Engineering Structures 30 (2008) 653663

www.elsevier.com/locate/engstruct
Shear deformation effect in nonlinear analysis of spatial beams
E.J. Sapountzakis

, V.G. Mokos
School of Civil Engineering, National Technical University, Zografou Campus, GR-157 80 Athens, Greece
Received 31 August 2006; received in revised form 2 May 2007; accepted 7 May 2007
Available online 18 June 2007
Abstract
In this paper a boundary element method is developed for the nonlinear analysis of beams of arbitrary doubly symmetrical simply or multiply
connected constant cross-section, taking into account shear deformation effect. The beam is subjected in an arbitrarily concentrated or distributed
variable axial loading, while the shear loading is applied at the shear centre of the cross-section, avoiding in this way the induction of a twisting
moment. To account for shear deformations, the concept of shear deformation coefcients is used. Five boundary value problems are formulated
with respect to the transverse displacements, the axial displacement and to two stress functions and solved using the analogue equation method, a
BEM-based method. Application of the boundary element technique yields a system of nonlinear equations from which the transverse and axial
displacements are computed by an iterative process. The evaluation of the shear deformation coefcients is accomplished from the aforementioned
stress functions using only boundary integration. Numerical examples are worked out to illustrate the efciency, the accuracy and the range of
applications of the developed method. The inuence of both the shear deformation effect and the variability of the axial loading are remarkable.
c 2007 Elsevier Ltd. All rights reserved.
Keywords: Transverse shear stresses; Shear centre; Shear deformation coefcients; Beam; Second order analysis; Nonlinear analysis; Boundary element method
1. Introduction
An important consideration in the analysis of the
components of plane and space frames or grid systems is
the inuence of the action of axial, lateral forces and end
moments on the deformed shape of a beam. Lateral loads
and end moments generate deection that is further amplied
by axial compression loading. The aforementioned analysis
becomes much more accurate and complex taking into account
that the axial force is nonlinearly coupled with the transverse
deections, avoiding in this way the inaccuracies arising from
a linearized second-order analysis.
Over the past twenty years, many researchers have
developed and validated various methods of performing a
linearized second-order analysis on structures. Early efforts led
to methods based on accounting for the aforementioned effect
by using magnication factors applied to the results obtained
from rst-order analyses [13]. An example of such a method
is the B
1
and B
2
factor approach provided in the AISC-
LRFD specication [4]. Since the modications used in this

Corresponding author. Tel.: +30 2107721718.

E-mail addresses: cvsapoun@central.ntua.gr (E.J. Sapountzakis),
vgmokos@central.ntua.gr (V.G. Mokos).
method are only applied to the moments of the columns and
not of the beams, the results obtained from this method are
often unsatisfactory especially for cases involving moderate to
large deformations [2]. Consequently, due to the demand of
more rigorous and accurate second-order analysis of structural
components several research papers have been published
including a nonlinear incremental stiffness method [5], closed-
form stiffness methods [6,7], the analysis of nonlinear effects
by treating every element as a beamcolumn one [8], a design
method for space frames using stability functions to capture
second-order effects associated with P (for one element)
and P (for multiple elements) effects [9], uniform formulae
restricted to a single bar of a skeletal structure and to only a
few loadings [10], the nite element method using linear and
cubic shape functions (approximate solutions) [11] and a 3-D
second-order plastic-hinge analysis accounting for material and
geometrical nonlinearities [12,13]. Recently, Katsikadelis and
Tsiatas [14] presented a BEM-based method for the nonlinear
analysis of beams with variable stiffness. In all these studies
shear deformation effect is ignored.
Moreover, Kim et al. presented a practical second-order
inelastic static [15] and dynamic [16] analysis for 3-D steel
frames, Machado and Cortinez [17] a geometrically nonlinear
0141-0296/$ - see front matter c 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2007.05.004
654 E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663
beam theory for the lateral buckling problem of bisymmetrical
thin-walled composite simply supported or cantilever beams,
taking into account shear deformation effect. Nevertheless, in
all of the aforementioned research efforts the axial loading
of the structural components is assumed to be constant.
Finally, the boundary element method taking into account
shear deformation effect has been employed only for rst-order
analyses [1820].
In this paper a boundary element method is developed
for the nonlinear analysis of beams of arbitrary doubly
symmetrical simply or multiply connected constant cross
section, taking into account shear deformation effect. The
beam is subjected in an arbitrarily concentrated or distributed
variable axial loading, while the shear loading is applied at
the shear centre of the cross-section, avoiding in this way the
induction of a twisting moment (the torsional loading of the
beam can be independently analysed). To account for shear
deformations, the concept of shear deformation coefcients
is used. Five boundary value problems are formulated with
respect to the transverse displacements, the axial displacement
and to two stress functions and solved using the analogue
equation method [21], a BEM-based method. Application of
the boundary element technique yields a system of nonlinear
equations from which the transverse and axial displacements
are computed by an iterative process. The evaluation of
the shear deformation coefcients is accomplished from
the aforementioned stress functions using only boundary
integration. The essential features and novel aspects of
the present formulation compared with previous ones are
summarized as follows:
i. The beam is subjected in an arbitrarily concentrated or
distributed variable axial loading.
ii. The beam is supported by the most general boundary
conditions including elastic support or restraint.
iii. The performed analysis is a nonlinear one, arising from the
fact that the variable axial force is nonlinearly coupled with
the transverse deections, taking into account additional
terms and avoiding the restrictions of a linearized second-
order analysis.
iv. Shear deformation effect is taken into account.
v. The shear deformation coefcients are evaluated using
an energy approach, instead of Timoshenkos [22] and
Cowpers [23] denitions, for which several authors [24,
25] have pointed out that one obtains unsatisfactory results
or denitions given by other researchers [26,27], for which
these factors take negative values.
vi. The effect of the materials Poisson ratio is taken into
account.
vii. The proposed method employs a pure BEM approach
(requiring only boundary discretization) resulting in line
or parabolic elements instead of area elements of the
FEM solutions (requiring the whole cross-section to be
discretized into triangular or quadrilateral area elements),
while a small number of line elements are required to
achieve high accuracy.
Fig. 1. Prismatic beam in torsionless bending (a) with an arbitrary doubly
symmetrical cross-section occupying the two-dimensional region (b).
Numerical examples are worked out to illustrate the
efciency, the accuracy and the range of applications of the
developed method. The inuence of both the shear deformation
effect and the variability of the axial loading are remarkable.
2. Statement of the problem
Consider a prismatic beam of length l with a doubly
symmetrical cross section of arbitrary shape, occupying the
two dimensional multiply connected region of the y, z
plane bounded by the K + 1 curves
1
,
2
, . . . ,
K
,
K+1
,
as shown in Fig. 1. These boundary curves are piecewise
smooth, i.e. they may have a nite number of corners. The
material of the beam, with shear modulus G and Poissons
ratio v is assumed homogeneous, isotropic and linearly elastic.
Without loss of generality, it may be assumed that the x-axis
of the beam principal coordinate system is the line joining
the centroids of the cross-sections. The beam is subjected to
an arbitrarily distributed axial loading p
x
and to torsionless
bending arising from arbitrarily distributed transverse loading
p
y
, p
z
and bending moments m
y
, m
z
along y and z axes,
respectively (Fig. 1(a)).
According to the nonlinear theory of beams for moderate
large deections ((u/x)
2
u/x, (u/x) (u/y)
(v/x) +(u/y) , (u/x) (u/z) (w/x) +(u/z))
and assuming small rotations [28,29], the angles of rotation
of the cross-section in the xz and xy planes of the beam
subjected to the aforementioned loading and taking into account
shear deformation effect satisfy the following relations:
cos
y
1 (1a)
cos
z
1 (1b)
sin
y

