Sei sulla pagina 1di 15

3-D beam element of composite cross section including warping

and shear deformation eects


E.J. Sapountzakis
*
, V.G. Mokos
Institute of Structural Analysis, School of Civil Engineering, National Technical University of Athens, Zografou Campus, GR 157 80 Athens, Greece
Available online 27 October 2006
Abstract
In this paper a boundary element method is developed for the construction of the 14 14 stiness matrix and the nodal load vector of
a member of arbitrary homogeneous or composite cross section taking into account both warping and shear deformation eects. The
composite member consists of materials in contact each of which can surround a nite number of inclusions. To account for shear defor-
mations, the concept of shear deformation coecients is used. In this investigation the denition of these factors is accomplished using a
strain energy approach. Seven boundary value problems are formulated and solved employing a pure BEM approach, that is only
boundary discretization is used. The evaluation of the shear deformation coecients is accomplished from stress functions using only
boundary integration. Numerical results are presented to illustrate the method and demonstrate its eciency and accuracy. The inuence
of the warping eect especially in composite members of open form cross section is analyzed through examples demonstrating the impor-
tance of the inclusion of the warping degrees of freedom in the analysis of a space frame. Moreover, the discrepancy of both the deec-
tions and the internal forces of a member of a spatial structure arising from the ignorance of the shear deformation eect necessitates the
inclusion of this additional eect, especially in thick walled cross section members.
2006 Elsevier Ltd. All rights reserved.
Keywords: Nonuniform torsion; Composite; Warping; Bar; Beam; Twist; Boundary element method; Stiness matrix; Shear deformation
1. Introduction
One of the problems often encountered in engineering
practice is the analysis of rectilinear or curved members
of structures of composite or homogeneous cross section
subjected to twisting moments. Accurate analysis of curved
box shaped bridges or composite concrete slab stiened by
steel beams is dicult to be achieved for two reasons.
According to the rst reason, generally commercial pro-
grams consider six degrees of freedom at each node of a
member of a space frame, ignoring in this way the warping
eects due to the corresponding restraint at the ends of the
member [13]. If the aforementioned structures are ana-
lyzed or designed for torsion considering only the eect
of Saint Venant torsion resistance, the analysis may under-
estimate the torsion in the members and the design may be
unconservative. Several researchers tried to overcome this
inaccuracy by developing a 14 14 member stiness matrix
including warping degrees of freedom at the ends of a
member with open thin-walled homogeneous cross section
and assuming simple [47] or more complicated torsional
boundary conditions [8,9].
According to the second reason, the aforementioned
commercial programs ignore shear deformations due to
the fact that they are unable to compute shear correction
factors. Though these deformations are quite small in most
civil engineering applications, they may be dominant in
some situations, where bending moments are small com-
pared to shear forces acting on the member. This is nor-
mally true in short span beams or in structural systems
such as curved box girder bridges.
In this paper a boundary element method (BEM) is
developed for the construction of the 14 14 stiness
matrix and the nodal load vector of a member of arbitrary
0045-7949/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2006.09.003
*
Corresponding author.
E-mail addresses: cvsapoun@central.ntua.gr (E.J. Sapountzakis),
vgmokos@central.ntua.gr (V.G. Mokos).
www.elsevier.com/locate/compstruc
Computers and Structures 85 (2007) 102116
homogeneous or composite cross section taking into
account both warping and shear deformation eects. The
composite member consists of materials in contact each
of which can surround a nite number of inclusions. To
account for shear deformations, the concept of shear defor-
mation coecients is used. There are several denitions of
these factors. In this investigation the approach of Bach
[10] and Stojek [11] is employed using a strain energy
approach, instead of Timoshenkos [12] and Cowpers
[13] denitions, for which several authors [14,15] have
pointed out that one obtains unsatisfactory results or de-
nitions given by other researchers [16,17], for which these
factors take negative values. Seven boundary value prob-
lems with respect to the variable along the bar angle of
twist, to the primary warping function, to a ctitious func-
tion, to two stress functions and to the beam transverse dis-
placements are formulated and solved employing a pure
BEM approach, that is only boundary discretization is
used. Numerical results are presented to illustrate the
method and demonstrate its eciency and accuracy. The
inuence of the warping eect especially in composite
members of open form cross section is analyzed through
examples demonstrating the importance of the inclusion
of the warping degrees of freedom in the analysis of a space
frame. Moreover, the discrepancy of both the deections
and the internal forces of a member of a spatial structure
arising from the ignorance of the shear deformation eect
necessitates the inclusion of this additional eect, especially
in thick walled cross section members.
2. Stiness matrix and nodal load vector formulation
Consider a prismatic element of length L with an arbi-
trarily shaped composite cross section consisting of materi-
als in contact, each of which can surround a nite number
of inclusions, with modulus of elasticity E
j
and shear mod-
ulus G
j
, occupying the regions X
j
(j = 1, 2, . . . , K) of the y; z
plane (Fig. 1). The materials of these regions are assumed
homogeneous, isotropic and linearly elastic. Let also the
boundaries of the nonintersecting regions X
j
be denoted
by C
j
(j = 1, 2, . . . , K.). These boundary curves are piecewise
smooth, i.e. they may have a nite number of corners.
Fig. 1. Prismatic beam (a) with a composite cross section of arbitrary shape occupying the two dimensional region X (b) C~y~z is the beam principal
centroidal system of axes, while Myz is the corresponding one through the shear center M.
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 103
In order to include the warping behaviour in the study
of the aforementioned element in each node at the element
ends a seventh degree of freedom is added to the well
known six DOFs of the classical three-dimensional frame
element. The additional DOF is the rst derivative of the
angle of twist h
0
x
dh
x
=dx denoting the rate of change of
the angle of twist h
x
, which can be regarded as the torsional
curvature (Fig. 2) of the cross section. Thus, the nodal dis-
placement vector in the local coordinate system, as shown
in Fig. 1, can be written as
fD
i
g
T
fu
~xj
u
~yj
u
~zj
h
xj
h
~yj
h
~zj
h
0
xj
u
~xk
u
~yk
u
~zk
h
xk
h
~yk
h
~zk
h
0
xk
g
1
and the respective nodal load vector as
fF
i
g
T
fN
j
Q
~yj
Q
~zj
M
tj
M
~yj
M
~zj
M
wj
N
k
Q
~yk
Q
~zk
M
tk
M
~yk
M
~zk
M
wk
g 2
where M
t
is the twisting moment at the ends of the element
given as
M
t
M
P
t
M
S
t
3
and M
w
the warping moment (bimoment) due to the tor-
sional curvature at the same sections given as
M
w
E
1
C
M
d
2
h
x
dx
2
4
In Eq. (3) M
P
t
is the primary twisting moment dened as the
resultant of the primary shear stress distribution and M
S
t
is
the secondary twisting moment dened as the resultant of
the secondary shear stress distribution due to warping
given as [18]
M
P
t
G
1
I
t
dh
x
dx
5a
M
S
t
E
1
C
M
d
3
h
x
dx
3
5b
where
C
M

