NONLINEAR ANALYSIS OF COMPOSITE BEAMS WITH DEFORMABLE

SHEAR CONNECTORS

By M. Reza Salari,t Enrico Spacone,2 Associate Member: ASCE

P. Benson Shing,3 Member, ASCE, and Dan M. Frangopol,4 Fellow, ASCE

~BS~R~CT: Co~posite floor systems, consisting of reinforced concrete slabs and steel girders, are common
m bUlldmgs and bndges. Partial restraints at the interface of the concrete slab and the steel girder have significant
effects ~m the response of the composite. beam. A few analytical models capable of accounting for these effects
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

are ava~lable; ho:vever, n~ne can be effiCIently used for the nonlinear analysis of composite beams. In this paper,
t~e basIc governmg equatIons are presented for a composite beam with deformable shear connectors under small
dlsplacen:ents. Based on these eq~ations, a new composite beam element is developed using the force method
of analySIS. The performance of thIS new element is compared with that of a previously developed displacement-
?ased compo~ite .beam ele';llent. Both.elements consist of two beam components connected through a deformable
mterface. A dlstnbuted spnng model IS used to account for the shear deformation at the interface. Two numerical
examples on the ~onlinear behavior. of composite beams are solved using both displacement- and force-based
elements. Companson of the numencal results shows the superior performance of the newly developed force-
based element over the displacement-based element.

INTRODUCTION of composite steel-concrete girder bridges. However, a perfect

Composite floor systems, consisting of reinforced concrete interaction was considered between the concrete and steel
slabs and steel girders, are common in buildings and bridges. components. Hirst and Yeo (1980) modeled composite beams
In the design of composite floors, one can take advantage of with eight-node isoparametric finite elements. Standard quad-
the compressive strength of concrete slabs, provided that a rilateral elements were used to model stud connectors.
composite action exists between the concrete and steel mem- Razaqpur and Nofal (1989) developed a three-dimensional bar
bers. A perfect composite action can be obtained by prohibit- element to model the nonlinear behavior of shear connectors.
ing the relative displacement of concrete and steel sections at Four-node plane quadrilateral elements were used to model the
their interface. The chemical and frictional bonds between the web of the I-beam. The slab and flanges of the I-beam were
concrete and steel are too weak to prevent slip. Therefore, modeled by four-node flat shell elements with six degrees of
sh~ar connectors are usually added to provide a composite
freedom per node. Each shell element was divided into a num-
actIOn. However, because of the deformability of the shear ber of layers through its thickness to capture the nonlinearity
connectors, a perfect composite action cannot be obtained in across the section. A two-dimensional finite-element analysis
practice. was performed by Bursi and Ballerini (1996) to study the be-
The strength and stiffness of a composite section depend on havior of composite frames under monotonic and cyclic load-
the degree of composite action between the concrete and steel ing.
components. The degree of composite action is related to the The use of two- and three-dimensional continuum elements
geometrical and mechanical properties of the shear connectors. for the analysis of composite beams requires the solution of a
The mechanical properties can vary during the loading history. large number of equations. To reduce the computational effort,
Therefore, different sections of a composite beam may have several researchers have used one-dimensional composite
different behaviors, ranging from that of a section with a per- beam elements. EI-Tawil and Deierlein have published a num-
fect connection to that of a section with no connection. In ber of papers on the application of nonlinear beam-column
general, assuming a perfect connection in the analysis of com- elements to the analysis and design of composite structures
posite beams with deformable shear connectors overestimates (among others, EI-Tawil and Deierlein 1996a,b). Their work,
their strength and stiffness properties. On the other hand, the however, assumes perfect compatibility between steel and con-
assumption of no connection at the interface between the con- crete components; thus, bond-slip is neglected. A composite
crete and steel components is too conservative. Therefore, in beam element with bond-slip was proposed by Arizumi et al.
a realistic analytical model, the actual strength and stiffness of (1981). It consists of two beam components for the concrete
the mechanical connectors must be accounted for. slab and the steel girder, respectively, and a continuous spring
To date, only a few analytical models have been developed for shear connectors. Cubic polynomials were used to describe
for composite systems. Wegmuller and Amer (1977) used a both the deflection and axial deformations of the beam. A com-
layered beam-plate model to describe the postelastic response posite beam element with a continuous spring model was used
by Daniel and Crisinel (1993) for the strength analysis of com-
IRes. Asst., Dept. of Civ., Envir. and Arch. Engrg., Univ. of Colorado, posite slabs. The same element was used by Amadio and Fra-
Boulder, CO 80309-0428. giacomo (1993) to study the creep and shrinkage effects in
2 Asst. Prof., Dept. of Civ., Envir. and Arch. Engrg., Univ. of Colorado, composite beams. Salari et al. (1997) employed the same el-
Boulder, CO. ement to study the behavior of composite structures under cy-
3Prof., Dept. of Civ., Envir. and Arch. Engrg., Univ. of Colorado, Boul-
der, CO.
clic loading. In this element, the beam deflection is approxi-
"Prof., Dept. of Civ.. Envir. and Arch. Engrg., Univ. of Colorado, Boul- mated by Hermitian cubic polynomials and quadratic functions
der, CO. are used for the axial displacements. Hajjar et al. (1997) ap-
Note. Associate Editor: W. Samuel Easterling. Discussion open until plied a similar element to study the effect of interlayer slip on
March I, 1999. To extend the closing date one monlh, a written requesl the response of composite frames with concrete-filled tubes. A
must be filed with the ASCE Manager of Journals. The manuscript for
this paper was submitted for review and possible publication on Decem-
simpler composite beam element was used by Aribert and Rag-
ber 8, 1997. This paper is part of the Journal of Structural Engineering, neau (1997) to investigate the moment redistribution in con-
Vol. 124, No. 10, October, 1998. ©ASCE, ISSN 0733-9445/98/0010- tinuous composite beams. The element consists of two Ber-
1148-1158/$8.00 + $.50 per page. Paper No. 17215. noulli beam elements. The axial and transverse displacements
1148/ JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158

 are interpolated using linear and cubic functions, respectively.
The shear connection between the concrete slab and the steel
girder is modeled by connector elements at the end nodes.
The use of Hermitian interpolation functions for beam de-
flection leads to a linear curvature distribution along the beam,
which reduces the accuracy of the element in the nonlinear
range. This problem is typical of displacement-based beam
elements when highly nonlinear curvature distributions are ex-
pected under high load levels. A study by Spacone et al. (1996)
on the nonlinear analysis of reinforced concrete frames indi-
cates that the use of force-based elements, with the elemental
internal forces as the principal unknowns, can improve the
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

