You are on page 1of 16

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2004; 61:657–672 (DOI: 10.1002/nme.1091)

A direct stiffness analysis of a composite beam


with partial interaction

G. Ranzi, M. A. Bradford∗, †, ‡ and B. Uy


School of Civil and Environmental Engineering, The University of New South Wales,
UNSW, Sydney, NSW 2052, Australia

SUMMARY
The use of the conventional semi-analytical stiffness method in finite element analysis, in which
interpolation polynomials are used to develop the stiffness relationships, leads to problems of curvature
locking when beam-type elements are developed for composite members with partial interaction
between the materials of which it is comprised. The curvature locking phenomenon that occurs for
composite steel–concrete members is quite well reported, and the general approach to minimizing
the undesirable ramifications of curvature locking has been to use higher-order polynomials with
increasing numbers of internal nodes. This paper presents an alternate formulation based on a direct
stiffness approach rather than starting from pre-defined interpolation polynomials, and which does not
possess the undesirable locking characteristics. The formulation is based on a more general approach
for a bi-material composite flexural member, whose constituent materials are joined by elastic shear
connection so as to provide partial interaction. The stiffness relationships are derived, and these are
applied to a simply supported and a continuous steel–concrete composite beam to demonstrate the
efficacy of the method, and in particular its ability to model accurately both very flexible and very stiff
shear connection that causes difficulties when implemented in competitive semi-analytical algorithms.
Copyright 䉷 2004 John Wiley & Sons, Ltd.

KEY WORDS: composite beams; curvature locking; direct stiffness method; elastic; finite elements;
partial interaction; shear connection

INTRODUCTION

Engineering structures are making increasing use of composite components, in which the at-
tributes that best suit a particular material are utilized in an optimal efficient and economic
manner. The most familiar composite member, and that which is considered tacitly in this
paper, is the steel–concrete composite beam [1, 2] where concrete is used in the compressive
portion of the member, and steel in the tensile portion of the member. Since concrete is weak

∗ Correspondence to: M. A. Bradford, School of Civil and Environmental Engineering, The University of New
South Wales, UNSW, Sydney, NSW 2052, Australia.
† E-mail: m.bradford@unsw.edu.au
‡ Professor

Received 23 September 2002


Published online 19 August 2004 Revised 22 September 2003
Copyright 䉷 2004 John Wiley & Sons, Ltd. Accepted 7 January 2004
658 G. RANZI, M. A. BRADFORD AND B. UY

in tension, and steel sections are prone to buckle in compression, the best use is made of
the two materials when they are utilized in this fashion. In order to provide a load path to
allow a composite member to be stressed in a way that is suited to the materials of which
it is comprised, it is imperative to provide a shear connection between the materials, and in
steel–concrete composite beams this shear connection is usually provided by headed stud shear
connectors, and in advanced composite materials this shear connection is often provided by
an epoxy resin. The structural mechanics of composite members is not only influenced by the
materials of which it is comprised, but also by the shear connection, and this often makes
the analysis of composite members more difficult than would appear at face value. This paper
addresses the mechanics of a quasi-elastic two-layered composite member with elastic shear
connection between the two materials.
The advantages that accrue to joining materials with a form of mechanical shear connec-
tion have long been known, and is not restricted to steel–concrete composite members, with
applications such as plywood predating the structural application of steel–concrete composite
beams. However, the seminal study and probably the most cited of the work on shear con-
nection is that of Newmark et al. [3], and this work which was motivated by the analysis of
steel–concrete composite beams quantified the influence of elastic shear connection between
two elastic materials (steel and concrete). By making highly simplifying assumptions in the
mechanics of materials, such as elastic behaviour, the Euler–Bernoulli theory of bending and
preventing vertical separation of the two materials, the solution for the midspan deflection of
a beam under uniformly distributed loading was shown to be far more complicated than the
5
familiar 384 wL4 /EI result, owing to the influence of elastic shear connection between the two
materials allowing them to slip relative to each other, so as to provide partial rather than full
interaction. Newmark’s solutions thus highlighted the complexities in the analysis of composite
beams that can arise when they are subjected to partial interaction, and indeed for more general
cases of loading and geometry closed form solutions are not available for the deflections and
stresses within composite beams.
Numerical techniques provide important tools for structural analysis, and the finite element
technique has been applied to the problem of analysing steel–concrete composite beams for
a number of years. Ansourian [4] presented an ambitious finite element technique that could
be applied to continuous steel–concrete composite beams, and that included realistic material
properties for the steel (elastic and plastic regions) and concrete (crushing and limited tensile
strength) and the non-linear load–slip characteristics of the stud shear connectors. However, his
use of master and slave nodes was problem-specific, and although giving valuable information
on ductility in continuous beams, the solution did not provide a general and robust technique.
More general displacement-type approaches have also been used, such as those described by
Arizumi et al. [5] and Daniels and Crisinel [6]. In one form or another, these approaches
start in the traditional finite element sense from interpolating polynomials that satisfy various
degrees of continuity in the displacements and their derivatives, and allow stiffness matrices
to be developed. However, the use of such an approach introduces a curvature-locking prob-
lem [7], which causes numerical instabilities for high and low values of the stiffness of the
shear connection. For example, in an 8-DOF element, the appropriate lengthwise interpolation
functions would be cubic for the transverse deformation v, and linear for the axial deforma-
tion w defined (without any loss of generality) at the centroids of the two materials (concrete
and steel). It has been shown [8, 9] that the expression for the interface slip s is also cubic,
but when a condition that approaches full interaction is enforced in the equation for the slip

