finding records that have similar shape to the response
spectrum provided by the code, obtained for a site with
similar characteristics. An efficient procedure may consist,
for example, of calculating the average root-mean-square
deviation (Drms ) of the spectrum of the tentative record
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819 1809

Table 1
Selected records and scaling factors

EQ ID Earthquake Date Country Magnitude (Mw ) Distance (km) Soil type PGA (g) Scaling factor
CLUC Campano Lucano 23/11/1980 Italy 6.9 10 Rock 0.0599 6.0
MHILL Morgan Hill 24/04/1984 US 6.5 28 Stiff 0.0340 10.0
LDRS Landers 28/06/1992 US 7.3 30 Stiff 0.1351 2.3
NRDG Northridge 17/01/1994 US 6.7 31 Stiff 0.0732 5.0
CHCH Chi-Chi 20/09/1999 Taiwan 7.6 39 Stiff 0.4145 1.1
TMRN Taumaranui 05/01/1973 New Zealand 6.6 65 Rock 0.0358 10.0
HMINE Hector Mine 16/10/1999 US 7.1 61 Rock 0.0568 5.3

from the target design spectrum. Alternatively, spectrum-

compatible artificial records may be generated from white
noise. This latter technique however tends to create unrealistic
records both in terms of frequency, phase content, number of
cycles and duration of motion [32]. On the other hand, for
situations where the site is well characterised in seismological
terms by either a deterministic or probabilistic seismic hazard
assessment, then one or more earthquake scenarios can be
established. Accordingly, the fault mechanisms and the ranges
of magnitudes and fault distances of the earthquakes affecting
the site are known [33]. In this case, the selection can be
conducted by searching a strong-motion database for records
that match those geophysical parameters.
The selection of records is carried out herein by combining
the two criteria discussed above. The records are selected
such that they correspond to the shape of the code-spectrum
employed (Type 1) as well satisfy the seismological criteria
inferred by this choice. Seven records, with the lowest Drms
from the target code-spectrum, were sought for this study from
the strong-motion database available at Imperial College, by
imposing the following conditions: (i) moment magnitudes
(Mw ) larger than 6 since Spectrum Type 1 corresponds to
high magnitude events [34]; (ii) records involving near-fault or
forward directivity effects are avoided; (iii) rock or stiff soil
sites, for consistency with the soil type assumed in design; (iii)
PGA larger than 0.03g, to avoid applying unrealistically high
scaling factors. The duration was not specifically considered
since the model structures do not exhibit significant degradation
effects. Records with forward directivity effects were not
considered although it should be noted that they can pose a high
damage potential to flexible structures [35]. However, these
types of records have been rarely observed in Europe with the
only relevant case recorded during the 1995 Aegion earthquake
Fig. 7. Response spectra of the selected records (all normalised to PGA) and
in Greece [36]. The seven records selected are listed in Table 1
target design spectrum (EC8-Type 1): (a) Initial records, (b) Wavelet-adjusted.
together with the scaling factors adopted in order to have a
good fit with the code-spectrum for the period range of interest,
which in this case is considered to be between 0.8 and 2.4 s. damping levels. The correction with RSPMATCH2005 is
The response spectra, for 5% damping, for all seven records as performed in two steps. In the first step, each record is modified
well as the target design spectrum are shown in Fig. 7(a). in order to match the target spectrum within the period range
The selection was then followed by another adjustment between 0 and 1 s. In the second step, wavelets are introduced
process for matching the records to the target design spectrum, in the time series in order to match the target spectrum for
based on the introduction of wavelets to the acceleration the whole period range, i.e. between 0 and 4 s. A maximum
time series [37]. For this purpose, a modified version [38] of tolerance of 5% for the mismatch between the record spectrum
the program RSPMATCH [39] was employed. The modified and the target spectrum is defined.
program incorporates new wavelet types that avoid the need The response spectra for the final adjusted records as well
for baseline corrections and permit adjustment for multiple as the target spectrum are shown in Fig. 7(b). On the other
1810 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819

(a) CLUC (before adjustment). (b) CLUC (after adjustment).

(c) MHILL (after adjustment). (d) LDRS (after adjustment).

(e) NRDG (after adjustment). (f) CHCH (after adjustment).

(g) TMRN (after adjustment). (h) HMINE (after adjustment).

