x
; S
y
rst moment of area above the x and y axes respectively
u
i
trial functions
x, y, z global coordinate system
d thickness of the beam
/
x
, /
y
bending rotations with respect to x and y axes
c
mx
, c
my
mean values of shear strains over a cross-section
z = constant
h rotation of the cross-section around its shear centre
q specic mass
q
x
, q
y
radii of curvature
n, g, f displacements in the directions x, y and z respectively
x angular frequency
Fig. 1. Denitions of terms.
F. de Borbn et al. / Computers and Structures 89 (2011) 14061416 1407
The physical model developed by Ambrosini [4,5] is formed by
the following three fourth order partial differential equations in
the generalized displacements n, g and h:
EJ
y
@
4
n
@z
4
@
3
c
mx
@z
3
_ _
qJ
y
@
4
n
@z
2
@t
2
@
3
c
mx
@z@t
2
_ _
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
q
x
;
1a
EJ
x
@
4
g
@z
4
@
3
c
my
@z
3
_ _
qJ
x
@
4
g
@z
2
@t
2
@
3
c
my
@z@t
2
_ _
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
q
y
;
1b
EJ
u
@
4
h
@z
4
qJ
u
@
4
h
@z
2
@t
2
qF
T
a
y
@
2
n
@t
2
a
x
@
2
g
@t
2
r
2
@
2
h
@t
2
_ _
GJ
d
@
2
h
@z
2
m
t
:
1c
In these equations, F
T
is the cross-sectional area, J
x
and J
y
are the
second moments of area of the cross-section in relation to the cent-
roidal principal axes, J
u
the sectorial second moment of area (warp-
ing constant), J
d
the torsion constant, a
x
and a
y
the coordinates of
the shear centre. q denotes the mass density of the beam material.
E and G are the Youngs and the shear modulus respectively. q
x
, q
y
and m
t
represent the externally applied loads per unit length in
the x, y and h directions, respectively. c
mx
and c
my
represent the
mean values of shear strains over a cross-section z = constant given
by:
c
mx
Q
x
k
x
F
T
G
; 2a
c
my
Q
y
k
y
F
T
G
; 2b
where Q
x
and Q
y
are the shear stress resultants on the cross section.
k
x
and k
y
denote the Cowpers shear coefcients that could be ob-
tained using the approximate equations given by Ebner and Billing-
ton [2]:
k
x
J
x
d
S
x
F
T
; 3a
k
y
J
y
d
S
y
F
T
; 3b
where S
x
and S
y
are the rst moment of area above the axis x and y
respectively and d is the thickness of the beam. It must be pointed
out that in the case of a doubly asymmetric cross-section, the prin-
cipal shear axes do not coincide with the principal bending ones
and, for this reason, the deection components in the x and y direc-
tions are in general coupled. However, in the present formulation,
the cross-shear coefcients, k
xy
and k
yx
, are neglected considering
that its inuence is secondary.
Finally,
r
2
a
2
x
a
2
y
J
x
J
y
F
T
: 4
In this paper, in order to incorporate the effect of an axial load to
Eqs. (1), a longitudinal force P is applied at an arbitrary point e
x
, e
y
.
If M
x
and M
y
are bending moments produced by the eccentric
force P (Fig. 2), the normal stress produced by P at any point of
coordinates x and y of the cross-section is given by:
r
p
P
F
T
M
y
J
y
x
M
x
J
x
y; 5
On the other hand, the displacement of any point M of the cross-
section is determined by:
n
M
n y a
y
h; 6a
g
M
g x a
x
h; 6b
where h is the angle of rotation around the pole.
The normal forces r
p
d
s
acting in the cross-section of an elemen-
tary strip, are projected on the directions of the axes x and y as the
result of the bending deformations. Denoting the magnitudes of
these projections p
x
and p
y
and considering that the normal stresses
do not vary throughout the beam we obtain from Fig. 2
p
x
dzds r
p
dds
dz
q
x
; 7a
p
y
dzds r
p
dds
dz
q
y
; 7b
where q
x
and q
y
are the radii of curvature of the projections of n
M
and g
M
on the coordinate planes coordinate O
xz
and O
yz
during the
deformation.
These curvatures can be expressed like
1
q
x
@
2
n
M
@z
2
@c
mx
@z
; 8a
1
q
y
@
2
g
M
@z
2
@c
my
@z
: 8b
Fig. 2. Axial load projections.
