Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Computers and Structures 84 (2006) 2373–2384

www.elsevier.com/locate/compstruc

General method of analysis for composite beams with longitudinal


and transverse partial interaction
a,* a,b a
G. Ranzi , F. Gara , P. Ansourian
a
School of Civil Engineering, The University of Sydney, NSW 2006, Australia
b
Department of Architecture, Construction and Structures, Universita’ Politecnica delle Marche, Ancona, Italy

Received 13 April 2006; accepted 24 July 2006


Available online 9 November 2006

Abstract

This paper presents an analytical formulation for the analysis of two-layered composite beams with longitudinal and vertical partial
interaction. The particularity of this model is its ability to incorporate an interface connection deforming both longitudinally, i.e., along
the beam length, and vertically, i.e., transverse to the connection interface, which is modelled by means of a uniformly distributed spring.
The novel formulation is based on the principle of virtual work expressed in terms of the displacement field consisting of the vertical and
axial displacements of the two layers; for completeness, the proposed model is presented in both its weak and strong forms. The partial
interaction problem is then solved by means of the finite element method. A parametric study is presented to investigate the effects of
different combinations of longitudinal and transverse connection rigidities on the overall structural response. For the purpose of these
simulations, a bi-linear constitutive model has been specified for the transverse interface connection to reflect the more realistic case in
which two different responses are observed in the transverse interaction, one in which one layer is bearing against the other one, one when
the two layers are separating. An iterative procedure has been proposed to obtain the convergence to the final solution.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Composite beams; Finite element method; Longitudinal partial interaction; Transverse partial interaction

1. Introduction the interfaces between components. These kinematical dis-


continuities are at the basis of the partial shear interaction,
Composite materials are widely used in engineering here also referred to as longitudinal partial interaction, and
applications due to their ability to exhibit the best proper- of the transverse partial interaction. The former originates
ties of its components. The composite action is provided by from the deformation at the interface connection along the
means of different types of adhesives, locking devices or member length and produces a relative longitudinal dis-
mechanical connectors. In the case of a strong connection placement between the two layers, commonly referred to
between the layers (connection classified as rigid), no dis- in the literature as slip, while the transverse partial inter-
placement discontinuities occur at the interface between action, also referred to as vertical partial interaction, is
components, and available formulations can be adopted produced by the deformation of the connection in the
for their analysis [1]. On the other hand, in the case of a direction normal to the interface, and depicts the relative
flexible connection, the kinematical implications of the displacement of the two layers leading to either separation
deformability becomes significant and an adequate model or bearing of the two layers. This vertical displacement
needs to account for the displacement discontinuities at produces a tensile vertical force per unit length along the
interface connection when the two layers are trying to sep-
arate; this force per unit length becomes compressive when
*
Corresponding author. Tel.: +61 2 9351 5215; fax: +61 2 9351 3343. the two layers are trying to penetrate, thus producing the
E-mail address: G.Ranzi@civil.usyd.edu.au (G. Ranzi). condition of bearing [2].

0045-7949/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2006.07.002
2374 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384

In the field of civil engineering, applications which Still in the 1980s, Robinson and Naraine [30] utilised the
require partial interaction considerations include, among governing system of differential equations presented by
others, steel–concrete composite beams and layered wood Adekola [26] and, revisiting the problem of a simply sup-
systems connected by nails or other mechanical devices. ported beam subjected to a point load applied at the quar-
While this paper is mainly concerned with the behaviour ter point, solved it analytically. Similarly to Adekola [26],
of steel–concrete composite beams, the proposed formula- they adopted the same value for the vertical connection
tion can be adopted for the analysis of generic composite stiffness. They were able to solve the partial interaction
systems which exhibit longitudinal and transverse partial problem, as the bending distribution along the beam was
interaction or to perform parametric studies to evaluate known a priori.
the influence of the partial interaction on the composite Recently, Nguyen et al. [31] utilised a transverse and
behaviour investigated. longitudinal partial interaction model for the analysis of
Historically, the first papers which dealt with the partial reinforced concrete beams retrofitted by steel plates bolted
shear interaction behaviour of composite beams (i.e., longi- to their side faces, where the partial interaction was pro-
tudinal partial interaction) appeared in the 1940s and 1950s, vided by the deformability of the bolted connection. In this
and the most cited of these is the one by Newmark et al. [3]. work, the longitudinal shear forces were assumed to be
This model, usually referred to as Newmark model, assumes resisted by the plastic capacity of the bolted connection
no vertical separation between the two layers and plane sec- while a linear-elastic model was adopted to describe the
tions to remain plane except for a discontinuity at the con- displacements in the transverse direction. The use of this
nection interface, i.e., slip. Since then, several studies have mixed approach (i.e., assuming a plastic behaviour to be
been published to investigate different aspects of the com- exhibited by the bolted connection in the longitudinal
posite partial shear interaction behaviour in the elastic direction and an elastic behaviour in the transverse one)
range [4–9], in the non-linear range [10–19] and accounting simplified the analytical problem and provided interesting
for time effects [20–25]. These studies relied on a variety of insight into the structural problem from a design view-
different formulations, i.e., finite difference, finite element point. As the stiffening steel plates were bolted to the sides
and analytical solutions. An extensive list of references is of the reinforced concrete beams no considerations over
beyond the scope of this paper. the transverse separation and penetration of the two com-
Adekola [26] was the first to extend the model proposed ponents were required.
by Newmark et al. [3] to account also for the transverse This paper proposes a novel formulation for the analysis
partial interaction, which was observed and measured by of generic composite beams with longitudinal and trans-
Chapman and Balakrishnan [27] a few years earlier in verse partial interaction, thereby further extending previ-
experimental tests of composite beams. Adekola [26] ous work to handle indeterminate beams and providing
derived the governing system of differential equations based the description of the partial interaction problem in both
on the equilibrium equations of the partial interaction prob- weak and strong forms. These are proposed in terms of
lem. In his formulation, the unknowns of the problem the internal stress resultants, and then, introducing the
included the tensile (vertical separation) or compressive assumptions of linear-elastic behaviour of the materials,
(vertical bearing) force per unit length and the shear flow are re-expressed in terms of the adopted displacement field.
related to the longitudinal slip at the interface. In this form, In all instances a complete set of boundary conditions is
the method was only suitable for the analysis of determinate provided enabling all possible support and loading condi-
structures, as it required the moment distribution to be tions to be considered. A finite element is then derived to
known a priori. The system of differential equations was model the longitudinal and transverse partial interaction
solved using the finite difference method adopting a solution behaviour. To gain a better insight into the influence of
procedure proposed by Fox [28] for two-point boundary different combinations of the longitudinal and transverse
value problems involving differential equations of orders connection stiffnesses, a parametric study is presented con-
higher than two. To achieve a good accuracy in the numer- sidering a simply supported beam subjected to a point load
ical solutions, the differential equations were re-arranged so applied to the lower layer at mid-span to enhance the effect
that no unknown variables appeared with a derivative of vertical separation. In these applications, a linear-elastic
higher than the second order. Adekola [26] applied this pro- longitudinal interface connection is adopted, while a bi-lin-
cedure to the case of a simply supported beam subjected to a ear model is preferred for the vertical interface connection
point load applied at the quarter point, and he adopted the for which two different stiffnesses are expressed to better
same value for the vertical connection stiffness in both cases represent the realistic response of the structure; one stiff-
of separation and bearing. ness is used when vertical separation occurs between the
Two decades later Aribert and Abdel Aziz [29] utilised two layers, and the second in the case of vertical bearing.
the transfer matrix method to solve the longitudinal and As the extent of beam which undergoes vertical separation
transverse partial interaction problem. This formulation or operates in bearing is not known a priori, an iterative
was utilised to model a discrete distribution of shear con- procedure is utilised to determine the distribution of the
nectors in the non-linear range of only statically determi- two. This problem was not addressed in the numerical
nate composite beams. solutions presented in previous work, as the same vertical
G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384 2375

