www.elsevier.com/locate/engstruct
Received 24 September 2006; received in revised form 21 May 2007; accepted 30 July 2007
Available online 10 September 2007
Abstract
In a steel–concrete composite beam section, part of the concrete slab acts as the flange of the girder in resisting the longitudinal compression.
The well-known shear-lag effect causes a non-uniform stress distribution across the width of the slab and the concept of effective width is usually
introduced in the practical design to avoid a direct analytical evaluation of this phenomenon. In the existing studies most researchers have adopted
the same definition of effective width which might induce inaccurate bending resistance of composite beam to sagging moments. In this paper, a
new definition of effective width is presented for ultimate analysis of composite beam under sagging moments. Through an experimental study
and finite element modeling, the distribution of longitudinal strain and stress across the concrete slab are examined and are expressed with some
simplified formulae. Based on these simplified formulae and some assumptions commonly used, the effective width of the concrete slab and the
depth of the compressive stress block of composite beams with varying parameters under sagging moments are analytically derived at the ultimate
strength limit. It is found that the effective width at the ultimate strength is larger than that at the serviceability stage and simplified design formulae
are correspondingly suggested for the ultimate strength design.
c 2007 Elsevier Ltd. All rights reserved.
Keywords: Steel–concrete; Composite; Effective width; Ultimate strength state; Experiment; Finite element analysis
stress distribution in the concrete slab, but there is no guarantee where As and A0s is tension and compression area of steel
that the resultant forces will be located at the same location. section, respectively.
Due to the shear-lag phenomenon the longitudinal stress in the In order to predict the actual strain and stress distributions
concrete slab decreases from y = 0 to y = b/2 and the depth across the section, finite element method was used to analyse
z c varies across the width of the concrete slab as shown in the composite beams with varying parameters under sagging
Fig. 1(a). Therefore, the stress block with a width of be from moments. The three variables β, ξ and z c0 were then solved
Eq. (1) may not have a total bending moment equivalent to that from Eqs. (2)–(4) after the distribution of σc in the concrete slab
based on the actual stress distribution in the concrete slab, i.e., was obtained and the bending resistance to sagging moment
the accurate value of Mu may not be obtained by the traditional of composite beams can then be obtained by a traditional
method when using the definition of effective width in Eq. (1). plastic beam approach. To confirm the numerical results, an
In order to ensure that the stress distribution in the concrete experimental study was also conducted and is described next.
slab as shown in Fig. 1(a) and (b) is equivalent to ultimate
strength analysis in terms of both axial force and moment, we 3. Experimental study
will have to derive the effective width and the depth of the stress
block by both force and moment equivalencies as: A steel–concrete composite floor model as shown in Fig. 2
Z h c Z b/2 was tested to investigate the shear-lag phenomenon in the
βbξ z c0 f c = σc (y, z)dydz (2) concrete slab of the composite beam in both elastic and inelastic
0 −b/2 stages and the experimental results were used to verify the
R h c R b/2
accuracy of the finite element model described in next section.
ξ z c0 0 −b/2 (h c − z)σc (y, z)dydz
= R h c R b/2 . (3) The model consists of three identical longitudinal girders and
−b/2 σc (y, z)dydz
2
0 two transverse girders at the ends of the longitudinal girders.
Considering the force equilibrium in the entire beam section A cast-in-place concrete slab with a height of 60 mm was
we have connected to the girders by head studs. The experimental model
Z h c Z b/2 with three girders represents more closely the real composite
structures and can give more realistic results than traditional
As f − A0s f = σc (y, z)dydz, (4)
0 −b/2 single-beam specimens.
J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407 1399
Fig. 3. Experimental stress–strain curve for steel materials. jacks in series with an increment of 2 kN. During the test both
global and local quantities, such as displacements, strains of
the concrete slab and steel beams, and slip at the concrete-steel
interface were monitored. Since the test specimen is designed
as a full composite section and the slip mainly affects the
serviceability behavior of beams and its effect on ultimate
strength is insignificant [16], no detailed slip information is
presented here for the sake of brevity. The mid-span vertical
displacement reached up to 160 mm at the ultimate load Pu =
256 kN, when the collapse happened due to the crushing
on the top surface of the concrete slab. Fig. 5 shows the
deformed shape of the specimen and loading frame used for
the experiment.