Fig. 8. Earthquake records adopted in the analysis.

hand, the acceleration time-histories for the final adjusted
records are shown in Fig. 8. It is worth mentioning that the
ground motion characteristics, as well as the acceleration and
velocity time series, are not significantly affected by the wavelet
adjustment. For example, this is illustrated in Fig. 8(a) and (b)
which depict the acceleration record of CLUC before and after
adjustment, respectively. However, some differences can occur
in the displacement series in some cases, where a reduction in
the number of ground displacement cycles is observed. This
situation develops in cases where the wavelet modification
algorithm requires several iterations to adjust the record.
4.3. Dynamic response
Using the seven modified records described in the previous
section, incremental dynamic analysis [40] was performed for
the reference frame. Each of the seven records was applied
with increasing ground motion intensity, resulting in over 100
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819 1811

Fig. 9. Time-history results vs pushover curves.

Fig. 10. Inter-storey drifts for 0.15 m global drift (time-history, linear and
analyses. In this study, the intensity measure was considered uniform pushover).
as the spectral acceleration at the fundamental period and, for
each analysis, the base shears as well as global and inter-storey
drifts were examined. A large amount of data was obtained
from the dynamic analysis hence, for brevity, only results which
are considered to be of direct relevance to the issues addressed
in this paper are discussed below.
The overall dynamic response is compared in Fig. 9 to that
obtained from pushover analysis, with both linear and uniform
patterns. The response is presented in terms of base shear versus
maximum drift at the top of the frame. A special procedure is
required to extract these results from the output of the dynamic
analysis since the maximum displacement does not necessarily
coincide with the time corresponding to the peak base shear. As
suggested in previous studies [41], the maximum displacement
was matched with the peak base shear within a specified time
interval, for which a value of ±0.5 s was found to produce
consistent results.
Fig. 9 indicates relatively low dispersion between the results Fig. 11. IDA curves of maximum base shear for control structure.
from the seven earthquakes, which is clearly aided by the
record-selection and spectrum-matching procedures described analysis with triangular loading pattern is more closely related
before. More importantly, the plot reveals relatively good to the mean of the time-history analyses. On the other hand,
correlation with the pushover results, particularly for the linear the uniform pattern idealisation tends to overestimate the inter-
pattern case. As shown in the figure, the average of the time- storey drifts at lower stories and to underestimate the demand
history results is more closely related to the linear pattern for at upper levels.
most of the response range, except perhaps at significant drift The maximum base shears and top frame drift obtained
levels when inelastic deformations may tend to concentrate in from the incremental dynamic analysis are depicted in Fig. 11
lower storeys. In contrast, the uniform pattern provides more and Fig. 12, respectively. These are shown in relation to the
of an envelope of the global behaviour rather than a realistic scaled spectral acceleration corresponding to the fundamental
representation of the mean level of dynamic response. period of vibration and 5% damping, i.e. Sa (T1 , 5%).
Pushover analysis using a linear triangular pattern of loading Whilst the maximum base shears do not show considerable
also provides better correlation with time-history results in variation for the various records, the dispersion of global
terms of inter-storey drifts. For example, Fig. 10 depicts the drifts increases with higher seismic intensity as the level
distribution of the maximum inter-storey drifts corresponding of inelastic deformations becomes more significant. More
to a top displacement of about 0.15 m, as this corresponds importantly, as illustrated by Fig. 12, the peak top displacement
approximately to average inter-storey drifts in-between yield is largely proportional to the seismic intensity. This suggests the
and ultimate. The same procedure adopted for extracting the general validity of the ‘equal displacement approximation’ rule
base shear versus top drift results was also used in this case commonly adopted in several seismic codes, including EC8, for
to obtain the inter-storey drifts corresponding to the specified the structure under consideration. It should be noted however
level of global deformation. The plot shows that the pushover that if the response is assessed on the basis of individual
1812 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819

Table 2
Summary of parameters considered and section sizes adopted

Study Designation Remarks Gravity loading design Seismic design

Primary Secondary External Internal External Internal GC*
beams beams columns columns columns columns
Reference CREF – IPE 500 IPE 330 HEB 300 HEB 450 HEB 340 HEB 450 c
frame
CR4 q=4 IPE 500 IPE 330 HEB 300 HEB 450 HEB 500 HEB 500 a
CRED30 Redistribution IPE 400 IPE 330 HEB 300 HEB 450 HEB 300 HEB 450 c
Design
CPZ No doubler IPE 500 IPE 330 HEB 300 HEB 450 HEB 340 HEB 450 c
considerations
plates
CSI015 PGA = 0.15g IPE 500 IPE 330 HEB 300 HEB 450 HEB 340 HEB 450 c
CSI050 PGA = 0.50g IPE 500 IPE 330 HEB 300 HEB 450 HEB 550 HEB 600 d