y
=
dw
dx
=
y

z
(1c)
sin
z

z
=
dv
dx
=
z

y
(1d)
E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663 655
where w = w (x), v = v (x) are the beam transverse
displacements with respect to z, y axes (Fig. 2(a)) [28,29],
respectively, while the corresponding curvatures are given as
k
y
=
d
y
dx
=
d
2
w
dx
2
+
d
z
dx
=
d
2
w
dx
2

p
z
GA
z
(2a)
k
z
=
d
z
dx
=
d
2
v
dx
2

d
y
dx
=
d
2
v
dx
2
+
p
y
GA
y
(2b)
where
y
,
z
are the additional angles of rotation of the cross-
section due to shear deformation and GA
y
, GA
z
are its shear
rigidities of the Timoshenkos beam theory, where
A
z
=
z
A =
1
a
z
A (3a)
A
y
=
y
A =
1
a
y
A (3b)
are the shear areas with respect to y, z axes, respectively
with
y
,
z
the shear correction factors, a
y
, a
z
the shear
deformation coefcients and A the cross section area. Referring
to Fig. 2(b) [28,29], the stress resultants R
x
, R
z
acting in the x,
z directions, respectively, are related to the axial N and shear
Q
z
forces as
R
x
= N cos
y
+ Q
z
sin
y
(4a)
R
z
= Q
z
cos
y
N sin
y
(4b)
which by virtue of Eq. (1) become
R
x
= N Q
z
dw
dx
(5a)
R
z
= Q
z
+ N
dw
dx
. (5b)
The second term in the right-hand side of Eq. (5a), expresses
the inuence of the shear force Q
z
on the horizontal stress
resultant R
x
. However, this term can be neglected since Q
z
w

is much smaller than N [28,29] and thus Eq. (5a) can be written
as
R
x
N. (6)
Similarly, the stress resultant R
y
acting in the y direction is
related to the axial N and shear Q
y
forces as
R
y
= Q
y
+ N
dv
dx
. (7)
The governing equation of the beam transverse displacement
w = w (x) will be derived by considering the equilibriumof the
deformed element in the xz plane. Thus, referring to Fig. 2(b)
we obtain
dR
x
dx
+ p
x
= 0 (8a)
dR
z
dx
+ p
z
= 0 (8b)
dM
y
dx
Q
z
+m
y
= 0. (8c)
Fig. 2. Displacements (a) and forces (b) acting on the deformed element in the
xz plane.
Substituting Eqs. (6) and (5b) into Eqs. (8a) and (8b), using
Eq. (8c) to eliminate Q
z
, employing the well-known relation
M
y
= EI
y
k
y
(9)
and utilizing Eq. (2a) we obtain the expressions of the angle of
rotation due to bending
y
and the stress resultants M
y
, R
z
as

y
=
dw
dx
+
1
GA
z
_
EI
y
d
3
w
dx
3

EI
y
GA
z
_
dp
z
dx
+ N
d
3
w
dx
3
2p
x
d
2
w
dx
2

dp
x
dx
dw
dx
_
+m
y
_
(10)
M
y
= EI
y
d
2
w
dx
2

EI
y
GA
z
_
p
z
+
dN
dx
dw
dx
+ N
d
2
w
dx
2
_
(11a)
R
z
= EI
y
d
3
w
dx
3

EI
y
GA
z
_
dp
z
dx
+ N
d
3
w
dx
3
2p
x
d
2
w
dx
2

dp
x
dx
dw
dx
_
+m
y
+ N
dw
dx
(11b)
and the governing differential equation as
EI
y
_
1 +
N
GA
z
_
d
4
w
dx
4
= p
z
p
x
dw
dx
+ N
d
2
w
dx
2
+
dm
y
dx

