Journal of Constructional Steel Research 114 (2015) 66–76

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Moment–shear–axial force interaction in composite beams

Brendan Kirkland a,c,⁎, Paul Kim b,c, Brian Uy c, George Vasdravellis d
a
School of Computing, Engineering and Mathematics, Univ. of Western Sydney, Penrith, NSW 2751, Australia
b
Hyder Consulting Pty Ltd, Level 5, 141 Walker St, North Sydney, NSW 2060, Australia
c
Centre for Infrastructure Engineering and Safety, Univ. of New South Wales, Sydney, NSW 2052, Australia
d
Institute for Infrastructure and Environment, Heriot-Watt Univ., Edinburgh EH14 4AS, UK

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

Article history: Composite steel–concrete beams are frequently used in situations where axial forces are introduced. Some
Received 13 March 2015 examples include the use in cable-stayed bridges or inclined members in stadia and bridge approach spans. In
Received in revised form 12 June 2015 these situations, the beam may be subjected to any combination of flexure, shear and axial loads. However,
Accepted 10 July 2015
modern steel and composite construction codes currently do not address the effects of these combined actions.
Available online xxxx
This study presents an analysis of composite beams subjected to combined loading. A finite element model
Keywords:
(FEM) has been developed and the results derived from the model show excellent agreement with existing
Combined actions FEM and experimental results. The effect of compression and tension loads on a member subjected to flexure
Composite beams and shear is also explored. Design models are proposed for estimating the flexure and shear interaction of an
Design model axially loaded member.
Finite element analysis © 2015 Elsevier Ltd. All rights reserved.
Moment–shear interaction

1. Introduction provides buckling resistance to the top flange should it be subjected to

Composite steel–concrete composite beams are one of the most
widely used methods of construction for steel-framed buildings, bridges
and stadia. These beams are ubiquitous structural elements in which a
steel beam and a solid or composite slab are interconnected by shear
connection. These elements act together to resist action effects as a
single structural member [25]. Generally, this connection is achieved
through headed shear connectors welded to the top flange of the steel
beam. The studs resist longitudinal slip and the vertical separation
between the two elements. In continuous or semi-continuous struc-
tures, members are subjected to either positive (sagging) or negative
(hogging) bending moments. The most efficient use of the materials'
strengths occurs when the beam is subjected to positive bending at
the mid-span. In this case, the steel component is subjected to tensile
forces and the concrete component primarily in compression, thus
utilising the favourable attributes of each material. A simple rigid plastic
analysis (RPA) of a section can show that the positive moment capacity
of a member can be increased by as much as 120% over the plain steel
beam through composite action. Baskar and Shanmugam [3] found
that composite action increased the positive moment capacity of a
girder by 132%. While under positive bending, the concrete slab also
⁎ Corresponding author at: School of Computing, Engineering and Mathematics, Univ.
of Western Sydney, Penrith, NSW 2751, Australia.
E-mail addresses: b.kirkland@uws.edu.au (B. Kirkland),
paul.kim@hyderconsulting.com (P. Kim), b.uy@unsw.edu.au (B. Uy),
g.vasdravellis@hw.ac.uk (G. Vasdravellis).
compression and the ductility of the steel allows for the achievement
of high curvatures.
At support locations, where negative bending moments are intro-
duced, the force distribution is the opposite. The concrete component
is treated as being cracked under the tensile load and contributes little
to the ultimate strength. The concrete still aids in the transfer of the
large tensile loads from the reinforcement through developing interac-
tion with the shear connection. The vulnerability of the steel section to
buckling under the compressive load and the relatively low tensile
strength of the concrete combine to significantly reduce the ultimate
moment capacity of the cross-section compared with the positive bend-
ing capacity.
For composite beams that are horizontal and free from any re-
straints, the design of such members simply requires addressing the in-
teraction of flexure, M and vertical shear force, V. Distinct requirements
for the design of beams subjected to flexure and shear are set by the
modern steel and composite construction codes including AS2327.1
[25] and Eurocode 4 [4]. However, despite experimental and analytical
evidence, the design ultimate shear strength of a composite beam is as-
sumed to be that of the steel web only. The shear strength of the con-
crete is neglected unless it can be shown that it contributes to the
ultimate shear strength of the beam. Liang et al. [13,14] showed that
disregarding the contribution of the concrete in assessing the ultimate
vertical shear strength results in conservative and inefficient designs.
Design models for the strength interaction of both sagging and hogging
bending in continuous beams and for sagging bending in simply sup-
ported beams were proposed. Their research showed that the ultimate

http://dx.doi.org/10.1016/j.jcsr.2015.07.011
0143-974X/© 2015 Elsevier Ltd. All rights reserved.
 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76
vertical shear strength of a beam may be increased by approximately
85% compared with current design models by including the shear
strength of the concrete slab. These results were later verified and re-
vised through experimental and numerical studies carried out by
Vasdravellis and Uy [31]. The design models previously proposed by
Vasdravellis and Uy [31] and Liang et al. [13] for sagging and hogging
bending respectively are:

3
6
M V
þ ≤1 ðSagging bendingÞ
M uo V uo

0:6
6
M V
þ ≤1 ðHogging bendingÞ
M uo V uo
where Muo and Vuo are the flexural and shear capacity of a member
respectively. Vasdravellis and Uy [31] also proposed an equation for
the shear capacity of a member which includes the contribution by
the concrete slab:
V uo ¼ V pl:Rd þ V slab
where Vpl.Rd is the shear capacity of the steel web as determined by AS
4100 [24] and Vslab is given by:
 0:7 qffiffiffiffi0ffi
V slab ¼ φs f ðλsd Þ b f Dslab f c:
Further information on the equations proposed can be found in the
relevant paper.
Composite beams often can be used in situations in which an axial
load may be introduced into the member. Some examples include the
use in cable-stayed bridges or inclined members in stadia and bridge
approach spans. The necessity to transfer diaphragm forces due to
wind and seismic loads will introduce an axial load into beams used in
floor systems for braced multi-storey buildings. Continuous members
may also incur axial loads due to thermal expansion or contraction of
materials and the restriction of their longitudinal displacement at the
supports. In these situations, the beam may be subjected to any combi-
nation of flexure, shear force and axial load, N (Fig. 1). However, modern
steel and composite construction codes currently do not address the ef-
fects of these combined actions.
The effect of axial load introduced through the installation of
prestressing cables was researched by Troitsky et al. [26], Saadatmanesh
et al. [21,22] and Ayyub et al. [1,2]. Similarly, Uy and Craine [28] and
Fig. 1. Member subjected to flexure, shear and axial loads.
67
Lorenc and Kubica [18] compared conventional steel–concrete compos-
ite beams with beams post-tensioned using steel prestressing cables.
Their studies noted a 15% and 25% increase in sagging flexural strength
due to the combined action of axial compression respectively. Chen and
Gu [7] studied prestressed beams subjected to positive moment and ob-
served an increase in the sagging flexural strength of 84% with a further
increase of 7% observed by using a draped tendon. Chen [6] tested
prestressed composite beams subjected to hogging moments. It was
found that the addition of the external tendons significantly increased
ð1Þ
the cracking moment resistance of the beams while only slightly lower-
ing its yield moment. Chen et al. [8] later tested two-span and three-
span continuous beams with post-tensioning tendons finding an 18% in-
ð2Þ
crease in the sagging moment capacity and a 262% increase in the crack-
ing moment at the supports. However, in all of these studies utilising
post-tensioning cables, the specimens were subjected to no more than
approximately 15% of the axial compressive strength.
Uy and Bradford [27] employed a cross sectional analysis method
(CSA) for their prestressed composite beam model. This extended
upon previous research by including a longitudinal discontinuity at
ð3Þ the beam–slab interface to represent the effects of partial shear connec-
tion (PSC). Loh et al. [16,17] also performed experimental and analytical
studies on the effect of PSC in hogging moment regions of composite
beams. It was found that, for beams using lower degrees of shear con-
nection, a significant increase in rotational capacity was achieved,
ð4Þ with only a slight reduction in peak moment resistance. Later studies
also confirmed the results that the ductility of the beam is considerably
increased when partial shear connection is used [19,29].
Shanmugam and Lakshmi [23] completed a thorough review of over
70 papers on steel–concrete composite columns covering both concrete-
encased and infilled sections. This showed the extensive research that
had been undertaken on axially loaded members. The composite col-
umns in these papers, however, like many recent studies including
Elghazouli and Treadway [10] and Dundar et al. [9] were doubly sym-
metric as opposed to a typical composite beam, such as the cross-
section used in this study, which is only symmetrical about its y-axis.
Uy and Tuem [30] were the first to address the effect of tension and
provide a full moment–axial load interaction diagram for composite
beams. A detailed analytical study of composite beams under combined
flexure and axial force was performed by two methods: a CSA and an
RPA. The CSA calculates the moment–curvature response of the com-
posite beam subjected to any combination of sagging or hogging bend-
ing and axial compression or tension. The RPA and CSA in this study
show almost identical results for most cases. The only variation oc-
curred for the sagging bending and axial compression combination
where the RPA results were greater than those of the CSA. This was
caused by the fact that part of the cross-section was not yielded at the
ultimate case, which is contrary to the RPA's fully yielded assumption.
Their model was limited to specimens with FSC and also placed the
axial load at the level of the plastic neutral axis. Thus it does not include
the additional moment induced by an eccentrically placed axial load or
its effect on the load carrying capacity.
The authors recently studied the behaviour and design of composite
beams under the four combinations of flexure and axial loads [32–35]
namely: tension and negative bending, tension and positive bending,
compression and negative bending and compression and positive bend-
ing respectively. Each of these studies contains an experimental series, a
finite element analysis and design recommendations for the M–N inter-
action of each respective loading combination. Kirkland and Uy [12]
performed a CSA and extended previous research with revised design
models for all four combinations of flexure and axial loads as well as
the influence of partial interaction. The proposed design models are
given below.
Quadrant I Sagging bending and compression:
M ≤M uo ; for N ≤0:6 Nuo ð5Þ
68 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76

Quadrant IV Hogging bending and compression

N M
þ ≤1 ð11Þ
N uo M uo

where Nuo is the axial capacity of the member.