CTB3 3 m spacing IPE 300 – HEA 220 HEA 240 HEA 320 HEA 360 c
Spacing and CTB12 12 m spacing IPE 550 IPE 400 HEB 400 HEB 650 HEB 400 HEB 650 b
geometry CPB6 6 m span IPE 330 IPE 330 HEB 260 HEB 320 HEB 450 HEB 500 c
CSTOR3 3-storey frame IPE 500 IPE 330 HEB 260 HEB 300 HEB 260 HEB 320 a
∗ Governing Criterion: a. seismic strength demand, b. gravity loading scenario, c. stability coefficient, d. serviceability drift.

on the performance. Using the reference frame described

Fig. 12. IDA curves of maximum top displacement for control structure.
records, rather than the mean or median, then significant
departure from the equal-displacement approximation would be
observed. This is illustrated in Fig. 12 by noting, for example,
the drift response of the structure under the TMRN record with
increasing seismic intensity.
After examining the various approaches for applying
the lateral seismic loading, subsequent parts of this paper
investigate the influence of other salient geometrical and design
parameters on the response. As discussed above, pushover
analysis based on the linear pattern was shown to be in good
agreement with the average results obtained from the dynamic
analysis. Accordingly, and given the relative simplicity of
performing and interpreting pushover analysis coupled with
the comparative nature of parametric and sensitivity studies, it
was decided to limit the discussions presented in subsequent
sections to results from pushover analysis with triangular load
patterns.
5. Design parameters and considerations
The design process of a composite moment frame involves
several assumptions and choices that can have a direct influence
before as a basis, a number of variations to several design
parameters and assumptions are considered in this section
in order to highlight important behavioural implications. The
modifications carried out to the reference frame are summarised
in Table 2, together with the resulting member sizes and
governing criteria. In all cases, only one design parameter or
assumption is varied within each specific frame. It should be
noted that, where possible, it was decided in most cases to retain
the beams sizes based on gravity design and adjust column sizes
to satisfy the various design criteria.
A significant number of parameters were varied but, for
brevity, only those which are thought to provide some insight
into key aspects of the behaviour are described herein. This
section deals with parameters related to design, for a structure
of the same geometry. In particular, focus is given to the
behaviour factor, moment redistribution, panel zone design and
seismic intensity. On the other hand, Section 6 is concerned
with variations to the structural configuration.
5.1. Behaviour factor
The recommended values for the behaviour factor ‘q’ in
EC8 are, in principle, intended as upper limits. However, in
practice, the recommended behaviour factor is often used as
a default with a view to exploiting available ductility and
energy dissipation capabilities. It is important to note, however,
that lower behaviour factors may be adopted and such a
choice could often be more rational [15] particularly in low-
to-moderate seismicity regions.
In order to examine some of the issues involved in using
a lower behaviour factor, the design of the reference frame
(CREF) is modified (CR4) in accordance with ‘q’ of 4 (DCM)
instead of 6.5 (DCH). This enables the use of Class 2 rather
than Class 1 cross-sections if necessary but, more importantly,
the change in ‘q’ results in the need for larger column sizes,
as indicated in Table 2, in order to satisfy the higher lateral
strength requirement. As indicated in Fig. 13(a), this leads to
an increase in the lateral capacity of the frame in comparison
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Fig. 13. Influence of design considerations on lateral response: (a) Pushover
curves, (b) Inter-storey drifts.
with the reference structure. Also, as depicted in Fig. 13(b),
due to the increased column-to-beam strength in the modified
frame, an improved distribution of inter-storey drift is obtained,
which is directly related to enhanced hinge formation and
reduced second-order effects. This is also supported by the
hinge pattern shown in Fig. 15(b), which shows more extensive
hinge formation in upper stories at ultimate in comparison with
the reference frame.
In general, it should be noted that the influence of the choice
of behaviour factor on the behaviour is inter-related to several
design parameters and criteria. A change in behaviour factor
can clearly lead to a modification of member sizes, and hence
directly influence the stiffness and capacity as well as the
overstrength and inelastic performance of the frame. However,
for a particular frame configuration, these effects are largely
dependent on a number of other parameters including the level
of seismicity, and the second-order design criteria such as the
stability coefficient in EC8.
5.2. Moment redistribution
Due to the relatively large spans often associated with
composite floors, the beam sizes are normally governed by
1813
the gravity design rather than the seismic loading combination.
Consequently, it may be beneficial to make use of the
redistribution ability of composite beams in the gravity design
situation to reduce the required beam size. The reduction
in beam size can be significant in composite beams due to
the characteristic difference between the positive and negative
moment capacity of the cross-section.
To examine the effect of gravity moment redistribution, the
design of the reference frame (CREF) is modified (CRED30).
In this case, the gravity design is based on a 30% redistribution
in accordance with the allowance in Eurocode 4 [12]. This
results in a reduction in beam size from IPE 500 to IPE 400.
However, this is also accompanied by a reduction in stiffness
and capacity which, depending on the seismic intensity and
behaviour factor as well as other design parameters, may still
be adequate. For the frame under consideration (CRED30),
although the reduced beam size enables a reduction in column
sizes from a capacity-design viewpoint, satisfaction of the
stability criterion (θ ) necessitates the use of at least the same
column sections.
Fig. 13(a) includes the pushover response of CRED30
in comparison with CREF, whilst the inter-storey drift
distributions at yield and ultimate are shown in Fig. 13(b).
On the other hand, the hinge pattern at ultimate is depicted in
Fig. 15(c). Whereas the stiffness and capacity of CRED30 are
clearly lower than the reference structure, the overall seismic
response is more favourable due to the more dominant beam
yielding as demonstrated in Figs. 13(b) and 15(c). This is
facilitated by the lower beam-to-column capacity ratio attained
in CRED30. In general, the benefits of employing gravity
moment redistribution depend on the specific case but it is an
issue that merits consideration within the design process.
5.3. Panel zone design
As noted previously, the column panel zone in the reference
frame was designed according to the provisions of EC8 which
do not imply significant yielding within this component. To
satisfy code requirements, each web panel was strengthened
with two doubler plates, as described before. Other seismic
codes (e.g. [42]) explicitly allow partial yielding in panel zones
in recognition of their inherent ductility and in order to alleviate
excessive demands on the plastic hinges in the beams. It was
deemed of interest therefore to examine the response of a
modified frame (CPZ) which has the same characteristics as
the reference case (CREF) but without strengthening the web
panels.
The lateral response of frame CPZ is included in Fig. 13(a),
whilst the drift distributions over the height are given in
Fig. 13(b). On the other hand, the hinge pattern and sequence
for this case are shown in Fig. 15(d). In comparison with
CREF, it is evident that the flexibility and yielding of the panel
zone leads to a significant reduction in the overall stiffness
and capacity of the frame. The panel zones in CPZ have
an average yield capacity of about 35% of the beam which
results in a reduction of about 30% in the overall frame
capacity. Importantly, this also results in a highly stable plastic
1814 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819