Fig. 6 displays the strain distribution along half of the slab
width (with the origin at the center of the deck as shown in
Fig. 2) on the top and bottom surfaces of the concrete slab.
Such curves are displayed under different loading levels for
Fig. 4. Stress–strain curve for steel materials used in FEM.
the mid-span section of the specimen where Pu is the ultimate
load from tests. In general, the compressive strain on the top
Table 1 surface of the portion of the slab remote from the steel beam
Results of compression tests on concrete
lags behind that of the portion near the beam (Fig. 6(a)), while
Cube no. 1 2 3 4 5 6 Average the tensile strain on the bottom surface of the portion of the
f cu (MPa) 34.9 34.7 44.0 41.9 34.7 37.2 37.9 slab remote from the steel beam are greater than that of the
portion near the beam (Fig. 6(b)). This high tensile strain shown
in the figure indicates cracking of concrete as observed in the
Coupons of steel beams and concrete blocks were tested in experiments. For the convenience of comparison, results from
order to determine the stress–strain curve, Young’s modulus, the finite element method discussed in the next section are
and compressive and tensile strength. For the reinforcement, also plotted in the figure. Reasonable agreements between the
three 6-mm-diameter bars were subjected to tensile tests and the experimental and FEM results are clearly observed in the figure.
results were averaged and plotted in Fig. 3 as the stress–strain The load vs. mid-span vertical displacement curve of
curve. For the steel beams, the tensile tests were performed the center longitudinal girder is plotted in Fig. 7. The
on six specimens and the averaged stress–strain curve is also load–displacement relationship is nearly linear up to the load of
displayed in Fig. 3. According to the experimental results a 100 kN, beyond which a sudden reduction of stiffness occurred
simplified stress–strain curve shown in Fig. 4 for steel beams due to the yielding of the steel beam. At the collapse load of
and rebars is used in the finite element analyses. For the 256 kN, the steel beam section at the mid-span yielded and
concrete materials, the cubic compressive strength f cu was significantly plasticized.
determined through six 15×15×15 cm3 cubes which were cast
and tested at the same time as the deck. The measured concrete 4. Finite element analysis
strength was shown in Table 1. The cylindrical compressive
strength f c was evaluated assuming f c = 0.8 f cu . In order to predict the distributions of the longitudinal
Each of the three longitudinal girders was subjected to strains and stresses in the concrete slab of composite beams
sagging moments through four-point loads (P/4 at each point) at the ultimate strength state, a finite element analysis
as shown in Fig. 2. The load was applied by three hydraulic through ANSYS
R
(2000) was carried out considering material
1400 J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407
(a) Compressive strain on top surface of concrete slab. (b) Tensile strain on bottom surface of concrete slab.
Fig. 6. Strain distribution along slab width (b = 1200 mm, y = 0–1800 mm).
The tested specimen in Fig. 2 was analysed by finite element In the ANSYS concrete model, a crack is a mechanism that
method. 4-node shell elements were used to mesh the steel transforms the behavior from isotropic to orthotropic, where
girders and 2-node link elements were used to mesh steel bars. the material stiffness normal to the crack surface becomes zero
The kinematic hardening rule including Bauschinger effect and while the full stiffness parallel to the crack is maintained. In this
von Mises yield criteria were used for the materials of steel bars smeared crack model, a smooth crack could close and all the
and beams. Multilinear stress–strain relationship of steel bars material stiffness in the direction normal to the crack may be
and beams obtained from tests as shown in Fig. 4 were adopted recovered. The uniaxial compressive stress–strain relationship
in the analysis. For all steel materials: Young’s modulus E s = of concrete used in the analysis is:
206,000 MPa, E t = 2000 Mpa, and Poisson ratio ν = 0.3; 2
ε ε
Steel beams: f y = 295 MPa, and f u = 448 MPa; Steel bars: σ 2 − , ε ≤ ε0
ε0 ε0
= (5)
f y = 380 MPa, and f u = 478 MPa. σ0
1, ε0 < ε ≤ εcu ,
The 8-node cubic (brick) elements for concrete material
available in ANSYS
R
were used for the concrete slab. The where σ0 = f c , and ε0 = 0.002.
