You are on page 1of 14

Engineering Structures, Vol. 20, No. 8, pp.

663-676, 1998

ELSEVIER

PII: S0141-0296(97)00107-7

1998 Elsevier Science Ltd


All rights reserved. Printed in Great Britain
0141-0296/98 $19.00+ 0.00

A distributed plasticity model for


concrete-filled steel tube beamcolumns with interlayer slip
Jerome F. Hajjar
Department of Civil Engineering, 500 Pillsbury Drive SE, University of Minnesota,
Minneapolis, MN 55455-0220, USA

Paul H. Schiller
Barr Engineering Company, 8300 Norman Center Drive, Minneapolis, MN 55437-1026, USA

Aleksandr Molodan
Department of Civil Engineering, University of Minnesota, Minneapolis, MN 55455-0220,
USA
(Received February 1997; revised version accepted April 1997)

A fiber-based distributed plasticity finite element formulation is


presented to perform three-dimensional monotonic analysis of
square or rectangular concrete-filled steel tube (CFT) beam-columns. This stiffness-based beam-column element fomulation
accounts comprehensively for all significant geometric nonlinearity
exhibited by CFT beam-columns as part of composite frame structures, and the steel and concrete constitutive models account for
the significant inelastic phenomena which are seen in CFT experiments. In addition, the finite element formulation accounts for slip
between the steel and concrete components of the CFT by incorporation of a nonlinear slip interface. This formulation is able to
capture behavior ranging from perfect bond to immediate slip. The
calibration and verification of the slip formulation are presented,
and the finite element model is verified against experiments of CFT
beam-columns subjected to monotonic loading. Results from a preliminary study are presented on the effect of slip on CFT beamcolumn and composite frame behavior. A related paper extends
this formulation to cyclic analysis of composite CFT frames, and
provides details of the steel and concrete constitutive models.
1998 Elsevier Science Ltd.

Keywords: concrete-filled steel tube, distributed plasticity finite


element, bond, slip

CFTs to date, which has focused mainly on computing the


ultimate axial and flexural capacity of individual members t,
the current formulation provides an efficient method to
study both the significant stress-strain behavior of a CFT
member and the load-deformation behavior of CFTs as
part of complete composite frames.
A stiffness-based beam-column finite element forms the
basis of the CFT model. This formulation utilizes the fiber
element approach for modeling distributed plasticity. The
model discretizes the CFT cross-sections at the beam ends
into a grid of fibers (as seen in Figure 1), and the stress-

I. Introduction
One- and two-way low- to moderate-rise unbraced composite structures, consisting of steel I-girders framing rigidly into concrete-filled steel tube (CFT) beam-columns,
exploit the many advantages that a CFT has to offer. This
paper presents an overview of a three-dimensional, geometrically and materially nonlinear distributed plasticity
finite element model for square or rectangular CFT beamcolumns used in these types of composite frames. In contrast to much of the computational research conducted on

663

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

664

Superscript c: concrete
Superscript s: steel

Concrete core
. J slips relative to
steel tube

...o

....

p*
..-.,--..~ .....,.....~

\pc
\

End J
Figure I

Distributed

plasticity

Origin, O , is in
plane of~steel
cross section

concrete-filled

tube

finite

element model

strain behavior is tracked explicitly at each fiber. Numerical


integration is used through the cross-section to yield a stiffness formulation based upon stress-resultants, such that the
element degress-of-freedom (DOFs) are all located at the
centroidal axis of the member at each of its ends. The use
of these types of finite elements has been well documented
for decades by many researchers and has been shown to
provide an accurate representation of the behavior of beamcolumn elements in structural frames. A wide variety of
fiber (or, in two-dimensions, 'layered') elements have been
used to model steel wide-flange beam-columns 2-7,
reinforced concrete beam-columns s t2, and steel reinforced
concrete beam-columns j3-~5, to name only a few. Very few
models of this type have been developed for CFT beamcolumns as part of frames. Bode 16, Bridge ~7, Tomii and
Sakino zS, Shakir-Khalil and Zeghiche ~9, Kawaguchi et al. 2,
and Tsuji et al. 2~ have presented similar formulations, predominantly for monotonic analysis of individual CFTs in
two dimensions. Fiber element formulations require less
computing time than three-dimensional continuum finite
elements, yet they permit direct modeling of the variation
of material properties across the element cross-section, and
they may account explicitly for such effects as residual
stresses or initial plastic strains.
The uniaxial steel and concrete constitutive models in
this work are based upon comprehensive multi-axial plasticity formulations which account for the significant inelastic phenomena exhibited in CFT experiments and which
greatly affect the behavior of CFTs. In addition, this CFT
finite element formulation accounts for slip between the
steel and concrete components of the CFT by incorporation
of a nonlinear slip interface. This interface allows axial
movement of the concrete core with respect to the steel
tube. It is able to capture behavior ranging from perfect
bond to immediate slip, and it permits modeling of the
gradual transfer of stress between the steel tube and concrete core in the connection region of CFT beam-columns.
The current formulation extends existing composite beam
computational models which include slip to elements which
incorporate both geometric and material nonlinearity. Calibration and verification of the slip formulation is presented,
as is verification of the complete beam-column model vs
monotonic beam-column experiments for CFTs having a
wide range of material strengths and cross-section dimensions to verify the accuracy and robustness of the formulation. A comparable steel wide-flange fiber element is also
implemented, based on the model of White 5, and a final
example of a multistory composite CFT frame is presented,
including an assessment of the effect of slip on the overall

frame behavior. Hajjar et al. 22 extend this model to cyclic


analysis of composite CFT frames, provide details of the
constitutive models and their calibration to material tests
and CFT experiments which yield moment-curvature
results, and compare the results to cyclic experiments of
CFT beam-columns and subassemblages of steel wideflange girders framing into CFTs. This fiber analysis formulation is geared for conducting monotonic or cyclic static
behavioral studies of composite CFT subassemblages or
complete composite CFT structures. It is also suited for
conducting comprehensive parametric studies which, in
conjunction with experiments, may provide the data
required to improve the accuracy and scope of current nonseismic and seismic design specification provisions for
CFTs.
1.1. Scope of the concrete-filled steel tube fiber model
The work presented here is part of an ongoing research
program to develop analysis formulations capable of modeling the geometrically and materially nonlinear behavior of
three-dimensional CFT beam-columns and their connections as part of composite frame systems. The fiber element
formulation complements a concentrated plasticity CFT
finite element model developed by Gourley and Hajjar,
which is suitable for conducting static, transient dynamic,
and eigenvalue buckling analysis of composite CFT
frames23 26. While less compact than the concentrated plasticity element, this distributed plasticity formulation provides more detailed information which is critical for
assessing the effect of slip on CFT member and frame
behavior, conducting parametric studies of CFT subassemblages to generate a comprehensive suite of axial-flexural
interaction diagrams, and assessing member and frame ductility.
The current research is limited to frame structures consisting of steel I-beams and/or square or rectangular CFTs.
The steel constitutive model calibration has been conducted
for ASTM A500 Grade B steel for CFTs, and ASTM A36
or A572 Grade 50 steel for wide-flange members. The concrete constitutive model calibration has been conducted for
concrete strengths (f') up to 50 MPa (see References 22
and 27 for calibration and verification of the constitutive
models). The CFTs are assumed to be completely filled
with concrete and to have no reinforcement or shear connectors. Confinement of the concrete core of the CFT by
the steel tube is accounted for through calibration of the
concrete constitutive model. Longitudinal residual stresses
produced in cold-formed, welded steel tubes vary through
the thickness 28 and are accounted for indirectly in the steel
constitutive model used in the fiber analysis through calibration to stress-strain curves obtained from tests of hollow
structural section tension coupon specimens, which retain
these residual stresses 22'29. Effects which are not modeled
directly in the current formulation include local buckling,
nonlinearity due to shear or torsion (since shear and torsional forces are expected to be small in CFT frame
members), shear deformations due to flexure (i.e. EulerBernoulli beam theory is assumed), time dependent effects
on the materials, and post-collapse behavior.