EI
y
GA
z
_
d
2
p
z
dx
2
3p
x
d
3
w
dx
3
3
dp
x
dx
d
2
w
dx
2

d
2
p
x
dx
2
dw
dx
_
. (12)
Moreover, the pertinent boundary conditions of the problem
are given as

z
1
w (x) +
z
2
R
z
(x) =
z
3
(13a)
656 E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663

z
1

y
(x) +
z
2
M
y
(x) =
z
3
at the beam ends x = 0, l (13b)
where
z
i
,
z
i
(i = 1, 2, 3) are given constants, while the angle
of rotation
y
and the stress resultants M
y
, R
z
are given as

y
=
EI
y
GA
z
_
1 +
N
GA
z
_
d
3
w
dx
3

dw
dx
(14a)
M
y
= EI
y
_
1 +
N
GA
z
_
d
2
w
dx
2
(14b)
R
z
= EI
y
_
1 +
N
GA
z
_
d
3
w
dx
3
+ N
dw
dx
at the beam ends x = 0, l. (14c)
Eq. (13) describe the most general boundary conditions
associated with the problem at hand and can include elastic
support or restrain. It is apparent that all types of the
conventional boundary conditions (clamped, simply supported,
free or guided edge) can be derived form these equations by
specifying appropriately the functions
z
i
and
z
i
(e.g. for a
clamped edge it is
z
1
=
z
1
= 1,
z
2
=
z
3
=
z
2
=
z
3
= 0).
Similarly, considering the beam in the xy plane we
obtain the boundary value problem of the beam transverse
displacement v = v (x) as
EI
z
_
1 +
N
GA
y
_
d
4
v
dx
4
= p
y
p
x
dv
dx
+ N
d
2
v
dx
2

dm
z
dx

EI
z
GA
y
_
d
2
p
y
dx
2
3p
x
d
3
v
dx
3
3
dp
x
dx
d
2
v
dx
2

d
2
p
x
dx
2
dv
dx
_
inside the beam (15)

y
1
v (x) +
y
2
R
y
(x) =
y
3
(16a)

y
1

z
(x) +
y
2
M
z
(x) =
y
3
at the beam ends x = 0, l (16b)
where
y
i
,
y
i
(i = 1, 2, 3) are given constants and the
expressions of the angle of rotation
z
and the stress resultants
M
z
, R
y
inside the beam are given as

z
=
dv
dx

1
GA
y
_
EI
z
d
3
v
dx
3

EI
z
GA
y
_
dp
y
dx
+ N
d
3
v
dx
3
2p
x
d
2
v
dx
2

dp
x
dx
dv
dx
_
m
y
_
(17a)
M
z
= EI
z
d
2
v
dx
2
+
EI
z
GA
y
_
p
y
+
dN
dx
dv
dx
+ N
d
2
v
dx
2
_
(17b)
R
y
= EI
z
d
3
v
dx
3

EI
z
GA
y
_
dp
y
dx
+ N
d
3
v
dx
3
2p
x
d
2
v
dx
2

dp
x
dx
dv
dx
_
m
y
+ N
dv
dx
. (17c)
In both of the aforementioned boundary value problems the
axial force N inside the beam or at its boundary is given from
the following relation [28,29]:
N = E A
_
du
dx
+
1
2
_
dw
dx
_
2
+
1
2
_
dv
dx
_
2
_
(18)
where u = u (x) is the bar axial displacement, which can be
evaluated from the solution of the following boundary value
problem:
E A
_
d
2
u
dx
2
+
d
2
w
dx
2
dw
dx
+
d
2
v
dx
2
dv
dx
_
= p
x
inside the beam (19)
c
1
u (x) +c
2
N(x) = c
3
at the beam ends x = 0, l (20)
where c
i
(i = 1, 2, 3) are given constants.
The solution of the boundary value problems prescribed
from Eqs. (12), (13a), (13b), (15), (16a) and (16b) presumes
the evaluation of the shear deformation coefcients a
z
, a
y
corresponding to the principal centroidal system of axes Cyz.
These coefcients are established equating the approximate
formula of the shear strain energy per unit length [25]
U
appr.
=
a
y
Q
2
y
2AG
+
a
z
Q
2
z
2AG
(21)
with the exact one given from
U
exact
=
_

(
xz
)
2
+
_

xy
_
2
2G
d (22)
and are obtained as [30].
a
y
=
1

y
=
A

2
_

[() e] [() e] d (23a)

a
z
=
1

z
=
A

2
_

[() d] [() d] d (23b)

where (
xz
) ,
_

xy
_
are the transverse (direct) shear stress
components, () i
y
(/y) + i
z
(/z) is a symbolic vector
with i
y
, i
z
the unit vectors along y and z axes, respectively,
is given from
= 2 (1 +) I
y
I
z
(24)
is the Poisson ratio of the cross-section material, e and d are
vectors dened as
e =
_
I
y
y
2
z
2
2
_
i
y
+
_
I
y
yz
_
i
z
(25a)
d = (I
z
yz) i
y
+
_
I
z
z
2
y
2
2
_
i
z
(25b)
and (y, z), (y, z) are stress functions, which are evaluated
from the solution of the following Neumann type boundary
value problems [30].