1408 F. de Borbn et al. / Computers and Structures 89 (2011) 14061416
Substituting (6) and (8) into (7) yields:
p
x
r
p
d
@
2
n
@z
2
y a
y
@
2
h
@z
2
@c
mx
@z
_ _
; 9a
p
y
r
p
d
@
2
g
@z
2
x a
x
@
2
h
@z
2
@c
my
@z
_ _
: 9b
The loads per unit length can be achieved performing integration of
p
x
and p
y
on ds:
q
px
_
r
p
d
@
2
n
@z
2
y a
y
_ _
@
2
h
@z
2
@c
mx
@z
_ _
ds; 10a
q
py
_
r
p
d
@
2
g
@z
2
x a
x
@
2
h
@z
2
@c
my
@z
_ _
ds; 10b
m
p
_
r
p
d
@
2
n
@z
2
y a
y
@
2
h
@z
2
@c
mx
@z
_ _
y a
y
ds
_
r
p
d
@
2
g
@z
2
x a
x
@
2
h
@z
2
@c
my
@z
_ _
x a
x
ds: 10c
Substituting (5) on (10) and keeping in mind that dds = dF and that
S
x
= S
y
= J
xy
= 0 in principal axes:
q
xp
P
@
2
n
@z
2
@c
mx
@z
_ _
@
2
h
@z
2
M
x
Pa
y
; 11a
q
yp
P
@
2
g
@z
2
@c
my
@z
_ _
@
2
h
@z
2
M
y
Pa
x
; 11b
m
p
@
2
n
@z
2
@c
mx
@z
_ _
M
x
Pa
y
@
2
g
@z
2
@c
my
@z
_ _
M
y
Pa
x
@
2
h
@z
2
Pr
2
M
y
b
x
M
x
b
y
_ _
; 11c
where b
x
and b
y
are geometric characteristic.
b
x
U
y
2J
y
a
x
; 12a
b
y
U
x
2J
x
a
y
12b
U
x
_
F
y
3
dF
_
F
x
2
ydF; 13a
U
y
_
F
x
3
dF
_
F
y
2
xdF: 13b
The right terms of expression(13) are thirdorder moments andprod-
ucts of inertia of the considered section. Introducing (11) in Eqs. (1)
EJ
y
@
4
n
@z
4
@
3
c
mx
@z
3
_ _
P
@
2
n
@z
2
@c
mx
@z
_ _
M
x
Pa
y
@
2
h
@z
2
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
qJ
y
@
4
n
@z
2
@t
2
@
3
c
mx
@z@t
2
_ _
q
x
; 14a
EJ
x
@
4
g
@z
4
@
3
c
my
@z
3
_ _
P
@
2
g
@z
2
@c
my
@z
_ _
M
y
Pa
x
@
2
h
@z
2
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
qJ
x
@
4
g
@z
2
@t
2
@
3
c
my
@z@t
2
_ _
q
y
; 14b
EJ
u
@
4
h
@z
4
GJ
d
@
2
h
@z
2
@
2
n
@z
2
@c
mx
@z
_ _
M
x
Pa
y
@
2
g
@z
2
@c
my
@z
_ _
M
y
Pa
x
Pr
2
M
y
b
x
M
x
b
y
_ _
@
2
h
@z
2
qF
T
a
y
@
2
n
@t
2
a
x
@
2
g
@t
2
r
2
@
2
h
@t
2
_ _
qJ
u
@
4
h
@z
2
@t
2
m
t
: 14c
Taking into consideration that
M
x
Pe
y
; 15a
M
y
Pe
x
; 15b
the exuraltorsional equations of motion of thin-walled beams
axially loaded are obtained.