K
j1
E
j
E
1
_
Xj
u
P
M

2
j
dX
j
6
I
t

K
j1
G
j
G
1
_
Xj
z
2
y
2
z
ou
P
M
oy
_ _
j
y
ou
P
M
oz
_ _
j
_ _
dX
j
7
are the warping and torsion constants of the composite cross
section, respectively and u
P
M
y; z
j
is the primary warping
function in the domain X with respect to the center of twist
M of the cross section of the bar (see Fig. 1), evaluated from
the solution of the following Neumann problem [18]:
r
2
u
P
M

j
0 in X
j
j 1; 2; . . . ; K 8
G
j
ou
P
M
on
_ _
j
G
i
ou
P
M
on
_ _
i

1
2
G
j
G
i

oq
2
M

os
_ _
j
on C
j
j 1; 2; . . . ; K 9
where r
2

j
o
2
=oy
2

j
o
2
=oz
2

j
is the Laplace opera-
tor; X [
K
j1
X
j
denotes the whole region of the composite
cross section; j and i indices in the boundary conditions de-
note jth and ith materials, respectively that have common
boundary C
j
(G
j
= 0 at a free boundary); q
M

y
2
z
2
_
is the distance of a point on the boundary C
j
from the cen-
ter of twist M; (o/on)
j
denotes the directional derivative
normal to the boundary C
j
and (o/os)
j
denotes dierentia-
tion with respect to its arc length s. The vector n normal
to the boundary C
j
is positive if it points to the exterior
of the X
j
region. It is worth here noting that the normal
derivatives across the interior boundaries vary discon-
tinuously.
It is worthnoting that in the case the origin O of the
coordinates is a point of the y; z plane other than the center
of twist, the warping function with respect to this point
u
P
O
y; z
j
is rst established from the Neumann problem
(8) and (9) substituting u
P
M

j
by u
P
O

j
. Using the evaluated
warping function u
P
O

j
; u
P
M

j
, is then established using the
transformation given by the following equation [19]:
u
P
M
y; z
j
u
P
O
y; z
j
zy
M
yz
M
c
p
10
where y y y
M
; z z z
M
; y
M
; z
M
are the coordinates of
the center of twist M with respect to Oyz system of coordi-
nates (see Fig. 1) and c
P
an integration constant.
Substituting Eqs. (5a and b) into Eq. (3) we obtain the
expression of the total twisting moment for the composite
element with constant cross section as
M
t
G
1
I
t
h
0
X
E
1
C
M
h
000
x
11
The nodal displacement and load vectors given in Eqs. (1)
and (2) are related with the 14 14 local stiness matrix of
the spatial beam element written as
Fig. 2. Torsional curvature of a rectangular and a hollow square cross section.
104 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
where the k
i
Tn
n 1; 2; 3; 4; 5; 6 coecients are given as
[4,20]
k
i
T1
k
0
kS; k
i
T2
k
i
T5
k
0
C 1 13a; b
k
i
T3
k
i
T6
k
0
CL
1
k
S
_ _
; k
i
T4
k
0
1
k
S L
_ _
13c; d
where
k

G
1
I
t
E
1
C
M
_
; S sinh e C cosh e 14a; b; c
e Lk; k
0

G
1
I
t
21 C eS
14d; e
According to the nodal load vector, assuming that the
beam member is subjected to the arbitrarily concentrated
or distributed twisting moment m
t
= m
t
(x), the evaluation
of the twisting and bending (due to the torsional curvature)
moment elements of the vector is accomplished using again
relations (4) and (11) employing the derivatives of the angle
of twist h
x
, obtained from the solution of the following
boundary value problem:
E
1
C
M
h
0000
x
G
1
I
t
h
00
x
m
t
inside the member 15
c
x1
h
x
c
x2
M
t
c
x3
16a
d
x1
h
0
x
d
x2
M
w
d
x3
at the member ends x 0; L 16b
for appropriate values of the c
xi
,d
xi
(i = 1, 2, 3) functions
and more specically for c
x1
= d
x1
= 1, c
x2
= c
x3
= d
x2
=
d
x3
= 0 at x = 0,L. It is worth here noting that Eqs.
(16a,b) give the most general linear torsional boundary
conditions, including also the elastic support. It is apparent
that all types of the conventional torsional boundary con-
ditions (clamped, simply supported, free or guided edge)
can be derived from these equations by specifying appro-
priately the functions c
xi
and d
xi
.
The k
i
lm
coecients (l,m = 1, 2, 3, 5, 6, 8, 9, 10, 12, 13) of
the matrix of Eq. (12) either come from the well known
12 12 stiness matrix of the classical three-dimensional
frame element, following in this way the classical theory
of beams or result from the so called Timoshenko theory
of beams, which takes into account the eect of shear
deformation. In this investigation the latter theory is
employed leading to the following systems of dierential
equations concerning beam bending with respect to ~y and
~z axes, respectively
d
d~x
u
~z
h
~y
M
~y
Q
~z
_