connection stiffness was adopted throughout the beam and it is further sub-divided into Ac and Ar, concrete com-
length. ponent and reinforcement, respectively, i.e., A1 = Ac [ Ar,
while A2 represents the cross-section of the steel joist only
and it is denoted as As.
2. Analytical model
2.2. Displacement and strain fields
2.1. Basic assumptions
The position of a generic material point P can be
A prismatic composite beam formed of two layers is expressed in the undeformed state of the beam by the vec-
shown in Fig. 1. The beam occupies the cylindrical region tor r as
V = A · [0, L] generated by translating its cross-section A,
with boundary XA, along a rectilinear axis which is orthog- rðx; y; zÞ ¼ xi þ yj þ zk; 8x; y 2 A ¼ A1 [ A2 ; z 2 ½0; L
onal to the cross-section and is assumed, in its undeformed ð1Þ
state, to be parallel to the Z-axis of the ortho-normal refer-
ence system {O; X, Y, Z}. For generality, the formulations in which i, j, k represent the unit vectors parallel to the axes
are derived for a beam segment of length L about an arbi- of the adopted ortho-normal reference system {O; X, Y, Z},
trary coordinate system as shown in Fig. 1. while the rectilinear line K representing the interface con-
The composite cross-section is represented as A = nection is located at x = 0 and y = yc with z 2 [0, L], where
A1 [ A2, where A1 and A2 are the cross-sections of the yc is defined in Fig. 1.
top and bottom elements which, for ease of reference, are The kinematical behaviour of the composite beam is
referred to as elements 1 and 2, respectively, and it is expressed in terms of u1(z), u2(z), v1(z), and v2(z) which rep-
assumed to be symmetric about the plane of bending, YZ resent the axial displacements at the fibres located at y1 and
being the plane of symmetry. y2, respectively, and the vertical deflections of layers 1 and
The two parts are connected along a rectilinear line K 2, respectively (Fig. 2). This displacement field is defined by
which describes the position of a continuously distributed the vector d as
connection between the two layers. This connection pro-
vides the composite action while allowing for relative dis- dT ðx; y; zÞ ¼ ½ u1 ðzÞ u2 ðzÞ v1 ðzÞ v2 ðzÞ  ð2Þ
placements in both longitudinal and transverse directions;
these will be referred to for clarity as longitudinal slip where u1(z), u2(z), v1(z) and v2(z) will be referred to as u1,
and vertical uplift. Positive uplift occurs when separation u2, v1 and v2 for ease of notation. For generality, the level
takes place between the two layers and negative uplift of the reference axes y1 and y2 can be defined arbitrarily.
depicts the case when one layer is bearing against the other, The expressions for the rotations h1 ð¼ v01 Þ, h2 ð¼ v02 Þ
thereby producing a compressive force per unit length and curvatures j1 ð¼ v001 Þ, j2 ð¼ v002 Þ along the beam can
along the vertical connection. then be obtained by differentiating the respective deflection
The two layers are assumed to obey the Euler–Bernoulli v1 and v2 with respect to the coordinate z along the beam
beam theory and therefore plane sections are assumed to length. The prime represents differentiation with respect
remain plane and orthogonal to the beam axis except for to z.
a discontinuity at the connection interface, while warping The admissible displacement of a generic point in the
of the cross-section (shear-lag effects) is not considered. composite beam is defined by the following vectors (Fig. 2):
Being a plane model, no torsional and out-of-plane flexural uðx; y; zÞ ¼ va j þ ½ua  ðy  y a Þv0a k; 8ðx; yÞ 2 Aa ; z 2 ½0; L
effects are considered.
ð3Þ
Without any loss of generality, the composite cross-sec-
tion here represented is that of a typical steel–concrete in which u(x, y, z) represents the displacement of the generic
composite beam in which A1 represents the reinforced slab point P of layer a (with a = 1, 2).

Z O
yc y1 X

O k y2
X
j i

L
Y Y

Fig. 1. Typical composite beam and cross-section.


