www.elsevier.com/locate/engstruct
Received 5 January 1999; received in revised form 1 April 1999; accepted 15 April 1999
Abstract
This paper is concerned with experimental study on the ultimate load behaviour of steel-concrete composite beams curved in plan.
Five beams of realistic dimensions built from rolled steel beam and concrete slab were tested to failure. Extensive measurements of
strain in both concrete and steel, and of displacements were made in order to obtain a complete picture of elastic and ultimate load
behaviour. Each of the beams was simply supported at the ends and was subjected to a concentrated load applied at mid-span. All
the beams tested were analyzed by using the finite element method and the results were compared with those obtained experimentally.
The test results indicate that the load-carrying capacity decreases with the increase in the “span/radius of curvature” ratio. The
experimental results for deformations, for stress distributions and for ultimate strengths were found to be in good agreement with
the corresponding values predicted by finite element analysis. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Experimental study; Steel–concrete composite beams; Beams curved in plan; Ultimate load behaviour; Finite element analysis
0141-0296/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 9 9 ) 0 0 0 4 6 - 2
878 V. Thevendran et al. / Engineering Structures 22 (2000) 877–889
Table 1
Dimensions of test specimens
(Uo)N (Uo)M
Specimen B (mm) Ds (mm) b (mm) D (mm) tf (mm) tw (mm) Ls (mm) L (mm) R (mm) L/R
(mm)a (mm)b
a
(Uo)N=L2/(8R)=Nominal initial lateral displacement of the centroid from the original position.
b
(Uo)M=Max lateral displacement at mid-span from the line joining the supports.
V. Thevendran et al. / Engineering Structures 22 (2000) 877–889 879
Tensile test coupons cut from the flanges and webs of 2.4. Test setup and instrumentation
the steel beam segments before and after cold-bending
were tested in order to obtain the material properties. The test setup was basically same for both straight
For each specimen, four tensile coupons —two from the and curved specimens. A test rig built on a strong floor
flange part and two from the web part—were cut out for and capable of applying up to a maximum load of 2000
the material tests. They were tested in an INSTRON 500 kN was used for testing the specimens. Specimens
kN Universal Testing Machine in accordance with the longer than 6 m could be tested on the rig. Roller sup-
ASTM specification (1979). The values obtained from ports were inserted between the ends of the specimen
the test for the modulus of elasticity and yield stress, and the rig to provide simply supported conditions.
before and after cold bending of steel girders are summa- Transverse beams were provided at the ends of the speci-
rized in Table 2. The material properties of concrete men to simulate the existence of transverse diaphragms
were measured by cube and cylinder tests. A summary at supports in practice. A transverse beam was also
of the concrete properties on the day of testing of the attached at the mid-span of the specimen for transmitting
respective beam specimens is presented in Table 3. concentrated load to the specimen. The specimen was
mounted on roller supports set at 6 m apart on the testing
2.3. Fabrication of test specimens rig. The test setup is shown in Fig. 2.
For the straight beam, instrumentation was provided
The I-girders were fabricated by welding the second- by means of transducers and strain gauges for measure-
ary beams to the main girder in a factory. Headed shear ments of vertical deflections along the length of the
studs (19 mm diameter, 75 mm long) were welded onto specimen, steel and concrete strains at mid-span and the
the top flanges of the main girder and of the secondary slip between concrete slab and steel beam. For the
beams by a stud welding machine, and were checked by curved beams, additional transducers were mounted to
regular bend testing. The formwork for the slab was built measure the horizontal deflection. Transducers were
in the laboraory using structural grade lumber and ply- mounted on specially prepared brackets. Concrete strains
form. The plyform was cut according to the curvature at quarter span were also measured for the specimens
of specimens and moved into position. The top of the SP3, SP4 and SP5. Fig. 3 shows the locations of trans-
plyform was adjusted to be in close contact and flush ducers and strain gauges for the specimens. The readings
with the top flange. The joints in the formwork were from these strain gauges were recorded using a data-
caulked and the formwork was oiled. Two layers of logger (TD-301) connected to a personal computer. The
welded mesh (8 mm diameter bars spaced at 150 mm load-displacement plot displayed by the computer
880 V. Thevendran et al. / Engineering Structures 22 (2000) 877–889
Fig. 1. Details of a typical specimen; (a) Structural steel components; (b) Detail A; (c) Detail B.