EJ
y
@
4
n
@z
4
@
3
c
mx
@z
3
_ _
P
@
2
n
@z
2
@c
mx
@z
_ _
Pe
y
a
y
@
2
h
@z
2
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
qJ
y
@
4
n
@z
2
@t
2
@
3
c
mx
@z@t
2
_ _
q
x
; 16a
EJ
x
@
4
g
@z
4
@
3
c
my
@z
3
_ _
P
@
2
g
@z
2
@c
my
@z
_ _
Pe
x
a
x
@
2
h
@z
2
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
qJ
x
@
4
g
@z
2
@t
2
@
3
c
my
@z@t
2
_ _
q
y
; 16b
EJ
u
@
4
h
@z
4
GJ
d
@
2
h
@z
2
P
@
2
n
@z
2
@c
mx
@z
_ _
e
y
a
y
P
@
2
g
@z
2
@c
my
@z
_ _
e
x
a
x
P r
2
2e
x
b
x
2e
y
b
y
_ _
@
2
h
@z
2
qF
T
a
y
@
2
n
@t
2
a
x
@
2
g
@t
2
r
2
@
2
h
@t
2
_ _
qJ
u
@
4
h
@z
2
@t
2
m
t
: 16c
The system (16a), (16b), (16c) represents a general model of beams
that take into account triply coupled exuraltorsional vibrations.
It must be pointed out that the longitudinal vibration equation re-
lated to the generalized displacement f (Fig. 1) is uncoupled with
the rest of the system (16a), (16b), (16c) and it was not taken in
consideration in the analysis [5,25]. In the case that the longitudi-
nal vibrations are of interest, this equation can be treated
independently.
3. Finite element model
3.1. Variational formulation
The equations of motion (16) are expressed in terms of dis-
placements replacing Eqs. (2) into Eqs. (16), eliminating the shear
strains:
EJ
y
@
4
n
@z
4
1
k
x
F
T
G
@
2
@z
2
@Q
x
@z
_ _
_ _
P
@
2
n
@z
2
1
k
x
F
T
G
@Q
x
@z
_ _
_ _
Pe
y
a
y
@
2
h
@z
2
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
qJ
y
@
4
n
@z
2
@t
2
1
k
x
F
T
G
@
2
@t
2
@Q
x
@z
_ _
_ _
q
x
; 17a
F. de Borbn et al. / Computers and Structures 89 (2011) 14061416 1409
EJ
x
@
4
g
@z
4
1
k
y
F
T
G
@
2
@z
2
@Q
y
@z
_ _
_ _
P
@
2
g
@z
2
1
k
y
F
T
G
@Q
y
@z
_ _
_ _
Pe
x
a
x
@
2
h
@z
2
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
qJ
x
@
4
g
@z
2
@t
2
1
k
x
F
T
G
@
2
@t
2
@Q
y
@z
_ _
_ _
q
y
; 17b
EJ
u
@
4
h
@z
4
GJ
d
P r
2
2e
x
b
x
2e
y
b
y
_ _ _
@
2
h
@z
2
P
@
2
n
@z
2
1
k
x
F
T
G
@Q
x
@z
_ _
_ _
e
y
a
y
P
@
2
g
@z
2
1
k
y
F
T
G
@Q
y
@z
_ _
_ _
e
x
a
x
qJ
u
@
4
h
@z
2
@t
2
qF
T
a
y
@
2
n
@t
2
a
x
@
2
g
@t
2
r
2
@
2
h
@t
2
_ _
m
t
: 17c
On the other hand, considering the equilibrium of forces acting on a
differential segment:
@Q
x
@z
dz qF
T
dxf
ix
; 18a
@Q
y
@z
dz qF
T
dxf
iy
; 18b
where f
ix
and f
iy
represents the distributed transverse inertia forces:
f
ix
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
; 19a
f
iy
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
: 19b
Substituting Eq. (19) into Eq. (18) yields