Expressions for the interaction of flexure, shear and axial load were
developed by Liu et al. [15] and Huang et al. [11] for steel members and
reinforced concrete beams respectively. However, the interaction of
Fig. 2. Finite element model and elements used for modelling. these combined forces on composite beams remains unexplored.
The study contained herein is the culmination of a larger research
project aimed to determine the effect of axial load on the flexural and
N M shear capacity of composite beams [12,31–35]. This paper presents the
þ 0:4 ≤1; for N N 0:6 N uo : ð6Þ results of a finite element model and proposes design models for the re-
Nuo Muo
duction in flexural and shear capacity of a member in the presence of
axial load.
Quadrant II Sagging bending and tension:

N ≤ Nuo ; for M ≤ 0:3 M uo ð7Þ 2. Finite element analysis

N M In order to produce analytical results comparable to the test data, a

0:7 þ ≤1; for M N 0:3 M uo : ð8Þ
Nuo Muo finite element model (FEM) was developed using the commercial finite
element software ABAQUS. The FEM used for this study was predomi-
nantly based on the ABAQUS model utilised by Vasdravellis et al.
Quadrant III Hogging bending and tension
[32–35] to analyse the effects of axial forces on the flexural strength of
composite beams. The cross sectional dimensions and component
M ≤ Muo ; for N ≤ 0:2 Nuo ð9Þ
sizes were kept identical to the test specimens in order to maintain con-
sistency. Further details on the model geometry, element type, mate-
N M rials and additional features have been provided within the following
þ 0:8 ≤ 1; for N N 0:2 Nuo : ð10Þ
Nuo Muo sections.

Fig. 3. Material stress–strain graphs used in the FEM.

 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76
2.1. Geometry and element types
The concrete slab was modelled using eight-node linear hexahedral
solid elements with reduced integration (C3D8R), which was utilised to
minimise computation time without significantly affecting the accuracy
of the results. The steel beam was modelled using eight-node linear
solid elements incorporating incompatible modes (C3D8I); this was
due to the detection of hourglass modes in the stud regions of the
steel beam, where high stress levels caused excessive distortion of the
element shapes. C3D8I elements were found to be more capable of sim-
ulating the behaviour of composite beams accurately compared with
C3D8R elements in these regions. The longitudinal steel reinforcement
was modelled using three-dimensional two-node truss elements
(T3D2) to accommodate for their larger length to thickness ratio. A
graphical illustration of the FEM is depicted in Fig. 2.
2.2. Interactions
Longitudinal reinforcement was included within the concrete slab
via the embedded element technique available in ABAQUS, enabling
perfect bond to exist between the surrounding concrete (host ele-
ments) and the steel surface (embedded region). It also allowed the re-
inforcement translational degrees of freedom to be constrained with
respect to the behaviour of the concrete slab. Vertical separation be-
tween the steel beam and concrete slab surface was restricted using a
contact interaction constraint, including the “hard” normal contact
with friction property specified between the two surfaces. A slightly
conservative friction coefficient value of 0.4 was used for this study in-
stead of the Eurocode 4 [4] recommended value of 0.5. Also, only half
the beam's span was modelled by imposing a symmetrical boundary
condition at the beam's mid-span cross section as annotated in Fig. 2.
A boundary condition preventing lateral movements of the FEM was
employed for analysing beams under a combination of axial loading
and hogging moment, thus effectively modelling the fly bracings used
in the experimental set-up by Vasdravellis et al. [32,34].
2.3. Material properties
The material stress–strain data used for the steel parts of the
ABAQUS model consisted of stress–strain curves obtained from the
steel material tests carried out by Vasdravellis et al. [33]. A logarithmic
and a simplified piecewise linear function was adopted for structural
steel and reinforcement steel respectively. The Young's modulus used
for the steel was 200 GPa and the yield stress values of 350 MPa and
540 MPa were used for the steel beam and reinforcement respectively.
Mises plasticity was defined with isotropic hardening law for all steel
parts of the FEM.
The model utilised for representing the stress–strain behaviour of
concrete is proposed by Carreira and Chu [36]. Flexural splitting test re-
sults were used to produce a bilinear tensile stress–strain curve which
peaked at a maximum stress value of 3.4 MPa and reached a zero stress
level at a strain of 0.1 incorporating the tension stiffening option
Fig. 4. Shear-span and loading variations.
69
Fig. 5. Web and flange stiffeners.
available in ABAQUS. This large value of strain at zero stress aims to
avoid early termination of the analyses due to numerical problems asso-
ciated with concrete in tension. In order to maintain consistency with
the original models [32–35], the damaged plasticity model was used
instead of the smeared cracked model to capture the behaviour of con-
crete when it was experiencing full inelasticity. The damaged plasticity
model was generally used for analysing concrete structures under static
or dynamic loadings more realistically despite being more computa-
tionally expensive compared to its companion model.
The most accurate way to model shear connectors in ABAQUS is to
use solid elements and incorporate contact interaction conditions be-
tween the concrete slab and the steel beam. However, this task is
time-consuming and may also lead to numerical convergence problems
during the simulations. The nonlinear spring model (SPRING 2 ele-
ments) utilised in the original beam model [33] was retained for this
study. The data obtained from three push-out tests for the shear studs
were used to create a Load vs. Slip curve with an average load capacity
of 110 kN per stud. This was incorporated into the model in the form
of a multi-linear function. The corresponding stress–strain graphs for
the above-mentioned materials have been provided in Fig. 3.
2.4. Loading and solution method
Vertical and horizontal loads were simulated in the form of an im-
posed controlled concentrated mid-span displacement mechanism
and an axial edge pressure at the beam-end respectively. Sagging and
hogging moments were induced by applying the vertical load on the
concrete slab and the steel beam flange respectively. Five different mag-
nitudes of edge pressure were simulated for the analyses (approximate-
ly equal to 0, 1, 2, 4 and 8 MPa corresponding to zero, low, moderate,
high and very high axial loading). Only the steel beam's area was
Fig. 6. First buckling mode for 400 mm shear-span beam. (a) Sagging moment.
(b) Hogging moment.
70 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76