2. CFT element formulation with interlayer slip


The occurrence of slip between the steel and concrete
components of rectangular CFTs has been noted by a number of researchers. Much of the past work has specifically

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.
studied whether a strong bond between the concrete and
steel is required for adequate CFT behavior, and whether
shear connectors, ribs in the tube, or some other explicit
means of achieving bond are required 3. The experiments
which have been conducted to determine slip and bond
characteristics in rectangular and circular CFTs have
included push-out tests 3] 35, tests of CFTs in pure flexure 36,
and tests steel I-girders framing into CFTs using simple
shear-tab connections 37-39. In addition, several building
codes for composite construction recommend using mechanical shear connectors if ~:he calculated interface stress in
a CFT column exceeds a limiting value 4'4~. However, the
effect of slip on the overall performance of CFTs remains
a subject of research, and i! is probable that for frame structures, the effect of loss of bond is relatively small,
especially with respect to the global behavior of the frame.
To permit comprehensive modeling of this phenomenon,
axial slip is permitted between the steel and concrete
materials in this fiber model, and behavior ranging from
perfect bond to immediate slip may be simulated.
The current formulation is intended to model the behaviour of CFT beam-columns in which the two materials are
allowed to slip with respect to each other along the member's centroidal axis. Similar formulations have been
applied to composite beams composed of a steel girder and
concrete deck connected by a line of shear connectors 42 44
For CFTs, the centroidal a~es of the steel tube and the concrete core coincide in the initial state. Slip is, therefore, not
induced due to flexure until the section becomes unsymmetric due to plastification or cracking of the materials. Slip
may also be induced whea shear force is transferred from
girders which frame into CFT columns at fully-restrained
and, especially, simple connections.
Amadio and Fragiacomo 43 presented a two-dimensional
layered composite beam element to model the slip between
an elastic steel beam and a concrete slab modeled as a viscoelastic material to capture creep and shrinkage. This formulation provides the basis for the current work, which
extends Amadio's model to CFTs which include threedimensional behavior, geometric nonlinearity, material
nonlinearity of both the slJtp interface and of the component
steel and concrete materials, and, in a related paper, cyclic
excitation =. Schiller and Hajjar4~ provide details of the
geometrically nonlinear stiffness formulation, the force
recovery procedure, the incremental/iterative NewtonRaphson nonlinear solution strategy, and results of CFT
analyses used to verify the accuracy of the geometrically
nonlinear element formulation. Features of this formulation
which relate directly to the inclusion of slip in the model
are outlined below.

2.1. Virtual work equation of equilibrium with interlayer


slip
For this work, the kinematic relationships for a CFT beamcolumn with interlayer slip are based on the assumption
that the steel and concrete are separated by a layer of
springs which determine the load transfer between the two
materials based on a nonlinear spring stiffness. Thus, to
track the differential movement between these materials for
a three-dimensional geometrically nonlinear CFT arbitrarily
oriented in space, and to allow for automated assembly of
CFT elements into the global stiffness matrix of a composite frame during geometrically nonlinear analysis, three
additional translational DOFs are added to each end of the
conventional 12 DOF fiber beam element (i.e. nine DOFs

665

are modeled at each joint - - see Figure 1). These additional


DOFs allow the steel and concrete to have independent
axial deformations at the element level, while allowing for
elements with different orientations of their longitudinal
axes to frame into a joint (these differing orientations may
be due either to the original topology of the frame, or to
geometric nonlinearity of CFT elements originally aligned
in a column stack). As will be described later, penalty functions are used in the global stiffness matrix to impose shear
constraints, such that the transverse displacements of the
steel and concrete at the ends of each element are constrained to be the same.
To establish this element formulation, the virtual work
equation of equilibrium 46 is modified to separate the steel
and concrete contributions to the internal strain energy, and
to include the strain energy associated with the deformation
of the layer of springs:

2v~

f
2.UO c 2dV~.+ /
2~eo
:2v'

2,,,g~j

2d}'~

"{"f 21 27r 28to Zdl = f 2s~ 2~ ~uc 2dSc


P
+|
J

2sS

(1)

zt~ 8u~ 2dgs

Left superscripts denote the configuration in which a quantity is measured, and left subscripts denote a reference state.
In this work, the current (unknown) state is referred to as
configuration C2, the most recent converged state is
referred to as configuration C 1, and the initial undeformed
state is referred to as configuration CO. In equation ( 1), "r is
the Cauchy stress tensor, 6e is the engineering strain tensor
corresponding to the virtual displacements, 8u is the virtual
displacement imposed in configuration C2, t is the surface
traction vector (body forces are neglected in this work), 7r
is the force per unit area transferred between the steel and
concrete through the slip interface, tO is the deformation of
the slip interface 43, V is the volume of the element, S is the
surface of the element, and I is the steel-concrete interface.
A right superscript c denotes concrete and a right
superscript s denotes steel.
An updated Lagrangian (UL) incremental formulation,
coupled with a corotational coordinate system, is used in
this work to account for all significant geometric nonlinear
behavior and path-dependent material behavior. This formulation follows the work of White s, Morales 47, and Yang
and Kuo48. Using a corotational approach, the element stiffness matrix is formed in a corotational coordinate system
which 'corotates' with the rigid body motion of the
element. In this system, only deformations which cause
element straining are considered. This corotational stiffness
matrix is then transformed to incorporate the rigid body
modes, resulting in the final 18 DOF local element tangent
stiffness matrix 45.

2.2. Slip kinematics and load-slip relation


The slip formulation is illustrated in Figure 1. Since the
concrete is encased within the steel, the two materials are
assumed to have the same transverse displacements and
rotations; slip stress is only caused by differential axial
strain between the steel and concrete layers. This differential axial strain can be computed directly by considering
the interpolation functions for axial displacement of the two

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

666

materials. Since the geometric centers of the two materials


coincide, the same local axes can be used for both
materials. The incremental deformation in the spring layer
is, thus, given by 43
A~O= Au~, - Au~

(2)

where Auc represents the incremental axial deformation


between the element ends along the centroidal axis. Equation (3) then relates the incremental force transferred
between the materials per area of interface, ATr, and the
incremental deformation of the interface, A~O:
2~_~_ I,jT.]_ A~.j. - I,jT .t_ ]~ml//

(3)

where ~: is the tangent stiffness of the interface (units of


force/length3). The incremental linearization of the nonlinear load-slip relationship allows the slip interface stiffness,
~:, to be updated at the beginning of each global load
increment or iteration.

2.3. Fiber element approach for modeling spread of


plasticity
To approximate the volume integrals which appear in the
virtual work equation of equilibrium using the fiber element
method, each integral of equation (1) is first decomposed
into area and length integrals. The area integrals are then
approximated by numerical integration over the fibers
which comprise each cross-section. By evaluating the area
integrals during incremental analysis using the current tangent modulus of each fiber in the cross-section, spread of
plasticity through the cross-section is accounted for. The
tangent modulus of each individual fiber is computed based
on the appropriate material constitutive model and the fiber
strain, which in turn is computed from the local element
displacements 45.
For the integral of the virtual work term incorporating
the increment in slip deformation, this decomposition may
be represented by:

/~.A~O.3,q/dl = f,L

La,/,.a, q,.p'dx

(4)

where p is the perimeter of the steel tube, L is the element


length, and dx is a differential element along the length of
the CFT.
In this work, only the torque and the rigidity term associated with torque are not calculated through numerical
integration, since this element formulation does not account
directly for shear strains. The torsional rigidity at a crosssection is assumed to remain constant over the entire load
history. This approximation is believed to be sufficient for
modeling the behavior of rectangular CFTs as members of
structural frames, as these members exhibit high torsional
rigidity 49.