_
_

_

0 1 0 1=G
1
A
~z
0 0 1=E
1
I
~y
0
0 0 0 1
0 0 0 0
_

_
_

_
u
~z
h
~y
M
~y
Q
~z
_

_
_

0
0
0
p
~z
_

_
_

_
17a
d
d~x
u
~y
h
~z
M
~z
Q
~y
_

_
_

_

0 1 0 1=G
1
A
~y
0 0 1=E
1
I
~z
0
0 0 0 1
0 0 0 0
_

_
_

_
u
~y
h
~z
M
~z
Q
~y
_

_
_

0
0
0
p
~y
_

_
_

_
17b
since the corresponding beam components of translation u
~z
and u
~y
except for the bending part take into account the
contribution of the shear deformation. Thus, in Eqs.
(17a,b) u
~z
, u
~y
are the components of total translation due
to bending and shear, h
~z
, h
~y
are the components of rotation
due to bending and p
~z
p
~z
~x; p
~y
p
~y
~x denote the arbi-
trarily distributed transverse loading with respect to ~z and
~y axes (see Fig. 1), respectively. Moreover, I
~y
; I
~z
are the
bending moments of inertia of the composite cross section
with respect to ~y and ~z axes, respectively dened as
I
~y

K
j1
E
j
E
1
_
Xj
~z
2
dX
j
; I
~z

K
j1
E
j
E
1
_
Xj
~y
2
dX
j
18a; b
The aforementioned contribution of shear deformation
according to Timoshenkos theory is taken into account
12
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 105
through the denitions of the shear areas with respect to ~y
and ~z axes given as
A
~y
j
~y
A
1
a
~y
A; A
~z
j
~z
A
1
a
~z
A 19a; b
where j
~y
; j
~z
are the shear correction factors, a
~y
; a
~z
are the
shear deformation coecients and A is the cross section
area given as
A

K
j1
G
j
G
1
_
Xj
dX
j
20
The shear deformation coecients a
~y
; a
~z
corresponding to
the principal centroidal system of axes C~y~z are evaluated
equating the approximate formula of the shear strain
energy per unit length [15]
U
appr

a
~y
Q
2
~y
2AG
1

a
~z
Q
2
~z
2AG
1
21
with the exact one given from
U
exact

k
j1
E
1
E
j
_
Xj
s
~x~z

2
j
s
~x~y

2
2G
l
dX
j
22
and are obtained as [21]
a
~y

1
j
~y

A
E
1
D
2

K
j1
_
Xj
E
j
rH
j
e rH
j
e dX
j
23a
a
~z

1
j
~z

A
E
1
D
2

K
j1
_
Xj
E
j
rH
j
d rU
j
d dX
j
23b
where ($)
j
i
y
(o/oy) + i
z
(o/oz) is a symbolic vector with
i
~y
; i
~z
the unit vectors along ~y and ~z axes, respectively, D is
given from
D 21 mI
~y
I
~z
24
m is the Poisson ratio of the cross section materials, which
for the torsionless bending problem is assumed to be com-
mon, e and d are vectors dened as
e mI
~y
~y
2
~z
2
2
_ _
i
~y
mI
~y
~y~zi
~z
25a
d mI
~z
~y~zi
~y
mI
~z
~z
2
~y
2
2
_ _
i
~z
25b
and H~y; ~z
j
; U~y~z
j
are stress functions, which are eval-
uated from the solution of the following Neumann type
boundary value problems [21]
r
2
H
j
2I
~y
~y in X
j
j 1; 2; . . . K 26a
E
j
oH
on
_ _
j
E
i
oH
on
_ _
i
E
j
E
i
n e on C
j
j 1; 2; . . . ; K
26b
r
2
U
j
2I
~z
~z in X
j
j 1; 2; . . . ; K 27a
E
j
oU
on
_ _
j
E
i
oU
on
_ _
i
E
j
E
i
n d on C
j
j 1; 2; . . . ; K
27b
where n is the outward normal vector to the boundary C
j
(E
i
= 0 at a free boundary). In the case of negligible shear
deformations a
~z
a
~y
0. It is also worth here noting that
the boundary conditions (9), (26b) and (27b) have been de-
rived from the following physical considerations:
The traction vector in the direction of the normal vector
n vanishes on the free surface of the beam.
The traction vectors in the direction of the normal vec-
tor n on the interfaces separating the j and i dierent
materials are equal in magnitude and opposite in
direction.
The displacement components remain continuous across
the interfaces, since it is assumed that the materials are
rmly bonded together.
Thus, taking into account the eect of shear deforma-
tion using Eqs. (17a,b), the K
i
lm
coecients (l, m = 1, 2,
3, 5, 6, 8, 9, 10, 12, 13) of the stiness matrix of Eq. (12) can
be written as
k
i
33
k
i
10;3
k
i
3;10
k
i
10;10
c
~z
12E
1
I
~y
L
3
28a
k
i
22
k
i
92
k
i
29
k
i
99
c
~y
12E
1
I
~z
L
3
28b
k
i
10;5
k
i
5;10
k
i
12;10
k
i
10;12
k
i
53
k
i
35
k
i
12;3
k
i
3;12
c
~z
6E
1
I
~y
L
2
28c
k
i
26
k
i
62
k
i
2;13
k
i
13;2
k
i
69
k
i
96
k
i
9;13
k
i
13;9
c
~y
6E
1
I
~z
L
2
28d
k
i
55
k
i
12;12
g
~z
4E
1
I
~y
L
; k
i
66
k
i
13;13
g
~y
4E
1
I
~z
L
28e; f
k
i
12;5
k
i
5;12
w
~z
2E
1
I
~y
L
; k
i
13;6
k
i
6;13
w
~y
2E
1
I
~z
L
28g; h
where
c
~z