accuracy of numerical results to a significant extent. More re-

cently, the theory has been extended to include geometrically
nonlinear problems (Neuenhofer and Filippou, 1998).
The results obtained by Spacone et al. (1996) suggest the
extension of force-based elements for the modeling of com- FIG. 2. Kinematics of Composite Beam
posite systems. A force-based composite beam element was
employed by Ayoub and Filippou (1997) for the nonlinear For the finite-element formulation of a composite beam, one
analysis of steel-concrete composite structures. The concrete has to assume the displacement and/or force fields that govern
and steel components were modeled with the same force-based the response of each beam component and the interface. In the
beam element developed by Spacone et al. (1996). The inter- composite beam element presented here, it is assumed that the
action between the concrete and steel components was mod- beam behavior is dominated by the axial and flexural defor-
eled with spring elements. The springs were lumped at the end mations of the concrete and steel sections and the shear de-
nodes, but the interface shear was linearly distributed over the formation of the interface. Shear deformations of the concrete
element length. and steel sections are neglected. With these assumptions, the
This paper proposes a new element formulation for the anal- following element displacements, forces, and section defor-
ysis of composite beams under small displacements. This el- mations are used in the formulation.
ement is superior to most existing elements in terms of the
accuracy and computational efficiency by modeling the inter- Element Displacements
nal forces in a more precise and consistent manner. First, the
basic equations that govern the behavior of a composite beam To simplify the modeling of a composite beam, one can
with deformable shear connectors are derived. The equations assume that the transverse deflections of the two beam com-
are then used to develop a composite beam element in a force- ponents, i.e., the concrete slab and the steel girder, are the
based finite-element framework. For comparison purposes, a same. However, different axial displacements may occur in the
displacement-based composite beam element is also consid- concrete and steel sections. Therefore, the following displace-
ered. Both elements consist of two beam components and an ment components are adopted here for the concrete and steel:
interface. The interface shear in the force-based element is y(x) = (v(x) uoC<X) uo.(x) f (1)
approximated with a cubic polynomial. The higher order shear
approximation and the continuity of interface flexibility are the in which L(x) is the transverse deflection of the composite
advantages of the proposed element over the existing force- beam; and uoc(x) and uo.(x) are the axial displacements of the
based element. Two numerical examples are solved by using concrete and steel components, respectively. These displace-
the force-based element and the results are compared with ments are shown in Fig. 2. Having the beam deflection and
those obtained using the displacement-based element. The rate axial displacements, the relative displacement between the
of convergence of numerical results with respect to the number concrete and steel components along the interface is
of elements is also investigated in this study. uc.(x) = Hv' + (uoc - uos) (2)
in which H = distance between the reference axes of concrete
BEAM DEFORMATIONS AND INTERNAL FORCES and steel cross sections; and the prime symbol indicates the
A composite beam generally consists of two components, derivative with respect to x. Because the two beam compo-
e.g., a concrete slab and a steel girder, connected by shear nents have the same transverse deflections, no relative dis-
connectors. Steel decking can be used to simplify the casting placements normal to the interface can be expected.
process of the concrete slab. It can also be designed to act Element Forces
with the concrete slab to reduce the amount of reinforcement.
A typical section of a composite beam is schematically rep- By ignoring the shear deformations of the beam compo-
resented in Fig. 1. The conventional beam theory can be used nents, only the following internal forces need to be considered
to model the concrete slab and steel girder. The effects of the for the concrete and steel sections:
deformable shear connectors on the response of a composite D(x) = (M(x) Nc(x) N.(x) f (3)
beam can be accounted for by using an interface model. Al-
though the shear connectors are generally discrete, a distrib- in which M(x) is the sum of the bending moments in the con-
uted interface with equivalent strength and stiffness properties crete and steel sections; and Nlx) and N.(x) are the axial forces
can be used to simplify the model. in the concrete and steel sections, respectively. In addition, by
ignoring the relative displacements normal to the interface,
only the tangential component of the bond force, Db(x), needs
to be considered. The element forces are shown in Fig. 3.
Steel Decking Section Deformations
Steel Girder
The two beam components have the same transverse de-
FIG. 1. Schematic Representation of Composite Section flections, and thus the same curvature distributions. However,
JOURNAL OF STRUCTURAL ENGINEERING /OCTOBER 1998/1149

J. Struct. Eng., 1998, 124(10): 1148-1158

 (9)
M fL----'-----I....-

~
(10)
N,; Nc+dNc
T ~ ~~~~~Vc+dVc Eqs. (8)-(10), which represent the equilibrium conditions

r ~f s
N - ~I
V,
D. - l~dM'
~~,
~+dV.
:Ns+dNs
for a composite beam, can be written in the following matrix
form:

in which P = {py 0 O} T. and the differential operators iJ and

(11)

I---dx--l iJ b are given as

Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Infinitesimal Segment of ComposIte Beam d2

0 0
dx 2
the axial strains in the concrete and steel sections are inde-
pendent. The section deformations corresponding to the sec- iJ= 0 -d 0 (12)
tion forces in (3) are as follows: dx
d
(4) 0 0 -
dx
in which <!>(x) is the curvature of the composite beam; and
£oe(x) and £olx) are the axial strains at the reference axes of
the concrete and steel sections, respectively. The other com- iJ=[H~ -1 ] (13)
ponent of section deformations that must be considered is the
bond slip db(x) along the interface, which is associated with It can be observed that any section of the composite beam
the bond force Db(x). contains four unknown forces: M, N e , N.. and Db' On the other
hand, only three independent equilibrium equations are avail-
BASIC GOVERNING EQUATIONS able at each section. Therefore, a composite beam is internally
indeterminate and its internal forces cannot be obtained
The behavior of a deformable body must satisfy three basic through equilibrium alone.
conditions. These conditions are equilibrium, compatibility,
and material constitutive laws. For the composite beam ele- Compatibility
ment under consideration, these conditions are given in the
following subsections. The flexural and axial deformations at any section of a com-
posite beam are related to the beam displacements through
Equilibrium compatibility conditions. Under small displacements, the com-
patibility conditions for the concrete and steel components are
To derive the differential equations of equilibrium for a
as follows:
composite beam, the infinitesimal beam segment shown in Fig.
3 is considered. Normal contact forces between the concrete
and steel components are neglected in the proposed formula- <l>} = {V"}
{ £oc u:x: (14)
tion. Equilibrium of the segment in the y-direction, which in- £05 U~s
cludes the concrete and steel components, leads to the follow-
ing expression: Furthermore, compatibility must hold between the interface
deformation, db, and the relative displacements of the two
(5) beam components at the interface, U c.' Using (2), this com-
in which Vc and Vs = shear forces in the concrete and steel patibility condition is
sections, respectively; and Py = vertical distributed load. The
moment equilibrium for the concrete and steel segments results
db = Ucs = Hv' + (u oe - uo.) (15)