2.4. Finite element interpolation functions


Once numerical integration is used to compute the element
forces, moments, and rigidities at the end nodes, interpolation functions are used to approximate the change in
these quantities along the element length. In conventional
beam finite element formulations, the axial force is assumed
to be constant in the element. This assumption remains true
in the current formulation for the total force at any point

along the element length, but the steel and concrete axial
forces, taken separately, may vary linearly along the length
of the member as they transfer force through the slip interface. The torque is assumed to remain constant along the
length, and a linear interpolation along the length is used
for the bending moments and higher-order force resultant
terms 4~. Linear interpolation functions are also chosen to
represent the change in axial, flexural and torsional rigidities along the element length.
Conventional cubic Hermitian shape functions are used
to describe the transverse deformations and rotations which
are caused by flexure, and a standard linear shape function
is used to describe the torsional rotation of the element.
A quadratic shape function is used to describe the axial
deformation, which requires the addition of midpoint axial
DOF for both the steel and concrete. The current formulation assumes that external loads are applied only at the
element ends. Thus, the midpoint DOFs are removed from
the element stiffness matrix through static condensation.
The formation of this local element tangent stiffness
matrix permits calculation of incremental deformations
based upon an increment in applied load. After the
incremental global deformations are computed, forces and
moments are recovered at the element ends. Strains at each
fiber are first computed, followed by assessment of the
stress at each fiber based on the constitutive formulation.
Stress-resultants are then computed using numerical integration. Schiller and Hajjar 45 present the details of the force
recovery procedure adopted for the current formulation,
including a discussion of the technique for including the
high-order geometrically nonlinear terms of the virual work
equation of equilibrium which are nonlinear in the displacements, recovery of element centroidal axis strain and curvature from the element end forces, and the technique used
for materially nonlinear force recovery. The terms retained
in the tangent stiffness matrix and force recovery procedure
account sufficiently for all significant P - A and P-6 effects
within the CFT beam-column, and the formulation is accurate for problems involving moderate rigid body rotations
and small incremental strains.

2.5. Constraint of concrete and steel transverse


displacements
If 18 DOF CFT beam-column elements are used as members in a structural frame model, shear deformation compatibility constraints are required between the steel and
concrete translational DOFs at each joint. For most practical structures, the CFT beam-columns will be aligned in
vertical column stacks, requiring two global compatibility
conditions (perpendicular to the axis of the column stack)
at each joint. This may not always be the case, however,
as other structural systems which may have arbitrary orientations of members entering each joint are possible. For
this reason, a general formulation for shear compatibility
constraints is used in this work.
Penalty functions imposed in the global coordinate system are used to enforce compatibility conditions on the
CFT steel and concrete translational DOFs 46. In this
approach, a large, but finite, stiffness is assigned between
corresponding steel and concrete transverse displacements,
thus constraining their values to be nearly identical. These
constraint equations are determined at the beginning of the
analysis, and the constraints are oriented in the directions
of the local element transverse displacements, perpendicular to the centroidal axis of each CFT at each joint.

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.
When conducting geometrically nonlinear analysis using
constraint equations, the directions of shear constraint are
assumed to rotate with the structural node at which they
act to reflect the changing geometry of the structure. This
approach assumes that all CFT members are free of internal
hinges or member releases, an assumption which is consistent with the scope of the current formulation.
In addition to the required shear constraints, there are
practical analysis situations when the steel and concrete
should not be allowed to slip in the axial direction at a
structural joint. These situations arise when mechanical
shear connectors are provided at the joint, when there is a
change in CFT cross-seclion size at a joint, when CFT
elements frame into a joint from different directions, or
when a beam-to-column connection includes diaphragms or
through-bolts such that they enforce axial strain compatibility of the steel and concrete at a joint. In order to prevent
slip at a joint in the computational model, an additional
penalty function constraint equation may be specified in the
direction of the element centroidal axis.

3. Constitutive modeling
The constitutive models for the steel and concrete simulate
the significant aspects of inelastic behavior found in CFT
members. These formulations and their calibration for the
current model are discussed in detail in References 22 and
27, and are summarized only briefly here. Key effects
which influence monotonic behavior of CFTs and which
are modeled in this work include a gradually changing
modulus of both steel and concrete with increased monotonic loading; retention of the elastic modulus upon
unloading for the steel; modulus degradation of the concrete upon load reversal, tensile behavior of the concrete;
and multiaxial stress effects such as confinement of the concrete core. In addition, the constitutive formulations have
been adopted for cyclic loading of CFTs, in which
additional key behavioral phenomena include strength
degradation; stiffness deterioration; a vanishing zone of linear behavior with cyclic loading; and the Baushinger

effect22.27,
The steel formulation, adapted from Shen et al. ~, models
the rounded shape of the stress-strain curve found in coldformed tube steel, a decreasing elastic zone with increased
plastic straining, and the different stress-strain behavior
exhibited in the corners and flanges of cold-worked steel
tubes. The concrete formulation, adapted from Chen and
Buyukozturk 5~, models strength and stiffness degradation
by means of a cumulative., damage parameter, and the postpeak behavior of the c,~ncrete is calibrated to account
implicitly for the effects of confinement of the concrete
core by the steel tube. The concrete formulation also works
well for fibers which cycle into tension and then back into
compression. Both of the material constitutive models are
formulated in multiaxial stress space, allowing for future
extension of the current tormulation to account directly for
confinement. In the present work, only the uniaxial components of these models are currently activated.

4. Calibration of the load-slip relation


The focus of experimental research related to slip in CFTs
has generally been to improve the recommendations which
account for bond in various building c o d e s 3 j - 3 4 ' 39 A
majority of the published bond experiments present the

667

results of push-out studies, and focus exclusively on bond


strength, rather than the full load-slip behavior. However,
tests performed by Shakir-Khali133'34 have provided insight
into the entire load-slip behavior of rectangular CFT sections. Graphs of the load-slip relationship for push-out tests
of CFTs both with and without shear connectors suggest a
simple bilinear relation. The spgcimens tend to lose their
initial slip stiffness suddenly, at a point which corresponds
to the bond strength of the interface. A very low stiffness
is observed after the interface bond breaks down; a value
near zero is assumed in this work. Consequently, only bond
strength and initial slip stiffness must be calibrated. The
calibration focuses on CFTs having no internal shear connectors or ribs. However, through recalibration or extension
of the slip constitutive relation, this formulation is equally
applicable to CFTs having explicit means of retaining bond.
Connection studies representing steel I-girders framing
into CFTs with simple shear tabs or tee-stubs have been
conducted by Dunberry et a l . 37 and Shakir-Khali139 to determine the effect of slip on CFT columns in which only a
portion of the column axial force enters at the connection,
with the remaining axial force introduced at the column
end. These studies provide detailed data about the load
transfer and slip around CFT connections, and are believed
to provide a more accurate representation of the load transfer at these connections than traditional push-out studies.
These tests often exhibit bond strength and initial slip stiffness which are higher than those observed in push-out tests
on similar CFT sections, due primarily to the added frictional resistance from pinching which occurs as the tube
starts to deform or buckle in the connection region. Consequently, rather than use the results of push-out studies for
calibration, the results of connection studies are utilized.
There are only a limited number of these types of connection studies documented thoroughly in the literature for
use in the calibration of the initial value of ~: from equation
(4). Thus, in this work, its value varies as an inverse function of the perimeter of the tube, such that ~: = k~v/p, where
k~,o is a calibrated constant value of slip stiffness per unit
length of interface, having units of force/length 2. Use of
this relation effectively makes the initial value of the term
k.p from equation (4) equal to a constant. The value of k~,o
is thus calibrated directly.

4.1. Calibration of initial slip stiffness


For calibration of initial slip stiffness, kslip, data is desired
in the materially elastic range in order to eliminate the
effects of the material model calibration on the slip calibration. Such data is available from Shakir-Khali139 from
strain measurements taken during connection testing. In this
publication, the strain profile along a CFT specimen is
presented for the entire loading, providing detailed data
about the straining effects near a CFT connection which
may result from slip. The test specimens were constructed
from square structural tubing with tee-cleat connections.
Loading and test parameters for specimen E6 are shown in
Figure 2. In the computational model, a node is assigned
at each of the strain gage positions shown in Figure 2. The
load applied at the connection is split between the three
nodes in the connection region (gages 5, 6 and 7) in the
ratio of 2 5 - 5 0 - 2 5 % to simulate the transfer of load along
the connection plate.
Figure 3 shows the measured strains in the steel at a level
of applied load, P, equal to 600 kN, vs the height along the
specimen for the computational model and the experiment.

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

668
P

Tube: 150x150x5

e = 1511

~" "'
--

"

'

~p

D= 150m~
k~

|775
[
150

t : 5.0mm
f~. = 386.9 MPa
f, = 465.0 MPa

150

E~= 3 .5GP

180

~ - y 0------, too.
i

1,

~ 5 ; 5 ~ d i s t a n c e between gages in mm
"~'gage number
925

Figure 2 Experimental setup of Shakir-KhaliP~, specimen E6


i

i Experiment (Shakir-Khalil,
1994, Specimen E6)
!--Analysis: kstip= 106 MPa
-~.-Analysis: k,,e = 104 MPa
Analysis: k,,e = 5x103 MPa
I.... Analysis: katie = 103 MPa .