failure surface is the modified William–Warnke criterion as The shear studs were modeled by nonlinear spring elements
shown in Fig. 9 in the biaxial principal stress space and (shown as Combin Element in Fig. 8). Typically, the actual
the crushing and cracking of concrete are considered in this load–slip curve of stud connectors was obtained by a push-out
element [17]. The material properties of the concrete slab used test. Previous studies have shown that the curve is generally
in the analysis are: f c = 30.3 MPa, tension strength f t = nonlinear even for low stress levels. It is thus reasonable to use
3.03 MPa, elastic modulus E c = 30,000 MPa, and Poisson’s a nonlinear spring in modeling the mechanical behavior of the
ratio ν = 0.17. connectors. The constitutive relationship of the spring is given
J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407 1401
by Aribert [18]:
(a) f y = 235 MPa, b/L = 0.3, h c = 90 mm, under different loading (b) f y = 235 MPa, b/L = 0.1–0.5, h c = 90 mm, under loading q.
types.
(c) f y = 235 MPa, b/L = 0.3, h c = 60–120 mm, under loading q. (d) f y = 235–400 MPa, b/L = 0.3, h c = 90 mm, under loading q.
Fig. 11. Strain distributions on top surface of concrete across slab width (y-axis) for different parameters.
where α is a parameter representing the degree of shear-lag with the finite element results under different loading types and
effect. b/L ratios, for the case of f y = 300 MPa and h c = 90 mm.
From Eq. (7) it is derived that εct (b/2) = εct (0)(1 − The accuracy of the simplified formulae is confirmed by this
0.25α); therefore α = 4(1 − εct (b/2)/εct (0)). When the strain comparison.
distribution is uniform across the slab width, then α = 0. Based As discussed earlier, the curvature φ(y) of the concrete slab
on the FEM results of εct (0) and εct (b/2), the values of α under remains constant from y = 0 to y = b/2 under uniform load
different parameters were obtained. According to the numerical and two-point load. By using equation φ(y) = εct (y)/z c (y)
results discussed earlier, the parameters h c and f y have a small (as shown in Fig. 13) and Eq. (7), z c (y) is derived as shown in
influence on α. For example, when b/L = 0.3, the value of α Eq. (9a) below. Meanwhile, since z c (y) remains constant from
increases by about 2% when h c increases from 60 to 120 mm, y = 0 to y = b/2 under one-point load, Eq. (9b) is adopted for
and α increases by about 4% when f y increases from 235 to this case as
400 MPa. Therefore, the influence of f y and h c on α can be
y2
ignored. The values of α under different loading types and b/L y
z c (y) = z c0 1 − α + α 2
ratios, when f y = 300 MPa and h c = 90 mm, were calculated b b
and plotted in Fig. 14. By curve fitting the numerical results of (for uniform load and two-point load) (9a)
α with b/L as the sole parameter, the α’s were derived below z c (y) = z c0 (for one-point load). (9b)
and are also plotted in Fig. 14 as
From Eqs. (7)–(9) and the assumption that the longitudinal
α = 2b/L − 0.075 strain at the mid-span of the concrete slab remains co-linear
(for uniform load and two-point load) (8a) along the z-axis, the strain at the mid-span of concrete slab
α = 7.5b/L − 0.5 (for one-point load). (8b) εc (y, z) as a function of y and z can be expressed as:
By using Eqs. (7) and (8) we can analytically obtain the z − h c + z c (y)
εc (y, z) = εct (y). (10)
values of εct (y) that are plotted in Fig. 15(a)–(c) to compare z c (y)
J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407 1403
(a) Longitudinal strain along z-axis under load q. Fig. 14. Values of α under different loading types and b/L ratios.
