org
Steel Structures 9 (2009) 29-37
1
Department of Civil and Structural Engineering, The University of Sheffield, Sheffield S1 3JD, UK
2
School of Architectural Studies, The University of Sheffield, Sheffield S10 2TN, UK
Abstract
The behaviour of composite cellular floor beams is becoming important as such members are increasingly used in multi-
storey buildings. In the event of fire, this issue becomes increasingly critical, particularly for exposed steelwork. In a fire
situation, a composite beam has a much higher perimeter area exposed to fire in its lower web-flange section than in the upper
web-flange section, and so the temperature distribution across a composite beam is usually non-uniform. The reduction in fire
of the strength and stiffness of the material properties of the perforated steel beam, as well as differential thermal expansion,
therefore becomes an important influence on the overall behaviour of the composite beam. The objective of this research is
to enhance the level of understanding of the generic behaviour of composite cellular floor beams in fire conditions. In this paper,
three-dimensional nonlinear finite element models of composite cellular floor beams have been developed, taking into
consideration the influence of the changes in material properties with temperature. Experimental data from furnace tests on
cellular composite floor beams obtained from previous research work has been used to validate the FE models. An analytical
model based on existing design guides is also presented in this paper. It is concluded that finite element analysis results are
in good agreement with the experimental data, and all the failure modes have been accurately predicted. The proposed
simplified analytical methods show reasonable agreement with the test and FE results, and are always conservative.
Keywords: cellular beam, structural fire engineering, modelling
the ultimate, serviceability and construction limit states. elements representing the shear connectors; no information
The design methods are based on plastic analysis of the is provided about the modelling of the concrete slab or
beam section at ULS and elastic analysis at SLS. The the material properties used for the concrete slab and steel
recommendations for the strength of web posts and the beam. Full-scale tests were carried out to investigate the
upper and lower web-flange sections were derived from a web-post buckling and Vierendeel bending of the
parametric study involving detailed linear and non-linear composite section. From the results, model factors for test
FEA. However, the analytical studies were only carried failures by web-post buckling ranged from 1.19 to 1.90,
out to investigate the capacity of non-composite cellular indicating that the proposed design method is conservative.
beams, and no studies were carried out on composite Long openings were shown to cause pull-out forces in the
cellular beams. shear studs caused by the development of local composite
Chung and Lawson (2001) presented simplified design action. It is recommended that the opening aspect ratio
tables for simply supported composite beams with (length:depth) should not exceed 2.5.
rectangular web openings under various common loads, These previous research studies, apply only to cross-
in the accordance with Eurocode 4. The proposed tables sections at ambient temperature; the investigations of
give general information on sizing of openings in terms composite cellular floor beams under fire conditions are
of a utilisation ratio for the bending resistance at mid- still rather limited in scope.
span and shear resistance at supports, as well as the type Nadjai (2007) described full-scale fire tests and
et al.
of loading of the member. The proposed design method analytical studies of the performance of cellular composite
has been validated against full-scale test results, indicating beams at elevated temperatures. Four full-scale fire tests,
that it is conservative as the moment ratio increases, but on composite beams consisting of 150 mm thick concrete
is reasonably accurate for high-shear regions. slabs and cellular steel beams of 4,500 mm length, were
Liu and Chung (2003) reported a non-linear FEA carried out at the University of Ulster. In these tests two
investigation on steel beams with large web openings of types of cellular beams were used; symmetric composite
various shapes and sizes. Eight types of web opening beams using UB406 × 140 × 39 as the upper and lower
shapes have been considered for three opening sizes. In web-flange sections, and asymmetric composite beams
the FE modelling, steel beams are formed by shell using UB406 × 140 × 39 as the upper web-flange section
elements, and a bi-linear stress-strain curve is adopted for and UB457 × 152 × 52 as the lower web-flange section.
steel. Comparisons of the global moment-shear interaction Each beam was tested at ambient and elevated temperatures
curves for these steel beams under a wide range of with various loading conditions. Simulation of the behaviour
applied moments and shear forces, have been presented. of the tested composite cellular beams was carried out
The results show that the beams behave very similarly in using the non-linear FEA software SAFIR. Shell elements
terms of failure modes and yield patterns. It was reported were used to model cellular beams, but no information
that the effective length of the opening significantly about the simulation of the concrete slab or the
controls the magnitude of the local Vierendeel moments connection between the slab and the steel beam. The
acting on the upper and lower web-flange sections. On comparison between the FE and test results shows good
the basis of this work Chung et al. (2003) developed an agreement in terms of failure modes, ultimate loads and
empirical design method for steel beams with large web load-deflection behaviour.