mechanism involving only the panel zones and the base of the
columns, as shown in Fig. 15(d), and hence more uniform inter-
storey distribution over the height (Fig. 13(b)).
Despite the stable inelastic behaviour of panel zones,
the rationale behind limiting energy dissipation within this
component in EC8 is based on merited concerns. The local
bending of column flanges associated with inelastic shear
deformations of the panel may cause excessive local strains
in the welded connection components, which can precipitate
early failure as suggested by several researchers [43,44]. On
the other hand, allowing yielding in panel zones can alleviate
inelastic demands on the beams and lead to favourable overall
inelastic frame response. Accordingly, there are also benefits
from allowing a degree of yielding of panel zones, as implied
in recent design documents [45]. More detailed examination
of this issue is beyond the scope of this investigation but it is
clearly an area that deserves further assessment.

5.4. Seismic intensity

In Eurocode 8, the seismic intensity is represented primarily

through the design peak ground acceleration (PGA). In order
to assess the influence of seismic intensity on the design of
the reference frame (CREF), for which a PGA of 0.30g was
assumed, two variations are considered (CSI015 and CSI050)
corresponding to PGA of 0.15g and 0.50g, respectively. Other
design parameters and criteria used for the reference frame are
retained.
In the case of CSI015, the design process results in identical
member sizes to those used in CREF. This is expected since
the criterion governing the design in both cases is the stability
coefficient (θ ) which, as evident by close observation of Eq.
(1), is independent of the seismic intensity. Since the design
base shear for CSI015 is half of that in CREF, whilst the Fig. 14. Influence of seismic intensity on lateral response: (a) Pushover curves,
member sizes are identical, the direct implication is that the (b) Inter-storey drifts.
frame overstrength is doubled. Clearly, this is a consequence of
also have a direct influence on the seismic response. In this
employing a high ‘q’ factor in a case of relatively low seismic
section, a number of geometric parameters related to the frame
demand, leading to undue levels of overstrength as reported in
configuration are considered, as summarised in Table 2. These
previous studies on framed structures [15].
include the frame spacing, primary beam span and the number
For CSI050, the increase in PGA changes the governing
of stories.
criterion in design. Similar effects can also be observed when
the soil classification is changed. In CSI050, the sizes of the
6.1. Transverse spacing
columns are governed by the inter-storey serviceability drift
limit assumed as 0.75% of the storey height. The inter-storey
Depending on the structural plan arrangement of the
drift increases with PGA, unlike the stability coefficient, and
building, the frame spacing may be considerably different
becomes more critical. If the inter-storey drift limit of 0.75% is
from that assumed in the reference structure selected. This,
relaxed to 1.0%, the seismic demand on the columns becomes
in turn, would have a direct effect on the level of gravity
the governing factor. In comparison with CREF, the larger
loading imposed on the primary beam which forms part of the
column sizes lead to relatively higher stiffness and capacity
composite moment frame. In order to examine the influence of
(Fig. 14(a)), as well as improved distribution of drift and hinge
these possible variations, two additional cases are considered,
formation over height (Fig. 14(b) and Fig. 15(f), respectively).
in which the frame spacing is changed from the 9.0 m assumed
Moreover, as expected, the level of global frame overstrength is
significantly lower than that in CREF. in the reference frame (CREF) to 3.0 m (CTB3) and 12.0 m
(CTB12). In the case of CTB3, and with reference to Fig. 1(a),
6. Frame spacing and geometry the slab is assumed to span directly onto the moment frames.
As shown in Table 2, for CTB12 the sizes of beams
Apart from the design parameters and considerations and columns are larger than those for CREF, as these are
discussed in the previous section, the frame configuration can governed by the gravity loading conditions. On the other
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819 1815

Fig. 15. Sequence of hinge formation for: (a) CREF, (b) CR4, (c) CRED30, (d) CPZ, (e) CSI015, (f) CSI050.