Fig. 7. Moment–deflection curve comparison. (a) Sagging moment–shear. (b) Hogging moment–shear.

subjected to axial loading, as large pressures applied to concrete ele-
ments were shown to disrupt solution convergence procedures during
the simulations.
In order to vary the level of shear force imposed on the beams, seven
different beam span cases were analysed while being simultaneously
subjected to both vertical and horizontal loads (shear-spans of 2000,
1600, 1200, 800, 600, 400 and 200 mm respectively). The spans were re-
duced by moving the beam's support mechanism (boundary condition
modelling a pinned support) inwards accordingly. This approach was
taken to avoid extra partitioning and re-meshing of the FEM, as well
as other complexities associated with re-modelling the beam. Hence,
thirty six sets of results were able to be produced for each quadrant of
the axial force-bending moment interaction diagram, each reflecting
the beam's ultimate state while exposed to axial tension or compres-
sion, shear force and sagging or hogging bending moment simulta-
neously. Fig. 4 illustrates the loading conditions considered for each
quadrant of this study. The ‘Free-Body cut’ tool in ABAQUS was
employed to evaluate the axial force and the bending moment at the
beam's support and mid-span region respectively (the ‘Free-Body cut’
tool utilises a section integration scheme to sum up the elemental
stresses present at a designated cross-sectional cut and calculates the
corresponding force and bending moment).
2.5. Additional features
In order to prevent the occurrence of premature failure, two web-
side plates were provided at the axial-load end of all beams to eliminate
localised buckling. For the analyses of beams under a combination of
hogging moment and very high axial compression, flange stiffeners
and extra plate parts were included to prevent premature buckling of
the steel beam and ensure that the ultimate capacity of the beam was
able to be obtained (Fig. 5).
A number of additional features were included for the analysis of
short beams with significantly higher levels of shear (600 mm,
400 mm and 200 mm shear-span beams): a linear buckling analysis
was carried out for the composite beam model and the first buckling
mode was assigned to produce more realistic deformation shapes
(depicting diagonal web buckling as shown in Fig. 6). Equivalent geo-
metric imperfection values, representing a local imperfection in the
steel web panel, were assigned in accordance with Eurocode 3 Part 1.5
Annex C — Clause C.5 [5]; therefore imperfection values of 1.036,
1.036 and 1 mm were calculated and used for the three beams respec-
tively. Additional transverse web stiffeners were also included directly
above the support and mid-span region to prevent premature local
buckling of the steel web.

Fig. 8. Moment–shear interaction comparison.

 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76 71

Fig. 9. Moment–axial force interaction diagram.(a) Compression case.(b) Tension case.

In order to maintain the full shear connection (FSC) assumption for
all simulations, the stiffness of the nonlinear spring located at the
beam-end (acting as a group of shear-studs) was raised accordingly,
as the degree of shear connection is directly proportional to the number
of springs and its strength capacity. For example, as the stiffness of an in-
dividual stud (spaced every 200 mm) was 200 kN/mm and the end-stud
was assigned a stiffness value of 400 kN/mm for the 4000 mm beam, the
end-stud stiffness was raised to 800 kN/mm for the 3200 mm beam
(substituted by the removal of two internal springs) and 1200 kN/mm
for the 2400 mm beam and so on. Partial shear connection (PSC) was
modelled with all spring stiffness values halved to produce a degree of
shear connection value between 0.5 and 0.6 when necessary. To further
ensure that no severe pull-out failure occurred with FSC, the slip in each
spring was monitored during every simulation to observe whether it
was approaching the stud's 5% characteristic slip capacity as given by
Oehlers and Bradford [20], calculated to be equal to 5.51 mm.
2.6. Failure criteria
To account for axial force, shear force and bending moment at the
beams' ultimate state, multiple failure modes are needed to be consid-
ered during the simulations. Thus, it was necessary to define limiting
values of some parameters during the analyses to clearly identify the ul-
timate state. Failure was considered when one or more of the following
criteria were satisfied:
• Structural steel fracture: Tensile strain values in the steel beam exceed
the ultimate plastic strain, corresponding to 0.2 for this study,
• Reinforcement fracture: Tensile strain values in the reinforcement
steel exceed the ultimate plastic strain, corresponding to 0.1 for this
study,
• Local buckling: Severe localised deformation of the flange or web of
the steel beam identified using the ABAQUS visualisation module,

Fig. 10. M–V interaction comparison for sagging bending. (a) Tension case. (b) Compression case.
72 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76

Fig. 11. M–V interaction comparison for hogging bending. (a) Low tension (600 mm SS). (b) Moderate compression (400 mm SS).