2
= 2I
y
y in (26a)

n
= n e on =
K+1
_
j =1

j
(26b)

2
= 2I
z
z in (27a)

n
= n d on =
K+1
_
j =1

j
(27b)
E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663 657
where n is the outward normal vector to the boundary . In
the case of negligible shear deformations a
z
= a
y
= 0. It is
also worth here noting that the boundary conditions (26b) and
(27b) have been derived fromthe physical consideration that the
traction vector in the direction of the normal vector n vanishes
on the free surface of the beam.
3. Integral representations numerical solution
According to the precedent statement, the nonlinear analysis
of a beam including shear deformation reduces in establishing
the transverse displacements w = w (x), v = v (x), the axial
displacement u = u (x) and the stress functions (y, z),
(y, z). The transverse and the axial displacements must
have continuous derivatives up to the fourth and second order,
respectively, with respect to x, while the stress functions must
have continuous partial derivatives up to the second order with
respect to y, z.
3.1. For the transverse displacements w, v
The numerical solution of the boundary value problems
described by Eqs. (12), (13a), (13b), (15), (16a) and (16b) is
similar. For this reason, in the following we will analyse
the solution of the problem of Eqs. (12), (13a) and (13b)
noting any alteration or addition for the problem of Eqs.
(15), (16a) and (16b). Eq. (12) is solved using the analogue
equation method [21]. This method has been developed for
the beam equation including axial forces by Katsikadelis and
Tsiatas [31]. However, another formulation is presented in this
investigation.
Let w be the sought solution of the boundary value problem
described by Eqs. (12), (13a) and (13b). Differentiating this
function four times yields
d
4
w
dx
4
= q
z
(x) . (28)
Eq. (28) indicates that the solution of the original problem
can be obtained as the deection of a beam with unit exural
rigidity and innite shear rigidity subjected to a exural
ctitious load q
z
(x) under the same boundary conditions. The
ctitious load is unknown. However, it can be established using
BEM as follows:
The solution of Eq. (28) is given in integral form as
w (x) =
_
l
0
q
z
w

dx
_
w

d
3
w
dx
3

dw

dx
d
2
w
dx
2
+
d
2
w

dx
2
dw
dx

d
3
w

dx
3
w
_
l
0
(29)
where w

is the fundamental solution, which is given as

w

=
1
12
l
3
_
2 +||
3
3 ||
2
_
(30)
with = r/l, r = x , x, points of the beam, which is a
particular singular solution of the equation
d
4
w

dx
4
= (x, ). (31)
Employing Eq. (30) the integral representation (29) can be
written as
w(x) =
_
l
0
q
z

4
(r)dx
_

4
(r)
d
3
w
dx
3
+
3
(r)
d
2
w
dx
2
+
2
(r)
dw
dx
+
1
(r)w
_
l
0
(32)
where the kernels
i
(r), (i = 1, 2, 3, 4) are given as

1
(r) =
1
2
sgn (33a)

2
(r) =
1
2
l(1 ||) (33b)

3
(r) =
1
4
l
2
|| (|| 2)sgn (33c)

4
(r) =
1
12
l
3
_
2 +||
3
3 ||
2
_
. (33d)
Notice that in Eq. (32) for the line integral it is r = x, x,
points inside the beam, whereas for the rest terms r = x , x
inside the beam, at the beam ends 0, l.
Differentiating Eq. (32) results in the integral representations
of the derivatives of the deection w as
dw(x)
dx
=
_
l
0
q
z

3
(r)dx
_

3
(r)
d
3
w
dx
3
+
2
(r)
d
2
w
dx
2
+
1
(r)
dw
dx
_
l
0
(34a)
d
2
w(x)
dx
2
=
_
l
0
q
z

2
(r)dx
_

2
(r)
d
3
w
dx
3
+
1
(r)
d
2
w
dx
2
_
l
0
(34b)
d
3
w(x)
dx
3
=
_
l
0
q
z

1
(r)dx
_

1
(r)
d
3
w
dx
3
_
l
0
. (34c)
The integral representations (32) and (34a) written for
the beam ends 0, l together with the boundary conditions
(13a) and (13b) can be employed to express the unknown
boundary quantities w, w

, w

and w

in terms of q
z
. This
is accomplished numerically. If L is the number of the nodal
points along the beam axis, this procedure yields the following
set of linear equations:
_
_
_
_
[E
11
] [E
12
] [E
13
] [E
14
]
[0] [E
22
] [E
23
] [0]
[E
31
] [E
32
] [E
33
] [E
34
]
[0] [E
42
] [E
43
] [E
44
]
_

_
_

_
{w}
_
w

_
_
w

_
_
w

_
_

_
=
_

_
_
a
z
3
_
_

z
3
_
{0}
{0}
_

_
+
_
_
_
_
[0]
[0]
[F
3
]
[F
4
]
_

_
{q
z
} (35)
in which [E
22
], [E
23
], [E
1i
], (i = 1, 2, 3, 4) are 2 2 matrices
including the nodal values of the functions a
z
1
, a
z
2
,
z
1
,
z
2
of
Eqs. (13a) and (13b) and
_
E
i j
_
, (i = 3, 4, j = 1, 2, 3, 4)
are square 2 2 known coefcient matrices resulting from the
values of the kernels
i
at the beam ends;
_
a
z
3
_
,
_

z
3
_
are 2 1
658 E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663
known column matrices including the boundary values of the
functions a
z
3
,
z
3
in Eqs. (13a) and (13b) and [F
i
] (i = 3, 4)
are 2 L rectangular known matrices originating from the
integration of the kernels on the axis of the beam. Finally, {w},
_
w

_
,
_
w

_
and
_
w

_
are vectors including the two unknown
nodal values of the respective boundary quantities and {q
z
} is
a vector including the L unknown nodal values of the ctitious
load.
The discretized counterpart of Eq. (32) when applied to all
nodal points in the interior of the beam yields
{W} = [F] {q
z
}
_
[E
1
] {w} +[E
2
]
_
w