hand, for CTB3 smaller member sizes are required due to the
considerably lower gravity loading present, but the column
sizes are ultimately governed by the stability criterion. As
shown in Fig. 16(a), the resulting response, in terms of stiffness
and capacity, reflects the difference in member sizes between
the three frames. Due to the high column-to-beam capacity
ratio in CTB3, it provides a favourable distribution of plasticity
over the height as shown in Fig. 16(b), as well dominant beam
hinging as indicated in Fig. 18(a). However, the significant
increase in column sizes from that required for gravity loading,
in order to satisfy the stability criterion rather than seismic
demands, results in relatively high frame overstrength in the
case of CTB3.
6.2. Primary beam span
In order to examine the influence of the span of the beam
on the seismic response of the composite frame, the beam
span is modified from 9.0 m in the reference frame (CREF)
to 6.0 m (CPB6). Due to the shorter length and reduced gravity
loading, the member sizes required to resist the gravity situation
are smaller than in CREF as listed in Table 2. However,
the column sizes are ultimately governed by the second-order
stability criterion. The adopted design procedure therefore leads
to larger column sizes in CPB6 compared to CREF.
The response of CPB6 in comparison with CREF is shown
in Fig. 17(a). The reduction in beam size leads to lower stiffness
1816 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Fig. 16. Influence of frame spacing on lateral response: (a) Pushover curves,
(b) Inter-storey drifts.
and capacity in CPB6. However, as expected, the higher
column-to-beam capacity ratio results in a more favourable
inelastic distribution over the height as shown in Fig. 17(b).
This is also illustrated in Fig. 18(c), indicating an ‘ideal’
hinge formation in the beams. It is interesting to note that,
in this case, the smaller beam spans and the prevention of
column hinging permit the formation of hinges at both sides of
most members. Clearly, the lower beam span in CPB6 enables
the use of smaller beam cross-section hence facilitating a
weak-beam/strong-column design within practical size ranges.
However, the observations made above in relation to inelastic
response characteristics could obviously change if the designer
decides to increase beam sizes, rather than column sizes only,
in order to satisfy the stability criterion.
6.3. Number of stories
Another variation of the reference frame (CREF) is
considered whereby the number of storeys is reduced to three
(CSTOR3). This case is selected to highlight an issue which
becomes particularly important in the case of frames with large
beam spans and/or low number of stories [15,46]. For CSTOR3,
the relatively low overall gravity loading applied to the ground
Fig. 17. Influence of beam span and number of stories on lateral response: (a)
Pushover curves, (b) Inter-storey drifts.
floor columns means that the stability coefficient does not
govern the size of the column and it is instead determined
by seismic strength demands. In principle, the capacity design
rules for columns should ensure that relatively strong columns
are provided such that more dominant beam hinging occurs.
However, direct implementation of the application rule of Eq.
(3) in isolation can lead to unsatisfactory performance.
Figs. 17(a), (b) and 18(d) depict the pushover response,
inter-storey drift and plastic hinge pattern, respectively,
obtained for CSTOR3. Clearly, the column sizes are inadequate
in this case to prevent the formation of an undesirable
column mechanism. This is largely an implication of Eq.
(3) which in its suggested form ignores the influence of
gravity loading in calculating the overstrength parameter Ω .
As a result, Ω is underestimated considerably for cases in
which the gravity moment in the beams (M Ed,G ) constitutes
a significant proportion of the total moment (M Ed ). This is
often the case in composite configurations as these typically
incorporate relatively large beam spans. Although this issue
is not necessarily related to the number of stories, the limits
imposed on the stability coefficient lessen its effect on the
required column sizes except in low-rise frames. In fact, a more