1
1 12l
~z
; c
~y

1
1 12l
~y
29a; b
g
~z

1 3l
~z
1 12l
~z
; g
~y

1 3l
~y
1 12l
~y
30a; b
w
~z

1 6l
~z
1 12l
~z
; w
~y

1 6l
~y
1 12l
~y
31a; b
and
l
~z
a
~z
E
1
I
~y
G
1
A
1
L
2
; l
~y
a
~y
E
1
I
~z
G
1
A
1
L
2
32a; b
while the coecients K
i
lm
(l, m = 1, 8) come from the classi-
cal three-dimensional frame element as
106 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
k
i
11
k
i
81
k
i
18
k
i
88

E
1
A
~x
L
33
where
A
~x

K
j1
E
j
E
1
_
Xj
dX
j
34
It is worthnoting that the reduction of Eqs. (6), (18a,b),
(34) using the modulus of elasticity E
1
and of Eqs. (7) and
(20) using the shear modulus G
1
of the rst material, could
be achieved using any other material, considering it as ref-
erence material.
According to the nodal load vector, assuming that the
span of the beam is subjected to the arbitrarily concen-
trated or distributed transverse loading p
~z
p
~z
~x and
p
~y
p
~y
~x, the evaluation of the elements concerning the
shear forces and bending moments is accomplished using
again the dierential Eqs. (17a,b) employing the derivatives
of the deections u
~z
and u
~y
, obtained from the solution of
the following boundary value problems for the two cases
of transverse loading
E
1
I
~y
u
0000
~z
p
~z
a
~z
E
1
I
~y
G
1
A
p
00
~z
inside the beam 35
c
~z1
u
~z
c
~z2
Q
~z
c
~z3
36a
d
~z1
h
~y
d
~z2
M
~y
d
~z3
at the beam ends x 0; L 36b
E
1
I
~z
u
0000
~y
p
~y
a
~y
E
1
I
~z
G
1
A
p
00
~y
inside the beam 37
c
~y1
u
~y
c
~y2
Q
~y
c
~y3
38a
d
~y
1
h~z d
~y2
M~z d
~y3
at the beam ends x 0; L 38b
for appropriate values of the c
~zi
; c
~yi
and d
~zi
; d
~yi
i 1; 2; 3
functions and more specically, for c
~z1
d
~z1
1,
c
~y1
d
~y1
1 and c
~z2
c
~z3
d
~z2
d
~z3
0, c
~y2
c
~y3

d
~y2
d
~y3
0 at x = 0 and x = L. It is worth noting that
relations (36a,b) and (38a,b) describe the most general lin-
ear boundary conditions associated with the problem at
hand including elastic support or restraint.
3. Integral representations numerical solution
3.1. The angle of twist h
x
and the deections u
~z
and u
~y
The evaluation of the angle of twist h
x
is accomplished
using BEM [22] as this is presented in Sapountzakis and
Mokos [23]. According to this formulation the integral rep-
resentations for the angle of twist and its second derivative
with respect to x
h
x
x
G
1
I
t
E
1
C
M
_
L
0
d
2
h
x
dx
2
K
4
rdz
1
E
1
C
M
_
L
0
m
t
xK
4
rdx
K
4
r
d
3
h
x
dx
3
K
3
r
d
2
h
x
dx
2
K
2
r
dh
x
dx
K
1
rh
x
_ _L
0
39a
d
2
h
x
x
dx
2

G
1
I
t
E
1
C
M
_
L
0
d
2
h
x
x
dx
2
K
2
rdz
1
E
1
C
M
_
L
0
m
t
xK
2
rdx
K
2
r
d
3
h
x
x
dx
3
K
1
r
d
2
h
x
x
dx
2
_ _L
0
39b
written for the boundary points of the beam x = 0,L to-
gether with the two boundary conditions (16a,b) and Eq.
(39b) constitute a system of ve simultaneous linear equa-
tions. These equations, using constant elements to approx-
imate the line integrals and a collocation technique, can be
solved to yield d
2
h
x
/dx
2
inside the beam and h
x
, dh
x
/dx,
d
2
h
x
/dx
2
, d
3
h
x
/dx
3
at the beam ends. It is worth here noting
that the kernels K
i
(r), (i = 1, 2, 3, 4) in Eqs. (39a,b) are given
as
K
1
r
1
2
sgnq 40a
K
2
r
1
2
L1 q 40b
K
3
r
1
4
L
2
jqjjqj 2sgnq 40c
K
4
r
1
12
L
3
2 jqj
3
3jqj
2
40d
with q = r/I and sgnq = 1, 1 or undened depending on
whether q is positive, negative or 0, respectively.
Subsequently, using the discretized form of Eq. (39a),
the angle of twist h
x
at any interior point of the beam is
computed as
h
x
x fC
x
gfH
00
x
g
ffA
x1
g fA
x2
g fA
x3
g fA
x4
gg ffh
x
g fh
0
x
g fh
00
x
g fh
000
x
gg
T
B
x
41
where {C
x
} is a 1 N known coecient row matrix origi-
nating from the integration of the kernels along the beam
(N is the number of nodal points along the beam axis),
fH
00
x
g is a N 1 column matrix including the values of
the second derivative of the angle of twist h
x
inside the
beam, {A
xi
} (i = 1, 2, 3, 4) are 1 2 known coecient row
matrices originating from the values of the kernels at the
beam ends, fh
x
g; fh x
0
gfh
00
x
gfh
000
x
g are 2 1 column matri-
ces including the values of the angle of twist h
x
and its
derivatives at the beam ends and B
x
is a known coecient
arising from the twisting loading of the beam m
t
.
The same numerical procedure can also be used for the
evaluation of the deections u
~z
and u
~y
.
3.2. The primary warping function u
P
M