_
+ [E
3
]
_
w

_
+[E
4
]
_
w

__
(36)
where [F] is an L L known matrix and [E
i
], (i = 1, 2, 3, 4)
are L 2 also known matrices. Elimination of the boundary
quantities from Eq. (36) using Eq. (35) for homogeneous
boundary conditions Eqs. (13a) and (13b) (a
z
3
=
z
3
= 0) yields
{W} =
_
B
z
_
{q
z
} (37)
where
_
B
z
_
is an L L matrix.
Moreover, the discretized counterpart of Eqs. (34a)(34c)
when applied to all nodal points in the interior of the beam, after
elimination of the boundary quantities using Eq. (35) yields
_
W

_
=
_
B

z
_
{q
z
} (38a)
_
W

_
=
_
B

z
_
{q
z
} (38b)
_
W

_
=
_
B

z
_
{q
z
} (38c)
where
_
B

z
_
,
_
B

z
_
,
_
B

z
_
are known L L coefcient matrices.
Note that Eqs. (37) and (38a)(38c) are valid for homogeneous
boundary conditions (a
z
3
=
z
3
= 0). For nonhomogeneous
boundary conditions, an additive constant vector will appear in
the right-hand side of these equations.
Finally, applying Eq. (12) to the L nodal points in the interior
of the beam we obtain the following linear system of equations
with respect to q
z
__
D

z
_

_
D

z
_ _
B

z
_

_
D

z
_ _
B

z
_

_
D

z
_ _
B

z
__
{q
z
}
= { p
z
} +
_
m

y
_

_
D
z
_ _
p

z
_
(39)
where
_
D

z
_
,
_
D

z
_
,
_
D

z
_
,
_
D

z
_
,
_
D
z
_
are diagonal L L
matrices whose elements are given from
_
D

z
_
i i
= EI
y
_
1 +
N
i
GA
z
_
(40a)
_
D

z
_
i i
=
3EI
y
GA
z
( p
x
)
i
(40b)
_
D

z
_
i i
=
3EI
y
GA
z
_
p

x
_
i
+ N
i
(40c)
_
D

z
_
i i
=
EI
y
GA
z
_
p

x
_
i
( p
x
)
i
(40d)
(D
z
)
i i
=
EI
y
GA
z
(40e)
at the L nodal points in the interior of the beam; {q
z
}, { p
z
}, {m

y
}
and
_
p

z
_
are vectors with L elements including the values of
the ctitious loading, the transverse loading, the rst derivative
of the bending moment distributed loading and the second
derivative of the transverse loading at the L nodal points in the
interior of the beam. The values of the
_
m
y
_

, ( p
z
)

, ( p
x
)

and
( p
x
)

quantities result after approximating the corresponding

derivatives with appropriate central, forward or backward nite
differences.
Similarly for the transverse deection v = v (x), application
of the Eq. (15) to the L nodal points in the interior of the beam
results the following linear system of equations with respect to
q
y
_
_
D

y
_

_
D

y
_ _
B

y
_

_
D

y
_ _
B

y
_

_
D

y
_ _
B

y
__
_
q
y
_
=
_
p
y
_

_
m

z
_

_
D
y
_
_
p

y
_
(41)
where
_
B

y
_
,
_
B

y
_
,
_
B

y
_
are known L L coefcient matrices
similar with those mentioned before for the deection w;
_
D

y
_
,
_
D

y
_
,
_
D

y
_
,
_
D

y
_
,
_
D
y
_
are diagonal L L matrices
whose elements are given from
_
D

y
_
i i
= EI
z
_
1 +
N
i
GA
y
_
(42a)
_
D

y
_
i i
=
3EI
z
GA
y
( p
x
)
i
(42b)
_
D

y
_
i i
=
3EI
z
GA
y
_
p

x
_
i
+ N
i
(42c)
_
D

y
_
i i
=
EI
z
GA
y
_
p

x
_
i
( p
x
)
i
(42d)
_
D
y
_
i i
=
EI
z
GA
y
(42e)
at the L nodal points in the interior of the beam;
_
q
y
_
,
_
p
y
_
,
_
m

z
_
and { p

y
} are vectors with L elements, similar with those
mentioned before for the deection w.
3.2. For the axial displacement u
Let u be the sought solution of the boundary value problem
described by Eqs. (19) and (20). Differentiating this function
two times yields
d
2
u
dx
2
= q
x
(x) . (43)
Eq. (43) indicates that the solution of the original problem
can be obtained as the axial displacement of a beam with unit
axial rigidity subjected to a exural ctitious load q
x
(x) under
the same boundary conditions. The ctitious load is unknown.
The solution of Eq. (43) is given in integral form as
u(x) =
_
l
o
u

q
x
ds +
_
u

du
dx

du

dx
u
_
l
0
(44)
E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663 659
where u

is the fundamental solution, which is given as

u

=
1
2
|r| (45)
with r = x , x, points of the beam.
Following the same procedure as in 3.1, the discretized
counterpart of Eq. (44) and its rst derivative with respect to
x, when applied to all nodal points in the interior of the beam
yields
{U} = [B
x
] {q
x
} (46a)
_
U