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
Fig. 18. Sequence of hinge formation for (a) CTB3, (b) CTB12, (c) CPB6, (d) CSTOR3.
realistic representation of the overstrength parameter Ω would
necessitate a modification of the code-specified relationship
to (M pl,Rd − M Ed,G )/M Ed,E . Such modification results in
a significantly higher value of Ω for CSTOR3, leading to
column sizes which are at least as large as those in CREF,
hence preventing the formation of a storey mechanism. The
increase in column sizes would evidently improve the seismic
performance of the structure and, importantly, it would have
an insignificant influence on the final cost of the structure.
Alternatively, instead of the suggested modification, the code
rules could be replaced by (or used in conjunction with) other
capacity design measures.
7. Concluding remarks
This paper assesses the inelastic seismic performance of
composite steel/concrete moment-resisting frames designed
according to the provisions of Eurocode 8. After discussing
the design procedures and assumptions, the numerical models
and response criteria adopted in this investigation are described.
The analysis accounts for material and geometric nonlinearities
and the models incorporate detailed representation of steel and
concrete elements as well as the panel zone components. A
number of studies are then carried out in order to examine
the influence of several key loading, geometric and design
parameters on the performance of composite moment frames.
1817
The nonlinear-static behaviour of a typical composite
frame is firstly assessed by comparing the response obtained
from triangular and uniform pushover loading patterns. The
validity of adopting nonlinear-static approaches for this type
of structure is evaluated by comparison against the results
of incremental dynamic analysis. For this purpose, seven
natural earthquake acceleration records, which are selected
and adjusted for compatibility with the design spectrum, are
utilised. It is shown that the results of the dynamic time-
history analysis are closely related to the nonlinear-static
response adopting triangular loading pattern, following the
trends observed in other frame types of regular configurations.
The suitability of pushover analysis with triangular loading,
in comparison with the uniform pattern, is supported by the
overall structural response as well as local inter-storey drift
distributions.
A degree of global overstrength is normally expected in
framed structures, particularly when relatively high behaviour
factors are adopted, since member sizes are often governed
by gravity considerations or by deformation-related criteria.
This overstrength becomes even more pronounced in composite
frames, due in-part to the typically large spans which increase
the dependence of beam sizes on gravity loading conditions.
The frames examined in this study exhibited strength levels
consistently exceeding four-fold the assumed design base shear.
The global overstrength is also contributed to by the significant
1818 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
redistribution in terms of the ultimate-to-yield resistance ratio,
referred to as αu /α1 in EC8. This ratio was higher than 1.5 in
all the study frames (which is well above the recommended
code value of 1.3) provided that local storey mechanisms are
prevented and a realistic level of gravity loading is considered.
It is shown in this paper that several design parameters
and assumptions have direct implications on the inelastic
behaviour of composite frames, as assessed through the overall
lateral response, inter-storey drift distribution and plastic hinge
patterns. In particular, the choice of behaviour factor in
relation to the seismic intensity, the incorporation of moment
redistribution under gravity conditions, and the panel zone
effects are all issues that merit careful consideration in design. It
is also shown that a number of geometric parameters, related to
the structural configuration, including frame spacing and beam
span, have a significant influence on the behaviour.
The plastic hinge patterns observed in the study frames
provide useful insight into the behaviour. In most cases, except
in frames with relatively short beam spans or when significant