2376 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384

In Eq. (7), XA represents the contour of the domain


yc y1 Z A = A1 [ A2, rz represents those stresses which produce
y2 u1k internal work, qz and qy are the force per unit length at
the connection interface in the longitudinal and vertical
v1j directions, respectively, and the third integral on the
u2k
right-hand side represents the work done by the surface
v2j forces applied at the cross-section at the ends of the seg-
ment of beam considered.
Γy v1'
Eq. (7) ensures that both compatibility and equilibrium
conditions are satisfied, and the static entities, which repre-
sent the stress resultant of the active stresses, are identified
v 2' in relation to their dual kinematical counterparts as out-
Γz lined in Eq. (8), following the integrations at the cross-
section:
Fig. 2. Displacement field. Z L
ðN 1 ^u01 þ N 2 ^u02  M 1^v001  M 2^v002 Þ dz
The kinematic discontinuities at the interface, which 0
include the longitudinal and transverse slips is expressed Z L
by the vector C as þ ½qy ð^v2  ^v1 Þ þ qz ð^u2  y c2^v02  ^u1 þ y c1^v01 Þ dz
0
CðzÞ ¼ uðx; y þ 
c ; zÞ  uðx; y c ; zÞ ¼ Cy ðzÞj þ Cz ðzÞk
Z L
¼ ðv2  v1 Þj þ ðu2  y c2 v02  u1 þ y c1 v01 Þk ð4Þ ¼ ðpz1 ^u1 þ pz2 ^u2  m1^v01  m2^v02 þ py1^v1 þ py2^v2 Þ dz
0
in which  1 ^u1 þ N
þ ½N  2 ^u2  M
 1^v0  M
 2^v0 þ T 1^v1 þ T 2^v2  ð8Þ
1 2 0;L
y c1 ¼ y c  y 1 ; y c2 ¼ y c  y 2 ð5a; bÞ
and Cy represents the transverse slip while Cz the longi- In Eq. (8), the internal axial forces and moments resisted
tudinal one. Note that the term y c1 and y c2 can assume neg- by elements 1 and 2, i.e., N1(z), N2(z), M1(z) and M2(z),
ative values depending upon the reference system adopted simply referred to as N1, N2, M1 and M2, are defined as
(Fig. 1). Z
The relevant strain field can then be obtained from the N 1 ðzÞ ¼ rz da ¼ N c ðzÞ þ N r ðzÞ ð9aÞ
A
displacement field given in Eq. (3), and the non-vanishing Z1
strain components can be obtained differentiating Eq. (3) N 2 ðzÞ ¼ rz da ¼ N s ðzÞ ð9bÞ
with respect to the beam coordinate as A
Z2
eza ðx; y; zÞ ¼ u0a  ðy  y a Þv00a ; 8ðx; yÞ 2 Aa ; z 2 ½0; L ð6Þ M 1 ðzÞ ¼ ðy  y 1 Þrz da ¼ M c ðzÞ þ M r ðzÞ ð9cÞ
A
Z 1
2.3. Global balance conditions M 2 ðzÞ ¼ rz ðy  y 2 Þ da ¼ M s ðzÞ ð9dÞ
A2
The partial interaction problem is solved using the prin-
ciple of virtual work. In the weak formulation, the global and a = 1, 2 refer to the two layers, i.e., elements 1 and 2,
equilibrium condition can be defined as outlined in Eq. respectively, Nc(z), Nr(z), Ns(z), Mc(z), Mr(z) and Ms(z) are
(7) below. This is accomplished by equating the work of the internal axial forces and moments resisted by the con-
internal stresses to the work of external actions for each vir- crete component, reinforcement and steel joist, respec-
tual admissible variation (expressed using a hat ‘‘ ˆ’’) of the tively, while T a , N
 a and M
 a are the concentrated vertical
displacement field ^ b
u as well as of the strain ^ez and slip C, and horizontal loads and the bending moment applied to
defined in Eqs. (3), (4) and (6). For generality, the beam is the end cross-sections, i.e., z = 0,L, of layer a defined as
Z Z
assumed to be subjected to quasi-static body forces, acting

T a0;L ðzÞ ¼ 
sy da; N a0;L ðzÞ ¼ sz da;
on the interior points, and surface forces, acting on the Aa0;L Aa0;L
boundary referred to as b = [bx, by, bz] and s = [sx, sy, sz], Z
respectively. The solution of the problem is sought in the  a0;L ðzÞ ¼
M ðy  y a Þsz da ð10a; b; cÞ
spaces of the regular functions fulfilling the kinematical Aa0;L

boundary conditions:
Z Z Z and, for ease of notation, the body and surface forces have
rz^ez da dz þ ðqz C ^ z þ qy C
^ y Þ dz been re-arranged (with a = 1, 2) as
L AZ Z L Z Z Z Z
¼ b^u da dz þ s^ u dl dz py a ðzÞ ¼ by da þ sy dl;
LZ A L XA A XA
Z a Z a
þ s^u da; 8^ u ð7Þ pza ðzÞ ¼ bz da þ sz dl ð11a; bÞ
A0;L Aa XAa
G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384 2377
Z Z
elastic moduli, respectively. The stresses resisted at the
ma ðzÞ ¼ ðy  y a Þbz da þ ðy  y a Þsz dl ð11cÞ
Aa XAa composite cross-section are then calculated as

