Journal of Constructional Steel Research 68 (2012) 20–33

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

The effects of axial tension on the hogging-moment regions of composite beams

G. Vasdravellis ⁎, B. Uy, E.L. Tan, B. Kirkland
School of Engineering & Civionics Research Centre, University of Western Sydney, Sydney, Australia

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

Article history: Structural parts commonly comprised of composite members such as bridge approaches, inclined parking
Received 20 December 2010 ramps and stadium beams, can be subjected to a combination of high axial loads and bending moments. Steel–
Accepted 29 June 2011 concrete composite construction is a popular solution for these types of structures due to the numerous
Available online 31 July 2011
advantages that they offer. Although, current design codes (e.g. Eurocode 4, American code AISC, Australian
codes AS2327 and AS5100) provide rules for the design of composite columns subjected to flexure and axial
Keywords:
Composite beams
load, however the design of composite beams, which are asymmetric in nature under the combined effects of
Negative bending tension and bending, is not yet fully addressed. This paper investigates the ultimate strength of composite
Partial interaction beams under the combined effects of axial tension and negative (hogging) bending moment. An experimental
Finite element analysis programme carried out in the laboratory of the University of Western Sydney comprised of a total of six
specimens representing composite beams and subjected to various levels of axial tension and bending
moment. Ultimate failure modes were identified and the resulting interaction diagrams were compared to the
results of sectional rigid plastic analysis. Following the tests, three-dimensional finite element models were
employed using the ABAQUS finite element software to further investigate the nonlinear behaviour of the
composite beams and extend the experimental observations by studying the effects of parameters such as the
span length and the effect of partial shear connection. Finally, simple design rules and formulae are proposed
for use in engineering practice.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction Composite steel–concrete beams are used in many types of

structures, including steel framed buildings, bridges, stadiums and
The connection of a steel beam to a concrete slab through shear car parks. Composite beams are being increasingly used in situations
connectors produces a composite beam that combines the best where combinations of bending, shear and axial force may be
attributes of each material. The most efficient use of the material's introduced into the beam [1]. This may occur via the installation of
strengths occurs when the beam is subjected to positive bending with post-tensioning cables; by the inclination of the beam such as in stadia
the steel section placed below a concrete section. This places the steel or bridge approaches; in cable-stayed bridges; or by the necessity to
component in tension and the concrete component in compression, transfer diaphragm forces from the floor plate into the shear core of a
making this type of construction both efficient and economical, while building. However, modern steel and composite construction codes,
under positive bending, the concrete slab provides buckling resistance including Eurocode 4 [2], Australian codes AS2327 and AS5100 [3,4]
to the steel in compression. However, at support locations where and American AISC [5], currently do not address the effects of
negative bending moments are introduced, the concrete component is combined actions in a unified fashion.
subjected to tension and the steel section subjected to compression. There has been significant research into the behaviour of composite
The vulnerability of the steel section to buckling and the relatively low beams under both sagging and hogging bending, and several in-
tensile strength of concrete significantly reduce the moment capacity vestigations into axial compression through both analytical and
of the cross section compared with positive bending. The introduction experimental investigations. The effect of pre-stressing on composite
of tensile forces to the beam will increase the tensile stress in the beams under positive bending was studied by Uy and Bradford [6] and
concrete component while reducing the compression in the lower Uy [7]. Chen [8] tested four groups of prestressed steel–concrete
steel, thus making the beam more prone to cracking but lessening the composite beams with external tendons in negative moment regions,
probability of buckling as a failure mode. and the cracking behaviour and the ultimate moment resistances were
investigated experimentally. Liang et al.[9] studied the ultimate strength
of continuous composite beams in combined bending and shear by
using the finite element analysis method. It was shown that the concrete
⁎ Corresponding author at: University of Western Sydney, School of Engineering &
Civionics Research Centre, Locked Bag 1797, Penrith NSW 2751, Australia. Tel.: + 61 2
slab and composite action significantly increase the ultimate strength of
47360119; fax: + 61 2 4736 0054. continuous composite beams. Loh et al.[10,11] conducted an extensive
E-mail address: g.vasdravellis@uws.edu.au (G. Vasdravellis). experimental and analytical study to investigate the behaviour of eight

0143-974X/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2011.06.017
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 21

composite beams subjected to hogging moment. They studied the Sydney. All specimens were constructed with a 600 mm wide and
effects of various parameters such as the degree of shear connection and 120 mm deep concrete slab connected to a 200UB29.8 steel section.
reinforcement ratio on the ultimate capacity and ductility of composite The beam–slab connection was achieved through 19 mm diameter,
beams. Elghazouli and Treadway [12] presented results from a series of 95 mm long headed shear studs welded in a single line along the
tests on partially-encased composite steel–concrete beam–columns. centre of the top flange of the steel beam. The number of the shear
The experimental inelastic behaviour of the specimens under lateral studs provided was adequate to ensure full shear connection between
loading and axial gravity loads was examined. The specimens in their the slab and the beam. The degree of shear connection in hogging
study, however, were symmetrical through both their x and y axes as moment regions of composite beams is defined as the ratio of the
opposed to a typical composite beam which is only symmetrical about shear connection strength provided by the studs to the strength of the
its y axis. Uy and Tuem [13] were the first to look at the effect of tension weakest component (steel reinforcement or steel beam), while the
on composite beams. An analytical study on combined axial load and tensile strength of the slab is neglected. That is:
bending was performed through a cross-sectional analysis and a rigid
plastic analysis. However, the combined effect of composite beams Νss Fstud
β= ð1Þ
under tension and bending remained untested. minfFr ; Fbeam g
The ultimate capacity of steel–concrete composite beams under the
combined effects of axial tension and negative bending moment is where β is the degree of shear connection, Nss is the number of studs
investigated in the present study. Firstly, details of an experimental in the shear span (half span), Fstud is the strength of an individual stud,
investigation into composite beams under combined axial tension and Fr is the axial strength of the reinforcement in the slab, and Fbeam is the
negative bending are provided. Six composite beams of identical cross- axial strength of the steel beam. In the experimental beams Nss = 8,
sections were tested in the laboratory of the University of Western Fstud = 90 kN [2], and min{Fr, Fbeam} = Fr = 250 kN, thus β = 2.88 N 1;
Sydney under different combinations of tension and bending and the therefore, a full shear connection was ensured.
plastic capacities and failure modes were identified. The results were A group of 3 studs was welded to the ends of each of the beams to
compared with the capacities provided by rigid plastic analysis through reduce slip and ensure full utilisation of the reinforcing bars.
the composite section and reasonable agreement was found. The Longitudinal and transverse reinforcement was placed in the concrete
following sections provide details of an extensive numerical study using slab. The composite cross section for all specimens and a plan view of
the finite element method. The tested beams were modelled by a a typical specimen showing the shear stud and reinforcement
nonlinear finite element model in the ABAQUS commercial software. arrangement are shown in Fig. 1. A detailed list of the dimensions of
The three-dimensional model was compared against experimental the tested specimens is given in Table 1.
results and found to be capable of simulating the inelastic behaviour of The beam section was selected to be a class 1 (or plastic) section in
the composite beams with efficiency and to trace the failure modes up to order to be capable of achieving plastic rotations without being
ultimate load levels. Further parametric studies were conducted on susceptible to premature local buckling, according to the current
composite beams with various sections and span-to-depth ratios and Australian structural code [3]. This was achieved by imposing the code
the influence of the partial shear connection was investigated. limits to the slenderness of the steel beam flange and web. Attention
should be paid in a composite section as the depth of the plastic
neutral axis usually is considerably larger than half of the beam depth,
2. Experimental study as the steel reinforcement in the slab leads the plastic neutral axis to
move near the top flange of the steel beam. Thus, the steel web should
2.1. Specimen description be classified taking into account the increased portion which is under
compression [14].
A total of six steel–concrete composite beams with identical cross- As the universal beam bottom flange was subject to compression
section were tested in the Laboratory of the University of Western forces caused by the negative moment it was susceptible to lateral