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 659

(i.e. s → 0), the expression for the transverse deformation becomes quadratic, so that the v has
been reduced from O(z3 ) to O(z2 ), with the result that the order of the curvature is zero, and
hence the curvature  is constant. The consequences of this locking phenomenon, which occurs
because of the combination of inconsistent field equations, have been described by Prathap and
Naganarayana [8], and the spurious oscillatory trends that occur in related multi-field problems
in engineering numerical methods have been quantified by Dall’Asta and Zona [9].
Some methods of circumventing the locking difficulty have been examined in Reference [9],
particularly trying to maintain parity with the order of the interpolation functions by introducing
internal nodes. One such scheme utilizes a 10-DOF element, with the inclusion of mid-length
nodes for w1 and w2 to arrive at axial displacements of order O(z2 ), or even using a 16-DOF
element with one internal node for v, , and three internal nodes for w1 and w2 . Dall’Asta
and Zona [9] highlighted difficulties that still remain in the use of these higher-order elements,
and suggested the use of a so-called displacement–strain mixed formulation, and described
an HR011 element with constant axial deformations, linear curvature and linear slip, and an
HR010 element with constant axial deformations, linear curvature and constant slip. Salari
et al. [10] and Salari and Spacone [11] presented a force-based formulation, while Ayoub and
Filippou [12] suggested a displacement–stress mixed formulation.
This paper proposes a different computational technique, in which a mixed type of element
is developed from a direct stiffness approach that utilizes an internal solution for the slip in the
same way as in Newmark’s work [3]. The result is an efficacious finite element stiffness matrix
for two quasi-elastic materials connected by an elastic shear connection. Firstly, a generic model
for shear connection between two elastic materials in a bi-material composite flexural member
is developed, and this is then used as a basis for the direct stiffness approach to determine
the stiffness matrix for an element. The use of the element is demonstrated for steel–concrete
composite beams, both statically determinate and statically indeterminate, for which the concrete
cracks and consequently requires a simple modification of the stiffness matrix to be performed.

PARTIAL INTERACTION ANALYSIS

General
The model of partial interaction developed in this section is based on the composite member
shown in Figure 1, and it forms the basis for the direct stiffness approach. The member is
subjected to a known loading regime, that produces a moment field M(z) along the beam which
is not necessarily known a priori, and for which the axial force is zero. The domain of the
cross-section  is comprised of two materials, each of which (i ) is elastic with the usual
constitutive relationship i = Ei i between the normal (longitudinal) stress and strain. These
materials are of arbitrary cross-section, but are symmetric about the axis of bending. The two
materials are connected by the elastic shear connection at the interface shown in Figure 1, in
which the constitutive relationship q = ks between the shear at their interface and the slip at
the interface is defined in terms of its elastic shear connection modulus k. In order to prevent
vertical separation of the materials, it will be assumed that the curvature  is equal for both,
that the strain–displacement relationship is linear, and that the Euler–Bernoulli assumption for
bending is valid. In the development of the model, recourse will be made to an invariant
reference position or fibre at the top level of the cross-section.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
660 G. RANZI, M. A. BRADFORD AND B. UY

εs (slip strain)

h1 y1 y

ELEMENT Ω1 y2

h
h2

ELEMENT Ω2

TYPICAL SECTION STRAIN DIAGRAM

Figure 1. Typical composite cross-section.