4
o

]
o

i
i
i
f

0
i~
tz:'/

connection l
re~ion
t

L
i

i.
t
i
t:
i
i

-0.0004

-0.0002

conducted by Schiller and Hajjar45 in which the concrete


core is pushed through the steel tube shows that values of
k,r~p above approximately 104 MPa provide nearly perfect
bond.
Similar correlation is achieved at a load, P, of 800 kN,
and from a second specimen, E8, which has the same
dimensions and strain gage positions as specimen E6, but
a slightly higher concrete compressive strength. These comparisons thus show evidence that the value of k,,p from
the push-out studies (approximately 103 MPa) consistently
overpredicts the length of the load transfer region around
the connections. The experimental results from this set of
test, as well as the observations of other researchers 37, suggests that there is little or no slip outside the load transfer
region. This condition is only predicted accurately by the
highest value of k~,p. The value of initial slip stiffness of
kslip of lO 4 MPa consistently provides the best balance of
mean error and standard deviation in the results, and this
value of initial slip stiffness is thus chosen for the current
computational model.

-0.0006

s
Figure3 Measured vs computed strain in the steel tube at a
load of 600 kN

The results for several different values of initial slip stiffness are shown. The introduction of the load to the steel
tube in the connection region is clearly seen, along with
localized strain concentration below the connection.
An analysis of the results indicates that the computational model matches the experimental strain measurements fairly well at this load level, especially outside the
connection region 45. The model does not predict the strains
within the connection region as accurately, due predominantly to the assumed pattern of load distribution in the
connection region (i.e. 25-50-25%), but the shape of the
strain distribution exhibits a similar rate of load transfer as
the experimental results. The strain measurements of gages
1, 2 and 3 above the connection show that the majority of
load is transferred to the concrete in the connection region
(since the strain differs little between these gages), a trend
which is only matched well by the highest of the k~p
values. All values of k~,v produce approximately the same
average percent error, between 6 and 7%, but the highest
value of slip stiffness produces the lowest standard deviation, indicating a better match to the shape of the strain
distribution. The value for ks,p of 103 MPa (which is representative of the results from push-out studies) produces
a strain diagram that over-predicts the length of the loadtransfer region. Note that computational parametric studies

4.2. Calibration of bond strength


Push-out studies may not be representative of the bond
strength at simple beam-to-column connections due to the
pinching action caused by connection rotation. Thus, in
order to study bond strength, connection specimens from
Dunberry et al. 37 are utilized. A much higher percentage of
the total load is applied at the connections for this series
of tests compared to Shakir-Khali139. This creates more slip
and bond loss and provides a better indication of the bond
strength at these connections. Dunberry et al. 37 present
graphs of the experimentally measured slip, strain, and total
load in the steel and concrete at a load level near the limit
point of the column. Note that material nonlinearity in the
steel and concrete contribute significantly to the straining
of these specimens near their ultimate loads.
Dunberry et al. 37 present extensive data for specimen D1.
This test used the same nominal tube size and had approximately the same material properties as the Shakir-KhaliP 9
specimens E6 and E8, which were used for calibration of
initial slip stiffness. Figure 4 shows the experimental setup
and test parameters. A detailed description of the strain
gage locations is not provided by the authors, but the speci-

!
t

_ _

t,

Tube: 152.4x 152.4x4.83


P = 1622 kN
D = 152.4 nun
t = 4.83 mm
590mm f~ = 443MPa
f~ = 588 MPa
180mm fc'= 29.6 MPa
180ram Ec= 19.3GPa

1050mm

T,
Figure 4 Experiment test setup of Dunberry et a/37, specimen
D1

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar


mens were instrumented to measure steel strain with strain
gages, and concrete strain and slip with demic gages.
The computational model uses 11 elements, with the
loading applied to the top, middle and bottom of the connection region in a ratio of 2 0 - 2 0 - 6 0 % to simulate the
gradual transfer of load in this region. This loading scheme
is different than the one used in the previous section, since
it better reflects the load transfer seen in this specific experiment (Schiller and Hajjar 45 present the loading scheme used
for the initial slip stiffness studies to demonstrate the effect
that the loading pattern has on the accuracy of the computational results for this type of connection). The initial elastic modulus of the concrete was not reported, and a value
of 19.3 GPa was selected for the analysis so as to match
the experimentally observed concrete rigidity45.
Figure 5 shows the comparison of the computational
model to the published experimental results for various
values of bond strength, J'bo,d" Slip between the concrete
core and the steel tube is shown at points along the length
of the steel tube.
The results show that the highest values offbo,a provide
the most accurate comparison to the experimental slip profile. However, the value of fbo,a of 0.6 MPa provides the
best comparison with the maximum slip, with an error of
2.5% 45. The slip profile in Figure 5 presents evidence that
the computational model predicts the shape and magnitude
of the slip along the columa with better accuracy below the
connection than above it. This phenomenon can again be
attributed to connection rotation and the pinching mechanism which results.
The distribution of load between the steel and concrete
portions of specimen D1 along the length of the column is
presented by Dunberry el al. 37 for a load level slightly
below the ultmate load of the c o l u m n . t Figure 6 shows the
177Y'A

'k',..

[:1

"

o=on

.::.-..~m;'~:~'.'5.2-'-- . . . .
~,~"~
~

Experiment (Dunberry et al., 1987,


Specimen D1)
~ - Analysis, fbo,~= 0.4 MPa
.... Analysis,fbo,,~= 0.5 MPa
...... Analysis,fbo,~= 0.6 MPa
..... Analysis, fbo,~= 0.7 MPa
.... Analysis, fbo,a= 0.8 MPa

,
*
i

0.05

0.1
Slip (mm)

I
a

!:

669

e t al.

Concrete,experimental (Dunberryet al., 1987,


Steel,experimental SpecimenD1)
- - Analysis,f0o.a= 0.4 MPa
"-- Analysis,fbo.a= 0.6 MPa
....Analysis,fbo.d---'-1.0 MPa
~..

connection~ - ~ load
on
i' [ transfer
~
region

~"Steel

Concrete""

Load(kN)

Figure 6 Measured vs computed axial force in the steel tube


and concrete core
comparison between the published results and the prediction of the computational model for various values o f bond
strength. A statistical analysis of the results suggests that
a bond strength of 1.0 MPa produces the best results when
compared to the experimental data (e.g. 6.86% error for the
concrete load) 45. However, the other values presented show
similar percent error (e.g. 7.29% error for fbon~ = 0.6 MPa
for the concrete load). A bond strength of 0.6 MPa gives
acceptable results and matches the maximum slip much better than the value of 1.0 MPa. Both of the higher values of
bond strength in the computational model match the overall
shape of the load diagram more accurately than the lowest
value (note that a value of bond strength of 0.4 MPa corresponds to that specified in Reference 40, while a value of
approximately 0.1 MPa corresponds to that specified in
Reference 41 ). This trend is especially apparent in the concrete load in the region above the connection. The experimental data points suggest that all the load transfer from
the steel to the concrete occurs in the region indicated on
the graph. Only the highest values of bond strength result
in a load transfer within this region, with the lowest bond
strength value showing transfer over a much larger region.
Based on the examples provided in this section, a bond
strength of 0.6 MPa provides the best correlation between
experimental and computational results. This value provides the best prediction of total slip, loss of bond around
the connection region, and transfer of load from the steel
tube to the concrete core after loss of bond occurs. The
calibration has also shown that the computational model is
more accurate for the region at the compression end of a
connection than the tension end because of localized
behavior in the experiments caused by connection rotation.

0.15

Figure 5

Measured vs computed slip between the steel tube


and concrete core

The total load in this specimen, and the proportion applied at the connection, may be calculated by adding the forces in the steel and concrete at
various points along the length. The published experimental graph indicates that the total applied loads are 680 kN at the top of the specimen
and 880 kN at the connection. This data contradicts the tabulated values
of Dunberry et al? 7, who state that 50% of the total load is applied at the
connection, and that the total load graphed is 1622 kN. In order to compare
to the published results, however, the applied loads in the analysis are

4.3. Verification o f slip formulation


As verification of these calibrated slip parameters, two
further experiments are investigated, holding all calibrated
parameters fixed. The series of tests conducted by Dunberry
et al. 37 included a specimen, C1, which was loaded only at
the connections, and had no cap or other means of pre-

assumed to be those which are computed by summing the steel and concrete loads in the published experimental graph.