• Shear connection failure: Recorded slip value is close to (within 10%)
or exceeds the failure slip value of 5.51 mm as recorded by the NL
springs,
• Shear failure: Diagonal web buckling and severe concrete cracking
identified using the visualisation module,
• Ductile failure (concrete crushing): No other failure modes are identi-
fied and the recorded maximum compressive strain in the concrete
slab exceeds the limiting strain value of 3500 με in compression,
• Deflection limit: When all failure modes were unable to be identified
and the beam continued to deflect vertically, the ultimate capacities
were recorded at a deflection value corresponding to 6% of the beam
span.
2.7. Assessment of the numerical model
The FEM was validated against existing analytical and experimental
data to ensure that all results obtained using the model were acceptably
accurate. It was essential to ensure that the results produced using the
FEM showed good agreement with previous work for the interaction
of sagging moment and shear, hogging moment and shear and moment
and axial forces as carried out by Liang et al. [13], Vasdravellis and Uy
[31] and Vasdravellis et al. [32–35] respectively.
Initially, moment–deflection curves were produced from simula-
tions of the FEM under pure sagging and hogging moment (Fig. 7).
The curves were compared with test results obtained by Vasdravellis
et al. [33,34] and the FEM was shown to behave in close agreement
with the test specimens with the initial stiffness and post-elastic behav-
iour being captured accurately. Simulation results obtained for all beam
spans only under a vertical mid-span loading yielded seven data points
which were able to be plotted and compared against previously pub-
lished research data and moment–shear interaction design models as
seen in Fig. 8.
The FEM was shown to produce a slightly more conservative sagging
moment–shear interaction curve as seen in Fig. 8(a). This difference is
due to an under-estimation of the ultimate bending moment capacity
for beams with shear-spans shorter than 1000 mm. Also, the pure
shear capacity under hogging bending was slightly over-estimated as
shown by the incomplete convergence of the results with the design
model proposed by Liang et al. [13] in Fig. 8(b). This difference is likely
to have been attributed to the use of solid elements in the FEM. Liang
et al. [13] and Vasdravellis and Uy [31] both used shell elements to
mesh their FEMs in order to maximise the accuracy of the results obtain-
ed. Despite these issues, the general trend and interaction behaviour
depicted by the results obtained from the FEM was in good agreement
with the analytical and experimental results produced by both studies
[13,31]. It was evident that the FEM can predict the ultimate bending