Horizontal equilibrium
The axial force in the two materials is

Ni = i dAi (i = 1, 2) (1)
i

where for each material


i = (y − yi ) (i = 1, 2) (2)
in which y is the distance below the reference position, and yi is the co-ordinate of the neutral
axis for material i .
From Equations (1) and (2)
Ni = (Bi − yi Ai )Ei  (i = 1, 2) (3)
where Bi is the first moment of area of i below the reference position. Since the beam is
subjected to bending only, horizontal equilibrium requires that

2
Ni = 0 (4)
i=1

so that

2
 (Bi − yi Ai )Ei = 0 (5)
i=1

which yields a relationship between the neutral axis depths given by



2
B Ẽ − yj (1 − ij )Aj Ej
j =1
yi = (6)
Ai Ei
2
where ij is the Kronecker delta, and B Ẽ = i=1 Bi Ei .

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 661

N1 + (dN1/dz)δz
N1

q (z)

δz

Figure 2. Free body diagram of an element of material 1 of length z.

Slip strain
The slip strain  between materials 1 and 2 is
ds
= = 1 − 2 = (y1 − y2 ) (7)
dz
so that
B Ẽ 1 
yi = − A1 E1 + i2 (i = 1, 2) (8)
AẼ AẼ 
2
where AẼ = i=1 Ai Ei .

Internal bending moment


The internal bending moment within a cross-section is given by
 
2
M =int
y dA = Ei i y dAi (9)
 i=1 i

Hence, substituting Equation (2) into Equation (9) gives



2
M int = I Ẽ −  Bi Ei yi (10)
i=1

where I Ẽ = 2i=1 Ii Ei and Ii is the second moment of area of the material i about the
reference position, and substituting Equation (8) into Equation (10) produces
 
B Ẽ 2 B Ẽ
M = I Ẽ −
int
+ A1 E1 − B1 E1  (11)
AẼ AẼ

Shear connection
Figure 2 shows an element of material 1 with its shear connection on the lower face. By
considering horizontal equilibrium,
 
dN1
N1 + z + qz − N1 = 0 (12)
dz

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
662 G. RANZI, M. A. BRADFORD AND B. UY

so that
dN1
+q =0 (13)
dz
which produces
dN1
= −ks(z) (14)
dz
If Equation (5) is substituted into Equation (3) then
B Ẽ AE
N1 = B1 E1  − A1 E1 −  (15)
AẼ AẼ
2
where AE = i=1 Ai Ei , and so
 
dN1 B Ẽ d AE d2 s
= B1 E1 − A1 E1 − (16)
dz AẼ dz AẼ dz2
which when using the equilibrium equation (14) gives
 
B Ẽ d AE d2 s
B1 E1 − A1 E1 − = −ks (17)
AẼ dz AẼ dz2

The internal moment M int is equal to the applied moment field M(z), so that from Equation (8)
 = M +  (18)
and hence
d dM d2 s
= + 2 (19)
dz dz dz
in which
B1 E1 AẼ − A1 E1 B Ẽ
= (20)
AẼI Ẽ − B Ẽ 2
and
AẼ
= (21)
AẼI Ẽ − B Ẽ 2
Substituting Equation (19) into Equation (27) produces the following generic linear differential
equation for partial interaction:
d2 s dM(z)
˜ − ks =  (22)
dz2 dz
in which
AEI Ẽ − (B1 E1 )2 A2 E2 − (B2 E2 )2 A1 E1
˜ = (23)
AẼI Ẽ − B Ẽ 2

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 663

The differential equation (22) may be solved to produce the slip and slip strain, respectively,
as
1 dM
s = C1 ez + C2 e−z −  (24)
k dz
1 d2 M
 = C1 ez − C2 e−z −  (25)
k dz2

where
k
2 = (26)