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar

670
uncapped

et al.

Tube: 101.8x101.8x4.82
P = 608 kN
D = 101.8nun
t = 4.82mm
fy = 374 MPa

220 mm
I j

180 mm
1'80 mm

]
l
0.8
i

L=

24.2 MPa
Ec= 17.9 GPa

m m

0.6

M/Mo

920 mm

ental

0.4

u and Kennedy, 1994,


Specimen CB33)

- - Analysis
0.2
Figure 7 Experimental parameters of Dunberry et aL37, specimen C1
0

venting slip at the top of the specimen. Figure 7 gives the


experimental parameters for this test. A plot of the experimentally observed load in the concrete over the height of
the column is presented for a given load of P = 608 kN.
Figure 8 shows the comparison of the experimental data
points to the computational model using the calibrated parameters. The shape and magnitude of the load distribution
is well predicted by the computational model, with the load
gradually entering the concrete core starting above the connection and continuing below it.
Lu and Kennedy 36 present the results of flexural testing
of a rectangular CFT, bent about its major axis, in which
the experimentally observed slip was reported for specimen
CB33. The CFT consisted of a 254 152 6.4 rectangular
steel tube, with a yield stress of 377 MPa, and a concrete
strength of 45.2 MPa. The specimen was simply supported,
3040 mm long, and had equal concentrated loads applied
766 mm from each end. The beam had no end plates or
other means of preventing slip along the specimen.
Figure 9 shows the comparison between the experimentally
observed slip and the slip predicted by the computational
model using the calibrated slip parameters (M,, is the ultimate moment strength of the CFT). This graph provides
evidence that the calibration based on connection tests is
accurate for other types of loading situations as well. The

"

50

Experiment (Dunberry et al., 1987,


Specimen C1)

100 150
Load (kN)

200

Figure8 Measured vs computed concrete axial force in


uncapped column specimen

0.05

0.1

0.15

Slip (mm)
Figure9

Measured vs computed slip in uncapped flexural

specimen

slip and the point where loss of bond occurs are predicted
well by the computational model.

5. Verification of the concrete-filled steel tube


fiber model
Verification of the CFT fiber element is performed by comparing the finite element results to over thirty different
experimental studies of CFT beams, eccentrically loaded
beam-columns, nonproportionally loaded beam-columns,
and composite frames composed of steel 1-girders rigidly
framing into CFT columns, holding all calibration parameters fixed45. Experimental verification studies were
selected to provide data for CFTs having a wide range of
parameters, such as material strengths, D/t ratio, L/D ratio,
and the method of applied loading j. Figures lOa-h show
the comparison between the fiber model and experimental
monotonic load-deflection curves for several of the verification studies (details of each test are provided in Table 1
and Reference 45). Each test is referenced by the nomenclature of the experimentalist. Figures lOa and b are beams
loaded in pure flexure by point loads, including both major
and minor axis bending. Figures lOc-f are eccentrically
loaded beam-columns~ including comparisons to tests from
three different experimentalists. These four examples
include both stocky and slender beam-columns subjected
to uniaxial or biaxial bending, for a range of material
strengths. Figures 10g and h n8 are annealed specimens
which are loaded nonproportionally, with varying levels of
constant axial force being applied in the different tests (in
the table, Po is the ultimate axial strength of the member),
followed by application of bending moment.
The percent error in load was computed for each of the
verification problems at the deformation level corresponding to the experimental maximum load, or at the level corresponding to the end of the analysis, if this occurs first,
and then again at half of that deformation level. The peak
strengths for all monotonic comparisons were underpredicted with an average error of 1.8%, with a standard deviation of 7.0%. At half of the deformation value at peak
strength, the average error in the strengths for all monotonic

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

25O

l,Oi

f""

2OO

Moment /

/.

Moment 150
(kN-m) 100

671

(kN-m)
50

Experimental

50

Experimental

- Computational

0.00000

- Computational
L

0.00010
0.00020
Curvature (rad/mm)
a) Lu and Kennedy (1994), Specimen CB33

0.00000
0.00010
0.00020
Curvature (rad/mm)
b) Lu and Kennedy (1994), Specimen CB53

lOOO ~
800

Load 1500

Load 600

(kN) 1000 "I~ t ~


500 ~ - ~
V
O f
0

.Experimental

(~,4) 4oo

- Computational
I

5
10
Deflection (ram)
c) Bridge 1'1976), Specimen SHC-1
2500 ~axial Bendi~
2000

~ ~ " - ~ ~

15

I
I
I
10
20
30
Deflection (mm)
d) Cederwall et al (1991), Specimen 10

Load
/"

500

/ /

- Computational
I

5
10
15
Deflection (mm)
e) Bridge (1976), Specimen SHC-5
15
....

20

"

200

150
(kN)100

Experimental

250

'

Load 1500
(kN) 1000

- Computational

200

50
0

- Computational

I
I
I
10
20
30
Deflection (mm)
f) Shakir-Khalil and Zeghiche (1989), Spec. 6

12

E~---~t~

10
10
Moment

(~'q-m) /
5

__

Moment
(l~/-m) 6
4

" ~ ~ ~ -t Computation
a l /
"al

2
0

t
t
I
0.005
0.01
0.015
Rotation (rad)
g) Tomii and Sakino (1979), Specimen 111-3
0

I
t
J
I
0.005 0 . 0 1 0 . 0 1 5 0.02
Rotation (rad)
h) Tomii and Sakino (1979), Specimen III-6
0

Figure 10 Comparison of experimental and computational monotonic Ioadin9 results

672

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

Table 1 Concrete-filled steel t u b e m o n o t o n i c a l l y loaded verification e x p e r i m e n t s


Test (specimen)

Actual tube
dimensions (mm)

D/t m a j o r
(minor)

L/D m a j o r
(minor)

f" (MPa)

fy (MPa)

Other data

Lu and Kennedy 36
(CB33)

253.4 152.0 x
6.17

39.7 (23.75)

12 (20)

45.2

377

Lu and Kennedy 36
(CB53)

253.4 152.0
6.17

39.7 (23.75)

8.8 (14.7)

42.1

377

Bridge 17 (SHC-1)

203.7 x 203.9 x
9.96
120 x 120 x 8

20

10.5

29.9

291

M a j o r axis bending;
P = 0 ; load applied
766 mm f r o m each end
of 3040 mm beam
M i n o r axis bending;
P= 0; load applied
463 mm f r o m each end
of 2231 mm beam
e = 38 mm 1

15

25

39

397

e = 20 mm

202.6 x 203.2 x
10.0
120 x 80 x 4.47

20

15

44.3

319

24 (16)

23 (34.5)

45

343.3

100 x 100 x 2.99

33

20.6

289

100 x 100 x 2.99

33

20.6

289

e = 38 mm
c~ = 30 2
er~.jor = 24 m m
eminor = 16 mm
P/Po = 0.30
A n n e a l e d tube
P/Po = 0.60
Anneal ed tube

Cederwall et al. 52
(10)
Bridge 17 (SHC-5)
Shakir-Khalil and
Zegiche 19 (6)
Tomii and
Sakino TM (111-3)
Tomii and
Sakino TM (111-6)

1Eccentricity o f applied axial load f r o m centroidal axis of m e m b e r


2Angle of applied axial load with respect to m a j o r principal axis o f cross-section (i.e. a = 0 induces m i n o r axis bendi ng)

comparisons was 0.15%, with a standard deviation of


6.7% 45.
For the verification studies, 10 fibers were used in each
flange of the steel tube (with one fiber through the tube
thickness), and the concrete core was meshed with a 10 by
10 grid of fibers. Four elements were used along the length
of each member.