@Q
x
@z
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
; 20a
@Q
y
@z
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
: 20b
Replacing Eqs. (20) into (17), the equations of motion are obtained
in terms of displacements:
EJ
y
@
4
n
@z
4
P
@
2
n
@z
2
Pe
y
a
y
@
2
h
@z
2
qF
T
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
Pq
k
x
G
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
qEJ
y
k
x
G
@
4
n
@z
2
@t
2
a
y
@
4
h
@z
2
@t
2
_ _
qJ
y
@
4
n
@z
2
@t
2
_ _
q
2
J
y
k
x
G
@
4
n
@t
4
a
y
@
4
h
@t
4
_ _
q
x
; 21a
EJ
x
@
4
g
@z
4
P
@
2
g
@z
2
Pe
x
a
x
@
2
h
@z
2
qF
T
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
Pq
k
y
G
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
qEJ
x
k
y
G
@
4
g
@z
2
@t
2
a
x
@
4
h
@z
2
@t
2
_ _
qJ
x
@
4
g
@z
2
@t
2
_ _
q
2
J
x
k
y
G
@
4
g
@t
4
a
x
@
4
h
@t
4
_ _
q
y
; 21b
EJ
u
@
4
h
@z
4
GJ
d
P r
2
2e
x
b
x
2e
y
b
y
_ _ _
@
2
h
@z
2
P
@
2
n
@z
2
e
y
a
y
P
@
2
g
@z
2
e
x
a
x
Pq
k
x
G
e
y
a
y
@
2
n
@t
2
a
y
@
2
h
@t
2
_ _
Pq
k
y
G
e
x
a
x
@
2
g
@t
2
a
x
@
2
h
@t
2
_ _
qJ
u
@
4
h
@z
2
@t
2
_ _
qF
T
a
y
@
2
n
@t
2
a
x
@
2
g
@t
2
r
2
@
2
h
@t
2
_ _
m
t
: 21c
The Eqs. (21) with the boundary conditions represent the strong
form of the problem. In order to develop a nite element, the differ-
entials equations can be replaced by an equivalent weak form fol-
lowing [26,27]. The previous equations are multiplied by test or
weight functions v
i
and they are integrated in the domain.
_
l
0
v
i
Au
i
dz 0 i 1; . . . ; 3 22
where A(u
i
) represents the differential equations, with u
1
= n, u
2
= g
and u
3
= h.
The weak problem is to nd:
u
i
2 Cin with i 1; . . . ; 3
Bu
i
; v
i
..
Internalvirtualwork
lv
i
..
Externalvirtualwork
8v
i
2 Cin
with i 1; . . . ; 3
Being
Cin u
i
z; t 2 L
2
X;
@u
i
z; t
@z
2 L
2
X and
@
2
u
i
z; t
@z
2
2 L
2
X
_ _
;
Cin
v
i
z; t 2 L
2
X;
@v
i
z; t
@z
2 L
2
X and
@
2
v
i
z; t
@z
2
2 L
2
X
_ _
:
23
While, the B(u
i
, v
i
) operator has stiffness and mass associated terms,
the l(v
i
) has the boundary condition terms.
The chosen trial functions u
i
have the following characteristics:
u
1
nz; t a
1
tnz;
u
2
gz; t a
2
tgz;
u
3
hz; t a
3
thz:
24
For the external loads:
q
x
z; t g
1
tq
x
z;
q
y
z; t g
2
tq
y
z;
m
t
z; t g
3
tm
t
z:
25
It must be noted that in the present FEM formulation, the external
loading need to be of the type of Eq. (25), i.e. with separate variables.
Then, the operators B(u
i
, v
i
) and l(v