where pya, pza and ma represent the distributed vertical and rzc ðx; y; zÞ ¼ Ec ½u01  ðy  y 1 Þv001  ð15aÞ
horizontal loads and the bending moment applied to the rzr ðx; y; zÞ ¼ Er ½u01  ðy  y 1 Þv001  ð15bÞ
domain a. The loading vectors p, Q0 and QL collect the ac- rzs ðx; y; zÞ ¼ Es ½u02  ðy  y 2 Þv002  ð15cÞ
tions applied along the beam and at the cross-sections at
the ends of the beam segment considered, i.e., at z = 0, in which rzc ðx; y; zÞ, rzr ðx; y; zÞ and rzs ðx; y; zÞ are the stresses
L, and are expressed as resisted by the concrete component, reinforcement and
steel joist.
pT ¼ ½ pz1 pz2 m1 m2 py1 py2  ð12aÞ It is also assumed that the interface connection between
QT0;L ¼ ½ N
1 2
N 1 M
M  2 T 1 T 2 0;L ð12bÞ elements 1 and 2 is elastic in both longitudinal and trans-
verse directions. The connection stiffness qz defines the rela-
tionship between the longitudinal flow per unit length qz
2.4. Local equilibrium conditions and the longitudinal slip at the interface Cz, while qy relates
the vertical force per unit length qy to the vertical uplift at
The strong form of the proposed formulation describing the interface Cy; these can be expressed as
the partial interaction problem is obtained integrating Eq.
qz ðzÞ ¼qz Cz ¼ qz ðu2  y c2 v02  u1 þ y c1 v01 Þ ð16aÞ
(7) by parts and using the fundamental lemma of varia-
tional calculus (Euler problem), which yields the governing qy ðzÞ ¼qy Cy ¼ qy ðv2  v1 Þ ð16bÞ
system of differential equations expressed in terms of the Eq. (16) are utilised in the derivation of the global and
internal stress resultants as local balance conditions while the parametric study pro-
N 01  qz ¼ pz1 ð13aÞ posed in the following adopts a bi-linear constitutive law
for the modelling of the vertical interface connection. For
N 02 þ qz ¼ pz2 ð13bÞ
this latter case, the model for the vertical connection
M 001  qy  y c1 q0z ¼ py1 þ m01 ð13cÞ (Fig. 3) can be expressed as
M 002 þ qy þ y c2 q0z ¼ py2 þ m02 ð13dÞ (
qys ðv2  v1 Þ ðv2  v1 Þ > 0
qy ðzÞ ¼ qy Cy ¼ ð17Þ
The relevant boundary conditions at the ends of the seg- qyb ðv2  v1 Þ ðv2  v1 Þ 6 0
ment considered, i.e., at z = 0, L, are expressed as
in which qys and qyb are the two vertical connection stiff-
 1 Þ^
ðN 1  N u1 j0;L ¼ 0; 8^
u10;L ð14aÞ nesses, the former being applicable when vertical separa-
 2 Þ^ tion occurs between the two layers, i.e., (v2  v1) > 0, and
ðN 2  N u2 j0;L ¼ 0; 8^
u20;L ð14bÞ
the latter in the case of vertical bearing, i.e., (v2  v1) 6 0.
 1 Þ^v0 j ¼ 0;
 ðM 1  M 8^v010;L ð14cÞ These two values have been introduced as, realistically, al-
1 0;L
ready in the linear-elastic range of the materials forming
 2 Þ^v0 j0;L ¼ 0;
 ðM 2  M 8v^0 20;L ð14dÞ
2 the cross-section the vertical behaviour of the connection
ðM 01 þ y c1 qz  T 1 þ m1 Þ^v1 j0;L ¼ 0; 8^v10;L ð14eÞ significantly varies for the cases of separation of and bear-
ing between the two layers, i.e., qyb  qys.
ðM 02 þ y c2 qz  T 2 þ m2 Þ^v2 j0;L ¼ 0; 8^v20;L ð14fÞ

The notation is as defined in Eqs. (9)–(12). In particular,


Eqs. (13) represent the four local equilibrium equations rel- qy
evant to the partial interaction problem considered, i.e.,
horizontal and vertical equilibrium at the cross-sections
of the two layers, while Eq. (14) are applied at the end
cross-sections of the beam segment considered; they ρys
express either the kinematical (essential) boundary condi- 1
tions when the generic displacements are known so that
its variation is nil, or the dual static (natural) boundary Γy
conditions when the displacements are unknown and the
dual static condition must equal zero.

3. Material properties ρyb


It is now assumed that the concrete, reinforcing bars and
1
steel joist behave in a linear-elastic manner in both com-
pression and tension, and Ec, Er and Es are their relevant Fig. 3. Bi-linear constitutive model for the vertical interface connection.
2378 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384
Z Z
4. Displacement-based formulation
Ar ¼ da; Sr ¼ ðy  y 1 Þ da;
Ar Ar
4.1. Global equilibrium conditions (linear-elastic modelling) Z
Ir ¼ ðy  y 1 Þ2 da ð22d; e; fÞ
The weak formulation of the partial interaction problem Ar
Z Z
is re-written based on Eq. (7) assuming all material pro-
A2 ¼ da; S2 ¼ ðy  y 2 Þ da;
perties to be linear-elastic including the longitudinal and A2 A2
vertical interface connection. In compact form: Z
2
Z Z I2 ¼ ðy  y 2 Þ da ð22g; h; iÞ
A2
½KDd  D^ d dz ¼ ðp  H^ dÞ dz þ ½Q  H^
d0;L ; 8^
d ð18Þ
L L In particular, Sc and Ic represent the first and second
moments of area of the concrete component, while Sr, Ir,
where the last term on the right-hand side of the equation S2 and I2 correspond to the equivalent cross-sectional
represents the external work produced by the actions ap- properties for the reinforcement and for the cross-section
plied at the cross-sections at the ends of the segment con- of element 2, i.e., the steel joist.
sidered (i.e., z = 0,L), while the stiffness matrix K, which
includes the axial and flexural rigidities, is defined as
2 3 4.2. Local equilibrium conditions (linear-elastic modelling)
qz 0 qz 0 0 qz y c1 0 0 qz y c2 0
6 ~1 e1 7
6 0 EA0 0 0 0 0
ES 0 0 7
6
6 qz
7 The strong formulation of the partial interaction prob-
6 0 qz 0 0 qz y c1 0 0 qz y c2 0 7 7
6 7 lem was previously stated in Eqs. (13) and (14) in terms
6 0 0 0 E2 A2 0 0 0 0 0 E2 S 2 7
6
6 0
7 of the internal stress resultants. It is worth noting that, in
6 0 0 0 qy 0 0 qy 0 0 7 7
K¼6 7 the linear-elastic range, these resultants can be re-stated
6 qz y c1 0 qz y c1 0 0 qz y 2c1 0 0 qz y c1 y c2 0 7
6 7 in terms of the variables of the displacement field as
6 0 Ee
S1 0 0 0 0 EeI 1 0 0 0 7
6 7
6 7 Z
6 0 0 0 0 qy 0 0 qy 0 0 7
6 7
6 7 N 1 ðzÞ ¼ rz da ¼ N c ðzÞN r ðzÞ
4 qz y c2 0 qz y c2 0 0 qz y c1 y c2 0 0 qz y 2c2 0 5
A1
0 0 0 E2 S 2 0 0 0 0 0 E2 I 2
ð19Þ ¼ Ec ½Ac u01  S c v001  þ Er ðAr u01  S r v001 Þ ð23aÞ
Z
and the formal differential operators D and H are deter- N 2 ðzÞ ¼ rz da ¼ E2 A2 u02  E2 S 2 v002 ð23bÞ
mined as (where o  dzd ) A2
Z
2 3
1 o 0 0 0 0 0 0 0 0 M 1 ðzÞ ¼ ðy  y 1 Þrz da ¼ M c ðzÞ þ M r ðzÞ
60 0 1 o 0 0 0 0 0 0 7 A1
6 7
DT ¼ 6 7;
4 0 0 0 0 1 o o2 0 0 0 5 ¼ Ec ½S c u01  I c v001  þ Er ðS r u01  I r v001 Þ ð23cÞ
0 0 0 0 0 0 0 1 o o2 Z
2 3 M 2 ðzÞ ¼ rz ðy  y 2 Þ da ¼ E2 S 2 u02  E2 I 2 v002 ð23dÞ
1 0 0 0 0 0 A2
60 1 0 0 0 07
6 7
HT ¼ 6 7 ð20a; bÞ Substituting Eq. (23) into Eqs. (13) and (14) yields the
4 0 0 o 0 1 0 5
governing system of differential equations and the relative
0 0 0 o 0 1
boundary conditions for the fundamental case of linear-
elastic material. The system becomes
The elastic cross-sectional properties, which include the
axial and flexural rigidities introduced in Eq. (19), are
~ 1 u00 þ E e
 EA S 1 v000 y c2 v02  u1 þ y c1 v01 Þ ¼ pz1
defined as 1 1  qz ðu2  