6. Effect of slip on concrete-filled tube beamcolumn behavior


6.1. Effect of bond and slip on flexural behavior
The purpose of including slip in the formulation of the CFT
element is primarily to determine whether or not loss of
bond occurs in CFT members used as part of structural
frames. It has already been seen in the verification and calibration of the slip parameters that slip may affect the region
near a connection consisting, for example, of a steel girder
and/or brace element framing into a CFT with a shear tab,
as may be found in braced frame structures. This section,
as well as the final example of a four-story frame, provide
the results of a preliminary study to determine the effect
of slip on behavior of CFT beam-columns, as may be found
in unbraced frames having fully-restrained connections.
Clearly if sufficient axial force is applied to only the steel
or concrete of a CFT, slip may be induced. The effect of
slip on flexural behavior is more difficult to quantify than
it is for axial behavior, because material nonlinearity is
required before any interface stress develops in flexure.
While complexities such as contact between the concrete
core and the steel tube are not modeled directly in this formulation, the calibrated slip model provides strong correlation with experimental results of CFTs subjected to flexure. Several comparisons to results from Lu and Kennedy 36
are presented in Schiller and Hajjar 45. The number of
elements along the length and fiber mesh densities at each
cross-section were similar to those used for the CFT verification studies. For each test, the parameters of the computational slip model were varied to study the effect of bond

strength on the moment-curvature behavior of the members. It was found that reducing the initial slip layer stiffness by two orders of magnitude reduced the initial flexural
stiffness of the members by less than 1%, and it did not
noticeably change the ultimate moment strength of these
members at the point corresponding to the maximum curvature obtained in each experiment, due largely to the fact
that the bond stress of 0.6 MPa was not breached in these
studies. Similarly, reducing the bond strength to one-third
of the calibrated value (0.2 MPa) decreased the maximum
strength attained in the specimens by an average of only
2%. In general, the magnitude of strength, stiffness and
deformation of the members changed very little for the full
duration of loading in all of these parametric studies, and
the computational curves all match the experimental results
equally well.

6.2. Composite frame example


The CFT fiber element formulation was used to analyze a
four-story composite CFT frame structure subjected to a
combination of gravity and wind loading. Figure 11 illustrates the members, geometry, structural parameters and
unfactored loading. The frame is proportioned so as to achieve all factored load combinations (specifically, a load factor of 1.0/0.9 = 1.1 1 times the factored loads to account for
the AISC LRFD 53 resistance factor in the analysis) without
collapsing due to combined geometric and material nonlinear behavior. In the analysis for the AISC LRFD 53 wind
load combination, which controls the design, the gravity
and lateral loads are increased proportionally, so that at a
load factor of 1.0, the applied loads equal the factored loads
for the load combination. The frame shown actually reaches
its structural limit point at a load factor of approximately
1.4, thus yielding a safe and serviceable design. The computational model utilizes the 18 DOF CFT beam-column
fiber element for the columns, and a comparable 12 DOF
steel beam-column fiber element for the girders 5,47. Twelve
elements are used to model each girder and column, and
the fiber mesh densities at each cross-section are similar to
those used for the CFT verification studies.

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.

673

V12
,

29,2 kN ~
27,3

v/21v~v

v/2jv
25.0 kN ~ '

/-/7

~v g~v ~ v ~ v

/
4 Stories
~vv/2 1@3.96m

~v~

~ v ~ _ ~ =15'84m

~v~v

4 Bays @ 6.10 m = 24.40 m

All loads shown are


unfactored. Gravity
load is divided equally
between dead and live.

Girders:
W410x60; fy= 248.2 MPa
CFT Sections:
~TS 177.8x177.8x6.35
@TS 152.4x152.4x4.76
~TS 254.0x203.2x6.35
~)TS 203.2x 152.4x6.35
TS 203.2x203.2x6.35

/~.nalysis Cases:
Case 1: No Slip Permitted
Case II: Slip Restrained at Bottom of
Column Stacks;
Calibrated Parameters

~TS 152.4xl 52.4x6.35

Case IV: Case II, fOond = 0.1 MPa

"

l.I
i

V = 186.8, kN

Scale in mm
i

C a s e II

Case HI: Case II, kslip = 102 MPa

- -

C a s e IH

....... C a s e I V

a) Slip in CFTs at a factored load level of 1.0

fy= 317.2 MPa, fc" = 34.5 MPa

Figure 11 Four-story frame computation model


,

The AISC LRFD interaction equation values for the individual members were also computed at a load factor of 1.0.
Even at this load level, almost all of the lower story columns have exceeded the allowable LRFD interaction equation value. The distributed plasticity analysis not only indicates that the frame has a higher ultimate capacity, but also
that the members are largely elastic at a load factor of 1.0.
This result illustrates the generally conservative approach
that design codes have adopted currently for CFT members.
Four cases were investigated to determine the effect of
slip parameters on the behavior of the frame. As seen in
Figure 1 I, in Case I the steel girder was assumed to engage
the concrete directly through the connection, and no slip
was permitted at the joints. In the other cases, the steel
girder was assumed to engage only the tube, load was transferred to the concrete through the slip interface, and the
concrete was free to protrude from the top of the column
stack. Case II used the calibrated parameters for fhond and
k~,p. In Case III, the value of k~,v was decreased to 10 2
MPa. In Case IV, the value of fbo.d was decreased to
0.1 MPa. Each analysis was carried out to the limit point
of the structure, There was no perceivable difference
between the entire load-deformation response of the analyses, thus indicating that slip has a negligible effect on the
overall load-deformation behavior of this frame.
However, the potential effects of slip on local behavior
of CFTs are more evident. Figure 12 shows the slip and
bond stress along the length of each CFT column stack for
Cases II, III, and IV. The slip plotted as positive to the
right of each column stack indicates that the concrete core
is moving vertically relative to the steel tube (i.e. it tends
to protrude through the top of the column stack). Figure 13
shows the axial force in tae concrete core and in the steel
tube along the length of each CFT column stack for the
same cases. All results are shown at a load factor of 1.0
for the wind load combination. The bond stress is not breached in Case II at this load level, although bond stresses
exceed 0.6 MPa at the top of each column stack at a load
level just over 1.1. The concentration of slip and bond
stress at each connection is evident (Figure 12), as is the
gradual transfer of load from the steel tube to the concrete
core both above and below each connection. The effects of
tension just above many of the connections are evident as

E>

>

>

C a s e II

Scale in M P a
', ', I ~ I ~
0
0.3
0.6

~CaseIII
.......

Case IV

Figure 12 Slip and bond stress in the composite frame at the


factored design load

well, with the compressive axial force in the steel tube


decreasing rapidly over a short distance as a portion of the
gravity load at the floor level is carried by the tube above
the connection as well as below (Figure 13b). However,
the magnitude of the slip is relatively small in this case, as
may be expected for unbraced framing action, and it does
not change significantly even up to the failure load.
For Case III, the slip is an order of magnitude larger than
for Case II because of the reduced slip stiffness
(Figure 12a). Correspondingly, the transfer of load from
the connection regions occurs over nearly the entire length
of each CFT column. As compared to that of Case II, the
concrete load for Case III is seen in Figure 13 to increase
consistently along the member length, while the steel tube
continuously sheds load to the concrete along the tube
length. In addition, because of this more gradual transfer
of load, the maximum interface stress values in the CFTs
of Case III are seen to be well below those of Case II

(Figure 12b).
For Case IV, the bond stress of 0.1 MPa is breached at
a load factor of 0.19. Since the slip stiffness is small at
locations which have breached fbo,d, the bond stress is
effectively capped at this value, as is seen by the dotted
line in Figure 12b. The magnitude of slip is more than double that of Case II at a factored level of 1.0, and the transfer
of loads from the connections takes place over nearly the

A distributed plasticity model for CFT beam-columns with slip: d. F. Hajjar et al.

674

7. Conclusions

ttl
Case II
P
0

Scale in kN
I
)
I
1500
Compression

a) A x i a l f o r c e in C F T concrete

cores

- -

Case I n

.......

Case IV

at a f a c t o r e d l o a d l e v e l o f 1.0

I
f

1
P
0

Scale in kin
I
I
I
1500

Case II
- -

Case HI

.......