in the following relations: Eqs. (14) and (15), which represent the compatibility con-
(6)
ditions for the composite beam under consideration, can be
written in the following matrix forms:
dMs - Db dx hs - Vs dx =0 (7)
(16a,b)
in which Me and M s are the bending moments in the concrete
and steel sections, respectively; and he and h s are the distances where y is the displacement vector defined in (1); and iJ and
from the reference axes of the concrete and steel sections to iJbare the differential operators given in (12) and (13), re-
the interface, respectively, with H = he + h•. Because the shear spectively.
deformations of the concrete and steel sections are neglected,
the shear forces Ve and V. need not be considered in the ele- Force-Deformation Relations
ment. Therefore, the shear forces should be eliminated from
the equilibrium equations. This can be done by dividing (6) The force-deformation relations for a composite section de-
and (7) by dx and taking derivative with respect to x and add- pend on the material properties and cross-sectional geometry
ing the resulting equations. With M = Me + M. and substituting of the beam. For a linearly elastic section, these relations are
dVe + dV. with (5), one obtains as follows:

M" - HD; - py = 0 (8) d =f D (17)

Finally, equilibrium of the horizontal forces in the two in- in which f is the section flexibility matrix for the concrete and
finitesimal concrete and steel segments leads to the following steel components. The flexibility matrix f at any section can
equations: be obtained by inverting the corresponding stiffness matrix, k.
1150/ JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158

 The section stiffness matrix for a linearly elastic composite
beam is given as V ={v 1 e1 V2 e2V
ue={uel U e2 ue3 V
LEcY~ + L
.4"
dA
A.
E,y; dA -L"', EeYc dA -L.... E,y, dA U,={U'1 U,2 us3 }T

k= -L"', EcYc dA L"', EcdA 0 (18)

FIG. 4. Displacement-Based Composite Beam Element

unknowns are expressed in terms of element nodal displace-

-L.... E,y, dA 0 L....
E,dA ments together with appropriate displacement shape functions.
Section deformations are obtained through the compatibility
conditions. Therefore, the compatibility conditions are satisfied
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

in which Ee, A e and E" A, are the elastic moduli and areas of
concrete and steel sections, respectively. As shown in Fig. 2, the in a strict sense along the element. By applying material con-
ordinates Ye and y, are measured from the reference axes of the stitutive laws at any section, section forces can be obtained.
concrete and steel sections, respectively. In an incremental non- Section forces must satisfy the equilibrium conditions in (11).
linear analysis, the elastic moduli Ee and E, are replaced with the To enforce these conditions, (11) can be multiplied by a kin-
corresponding tangent moduli. In this case, the integrations over ematically admissible virtual displacement field 8y, and inte-
the concrete and steel sections are performed numerically by dis- grated over the length of the element as follows:
cretizing the beam cross section into fibers (layers).
Finally, the relation between bond force and bond slip at
the interface between the beam components can be introduced
i 8y T(81» - 8rDb - P) dx =0 (20)

through the bond constitutive law as follows: Eq. (20), which is the integral form of the equilibrium equa-
db =/J, Db (19) tions, is the basis for a displacement-based finite-element for-
mulation. A composite beam element of this type is shown in
in which fb is the bond flexibility. Fig. 4. The unknown displacements are those presented in (I),
and the relative displacement along the interface can be ob-
ELEMENT FORMULATION tained from (2). The interface is modeled by a continuous
The basic governing equations previously derived represent spring element. To approximate the displacements of the com-
the equilibrium and compatibility conditions and the material posite beam, cubic Hermitian polynomials are employed for
constitutive laws for a composite beam. These equations, along the transverse deflection v and quadratic functions are used for
with appropriate static and kinematic boundary conditions, the axial displacements Uoe and Uo,' These assumptions lead to
govern the response of a composite beam to applied loads. a quadratic bond slip distribution along the interface. The
However, different combinations of these equations may lead quadratic axial displacements are achieved by introducing an
to different solution schemes. Starting from the equilibrium extra node in the .middle of the element with two axial degrees
equations, with displacement fields as principal unknowns, one of freedom.. one m each beam component. The displacements
can obtain a displacement-based finite-element formulation. at any sectIOn of the element are written in terms of the nodal
On the other hand, assuming the internal forces as the principal displacements as follows:
un~owns and starting with t~e compatibility equations, one vex) = avv; uoc(x) = aelle; uo.(x) = a,ll, (21a-c)
obtams a force-based formulatIOn.
in which a", a e, and a, are the displacement shape functions
In this paper, attention is focused on a force-based finite-
that describe the beam deflection and axial displacements of
element formulation of composite beams. For comparison pur-
the concrete and steel components, respectively. The nodal dis-
poses, a displacement-based composite beam element is also
presented. The following two sections deal with the two ele- placements v, U e, and u, are shown in Fig. 4. A detailed de-
viation of the element formulation is presented by Amadio and
ment formulations.
Fragiacomo (1993). In the present paper, the element is used
DISPLACEMENT-BASED ELEMENT to compare the performance of displacement-based and force-
based composite beam elements. The element stiffness matrix
In a displacement-based finite-element formulation element in terms of the displacement shape functions is given in the
displacements are considered as principal unknow~s. These following:

i (y~a~TEea~ y;a~TE.a~) + dv + -i Yc8~TEc8; dv + -Ly.a~TE.a: dv -

i a~TEba~
H
2
dx
•
H1 L
a:TEbae dx H L a:TEba, dx
........"-"J'.v. y" ~;'T'E" ~ 'v','d"v"+'
e- e- : . ..