Fig. 12. Failure modes predicted by the FEM. (a) Compression vs shear force. (b) Tension vs shear force.
 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76
Fig. 13. Axial load vs shear force interaction.
and shear capacity of the composite beam section with reasonable accu-
racy when subjected to a combination of bending moment and shear
force.
The moment–axial force interaction validation was completed by
plotting the results obtained from relevant simulation runs of the FEM
against results produced by Kirkland and Uy [12]. Quadrant I and II sim-
ulations were carried out for both PSC and FSC cases as past experiments
were conducted using specimens with a degree of shear connection of
0.6 while this study was focused on the behaviour of composite beams
with FSC. As shown in Fig. 9, the behaviour of composite beams under
a combination of sagging or hogging moment with tension was in
close agreement with the trends identified from previous studies. The
pure compressive axial capacity of the FEM was shown to be 25%
lower capacity than that obtained via tests and this was attributed to
the steel beam only being axially loaded during the simulations, hence
both the ultimate tensile and compressive capacity of the composite
section was limited to the strength of the steel section. This affected
the flexural capacity of the beam for quadrants I and II, as premature
shear connection failure was predicted to occur for high and very high
levels of axial compression. Despite the discrepancies mentioned
above, the FEM was still clearly capable of predicting the behaviour of
the composite beam under all combinations of bending moment and
axial forces with reasonable accuracy. Considering the complexity of
the beam behaviour being modelled in accordance with the various
combined loading scenarios, the FEM was deemed suitable for further
use in this study.
3. Results and discussion
In order to quantify the moment–shear interaction curves for vary-
ing levels of axial loads, all beam spans were tested under a combination
of vertical and horizontal loading as described in Section 2.4. The follow-
ing sections will discuss the analytical results obtained from this study:
3.1. Quadrant I — sagging moment and compression
All beams with a shear-span greater than or equal to 800 mm were
shown to experience ductile failure with increasing axial compression,
concrete crushing occurred more rapidly and the beams became more
prone to premature shear connection failure. The 200–600 mm shear-
span beams were observed to fail by shear as indicated by high stress
levels present in the steel web panel along with severe diagonal web-
buckling. Other common observations included local buckling of the
mid-span-flange for high and very high axial compression cases and ex-
cessive plastic steel strain values above 0.15 near the connector regions.
The FEM results presented in Figs. 8(a) and 10(a) indicate that when
the shear force present is less than 0.6 Vuo, the flexural capacity may be
slightly increased when exposed to compressive forces less than or
73
equal to 0.6 Nuo. For shear forces greater than 0.6 Vuo, the flexural capac-
ity drops rapidly with increasing axial compression. The experimental
results were shown to have lower N/Nuo values compared to the FEM
data; this is attributed to the tests being conducted for beams with
PSC, hence a faster deterioration of the beam's flexural and shear capac-
ity for lower axial loads.
3.2. Quadrant II — sagging moment and tension
The results showed that 800–2000 mm shear-span beams, subjected
to low to high levels of axial tension, experienced ductile failure due to
concrete crushing at the mid-span region. However, for all 200 mm and
400 mm shear-span beam simulations, the ultimate capacity was
governed by shear failure. A change in failure mode was identified for
the 600 mm shear-span beam from shear-failure to ductile failure
when axial tension was introduced into the steel beam. This is mainly
attributed to tensile forces restricting the initiation of diagonal web
buckling in the steel web panel enabling the beam to continually deflect
downwards and undergo ductile failure. For all beams exposed to very
high axial tension (with the exception of the 200 mm shear-span
beam), the governing failure mode was steel fracture accompanied by
severe concrete cracking in the stud regions.
The FEM results seen in Figs. 8(a) and 10(b) show a slight increase in
the bending moment capacity (around 5%) with V/Vuo less than 0.6 be-
fore a rapid decline in the flexural capacity. This V/Vuo value where the
flexural capacity first begins its rapid decline is shown to change in an
almost linear manner with increasing axial load. The FEM results are
in good agreement with the experimental results.
3.3. Quadrant III — hogging moment and tension
The results of the analyses showed that 800–2000 mm shear-span
beams were predicted to experience very large vertical deflections.
Fig. 14. Composite beam section.
74 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76
Fig. 15. Design example. (a) Compression case. (b) Tension case.
This is attributed to the opposing effects between the axial tension and
the vertical mid-span load where the tension load offsets the buckling of
the steel flange. For beams subjected to low/moderate axial tension, the
longer beams showed signs of local buckling accompanied by large plas-
tic strain values in the steel reinforcement. However, the ultimate state
was not governed by buckling or the expected fracture of the reinforce-
ment, rather the deflection limit previously described. The deflection
limit did not govern when these beams were exposed to very high
axial tension but rather, the reinforcement fracture failure mode was
found to be more common. The 200–600 mm shear-span beams were
not affected by the deflection limit failure criteria. These beams failed
by shear, often accompanied by local buckling of the top flange in the
mid-span region for low to moderate levels of axial tension (Fig. 12a).
The FEM results shown in Fig. 11(a) show an almost bilinear rela-
tionship for the M–V interaction, with the flexural capacity reduction
becoming severe for shear forces greater than 0.4 Vuo. This characteristic
behaviour is reasonably similar to the design model proposed by Liang
et al. [13] although some results were shown to lie beyond the boundary
of the design curve. This is attributed to the over-estimation of the
beam's pure shear capacity in the presence of hogging moment, as
discussed in Section 2.7. The FEM results were in close agreement
with the experimental results, however it must be noted that the ex-
perimental series did not consider the effect of shear and did not
achieve V/Vuo values greater than 0.21. Similar to quadrant II, the CSA
Fig. 16. Design models for sagging bending.(a) Tension case.(b) Compression case.
results indicated that the beam section had a slightly higher flexural
resistance for a given axial tension load than the capacities predicted
by the FEM.
3.4. Quadrant IV — hogging moment and compression
The results indicated that most beams were to experience failure
due to local buckling, although a shear-dominant failure mode ac-
companied by local buckling was shown to be more likely for the
200–600 mm shear-span beams. This is shown in Fig. 12(b) for the fail-
ure of a 400 mm shear-span beam under moderate axial compression.
For low to moderate axial compression, local buckling was observed in
the top flange of the mid-span region for the 800–2000 mm beams.
This was occasionally accompanied by a slight deformation of the
web. However, when the compressive loads were high, local buckling
occurred more regularly in the vicinity of the quarter-span points,
with both the flange and web undergoing critical deformations.
Similar to quadrant III, the FEM results in Fig. 11(b) show an almost
bilinear relationship for the M–V interaction with the flexural capacity
reduction becoming severe for shear forces greater than 0.4 Vuo. This
V/Vuo value where the flexural capacity first begins its rapid decline is
shown to change in an almost linear manner with increasing axial
load similar to the results from quadrant II. Comparison of the ex-
perimental and CSA results from Kirkland and Uy [12] and FEM results
indicates a good agreement. However, similar to Quadrant I, the exper-
iments were carried out with axial loading imposed on both the steel
beam and the concrete slab resulting in a higher Nu value than the
FEM was able to achieve.
4. Design models
Design models for the interaction of moment and shear force in com-
posite beams have been proposed by Liang et al. [13] and Vasdravellis
and Uy [31] for hogging bending and sagging bending respectively.
Kirkland and Uy [12] recently proposed equations for the interaction of
moment and axial load extending upon the work of Vasdravellis et al.
[32–35]. Combining these models for the third dimension first requires
knowledge of the effect of axial load on the shear capacity of a member.
The FEM results show that decreasing the shear force linearly with the in-
creasing axial load can produce a mostly conservative interaction model
for both compression and tension (Fig. 13).
The simplest way to approach design models for the interaction of
simultaneous moment, shear and axial load is to utilise the existing
 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76 75

Fig. 17. Design models for hogging bending.