_
=
_
B

x
_
{q
x
} (46b)
where [B
x
],
_
B

x
_
are known matrices with dimensions L L,
similar with those mentioned before for the deection w and
the following system of equations with respect to q
x
, q
y
and q
z
is obtained
_
D

x
_
{q
x
} +
_
D

x
_ __
B

z
_
{q
z
}
_
dg.
_
B

z
_
{q
z
}
+
_
D

x
_
__
B

y
_
_
q
y
_
_
dg.
_
B

y
_
_
q
y
_
= { p
x
} (47)
where the symbol []
dg.
indicates a diagonal matrix with the
elements of the included column matrix, {q
x
}, { p
x
} are vectors
with L elements, similar with those mentioned before for the
deection w, while the axial force at the neutral axis of the
beam can be expressed as follows
{N} =
_
D

x
_ _
B

x
_
{q
x
} +
_
D

x
_ __
B

z
_
{q
z
}
_
dg.
_
B

z
_
{q
z
}
+
_
D

x
_ __
B

z
_
{q
z
}
_
dg.
_
B

z
_
{q
z
} . (48)
In Eqs. (47) and (48)
_
D

x
_
,
_
D

x
_
are diagonal LL matrices
whose elements are given from
_
D

x
_
i i
= E A (49a)
_
D

x
_
i i
=
E A
2
(49b)
at the L nodal points in the interior of the beam. Eqs. (39),
(41), (47) and (48) constitute a nonlinear coupled system
of equations with respect to q
x
, q
y
, q
z
and N quantities.
The solution of this system is accomplished iteratively by
employing the two term acceleration method [32,33].
3.3. For the stress functions (y, z) and (y, z)
The evaluation of the stress functions (y, z) and (y, z)
is accomplished using BEM as this is presented in Sapountzakis
and Mokos [30].
Moreover, since the torsionless bending problem of beams
is solved by the BEM, the domain integrals for the evaluation
of the area, the bending moments of inertia and the shear
deformation coefcients (Eqs. (23a) and (23b)) have to be
converted to boundary line integrals, in order to maintain the
pure boundary character of the method. This can be achieved
using integration by parts, the Gauss theorem and the Green
identity. Thus, the moments of inertia and the cross section area
can be written as
I
y
=
_

_
yz
2
cos
_
ds (50a)
I
z
=
_

_
zy
2
sin
_
ds (50b)
A =
1
2
_

(y cos + z sin ) ds (50c)

while the shear deformation coefcients a
y
and a
z
are obtained
from the relations
a
y
=
A