L a;TEea; dv +
K=
Ia~Eba:
H dx r
JL a~Ebae dx
(22)

........................................... ............................................ : .
-Ly.a:TE,a~ dv - L a:TE.a: dv +

H La;Eba~ dx L a;Eba, dx

JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998/1151

J. Struct. Eng., 1998, 124(10): 1148-1158

 where Eb is the bond stiffness. in the beam components, D, and the bond force, Db, are in-
terpolated in terms of the nodal forces Q, and the bond forces
FORCE-BASED ELEMENT Qb' at selected reference points along the beam. The resulting
expression is
In a force-based finite-element formulation, the element in-
ternal forces are the principal unknowns. In general, the un-
(28)
knowns are interpolated in terms of the element nodal forces.
By deriving the interpolation functions from the equilibrium
conditions of an element, equilibrium is strictly satisfied along where b, C, bb' and Cb are the force shape functions and are
the element. defined later on in the paper.
To satisfy the compatibility conditions, a statically admis- Next, (28) is substituted into (27). The boundary terms in
sible virtual force field is applied. Using the virtual forces 8D, (27) represent the external virtual work and are written in
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

and 8Db for the beam components and interface, respectively, terms of virtual nodal forces 8Q and nodal displacements q.

i {:8h r[:b :JT[~ 1] [:b :J {~}

one can enforce the compatibility conditions of (16) in the With these substitutions, the following expression is obtained:
following integral form:

i 8DT (d - iJy) dx + i 8Db(db - iJby) dx =0 (23)

= {:8JT {~}
dx

(29)
This implies that in a force-based formulation, compatibility
is satisfied in a weak sense. To satisfy the constitutive laws Performing matrix multiplication and noting that the virtual
for the concrete, steel, and bond materials, (17) and (19) are force field is arbitrary, (29) reduces to the following expres-
used. Therefore, (23) takes the following form: sion:

i T
8D fD dx + i 8DbfbDb dx - i T
8D iJy dx - i 8DbiJby dx =0
(24)
Rewriting the last two integrals of (24) in terms of the com-
ponents of internal forces and displacements and performing or
integration by parts, we have

i 8D TfD dx + i 8Dbf~b i dx - (8M" - H8D~)v dx

(31)

+ i (8N; - 8Db)uoc dx i + (8N; + 8Db)uos dx

The zero term on the right side of (31) indicates that the
relative slips corresponding to the reference bond forces Qb
are equal to zero. This is equivalent to the compatibility con-
ditions of bond deformations, which can be used to calculate
- [8Mv' - (8M' - H8D b)v + 8Ncuoc + 8NsUos]~ =0 (25) the redundant bond forces Qb' This is done by static conden-
sation of (31) as follows:
Eq. 25 represents the integral form of the compatibility con-
ditions including the constitutive laws. Now, by applying the (32)
equilibrium conditions of (8)-(10), and setting 8py = 0, the
variational expression reduces to the following form: By substituting (32) into the first row of (31), the governing
equation of the element is obtained as
i 8D TfD dx + i 8Dbf~b dx - [8Mv' - (8M' - HBDb)v FQ=q
where F is the element flexibility matrix, given by
(33)

(26)
F = f QQ - fQQJa:QJQ.Q (34)
Eq. 26, which satisfies the compatibility conditions in in-
tegral form and the equilibrium and constitutive laws point-
wise, is the basis for the force-based composite beam element.
This equation is in fact the expression of the complementary
virtual work theorem, where the integral terms represent the
internal virtual work corresponding to the beam components
and interface, respectively, and the boundary terms indicate
the external virtual work. The equation should be discretized (a)
in terms of the element nodal forces and displacements to ob-
tain the governing equations for the element. To simplify the

i {:gb r[~ X] {~}

presentation, (26) is rewritten in the following matrix form:

dx - [Boundary Terms] =0 (27)

Finite-Element Discretization (b)

Eq. 27 represents the continuous variational form of the FIG. 5. Force-Based Composite Beam Element with Deforma-
governing equations for the beam element. To discretize the ble Shear Connectors; (a) without Rigid Body Modes; (b) with
equation in terms of the nodal quantities, the internal forces Rigid Body Modes

1152/ JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158

 To use the present fonnulation in a general purpose dis- By substituting D/,(x) from (35) into (36) and integrating the
placement-based finite-element analysis code, the flexibility left side, the bond force Qb4 is
matrix F must be inverted to obtain the element stiffness ma-
trix. This can be done by excluding the element rigid body
modes. The rigid body modes are added after inverting the
flexibility matrix. The force-based composite beam element
with and without rigid body modes is shown in Fig. 5. By substituting (37) into (35), the bond force distribution in
tenns of the element nodal forces and bond reference forces
Force Shape Functions is obtained as follows:
In a force-based beam element, the internal forces are in- (38)
terpolated using force shape functions. For a statically deter-
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

minate beam, the force shape functions obtained from equilib- in which Q is the vector of nodal forces corresponding to the
rium represent the exact distribution of internal forces. For the degrees of freedom shown in Fig. 5(a), Qb = {Qbl Qb2 Qb3}T,
composite beam element, which is internally indetenninate, and the bond force shape functions bb and Cb are
the exact distributions of internal forces are not available, ex-
cept for some special cases with linear elastic behavior. How-
ever, if the bond force distribution is known, all other internal
forces can be obtained through equilibrium. In the present for-
mulation, the bond force is treated as a redundant force and
must be detennined using compatibility conditions. The num-
ber of compatibility equations is equal to the degree of inde- 0 0]
.~ .~
(39a)
tenninacy. However, for a composite beam with a continuous
bond, the degree of internal indetenninacy is infinity. There-
fore, some approximations are required to simplify the for- -9 27 13 : -9 6:
c(x)= [ -K+-r--x+l'-~+-x'
mulation. This can be done by interpolating the continuous b L3 2e 2L : L2 L:
bond force in tenns of the bond forces at a finite number of
reference points along the element.
In the present fonnulation, the bond force distribution along
(39b)
the element is approximated with a cubic interpolation func-
tion. As shown in Fig. 6, a cubic bond force distribution re- With the above bond force distribution, the internal forces
quires four reference forces, Qbh ... , Qb4. The bond force in each beam component can be obtained from equilibrium as
distribution Db(x) in tenns of these reference forces is follows:
-9 9 11 : 27 45 9:
D(x)=
b [ -x3+-x2--x+l'-x3--~+-x'
2L3
L 2
2L : 2L 3
2L 2
L: (40)

-27
--K+-X
3 2
2L
--X'-X
18
3 --x
L 2
2 9
2L:
: 9
2L
3 9
2L
2
+-xJ1
L
Qbl}
Qb2
{ Qb3
Qb4
(35)

However, the bond force at one reference point, say Qb4' is

M(x) = H(1 - i) Q. - (1 - i) Q2 - Hi Q3 + i Qs
related to the bond forces at the remaining three points and
the nodal forces through equilibrium. This relation is obtained + HNc(x) (42)
by enforcing equilibrium in the x-direction of the top or bot-
tom beam component of the composite beam element shown Upon substitution of the bond force from (38) into the above
in Fig. 7. It follows that: three equations, the expressions for the internal forces are

(43a)
(36)
(43b)

M(x) = b..,(x)Q + C..,(X)Qb (43c)