Hence from Equation (18)
2 d 2 M
 = M + C1 ez − C2 e−z − (27)
k dz2
which when integrated with respect to z yields the slope as
  2
z −z 2 d M
 =  M dz + C1 e + C2 e − dz + Ĉ1 (28)
L k L dz2
and which when again integrated with respect to z produces the deflection of the beam as
    2
C1 ez C2 e−z 2 d M
v= M dz dz + − − dz dz + Ĉ1 z + Ĉ2 (29)
L   k L dz
2

The axial force in material i = 1 is also needed for the direct stiffness technique. This is given
after appropriate substitution into Equation (15) as
N1 = 1  + 2  (30)
in which
B Ẽ
1 = B1 E1 − A1 E1 (31)
AẼ
A1 E1 A2 E2
2 = − (32)
AẼ

DIRECT STIFFNESS APPROACH

In formulating the direct stiffness approach that is underpinned by the partial interaction analysis
derived in the previous section, the vectors of the nodal freedoms and actions at the ends are
given, respectively, by

Q = V0 , M0 , N0 , VL , ML , NL T (33)
q = v0 , 0 , s0 , vL , L , sL T (34)

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
664 G. RANZI, M. A. BRADFORD AND B. UY

where the subscripts denote the ends z = 0, L and in which the slip may be thought of being
conjugate in an abstract sense to the axial force within the material 1 . The direct stiffness
procedure assembles the stiffness matrix from an initially unloaded element and whose freedoms
are all restrained except for one, for which a unit value is imposed. When the moment is then
expressed as

M = −M0 + R0 z (35)

the expressions in Equations (27)–(29) become

 = (R0 z − M0 ) + C1 ez − C1 e−z (36)


 
R0 z2
= − M0 z + C1 ez + C1 e−z + Ĉ1 (37)
2
 
R0 z3 M0 z 2 C1 ez C1 e−z
v= − + + + Ĉ1 z + Ĉ2 (38)
6 2  

and the slip in Equation (24) becomes



s = C1 ez + C2 e−z − R0 (39)
k
The imposition of a unit vertical displacement at z = 0 produces the kinematic state described by

q = 1, 0, 0, 0, 0, 0T (40)

and when substituted into Equations (37) to (39) allows the constants of integration C1 and C2
(by the use of s0 = sL = 0), Ĉ1 and Ĉ2 (by the use of v0 = 1, vL = 0) and R0 and M0 (by the
use of 0 = L = 0) to be obtained. Similar sequential impositions of q = 1n , 2n , . . . , 6n T
(where in is the Kronecker delta) from n = 2 through to n = 6 allow the six constants of
integration to be obtained at each step, and hence the corresponding entries in the vector Q.
The terms in the stiffness matrix defined by
   

 −wL/2   k11 k12 k13 k14 k15 k66

 


   

 −wL2 /12 
  k22 k23 k24 k25 k26 

  
 

 −wL    
 k33 k34 k35 k36 
Q+ =  q (41)
 −wL/2 
    k k k

46 

 
 
44 45


 
  

 wL 2
/12 
  sym k k 

 

55 56

 

wL k66

are given in Appendix 1 of Reference [13], in which

(e−L + 1)(1  + 2 )/2 − (e−L − 1)(Lk/12 − 1 /L − 2 /L)


= (42)
k(e−L − 1)

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 665

and w is the uniformly distributed loading on the element. The second vector on the left hand
side of Equation (41) is the vector of fixed end actions resulting from the imposition of w.
Similar stiffness coefficients have been derived in Faella et al. [14] by inverting the flexibility
matrix for a composite simply supported beam. The derivation of the flexibility coefficients is
also presented in Cosenza and Pecce [15]. In both References [14, 15], the differential equation
describing the behaviour of the composite beam is expressed and solved with respect to the
curvature while the slip expression is written in terms of hyperbolic functions. In this paper,
however, the differential equation is expressed and solved with respect to the slip while two
different expressions of the slip are utilized, one in terms of exponentials as in Equation (39)
(equivalent to the one in terms of the hyperbolic functions) and one in terms of the truncated
Taylor series of these exponentials, in order to overcome numerical instabilities as outlined
below.