2
_
(4v +2) I
y
I
y
+
1
4
v
2
I
2
y
I
ed
I
e
_
(51a)
a
z
=
A

2
_
(4v +2) I
z
I
z
+
1
4
v
2
I
2
z
I
ed
I
d
_
(51b)
where
I
e
=
_

(n e) ds (52a)
I
d
=
_

(n d) ds (52b)
I
ed
=
_

_
y
4
z sin + z
4
y cos +
2
3
y
2
z
3
sin
_
ds (52c)
I
y
=
1
6
_

_
2I
y
y
4
z sin +(3 cos
y (n e)) y
2
_
ds (52d)
I
z
=
1
6
_

_
2I
z
z
4
y cos +(3 sin
z (n d)) z
2
_
ds. (52e)
4. Numerical examples
On the basis of the analytical and numerical procedures
presented in the previous sections, a FORTRAN program has
been written and representative examples have been studied
to demonstrate the efciency, wherever possible the accuracy
and the range of applications of the developed method. In
all the examples treated each cross section has been analysed
employing N
CS
= 300 constant boundary elements along
the boundary of the cross section and N
Beam
= 29 constant
elements along the axis of the beam, which are enough to ensure
the convergence of the solution procedure. Moreover, the CPU
time at a Personal Computer Intel(R) 2.00 GHz for the analysis
of each example is less than 6 s.
Example 1. A cantilever beam of a hollow rectangular
cross section b h loaded axially (either tensionally or
compressively) and transversely in both directions, as shown
in Fig. 3 has been studied. In Figs. 4 and 5 the transverse
deections w, v and in Fig. 6 the axial displacement u along
the beam axis are presented for both cases of tensile or
compressive axial loading as compared with those obtained
from a linear analysis taking into account or ignoring shear
660 E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663
Fig. 3. Cantilever beam of Example 1.
Fig. 4. Deections w of the beam of Example 1.
deformation effect. As expected, from the rst two gures
it is easily veried that tensile axial loading decreases while
compressive increases the resulting deections. From the last
gure it comes up that even for tensile axial loading the results
of the nonlinear analysis show a remarkable beam shortening
coming from the intense transverse load. Moreover, from the
aforementioned gures the increment of all the deections
and the axial displacement due to the inuence of the shear
deformation effect is remarkable and should not be ignored in
nonlinear analysis. Furthermore, in order to demonstrate the
accuracy of the proposed method, in Fig. 7 the obtained beam
transverse deections v at the free end of the cantilever beam
for tensile axial loading are presented as compared with those
obtained from a FEM solution [34] using 60 beam elements and
applying the Full NewtonRaphson Method [28]. The results
are in excellent agreement. Finally, in Table 1 the transverse
deections w, v and the axial displacement u at the free end of
the cantilever beam for compressive axial loading are presented
for various discretization schemes of N
Beam
constant elements,
demonstrating the convergence and stability of the proposed
method. It is apparent that 29 elements give good results.
Example 2. For comparison reasons, a clamped beam of length
l = 0.508 m, of a rectangular cross section (E = 2.0688
10
8
kPa, v = 0.3) b
y
h
z
= 0.0254 m0.003175 m subjected
Fig. 5. Deections v of the beam of Example 1.
Fig. 6. Axial displacement u of the beam of Example 1.
Fig. 7. Deections v at the free end of the cantilever beam of Example 1.
E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663 661
Table 1
Transverse deections w (cm), v (cm) and axial displacement u (cm) at the free
end of the cantilever beam of Example 1 for compressive axial loading
N
Beam
Nonlinear analysis with shear deformation
w (cm) v (cm) u (cm)
5 29.9960 48.7650 9.6267
9 29.5840 45.6095 8.7167
13 29.5005 45.0107 8.5505
17 29.4696 44.7923 8.4904
21 29.4548 44.6882 8.4619
25 29.4466 44.6305 8.4461
29 29.4416 44.5952 8.4364
50 29.4320 44.5277 8.4179
100 29.4283 44.5022 8.4110
Fig. 8. Deections w (cm) at the mid-span of the beam of Example 2.
to a concentrated transverse load P
z
( kN) at its mid span has
been studied. In Table 2 the deections w at the loading point
for various values of the concentrated load P
z
are presented as
compared with those obtained from a linear analysis taking into
account or ignoring shear deformation effect. From this table
the signicant inuence of the nonlinear analysis effect is easily
veried especially in the case of intense loading, while as it was
expected (since the cross sections dimensions are very small
compared with the beams length) the shear deformation effect
in this case could be ignored. Moreover, in Fig. 8 the deections
w at the mid span for various values of the concentrated load
P
z
are presented as compared with those obtained from a FEM
solution using 8-noded plane stress rectangular elements [35].
The accuracy of the results is remarkable.
In order to establish the inuence of the shear deformation
effect the aforementioned clamped beam of l/4 length
subjected to a concentrated transverse load P
y
at its mid span
has been studied. In Table 3 and in Fig. 9 the deections v at
the loading point for various values of the concentrated load P
y
are presented as compared with those obtained from a linear
analysis taking into account or ignoring shear deformation
effect. The increment of all the deections due to the inuence
of the shear deformation effect is noteworthy.
Table 2
Transverse deections w (cm) at the mid-span of the clamped beam of
Example 2
P
z
(kN) Linear analysis Nonlinear analysis
Ignoring
shear
deformation
(a
z
= 0)
With shear
deformation
(a
z
= 3.8741)
Ignoring
shear
deformation
(a
z
= 0)
With shear
deformation
(a
z
= 3.8741)
0.044 0.2166 0.2168 0.1772 0.1773
0.089 0.4332 0.4336 0.2799 0.2801
0.222 1.0832 1.0841 0.4522 0.4523
0.356 1.7332 1.7346 0.5594 0.5595
0.445 2.1665 2.1682 0.6157 0.6158
0.890 4.3331 4.3365 0.8175 0.8177
1.334 6.4997 6.5048 0.9581 0.9583
1.779 8.6662 8.6731 1.0696 1.0699
2.224 10.832 10.8413 1.1638 1.1642
2.669 12.9994 13.0096 1.2460 1.2465
3.114 15.1660 15.1779 1.3197 1.3202
Table 3
Transverse deections v (cm) at the mid span of the clamped beam of l/4 length
of Example 2
P
y
(kN) Linear analysis Nonlinear analysis
Ignoring
shear
deformation
(a
z
= 0)
With shear
deformation
(a
z
= 3.8741)
Ignoring
shear
deformation
(a
z
= 0)
With shear
deformation
(a
z
= 3.8741)
50 0.05944 0.07428 0.05942 0.07422
100 0.11888 0.14856 0.11869 0.14809
200 0.23776 0.29587 0.23629 0.29343
400 0.47552 0.59424 0.46439 0.56759
600 0.71328 0.89137 0.67859 0.81321
800 0.95104 1.18849 0.87640 1.03012
1000 1.18879 1.48561 1.05776 1.22204
1250 1.48599 1.85701 1.26321 1.43333
1500 1.78319 2.22841 1.44816 1.61947
2000 2.37758 2.97122 1.76910 1.93654
2500 2.97199 3.71402 2.04101 2.20166
5. Concluding remarks
In this paper a boundary element method is developed
for the nonlinear analysis of beams of arbitrary doubly
symmetric simply or multiply connected constant cross
section, taking into account shear deformation effect. Five