The shape functions corresponding to these internal forces

are given in the following. It can be observed that both axial
forces and bending moment have fourth-order distributions.
I" L/3 "I" £/3 "I" £/3 _I By increasing the degree of bond force interpolation function,
the axial and flexural behavior of the element will also im-
FIG. 6. Cubic Bond Force Distribution along Composite Beam prove.

bNC<x) = [ L49 X
4
-
12
L3
.3
X
4 2
+ L2 x - I :
::

:
9 4 12
0 : 4 x - 3' K
:L L
4
+ - 2 x2
L

FIG. 7. Forces on Beam Components

.~ 0 .~ oJ
(44a)

JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998/1153

J. Struct. Eng., 1998, 124(10): 1148-1158

 -9 9 13 Filippou (1997) can be used to obtain the element resisting
= [ -4L3 x + -2L r - - x +
4 2
c (x) 2 X forces. Both methods use the element stiffness matrix from the
Ne 4L
last iteration to calculate the nodal force increments. The sec-
tion force increments corresponding to the nodal force incre-
-27 x4 + ~r _ ~X2] ments are obtained using force shape functions. The section
4L3 2L2 4L (44b) flexibility matrix from the last iteration and the section force
increments are used to calculate the increments in section de-
b (x)
Ns
= [_.2..
L4
x4 + 12
L3
r - ..±.
L
x 2
2 + 1:0 formations. The section deformations are then updated and the
section resisting forces and section flexibility matrix are ob-
tained using the material constitutive laws. In general, the sec-
9 4 tion forces obtained from section constitutive laws do not sat-
- -4 x + -123 X 3 -
4
- X
2
+ 1: 1 : 0
: ]
isfy the equilibrium conditions. The unbalanced section forces
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

L L L2 (44c)
introduce residual section deformations, which are integrated
to obtain the corresponding nodal residual displacements. The
C
Ns
(x) = [ -4L9X
3
4
--X
2
2L
9 3 13
+ - x - x ' -2x
4L :L
2 : 3 _3 3
--X
L
2 nodal residual displacements, which violate the nodal com-
patibility conditions, are used to calculate the nodal residual
forces. In the method developed by Spacone et al. (1996),
these residual forces are eliminated by carrying out the afore-
: 27 4- -21X3+15
'-x -x
2] mentioned process iteratively inside the element. However, in
: 4L3
2
2L 4L (44d) the simplified method of Neuenhofer and Filippou (1997), the
element residual forces are transferred to the structure level
9 x4 - -
12x3 and are eliminated in the global iterations.
b fX) =[H (- + -42x 2 1 ) :. -
- -x 1x-
M\ L4 L3 L L :L Calculation of the nodal residual displacements is performed
using the virtual work theorem. A statically admissible virtual
force system is applied and the resulting external virtual work,
9 x4 - -
H (- 12x3 +-x
4 2 - -1x) :'O'-x
: 1 ] SWE' is obtained as:
4 3 2
L L L L : :L (44e)
(47)

c =[H -9 4 + -9 3 13 2
+ X) : where SQ and qr are the virtual nodal loads and the nodal
M\
fX)
( -4L2 x 2e X - -
4L
X .
: residual displacements, respectively. For the composite beam
element under consideration, the internal virtual work, SWlo is
obtained in terms of internal virtual forces SD and SD b and
-3
H (- X
3+ 3 2): H (-273 4+
-X . --X
21
-X
3- -15x2)] their corresponding section residual deformations d r and dbr as
L2 L : 4L 2
2L 4L (44f) follows:

ELEMENT RESISTING FORCES

In an iterative nonlinear analysis, resisting forces are re-
SWI = i T
8D dr dx + i 8Dbdbr dx (48)

quired to calculate the unbalanced loads acting on the struc- By substituting D and Db with (28) and using (32), the in-
ture. Convergence is achieved if the unbalanced loads are ternal virtual work takes the following form:
within a specified tolerance. The resisting forces of a structure
can be obtained by assembling the resisting forces of its ele-
ments. 8 WI = i (b8Q + C8Qb/dr dx + i (bb 8 Q + Cb 8 Qb)dbr dx

(49a)
Displacement-Based Element
In the displacement-based composite beam element, the el-
ement resisting forces are calculated by integrating the section
8WI = i (b8Q - cfii:Q.fQ.Q8Q/dr dx

resisting forces as follows:

Q= i aTD dx + i arDb dx (45)

+ i (bb8Q - cb f ii:Q.fQ.Q8Q)dbr dx
(49b)

Finally, using the virtual work theorem and considering that

in which the deformation shape functions a and ab, corre- the virtual nodal loads 8Q are arbitrary, the nodal residual
sponding to the beam components and interface, respectively, displacements can be obtained as follows:
are given as:

a =
a"
0·
0 0]
a; 0 , (46a,b)
qr = i (b - cfii:Q.fQ.Q/dr dx + i (bb - cbfii:Q.fQ.Q/dbr dx
(50)
[ o 0 a;
NUMERICAL EXAMPLES
Force-Based Element
The performances of displacement-based and force-based
Calculation of the element resisting forces in force-based composite beam elements are compared in two numerical ex-
elements is not as straightforward as in displacement-based amples. In these examples, the simply supported composite
elements. The reason is that the deformation shape functions beam of Fig. 8 is considered. The beam consists of two steel
used in (45) are not available in force-based elements. For this components with 20 X 50 mm rectangular cross sections. The
reason, the iterative procedure developed by Spacone et al. steel behavior is elastic-perfectly plastic with yield stress h =
(1996) or its simplified version introduced by Neuenhofer and 300 MPa and elastic modulus Es = 2 X lOS MPa. An elastic-
1154/ JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158