equations for M–V [13,31] and reduce Muo and Vuo with the increasing Using this method, design models for the four combinations of mo-
level of axial load. i.e. reduce Muo according to the design models pro- ment, shear and axial load are illustrated in Figs. 16 and 17. With
posed by Kirkland and Uy [12] and reduce Vuo linearly proportionally comparison to the FEA, 93% of the results exceed the strength esti-
to N/Nuo (Fig. 13). The shear capacity, Vuo may be calculated by using mated by the proposed design models. These results exceeded the
Eq. (3) proposed by Vasdravellis and Uy [31]. Muo and Nuo may be calcu- proposed design models by an average of 20% indicating some un-
lated simply by RPA. derestimation of strength in the design models. This may be partially
The specimen used during the experiments by Vasdravellis et al. [32] due to the conservative linear estimate of the N–V interaction
(Fig. 14) may be used as an example of the design model for a beam sub- (Fig. 13). The few results that did not achieve the strength predicted
jected to tension, shear force and hogging bending. An RPA is used to de- by the design model strength all occur when large axial compression
termine the tensile capacity, Nuo = 1428 kN and the hogging moment is present. This is expected due to limitations in the FEM previously
capacity, Muo = 134 kNm. Using Eq. (3), Vuo is calculated as 475 kN. Sup- discussed in Section 2.7.
pose design loads on the section are M = 70 kNm (M/Muo = 0.52) and
V = 200 kN (V/Vuo = 0.42). I.e. short span and large shear loading. With
no axial load applied, Eq. (2) is used and the loads are calculated to be 5. Conclusions and further research
within the design model:
The finite element model developed herein was shown to provide

0:6

70 200 6 accurate estimates of the behaviour and strength for composite beams
þ ¼ 0:68 ≤ 1: ð12Þ
134 475 subjected to the combination of flexure, shear and axial loads. The effect
of axial load on the moment–shear interaction for each of the four com-
With an axial load of say, 500 kN applied, Muo and Vuo are reduced. binations of flexure and axial loads has been determined.
The reduction is determined by N/Nuo = 0.35 (35% of the axial capacity). The RPA has been shown by previous researchers to be a feasible de-
Vuo is reduced by 35% to 308 kN and Muo is reduced according to the de- sign tool for axially loaded members. Use of an RPA in combination with
sign model proposed by Kirkland and Uy [12] in Eq. (10) for hogging the equations for Vuo and the M–V interaction provided by previous re-
bending and tension: searchers, design models have been proposed for the interaction of si-

 multaneous moment, shear and axial load. These design models are
M simple, easy to apply and capable of producing generally conservative
0:35 þ 0:8 ≤1 →M ≤ 108: ð13Þ
134 estimates of the M–N interaction of a member subjected to axial load.
These models are considered suitable for inclusion in future internation-
Therefore M is reduced to 108 kNm. Again, the loads are within the al design standards of composite structures.
design model: The proposed design models are based on experiments and FEM

0:6
 analyses using composite sections with relatively large slab to beam as-
70 200 6 pect ratios and compact steel sections, which are commonly used in
þ ¼ 0:84 ≤ 1: ð14Þ
108 308 buildings. More tests and numerical analyses should be done to safely
generalise the results with an extended parametric study. This should
With an axial load of say 800 kN (0.56 Nuo), Muo is reduced using include composite girders having different slab to beam aspect ratios
Eq. (10) to 73 kN and Vuo is reduced by 56% to 208 kN. The interaction and slenderness, which are commonly used in bridges.
is now greater than the proposed design model:

0:6

70 200 6 Acknowledgements
þ ¼ 1:74 N 1: ð14Þ
73 208
The authors wish to acknowledge the Australian Research Council
The design loads are plotted in Fig. 15 along with curves for Eq. (2) (ARC) Discovery Project DP0879734 for the funding provided to
adjusted to include the applied axial loads given above. The same meth- complete the experimental and analytical work involved in this
od can be applied to the other combinations of flexure and axial loads. project.
76 B. Kirkland et al. / Journal of Constructional Steel Research 114 (2015) 66–76
References
[1] B.M. Ayyub, Y.G. Sohn, H. Saadatmanesh, Prestressed composite girders. I: experi-
mental study for negative moment, J. Struct. Eng. 118 (10) (1992) 2743–2762.
[2] B.M. Ayyub, Y.G. Sohn, H. Saadatmanesh, Prestressed composite girders. II: analytical
study for negative moment, J. Struct. Eng. 118 (10) (1992) 2763–2783.
[3] K. Baskar, N.E. Shanmugam, Steel–concrete composite plate girders subject to com-
bined shear and bending, J. Constr. Steel Res. 59 (2003) 531–557.
[4] British Standard Institution, Eurocode 4, Design of Composite Steel and Concrete
Structures, Part 1.1 General Rules and Rules for Buildings, British Standard Institu-
tion, London, 2004.
[5] British Standard Institution, Eurocode 3, Design of Steel Structures, Part 1.1 General
Rules and Rules for Buildings, British Standard Institution, London, 2005.
[6] S. Chen, Experimental study of prestressed steel–concrete composite beams with
external tendons for negative moments, J. Constr. Steel Res. 61 (2005) 1613–1630.
[7] S. Chen, P. Gu, Load carrying capacity of composite beams prestressed with external
tendons under positive moment, J. Constr. Steel Res. 61 (2005) 515–530.
[8] S. Chen, X. Wang, Y. Jia, A comparative study of continuous steel–concrete beams
prestressed with external tendons: experimental investigation, J. Constr. Steel Res.
65 (7) (2009) 1480–1489.
[9] C. Dundar, S. Tokgoz, A.K. Tanrikulu, T. Baran, Behaviour of reinforced and concrete-
encased composite columns subjected to biaxial bending and axial load, Build. Envi-
ron. 43 (2008) 1109–1120.
[10] A.Y. Elghazouli, J. Treadway, Inelastic behaviour of composite members under com-
bined bending and axial loading, J. Constr. Steel Res. 64 (2008) 1008–1019.
[11] L. Huang, Y. Lu, C. Shi, Unified calculation method for symmetrically reinforced con-
crete section subjected to combined loading, ACI Struct. J. 110 (2013) 127–136.
[12] B. Kirkland, B. Uy, Behaviour and Design of Composite Beams Subjected to Flexure
and Axial Load, 2015. (in press).
[13] Q.Q. Liang, B. Uy, M.A. Bradford, H.R. Ronagh, Ultimate strength of continuous com-
posite beams in combined bending and shear, J. Constr. Steel Res. 60 (2004)
1109–1128.
[14] Q.Q. Liang, B. Uy, M. Bradford, H.R. Ronagh, Strength analysis of steel–concrete com-
posite beams in combined bending and shear, J. Struct. Eng. 131 (2005) 1593–1600.
[15] Y. Liu, L. Xu, D. Grierson, Combined MVP failure criterion for steel cross-sections, J.
Constr. Steel Res. 65 (2009) 116–124.
[16] H.Y. Loh, B. Uy, M.A. Bradford, The effects of partial shear connection in the hogging
moment regions of composite beams. Part I—experimental study, J. Constr. Steel Res.
60 (2004) 897–919.
[17] H.Y. Loh, B. Uy, M.A. Bradford, The effects of partial shear connection in the hogging
moment regions of composite beams. Part II—analytical study, J. Constr. Steel Res. 60
(2004) 921–962.
[18] W. Lorenc, E. Kubica, Behaviour of composite beams prestressed with external ten-
dons: experimental study, J. Constr. Steel Res. 62 (2006) 1353–1366.
[19] Q.H. Nguyen, M. Hijaj, B. Uy, S. Guezouli, Analysis of composite beams in the hogging
moment regions using a mixed finite element formulation, J. Constr. Steel Res. 65
(2009) 737–748.
[20] D.J. Oehlers, M.A. Bradford, Composite Steel and Concrete Structural Members: Fun-
damental Behaviour, Pergamon Press, Oxford, 1995.
[21] H. Saadatmanesh, P. Albrecht, B.M. Ayyub, Experimental study of prestressed com-
posite beams, J. Struct. Eng. 115 (9) (1989) 2348–2363.
[22] H. Saadatmanesh, P. Albrecht, B.M. Ayyub, Analytical study of prestressed composite
beams, J. Struct. Eng. 115 (9) (1989) 2364–2381.
[23] N.E. Shanmugam, B. Lakshmi, State of the art steel–concrete composite columns, J.
Constr. Steel Res. 57 (2001) 1041–1080.
[24] Standards Australia, AS 4100–1998 Australian Standard: Steel Structures, Standards
Australia International Ltd, 1998.
[25] Standards Australia, AS 2327.1-2003 Australian Standard: Composite Structures,
Part 1: Simply Supported Beams, Standards Australia International Ltd, 2003.
[26] M.S. Troitsky, Z.A. Zielinski, A. Nouraeyan, Pre-tensioned and post-tensioned com-
posite girders, J. Struct. Eng. 115 (12) (1989) 3142–3153.
[27] B. Uy, M.A. Bradford, Cross-sectional deformation of prestressed composite tee-
beams, Struct. Eng. Rev. 5 (1) (1993) 63–70.
[28] B. Uy, S. Craine, Static flexural behaviour of externally post-tensioned steel–concrete
composite beams, Adv. Struct. Eng. 7 (1) (2004) 1–20.
[29] B. Uy, D.A. Nethercot, Effects of partial shear connection on the required and avail-
able rotations of semi-continuous composite beam systems, Struct. Eng. 83 (4)
(2005) 29–39.
[30] B. Uy, H.S. Tuem, Behaviour and design of composite steel–concrete beams under
combined actions, ASCCS'2006 Eighth International Conference on Steel–Concrete
Composite and Hybrid Structures, Harbin, China, August 2006.
[31] G. Vasdravellis, B. Uy, Shear strength and moment–shear interaction in steel–
concrete composite beams, J. Struct. Eng. 140 (11) (2014).
[32] G. Vasdravellis, B. Uy, E.L. Tan, B. Kirkland, The effects of axial tension on the
hogging-moment regions of composite beams, J. Constr. Steel Res. 68 (1) (2012)
20–33.
[33] G. Vasdravellis, B. Uy, E.L. Tan, B. Kirkland, The effects of axial tension on the
sagging-moment regions of composite beams, J. Constr. Steel Res. 72 (2012)
240–253.
[34] G. Vasdravellis, B. Uy, E.L. Tan, B. Kirkland, Behaviour and design of composite beams
subjected to negative bending and compression, J. Constr. Steel Res. 79 (2012)
34–47.
[35] G. Vasdravellis, B. Uy, E.L. Tan, B. Kirkland, Behaviour and design of composite beams
subjected to sagging bending and compression, J. Constr. Steel Res. 110 (2015)
29–39.
[36] D. Carreira, K. Chu, Stress–strain relationship for plain concrete in compression, J.
ACI Struct 82 (11) (1985) 797–804.