ð24aÞ
EA~ 1 ¼ E c Ac þ E r Ar ; Ee
S 1 ¼ Ec S c þ Er S r ;
 E2 A2 u002 þ E2 S 2 v000 y c2 v02  u1 þ y c1 v01 Þ ¼ pz2
2 þ qz ðu2  
EeI 1 ¼ Ec I c þ Er I r ð21a; b; cÞ
ð24bÞ
The first and second moments of area, S and I, are 0000
defined as  Ee
S 1 u000 e
1 þ E I 1 v1  qy ðv2  v1 Þ
Z Z
 qz y c1 ðu02  y c2 v002  u01 þ y c1 v001 Þ ¼ py1 þ m01 ð24cÞ
Ac ¼ da; S c ¼ ðy  y 1 Þ da;
A Ac 0000
Z c  E2 S 2 u000
2 þ E2 I 2 v2 þ qy ðv2  v1 Þ
2
Ic ¼ ðy  y 1 Þ da ð22a; b; cÞ
Ac þ qz y c2 ðu02  y c2 v002  u01 þ y c1 v001 Þ ¼ py2 þ m02 ð24dÞ
G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384 2379

while the relevant boundary conditions: and Ne is the interpolation matrix collecting the shape
functions as
~ 1 u0  E e
½EA  1 ^
S 1 v001  N u1 j0;L ¼ 0; 8^
u10 ;L ð25aÞ 2 3
1 ll lr 0 0 0 0 0 0 0 0 0 0 lm 0
½E2 A2 u02  E2 S 2 v002  2 ^
N u2 j0;L ¼ 0; 8^
u20 ;L ð25bÞ 60 0 l l 0 0 0 0 0 0 0 0 0 l 7
6 l r m7
Ne ¼ 6 7
 ½E e
S 1 u01  EeI 1 v001  M
 1 v0 j ¼ 0; 8^v010 ;L ð25cÞ 4 0 0 0 0 gl gr t l t r 0 0 0 0 0 0 5
1 0;L

 2 ^v0 j ¼ 00;L ; 0 0 0 0 0 0 0 0 gl gr t l t r 0 0
 ½E2 S 2 u02  E2 I 2 v002  M 2 8^v020 ;L ð25dÞ
ð28Þ
½E e
S 1 u001  EeI 1 v000 y c1 ðu2  y c2 v02  u1 þ y c1 v01 Þ
1 þ qz 
The polynomials adopted in the approximated displace-
 T 1 þ m1 ^v1 j ¼ 0; 8^v1
0;L 0;L
ð25eÞ ment field are specified as
½E2 S 2 u002  E2 I 2 v000 y c2 ðu2  y c2 v02  u1 þ y c1 v01 Þ
2  qz  ll ¼ 1  3n þ 2n2 ; lr ¼ n þ 2n2 ð29a; bÞ
 T 2 þ m2  ^v2 ¼ 0; 8^v2
0;L 0;L 0;L
ð25fÞ lm ¼ 4n  4n 2
ð29cÞ
2 3 2 3
Eqs. (25) express the kinematical boundary conditions gl ¼ 1  3n þ 2n ; gr ¼ 3n  2n ð29d; eÞ
when ^ d ¼ 0 and their dual static boundary conditions when tl ¼ Le ðn  2n2 þ n3 Þ; tr ¼ Le ðn2 þ n3 Þ ð29f; gÞ
^
d 6¼ 0.
where n = z/Le represents the dimensionless coordinate
along the element, and Le represents the finite element
5. Finite element formulation length.
The proposed element has been derived substituting Eq.
A finite element is described in the following for the (26) into the weak formulation of the partial interaction
analysis of composite beams with longitudinal and vertical problem of Eq. (18), which yields the following expression
partial interaction. The proposed element possesses 14dof for an element
(where dof stands for degrees of freedom), as depicted in Z
T
Fig. 4, and these include the vertical displacements and ½ðDNe Þ KðDNe Þde   ^de dz
the rotations at the element ends of both layers, as well Le
Z
as the longitudinal displacements at the level of the refer-
¼ ½ðHNe ÞT p  ^de dz þ ½ðHNe ÞT Q0;Le  ^de ; 8^
de ð30Þ
ence axis for elements 1 and 2. Le
The shape functions adopted for the vertical and axial
This is re-arranged at element level in a more compact
displacements of the two layers consist of third and second
form as
order polynomials, respectively. This was implemented to
ensure consistent contributions from the axial and trans- Ke de ¼ pe ð31Þ
verse displacement fields to the expression for the longitu-
where Ke is the stiffness matrix of the element and pe is the
dinal slip. The second order polynomials for the axial
vector of the nodal external actions accounting for the
displacements require two internal nodes as shown in
forces distributed along the element and those concen-
Fig. 3. The unknown displacements are approximated by
trated at the element ends.
a linear combination of interpolating shape functions
According to the principle of virtual work, the equilib-
according to
rium condition for the entire structure can then be
d ffi N e de ð26Þ expressed by assembling the vectors and matrices defined
for each element following the usual solution procedure
where de is the vector of the unknown nodal displacements of the finite element formulation [32].
dTe ¼ ½ u1l u1r u2l u2r v1l v1r v01l v01r v2l v2r v02l v02r u1m u2m 
6. Applications
ð27Þ
The proposed finite element formulation is utilised in this
section for the analysis of a simply supported beam subjected
θ1l u1l u1m u1r θ1r to a point load applied to the bottom element at mid-span. A
v1l v1r 10 m long beam has been adopted with a cross-section
formed by a rectangular slab (2300 mm · 200 mm) and a
fabricated steel joist which has a top flange 300 mm ·
u2l u2m u2r 20 mm, a web 1550 mm · 15 mm and a bottom flange
θ2l θ2r 450 mm · 30 mm. [33] The elastic modulus adopted for the
v2l v2r steel reinforcement and joist is 210,000 MPa. A concrete
strength fc0 of 32 MPa has been utilised, and the relevant elas-
Fig. 4. 14dof finite element. tic material properties of the concrete have been calculated
2380 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384