 sommI
I¥F I"
~~~~

L = 2000 mm
1 ·1
sommI
D by = 200 N/mm
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

FIG. 8. Simply Supported Composite Beam

1.6E+07
_~rf.1t ~~d_ 6 ,,
.. -. .... - --n·1 Ht 8
I
I
UE+07 .· fI--
I
- 9 = 0.01 _Numerical
--9 = 0.10 Numerical
ll
•
.LJ.. 1I.8~H
II,
.,.··
--- n.2 I
-9=0.01 Exact 21Iemen"~1fSEY n
1.2E+07 ······n.4 "---
I
I
--9=0.10 -Exact ,
'

-'-'n-16
, l\ l
1.0E+07 , ,..... _ .. - n·32 , 0,.
,, }" I ,
f 8.OE+06
·
, ~ .--_
,"'-'-............ -...----
.e'
....
----
........ 0 •• 0 ...........
I
I

\
,
II'
/1
I

/1 :;.,--- l'
~_.t.lru ~ ,..". .... ~r

. .. -.. - - ••• 'r- Io'Bo~d

~ ,I
\ /
:II ' .... ~'"

·.
6.0E+06
-~ "',Qo.:- bL-~
4.0E+06
f/ ·
I

C11,9
o
o ~ ~ ~ 1~
_", ..,so",

1~ 1~ 1m ~

:I
1I,9
I
I
j
2.0E+08
I ;ti 4 x (mm)
~/ (a)
O.OE+OO 6
0.00 0.10 0.16 0.20 0.26 I
0.06
,I
11.8.1 ""
H(~8
" -9=0.01 Numerical
r}-
9 (Red)
--9=0.10 Numerical Illi"'I ',
(a) I" ,
~
-9=0.01 Exact 21 lementll Half8EYft :'
1.8E+07
I:
,I
--9=0.10 Exact
I,""
.. , .... - -'--
Perte tBond I ,I

.
, I
.'I
..·.·
UE+07 : I I
-
--n·1 \ I
~"
-
\I
1.2E+07 \l ~

,.
--- n - 2 -
..... --'"-,
{
1.0E+07
··
··· .. ·n·4
o
o
' ':.'':.111

~
~ ..
~
1---

m 1~ 1~
-- \.-_

1~ 1m
~'"

~
:II
8.OE+08 .i f
I

.. ...- - .,. 'fI o.-. -

Bond
x (mm)

. .'
(b)
8.0E+06

4.0E+06
f/ · I
# FIG. 10. Curvature Distribution under Pure Bending with: (a)
Displacement-Based Elements; (b) Force-Based Elements

2.0E+08 :I - - I
I
I ~ 11,9
A
11,9 ~
4 The composite beam is analyzed under two different load
11/
I
cases. The first analysis deals with the behavior of the beam
O.OE+OO under pure bending. Two different boundary conditions can be
0.00 0.06 0.10 0.15 0.20 0.26 used to impose bending moments at the ends of the beam. One
is to apply a bending moment at the rotational degree of free-
9 (Rad)
dom. The other is to impose two equal and opposite axial
(b) forces at the two axial degrees of freedom at each end. Even
though the two aforementioned boundary conditions are stat-
FIG. 9. Moment-Rotation Diagrams under Pure Bending with: ically equivalent, they result in different responses for the
(a) Displacement-Based Elements; (b) Force·Based Elements
composite beam. This can be attributed to the different relative
displacements developed between the concrete and steel sec-
perfectly plastic constitutive law is considered for the interface tions under the two conditions. In this study, the former
with elastic stiffness E b = 1,000 MPa, and shear strength Dby method is used to apply the state of pure bending.
= 200 N/mm. Each steel section is divided into 15 layers to The moment-rotation diagrams obtained by the displace-
perform the numerical integrations over the section. ment-based and force-based elements are shown in Figs. 9(a)
JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998/1155

J. Struct. Eng., 1998, 124(10): 1148-1158

 and (b), respectively. The closed fonn solutions for the two The bond force distributions along the beam at different
extreme cases that correspond to a perfect bond and no bond levels of end rotation 9 are shown in Fig. 11. For both dis-
are also presented. Due to symmetry, only one-half of the placement- and force-based elements the numerical results are
beam is analyzed. As shown in Fig. 9, for small end rotations, obtained using two elements per half span. 1\\'0 displacement-
say 9 < 0.04 rad, either a single displacement-based or a single based elements cannot accurately capture the overall behavior
force-based element is able to predict the response of the beam of the beam and this mesh is used here for comparison with
accurately. For large values of 9, however, the two element the same number of force-based elements. A discontinuity can
types show different perfonnances. A single force-based ele- be observed in the bond force distribution between two adja-
ment provides almost the same precision as 32 displacement- cent force-based elements. This happens because the bond
based elements. This is essentially because the force-based el- forces at the reference points are considered as internal degrees
ement provides a better estimation of the curvature variation of freedom and are condensed out in the element fonnulation.
in the nonlinear defonnation of the composite beam than the This discontinuity does not exist for the axial force and bend-
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

displacement-based element. The curvature variations shown ing moment variations because their corresponding nodal val-
in Fig. 10 are obtained with two elements per half span. As ues are considered as external degrees of freedom. In spite of
shown in Fig. 10, for low values of 9, both elements can ap- the bond force discontinuity, force-based elements provide a
proximate the curvature variation along the beam with a good more accurate moment-rotation behavior than the displace-
accuracy. For large values of 9, however, the highly nonlinear ment-based elements. It should be mentioned that even though

--
curvature variation developed at the two ends of the beam
cannot be adequately captured by the linear approximation as- 3.6E+04
~
sumed in the displacement-based element. A discontinuity is
also observed in the curvature variation between two adjacent
3.0E+04
Perfect IBond.......
,--- ---- ----
displacement-based elements. For the force-based element,
I' ~-
••• _ ••••• 6 •

~~l"_e
.:.0;;;;:;;';; 1:0=1::= :::::=:.
however, the curvature variation that is obtained from the mo-
ment-curvature relationship is more accurate. In this case, the 2.6E+04 i~ --n·1
--- n. 2
bending moment continuity between two adjacent elements re-
sults in a continuous curvature distribution.
2.0E+04
:'/ .... ··n·4
-·-·n·8

250 ~
Il.
I, I:11 - .. -n·18

200 - 9 0.01 _.- -- 1.5E+04 .....

.'
•-., I
I ,, Ne Bond
,,
..,.
--9=0.03
150 ·.·9 = 0.05 /
I I Ii
,
E 100 """'9 0.10 1.0E+04 ,
1/ ,,
'
l
j
60
0
, .. - V ........ -; .,
~
/.
."" ....
!:.-
.... J

/"
5.0E+03 J
If",,, " A
t P,.1
-60
'a
j 1'-

! -100
I I
,"/
p
Iq Af~'t=
i
O.OE+OO
I I I

-150 - o 10 20 30 40 50 80
-200
V I . " -'.... 21flement8(Half SPllin
A (mm)
-250 (a)
o 260 500 750 1000 1260 1600 1760 2000
x (mm) 3.5E+04

Perfec Bond
(a) 3.0E+04
.I'
2.5E+04
// P""""
n·1
260
200 - 9 0.01
--9=0.03
. .4
Ioo-~-

/ 2.0E+04
:/ --- n·2
.. ·· .. n·4

E
i
160
100
···9 = 0.05
.....,. 9 = 0.10
_.
I I
I

V
/ -
Z