600.0
4N12

120.0

9.6

188.0 6.4

9.6 134.0

N12 BARS @ 200MM

4N12 BARS @ 120MM
19/95 STUDS CENTRES
CENTRES

600

100 350 400 400 400 400 400 400 400 400 400 350

4500 100

Fig. 1. Composite beam cross-section (dimensions in mm) and plan view of slab reinforcement and shear stud layout.
22 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

Table 1
Specimen details.

Specimens CB1 to CB5 Specimen CB6

Slab size 4500 × 600 × 120 mm 4000 × 600 × 120 mm

Longitudinal 4N12 × 4450 mm 4N12 × 3950 mm
reinforcement
Transversal 550 mm N12 at 400 550 mm N12 at 400
reinforcement
Shear connectors 19 mm headed studs, 95 mm pre-welded length
Number of studs 12 at 400 with end spans 10 at 430 with end spans
per beam of 350 of 395

torsional buckling. Lateral buckling of the steel beam can cause

premature failure of the composite beam and interruption of the test.
As this was not under investigation in the present study, lateral Fig. 3. Steel beam, shear studs, reinforcing steel and formwork.
movement and buckling of the steel beam were minimised by using a
bracing system consisting of five sets of fly bracing located at the ends,
quarter points and at the midspan. The distance of the bracing was
designed according to the current Australian structural code [3], 2.3. Material properties
which gives the maximum distances between two lateral restraints in
order to eliminate the effects of distortional instabilities. The braces Material tests were performed prior to the tests in order to
were designed to withstand 5% of the force within the flange of the determine the actual strength and stiffness of steel and concrete. All
beam acting horizontally at a right angle to the beam. Web stiffeners materials were ordered from the same set, therefore it was assumed
were welded to each side of the beam and the fly bracing was attached that the properties would be uniform. Tensile tests were performed
using grade 8.8 high tensile bolts. The connection to the slab also on coupons cut from the steel web, flange and the reinforcing bars and
utilised grade 8.8 high tensile bolts which were cast in place. This the corresponding stress–strain laws were obtained. The concrete
arrangement can be seen in Fig. 2. tests comprised of standard cylinder compression tests and flexural or
splitting tests to determine the tensile strength of concrete according
to Australian Standards [15]. The resulting values of the material tests
2.2. Fabrication of tested specimens and the specified characteristic strengths of each material can be seen
in Table 2.
All specimens were fabricated in the Structures Laboratory of the
University of Western Sydney. All specimens were cast with the steel 2.4. Instrumentation
beam on the bottom and plywood formwork and shuttering
constructed to ensure stability during the concrete pouring. A Various measurement devices were employed to record the
qualified stud fabricator was employed to install the shear connectors relevant parameters and to obtain the experimental behaviour of
using a stud welding machine. The reinforcing bars were then the beams. An automatic data acquisition system was employed to
positioned using steel chairs and tie wire and the reinforcement automatically record data from all measuring devices including load
strain gauges were attached. The composite beam formwork setup cells, strain gauges and linear potentiometers throughout the test.
can be seen in Fig. 3. Concrete casting was carried out next with the Measurements were recorded once every second. Strain gauges were
casting of each beam specimen and the standard cylinders for material used to measure strains of the steel beam and reinforcing bars and to
tests at the same time. A standard mix of normal weight concrete with determine at what load the components yield. A total of 21 strain
a specified characteristic strength of 25 MPa at 28 days was used. The gauges were used for each beam specimen. Strain gauges were located
concrete was poured using standard levelling and compaction in sets of seven through each cross-section with one set at midspan
techniques. The surface of the wet concrete was then covered with and one set at each quarter point, as shown in Fig. 4. At each section
polythene sheets and cured under wet condition for 14 days. The there was one strain gauge attached to each longitudinal reinforcing
beams were then flipped and fly bracing was attached, as were strain bar, numbered 1 to 4, one each on the top and bottom flanges and one
gauges on the universal beam section. at the centre of the web. The strain gauges at midspan were also
intended to establish the position of the neutral axis and the value of
curvature under the application of load.
Linear potentiometers were used for measuring the deflection of
the beam. These were placed at the midspan and at the quarter points.
The connector slip and interface slip were also measured with linear
potentiometers. The slip was measured at the ends, quarter points and
midpoint, as indicated in Fig. 4.

Table 2
Material test results.

Material strength (MPa)

Specified Measured yield Measured ultimate

Concrete compression (28 days) 25 n/a 24.1

Concrete splitting n/a n/a 4.3
Steel web 320 359 510
Steel flange 320 333 499
Reinforcement bars 500 541 652
Fig. 2. Fly bracing and web stiffeners in place.
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 23

1 2 3

1000 1000 1000 1000

STRAIN SECTIONS 1, 2, 3
GAUGE
B1
LVDT

B2
B3

C1 C2 C3 C4
Fig. 4. Specimen instrumentation.