at 28 days following design guidelines [34]. In the applica- the length of the beam, h represents the distance between
tions it has been assumed that the restraints provided by the centroids, and the subscripts ‘L’, ‘Ts’ and ‘Tb’ refer
the pinned or roller supports of the simply supported beam to the longitudinal interface connection and to the trans-
are applied to the bottom element, while the top element is verse one with vertical separation and bearing, respectively.
supported by element 2 by means of the transverse connec- In order to provide a common basis of comparisons, it
tion. Any other support conditions can be easily imple- has been assumed that the transverse bearing connection
mented in the solution procedure. stiffness adopted is such as to yield a very high value of
The bi-linear constitutive law outlined in Eq. (17) is aTbL, i.e., equal to 1000 in Eq. (32c); this represents a more
adopted to model the behaviour of the transverse interface realistic situation in which the bearing stiffness is usually
connection. As the extent of separation and bearing along very high.
the beam is not known a priori due to the bi-linear nature Several simulations have been carried out to evaluate
of the vertical connection, an iterative procedure is how different levels of the transverse separation connection
required to converge to the final solution in which the seg- stiffness affect the overall response. For this purpose, a
ments of beam undergoing separation are modelled using point load equal to the self-weight of the beam has been
qys, and qyb is adopted in the remaining parts of the beam. arbitrarily applied to emphasise the partial interaction
The iterative procedure starts by assuming the separation behaviour. Due to the non-linearity of the vertical interface
between the two layers to occur throughout the beam connection the results depend on the magnitude of the
length. In successive iterations, the appropriate transverse applied loading.
stiffnesses are specified along the segments of separation Figs. 5 and 6 depict the differences between the values of
and bearing, respectively. To better depict the final solu- the mid-span deflection and curvature calculated using the
tion, additional nodes are inserted at the location of tran-
sitions between separation and bearing; therefore, the
mesh is modified during the solution process. In adjacent 1 10 100 1000
iterations the additional nodes are relocated till conver- 0%
gence is established when the Euclidean norm of these k Ts /k L
adjustments is negligible. This approach has proven to be
-1%
robust and able to converge to the final solution in a lim-
ited number of iterations. The main advantage of this solu-
tion strategy over more generic non-linear iterative -2%
procedures, i.e. Newton–Raphson, secant stiffness, etc., is
that the transitions between the segments with transverse -3%
bearing and separation are accurately modelled by intro-
ducing new nodes where required, therefore avoiding the α L L =1
changes in the vertical connection stiffnesses to be -4% α L L =5
v 1 -v Newmark α L L =10
smoothed along the length of the elements containing these
v Newmark z =L /2 α L L =20
transitions. By doing this, each element of the mesh experi- -5%
ences either separation or bearing between the two layers
Fig. 5. Differences in the values of the mid-span deflection of the top
over its whole length. element calculated using the proposed and Newmark models.
The results of the proposed applications have been pre-
sented as a function of the dimensionless stiffness parame-
ter noted by Girhammar and Pan [9] for the case of
longitudinal partial interaction only. In particular, this 1 10 100 k Ts1000
/k L
parameter has been re-utilised to depict the level of rigidity 0%
for the longitudinal and for both transverse interface con-
nections as follows: -5%
pffiffiffiffiffiffiffi
aL L ¼ ~ aqz L ð32aÞ
qffiffiffiffiffiffiffiffi
-10%
aTs L ¼ ~ aqys L ð32bÞ
qffiffiffiffiffiffiffiffiffi
aTb L ¼ ~ aqyb L ð32cÞ -15%

and α L L =1
-20%
κ 1 - κ Newmark α L L =5
1 1 h2 α L L =10

~ þ þ ð33Þ κ Newmark z =L /2
A1 E 1 A2 E 2 I 1 E 1 þ I 2 E 2 -25% α L L =20
where the cross-sectional properties included in ~
a are calcu- Fig. 6. Differences in the values of the mid-span curvature of the top
lated about the centroidal axes of the two components, L is element calculated using the proposed and Newmark models.
G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384 2381

s h (z =0) proposed model and Newmark model. For the longitudinal


35% max (a L L= 1) connection stiffnesses common values equivalent to aLL =
vNewmark
1, 5, 10, 20 have been adopted. The transverse separation
30%
stiffnesses has been varied as a function of the longitudinal
α L L =1
25% α L L =5 connection stiffness considering rations of the two in the
α L L =10 range between 1 and 1000 (where the ratio is calculated
20%
α L L =20 dividing the transverse separation stiffness by the longitudi-
15% nal one). This selected range reflects the fact that the

10%

5% κ 1(α L L =1) k Ts /k L =1
k Ts /k L 1.0 k Ts /k L =5
κ Newmark
max (a L =1)
L

0% k Ts /k L =10
1 10 100 1000 0.8 k Ts /k L =20

Fig. 7. Variation of the longitudinal end slip calculated using the


proposed model and non-dimensionalised with the mid-span deflection 0.6
of Newmark model.