Numerical instability
The exponential terms in Equation (39) are a potential source of numerical instability for low
values of the shear connection parameter  in Equation (31), which requires a consideration
of large terms when exponentials in  are inverted within the stiffness relationships. Figure 3
illustrates this behaviour, where the normalized slip freedom ks /ks1 on the diagonal of the
stiffness matrix (corresponding to k(3, 3) of Equation (41)) is plotted as a function of the
dimensionless
L stiffness defined by Girhammar and Pan [16] as
  1/2
1 1 h2

= k + + = (43)
A1 E1 A2 E2 I1 E1 + I2 E2

in which  was defined in Equation (26) and ks1 is the slip freedom determined for
L = 1.
The value of
L is here calculated for each element and the cross-sectional properties in
Equation (43) are calculated about the centroid of each element i (i = 1, 2) and h represents
the distance between these centroids, while the cross-sectional properties used to calculate 
defined in Equation (26) are calculated about an invariant reference position. It can be seen
from Figure 3 that an oscillatory trend is present when
L reaches a small (indeed infinitesimal)
value, and reliable results are obtained when
L > 0.05.
In order to retain the robustness of the algorithm, the stiffness matrix needs to be mod-
ified for low values of the shear connection. This has been achieved following the same
procedure of the direct stiffness method previously outlined, where the exponentials appearing
in the expression of the slip of Equation (39) have been expanded as a Taylor series trun-
cated at the eighth term. The stiffness coefficients so derived are given in Appendix II of
Reference [13].

7 (z)j 7 (z)j 
s = C1 + C2 (−1)j − R0 (44)
j =0 j ! j =0 j! k

In the proposed modelling procedure the element stiffness matrix is calculated by means of the
direct stiffness approach expressing the slip with the exponentials as in Equation (39) or with
their truncated Taylor series as in Equation (44) depending upon the value of
L calculated

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
666 G. RANZI, M. A. BRADFORD AND B. UY

Stiffness coefficient on the diagonal of the element stiffness matrix


related to the slip degree of freedom
800

Ratio k (3,3) over [k (3,3) for νL=1] 600

400

200

0
-50 50 150 250 350 450 550 650 750 850 950
-200

-400

-600
(a) νL

Stiffness coefficient on the diagonal of the element stiffness matrix


related to the slip degree of freedom
12
Ratio k (3,3) over [k (3,3) for νL=1]

10

0
0 0.001 0.002 0.003 0.004 0.005
-2

-4
(b) νL

Figure 3. Behaviour of stiffness coefficient k(3, 3) using the exponential approach for varying L.

for the element. In the case when


L is equal or greater than 0.05 the former formulation is
adopted, while in the case when
L is less than 0.05 the latter formulation is adopted.
The robustness of this modelling procedure is verified ensuring that the coefficients of both
stiffness matrices (the one based on the exponential and the one based on the truncated Taylor
series respectively) are identical for
L equal to 0.05 regardless of the actual value of
L
for the element. This ensures that the two matrices can be interchanged for
L equal to 0.05
(and in all applications of the method they were). It has been observed in the applications
performed that only two stiffness coefficients, viz. k(3, 3) and k(3, 2) in Equation (41), need
to be compared for
L equal to 0.05 to ensure the validity of this approach.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 667