panel zone yielding is permitted, hinges form primarily
on one side of the beam spans under pushover loading.
This effect is particularly significant in composite frames
due to the characteristic asymmetry in beam cross-section
capacity, coupled with the influence of gravity moments. More
importantly, local storey mechanisms did not occur in the
study frames except in the case of the three-storey structure.
This undesirable performance, to which low-rise and/or large
span configurations are particularly vulnerable, highlights the
need for careful interpretation or modification of the suggested
capacity design provisions.
References
[1] Broderick BM, Elnashai AS. Seismic response of composite frames:
Response criteria and input motion. Engineering Structures 1996;18(9):
696–706.
[2] Leon RT. Analysis and design problems for PR composite frames
subjected to seismic loads. Engineering Structures 1998;20(4–6):364–71.
[3] Spacone E, El-Tawil S. Nonlinear analysis of steel-concrete composite
structures: State of the art. Journal of Structural Engineering 2004;130(2):
159–68.
[4] Leon RT, Hajjar JF, Gustafson MA. Seismic response of composite
moment-resisting connections. I: Performance. Journal of Structural
Engineering 1998;124(8):868–76.
[5] Hajjar JF, Leon RT, Gustafson MA, Shield CK. Seismic response
of composite moment-resisting connections. II: Behaviour. Journal of
Structural Engineering 1998;124(8):877–85.
[6] Bursi OS, Caldara R. Composite substructures with partial shear
connection: Low cycle fatigue behaviour and analysis issues. In: 12th
world conference on earthquake engineering. 2000.
[7] Plumier A, Doneux C, Bouwkamp JG, Plumier C. Slab design in
connection zones of composite frames. In: Proceedings of the 11th ECEE
conference. 1998.
[8] Elnashai AS, Broderick BM. Seismic response of composite frames:
Calculation of behaviour factors. Engineering Structures 1996;18(9):
707–23.
[9] Thermou GE, Elnashai AS, Plumier A, Doneux C. Seismic design
and performance of composite frames. Journal of Constructional Steel
Research 2004;60(1):31–57.
[10] CEN. EN 1998-1, Eurocode 8: Design provisions for earthquake
resistance of structures, Part 1: General rules, seismic actions and rules
for buildings. Brussels: European Committee for Standardization; 2004.
[11] CEN. EN 1993-1-1, Eurocode 3: Design of steel structures — Part 1.1:
General rules and rules for buildings, Brussels: European Committee for
Standardization; 2005.
[12] CEN. EN 1994-1-1, Eurocode 4: Design of composite steel and concrete
structures — Part 1.1: General rules and rules for buildings. Brussels:
European Committee for Standardization; 2004.
[13] Bursi OS, Sun FF, Postal S. Non-linear analysis of steel-concrete
composite frames with full and partial shear connection subjected to
seismic loads. Journal of Constructional Steel Research 2005;61(1):
67–92.
[14] CEN. EN 1991-1-1, Eurocode 1: Actions on structures — Part 1.1:
General actions — Densities, self-weight, imposed loads for buildings,
Brussels: European Committee for Standardization; 2002.
[15] Elghazouli AY. Assessment of capacity design approaches for steel
framed structures. Journal of Steel Structures 2005;5(5):465–75.
[16] Izzuddin BA. Nonlinear dynamic analysis of framed structures. Ph.D.
thesis. University of London; 1991.
[17] Izzuddin BA, Elnashai AS. Adaptive space frame analysis. 2 — A
distributed plasticity approach. Proceedings of the Institution of Civil
Engineers — Structures and Buildings 1993;99(3):317–26.
[18] Elnashai AS, Elghazouli AY. Seismic behaviour of semi-rigid steel
frames. Journal of Constructional Steel Research 1994;29:149–74.
[19] Song L, Izzuddin BA, Elnashai AS, Dowling PJ. An integrated adaptive
environment for fire and explosion analysis of steel frames — Part 1:
Analytical models. Journal of Constructional Steel Research 2000;53(1):
63–85.
[20] Izzuddin BA, Song L, Elnashai AS, Dowling PJ. An integrated adaptive
environment for fire and explosion analysis of steel frames — Part II:
Verification and application. Journal of Constructional Steel Research
2000;53(1):87–111.
[21] Elghazouli AY, Izzuddin BA. Realistic modeling of composite and
reinforced concrete floor slabs under extreme loading: Verification and
application. Journal of Structural Engineering 2004;130(12):1985–96.