enabled the yielding of the steel and the onset of failure applied to test all specimens was the same. The specimen
of the specimen to be monitored. was carefully positioned in the test frame such that the
midpoint of the beam was under the actuator. Before
2.5. Test procedure applying any load on the specimen, all the electrical
resistance strain gauges and transducers were connected
The tests were carried out at the age of seven days or to a data acquisition system (TDS-30) which had been
more (after the casting of slab). The test procedure programmed to record the outputs on a floppy diskette
V. Thevendran et al. / Engineering Structures 22 (2000) 877–889 881
coupling of bending stiffnesses while the steel flange and the tensile behaviour of the concrete. The first part joins
web were modelled by four-node isoparametric thin shell the origin (zero stress at zero strain) to the maximum
elements with the coupling of bending and stiffnesses. uniaxial tensile stress, ftu, at strain at which the concrete
The shear connectors between concrete slab and steel cracks. Beyond this strain value the tensile stress is
flange were modelled by rigid beam elements. Rigid assumed to decrease linearly from ftu in order to reflect
connection beam elements were used to model the shear the softening of concrete due to crack. The loss of shear
studs based on the assumption that no slip occurs modulus due to cracks was accounted for by using a
between the concrete slab and the steel girder. During multiplying factor, which defined the modulus for shear-
the experiments, the interfacial slip between the slab and ing of cracks as G=rGc, where Gc is the elastic shear
the top flange of the steel girder was measured at both modulus of the uncracked concrete. The shear retention
ends. The relative displacements at failure were found model assumes that the shear stiffness of open cracks
to be negligibly small in all specimens and the maximum reduces linearly to zero as the crack opening increases.
value recorded was 0.09 mm and hence the slip could The multiplying factor, r, is given by
be ignored. The assumption of perfect bonding between
concrete slab and steel beam is thus justified.
Steel was assumed to behave as an elastic-plastic
冋 册
冦 冧
material with strain hardening in both tension and com-
⑀
pression. The idealised stress-strain curve adopted for 1− for 0ⱕ⑀ⱕ⑀max
steel in the finite element analysis is shown in Fig. 5(a). r⫽ ⑀max
The stress-strain relationship of concrete was input by 0 for ⑀ⱖ⑀max
the concrete option in ABAQUS. The constitutive
relation curve adopted for concrete both in compression
and in tension is shown in Fig. 5(b). Concrete in com- where ⑀ is the direct strain across the crack and ⑀max is
pression is considered to be elasto-plastic and strain the value given on the data card of the option. Cedolin
hardening material. Its uniaxial compression stress-strain and Poli [10] proposed that the slope of the shear stress-
curve is assumed to follow the expression [9] shear displacement curve decreases with crack width.
fc⫽f⬘ 冋 冉 冊册
2⑀c ⑀c
⑀0
⫺
⑀0
2
,
The crack width for which the slope becomes zero is
given as cw=0.75 mm. Referring this magnitude to a typi-
cal distance between cracks in the real structures, lc=150
where f⬘ is the cylinder compressive strength of concrete mm, ⑀max=cw/lc=0.005. The model also assumes that
in MPa. The strain, ⑀0, at which the maximum compress- cracks that subsequently close have a reduced shear
ive stress is reached, has been taken as 0.002, while the modulus where r is assumed as 0.95. The nonlinear
strain, ⑀cu, at which the concrete crushes, has been taken response of the beam in the loading path is solved by
as 0.0038. A bilinear approximation was used to model using the Newton iterative technique. Further details of
V. Thevendran et al. / Engineering Structures 22 (2000) 877–889 883
Fig. 5. (a) Idealised uniaxial stress-strain relationships for steel; (b) Idealised uniaxial stress-strain relationships for concrete.