2.5. Test setup the beam. An overview of the test setup in the laboratory is shown in
Fig. 6.
A combination of load actuators was used in order to produce axial In order to prevent any induced moment created by the
tension loads and bending moments in the composite beam eccentricity of the axial load the centre of the pin had to be located
specimens. The UWS testing laboratory has a test bed for testing at the plastic centroid. Restrictions such as the size of the pin distance
specimens in axial compression consisting of two large universal between beam flanges meant that the pin had to be located at the
columns with crossheads. To create the axial load two 800 kN and centre of the Universal Beam section. The centre of the Universal
160 mm-stroke hydraulic actuators were used, attached to a large Beam section has an eccentricity, e, of approximately 30 mm below
crosshead between the two universal columns. the plastic centroid. This eccentricity imposes initial second order
To apply axial load to the beam a tensile load applicator was positive moments in the beam. Furthermore, the second order
designed. The tensile load applicator was required to withstand a moments are increased as the eccentricity is increased due to the
tensile force of 1600 kN, the maximum capacity of the horizontal beam vertical deflection. This created a problem as the moment
actuators. These load applicators are shown in Fig. 5. A problem within the beams caused by the point load would be opposed by the
encountered was the need for the connection between the tensile axial load. However this problem was overcome by using further
load applicator and the composite beam to be free to rotate. This is calculations to find the actual moment caused by the forces on the
necessary since, as the beam is simply supported, rotation at the beam. The following formula was used to calculate the actual moment
supports is inevitable and resistance to rotation would affect results in the beam:
by causing semi rigid supports, making moment within the beam
very difficult to calculate. To prevent rigidity in the beam to load
applicator connection one large greased pin was used to connect the PL
M= −Pa ðe + ed Þ ð2Þ
tensile load applicators to either end of the composite beam 4
specimen, as shown in Fig. 5. The web of the universal beam section
of the composite beams was reinforced by welding two 10 mm thick
plates to each side of the web in order to prevent any undesired web
yielding. These plates increase the bearing and tear out capacity of

Fig. 5. Load applicator used to apply the axial force. Fig. 6. General view of the test setup.
24 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

where M is the moment in the beam, P is the vertical applied point

load, L is the beam span, Pa is the axial point load, e is the eccentricity
of axial point load, and ed is the eccentricity of axial load due to beam
midspan deflection.

2.6. Loading procedure

Composite beam specimen CB1 was only loaded in bending

therefore a vertical load was applied at a constant rate to produce a
bending moment. The load was increased until failure or until
maximum stroke of the actuator. Specimens CB2 to CB5 were tested
in combined bending and tension therefore the vertical force and the
tensile force were applied in steps of approximately 10% of design
strength. Both loads were increased separately until failure or
maximum stroke of the actuator. Specimen CB6 was only loaded in Fig. 8. Local buckling initiation in the compressive flange of specimen CB1.
tension and in order to reduce the moment created by eccentricity of
the applied axial load the vertical actuator was set up to just touch the the beam through the pin, as shown in Fig. 5, and they impose
beam. This beam was loaded at a constant rate until failure. During additional resistance to the vertical load as the beam deflects. Beams
loading a complete set of readings, including the jack force, strains, CB2 and CB3 have similar load–deflection curves up to the point
and deflections, were recorded by the data acquisition system and where CB3 fails, although specimen CB2 has a considerably higher
simultaneously stored in the computer once every second. Cracking of ductility. Specimen CB4 has a slightly lower initial stiffness and it
the slab was monitored from the sides and the load at which they arrives at the same ductility levels as CB3, whereas CB5 has a
occurred was noted. considerably lower initial stiffness compared to CB1, but it possesses a
ductility similar to the beam CB2. The implications of the tensile axial
3. Experimental results load level acting in the beam together with the negative bending
moment on the failure behaviour of the composite beams are
3.1. Observations and failure modes discussed in the following paragraphs.
Composite beam CB1 was tested in pure negative bending. There
The six composite beams were tested under combined bending was no failure of the shear connection and interface slip was very
moment and axial tension and the ultimate strength achieved by each small, not exceeding 0.5 mm. In the final stage of loading, there was
specimen was determined. The main objective of the tests was to load only a small amount of visible local buckling present in the top flange,
the composite beams until collapse is reached by failure of one or which can be seen in Fig. 8, indicating that both the shear connection
more components. The lateral torsional buckling of the compressive and the fly bracing system performed well. The first cracking of the
flange was prevented by the use of lateral restraints, as described in concrete slab occurred at a load approaching 60 kN, which is about
Section 2.1, thus the ultimate capacity of each beam was determined 38% of the ultimate load, and cracking continued to enlarge until the
by one of the four possible failure conditions: 1) excessive yielding or ultimate load, as seen in Fig. 9. The beam specimen behaved linearly
fracture of the reinforcement, 2) local buckling of the compressive up to a load of approximately 130 kN with a deflection of 23 mm, at
flange, 3) failure of the shear connection, or 4) ductile failure. The which point the force deflection curve began to plateau to a peak of
ductile failure mode is defined as the failure mode of a composite around 157 kN, as can be seen in Fig. 7. Thus the failure mode of the
beam when none of the first three conditions is met and the beam composite beam was the local buckling of the compressive flange,
deformations are excessive, while load applicators reach their although this was met at the final stages of loading and after the beam
maximum strokes. has reached its plastic moment capacity.
The vertical load–deflection curves for the specimens CB1 to CB5 Composite beam CB2 was subjected to negative bending while a
are drawn in Fig. 7. As it can be seen in the figure, the load–deflection mild axial tension was applied. The measured interface slip has not
curve for specimen CB1 is almost perfectly plastic in the post-elastic exceeded 1 mm in this case. As in the case of beam CB1, there was a
range, while the curves for the rest of the specimens show a hardening small amount of local buckling in the compression flange, however
behaviour which increases as the axial load level increases. This this was reduced compared to CB1 as the applied tensile force resisted
behaviour is due to the axial loading plates which are connected on the buckling forces caused by compression within the top flange. The
first cracking of the concrete slab occurred at a point load approaching
30 kN, with an axial load of 200 kN, and continued to enlarge until the

Fig. 7. Vertical load–deflection curves for all tested beams. Fig. 9. Cracking in the slab of specimen CB1.
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 25

Fig. 10. Steel flange fouling on the load applicator.

ultimate load. The beam specimen behaved linearly up to a load of Fig. 12. Axial load–axial displacement diagram for specimen CB6.