j
The integral representations and the numerical solution
for the evaluation of the angle of twist h
x
assume that the
warping C
M
and torsion I
t
constants given from Eqs. (6)
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 107
and (7), respectively are already established. Eqs. (6) and
(7) indicate that the evaluation of the aforementioned con-
stants presumes that the primary warping function u
P
M

j
at
any interior point of the domain X of the cross section of
the bar is known. Once u
P
M

j
is established, C
M
and I
t
con-
stants are evaluated by converting the domain integrals
into line integrals along the boundary using the following
relations:
C
F
y
z
x
P=40kN
e=b/4
P=40kN
e=b/4
L=15.0m
4.0m
3.5m
3.5m
2.0m
2.0m
D
E
y
z
Steel: E
S
Concrete: E
C
b=0.30m
t=0.02m
t=0.02m
h
=
0
.
3
0
m
e=b/4
P=40kN
y
z
b=0.30m
t=0.02m
t=0.02m
h
=
0
.
3
0
m
e=b/4
P=40kN
Steel: E
S
A
Fig. 3. View (a) and transverse sections of the rst, third (b) and of the second (c) span of the composite structure of Example 1.
Table 1
Cross section rigidities of the composite structure of Example 1
Cross section E
S
A
(kN)
G
S
A
~z
(kN)
E
S
I
~y
(kNm
2
)
G
S
I
t
(kNm
2
)
E
S
C
M
(kNm
4
)
Homogeneous 3,612,000 446,822 55627.6 206.815 368.606
Composite 5,796,000 1,464,960 67930.8 24107.4 108.124
Fig. 4. Structural model and loading of the composite structure of Example 1 (supported nodes 1, 9, 23, 31).
108 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
C
M

1
E
1

K
j1
_
Cj
B E
j
ou
P
M
on j
_ _
E
i
ou
P
M
on
_ _
i
_ _
ds
on C
j
j 1; 2; . . . ; K 42a
I
t

1
G
1

K
j1
_
Cj
G
j
G
i
zy
2
yu
P
M

j
_ _
cos b
_
z
2
y zu
P
M

j
_ _
sin b
_
ds on C
j
j 1; 2; . . . ; K
42b
Eqs. (42a,b) C
j
(j = 1, 2, . . . , K), is an interface between re-
gions X
j
and X
i
, while E
j
= G
j
= 0 in the case C
j
is a free
boundary. Moreover, in these equations the normal n to
the boundary C
j
points to the exterior of the region X
j
and C
j
is traveled only once, while (B(y, z))
j
is a ctitious
function dened as the solution of the following Neumann
problem:
r
2
B
j
u
P
M

j
in X
j
j 1; 2; . . . ; K 43
E
j
oB
on
_ _
E
i
oB
on
_ _
0 on C
j
j 1; 2; . . . ; K 44
Finally, Eq. (10) can give the primary warping function
u
P
M

j
with respect to the center of twist M, if the warping
function u
P
O

j
with respect to the origin of the coordinates
is rst established. The evaluation of the warping function
u
P
O

j
and the ctitious function (B(y,z))
j
is accomplished
using BEM as this is presented in Sapountzakis [24] and
in Sapountzakis and Mokos [18], respectively.
3.3. The stress functions (H(y,z))
j
and (U(y,z))
j
The evaluation of the stress functions (H(y,z))
j
and
(U(y,z))
j
is accomplished using BEM as this is presented
in Mokos and Sapountzakis [21].
Moreover, since the torsionless bending problem of
beams is solved by the BEM, the domain integrals in Eqs.
(18a,b), (20) and (23a,b) have to be converted to boundary
line integrals, in order to maintain the pure boundary char-
acter of the method. This can be achieved using integration
by parts, the Gauss theorem and the Green identity. Thus,
for the bending moments of inertia and the cross section
area Eqs. (18a,b), (20) can be written as
I
~y

1
E
1

K
j1
_
Cj
E
j
E
i
~y~z
2
cos b ds 45a
I
~z

1
E
1

K
j1
_
C
j
E
j
E
i
~z~y
2
sin b ds 45b
A
1
2G
1

K
j1
_
Cj
G
j
G
i
~y cos b ~z sin b ds 45c
while the shear deformation coecients a
~y
and a
~z
are ob-
tained from the relations
a
~y

A
E
1
D
2
4v 2I
~y
I
H~y

1
4
v
2
I
2
~y
I
ed
I
He
_ _
46a
a
~z

A
E
1
D
2
4v 2I
~z
I
H~z

1
4
v
2
I
2
~z
I
ed
I
Ud
_ _
46b
where
I
He

K
j1
_
Cj
E
j
E
i
H
j
n e ds 47a
I
Ud

K
j1
_
Cj
E
j
E
i
H
j
n d ds 47b
I
ed

K
j1
_
Cj
E
j
E
i
~y
4
~z sin b ~z
4
~y cos b
2
3
~y
2
~z
3
sin b
_ _
ds
47c
I
H~y