i
) have the following
expressions:
Bu
1
; u
3
; v
1
a
1
_
l
0
d
2
v
1
dz
2
EJ
y
d
2
n
dz
2
dz a
1
_
l
0
dv
1
dz
P
dn
dz
dz
a
3
_
l
0
dv
1
dz
Pe
y
a
y
dh
dz
dz
d
2
a
1
dt
2
_
l
0
v
1
qF
T
ndz
d
2
a
3
dt
2
_
l
0
v
1
qF
T
a
y
hdz
d
2
a
1
dt
2
_
l
0
v
1
Pq
k
x
G
ndz
d
2
a
3
dt
2
_
l
0
v
1
Pq
k
x
G
a
y
hdz
d
2
a
1
dt
2
_
l
0
dv
i
dz
qJ
y
dn
dz
dz
d
2
a
1
dt
2
_
l
0
dv
1
dz
qEJ
y
k
x
G
dn
dz
dz
d
2
a
3
dt
2
_
l
0
dv
1
dz
qEJ
y
a
y
k
x
G
dh
dz
dz
d
4
a
1
dt
4
_
l
0
v
1
q
2
J
y
k
x
G
ndz
d
4
a
3
dt
4
_
l
0
v
1
q
2
J
y
k
x
G
a
y
hdz; 26a
1410 F. de Borbn et al. / Computers and Structures 89 (2011) 14061416
Bu
2
; u
3
; v
2
a
2
_
l
0
d
2
v
2
dz
2
EJ
x
d
2
g
dz
2
dz a
2
_
l
0
dv
2
dz
P
dg
dz
dz
a
3
_
l
0
dv
2
dz
Pe
x
a
x
dh
dz
dz
d
2
a
2
dt
2
_
l
0
v
2
qF
T
gdz
d
2
a
3
dt
2
_
l
0
v
2
qF
T
a
x
hdz
d
2
a
2
dt
2
_
l
0
v
2
Pq
k
y
G
gdz
d
2
a
3
dt
2
_
l
0
v
2
Pq
k
y
G
a
x
hdz
d
2
a
2
dt
2
_
l
0
dv
2
dz
qJ
x
dg
dz
dz
d
2
a
2
dt
2
_
l
0
dv
2
dz
qEJ
x
k
y
G
dg
dz
dz
d
2
a
3
dt
2
_
l
0
dv
2
dz
qEJ
x
a
x
k
y
G
dh
dz
dz
d
4
a
2
dt
4
_
l
0
v
2
q
2
J
x
k
y
G
gdz
d
4
a
3
dt
4
_
l
0
v
2
q
2
J
x
k
y
G
a
x
hdz; 26b
Bu
1
; u
2
; u
3
; v
3
a
3
_
l
0
d
2
v
3
dz
2
EJ
u
d
2
h
dz
2
dz a
3
_
l
0
dv
3
dz
GJ
d
dh
dz
dz
a
1
_
l
0
dv
3
dz
Pe
y
a
y
dn
dz
dz
a
2
_
l
0
dv
3
dz
Pe
x
a
x
dg
dz
dz
d
2
a
1
dt
2
_
l
0
v
3
qF
T
a
y
ndz
d
2
a
2
dt
2
_
l
0
v
3
qF
T
a
x
gdz
d
2
a
3
dt
2
_
l
0
v
3
qF
T
r
2
hdz
a
3
_
l
0
dv
3
dz
PR
2
dh
dz
dz a
3
_
l
0
dv
3
dz
qJ
u
dh
dz
dz
d
2
a
1
dt
2
_
l
0
v
3
Pq
k
y
G
ne
y
a
y
dz
d
2
a
3
dt
2
_
l
0
v
3
Pq
k
y
G
a
y
he
y
a
y
dz
d
2
a
2
dt
2
_
l
0
v
3
Pq
k
x
G
ge
x
a
x
dz
d
2
a
3
dt
2
_
l
0
v
3
Pq
k
x
G
a
x
he
x
a
x
dz; 26c
lv
1
a
1
v
1
lEJ
y
d
3
nl
dz
3
a
1
v
1
0EJ
y
d
3
n0
dz
3
a
1
dv
1
l
dz
EJ
y
d
2
nl
dz
2
a
1
dv
2
0
dz
EJ
y
d
2
n0
dz
2
a
1
v
1
lP
dnl
dz
a
1
v
1
0P
dn0
dz
a
3
v
1
lPe
y
a
y
dhl
dz
a
3
v
1
0Pe
y
a
y
dh0
dz
d
2
a
1
dt
2
v
1
lqJ
y
dnl
dz
d
2
a
1
dt
2
v
1
0qJ
y
dn0
dz
d
2
a
1
dt
2
v
1
l
qEJ
y
k
x
G
dnl
dz
d
2
a
3
dt
2
v
1
l
qEJ
y
k
x
G
a
y
dhl
dz
d
2
a
1
dt
2
v
1
0
qEJ
y
k
x
G
dn0
dz
d
2
a
3
dt
2
v
1
0
qEJ
y
k
x
G
a
y
dh0
dz
g
1
_
l
0
v
1
q
x
dz;
27a
lv
2
a
2
v
2
lEJ
x
d
3
gl
dz
3
a
2
v
2
0EJ
x
d
3
g0
dz
3
a
2
dv
2
l
dz
EJ
x
d
2
gl
dz
2
a
2
dv
2
0
dz
EJ
x
d
2
g0
dz
2
a
2
v
2
lP
dgl
dz
a
2
v
2
0P
dg0
dz
a
3
v
2
lPe
x
a
x
dhl
dz
a
3
v
2
0Pe
x
a
x
dh0
dz
d
2
a
2
dt
2
v
2
lqJ
x
dgl
dz
d
2
a
2
dt
2
v
2
0qJ
x
dg0
dz
d
2
a
2
dt
2
v
2
l
qEJ
x
k
y
G
dgl
dz
d
2
a
3
dt
2
v
2
l
qEJ
x
k
y
G
a
x
dhl