[22] Castro JM, Elghazouli AY, Izzuddin BA. Modelling of the panel zone in
steel and composite moment frames. Engineering Structures 2005;27(1):
129–44.
[23] Castro JM. Seismic behaviour of composite moment-resisting frames.
Ph.D. thesis. University of London; 2006.
[24] Bursi OS, Caramelli S, Fabbrocino G, Molina J, Salvatore W, Taucer F,
Zandonini R. 3D full-scale seismic testing of a steel-concrete composite
building at ELSA, Technical report EUR 21299 EN. Joint Research
Centre; 2004.
[25] Castro JM, Elghazouli AY, Izzuddin BA. Assessment of effective slab
widths in composite beams. Journal of Constructional Steel Research
2007;63(10):1317–27.
[26] Elghazouli AY, Elnashai AS. Performance of composite steel/concrete
members under earthquake loading: Parametric studies and design
considerations. Earthquake Engineering & Structural Dynamics 1993;
22(4):347–68.
[27] SEAOC. Vision 2000 — Performance based seismic engineering
of buildings. Sacramento (CA): Structural Engineers Association of
California; 1995.
[28] FEMA. FEMA 356 — Prestandard and commentary for the seismic
rehabilitation of buildings. Washington (DC): Federal Emergency
Management Agency; 2000.
[29] Mwafy AM, Elnashai AS. Static pushover versus dynamic collapse
analysis of RC buildings. Engineering Structures 2001;23(5):407–24.
[30] Krawinkler H, Seneviratna GDPK. Pros and cons of a pushover analysis
of seismic performance evaluation. Engineering Structures 1998;20(4-6):
452–64.
[31] Gupta A, Krawinkler H. Seismic demands for performance evaluation
of steel moment resisting frame structures, John A. Blume earthquake
engineering Center report No. 132. Department of Civil Engineering,
Stanford University; 1999.
[32] Bommer JJ, Acevedo AB. The use of real earthquake accelerograms as
input to dynamic analysis. Journal of Earthquake Engineering 2004;8:
43–91.
 A.Y. Elghazouli et al. / Engineering Structures 30 (2008) 1802–1819
[33] Bommer JJ, Scott SG, Sarma SK. Hazard-consistent earthquake scenarios.
Soil Dynamics and Earthquake Engineering 2000;19(4):219–31.
[34] Rey J, Faccioli E, Bommer JJ. Derivation of design soil coefficients (S)
and response spectral shapes for Eurocode 8 using the European strong-
motion database. Journal of Seismology 2002;6(4):547–55.
[35] Alavi B, Krawinkler H. Behavior of moment-resisting frame structures
subjected to near-fault ground motions. Earthquake Engineering &
Structural Dynamics 2004;33(6):687–706.
[36] Bouckovalas GD, Gazetas G, Papadimitriou AG. Geotechnical aspects
of the 1995 Aegion, Greece, earthquake. In: Proceedings of the second
international conference on earthquake geotechnical engineering. 1999.
[37] Lilhanand K, Tseng W. Development and application of realistic
earthquake time histories compatible with multiple-damping design
spectra. In: Proceedings of the 9th world conference on earthquake
engineering, vol. II. 1988. p. 819–1324.
[38] Hancock J, Watson-Lamprey J, Abrahamson NA, Bommer JJ, Markatis A,
McCoy E, et al. An improved method of matching response spectra of
recorded earthquake ground motion using wavelets. Journal of Earthquake
Engineering 2006;10(Special issue no. 1):67–89.
1819
[39] Abrahamson NA. Non-stationary spectral matching. Seismological
Research Letters 1992;63(1):30.
[40] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake
Engineering & Structural Dynamics 2002;31(3):491–514.
[41] Antoniou S, Pinho R. Advantages and limitations of adaptive and
non-adaptive force-based pushover procedures. Journal of Earthquake
Engineering 2004;8(4):497–522.
[42] AISC. ANSI/AISC 341-05 — Seismic provisions for structural steel
buildings. Chicago (IL): American Institute of Steel Construction; 2005.
[43] El-Tawil S, Vidarsson E, Mikesell T, Kunnath SK. Inelastic behaviour
and design of steel panel zones. Journal of Structural Engineering 1999;
125(2):183–93.
[44] Krawinkler H, Bertero VV, Popov EP. Shear behavior of steel frame
joints. Journal of the Structural Division (ASCE) 1975;101(11):2317–36.
[45] FEMA. FEMA 350 — Recommended seismic design criteria for new
steel moment-frame buildings. Washington (DC): Federal Emergency
Management Agency; 2000.
[46] Elghazouli AY. Seismic design of steel structures to Eurocode 8. The
Structural Engineer 2007;85(12):26–31.