boundary value problems are formulated with respect to
the transverse displacements, the axial displacement and
to two stress functions and solved using the Analog
Equation Method, a BEM based method. The evaluation
of the shear deformation coefcients is accomplished from
the aforementioned stress functions using only boundary
integration. The main conclusions that can be drawn from this
investigation are
a. The numerical technique presented in this investigation
is well suited for computer aided analysis for beams of
arbitrary doubly symmetric simply or multiply connected
cross section.
662 E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663
Fig. 9. Deections v (cm) at the mid-span of the clamped beam of l/4 length
of Example 2.
b. The signicant inuence of geometrical nonlinear analysis
in beam elements subjected in intense transverse loading is
veried.
c. The discrepancy between the results of the linear and the
nonlinear analysis demonstrates the signicant inuence of
the axial loading.
d. The accuracy of the proposed method is noteworthy even for
few boundary elements.
e. The remarkable increment of all the deections and the axial
displacements due to the inuence of the shear deformation
effect demonstrates its signicant inuence in nonlinear
analysis.
f. The developed procedure retains the advantages of a BEM
solution over a pure domain discretization method since it
requires only boundary discretization.
Acknowledgments
Financial support provided by the HRAKLEITOS Research
Fellowships with Priority to Basic Research, an EU funded
project in the special managing authority of the Operational
Program in Education and Initial Vocational Training. The
Project HRAKLEITOS is co-funded by the European Social
Fund (75%) and National Resources (25%).
References
[1] Chen WF, Atsura T. Theory of beamcolumn, vol. I. New York (NY):
McGraw-Hill Inc; 1977.
[2] Goto Y, Chen WF. Second-order elastic analysis for frame design. ASCE
Journal of Structural Engineering 1987;113(7):150129.
[3] Rutenberg A. A direct P-delta analysis using standard plane frame
computer programs. Computers and Structures 1981;14(12):97102.
[4] L.R.F.D. Load and resistance factor design specications for buildings.
Chicago (IL): American Institute of Steel Construction (AISC); 1994.
[5] Chajes A, Churchill JE. Nonlinear frame analysis by nite element
methods. ASCE Journal of Structural Engineering 1987;113(6):122135.
[6] Liew JYR, Chen WF. Trends toward advanced analysis. In: Chenand WF,
Tom S, editors. Advanced analysis of steel frames. Boca Raton (FL): CRC
Press, Inc.; 1994. p. 145.
[7] Goto Y. Second-order elastic analysis of frames. In: Chen WF, Toma S,
editors. Advanced analysis of steel frames. Boca Raton (FL): CRC Press,
Inc.; 1994. p. 4790.
[8] Vasek M. The non-linear behaviour of large space bar and beamstructures.
In: Proc. space struct. 4 conf. Thomas Telford series, vol. 1. London
(England). 1993. p. 66573.
[9] Kim E-S, Park M, Choi S-H. Direct design of three dimensional
frames using practical advanced analysis. Engineering Structures 2001;
23:1491502.
[10] Rubin H. Uniform formulae of rst- and second-order theory for skeletal
structures. Engineering Structures 1997;19(11):9039.
[11] Torkamani M, Sommez M, Cao J. Second-order elastic plane-rame
analysis using nite-element method. Journal of Structural Engineering
1997;122535.
[12] Kim E-S, Lee J, Park J-S. 3-D second-order plastic analysis accounting
for local buckling. Engineering Structures 2003;25:8190.
[13] Kim E-S, Choi S-H. Practical second-order inelastic analysis for three-
dimensional steel frames subjected to distributed load. Thin-Walled
Structures 2005;43:13560.
[14] Katsikadelis JT, Tsiatas GC. Large deection analysis of beams with
variable stiffness. Acta Mechanica 2003;164:113.
[15] Kim E-S, Lee J, Kim E-S. Practical second-order inelastic analysis for
steel frames subjected to distributed load. Engineering Structures 2004;
26:5161.
[16] Kim E-S, Ngo-Huu C, Lee D-H. Second-order inelastic dynamic analysis
of 3-D steel frames. International Journal of Solids and Structures 2005;
117.
[17] Machado PS, Cortinez VH. Lateral buckling of thin-walled composite
bisymmetric beams with prebuckling and shear deformation. Engineering
Structures 2005;27:118596.
[18] Rong G, Kisu H, Huang C. A new algorithm for bending problems
of continuous and inhomogeneous beam by the BEM. Advances in
Engineering Software 1999;30:33946.
[19] Schanz M, Antes H. A boundary integral formulation for the dynamic
behaviour of a Timoshenko beam. Electronic Journal of Boundary
Elements 2002;3:34859. BETEQ 2001.
[20] Antes H. Fundamental solution and integral equations for Timoshenko
beams. Computers and Structures 2003;81:38396.
[21] Katsikadelis JT. The analog equation method. A boundary-only integral
equation method for nonlinear static and dynamic problems in general
bodies. Theoretical and Applied Mechanics 2002;27:1338.
[22] Timoshenko SP, Goodier JN. Theory of elasticity. 3rd ed. New York:
McGraw-Hill; 1984.
[23] Cowper GR. The shear coefcient in Timoshenkos beam theory. Journal
of Applied Mechanics ASME 1966;33(2):33540.
[24] Schramm U, Kitis L, Kang W, Pilkey WD. On the shear deformation
coefcient in beam theory. Finite Elements in Analysis and Design 1994;
16:14162.
[25] Schramm U, Rubenchik V, Pilkey WD. Beam stiffness matrix based on
the elasticity equations. International Journal for Numerical Methods in
Engineering 1997;40:21132.
[26] Stephen NG. Timoshenkos shear coefcient from a beam subjected to
gravity loading. ASME Journal of Applied Mechanics 1980;47:1217.
[27] Hutchinson JR. Shear coefcients for Timoshenko beam theory. ASME
Journal of Applied Mechanics 2001;68:8792.
[28] Ramm E, Hofmann TJ. In: Mehlhorn G, editor. Stabtragwerke, Der
Ingenieurbau. Berlin: Band Baustatik/Baudynamik, Ernst & Sohn; 1995.
[29] Rothert H, Gensichen V. Nichtlineare Stabstatik. Berlin: Springer-Verlag;
1987.
[30] Sapountzakis EJ, Mokos VG. A BEM solution to transverse shear loading
of beams. Computational Mechanics 2005;36:38497.
[31] Katsikadelis JT, Tsiatas GC. Nonlinear dynamic analysis of beams with
variable stiffness. Journal of Sound and Vibration 2004;270:84763.
E.J. Sapountzakis, V.G. Mokos / Engineering Structures 30 (2008) 653663 663
[32] Isaacson E, Keller HB. Analysis of numerical methods. New York: John
Wiley and Sons; 1966.
[33] Sapountzakis EJ, Katsikadelis JT. Unilaterally supported plates on
elastic foundations by the boundary element method. Journal of Applied
Mechanics Transactions of ASME 1992;59:5806.
[34] MSC/NASTRANfor Windows, Finite element modeling and postprocess-
ing system. Help system index, version 4.0. USA; 1999.
[35] Mondkar DP, Powell GH. Finite element analysis of nonlinear static
and dynamic response. International Journal for Numerical Methods in
Engineering 1977;11:499520.