dz
d
2
a
2
dt
2
v
2
0
qEJ
x
k
y
G
dg0
dz
d
2
a
3
dt
2
v
2
0
qEJ
x
k
y
G
a
x
dh0
dz
g
2
_
l
0
v
2
q
y
dz; 27b
lv
3
a
3
v
3
lEJ
u
d
3
hl
dz
3
a
3
v
3
0EJ
u
d
3
h0
dz
3
a
3
dv
3
l
dz
EJ
u
d
2
hl
dz
2
a
3
dv
3
0
dz
EJ
u
d
2
h0
dz
2
a
3
v
3
lGJ
d
dhl
dz
a
3
v
3
0GJ
d
dh0
dz
a
1
v
3
lPe
y
a
y
dnl
dz
a
1
v
3
0Pe
y
a
y
dn0
dz
a
2
v
3
l
Pe
x
a
x
dgl
dz
a
2
v
3
0Pe
x
a
x
dg0
dz
a
3
v
3
l
PR
2
dhl
dz
a
3
v
3
0PR
2
dh0
dz
d
2
a
3
dt
2
v
3
lqJ
u
d
2
hl
dz
2
d
2
a
3
dt
2
v
3
0qJ
u
d
2
h0
dz
2
g
3
_
l
0
v
3
mdz: 27c
3.2. Beam element
When the axial load, shear deformations and rotatory inertias
are null, the operator B(u
i
,v
i
) is reduced to the operator for linear
theory of beams. For that theory, Hermite cubic polynomials are
chosen as shape functions (Fig. 3). From the bilinear operator, it
can be seen that all terms have equal or less exigency of derivation
than the linear beam theory, then Hermite cubic polynomials are
chosen as shape functions as well.
Even more, these polynomials are adopted to interpolate the
displacements of the test and weight functions. The shape func-
tions for an element of length h
e
are:
N
1
s
1
4
2 s1 s
2
;
N
2
s
h
e
8
1 s1 s
2
;
N
3
s
1
4
2 s1 s
2
;
N
4
s
h
e
8
s 11 s
2
;
28
s
2z
h
e
1 1 < s < 1: 29
The degrees of freedom of an element of length h
e
are dened in
Fig. 4.
Then, the displacements of an element are dened by:
u
i
v
i
N
e
i
a
e
i
; 30
a
e
a
1
n; a
2
n
0
; a
3
g; a
4
g
0
; a
5
h; a
6
h
0
; a
7
n; a
8
n
0
; a
9
g; f
a
10
g
0
; a
11
h; a
12
h
0
g; 31
where:
N
e
i
= Shape functions matrix for an element of length h
e
a
e
i
= Nodal vector with the degrees of freedom of the element.
Using the following nomenclature to indicate second and rst
derivate of the shape functions
B
e
@
2
N
e
1
s
@s
2
@
2
N
e
2
s
@s
2
@
2
N
e
3
s
@s
2
@
2
N
e
4
s
@s
2
_ _
ds
dz
_ _
2
;
G
e
@N
e
1
s
@s
@N
e
2
s
@s
@N
e
3
s
@s
@N
e
4
s
@s
_ _
ds
dz
_ _
:
32
The equilibrium equations can be expressed in matrix form
(Zienkiewicz and Taylor [27]).
nel
i1
a
i
tK
e
ij
d
2
a
i
t
dt
2
M
e
1ij
d
4
a
i
t
dt
4
M
e
2ij
_ _
a
e
_ _
nel
i1
pgta
e
: 33
When the external loads are p = 0, a
j
has the following solution:
F. de Borbn et al. / Computers and Structures 89 (2011) 14061416 1411
a
j
e
ixt
j 1; . . . ; 3: 34
Replacing (34) into (33) and as the nodal vector of displacements a
e
is equal for all the elements
K
ij
x
2
M
1ij
x
4
M
2ij
_ _
a 0: 35
The Eq. (35) represents a classic eigenvalue problem.