openings of various shapes and sizes, using a generalised Bailey (2004) presented the results from a preliminary
moment-shear interaction curve to determine load indicative fire test on unprotected and protected cellular
capacities. The simplified design method includes the and solid-web beams, to investigate the temperatures
shear areas of the upper and lower flanges, because the experienced by the web-posts of a cellular beam. The
FEA results show that the basic shear capacities of specimens consisted of unprotected cellular and solid-
perforated beam are consistently greater than those web beams, and intumescent-protected versions of the
predicted by the current design rules. The proposed same beams. All the beams were subjected to the same
design method considers the coupled shear capacity ratio, furnace conditions. On comparing the results obtained
the coupled moment capacity ratio and the Vierendeel from the fire tests, it was found that no increase in web-
mechanism of the perforated beam. post temperatures occurred for the unprotected cellular
Lawson et al. (2006) introduced a design method for beam, but that web-post temperatures for the protected
composite asymmetric cellular beams, which is not fully cellular beam were higher than the web temperatures of
covered by existing guidance. A simplified method is the solid beam. This suggested that the increases in web-
presented for web-post buckling, based on an effective post temperatures were due to the intumescent coating
length of the compression field, modelled as a strut. The surrounding the web opening being ‘pulled-back’ as the
proposed design method is calibrated against a result of fire temperature increased. This mode of behaviour needs
series of FEA and full-scale tests, with various openings, to be investigated further. The tests also indicated that the
spacings, and web thicknesses. The non-linear FEA difference between web-post and bottom flange
package ANSYS was used to build the FE model. Shell temperatures of cellular beam is dependent on the type
elements were used to form the cellular beam, with spring and thickness of intumescent coating used.
Behaviour of Composite Cellular Steel - Concrete Beams at Elevated Temperatures 31
Lawson et al. (2006) prepared a cellular beam design 2. Finite Element Analysis
guide for Fabsec Ltd , to assist in the design of cellular
beams in non-composite and composite applications, both 2.1. General
at ambient temperature and in the fire situation. The Based on the results obtained from the available full-
design guide is in accordance with BS 5950-1 (2000), BS scale fire tests (Nadjai 2007), geometrically non-
et al.,
5950-3.1 (1990) and BS EN 1994-1-1 (2004). The linear finite element simulations have been carried out
structural model used to assess the resistance to local and using the FEA package ABAQUS with non-linear material
global actions and development of the thermal model are properties, on composite beams including downstand
described. The design model for web-post buckling, steel beams incorporating circular web openings. The
which is based on a modified strut analogy, has been deflected profiles around openings and load-deflection
introduced. Development of the structural model at high curves have been compared with the corresponding
temperatures is based on modification of the design experimental results. The finite element modelling results
strength and stiffness used at ambient temperature. The are found to be in good agreement with the full-scale tests
material properties used are those appropriate to elevated in terms of failure modes and load-deflection curves.
temperatures according to BS 5950-8 (2003). Loaded fire
resistance tests on protected composite beams and unloaded 2.2. Experimental investigation
short sections have been carried out at Warrington Fire The full-scale fire tests performed by Nadjai et al.
Research Centre. Based on these tests, a mathematical considered both symmetric and asymmetric composite
model of the behaviour protected steel sections has been cellular beams at elevated temperatures. The symmetric
developed. The temperature distribution across any composite beams (Type A) were formed using UB406
section may be obtained by using this model. The design × 140 × 39 as upper and lower web-flange sections; the
model at ambient temperature has been developed by asymmetric composite beams (Type B) were formed
using the results obtained from FEA modelling, which using UB406 × 140 × 39 as the upper web-flange section
was carried out using the non-linear software package and UB457 × 152 × 52 as the lower web-flange section.
ANSYS. These analyses were carried out for non- Both tests used a 150 mm thick × 1,200 mm wide C35
composite beams, and hence conservative results were concrete slab with A142 reinforcement mesh having a
given for composite beams. yield strength of 460 N/mm2. Shear connectors of 19mm
In this paper, three-dimensional nonlinear finite element diameter × 120 mm height were equally distributed at 150
models of composite cellular floor beams have been mm spacing along the longitudinal span of the test
developed, taking into consideration the influence of the specimens; this is a high density in order to ensure full
changes in material properties with temperature. An interaction between the slab and the beam. 1.25 mm thick
analytical model based on existing design guides is also HR 51/150 Holorib steel decking having a yield strength
presented in this paper. The FE and analytical results of 327 N/mm2 was used in the tests. The span of both
obtained are compared with the experimental data from specimens was 4,500 mm, and the steel downstand sections
furnace tests on cellular composite floor beams obtained had a yield strength of 355 N/mm2. Load was applied to
from previous research work. the top of the concrete slab through distribution beams as
beam test (Type B). were two in Type A and one in Type B, were simulated
by 5 mm×1,200 mm patches of distributed load of intensities
15 N/mm2 and 30 N/mm2, applied to beams of Types A
shown in Fig. 1. Two point loads, each of 90 kN, and a and B respectively. This was to simulate the load applied
single point load of 210 kN were applied on both the through the load distribution beams shown in Fig. 1. Full
symmetric and asymmetric beams. Table 1, Fig. 2 and composite action between the concrete slab and the
Fig. 3 illustrate the detailing of the composite beams used cellular steel beam was achieved by using a tying
in the experiments. constraint to tie the surfaces of both components together.