1
6

K
j1
_
Cj
E
j
E
i
2I
~y
~y
4
~z sin b
3H
j
cos b ~yn e~y
2
ds 47d
I
U~z

1
6

K
j1
_
Cj
E
j
E
i
2I
~z
~z
4
~y cos b
3U
j
sin b ~zn d~z
2
ds 47e
-0.022
-0.016
-0.010
-0.004
0.002
0.008
0.014
0.020
-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
P 2
M
max 0.02197m =
-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
-0.014
-0.010
-0.006
-0.002
0.002
0.006
0.010
P 2
M
max 0.01226m =
Fig. 5. Distribution of the primary warping function u
P
M
of the homo-
geneous (a) and the composite (b) cross section of Example 1.
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 109
4. Numerical examples
On the basis of the analytical and numerical procedures
presented in the previous sections, a computer program has
been written and representative examples have been studied
to demonstrate the eciency, wherever possible the accu-
racy and the range of applications of the developed
method. The appropriate number of beam elements results
as the minimum one able to describe accurately the geom-
etry and loading of the structure (the reason for the
employed beam elements in examples 1 and 3 is the clear
representation of the deection lines and moment diagrams
in these examples). In all the examples treated 300 constant
boundary elements were employed for the discretization of
the boundary of the cross section.
Example 1. In order to demonstrate the eciency and the
range of applications of the developed method in our rst
example a three span of L = 4m + 7m + 4m = 15m com-
posite beam resting on four simply supported torsional-
bending boundary conditions and subjected in eccentric
loading, as shown in Fig. 3 has been studied. The cross
section of the midspan of the structure is a homogeneous
one consisting of an I-steel (E
S
= 2.1 10
8
kPa, v = 0.20)
beam, while the cross section of the rst and the third spans
is a composite one consisting of the aforementioned I-steel
beam (reference material) lled with concrete (E
C
=
3.0 10
7
kPa, v = 0.20). In Table 1 the rigidities of the
homogeneous and the composite cross sections and in
Fig. 4 the structural model and the loading of the
composite structure are presented. In Fig. 5 the contour
lines of the primary warping function u
P
M
in the interior of
both the homogeneous and the composite cross sections
are presented. From this gure it follows that warping is
not constant along the thickness of the cross section walls
as it is assumed in Thin Tube Theory for thin walled
beams. Moreover, in Fig. 6 the twisting moment diagrams
of the composite beam and in Table 2 the support reactions
including (14 14) or ignoring (12 12) the torsional
Fig. 6. Twisting moment diagrams of the composite structure of Example 1, (a) including (14 14) or (b) ignoring (12 12) the torsional curvature of the
members.
Table 2
Support reactions of the composite beam of Example 1
Support 14 14 & 12 12 member
stiness matrix
Including & Ignoring shear
deformation
Including
shear
deformation
Ignoring
shear
deformation
14 14
member
stiness
matrix
12 12
member
stiness
matrix
Reaction R
z
(kN) Reaction R
xx
(kNm)
A 0.5768E+01 0.5748E+01 0.3708E+00 0.0000E+00
B 0.2374E+02 0.2368E+02 0.1874E+01 0.1500E+01
D 0.5135E+02 0.5145E+02 0.3623E+00 0.0000E+00
F 0.1068E+02 0.1063E+02 0.1865E+01 0.1500E+01
110 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
curvature of the members are presented. From the
obtained results the signicant role of the torsional
curvature is remarkable, while in this case the shear
deformation effect does not have any inuence since shear
and torsion in this example are independent.
Example 2. A composite beam of length L = 80.0m, with
curved plan view of a quadratic parabola (Fig. 7a),
clamped at its end points A, E and inhibiting the deection
at points B, D, having a cross section consisting of a con-
crete (E
C
= 3.0 10
7
kPa, m = 0.20) rectangular part (refer-
ence material) stiffened by steel beams (E
S
= 2.1 10
8
kPa,
m = 0.20), forming either a closed box shaped section
(Fig. 7b) or an open P-section (Fig. 7c), subjected to a ver-
tical load P = 600kN at point C (Fig. 7a) has been studied.
In Table 3 the constants of the closed and the open cross
sections and in Fig. 8 the structural model and the loading
of the composite structure are presented. in Fig. 9 the
boundary distributions of the primary warping function
u
P
M
of the composite beam with the closed cross section
are presented as compared with those of the beam with
the open cross section. The advantages of the closed cross
section beam are easily veried. Moreover, in Table 4 the
maximum deections of the composite beam and in Figs.
10 and 11 the deections along the composite beam of
the closed and open form cross section, respectively are
presented as compared with those obtained taking into
account or ignoring both the shear deformation effect
and the torsional curvature. From the obtained results,
both the negligible inuence in the closed cross section
and the signicant role in the open one of the torsional cur-
vature and the remarkable inuence of the shear deforma-
tion effect in all cases are pointed out. In Fig. 12 the
bending moment diagrams of the composite beam with
open cross section, taking into account the shear deforma-
tion effect and including (14 14) or ignoring (12 12) the
torsional curvature of the members and in Fig. 13 the
bending moment diagram due to the torsional curvature
L=80.00m
Quadratic parabola
A E
h1=20.00m
2
1
, 0,1
( 4 4 )
x L
y h
=
= +
.
String length LDE: 91.82m
C
P=600kN
.
20.0m 20.0m
.
B
D
.
.
15.00m
x
y
15.00m
A B
t
t
0.25 0.25
0.75 0.75 0.75 0.75
2.40
5.50m
1.50
0.20
3.00
1
.
4
9
4
5
m
7.00m
Steel: ES
Concrete: EC
t=0.05m
C: Centroid
M: Center of twist
0.2814m
A B
t
0.25 0.25
0.75 0.75 0.75 0.75
2.40
5.50m
1.50
0.20
3.00
1
.
7
4
6
4
m
7.00m
Steel: ES
Concrete: EC
t=0.05m
C: Centroid
M: Center of twist
2.6016m
A B