I l.
1.6E+04
.
,,
:11
.....
....'" j,!..../ N Bond

I , , ...
60
~~
~ /-t -".;;.
..... .
,
• 0 1.0E+04
-
]
-50
-100
-160 f-
/
I .1
--~
I :/
1~,8
..... Af,8'
Ii1i t= 5.0E+03 I/,,/ ,"
,"

:rr
t P,.1

·200
·260
o
- 260
,f/

600 760
-----~~~~~-
1000 1260 1600 1750 2000
O.OE+OO
,
o 10 20 30
I I
40 50
I I
80

x (mm) A (mm)

(b)
(b)
FIG. 12. Load-Deflection Diagrams under Concentrated Load
FIG. 11. Bond Force Distribution under Pure Bending with: (a) with: (a) Displacement-Based Elements, (b) Force-Based Ele-
Displacement-Based Elements; (b) Force-Based Elements ments

1156/JOURNALOF STRUCTURAL ENGINEERING/OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158

 250
the constitutive behaviors of the steel and bond are assumed
to be elastic-perfectly plastic, a small hardening is introduced
into the material models in the force-based element to prevent
numerical problems in flexibility calculation in the plastic E
200
150
100
-A=IO
--A=15
.• - A=20 .
~
I : .....
J '/
"
. -- _.. - ~--

_. A=30
range. As a result, the bond force at the two ends of the beam, 1://
which experience very high levels of relative displacements, ~ 50
• '1'

exceeds the intended shear strength of the bond. ~ 0 P

The behavior of a simply supported composite beam under {l -60 A
a concentrated load at midspan is also studied with both dis-
l -- ---- Vl:'j
t
-.
-100
placement-based and force-based elements. The load-deflec- /o' I It +P,4
1
tion diagrams are shown in Fig. 12. Due to symmetry, only
one-half of the beam is analyzed. As shown in Fig. 12, only
-150
·200 ~-- --- -",
.....
...I 2 ~1eme~81f SP'In
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

one force-based element is needed to obtain almost the same

-260
accuracy as 16 displacement-based elements. The curvature
variations along the beam obtained with displacement-based o 250 600 750 1000 1260 1600 1760 2000
and force-based elements are shown in Fig. 13. Both distri- x(mm)
butions are obtained using two elements per half span.
As shown in Fig. 13, for small values of a, both elements (a)
can capture the curvature variation along the element success-
fully. For large values of a, a very large curvature is developed
under the concentrated load, which is successfully approxi- 250 -r--"",,-""""--r--....,..----,---r----,--.....,
mated with the force-based element. However, a large error is
obtained with the displacement-based element. The bond force 200 1-1;:==±:::::::::;-r--r~,:1. -"'""j~=:J:-:::~
---4= 10 ~-- --... ......-
distribution along the beam is shown in Fig. 14. The distri-
bution obtained with the force-based element indicates a large E 150 +---+-1 .••
100 -- A = 20
A= it===J:2.2J1~J:~~f::~;f:::~
IS 1- .,
discontinuity at mid span for high levels of a. However, the E --A=30:/,
60 -l------t-'====t==
overall distribution of bond force is very close to the one ob-
tained by displacement-based elements. The above numerical
~
-
~ 0 f----.
V/
.------I-----r---+----1I---+---I
{l It
-60 - t - - - t - - - t - - t - - r - . ' - t - - - + - - - - 1 I - - - + - - - I
4.5 ~_ ·100 -I---t--+---=~/-~).I./~:t.----r~_~:-.....Jl=
-, I ~ .P,4
4.0 ---_:-t"=-::~;;'1;(t'-' 1--1:-.:;II,---,.....,....,.,..;:;IJ
·160 t"-------t+-__
b....
E
)(
3.5 r----d=10
3.0 -
--d=30
-d=10
I
Numerical ~
Numerical : :
Exact Cj I,
b~
II

lli
~P.4 II t= ~ -... ".-- . . . . . . .
.200

-250
~ -+-......."_'"'lj_~~I-' _

+--+---+---+---+--4--+-~I----t
2"al ••..uu. If S'ft.l..
_..&.IjI--'++-.,eme--n_+_"'fnc:....8'-'--'>...
+.n-'--__1

E 2 E amenta lHalf SPl n o

i 2.5 - --d=30 Exact I 250 500 760 1000 1250 1600 1760 2000
I I x (mm)
$.
l! 20
.
~ 1.5
:
I
I
I - --, ..

:s 1.0
_:.... ....t (b)
(J v . . . / , ........
0.5 '" ,- " 1"- FIG. 14. Bond Force Distribution under Concentrated Load
~ ....4 - -
--..=.."" """- with: (a) Displacement-Based Elements; (b) Force-Based Ele-
0.0 ments
o 250 500 750 1000 1260 1500 1760 2000
x (mm) studies indicate that the force-based composite beam element
(a) is far more accurate than the displacement-based element for
the analysis of nonlinear composite beams.
4.5 Another interesting result of this study is the difference in
4.0
the behavior of a composite beam under different load cases.
I I I For the case of pure bending, the maximum slip occurs at the
T
i3.5 -
E 3.0
-d=10I---

--d=30
-d=IO
Numerical
Numerical
Exact
r-
t1 it
r
~P.4
Ii
t end sections, which do not benefit from the composite action.
Because the total bending moment is constant along the ele-
E ' 2 Elamentl Half Sp n
ment, the end section is therefore the critical section and the
-._- --d=30
Exact
!2.5
I; moment capacity converges to that of a beam with no bond.
i 2.0 .---""t- On the other hand, for the composite beam under concentrated
J i load, the maximum moment occurs at midspan, where no slip
i1.6
~
61.0 : -\. ~ can occur and the full composite action is retained. Therefore,
the response of the composite beam to the concentrated load
7.- -~ is close to that of a beam with perfect bond.
0.5
0.0
..Iea:'" --- I
... -"...... ....._-
SUMMARY AND CONCLUSIONS
o ~ ~ m 1~ 1~ 1~ 1m ~
x (mm) The basic equations that govern the behavior of a composite
beam with deformable shear connectors under small displace-
(b) ments are derived in this paper. The equations are used to
FIG. 13. Curvature DistributIon under Concentrated Load develop a new force-based composite beam element. The el-
with: (a) Displacement-Based Elements; (b) Force-Based Ele- ement consists of two beam components connected by a con-
ments tinuous interface, which accounts for the deformations of the
JOURNAL OF STRUCTURAL ENGINEERING I OCTOBER 1998/1157

J. Struct. Eng., 1998, 124(10): 1148-1158

 shear connectors. The bond force along the interface is ap- composite substructure with full and partial shear connection." Proc.•
proximated by a cubic interpolation function. I Ith World Conf on Earthquake Engrg., Pergamon, Disk 2. Paper No.
Two numerical examples on the nonlinear behavior of com- 77I, Elsevier Science, Oxford, England.
Daniel, B. J., and Crisinel, M. (1993). "Composite slab behavior and
posite beams are solved using both force-based and displace- strength analysis. Part I: Calculation procedure." J. Struct. Engrg.,
ment-based composite beam elements. The results clearly ASCE, 119(1), 16-35.
demonstrate the superiority of the force-based element over a EI-Tawil, S., and Deierlein, G. G. (l996a). "Fiber analysis of composite