2.3. Finite Element (FE) modelling 2.4. Material modelling
The commercial FEA package ABAQUS/Explicit was Concrete is assumed to be an isotropic material prior to
used to carry out the simulation work. This uses an cracking at its ultimate stress, after which the compressive
explicit dynamic FE formulation, and is suitable for stress is assumed decrease linearly to zero. In this paper,
overcoming problems with convergence difficulties in it is assumed that compressed concrete behaves linearly
complex composite structures. For the concrete slab, a up to a stress of 0.3f . Beyond this point, the compressive
c
three-dimensional eight-noded solid element (C3D8R) stress σ in concrete is represented as a function of strain
c
was used to improve the rate of convergence. Reinforcing according to Eq. (1), as suggested by Carreira and Chu
mesh in the solid slab element was defined as a layer of (1985), in which γ is a parameter used to define the
steel of equivalent area in each direction. The position of stress- strain curve for concrete.
the reinforcement layer was measured from the mid-depth f c γ ( ε ⁄ εc 1 )
of the concrete slab. Four-noded quadrilateral shell elements σc = ------------------------------ (1)
(S4R) with reduced integration were used to represent the γ – 1 + ( ε ⁄ εc 1 )γ
cellular steel beam. This is the most appropriate type of in which, according to EN 1992-1-2 (2004);
Figure 5. Boundary conditions and loaded areas for beams of Types A and B.
Behaviour of Composite Cellular Steel - Concrete Beams at Elevated Temperatures 33
γ =3 and εc =0.0025
1
2007).
of its compressive strength and 0.2 respectively. At zero
stress, the value of strain is assumed to be 8 times the 355 N/mm at 0.002 strain, and its elastic modulus is
2
strain at failure. Figure 6 presents the stress-strain 210,000 N/mm at temperature 20 C. In fire, the strength
2 o
relationship for concrete (BS EN 1994-1-2, 2005) which and stiffness of the material are reduced as its temperature
has been used in the finite element models. increases. The strength and modulus reduction factors
The damaged plasticity concrete model has been used are based on EN 1994-1-2.
adopted in the ABAQUS FE modelling. This concrete
model uses the concept of isotropic damaged elasticity in 2.5. Analysis procedure
combination with isotropic tensile and compressive plasticity The mechanical strength and stiffness properties of
to represent the inelastic behaviour of concrete. The two most materials decrease as their temperature rises, and so
main failure mechanisms are tensile cracking and compressive the structural resistance of a member decreases as it is
crushing of the concrete material. The damaged plasticity heated. The structural analysis simulates structural behaviour
concrete model is intended to provide a general capability at ambient temperature (Step 1), and at elevated temperatures
for the analysis of concrete structures subjected to static (Step 2) as a function of time. The beams were tested
and dynamic loading under low confining pressures. under heating by a slow fire curve, as illustrated in Fig. 8.
The steel material models for structural steel and rebar In the structural analysis at ambient temperature, the
are represented as elastic-plastic with strain hardening, as loads applied to beams of Type A and Type B are 180 kN
illustrated in Fig. 7 (BS EN 1994-1-2, 2005). In ABAQUS, and 210 kN respectively. Both beam types were loaded to
the behaviour of steel is been considered using the true their respective applied loads at 20 C in Step 1. In the
o
stress – true strain curve. The conversions to true stress second step, the temperatures measured in the tests at time
(σtrue) and true strain (εtrue) are expressed in terms of the 80 minutes, when they were terminated, were introduced
nominal stress (σnorm) and nominal strain (εnorm) as into the finite element models. For the concrete slab, an
σtrue
εtrue = ln (1 + εnom) – ---------- (3) effective flat slab thickness and an assumed linear
E distribution of temperature through the slab have been
σtrue = σnom(1 + εnom) (4) adopted in the FEA models. The temperatures at the top
and bottom surfaces of the slab obtained from experiment
The yield strength of the structural steel used here is were directly applied in the FEA model. Figure 9 illustrates
34 Vuiyee Bernice Wong et al.
diameter.
and
MN,Rd,t are the plastic bending resistances of
MN,Rd,b
MN Rd = Mv Rd 1 – ⎛⎝--------
, ,
N ⎠ (6)
the effective slab thickness of the composite slab. Figure Rd
10 and Table 2 show how temperature is distributed. Mv,Rd is the local bending resistance of the web-flange
section. Plastic resistance should be used for Class 1 and
3. Analytical model 2 sections; elastic section properties are used for Class 3
and 4 sections.