Fig. 7. Plan view (a) and transverse sections of the closed (b) and open (c)
beam of Example 2.
Table 3
Cross section constants of the composite structure of Example 2
Cross section A (m
2
) A
~z
(m
2
) I
~y
(m
4
) I
t
(m
4
) C
M
(m
6
)
Closed 5.74 1.8691 10.0374 12.7531 1.93659
Open 4.90 1.9646 7.9123 0.60625 33.4857
Fig. 8. Structural model and loading of the composite structure of Example 2 (clamped nodes 1, 61, nodes with internal support 16, 46, loaded node 31).
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 111
are presented. It is worth here noting the magnitude of the
discrepancy (22.5%) of the maximum normal stress at sec-
tion C of the composite beam with open cross section tak-
ing into account the shear deformation effect and including
(14 14) (maxr
x
= 16.86 10
3
kpa) or ignoring (12 12)
(maxr
x
= 13.07 10
4
kpa) the torsional curvature of the
members. This discrepancy necessitates the inclusion of this
additional degree of freedom, especially in beams with
open form cross sections.
Fig. 9. Distribution of the primary warping function u
P
M
of the closed (a) and open (b) cross section of Example 2.
Table 4
Maximum deections of the composite beam of Example 2
Beams 14 14 member stiness
matrix
12 12 member stiness
matrix
Including
shear
deformation
Ignoring
shear
deformation
Including
shear
deformation
Ignoring
shear
deformation
Close 0.1822 0.1464 0.1823 0.1465
Open 0.2640 0.2272 0.3600 0.3207
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
Beam length (m)
-0.20
-0.18
-0.15
-0.13
-0.10
-0.08
-0.05
-0.03
0.00
0.03
D
e
f
l
e
c
t
i
o
n

(
c
m
)
Closed Cross section (14x14 & 12x12 Matrix)
Without Shear Deformation
Including Shear Deformation
Fig. 10. Deections along the composite beam with closed cross section of
Example 2.
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
Beam length (m)
-0.38
-0.35
-0.32
-0.30
-0.27
-0.25
-0.22
-0.20
-0.17
-0.15
-0.12
-0.10
-0.07
-0.05
-0.02
0.00
0.03
0.05
D
e
f
l
e
c
t
i
o
n