approximately 150 kN with a deflection of 26 mm at which point the

force deflection curve began to harden, as shown in Fig. 7. was removed from the test rig it was evident that the holes for point
Composite beam CB3 was tested under a combination of high load application had elongated due to bearing.
negative bending and moderate axial tension. There was no failure of Finally, specimen CB6 was loaded in pure axial tension; however
the shear connection and the maximum measured interface slip was the eccentricity of the tensile load applicators would cause a positive
2 mm at the ends of the beam. No buckling indication was observed in moment which would result in an upward deflection. To oppose this
the compressive flange, implying that the tensile force counteracted moment and the associated upward deflection, the vertical load
the buckling forces caused by compression within the compression actuator was used. The vertical load actuator was positioned to just
flange. The first cracking of the slab was noticed at a point load of touch the composite beam specimen and the horizontal load was then
50 kN and a horizontal load of 75 kN. When the vertical point load was increased and both the horizontal and vertical load cells were
being raised past 160 kN it began to increase the axial tension load monitored. The axial load–axial deformation diagram for specimen
and when raised from 180 to 195 kN the horizontal load increased by CB6 is plotted in Fig. 12, from which it can be stated that the
approximately 25 kN from 300 to 325kN. This was due to the beam composite section exhibited high ductility levels up to an elongation
bending thereby bringing the ends closer together. The test was of 100 mm. Cracking appeared at a force of 450 kN axial tension and
discontinued at the maximum deflection as the steel beam flange an elongation of 13.5 mm, and cracks continued to appear throughout
fouled on the load applicator (Fig. 10). the length. Compared to CB4 and CB5 the cracks which formed around
Specimens CB4 and CB5 were tested under the effect of a moderate the midspan were much less severe as there was no negative moment
and a high tensile axial load, respectively. There was no evidence of and midspan deflection. At axial loads approaching 1400 kN cracks at
local buckling or shear connection failure after the completion of the either end were forming around the shear stud location as seen in
tests. In fact, the failure mode of the beams was the excessive Fig. 13. Interface slip at the east end was highest, measuring 20 mm,
reinforcement strain which caused the fracture of the rebars in the followed by the west end which had a slip of 12 mm, although the
slab. Very severe cracking was observed in the slab of both specimens, splitting of the slab at the ends may have affected these two results. At
as shown in Fig. 11 for specimen CB5. The reinforcement fracturing maximum load these cracks opened up considerably, as can be seen in
occurred for the specimen CB4 at a combination of 700 kN axial tensile Fig. 13, indicating shear connection failure accompanied by drops in
load and 226 kN vertical load and a midspan deflection of about load.
113 mm, while for specimen CB5 the failure combination was
1200 kN axial force and 330 kN vertical load at a midspan deflection 3.2. Experimental strength and rigid plastic analysis
of 165 mm. Upon inspection a crack which had formed through the
midspan had increased and opened up to around 20 mm on the south The maximum values of vertical and axial loads achieved at the
side. The slip observed was minimal not exceeding 2 mm except for failure point of each specimen were recorded and the ultimate
the slip at the midspan which measured 10.5 mm; this was due to the moment capacity of the beams under the given axial tension level was
severity of the cracks in this area. After the composite beam specimen then calculated using Eq. (2). Table 3 presents a summary of the

Fig. 11. Cracking in the slab of specimen CB5. Fig. 13. Shear stud failure at the end of specimen CB6.
26 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

Table 3 with the experimental values, no partial safety factors have been used
Experimental failure modes. and the average yield strength resulted from the material tests was
Specimen Loading Failure mode assumed for the steel parts. The resulting moment–axial load
combinations from the RPA together with the experimental ultimate
CB1 Pure negative bending Buckling of compressive
flange strengths are summarised in Table 4.
CB2 Negative bending and mild axial tension Buckling of compressive The values obtained by the tests and those calculated by the RPA
flange are plotted in the same graph in the interaction diagram of Fig. 15.
CB3 Negative bending and moderate axial tension Ductile failure
This is the lower left quadrant of the complete moment–axial force
CB4 Negative bending and high axial tension Reinforcement fracture
CB5 Negative bending and very high axial Reinforcement fracture interaction diagram of a composite section. It is observed from the
tension same graph that the experimental data points follow the same basic
CB6 Axial tension only Shear connection failure shape as the theoretical interaction diagram, but are greater in most
cases including negative bending, combined negative bending and
tension, and less than for the case of pure axial tension. The increase
various failure modes observed in each test. It is concluded that, while can partially be explained by the strain hardening of the steel
at low axial tensile load levels the failure mode is primarily the local components and ultimate strengths found to be greater than yield
buckling of the compression flange, at moderate and high axial loads strengths in the tests. The decrease can be explained by premature
acting in the section the failure mode becomes the ductile failure failure of the shear connection during the test of beam CB6 which is
mode by excessive elongation of rebars. not considered in the sectional analysis.
An analytical calculation of the composite beam capacities was From the interaction diagram and the values of Table 4 it can be
also conducted by means of rigid plastic analysis (R.P.A.) within the concluded that the flexural strength of a composite section is not
section. In this analysis only the steel parts (reinforcement bars and affected or is even slightly increased under the simultaneous action of
steel beam section) of the composite section are considered to a relatively low axial tensile force. More specifically, the experimental
contribute to the section capacity, while the concrete in tension is bending capacity of beam CB2 is increased by 6% over the capacity of
neglected. Fig. 14a shows the stress distribution through the beam CB1 (from 157 kN m to 166 kN m in Table 4), when an axial
composite section as it is assumed by the R.P.A. for the case of force equal to 460 kN (about 30% of the plastic axial capacity, Nu) was
negative bending without axial force. For the case of axial force the applied. A similar behaviour is observed in the RPA results: the
plastic neutral axis is assumed to lie in several points within the negative bending strength was not reduced under the simultaneous
section height and the resulting moment and axial tension are action of axial forces equal to 116 and 242 kN (8 and 16% of Nu,
summed taking as centre of rotation the plastic centroid of the respectively, Table 4). As the axial load level applied in the section is
composite section, as shown in Fig. 14b. For comparison purposes increased, the flexural capacity reduces significantly, as can be verified
from inspection of the interaction diagram.
Furthermore, experimental study has confirmed that sectional
a) rigid plastic analysis is a reasonable assumption for the ultimate
design strength of composite beams under the effects of combined
Tr
negative bending and axial tension. A rigid plastic analysis could be
Dpna used for design of composite beams subject to these forces since it is
Ts fairly accurate and will produce a fairly conservative figure for
Mpl
P.N.A. P.N.A.

composite beam strength under the combination of negative bending

and axial forces. When under pure tension rigid plastic analysis will
Cs produce a value which is slightly greater than that found by practical
testing.

b)
Tr
4. Finite element analysis
Dpc
4.1. Finite element model description
Plastic N
centroid
The experimental programme described in the previous sections
P.N.A.

M provided data on the ultimate strength of composite beams under the

Ts
P.N.A. Cs combined effects of negative bending and various levels of axial
tensile forces. Nevertheless, the test results regard only one specific
Fig. 14. Stress distribution according to R.P.A. in a) pure bending and b) assumed stress
composite section. In order to generalise the results and to study a
distribution resulting to axial tension and negative moment in the section (P.N.A. = broader range of sections, the finite element method has been
plastic neutral axis). employed. For this purpose, a three-dimensional finite element model

Table 4
Experimental ultimate strengths of composite beams.