Finally, the expressions of the element stiffness and mass
matrices are the following ones:
The mass matrix is shown like a sum of three matrices only with
the purpose of clarifying its content. The rst matrix has the terms
of mass themselves. The second has the terms that affect the accel-
erations originated by the shear deformations. Finally, the third
matrix has terms that affect the accelerations originated by the
shear deformations but associated to fourth order time derivative.
In slender beams, this last matrix can be neglected (Clough and
Penzien [28]).
The stiffness and mass matrices have been ordered showing the
exural torsional coupled terms. The matrices are solved using
numerical integration, in this case Gauss quadrature.
3.3. The eigenvalue problem
The eigenvalue problem is:
K kM
1
kM
2
k
2
M
3
_ _
a 0; 40
where:
k x
2
: 41
If it is called:
b ka: 42
The Eq. (35) can be written:
Ka kaM
1
kaM
2
kbM
3
0: 43
The Eq. (38) can be expressed in matrix form
K 0
0 I
_ _
a
b
_ _
k
M
1
M
2
M
3
I 0
_ _
a
b
_ _
0: 44
37
38
39
36
Fig. 3. Shape functions. Hermite cubic polynomials.
1412 F. de Borbn et al. / Computers and Structures 89 (2011) 14061416
The angular frequency is obtained
x
k
p
45
and frequency
f
x
2p
: 46
4. Numerical solution
The theory presented in this paper was implemented in a nite
element program taking as a base the structure presented by Fish
and Belytschko [29]. The mass and stiffness matrices presented in
this paper can be easily obtained and programmed in any compu-
tational frame (e.g. MATLAB) independently of the boundary
conditions.
4.1. Example 1. Experimental validation
In the rst example, presented in [30], experimental results
were compared with a state variables approach and with a com-
mercial nite element code with 900 shell elements (SAP2000
[31]). The cross-section is shown in Fig. 5 and the results obtained
with the beam element developed are presented in Table 1. The
clampedclamped beam was modeled with 20 beam elements.
The beam has the following geometrical and mechanical
properties:
F 3:12e
4
m
2
; q 2650:0 kg=m
3
;
J
x
3:024e
8
m
4
; J
y
4:595e
8
m
4
;
J
d
4:270e
10
m
4
; J
u
2:838e
11
m
6
;
a
x
2:67e
2
m; a
y
1:30e
2
m;
E 4:5e
10
N=m
2
; m 0:25;
l 2:00 m; P 0:
Because of the slenderness of the beam, the shear deformations are
not considered.
The differences between the numerical and experimental re-
sults are not superior to 3%, which is acceptable. The values
achieved with 20 beam elements are better than those obtained
with shell elements even when the shell model is rened. Then,
a lower computational cost is achieved.
4.2. Example 2. Numerical validation
A non symmetric cross section beam is analyzed [8]. The beam
is simply supported. The cross-section is shown in Fig. 6 and
numerical results are presented in Table 2. The beam is modeled
with 10 elements. The beam lengths are 3, 10 and 50 m.
The geometrical and mechanical properties are:
F 2:4e
4
m
2
; q 8000:0 kg=m
3
;
J
x
6:658e
4
m
4
; J
y
1:821e
4
m
4
;
J
d
5:20e
6
m
4
; J
u
3:943e
6
m
6
;
a
x
0:0548 m; a
y
0:0865 m;
E 2:1e
8
kN=m
2
; G 8:07e
7
kN=m
2
;
k
x
0:474; k
y
0:306;
P 0:
Table 1
Natural frequencies of the beam, Example 1.
Mode Frequencies (Hz)
Test FEM Beam 20 elem Diff (%) Ref. [26] 60 elem Diff (%) Ref. [26] 900 elem Diff (%)
1 Flexural vertical 24.5 24.8 1.21 22.2 9.4 24.7 0.8
2 Flexural lateral 49.1 47.8 2.72 40.5 17.5 44.9 8.6
3 Flexural vertical 63.5 63.3 0.35 55.2 13.1 61.9 2.5
4 Flexural vertical 122.5 119.1 2.85 102.8 16.1 114.1 6.9
Fig. 5. Cross-section example 1.
Fig. 4. Degrees of freedom.
Fig. 6. Cross-section example 2.
F. de Borbn et al. / Computers and Structures 89 (2011) 14061416 1413
Even when the approaches of Ref. [8] and this paper are different,
the numerical results are in excellent agreement for every beam
length. This is a remarkable fact because the nite element model
is compared with analytical results.