displacement-based element in the nonlinear analysis of com- beam-column cross sections." Struct. Engrg. Rep. No. 96-06, School
posite beams with deformable shear connectors. of Civ. and Envir. Engrg., Cornell University, Ithaca, N.Y.
The main conclusions drawn from this study are as follows: EI-Tawil, S., and Deierlein, G. G. (1996b). "Inelastic analysis of mixed
steel-concrete frames." SSRC-ASCE Struct. Congr. XIV, S. K. Ghosh
and J. Mohammadi, eds., ASCE, New York.
1. At low load levels, both the displacement- and force- Hajjar, J. F., Schiller, P. H., and Molodan, A (1997). "A distributed plas-
Downloaded from ascelibrary.org by Nanyang Technological University- Library on 10/11/17. Copyright ASCE. For personal use only; all rights reserved.

based elements can predict the response of composite ticity model for concrete-filled steel tube beam-columns with interlayer
beams with enough accuracy. However, at high load lev- slip. Part I: Slip formulation and monotonic analysis." Struct. Engrg.
els, where the response is nonlinear, the force-based el- Rep. ST-97-I, Dept. of Civ. Engrg., University of Minnesota, Minne-
apolis, Minn.
ement is far more accurate than the displacement-based Hirst, M. J. S., and Yeo, M. F. (1980). "The analysis of composite beams
element considered here. using standard finite element programs." Compo Struct., 11(3), 233-
2. The force-based element provides a much better curva- 237.
ture approximation than the displacement-based element. Neuenhofer, A., and Filippou, F. C. (1997). "Evaluation of nonlinear
This is especially true for high load levels, where the frame finite-element models." J. Struct. Engrg., ASCE, 123(7), 958-
response is nonlinear. A discontinuity can be observed in 966.
Neuenhofer, A., and Filippou, F. C. (1998). "Geometrically nonlinear
the curvature variation between two adjacent displace- flexibility-based frame finite element." J. Struct. Engrg., ASCE,
ment-based elements. The force-based element, however, 124(6), 704-711.
retains a continuous curvature variation. Razaqpur, A. G., and Nofal, M. (1989). "A finite element for modeling
3. The force-based composite beam element results in a the nonlinear behavior of shear connectors in composite structures."
bond force discontinuity between adjacent elements. The Compo Struct., 32(1), 169-174.
discontinuity is large when the bond force distribution is Salari, M. R., Spacone, E., Shing, P. B., and Frangopol, D. M. (1997).
"Behavior of composite structures under cyclic loading." Build. to
steep. This discontinuity, however, has no significant Last, Proc., ASCE Struct. Congr. Xv. Vol. I, L. Kempner Jr. and C. B.
effect on the overall performance of the force-based Brown, eds., ASCE, New York, 731-735.
element. No discontinuity exists in the axial force and Spacone, E., Filippou, F. C., and Taucer, F. F. (1996). "Fiber beam-col-
bending moment distributions between the adjacent umn model for nonlinear analysis of RIC frames. I: Formulation."
force-based elements. Earthquake Engrg. and Struct. Dyn., 25(7), 711-725.
4. Two horizontal degrees of freedom are used at midspan Wegmuller, A W., and Amer, H. N. (1977). "Nonlinear response of com-
posite steel-concrete bridges." Compo Struct., 7(2), 161-169.
of the displacement-based element to improve its axial
behavior. These degrees of freedom have no direct effect APPENDIX II. NOTATION
on the flexural behavior of the element. For the force-
based element, the bond force distribution can be im- The following symbols are used in this paper:
proved by increasing the number of bond reference a displacement shape functions;
points. This results in a higher order approximation of b, C force shape functions;
bending moment and axial forces in the composite beam D section forces;
components. Db bond force;
d section deformations;
Additional numerical studies are needed to verify the perfor- db bond deformation;
mance of the force-based element under cyclic loading with dbr = bond residual deformation;
experimental results. Furthermore, the bond constitutive law dr = section residual deformations;
can be improved to model bond deterioration under cyclic Eb bond stiffness;
loading. Ee , Es = ~tiffnesses of concrete and steel sections, respectively;
F = element flexibility matrix;
ACKNOWLEDGMENTS f = section flexibility matrix;
h bond flexibility;
This work is supported by National Science Foundation Grant CMS- k section stiffness matrix;
9520282. This support is gratefully acknowledged. = bending moments in concrete and steel sections, re-
spectively;
APPENDIX I. REFERENCES = axial forces in concrete and steel sections, respec-
tively;
Amadio, C., and Fragiacomo, M. (1993). "A finite element model for the py = vertical distributed load;
study of creep and shrinkage effects in composite beams with deform-
able shear connections." Costruzioni Metalliche, Vol. 4, Associazione Q nodal forces;
Costruttori Acciaio Italiani, Milan, Italy, 213-228. Qb bond forces at reference points;
Aribert, J. M., and Ragneau, E. (1997). "Theoretical investigation of q nodal displacements;
moment redistribution in composite continuous beams of different qr = nodal residual displacements;
classes." Compos. Constr. in Steel and Concrete III, Proc.. Engrg. uc , relative displacement at interface;
Found. Conf, C. D. Buckner and B. Shahrooz, eds., ASCE, New York, uoc , uos axial displacements in reference axes of concrete and
392-405. steel sections, respectively;
Arizumi, Y., Hamada, S., and Kajita, T. (1981). "Elastic-plastic analysis
of composite beams with incomplete interaction by finite element
Vc, V, = shear forces in concrete and steel sections, respec-
tively;
method." Compo Struct., 14(5-6),453-462.
Ayoub, A, and Filippou, F. C. (1997). "A model for composite steel- v = beam transverse deflection;
concrete girders under cyclic loading." Build. to Last. Proc.• ASCE y = vector of element displacements;
Struct. Congr. Xv. Vol. I, L. Kempner Jr. and C. B. Brown, eds., ASCE, coe' EoJ' axial strains in reference axes of concrete and steel
New York, 721-725. sections, respectively; and
Bursi, O. S., and Ballerini, M. (1996). "Behavior of a steel-concrete <l> curvature.

1158/ JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1998

J. Struct. Eng., 1998, 124(10): 1148-1158