APPLICATION

Simply supported beam

A simply supported composite beam of length 10m has been modelled using the direct stiffness
approach using one element. The cross-section  comprised of two rectangular sub-domains,
with 1 having a width of 600mm, a depth of 300mm and an elastic modulus E1 = 20kN/mm2 ,
while the corresponding values for domain 2 were 60 mm width, 300 mm depth and E2 =
200kN/mm2 . The beam was assumed be cast shored [1,2], and supports a uniformly distributed
load of 35 kN/m.
The effect of the degree of interaction has been investigated by varying the value of
L =
L given in Equation (43), with particular emphasis on the ability of the element
to overcome the locking phenomenon for large shear connection stiffnesses, which has been
demonstrated [9] to occur when L > 10. Figure 4(a) presents a plot of the slip along
the beam for a variety of dimensionless stiffnesses of the shear connection, and in partic-
ular for the value of L = 13.61 that has been worked as a practical example in Refer-
ence [17] using a stiffness of 150 000 kN/m and a connector spacing of 180 mm, so that
a value of k = 150 000/0.18 = 833 × 106 kN/m2 has been adopted in Figure 4(a). The
example in Reference [17] is based directly on the closed form solution based on New-
mark’s approach [3], and the results of this solution and those obtained using the method
of this paper are identical. The equation for the slip derived in Reference [17] has also
been used to establish the efficacy of the direct stiffness method for very high values of the
shear connection stiffness, and this is demonstrated in Figure 4(b) for L = 100 and 500.
The use of interpolation functions leads to curvature locking for the cases in Figure 4(b)
[11], but it can be seen from the figure that this difficulty does not occur for the direct
stiffness approach of this paper, whose solutions coincide with the closed form solution in
Reference [17].
An interesting consequence of partial shear connection is the variation of the depth to the
neutral axis of the sub-domains along the length of the member. Figure 5 plots the depth
from the top reference level to the neutral axes y1 and y2 of the components 1 and 2 ,
respectively, along the length of the member for the beam considered in Figure 4. For low
levels of interaction (L = 1) the beam behaves as if made of two components 1 and 2 , who
apart from having the same curvature do not interact in any way, and so the respective neutral
axes are at the mid-depth of 1 and of 2 . On the other hand, for L = 10, a condition of
partial interaction occurs and the neutral axes diverge from each other from midspan towards
the beam’s ends. For L = 100, a condition of full interaction is achieved, and the two neutral
axes coincide at the position of that of the neutral axis of the cross-section  determined in
accordance with transformed modulus theory (which assumes full interaction) throughout the
length of the beam.
Another simply supported composite beam that is 10 m long and with the previous cross-
sectional properties has been analysed subjected to a point load of 200 kN applied at mid-
span. Two elements were utilized to model the simply supported beam and the point load
was applied at the middle node. Different values of L were considered, and the results
for the slip are compared in Figure 6 with those obtained analytically with the expres-
sions presented by Newmark et al. [3]. It can be seen from Figure 6 that the agreement
is excellent.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
668 G. RANZI, M. A. BRADFORD AND B. UY

10.00 νL = 1 - Direct Stiffness Method


νL = 10 - Direct Stiffness Method
8.00
νL = 13.61 - Direct Stiffness Method
6.00 νL = 1 - based on Johnson (1994)
νL = 10 - based on Johnson (1994)
4.00 νL = 13.61 - based on Johnson (1994)
2.00
Slip (mm)

0.00
0 1 2 3 4 5 6 7 8 9 10
-2.00
-4.00
-6.00
-8.00
Coordinate along beam (m)
-10.00
(a)

0.015 νL = 100 - Direct Stiffness Method


νL = 500 - Direct Stiffness Method
0.010 νL = 100 - based on Johnson (1994)
νL = 500 - based on Johnson (1994)

0.005
Slip (mm)

0.000
0 1 2 3 4 5 6 7 8 9 10

-0.005

-0.010

Coordinate along beam (m)


-0.015
(b)

Figure 4. Slip along a simply supported beam 10 m long subjected to a uniformly distributed load.

Propped cantilever
A propped cantilever of length 10 m subjected to a uniformly distributed load of 10 kN/m
has been investigated using the direct stiffness method, with one element being used. The two
materials were rectangular, with the same dimensions and properties as those given in the
previous example. The numerical results for this statically indeterminate beam are compared
in Figure 7 with those of Ranzi et al. [18], where excellent agreement is evident between the
two.

Continuous steel–concrete composite beam


In a practical continuous composite steel–concrete tee-section beam, the portion of the member
in the vicinity of the internal support undergoes negative (hogging) bending that subjects the

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 669

Height of neutral axis from top fibre (m)


0.60 y2 νL = 1 y2 νL = 10 y2 νL = 100
y1 νL = 1 y1 νL = 10 y1 νL = 100
0.50

0.40

0.30

0.20

0.10

0.00
1 2 3 4 5 6 7 8 9
Coordinate along beam (m)

Figure 5. Location of neutral axes of top (y1 ) and bottom (y2 ) elements
of a composite beam subjected to a uniformly distributed load.