Specimen Experimental results Rigid plastic analysis

Ultimate vertical load (kN) Ultimate moment (kNm) Axial Force (kN) Moment (kNm) Axial Force (kN)

CB1 157 − 157 0 − 145 0

CB2 263 − 166 − 460 (30%)a − 147 − 116 (8%)a
CB3 214 − 130 − 556 (37%)a − 145 − 242 (16%)a
CB4 226 − 126 − 703 (47%)a − 114 − 652 (42%)a
CB5 371 − 84 − 1418 (94%)a − 58 − 1097 (71%)a
CB6 46 −2 − 1497 0 − 1542
a
Percentage of applied axial force to the plastic axial capacity of the composite section.
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 27

exists between the rebars and the surrounding concrete. In addition, a

contact interaction was applied in the beam–slab interface which did
not allow separation of the surfaces after contact in order to prevent
uplift. The node-to-surface contact with small sliding was used while
the hard contact without friction was specified as the contact
property.

4.1.3. Materials
The stress–strain relations obtained from material tests were
converted to piecewise linear curves and used for the modelling of the
steel material for the beam and the reinforcing bars, as shown in Fig.
18b and c. A plastic material with isotropic hardening law was used as
the constitutive law for all the steel parts of the model.
Fig. 15. Interaction diagram and comparison between experimental and analytical
values. The concrete material stress–strain relationship was calibrated
according to the values obtained from the concrete cylinder and
splitting tests. The stress–strain curve for compression follow the
was constructed in order to reproduce the tests on the composite formula proposed by Carreira and Chu [17], while the tensile
beams. The model is relying on the use of the commercial software behaviour is assumed to be linear up to the uniaxial tensile stress
ABAQUS [16]. A detailed description of the model geometry, element provided by the material test. The stress–strain law used is plotted in
types, materials and solution method is given in the following Fig. 18a. The post-failure behaviour for direct straining across cracks is
sections. modelled using the tension-stiffening option and determining a linear
relation until stress is zero at a strain value of 0.05. This value is used
4.1.1. Geometry and element types for avoiding numerical problems in the computational procedure
The concrete slab and the steel beam were modelled using eight- while accuracy is not affected considerably. There are two plastic
node linear hexahedral solid elements with reduced integration, models available in Abaqus/Standard for modelling the concrete
namely C3D8R in Abaqus. The reinforcing rebars were modelled as behaviour. In the present analysis the damaged plasticity model was
two-node three-dimensional linear truss elements, T3D2. An over- preferred over the smeared cracked model. This model provides a
view of the finite element discretization is shown in Fig. 16. The fly general capability for the analysis of concrete structures under static
bracing was modelled indirectly by applying boundary conditions or dynamic and monotonic or cyclic loading based on a damaged
which prevent lateral displacement at the same points on the beam plasticity algorithm. Compared to the companion model (smeared
compression flange as the fly bracings were located in the tests. Due to crack model), it models concrete behaviour more realistically but it is
the symmetrical geometry and loading, only half of the beam was computationally more expensive. Although, this model was chosen
modelled, while appropriate boundary conditions were applied on the for monotonic loading due to its numerical efficiency when full
plane of symmetry. A schematic representation of the various inelastic response has to be captured.
modelling assumptions is depicted in Fig. 17. The shear connection modelling in a composite beam is probably
the most complicated task due to the complex interactions between
4.1.2. Interactions the studs and the concrete slab. The most accurate way is to model the
For modelling the reinforcement in the slab the embedded studs using solid elements and applying contact interactions with the
element technique was employed. The embedded element technique elements of the slab. Although this technique is quite cumbersome
in Abaqus is used to specify an element or a group of elements that lie both in terms of geometry construction and computational time,
embedded in a group of host elements whose response will be used to while convergence problems during the solution are likely to arise
constrain the translational degrees of freedom of the embedded due to the severe immediate cracking of concrete in the region
nodes. In the present case, the truss elements representing the surrounding the studs. A spring model representation of shear studs is
reinforcement are the embedded region while the concrete slab is the chosen to simulate the interface slip in the present study. The
host region. Using this technique, it is assumed that perfect bond nonlinear spring element SPRING2 was adopted to connect a beam
flange node with a slab node in the interface at the same positions
where studs were welded on the specimen, as schematically shown in
Fig. 17. The force slip law for the spring element is derived by the
standard push-out tests on 19 mm-diameter shear studs. A piecewise
linear curve was fitted to the experimental diagram and defined as the
force-slip law for the springs, as shown in Fig. 18d.

4.1.4. Loading and solution method

The vertical load was applied by a concentrated force on the top of
the beam web, while the axial load was applied as an edge pressure on
the steel beam section. The analysis consisted of two steps. In the first
step the contact was established, ensuring that numerical problems
due to contact formulation will not be encountered during the next
steps, while in the second step the vertical and the axial loads were
applied simultaneously as in the experimental specimens. The
modified Riks solution algorithm was employed to solve for the
nonlinear response of the composite beams and to capture any
possible local instabilities. Finite element analysis with concrete
elements mainly in tension may give convergence problems. To avoid
Fig. 16. An overall view of the finite element model used for the experimental these, the discontinuous analysis option was used in the general
simulation. solution control options.
28 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

STEEL BEAM
C3D8R elements
Lateral restraints Symmetry axis
Steel Beam

Contact
Interface
RC slab
Shear connection V

CONCRETE SLAB
C3D8R elements
"zero"
Nonlinear spring

Fig. 17. Finite element model details.

4.1.5. Failure criteria ultimate load achieved was easily identified by the load–deflection
In order to monitor the behaviour of the finite element model and curve.
to identify the modes of failure at ultimate loading, specific failure
criteria were defined corresponding to the ultimate strength of the 4.2. Model validation
various section components. In fact, failure of the composite beam in
the simulation was identified by one of the following situations: a) In order to validate the accuracy of the finite element model the
flange or web buckling, b) reinforcement fracture, and c) shear experimental loading of the beams was reproduced in Abaqus. In the
connection failure. While local instabilities can be captured accurately pure moment case the FE model stiffness and ultimate moment
by Abaqus using the modified RIKS method of solution, the second and capacity are in very good agreement with the experimental curve, as it
third conditions should be identified by monitoring some specific can be seen from Fig. 19. The FE model reached an ultimate moment of
response quantities. To this end, the strain of the reinforcing bars and 160 kN m, which is only 1.9% greater than the experimental value of
the relative displacements in the nonlinear springs representing the 157 kN m. The numerical response is ductile up to failure with an
studs were monitored during the analysis. An excessive strain in the indication of local buckling in the compression flange, as can be seen
rebars reaching the experimental rapture values should indicate a from Fig. 20. The yielding sequence of the basic components of the
reinforcement fracture, while e measured slip near or above the composite section and the load level where yielding occurs in the test
failure slip of the push-out tests was a criterion for shear connection and in the numerical analysis was also compared and is shown in
failure. Finally, a ductile failure mode was defined when none of the Fig. 19. The yielding sequence in the experiment is labelled as A, B, and
above criteria was met and the analysis was ended by numerical C on the diagram and corresponds to yielding of compressive flange,
problems due to excessive distortion of concrete in tension, while the reinforcement and tension flange, respectively, while the respective

Fig. 18. Material stress–strain laws adopted in the finite element model a) concrete, b) structural steel parts, c) reinforcement, and d) shear stud force-slip law.
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 29

Fig. 21. Slip distribution along the beam length in the FE model.