(
c
m
)
Open Cross section
Without Shear Deformation (12x12 Matrix)
Without Shear Deformation (14x14 Matrix)
Including Shear Deformation (12x12 Matrix)
Including Shear Deformation (14x14 Matrix)
Fig. 11. Deections along the composite beam with open cross section of
Example 2.
112 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
Example 3. A clamped slab and beam structure of length
L = 20.0m of a composite cross section consisting of three
concrete C35/45-I beams (E
c35
= 3.35 10
7
kpa, m = 0.20)
joined on top with a concrete C20/25 (reference material)
slab (E
c20
= 2.90 10
7
kpa, m = 0.20), at their ends with a
transverse concrete C20/25 beam of thickness 0.60m and
at their midspan with a transverse concrete C20/25 beam
of thickness 0.30m, as shown in Fig. 14 has been studied.
In Table 5 the constants of the transverse and the longitu-
dinal cross sections and in Fig. 15 the structural model and
Fig. 12. Bending moment diagrams of the composite beam of Example 2 with open cross section (a) including (14 14) or (b) ignoring (12 12) the
torsional curvature.
Fig. 13. Bending moment diagram of the composite beam of Example 2 with open cross section due to the torsional curvature.
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 113
the loading of the composite structure are presented. It is
worth here noting that the constants of the members of
the grid are estimated as the corresponding part of the
whole cross section taking into account the ratio of the
area of the member with respect to the area of the whole
cross section. In Figs. 16 and 17 the deection surfaces
and the bending moment diagrams together with their
maximum values are presented, respectively taking into
account the shear deformation effect and including
(14 14) or ignoring (12 12) the torsional curvature of
the grid members. Moreover, in Table 6 the maximum
deections of the longitudinal beams I, II, III (Fig. 15)
are presented as compared with those obtained taking into
account or ignoring both the shear deformation effect and
the torsional curvature. From the obtained results both the
signicant role of the torsional curvature and the inuence
of the shear deformation effect are remarkable.
5. Concluding remarks
In this paper a boundary element method is developed
for the construction of the 14 14 stiness matrix and
the nodal load vector of a member of arbitrary homoge-
neous or composite cross section taking into account both
warping and shear deformation eects. The composite
member consists of materials in contact each of which
1.20m
0.10m
0.05m
0.10m
0.15m
0.20m
0.60m
0.50m
0.20m
0.10m
C: Centroid
M: Center of twist
0.5926.m
0.8192.m
3.60m
1.30m
1.20m
0.10m
0.60m
C20/25
C35/45
4.80m
L=20.0m
C: Centroid
M: Center of twist
0.20m
0.30m
0.60m
1.30m
C20/25
1.0102m
0.4799m
Fig. 14. Transverse (a) and longitudinal (b) sections of the composite slab and beam structure of Example 3.
Table 5
Cross section constants of the composite structure of Example 3
Cross section A (m
2
) A
~z
(m
2
) I
~y
(m
4
) I
t
(m
4
) C
M
(m
6
)
Transverse 2.14284 0.81179 0.41685 0.04969 0.49530
Longitudinal 5.65000 0.19170 0.67325 0.21424 40.12990
62
42
31
11
53
30
1
2
22
43
32
21
63
23
44
45
24
3
4
5
6
7
8
9
10
12
13
14
15
16
17
18
19
20
25
26
27
28
29
52
46
47
48
49
50
51
33
34
35
36
37
38
39
40
41
54
55
56
57
58
59
60
61
20m
1.2m
1.2m
P=100kN
P
P
P
P
P
Beam III
Beam I
Beam II
Fig. 15. Structural model and loading of the composite slab and beam structure of Example 3.
114 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116
can surround a nite number of inclusions. To account for
shear deformations, the concept of shear deformation coef-
cients is used. Seven boundary value problems are formu-
lated and solved employing a pure BEM approach, that is
Fig. 16. Deection surfaces and corresponding maximum values of the composite slab and beam structure of Example 3, (a) including (14 14) or (b)
ignoring (12 12) the torsional curvature of the grid members.
Fig. 17. Bending moment diagrams of the composite slab and beam structure of Example 3, (a) including (14 14) or (b) ignoring (12 12) the torsional
curvature of the grid members.
E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116 115
only boundary discretization is used. The main conclusions
that can be drawn from this investigation are
(a) The presented numerical technique is well suited for
computer aided analysis for beams of arbitrary
homogeneous or composite cross section, subjected
to any linear boundary conditions and to an arbi-
trarily distributed or concentrated transverse or twist-
ing loading.
(b) The discrepancy of the deections and the internal
stress resultants arising from the ignorance of the
warping degrees of freedom at the ends of a member
and the magnitude of the normal stresses due to
warping necessitate the utilization of the 14 14
member stiness matrix, especially for beams with
open form cross sections.
(c) The remarkable inuence of the shear deformation
eect in the deection and the internal stress resul-
tants is pointed out.
(d) The advantages of a box shaped closed cross section
beam subjected in torsional loading compared with
that of an open one are easily veried.
(e) The magnitude of the evaluated normal stresses due to
warping necessitates the consideration of these addi-
tional normal stresses, especially near the restrained
edges.
(f) Warping is not constant along the thickness of the
cross section walls as it is assumed in Thin Tube The-
ory for thin walled beams.
(g) The developed procedure retains the advantages of a
BEM solution over a pure domain discretization
method since it requires only boundary discretization.
Acknowledgements
Financial support of this research was provided by the
Iraklitos Research Fellowships with Priority to Basic Re-
search, an EU funded project in the special managing
authority of the Operational Program in Education and
Initial Vocational Training.
References
[1] Murin J, Kutis V. 3D-beam element with continuous variation of the
cross-sectional area. Comput Struct 2002;80:32938.
[2] Mur n J. 3D beam element with changing cross sectional area. Eng
Mech 1999;1:2535 [in Slovak].
[3] Mur n J. Beam element with varying cross-section satisfying local and
global equilibrium conditions. Mech Eng 1998;49(3):20823 [in
Slovak].
[4] Reilly RJ. Stiness analysis of grids including warping. ASCE J Struct
Div 1972;7:151123.
[5] Barsoum RS, Gallagher RH. Finite element analysis of torsional-
exural stability problems. Int J Numer Methods Eng 1970;2:33552.
[6] Waldron P. Elastic analysis of curved thin-walled girders including
the eect of warping restraint. Eng Struct 1985;7:93104.
[7] Waldron P. Stiness analysis of thin-walled girders. ASCE J Struct
Div 1986;6:136684.
[8] Yang Y, McGuire W. A procedure for analyzing space frames with
partial warping restraint. Int J Numer Methods Eng 1984;20:137798.
[9] Ahmed MZ, Weisgerber FE. Torsion constant for matrix analysis of
structures including warping eect. Int J Solids Struct 1996;33(3):
361374.
[10] Bach C, Baumann R. Elastizitat und Festigkeit. 9th ed. Ber-
lin: Springer; 1924.
[11] Stojek D. Zur Schubverformung im Biegebalken. Zeitschrift fu r
Angewandte Mathematik Mech 1964;44:3936.
[12] Timoshenko SP, Goodier JN. Theory of Elasticity, 3rd ed. New
York: McGraw-Hill.
[13] Cowper GR. The Shear Coecient in Timoshenkos Beam Theory. J
Appl Mech, ASME 1966;33(2):33540.
[14] Schramm U, Kitis L, Kang W, Pilkey WD. On the shear deformation
coecient in beam theory. Finite Elements Anal Des 1994;16:14162.
[15] Schramm U, Rubenchik V, Pilkey WD. Beam stiness matrix based
on the elasticity equations. Int J Numer Methods Eng 1997;40:
211232.
[16] Stephen NG. Timoshenkos shear coecient from a beam subjected
to gravity loading. ASME J Appl Mech 1980;47:1217.
[17] Hutchinson JR. Shear coecients for Timoshenko beam theory.
ASME J Appl Mech 2001;68:8792.
[18] Sapountzakis EJ, Mokos VG. Warping shear stresses in nonuniform
torsion of composite bars by BEM. Comput Methods Appl Mech
Eng 2003;192:433753.
[19] Windisch E. Eine Numerische Methode zur Lo sung des Torsions-
problems I. Acta Mech 1967;4:1919.
[20] Friemann H. Schub und torsion in gerade Staben. Du sseldorf: Wer-
ner-Verlag; 1993.
[21] Mokos VG, Sapountzakis EJ. A BEM solution to transverse shear
loading of composite beams. Int J Solids Struct 2005;42:326187.
[22] Katsikadelis JT. Boundary Elements: Theory and Applica-
tions. Amsterdam-London: Elsevier; 2002.
[23] Sapountzakis EJ, Mokos VG. Nonuniform torsion of composite bars
by boundary element method. J Eng Mech, ASCE 2001;127(9):
94553.
[24] Sapountzakis EJ. Solution of nonuniform torsion of bars by an
integral equation method. Comput Struct 2000;77:65967.
Table 6
Maximum deections of the longitudinal beams I, II, III of the composite
structure of Example 3
Beams 14 14 member stiness
matrix
12 12 member stiness
matrix
Including
shear
deformation
Without
shear
deformation
Including
shear
deformation
Without
shear
deformation
I 0.2183 0.2020 0.1254 0.1133
II 0.2404 0.2031 0.2426 0.2035
III 0.2475 0.2038 0.3382 0.2921
116 E.J. Sapountzakis, V.G. Mokos / Computers and Structures 85 (2007) 102116

Potrebbero piacerti anche