Table 2
Results of z c0 , β and ξ with various b/L and h c / h s (under uniform load and two-point loads)
hc / hs β ξ
b/L b/L
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
0.1 1.000 1.000 0.999 0.996 0.993 0.285 0.886 0.910 0.882 0.852
0.2 1.000 1.000 0.998 0.996 0.993 0.886 0.942 0.914 0.883 0.852
0.3 1.000 0.999 0.998 0.996 0.993 0.913 0.946 0.914 0.883 0.852
0.4 1.000 0.999 0.998 0.996 0.993 0.927 0.946 0.914 0.883 0.852
hc / hs z c0 βbh f c /As f
b/L b/L
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
0.1 315.53 101.57 93.97 84.15 69.97 0.304 0.608 0.913 1.217 1.521
0.2 101.54 93.04 75.13 58.48 48.64 0.438 0.875 1.313 1.750 2.188
0.3 98.54 90.46 64.14 49.93 41.53 0.513 1.025 1.538 2.050 2.563
0.4 97.04 84.90 58.66 45.66 37.97 0.560 1.121 1.681 2.242 2.802
Table 3
Results of z c0 , β and ξ with various b/L and h c / h s (under one-point load)
hc / hs β ξ
b/L b/L
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
0.1 1.000 1.000 0.998 0.999 1.000 0.285 0.883 0.954 0.994 1.000
0.2 1.000 0.999 1.000 1.000 1.000 0.886 0.966 1.000 1.000 1.000
0.3 1.000 1.000 1.000 1.000 1.000 0.913 0.995 1.000 1.000 1.000
0.4 1.000 1.000 1.000 1.000 1.000 0.927 0.998 1.000 1.000 1.000
hc / hs z c0 βbh f c /As f
b/L b/L
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
0.1 315.53 101.63 93.01 74.44 59.17 0.304 0.608 0.913 1.217 1.521
0.2 101.54 92.67 68.59 51.41 41.13 0.438 0.875 1.313 1.750 2.188
0.3 98.55 88.31 58.54 43.90 35.12 0.513 1.025 1.538 2.050 2.563
0.4 97.05 80.43 53.53 40.14 32.12 0.560 1.121 1.681 2.242 2.802
b, b/L ≤ 0.5 axis, the results in Tables 2 and 3 can be distinguished into two
be = (12)
L/2, b/L > 0.5, situations, namely (a) and (b) as follows:
(a) When βbh c f c > As f (refer to the cases of normal fonts
where L is the span length for simply supported beam and the in Tables 2 and 3), then z c0 < h c ; at this situation the neutral
distance between the points of zero bending moments under axis lies in the concrete slab as shown in Fig. 16(a). We then
dead load for continuous beams. have force equilibrium as
For the ultimate strength analysis AISC [9] and AASHTO
codes [19,20] adopt the same effective width as that used for βbξ z c0 f c = As f. (14)
elastic analysis shown in Eq. (13) below, which is based on the (b) When βbh c f c ≤ As f (refer to the cases of bold fonts in
traditional definition of effective width shown in Eq. (1). Take Tables 2 and 3) or βbh c f c ≈ As f (the italic font in Tables 2 and
an interior girder for example, the effective width is 3, if any), then z c0 > h c ; in this situation the neutral axis lies
be = min {b, L/4, 12ts } , (13) below the concrete slab as shown in Fig. 16(b). We then have
(a) f y = 300 MPa, h c = 90 mm, under loading q. (b) f y = 300 MPa, h c = 90 mm, under loading P1 .
Fig. 15. Comparison between εct (y) values from finite element method (FEM) and simplified formulae (SF).