3.1. General Mvc is the local bending resistance of the plastic hinge
Based on the available design guides (BS 5950-1, 2000; above the opening at which the concrete slab is further
BS EN 1992-1-2, 2005; Lawson et. , 2006; Steel
al. compressed by the hinge rotation due to local composite
Construction Institute, 2007a and 2007b) an analytically- action between the upper web-flange section and the
based model is now presented for determining the critical concrete slab. This may be approximated by:
temperatures of composite cellular floor beams at elevated Mvc = nsPRd(h + yt) (7)
temperature. The model considers shear force transfer by
Vierendeel action across the openings and the forces in which is the number of shear connectors provided
ns
generated across web-posts. The high-temperature limiting above the opening in a length (0.5 ), where is the
do + h h
cases of the composite beam are assessed by applying the thickness of concrete slab, is the design shear
PRd
material properties appropriate to elevated temperature. resistance of a shear connector based on EN 1994-1-1,
Assuming that the design of the composite beam with and is the distance of the centroid of the upper web-
yt
openings is adequate for its overall shear and bending flange steel tee-section from the top of the flange.
resistance requirements, the following simple method is
proposed for calculating the resistance of the zone influenced 3.2.2. Buckling resistance of web-post
by local Vierendeel action and web-post buckling. For a simply supported cellular beam subject to uniform
loading, beam web-posts are subject to high combined
3.2. Structural model at ambient temperature stresses generated by horizontal forces from global bending
3.2.1. Resistance to Vierendeel bending action, shear forces and local Vierendeel bending. Therefore,
In general, the shear force at an opening is mainly in cellular beam design, the web-posts of the perforated
2
E-
fE = π--------
2
λ
12l
λ = -------------e
tw
in which, λ is the slenderness of the web-post and l is its
e
Web-post buckling check. effective length as an equivalent strut. For a circular web
opening, l is given by:
Figure 12.
e
section should be checked for resistance to local buckling. le = 0.5 S2o + d2o (10)
Each web post is checked as a strut, by considering a
compression force acting over its minimum width S (Fig. o 3.3. Structural model at high temperature
12). The capacity of a web-post should satisfy The design of the composite section for fire resistance
VED generally follows the design procedure at ambient temperature.
Vt = --------
- ≤ Vh buck = χfySotw (8) The effect of fire is accounted for by calculating the
2 ,
Figure 14. Beam Type A; web-post buckling failure in FE and test (Nadjai et al. , 2007).
36 Vuiyee Bernice Wong et al.
Figure 15. Beam Type B; web-post buckling failure in FE and test (Nadjai et al., 2007).
Table 3. Experimental results, FE and analytical models
Mid-span deflection (mm) Bottom Flange temperature Bottom Flange critical
Beam type At ambient temperature At elevated temperature (°C) at 80 minutes temperature (°C)
Test FEA Test FEA Test FEA Analytical model
A 4.21 4.25 200 210.1 667 667 604.2
B 4.71 4.14 193.8 206.7 635 635 555.2
Figure 16. Experimental results and FE model for Beam Figure 17. Experimental results and FE model for Beam
Type A. Type B.
structural behaviour of the composite perforated sections the prediction was only between 8-10% in term of critical
observed from the experiments are in good agreement temperature, indicating reasonably close agreement.
with the finite element results in terms both of failure
modes and overall behaviour. Web-post buckling is clearly 5. Conclusions
observed in beams of both Types A and B, as illustrated
in Figs. 14 and 15, respectively. Comparison of the mid- A finite element model has been developed to investigate
span displacements at ambient and elevated temperatures the behaviour of cellular composite floor beams at elevated
has been made between the results obtained from tests temperatures. A simplified analytical model has also been
and FEA, as presented in Table 3. Figures 16 and 17 presented. Comparisons with the results obtained from
illustrate the comparison between FEA and experiment of some available experimental fire tests have shown good
the mid-span displacements in terms of the beam bottom- agreement; all failure modes are accurately predicted by
flange temperature. The critical temperatures generated the FE models. The simplified analytical model has also
by the proposed analytical model have also been shown. shown good, though conservative, agreement with both
The experimental results and FEA appear to be in good the FEA and the furnace tests in terms of critical bottom-
agreement. The simplified analytical model shows a slightly flange temperature. The proposed FE model can be utilised
conservative prediction, in which the critical bottom- to investigate the high-temperature behaviour of composite
flange temperatures were 604.2oC for the beam of Type A cellular beams in far more detail than is possible by testing
and 555.2oC for Type B. However, the safety margin of alone.
Behaviour of Composite Cellular Steel - Concrete Beams at Elevated Temperatures 37