0.4

s v (z =L /2) 0.2
5%
max (a L =1)
L α L L =1
vNewmark
α L L =5 0.0
4% α L L =10 0 0.2 0.4 0.6 0.8 1
α L L =20 z/L
3%
κ 1(α L L =5) k Ts /k L =1
1.0 max (a L L =5) k Ts /k L =5
κ Newmark
2% k Ts /k L =10
0.8 k Ts /k L =20
1%
0.6
k Ts /k L
0% 0.4
1 10 100 1000
Fig. 8. Variation of the transverse mid-span slip calculated using the 0.2
proposed model and non-dimensionalised with the mid-span deflection of
z/L
Newmark model. 0.0
0 0.2 0.4 0.6 0.8 1
-0.2

0.8 q v (z =L /2) α L L =5 κ 1(α L L =10) k Ts /k L =1


α L L =10 1.0 k Ts /k L =5
qh(z=0)
Newmark
max (a L L =10)
κ Newmark
0.7 α L L =20 k Ts /k L =10
0.8 k Ts /k L =20
0.6

0.5 0.6
0.4
0.4
0.3

0.2 0.2

0.1 z/L
k Ts /k L 0.0
0.0 0 0.2 0.4 0.6 0.8 1
1 10 100 1000 10000 -0.2 /L

Fig. 9. Variation of the mid-span transverse distributed connection force Fig. 10. Variation of the curvature of the top element along the member
calculated using the proposed model and non-dimensionalised with the length calculated with the proposed and Newmark models: (a) aLL = 1,
longitudinal distributed connection force of Newmark model. (b) aLL = 5, (c) aLL = 10.
2382 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384

transverse separation stiffness is usually expected to be Newmark model and, therefore, its behaviour has not been
greater than the longitudinal one; this would certainly be plotted. Similarly to the deflection case, the mid-span cur-
the case for common shear connectors. Fig. 5 highlights vature of the top layer calculated with the proposed model
that slightly smaller values for the deflection of the top appears to yield smaller values than those obtained with
component are obtained using the proposed method as Newmark model for a wide range of connection stiffnesses
the overall behaviour is affected by the transverse connec- as depicted in Fig. 6; the trend of the curvature of the bot-
tion stiffness. On the other hand, the behaviour of the bot- tom layer is similar to the one obtained with Newmark
tom layer is very similar to the one observed with model and, therefore, this comparison has not been
presented.
The variation of the longitudinal end slip and the mid-
s v (α L L =1) k Ts /k L =1 span vertical separation are illustrated in Figs. 7 and 8.
5% max (a L L =1) k Ts /k L =5
v Newmark These values have been non-dimensionalised against the
k Ts /k L =10
4%
mid-span deflection calculated using Newmark model.
k Ts /k L =20
Fig. 7 well depicts that the longitudinal slip is not affected
by the rigidity of the vertical interface connection. Even if
3%
the magnitude of the transverse slip tends to reduce while
increasing aTsL, the actual transverse action resisted by
2%
the connection increases with the transverse connection
rigidity. This trend is plotted in Fig. 9 where the transverse
1%
z/L
distributed force calculated at mid-span is non-dimensiona-
lised against the maximum longitudinal distributed force
0%
obtained with Newmark model (i.e., at the extremes of
0 0.2 0.4 0.6 0.8 1 the beam). This consideration emphasises that the main
-1% advantage of the proposed method relies on the fact that
the combined stress-state, i.e., shear and axial forces,
s v (α L L =5) k Ts /k L =1 resisted by the interface connection can be accurately deter-
1.4% k Ts /k L =5
v max (a L L =5)
Newmark
mined. In the context of the applications considered, the
1.2% k Ts /k L =10 magnitude of the transverse distributed force tends to
k Ts /k L =20 approach the one measured longitudinally with Newmark
1.0%
model for increasing values of the connection stiffness.
0.8% Due to the non-linearity of the problem, the proposed
results are load dependent and, therefore, varying the mag-
0.6% nitude and nature of the applied loading would affect the
0.4% significance of the transverse partial interaction; also, mem-
ber loadings would produce more pronounced effects over
0.2% the whole beam length.
z/L
0.0% The variation of the curvature of the top layer and of the
0 0.2 0.4 0.6 0.8 1 vertical separation along the beam lengths have been plot-
-0.2% ted in Figs. 10 and 11 for different levels of the longitudinal
connection stiffnesses, i.e., aLL = 1, 5, 10. It is clear that the
s v (α L L =10) k Ts /k L =1 proposed model enables to better depict the behaviour in
0.8% k Ts /k L =5
v max (a L L =10) the central region of the beam and at the supports, where
Newmark
k Ts /k L =10 the latter tend to be subjected to negative curvature in
0.6% k Ts /k L =20 the top element when compared with the results obtained
using Newmark model.
0.4%
7. Conclusions

0.2% This paper proposed an analytical model for the analysis


of two-layered composites with longitudinal and transverse
z/L
0.0%
partial interaction, capable of accounting for both vertical
and horizontal relative displacements which occur at the
0 0.2 0.4 0.6 0.8 1
interface connection between the two layers.
-0.2% The proposed analytical formulation has been outlined
Fig. 11. Variation of the transverse slip along the member length for completeness in both its weak and strong forms. The
calculated with the proposed model and non-dimensionalised with the governing system of differential equations has been pre-
mid-span deflection of Newmark model. sented as a function of both the internal stress resultants
G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384 2383