Table 4
Comparison of effective width at the ultimate strength state
supported project [22]. The comparison between the proposed However, the proposed formula is still more conservative than
effective width in Eq. (12), that from [21], and that specified that of Chen et al. [21] whose proposed be is b for all girders.
in the US codes (AISC and AASHTO) in Eq. (3) is listed Again, the AISC (AASHTO) is based on elastic analysis, while
in Table 4. The majority (perhaps 99%) practical structures the present study is based on the ultimate analysis that is
fall into the case of b/L ≤ 1/2. In this range the proposed supposed to be more accurate and less conservative. Due to the
effective width be in the present study is exactly the same rareness of the practical cases in the range of b/L > 1/2, there
as that of Chen et al. [21], though a different approach was is no available experimental data to directly verify the proposed
used in the two studies. By using their effective width and that formula in the range of b/L > 1/2.
specified in AASHTO code [20], Chen et al. [21] have shown
7. Discussion of the effective width
that the difference is less than 4% in terms of ultimate capacity.
Therefore, the proposed effective width and that of the AISC A few special notes are of worth and are mentioned below:
(AASHTO) are indirectly shown to be basically the same for (1) Theoretically, the obtained formulae for the effective
the case of b/L ≤ 1/2. For b/L > 1/2, while the proposed width and depth are only valid for the evaluation of ultimate
effective width could be twice that of the AISC (AASHTO), the strength. However, the effective width is also traditionally used
practical structures rarely fall into this category (perhaps <1%). for stress and deflection calculation.
1406 J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407
Fig. 16. Situations (a) and (b) for ultimate strength analysis (dashed rectangular means stress block).
(2) In the present study only simply supported decks 2. Through an experimental study and finite element analysis
are studied, and other end boundary conditions have not the distributions of longitudinal strain and stress across the
been considered. However, engineers are more concerned concrete slab at ultimate strength state are examined and
about the mid-span section that is less affected by the expressed by simplified formulae, which makes it possible to
continuity/boundary conditions. Traditionally, for simplicity, analytically derive the effective width.
engineers only distinguish between positive and negative 3. For composite beams at the ultimate strength state with
moment sections without considering the changes of effective various loading types, β, z c0 and ξ are solved from a set of
width along the span and without considering many other simultaneous equations based on the new definition of effective
factors that may affect the effective width to different extents. width and simplified formulae of stress distributions across the
(3) The effective width depends on the level of stress and concrete slab.
type of loading at the section. Therefore, a more general case 4. The effective width for the ultimate strength state is found
of stress resultants, i.e. a case of simultaneous application to be nearly the same as the physical width for the cases
of bending and axial forces, should be analysed. However, examined in the present study and a simplified effective width
considering the axial force will make the problem much more be for composite beam sections subjected to sagging moment
complicated since axial forces are variable. If axial force is an is thus proposed. Simplified formulae for calculating the depth
important component of the section forces, then we suggest of the rectangular-stress block ξ z c0 are also presented for the
using 3D finite element analysis directly. ultimate strength design of composite beams. Once both the
(4) Theoretically, effective width varies along the span width and depth of the stress block are known, the moment
length of the composite deck. Thus, the computation of resistance of composite beam sections at the ultimate strength
deflections in simply supported decks or of stress resultants state can thus be obtained by the traditional plastic section
in continuous beams can be a rather complex task due method specified in any design codes.
to the longitudinal variation of the cross-section properties.
Using a variable effective width along the span length is too Acknowledgments
troublesome and is not practical for routine application.
(5) The proposed effective width is based on limited The first two authors gratefully acknowledge the financial
finite element and experimental studies. A more meaningful support provided by the National Natural Science Foundation
verification would be a comprehensive one that should include of China (# 50438020) and the third author appreciates
many cases considering different parameters such as arbitrary the financial support from the Louisiana State University
loading, which is out of the scope of the present study and for international travel and collaboration. The authors also
perhaps should be pursued in a separate study. appreciate the constructive comments from the reviewers.
8. Conclusions
References
1. In the traditional definition, the effective width of concrete
slab is determined based on the equivalence of axial force [1] Adekola AO. Effective widths of composite beams of steel and concrete.