Fig. 19. Moment deflection curves for specimen CB1.

actual behaviour with fairly good accuracy and can be used for further
studies on the behaviour of this kind of structures.
yielding sequence in the FEM is labelled as 1, 2, and 3 on the diagram.
It is observed that, as in the experiment, yielding of the compressive 5. Parametric studies
flange occurs first at a similar load level, followed by yielding of
reinforcing rebars and tension flange, which occur almost simulta- The finite element model presented in the previous section was
neously. In the same diagram, the point at which initiation of local validated against experimental data and was proved to be reliable and
buckling has occurred is labelled as “LB”. Fig. 21 presents a plot of the accurate to capture the ultimate strength of composite sections under
slip distribution along the beam length at ultimate load. It is in good the combined effects of axial tension and negative bending. In order to
agreement with the experimentally measured values, which are about generalise the results and to specify a reliable M–N interaction
0.5 mm along the beam length. equation for design, a parametric study was conducted using a series
The pure axial tension test of specimen CB6 is compared with the of beams with different design parameters. More specifically, the span
FE result in terms of axial load–axial displacement curve in Fig. 22. The length was varied and accordingly the slab effective width and
numerical response is considerably stiffer initially than the experi- reinforcement ratio were designed for each parametric case according
mental one, while the ultimate axial strength values are in good to current codes. Furthermore, the effect of partial shear connection
agreement. The greater numerical stiffness can be explained by noting on the interaction diagram was also evaluated. Details and results of
that the experimental horizontal frame which supported the the parametric study are presented within the next sections.
horizontal load applicators had a limited stiffness and due to the
very high axial loads exhibited a minimum deflection which in turn 5.1. Parametric beam designs
has softened the initial stiffness of the load–deflection curve. The
numerical axial capacity is 1590 kN which is 3% higher than the The design example is taken as an inclined continuous composite
experimental value of 1542 kN. This over-estimation of the FE model ramp consisting of beams in 2 m distance to each other and with
can be partly attributed to the introduction of a rather increased value simple supports at three equally-spaced internal points. A 120 mm
of tension stiffening in the concrete model, which is necessary for slab is assumed to act compositely with the steel sections, thus the
avoiding convergence problems. beams to be designed are four-span continuous composite beams. It is
Fig. 23 illustrates the resulting data in the M–N interaction assumed that the dead load of the ramp is DL = 4.0 kN/m 2 and the live
diagram resulting from the tests, the RPA and the FE model. The FE load is LL = 5.0 kN/m 2. The design load combination is taken
points follow the same trend as the RPA analysis with a small over- according to Eurocode 1 as qdes = 1.35*DL + 1.5*LL, giving a design
prediction, which is expected mainly due to the strain hardening load of 25.8 kN/m on each internal beam. The moment in the middle
effects which are not present in the RPA. For intermediate axial load support is equal to Mdes = −0.071qdesL 2 while the shear force is equal
levels, as for specimens CB3 and CB4, the resulting points are in very to Vdes = 0.464qdesL, where L is the span length. The composite beam
good agreement, whereas there is some discrepancy observed for the section under consideration is the section under negative moment at
CB5 and CB2 case. Nevertheless, considering the complexity of the the middle support and is shown in Fig. 24. By varying the span length,
experimental setup it can be stated that the FE model can capture the the effective width of the slab is also affected, ranging from 1 m for the
8 m span length and assuming a maximum value of 2 m. The
reinforcement in the slab consists of 12 mm-diameter bars at
150 mm distance to each other. A summary of the parametric beam

Fig. 20. Deformed shape of FE model at ultimate load (equivalent plastic strain, PEEQ,
contour plot). Fig. 22. Axial load versus axial displacement for specimen CB6.
30 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

Fig. 23. Comparison of M–N interaction diagrams resulting from experimental and Fig. 25. Comparison of FE results with RPA analysis for IPE240 and IPE300 beams.
numerical data.

Composite described in Section 4.1.5 was met or until the ultimate load level
beam design
reached was obvious by the curve shape.
L
Fig. 25 presents a comparison of the negative moment–tensile
axial force interaction diagrams between the results of the finite
[M] element analyses and the RPA for beams IPE240 and IPE300 and
Le Fig. 26 compares the same results for the beams IPE360, IPE450 and
IPE500. The main observation is that the interaction diagram follows
L the same trend in both analyses, although the bending strength
resulting from the model is greater than the bending resistance
predicted by the RPA, while the corresponding axial strengths have
[N]
similar values. This outcome was expected for the bending case due to
the hardening behaviour introduced in the material constitutive
model of the steel material and the increased tension stiffening effect
introduced in the concrete material model. In fact, the stresses in the
steel flanges at the end of the analyses have values larger than the
Fig. 24. The design example. initial yield strength, while the initial yield value is assumed along the
composite section height in a RPA. Furthermore, in the axial tension
case the above behaviour does not occur and the results of the model
designs is given in Table 5. For comparison reasons, the steel sections are in most cases slightly smaller than those from the RPA. This is
are given in both European shapes (IPE) and Universal Beam shapes mainly attributed to the fact that in axial loading the numerical model
(UB), although the IPE sections are used in the finite element analyses. stops earlier than in a bending analysis due to excessive distortions
Also presented in Table 5 are the moment-to-shear ratio, m/q, and the and numerical problems involved in the axial force transfer and
shear stud number and spacing needed to achieve a full shear concrete in tension. Nevertheless, it can be concluded that the RPA
connection degree employing 19 mm-diameter studs. It has to be gives fairly accurate results for composite beams and can be safely
noted that the moment-to-shear ratio is large for all beams, ranging used in design practice.
from 1.22 to 3.67, which means that the shear–moment interaction Fig. 27 plots the interaction data points resulting from the FE
effects are neglected in the present study [9]. analyses of the five parametric beams with full shear connection in
the same graph. The parametric analyses verify the shape of the
interaction diagram for the fourth quadrant observed in the
5.2. Parametric beam results experimental study. The main outcome is that the bending strength
of a composite beam is not affected or is slightly increased in the
All the parametric beams were subjected to combined negative presence of an axial tensile force up to 20% of the tensile strength of
bending and axial tension. First, a pure moment and a pure axial the composite section. The above result is more obvious if the
analyses were performed in order to specify the ultimate resistances interaction curves are plotted in non-dimensional form by dividing
of the sections. After that, the model was subjected to a constant the x and y values by the ultimate bending and axial strengths,
vertical load at the midspan and various levels of axial tension acting respectively, as shown in Fig. 28. For tensile axial loads greater than
simultaneously, ranging from 0.1 to 0.8 of the ultimate axial 20% of the plastic axial resistance the negative moment capacity is
resistance. The analysis was continued until one of the failure criteria reduced almost linearly by increasing the axial tension in the beam.