Case IV

Compression
b) A x i a l f o r c e in C F T steel t u b e s at a f a c t o r e d l o a d l e v e l o f 1.0

Figure 13 Axial force in the concrete core and steel tube in the
composite frame at the factored design load

entire length of the CFT, as in Case Ill (Figure 13). The


bond stress and slip profile at the failure load of the structure (not shown) indicate substantially more slip and breaching of the bond strength occurring along nearly the
entire length of the three central CFT column stacks, indicating that continued loading substantially increases the
bond loss in these members.
The frame was also analyzed for the AISC LRFD 53 gravity load combination using the calibrated parameters to
determine if loss of bond is detected under the increased
vertical loading. For this loading combination, the interface
stress of 0.6 MPa was breached at 0.68 times the factored
load, suggesting that the increased vertical load that must
be transferred at connections due to factored gravity loading may cause bond loss in CFT members before the design
load level of the frame is reached.
For this specific frame, the top of each column stack
consistently has the highest magnitudes of slip and bond
stress for all loading combinations. Capping the CFTs at
the top of the column stacks alleviates the slippage at that
location, but localized slip in the connection regions at each
floor level is still seen when using the calibrated parameters. Global behavior, however, is unaffected by slip in
these analyses for all load combinations. Of course, engaging the concrete directly in each connection, and thus
restraining slip, facilitates immediate transfer of the load to
the connection region.

This paper summarizes the development and use of a fiberbased distributed plasticity finite element formulation for
three-dimensional concrete-filled steel tube beam-columns.
The formulation may be used to simulate the behavior of
composite frames consisting of steel I-girders framing into
CFTs, subjected to monotonic static loading; a companion
paper 2~ outlines the formulation for cyclic loading. The
fiber element approach discretizes the CFT element end
cross-sections into a grid of fibers, and the steel and concrete stress-strain behavior is tracked explicitly at each
fiber. In addition, this formulation permits axial slip
between the concrete core and steel shell of the CFT, so
as to permit study of the effect of slip on CFT beam-column
behavior as part of braced or unbraced frame structures.
Details of the calibration and verification of the slip parameters are presented, based on comparison to tests of steel
I-girders framing into CFTs with shear connections. The
calibrated parameters suggest that little slip is experienced
in a CFT member before the bond strength of the slip interface is breached. In addition, the calibrated value of bond
strength used for analysis is higher than the value recommended by design codes, suggesting that the recommended design values may be conservative. The beamcolumn element is verified against experiments of individual CFTs having a wide range of cross-section properties,
material properties and lengths, subjected to combined axial
force and uniaxial or biaxial bending moment, and loaded
either proportionally or nonproportionally.
A preliminary investigation is included which indicates
that bond strength may be breached in unbraced frames,
possibly before the design load level is reached, if slip is
not restrained at the joints (e.g. if the concrete core is not
engaged directly at the connection), due either to gravity
load or wind load combinations. However, in these studies,
even for the more extreme conditions, slip is seen to have
little effect on the global behavior of a composite CFT
frame, or on the strength achieved by a CFT member subjected to flexure. Nevertheless, understanding the effect of
slip more fully on the behavior of CFTs in composite frame
structures warrants further comprehensive parametric studies; experimental and analytical research in this area is
ongoing 54. Documenting the effect of varying CFT perimeters or member lengths on the slip calibration requires
additional experimental tests of steel I-beams framing into
CFTs for comparison, and establishing the effect of slip
on a wider variety of frame configurations and boundary
conditions is needed.
This stress-resultant based distributed plasticity formulation provides detailed information on CFT behavior, yet
remains efficient and suitable for studying complete composite CFT frames. The formulation is especially amenable
to conducting static, nonlinear seismic 'push-over' analysis
of composite CFT frames, or for conducting advanced
analysis directly for static design of frames z6. Cyclic
behavioral studies may also be undertaken; transient
dynamic analysis is pending for this CFT formulation.

Acknowledgements
The authors would like to thank Professor H. Shakir-Khalil,
University of Manchester, Professor T. Usami, Nagoya
University, and Professor O. Buyukozturk, Massachusetts
Institute of Technology, for their generous sharing of infer-

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.
mation relevant to this research, and Katherine A. Fetterer
for assisting with the analysis of the four-story frame.
Funding for this research was provided by the National
Science Foundation (Grant no. CMS-9410473) under Dr
Shih-Chi Liu and Dr M. P. Singh, and by the University
of Minnesota Department of Civil Engineering through a
Sommerfeld Fellowship for the third author and through
additional research funding. The authors gratefully
acknowledge this support.

21

22
23

References
1 Gourley, B. C., Hajjar, J. F. and Schiller, P. H. 'A synopsis of studies
of the monotonic and cyclic behavior of concrete-filled steel tube
beam-columns', Report no. ST-93-5.2, Department of Civil Engineering, University of MinneapoLis, Minneapolis, MN, 1995
2 Wright, E. W. and Gaylord, E. H. 'Analysis of unbraced multistory
steel rigid frames', J. Struct. Div., ASCE 1968, 94(ST5), 1143-1163
3 Alvarez, R. J. and Birnstiel, C. qnelastic analysis of multistory multibay frames', J. Struct. Div., ASCE 1969, 95(STll), 2477-2503
4 Kanchanalai, T. 'The design and behavior of beam-columns in
unbraced
steel
frames',
AISI
Proj.
no.
189,
Civil
Engineering/Structural Research Laboratory (CESRL) Rep. no. 772, Department of Civil Engineering, Structural Research Laboratory,
University of Texas, Austin, TX, 1977
5 White, D. W. 'Material and geometric nonlinear analysis of local
planar behavior in steel frames using interactive computer graphics',
M.S. thesis, Cornell University, Ithaca, NY, 1985
6 Izzuddin, B. A. and Elnashai, A. S. 'Adaptive space frame analysis.
Part II: a distributed plastici:y approach', Proc. Inst. of Civil Engng
1993, 99, 317-326
7 Challa, V. R. M. and Hall, J. F. 'Earthquake collapse analysis of steel
frames', Earthq. Engng Struct. Dyn. 1994, 23( I 1), 1199-1218
8 Krishnamoorthy, C. S. 'Finite element models for inelastic analysis
of reinforced concrete frames and application to limit state design',
Computational plasticity: models, software and applications, Owen,
D. R. J., Hinton, E. and Onate, E. (eds), Pineridge Press, Swansea,
1987
9 Taucer, F., Spacone, E. ancl Filippou, F. C. 'A fiber beam-column
element for seismic response analysis of reinforced concrete structures', Report no. UCB/EERC-91/17, Earthquake Engineering
Research Center, University of California, Berkeley, CA, 1991
10 Chang, G. A. and Mander, J. B. 'Seismic energy based fatigue damage analysis of bridge columns: Part I - - evaluation of seismic
capacity', Rep. no. NCEER-94-0006, National Center for Earthquake
Engineering Research, State University of New York, Buffalo, NY,
1994
11 lzzuddin, B. A., Karayanni,;, C. G. and Elnashai, A. S. 'Advanced
nonlinear formulation tor reinforced concrete beam-columns', J.
Struct. Engng, ASCE 1994, 120(10), 2913-2934
12 Ahmad, S. H. and Weerakoon, S. L. 'Model for behavior of slender
reinforced concrete column:~ under biaxial bending', ACI Struct. J.
1995, 92(2), 188-198
13 Roik, K. and Bergmann, R. 'Composite columns', Const. Steel Des.:
An lnt Guide, Dowling, P. J, Harding, J. E. and Bjorhovde, R. (eds),
Elsevier Applied Science, London, 1992, pp. 443-469
14 Elnashai, A. S. and Elghazouli, A. Y. 'Performance of composite
steel concrete members under earthquake loading: 1. Analytical model', Eorthq. Engng Struct. Dyn. 1993, 22(4), 315-344
15 EI-Tawil, S. and Deierlein, G. G. 'Inelastic dynamic analysis of
mixed steel-concrete space frames', Structural Engineering Rep. no.
96-5, School of Civil and F,nvironmental Engineering, Coruell University, Ithaca, NY, 1996
16 Bode, H. 'Columns of steel tubular sections filled with concrete
design and applications', Acier Stahl 1976, 11(12), 388-393
17 Bridge, R. Q. 'Concrete filled steel tubular columns', Rep. R283,
School of Civil and Mining Engineering, University of Sydney, Sydney, Australia, 1976
18 Tomii, M. and Sakino, K. 'Experimental studies on the ultimate
moment of concrete-filled square tubular beam-columns', Trans.,
Arch. Inst. Japan 1979, 27S, 55-63
19 Shakir-Khalil, H. and Zeghiche, Z. 'Experimental behavior of concrete-filled rolled rectangular hollow section columns', The Struct.
Engineer 1989, 67(19), 345-353
20 Kawagucbi, J., Morino, S., Atsumi, H. and Yamamoto, S. 'Strength
deterioration behavior of concrete-filled steel tubular beam-columns',
Proc. Third Int. Conf on S,'eel-Concrete Comp. Struct., Wakabaya-