Structural Engineer 1968;46(9):285–9.
between the actual stress distribution and the simplified stress
[2] Adekola AO. The dependence of shear lag on partial interaction in
block. In the present study, a new definition of the effective composite beams. International Journal of Solids Structures 1973;10(4):
width is presented for ultimate strength state of steel–concrete 389–400.
composite beams under sagging moments. The effective width [3] Ansourian P, Aust MIE. The effective width of continuous composite
factor β, the position of neutral axis z c0 , and the depth of beams. Civil Engineering Transitions 1983;25(1):63–9.
the rectangular-stress block ξ z c0 are solved from a set of [4] Johnson RP. Research on steel–concrete composite beams. Journal of
Structural Division, ASCE 1970;96(3):445–59.
simultaneous equations based on the equivalencies of both the
[5] Heins CP, Fan HM. Effective composite beam width at ultimate load.
total axial force and the moment resistance, which ensures that Journal of Structural Division, ASCE 1976;102(11):2163–79.
the simplified stress distribution within the effective width will [6] Elkelish S, Robison H. Effective widths of composite beams with ribbed
represent the actual moment resistance of the original beam. metal desk. Canadian Journal of Civil Engineering 1986;13(2):66–75.
J.-G. Nie et al. / Engineering Structures 30 (2008) 1396–1407 1407
[7] Amadio C, Fragiacomo M. Effective width evaluation for steel–concrete [15] Johnson RP. Composite structure of steel and concrete, 2nd ed. vol. 1.
composite beams. Journal of Constructional Steel Research 2002;58(3): London: Blackwell Scientific Publications; 1994.
373–88. [16] Nie JG, Cai CS. Steel–concrete composite beams considering shear slip
[8] Amadio C, Fedrigo C. Experimental evaluation of Effective width in effect. Journal of Structural Engineering, ASCE 2003;129(4):495–506.
steel–concrete composite beams. Journal of Constructional Steel Research [17] ANSYS Inc. ANSYS theory reference. 2000.
2004;60(2):199–220. [18] Aribert JM. Slip and uplift measurements along the steel and concrete
[9] AISC. Load & resistance factor design, Volume 1, Part 5: Composite interface of various types of composite beams. In: Proceedings of the
design. 1998. international workshop on needs in testing metals: Testing of metals for
[10] CEN. 1994. Commission of the European communities. ENV 1994-1-1. structures. London: E. &FN Spon; 1992. p. 395–407.
Eurocode 4-Design of composite steel and concrete structures-Part 1-1: [19] AASHTO. Standard specification for highway bridges. Washington (DC):
General rules and rules for buildings, Bruxelles. American Association of State Highway and Transportation Officials,
[11] Song QG, Scordelis AC. Formulas for shear-lag effect of T-, and I-, AASHTO; 2002.
and box beams. Journal of Structural Engineering, ASCE 1990;116(5): [20] AASHTO. LRFD bridge design specifications. Washington (DC):
1306–18. American Association of State Highway and Transportation Officials,
[12] Song QG, Scordelis AC. Shear-lag analysis of T-, I-, and box beams. AASHTO; 2002.
Journal of Structural Engineering, ASCE 1990;116(5):1290–305. [21] Chen SS, Aref AJ, Chiewanichakorn M, Ahn II S. Proposed effective
[13] Elhelbawey M, Fu CC, Sahin MA, Schelling DR. Determination of slab width criteria for composite bridge girders. Journal of Bridge Engineering,
participation from weigh-in-motion bridge testing. Journal of Bridge ASCE 2007;12(3):325–38.
Engineering, ASCE 1999;4(3):165–73. [22] Chen SS, Aref AJ, Ahn I-S, Chiewanichakorn M, Carpenter JA,
[14] Chiewanichakorn M, Aref AJ, Chen SS, Ahn II S. Effective flange Nottis A et al. Effective slab width for composite steel bridge members
width for steel–concrete composite bridge girder. Journal of Structural NCHRP Report 543. Washington (DC): Transportation Research Board;
Engineering, ASCE 2004;130(2):2016–30. 2005.