8.00 νL = 1 - Direct Stiffness Method


νL = 5 - Direct Stiffness Method
νL = 10 - Direct Stiffness Method
6.00
νL = 1 - based on Newmark, Siess & Viest (1951)
νL = 5 - based on Newmark, Siess& Viest(1951)
4.00 νL = 10 - based on Newmark, Siess & Viest (1951)

2.00
Slip (mm)

0.00
0 1 2 3 4 5 6 7 8 9 10
-2.00

-4.00

-6.00
Coordinate along beam (m)
-8.00

Figure 6. Variation of slip along a simply supported composite beam with a point load.

top material (1 ) to tensile actions, and because of the inability of the concrete slab material
to resist tension, it is cracked and its elastic modulus E1 can be taken as zero. The location
of the point of inflexion that defines the reduction in the elastic modulus of the material
1 is, however, unknown a priori, and depends on the geometry of the cross-section , the
constitutive constants E1 , E2 and k, as well as the span lengths. In using the direct stiffness
approach, it is assumed firstly that the moduli E1 and E2 are constant, and the points of
inflexion are then determined using a simple algorithm. With the hogging region being thus
determined, element nodes are placed at these points of inflexion, and the modulus of 1 (E1 )
is set as zero for the element that is in the hogging region (in the modelling procedure an
infinitesimal value approaching zero has been adopted), and the process repeated. This simple
algorithm is able to converge upon the points of inflexion after only a few iterations.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
670 G. RANZI, M. A. BRADFORD AND B. UY

1.00

0.50

0.00
Slip (mm)

0 1 2 3 4 5 6 7 8 9 10

-0.50
νL = 1 - Direct Stiffness Method
νL = 10 - Direct Stiffness Method
νL = 100 - Direct Stiffness Method
-1.00
νL = 1 - Analytical results
νL = 10 - Analytical results
νL = 100 - Analytical results
-1.50
Coordinate along beam (m)

Figure 7. Variation of slip along a propped composite cantilever with a uniformly distributed load.

The analysis of a three-span composite beam (each span of length 10 m) that is continuous
over its two internal supports, encastré at one end and simply supported at the other, has been
undertaken using the direct stiffness model. The beam supports a uniformly distributed load
of 10 kN/m. The cross-section used was the same as that in the previous examples with a
shear connection stiffness equal to 12 000 kN/m2 , and the results are shown in Figure 8. It can
be seen from this figure that the iterative technique to identify the inflexion points converges
sensibly after only four iterations.

CONCLUSIONS

This paper has presented a direct stiffness method for deriving the stiffness matrix, with six
degrees of freedom, of a composite beam with partial interaction for implementation into
line-type finite element algorithms. This matrix is not affected by the undesirable locking
phenomenon that occurs in more traditional semi-analytical finite element approaches using
pre-prescribed interpolation polynomials, and which is manifest for high values of the stiffness
of the shear connection. The simple 8-DOF line element is an example of such an element that
possesses this undesirable characteristic. It is usual to try to minimize (rather than to eliminate)
the locking phenomenon by using elements with larger numbers of degrees of freedom (10-
DOF and 16-DOF elements), but these elements only produce a gradual improvement on the
results as the degrees of freedoms increase.
While the direct stiffness technique completely eliminates the locking phenomenon, it can
produce numerical instabilities for low values of the shear connection stiffness, owing to the
manipulation of exponential terms of highly disparate magnitude in the stiffness relationships.
This difficulty was shown to be overcome adequately by recourse to a truncated Taylor series
expansion of the exponential function for the slip within the stiffness matrix for an element.
The efficacy of the direct stiffness approach was demonstrated by considering simply sup-
ported steel–concrete composite beams and propped cantilevers, for which independent solu-
tions have been obtained. It was also shown how the technique may easily handle continuous

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
DIRECT STIFFNESS ANALYSIS 671

100

Bending moment (kNm)


50

0
0 5 10 15 20 25 30
-50

-100 Iteration 1
Iteration 2
Iteration 3
Coordinate along beam (m) Iteration 4
-150
(a)

0.0020
0.0015
0.0010
0.0005
Curvature (1/m)

0.0000
-0.0005 0 5 10 15 20 25 30
-0.0010
-0.0015
-0.0020 Iteration 1
Iteration 2
-0.0025 Iteration 3
-0.0030 Iteration 4
Coordinate along beam (m)
-0.0035
(b)
0.002
0.000
0 5 10 15 20 25 30
-0.002
Deflection (m)