the finite element analyses for all the specimens may be test specimens. The load-vertical displacement curve and
found elsewhere [11]. load-lateral displacement curve at mid-span of the beams
are plotted against the applied load in Figs. 6 and 7
respectively. Maximum lateral displacement in the case
4. Results and discussion of SP2 was very small, less than 2 mm and its variations
with load is not presented in Fig. 7. In the nonlinear
Experimental failure loads along with the correspond- region, the stiffness of SP5 reduces comparatively at a
ing analytical values are summarized in Table 4 for all quicker rate than that for the other beams. Comparing
Table 4
Comparison of ultimate loads predicted by ABAQUS with experimental values
Fig. 9. (a) Variation of tangential stress across width of concrete Fig. 10. (a) Variation of radial stress across width of concrete slabs
slabs at mid-span of specimens at the load of 200 kN; (b) Variation at mid-span of specimens at the load of 200 kN; (b) Variation of radial
of tangential stress across width of concrete slabs at quarter-span of stress across width of concrete slabs at quarter-span of specimens at
specimens at the load of 200 kN. the load of 200 kN.
Fig. 11. Variation of tangential stress in steel beams at mid-span of specimens at applied load of 200 kN.
sis is reliable in predicting the ultimate strength of com- plifies significantly the actual behaviour. Furthermore,
posite curved beams. the smeared “crack” concept is used for numerical mode-
A comparison between typical load-deflection curves ling of crack initiation and crack propagation. This sme-
(of specimens SP3, SP4 and SP5) obtained from the ared “crack” model does not track individual “macro”
finite element analysis and experiments is shown in Fig. cracks. Instead, constitutive calculations are performed
12. Satisfactory agreement between the analytical and independently at each integration point of the finite
experimental values for deflections is observed. Slight element model. However, when the crack developed in
discrepancies between experimental curves and analyti- the concrete during the experiment, the moment of iner-
cal curves may be due to the approximation in concrete tia of the whole section decreased and the ratio of defor-
modeling by ABAQUS. The concrete is not a homo- mation to loading increased considerably.
geneous material. The concrete model in ABAQUS sim- A comparison of initial crack loads and yield loads
V. Thevendran et al. / Engineering Structures 22 (2000) 877–889 887
obtained experimentally with the corresponding values Experiments carried out on steel-concrete composite
obtained from the finite element analysis is shown in beams that are curved in plan are described. Beams with
Table 5. It can be observed that the yield loads predicted different values of radius of curvature have been tested
by ABAQUS are somewhat larger than the values pre- to failure. Test results show that the ultimate load
dicted experimentally and the maximum deviation is capacity decreases with increase in the L/R ratio. Failure
about 10%. The deviation for initial cracking load is modes are obviously different for each of the specimens.
between ⫺29% and 23%. The deviation is due to the With the increase in the L/R ratio, the main cause of
fact that the cracking model in ABAQUS is not meant failure changes from bending to the combined action of
to trace the development of individual cracks. bending and twisting. Crack patterns observed on the
Table 5
Comparison of initial crack loads and yielding loads predicted by ABAQUS with experimental values
Fig. 14. (a) Comparison of experimental and theoretical tangential stress distributions across width of concrete slab at mid-span of SP4 at the
load of 200 kN; (b) Comparison of experimental and theoretical tangential stress distributions across width of concrete slab at quarter-span of SP4
at the load of 200 kN; (c) Comparison of experimental and theoretical tangential stress distributions in steel beam at mid-span of SP4 at the load
of 200 kN.
[9] Park R, Paulay. Reinforced concrete structure. Canada: John [11] Chen S. Steel-concrete composite beams curved in plan. MEng
Wiley and Sons, 1975. dissertation submitted to the National University of Singapore,
[10] Cedolin L, Poli SD. Finite element studies of shear-critical R/C 1997.
beams. J Engng Mech Div ASCE 1977;107:395–410.