Table 5
Details of the parametric beam designs.

L (m) qdes (kN/m) Mdes (kNm) Vdes (kN) m/q beff (mm) As (mm2) Beam section Shear studs (no/spacing in mm)

8 25.8 − 117 96 1.22 1000 791 IPE240 (UB254 × 102 × 28) 14/307
12 25.8 − 264 144 1.84 1500 1130 IPE300 (UB305 × 165 × 40) 20/315
16 25.8 − 469 192 2.45 2000 1469 IPE360 (UB356 × 171 × 57) 24/347
20 25.8 − 733 239 3.06 2000 1469 IPE450 (UB457 × 152 × 82) 24/434
24 25.8 − 1055 287 3.67 2000 1808 IPE500 (UB457 × 191 × 98) 24/521
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 31

5.3. Effect of partial shear connection

As mentioned before, spring elements were used in place of studs

to control the relative movement in the beam–slab interface. In order
to account for a partial shear connection (PSC), the spring force-slip
law was modified by reducing the stud ultimate capacity. By defining
a shear stud strength of 55 kN and maintaining the same number of
studs, shear connection degrees between 0.5 and 0.6 were achieved
and the analyses were repeated on the parametric beams.
Fig. 29 compares the ultimate load–deflection curves of the
parametric beams with full and partial shear connection for the case
of pure negative bending. It can be seen that partial interaction results
in a reduction of the yielding point, although the ultimate load
Fig. 26. Comparison of FE results with RPA analysis for IPE360, IPE450 and IPE500 compromise is smaller. This is attributed to a hardening of the
beams. behaviour after the onset of yielding in the case of partial shear
connection. Furthermore, the present analyses have confirmed the
results from numerous relative studies [e.g. 18,19] that the ductility of
the beam is considerably increased when partial shear connection is
used. In addition, it is concluded that partial shear connection
prevents local buckling of the compressive flange. This is more
obvious from the load deflection curves of beams IPE300, IPE360 and
IPE450. For these beams, a drop in the load at the end of the curve
indicates a local buckling initiation, while this behaviour is absent in
the PSC case, where the behaviour is continuously hardening.
The axial capacity of a composite beam is not seriously affected by
the use of partial shear connection, as it is seen in Fig. 30. The steel
beam is still capable of transferring the axial tension to the
reinforcement in the slab, although this may happen at later loading
stages. The shear connection is always the failure mode and the
reason that the numerical analysis ends, as the studs which are
located near the beam ends exhibit excessive slip values.
The shape of the negative moment–axial tension interaction
Fig. 27. Interaction curves for the parametric beams.
diagram does not seem to be affected by the use of lower degrees of
shear connection. This can be verified by studying Fig. 31, where the
interaction curves for the case of partial and full shear connection are
plotted together. The axial tension resistance is slightly lower in the
PSC case, as the studs are not able to transmit efficiently all the axial
load to the reinforcing bars at the ultimate loading levels.

5.4. Proposed design equations for composite beams under hogging

moment and axial tension

Current structural codes do not provide explicit methodologies for

the design of composite beams, which are asymmetric in nature,
under the interaction of bending and axial load. In the present section,
a design formula is proposed for the design of composite beams under
the combined effects of negative bending moment and axial tension
acting in the steel section. The design formula is based on the results
Fig. 28. Non-dimensional interaction curves for the parametric beams. of the experimental study presented in the previous sections and the

Fig. 29. Load–deflection curves for parametric beams with full and partial interaction.
32 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33

Fig. 30. Axial load–axial displacement curves for parametric beams with full and partial shear connection.

finite element analysis, which is considered by the authors as a formula is valid for composite beams with full shear connection and
reasonable and accurate prediction of the composite beam section up partial shear connection not less than 0.5, provided that premature
to ultimate loads. shear connection failure is avoided by using ductile shear connectors.
The non-dimensional moment–axial force interaction data points
for all parametric beams are plotted in Fig. 32, both for full and partial 6. Conclusions
shear connection. A design curve is superimposed on the analysis data
points. Taking into account the various safety factors that are The ultimate capacity of composite beams under the combined
introduced in the practical design by all structural codes, the following effects of negative (or hogging) bending moment and axial tension
design equation is proposed for the design of a composite beam under was investigated by means of an experimental programme, a rigid-
the combined effects of negative (hogging) bending and axial tension: plastic cross-sectional analysis and nonlinear finite element models. It
was assumed that adequate lateral restraints are provided to the
N 8 M compression flange so that the beam is not susceptible to lateral
+ = 1; for N N 0:20 × NU ðaÞ
NU 10 MU ð3Þ torsional buckling.
M = MU ; for N ≤ 0:20 × NU ðbÞ The experimental study has demonstrated that the negative
moment capacity of composite beams is not affected or is even slightly
where N and M are the design axial force and bending moment and Nu increased when a relatively low axial tension force is simultaneously
and Mu are the plastic axial and bending capacities of the composite acting in the composite section. For higher values of axial tension, the
beam, which can be calculated by means of the rigid plastic analysis moment capacity is reduced significantly by increasing the tension
method, as described in various structural codes and is proven to be a level.
reasonably conservative method. According to this formula, the Sectional rigid plastic analysis in accordance to current design
bending capacity at the hogging moment regions of composite codes has been proven to be reasonably accurate and to provide with
beams is not reduced when an axial tensile force up to 20% of the plastic capacities which can be used in the design practice safely. The
axial strength is applied in the beam, but it reduces linearly to zero general shape of the moment–axial force interaction curves resulting
when larger values of axial force are present. Table 6 presents a from the experiments and from the rigid plastic analysis are in good
comparison between the bending strength predicted by the FEM agreement.
results and the design moment capacity calculated by Eq. (3). The The finite element model developed has been validated against the
associated error is also reported. It can be seen that the proposed experimental results and the very good agreement between the tests
equation is fairly conservative for the design of composite beams in and the numerical outcome has proved the efficiency of the finite
hogging moment regions under the effects of axial tension. The error element model to capture the nonlinear behaviour of the composite
is small for low axial loads and becomes larger as the axial load
increases. Finally, it has to be pointed out that the proposed design