24

25

26

27

28

29
30

31

32

33
34
35

36

675

shi, M. (ed.), Fukuoka, Japan, 1991, Association of International


Cooperation and Research in Steel-Concrete Composite Structures,
1991, pp. 119-124
Tsuji, B., Nakashima, M. and Morita, S. 'Axial compression behavior
of concrete filled circular steel tubes', Proc. Third Int. Conf. SteelConcrete Composite Struct., Wakabayashi, M. (ed.), Association
International Cooperation and Research in Steel-Concrete Composite
Structures, 1991, pp. 19-24
Hajjar, J. F., Molodan, A. and Schiller, P. H. 'A distributed plasticity
model for cyclic analysis of concrete-filled steel tube beam-columns
and composite frames', Engng Struct. 20(4-6), 398412
Gourley, B. C. and Hajjar, J. F. 'Cyclic nonlinear analysis of threedimensional concrete-filled steel tube beam-columns and composite
frames', Report no. ST-94-3, Department of Civil Engineering, University of Minneapolis, Minneapolis, MN, 1994
Hajjar, J. F. and Gourley, B. C. 'Representation of concrete-filled
steel tube cross-section strength', J. Struct. Engng, ASCE 1996,
122( I 1), 1327-1336
Hajjar, J. F., Gourley, B. C. and Olson, M. C. (for Part II) 'A cyclic
nonlinear model for concrete-filled tubes. I: formulation. II: verification', J. Struct. Engng, ASCE 1997, 123(6), 736-754
Hajjar, J. F., Gourley, B. C., Schiller, P. H., Molodan, A. and
Stillwell, K. A. "Seismic analysis of concrete-filled steel tube beamcolumns and three-dimensional composite frames', Comp. Const. HI,
Buckner, C. D. and Shahrooz, B. M. (eds), Engineering Foundation,
ASCE, NY 1997, pp. 75-88
Molodan, A. and Hajjar, J. F. 'A cyclic distributed plasticity model
for three-dimensional concrete-filled steel tube beam-columns and
composite frames', Rep. no. ST-96-6, Department of Civil Engineering, University of Minneapolis, Minneapolis, MN, 1997
Sherman, D. R. 'Tubular members', Const. Steel Des.: An Int. Guide,
Dowling, P. J., Harding, J. E. and Bjorhovde, R. (eds), Elsevier
Applied Science, London, 1992, pp. 91-104
Sully, R. M. and Hancock, G. J. 'Behavior of cold-formed SHS beamcolumns', J. Struct. Engng, ASCE 1996, 122(3), 326-336
Goel, S. C. and Yamanouchi, H. (eds) Proc. 1992 US-Japan Workshop on Composite and Hybrid Structures, Berkeley, CA, 1992,
Research Report, Department of Civil and Environmental Engineering, University of Michigan, Ann Arbor, MI, 1993
Virdi, K. S. and Dowling, P. K. 'Bond strength in concrete-filled
steel tubes', IABSE Periodica 3/1980, IABSE Proc. P-33/80, 1980,
pp. 125-139
Tomii, M. 'Bond check for concrete filled steel tubular columns',
Comp. and Mixed Constr., Proc. of the US-Japan Joint Sere. on
Comp. and Mixed Constr., Roeder, C. W. (ed.), University of Washington, Seattle, WA, ASCE, NY, 1985, pp. 195-204
Shakir-Khalil, H. 'Pushout strength of concrete-filled steel hollow
sections', The Struct. Engineer 1993, 71(13), 230-233
Shakir-Khalil, H. 'Resistance of concrete-filled steel tubes to pushout
forces', The Struct. Engineer 1993, 71(13), 234-243
Shakir-Khalil, H. and Hassan, N. K. A. 'Push-out resistance of concrete-filled tubes', Tubular structures VI, Grundy, P., Holgate, A. and
Wong, W. (eds), Melbourne, Australia, A. A. Balkema, Rotterdam,
The Netherlands, 1994, pp. 285-291
Lu, Y. Q. and Kennedy, D. J. L. 'The flexural behaviour of concretefilled hollow structural sections', Can. J. Civil Engng 1994, 21(1),
I 11-130

37

38

39

40

41
42
43

Dunberry, E., LeBlanc, D. and Redwood, R. G. 'Cross-section


strength of concrete-filled HSS columns at simple beam connections',
Can. J. Civil Engng 1987, 14, 408-417
Shakir-Khalil, H. 'Connection of steel beams to concrete-filled tubes',
Proc. 5th Int. Symp. Tubular Struct., Nottingham, UK, 1993,
pp. 195-203
Shakir-Khalil, H. 'Beam connection to concrete-filled tubes', Tubular
structures VI, Grundy, P., Holgate, A. and Wong, W. (eds), Melbourne, Australia, A. A. Balkema, Rotterdam, The Netherlands, 1994,
pp. 357-364
British Standards Institute (BSI) BS 5400: Steel, concrete and composite bridges. Part 5: Code of practice for design of composite
bridges, London, British Standards Institution, 1979
Architectural Institute of Japan (AIJ) AIJ Standard for structural calculation of mixed tubular steel-concrete composite structures, AIJ,
Tokyo, Japan, 1980
Bradford, M. A. and Gilbert, R. I. 'Composite beams with partial
interaction under sustained loads', J. Struct. Engng, ASCE 1992,
118(7), 1871-1883
Amadio, C. and Fragiacomo, M, 'A finite element model for the study
of creep and shrinkage effects in composite beams with deformable
shear connections', Construzioni Metalliche, 1993, 4, 213-228

676
44

45

46
47

48
49

A distributed plasticity model for CFT beam-columns with slip: J. F. Hajjar et al.
Aribert, J.-M. 'Influence of slip on composite joint behaviour',
Conns. in Stl Structs. I11: Beh., Strength and Des. R. Bjorhovde, A.
Colson and R. Zandonini (eds), Pergamon, Amsterdam, 1995,
pp. 11-22
Schiller, P. H. and Hajjar, J. F. 'A distributed plasticity formulation
for three-dimensional rectangular concrete-filled steel tube beam columns and composite frames', Report no. ST-96-5, Department of
Civil Engineering, University of Minneapolis, Minneapolis, MN,
1996
Bathe, K.-J. Finite element procedures in en~,ineering analysis, Prentice-Hall, Englewood Cliffs, NJ, 1982
Morales, L. E. 'Object-oriented software for advanced analysis of
steel frames', M.S. thesis, Department of Civil Engineering, Purdue
University, West Lafayette, IN, 1994
Yang, Y. B. and Kuo, S. R. Theory and analysis of nonlinear framed
structures, Prentice-Hall, Englewood Cliffs, NJ, 1994
Kitada, T. 'Ductility and ultimate strength of concrete-filled steel
members', Stability and ductility of steel structures under cyclic loading, Fukumoto, Y. and Lee, G. C, (eds), CRC Press, Boca Raton,
FL, 1992, pp. 139-148

50

51
52

53

54

Shen, C., Mamaghani, I. H. P., Mizuno, E. and Usami, T. 'Cyclic


behavior of structural steels. II. Theory', J. Engng Mech., ASCE 1995,
121(11), 1165-1172
Chert, E. S. and Buyukozturk, O. "Constitutive model for concrete in
cyclic compression', J. Engng Mech., ASCE 1985, I l l , 797-813
Cederwall, K., Engstrom, B. and Grauers, M. 'High strength concrete
used in composite columns', 2nd h~t. Syrup. on Utilization t~f HighStrength Conc., American Concrete Institute, 1991, pp. 195-214
American Institute of Steel Construction, Load and resistance.tactor
design specification for structural steel buildings, Second edn, AISC,
Chicago, IL, 1993
Roeder, C. W. 'Design requirements for shear connectors in encased
steel (SRC) and concrete filled tube (CFT) construction'. Presentation made at the Third JTCC, U.S./Japan Cooperative Research
Program on Composite and Hybrid Structures, Hong Kong, 1996,
Department of Civil Engineering, University of Washington, Seattle,
WA, 1996

You might also like