-0.004
-0.006
-0.008
-0.010 Iteration 1
Iteration 2
-0.012 Iteration 3
Iteration 4 Coordinate along beam (m)
-0.014
(c)
1.50

1.00

0.50
Slip (mm)

0.00
0 5 10 15 20 25 30
-0.50
Iteration 1
Iteration 2
-1.00 Iteration 3
Iteration 4 Coordinate along beam (m)
-1.50
(d)

Figure 8. Results for a continuous three span composite beam with a uniformly distributed load:
(a) bending moment; (b) curvature; (c) deflection; and (d) slip.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672
672 G. RANZI, M. A. BRADFORD AND B. UY

composite steel–concrete beams in which the concrete is cracked in tension, for which the
points of inflexion are unknown initially, by use of a simple iterative algorithm.

REFERENCES
1. Oehlers DJ, Bradford MA. Steel and Concrete Composite Structural Members: Fundamental Behaviour.
Pergamon Press: Oxford, 1995.
2. Oehlers DJ, Bradford MA. Elementary Behaviour of Composite Steel and Concrete Structural Members.
Butterworth-Heinemann: Oxford, 1999.
3. Newmark NM, Siest CP, Viest IM. Tests and analysis of composite beams with incomplete interaction.
Proceedings of the Society of Experimental Stress Analysis 1952; 9(1):75 –92.
4. Ansourian P. Experiments on continuous composite beams. Proceedings of the Institution of Civil Engineers,
London. Part 2 1981; 71:25 –51.
5. Arizumi Y, Hamada S, Kajita T. Elastic–plastic analysis of composite beams with incomplete interaction by
finite element method. Computers and Structures 1981; 14(5–6):453– 462.
6. Daniels BJ, Crisinel M. Composite slab behavior and strength analysis. Part I: Calculation procedure. Journal
of Structural Engineering (ASCE) 1993; 119(1):16 –35.
7. Dall’Asta A, Zona A. Non-linear analysis of composite beams by a displacement approach. Fifth International
Conference on Computational Structures Technology, Leuven, Belgium, Saxe-Coburg, 2000; 337–348.
8. Prathap G, Naganarayanan BP. Stress oscillations and spurious load mechanisms in variationally inconsistent
assumed strain formulations. International Journal for Numerical Methods in Engineering 1992; 33:2181–2197.
9. Dall’Asta A, Zona A. Finite elements for the analysis of composite beams with interlayer slip. XVIII Italian
Workshop on Steel (CTA)—New Directions for Steel, Venice, Italy, 2001.
10. Salari MR, Spacone E, Shing PB, Frangopol DM. Nonlinear analysis of composite beams with deformable
shear connectors. Journal of Structural Engineering (ASCE) 1998; 124(10):1148 –1158.
11. Salari MR, Spacone E. Finite element formulations of one-dimensional elements with bond-slip. Engineering
Structures 2001; 23:815 – 826.
12. Ayoub A, Filippou FC. Mixed formulation of nonlinear steel–concrete composite beam element. Journal of
Structural Engineering (ASCE) 2000; 126(3):371–381.
13. Ranzi G, Bradford MA, Uy B. A direct stiffness analysis of a composite beam with partial interaction.
UNICIV Report, School of Civil and Environmental Engineering, The University of New South Wales,
Australia, 2002.
14. Faella C, Martinelli E, Nigro E. Steel and concrete composite beams with flexible shear connection: ‘exact’
analytical expression of the stiffness matrix and applications. Computers and Structures 2002; 30:1001–1009.
15. Cosenza E, Pecce M. Structural analysis of linear-elastic continuous composite beams with partial interaction.
XVI Italian Workshop on Steel (CTA)—Steel in a Sustainable Development, Ancona, Italy, 1997.
16. Girhammar UA, Pan D. Dynamic analysis of composite members with interlayer slip. International Journal
of Solids and Structures 1993; 30(6):797– 823.
17. Johnson RP. Composite Structures of Steel and Concrete. Blackwell Scientific Publishers: Oxford, 1994.
18. Ranzi G, Bradford MA, Uy B. A general method of analysis of composite beams with partial interaction.
Steel and Composite Structures 2003; 3(3):169–184.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 61:657–672

You might also like