Fig. 31. Moment–axial load interaction curves for the cases of full and partial shear
connection. Fig. 32. Non-dimensional interaction diagram and proposed design equation.
 G. Vasdravellis et al. / Journal of Constructional Steel Research 68 (2012) 20–33 33

Table 6 the design resistances are computed according to rigid plastic theory,
Formula error on the prediction of moment capacity. in accordance to current structural codes.
Beam N/Nu M (FEM) M (equation) Error (%)

IPE240 0.00 − 208 − 208 0.00 Acknowledgements

0.13 − 212 − 208 − 1.89
0.19 − 213 − 208 − 2.35 The experimental research presented in this paper was funded by
0.38 − 192 − 160 − 16.67
the Australian Research Council (ARC) Discovery Project DP0879734.
0.54 − 157 − 120 − 23.39
0.64 − 127 − 92 − 27.32 The authors would also like to thank Dr Mithra Fernando and all the
1.00 0 0 0.00 technical staff of the Structures Laboratory at University of Western
IPE300 0.00 − 378 − 378 0.00 Sydney for their valuable assistance with the experimental work.
0.14 − 379 − 378 − 0.26
0.31 − 350 − 325 − 7.24
0.47 − 296 − 250 − 15.70 References
0.60 − 240 − 189 − 21.25
1.00 0 0 0.00 [1] Banfi M. Composite beams with web openings subject to axial load. EUROSTEEL
2008; September 3–5 2008. Graz, Austria.
IPE360 0.00 − 607 − 607 0.00
[2] Eurocode 4. Design of composite steel and concrete structures, part 1.1: general
0.14 − 610 − 607 − 0.49
rules and rules for building. BS EN 1994-1-1: 2004. London (UK): British Standards
0.30 − 574 − 532 − 7.24
Institution; 2004.
0.58 − 396 − 317 − 19.98 [3] Standards Australia. Composite structures part 1: simply supported beams. AS
0.76 − 205 − 183 − 10.52 2327.1-2004; 2004. Sydney (Australia).
1.00 0 0 0.00 [4] Standards Australia. Bridge design, part 6: steel and composite construction. AS
IPE450 0.00 − 915 − 915 0.00 5100.6-2004; 2004. Sydney (Australia).
0.14 − 912 − 915 0.33 [5] ANSI/AISC 360-05. Specification for structural steel buildings. Chicago (IL, USA):
0.21 − 896 − 904 0.92 American Institute of Steel Construction; 2005.
0.46 − 680 − 613 − 9.90 [6] Uy B, Bradford MA. Cross-sectional deformation of prestressed composite tee-
0.65 − 490 − 396 − 19.28 beams. Struct Eng Rev 1993;5(1):63–70.
0.78 − 302 − 257 − 14.85 [7] Uy B. Long term behaviour of composite steel–concrete beams prestressed with high
1.00 0 0 0.00 strength steel tendons. Advances in Steel Structures (ICASS '96); 1996. p. 419–24.
IPE500 0.00 − 1200 − 1200 0.00 [8] Chen S. Experimental study of prestressed steel–concrete composite beams with
external tendons for negative moments. J Constr Steel Res 2005;61(12):1613–30.
0.14 − 1205 − 1200 − 0.41
[9] Liang QQ, Uy B, Bradford MA, Ronagh HR. Ultimate strength of continuous
0.20 − 1199 − 1197 − 0.20
composite beams in combined bending and shear. Journal of Constructional Steel
0.44 − 988 − 834 − 15.58 Research 2004;60(8):1109–28.
0.63 − 690 − 553 − 19.90 [10] Loh HY, Uy B, Bradford MA. The effects of partial shear connection in the hogging
0.75 − 447 − 373 − 16.51 moment regions of composite beams part I — experimental study. Journal of
1.00 0 0 0.00 Constructional Steel Research 2004;60(6):897–919.
[11] Loh HY, Uy B, Bradford MA. The effects of partial shear connection in the hogging
moment regions of composite beams part II — analytical study. Journal of
Constructional Steel Research 2004;60(6):921–62.
section and any possible local instabilities. The finite element analyses [12] Elghazouli AY, Treadway J. Inelastic behaviour of composite members under
combined bending and axial loading. Journal of Constructional Steel Research
have confirmed the interaction curve shape for composite beams with 2008;64(9):1008–19.
a broader range of steel sections, span lengths and slab widths. [13] Uy B, Tuem HS. Behaviour and design of composite steel–concrete beams under
Furthermore, the partial shear connection effect has been studied. It is combined actions 2006. Eighth international conference on steel–concrete
composite and hybrid structures (ICSCCS 2006), Harbin, China: 286–91; 2006.
concluded that the use of partial shear connection not lower than 0.5 [14] Johnson RP. Composite structures of steel and concrete. Volume 1: beams, slabs,
does not affect the plastic capacity of a composite beam. In addition, columns, and frames for buildings2nd ed. . Oxford: Blackwell Scientific
partial shear connection has a positive effect on the failure mode of a Publications; 1994.
[15] Standards Australia. Methods of testing concrete. AS1012.9-1999. Standards
composite beam by alleviating the eventual local instabilities in the Australia International Ltd.
compression flange and introducing higher ductility levels. As a [16] ABAQUS user's manual, version 6.9. Pawtucket, RI: Hibbit, Karlsson & Sorenson;
consequence, the general shape of the interaction curve is not 2005.
[17] Carreira DJ, Chu KH. Stress–strain relationship for plain concrete in compression.
significantly changed with the use of partial shear connection levels.
ACI J Proc 1985;82(11):797–804.
Finally, a design formula has been proposed for composite beams [18] Uy B, Nethercot DA. Effects of partial shear connection on the required and
subjected to a combination of negative bending moment and axial available rotations of semi-continuous composite beam systems. Struct Eng
tensile loads. The design formula is the same for composite beams 2005;83(4):29–39.
[19] Nguyen QH, Hjiaj M, Uy B, Guezouli S. Analysis of composite beams in the hogging
with full and partial shear connection and is based on the results of moment regions using a mixed finite element formulation. Journal of Construc-
both the experimental study and the finite element analyses, while tional Steel Research 2009;65(9):737–48.