and of the variables included in the adopted displacement [7] Faella C, Martinelli E, Nigro E. Steel and concrete composite beams
field; the complete set of boundary conditions has been with flexible shear connection: ‘‘exact’’ analytical expression of
the stiffness matrix and applications. Comput Struct 2002;80:
included in both instances. 1001–9.
Based on this formulation a novel finite element has [8] Cosenza E, Mazzolani S. Linear-elastic analysis of composite beams
been proposed to enable the numerical applications to be with partial shear interaction. In: Proceedings of the first Italian
easily carried out. This element possesses fourteen degrees workshop on composite structures. University of Trento; June 1993
of freedom, which include the deflections, axial displace- [in Italian].
[9] Girhammar UA, Pan D. Dynamic analysis of composite members
ments and rotations of both layers at the element ends. A with interlayer slip. Int J Solids Struct 1993;30(6):797–823.
cubic polynomial and a parabolic polynomial have been [10] Dall’Asta A, Zona A. Three-field mixed formulation for the non-
adopted to describe the shape functions of the unknown linear analysis of composite beams with deformable shear connection.
displacements, which are the vertical and axial displace- Finite Elements Anal Design 2004;40:425–48.
ments; the latter parabolic approximation required an [11] Loh H, Uy B, Bradford MA. The effects of partial shear interaction in
the hogging moment regions of composite beams: Part II – analytical
additional internal node to be inserted at mid-span of the study. J Construct Steel Res 2004;60:921–62.
element for each layer. [12] Faella C, Martinelli E, Nigro E. Steel connection nonlinearity and
The method has been applied to a simply supported deflections of steel–concrete composite beams: a simplified approach.
beam subjected to a point load applied to the bottom layer J Struct Eng ASCE 2003;129(1):12–20.
at mid-span. Different combinations of stiffnesses have [13] Ayoub A. A two-field mixed variational principle for partially
connected composite beams. Finite Elements Anal Design 2001;37:
been considered in the applications for the longitudinal 929–59.
and transverse interface connections. To provide a com- [14] Salari MR, Spacone E. Finite element formulations of one-dimen-
mon basis for comparison a very high level of connection sional elements with bond-slip. Eng Struct 2001;23(7):815–26.
stiffness has been adopted for the transverse bearing stiff- [15] Ayoub A, Filippou FC. Mixed formulation of nonlinear steel–
ness. The iterative procedure utilized in the numerical solu- concrete composite beam element. J Struct Eng ASCE 2000;126(3):
371–81.
tions has been presented. This is required as the extent of [16] Fabbrocino G, Manfredi G, Cosenza E. Analysis of continuous
the bearing and separation regions along the structural ele- composite beams including partial interaction and bond. J Struct Eng
ment is not known a priori. ASCE 2000;126(11):1288–94.
Based on the results obtained for the applications con- [17] Gattesco N. Analytical modelling of nonlinear behaviour of com-
sidered, the significance of adopting the proposed posite beams with deformable connection. J Construct Steel Res 1999;
52:195–218.
approach has been emphasized as necessary to well depict [18] Nguyen NT, Oehlers DJ, Bradford MA. A rational model for the
the combined stress-state, i.e., longitudinal and transverse degree of interaction in composite beams with flexible shear connec-
forces, resisted by the interface connection as its effects tors. Mechanics Struct Machines 1998;26(2):175–94.
have been observed to be potentially important. For this [19] Oehlers DJ, Sved G. Composite beams with limited-slip-capacity
reason, further numerical and experimental work is connectors. J Struct Eng ASCE 1995;121(6):932–8.
[20] Virtuoso F, Vieira R. Time dependent behaviour of continuous
required to better understand this composite behaviour composite beams with flexible connection. J Construct Steel Res 2004;
and, in particular, in indeterminate structural systems as, 60:451–63.
to date, available modelling techniques have focused on [21] Fragiacomo M, Amadio C, Macorini L. Influence of viscous
determinate structures. Parametric studies need to be car- phenomena on steel–concrete composite beams with normal or high
ried out for several structural typologies to determine the performance slab. Steel Composite Struct 2002;2(2):85–98.
[22] Kwak HG, Seo YL. Time-dependent behaviour of composite beams
significance of the transverse partial interaction, and par- with flexible connectors. Comput Methods Appl Mech Eng 2002;191:
ticularly its effect on the distributions of the internal 3751–72.
actions. [23] Dezi L, Gara F, Leoni G, Tarantino AM. Time dependent analysis
of shear-lag effect in composite beams. J Eng Mech 2001;127(1):
71–9.
References [24] Dezi L, Leoni G, Tarantino AM. Creep and shrinkage analysis of
composite beams. Progr Struct Eng Mater 1998;1(2):170–7.
[1] McGuire W, Gallagher RH, Zieman RD. Matrix structural analysis. [25] Gilbert RI, Bradford MA. Time-dependent behavior of continuous
2nd ed.. New York: John Wiley; 2000. composite beams at service loads. J Struct Eng ASCE 1995;121(2):
[2] Oehlers DJ, Bradford MA. Composite steel and concrete structural 319–27.
members: fundamental behaviour. Oxford: Pergamon Press; 1995. [26] Adekola AO. Partial interaction between elastically connected
[3] Newmark NM, Siess CP, Viest IM. Tests and analysis of composite elements of a composite beam. Int J Solids Struct 1968;4:1125–35.
beams with incomplete interaction. Proc Soc Exp Stress Anal 1951; [27] Chapman JC, Balakrishnan S. Experiments on composite beams.
9(1):75–92. Struct Eng 1964;42(11):369–83.
[4] Ranzi G, Bradford MA, Uy B. A direct stiffness analysis of a [28] Fox L. The numerical solution of two-point boundary problems in
composite beam with partial interaction. Int J Numer Methods Eng ordinary differential equations. Oxford: Oxford University Press;
2004;61:657–72. 1957.
[5] Seracino R, Lee CT, Lim TC, Lim JY. Partial interaction stresses in [29] Aribert JM, Abdel Aziz K. Modelling of composite beams in
continuous composite beams under serviceability loads. J Construct ultimate state with vertical uplift. Construct Métall 1985;4:3–36 [in
Steel Res 2004;60:1525–43. French].
[6] Wu YF, Oehlers DJ, Griffith MC. Partial-interaction analysis of [30] Robinson H, Naraine KS. Slip and uplift effects in composite beams.
composite beam/column members. Mech Struct Machines 2002;30(3): In: Proceedings of the engineering foundation conference on com-
309–32. posite construction (ASCE). 1988. p. 487–97.
2384 G. Ranzi et al. / Computers and Structures 84 (2006) 2373–2384

[31] Nguyen NT, Oehlers DJ, Bradford MA. An analytical model for [33] Tarantino AM, Dezi L. Creep effects in composite beams with
reinforced concrete beams with bolted side plates accounting for flexible shear connectors. J Struct Eng (ASCE) 1992;118(8):2063–
longitudinal and transverse partial interaction. Int J Solids Struct 81.
2001;38:6985–96. [34] Comité Euro-International du Béton-Fédération International de la
[32] Cook R, Malkus D, Plesha M, Witt R. Concepts and applications of Précontrainte (CEB-FIB). Model Code 1990: Design Code. London:
finite element analysis. 4th ed. New York: Wiley; 2001. Thomas Telford; 1993.

You might also like