4.3. Example 3. Validation of axial loading
To validate the model with an axial load, a problem studied by
Jun et al. [15] among other authors was chosen. The cross-section
is shown in Fig. 7. The axial load is applied in the centroid. The
shear deformations were not considered.
The geometrical and mechanical properties are:
Table 2
Natural frequencies of the beam, Example 2.
Mode Frequency [Hz]
L = 3 m L = 10 m L = 50 m
FEM Ref. [8] Diff. (%) FEM Ref. [8] Diff. (%) FEM Ref. [8] Diff. (%)
1 Flexural lateral torsional 58.98 57.97 0.02 6.74 6.74 0.02 0.28 0.28 0.00
2 Flexural vertical torsional 90.38 90.41 0.03 11.85 11.85 0.02 0.53 0.53 0.14
3 Flexural lateral torsional 139.41 139.75 0.25 15.75 15.75 0.01 1.12 1.12 0.06
4 Flexural vertical torsional 205.27 205.28 0.00 23.91 23.90 0.02 2.12 2.12 0.03
5 Flexural lateral torsional 325.51 326.46 0.29 35.89 35.89 0.01 2.49 2.49 0.03
6 Flexural vertical torsional 435.35 441.53 1.40 48.22 48.19 0.06 2.51 2.51 0.04
Fig. 7. Cross-section example 3.
Table 3
Natural frequencies of the beam axially loaded, Example 3.
Mode Frequency [Hz] P = 1790 N
Simply supported Cantilever Clampedclamped
FEM Ref. [15] Diff. (%) FEM Ref. [15] Diff. (%) FEM Ref. [15] Diff. (%)
1 Flexural lateral 84.52 84.69 0.20 24.99 25.01 0.08 196.64 196.55 0.05
2 Flexural torsional 147.12 147.77 0.44 61.16 61.31 0.25 199.31 199.91 0.30
3 Flexural torsional 316.60 319.07 0.78 136.33 136.15 0.13 420.46 420.89 0.10
4 Flexural lateral 350.47 352.62 0.61 191.24 192.62 0.72 548.15 554.53 1.16
5 Flexural torsional 360.40 361.42 0.28 275.04 275.03 0.00 605.94 616.77 1.79
6 Flexural torsional 599.27 598.16 0.19 476.29 479.40 0.65 690.18 690.47 0.04
Fig. 8. Mode-shapes example 3.
1414 F. de Borbn et al. / Computers and Structures 89 (2011) 14061416
F 3:08e
4
m
2
; q 2711:0 kg=m
3
;
J
x
9:26e
8
m
4
; J
y
1:77e
8
m
4
;
J
d
1:64e
9
m
4
; J
u
1:52e
12
m
4
;
a
x
1:55e
2
m; a
y
0:00 m;
E 6:89e
10
N=m
2
; m 0:30;
l 0:82 m; P 1790:0 N:
Three boundary conditions were analyzed: Simply supported, canti-
lever and clampedclamped. For the two rst cases, 4 elements
were used. The clampedclamped case was analyzed with 10 ele-
ments. The results are presented in Table 3. The numerical results
are in good agreement of those of Li et al. [15] for all boundary
conditions.
Finally, the mode-shapes obtained for this case are presented in
Fig. 8.
4.4. Example 4. Asymetrical cross section with axial load
In this example, a non symmetric cross section with axial load is
analyzed. The axial load is applied at the centroid. The beam is
modeled with 10 elements. The geometrical and mechanical prop-
erties are those of the rst example, but in order to consider shear
deformations the beam length is 1 m. The Cowpers shear coef-
cients obtained with (3) are k
x
= 0.0945 and k
y
= 0.2051. Two
boundary conditions are considered: clamped and cantilever. In
both cases, the axial load is a compression load of 2000 N. The re-
sults are presented in Table 4.
From Table 4, it can be seen that the inuence of the axial load
cannot be neglected in the rst mode of vibration for both load
conditions, in particular for the cantilever beam. It is also observed
that the compression load diminishes the value of the frequency.
The fact of not considering the axial load in cantilever beam leads
to a difference of 54% in the value of the frequency. This inuence
is minor in the superior modes.
5. Conclusions
The triple exural torsional coupled equations of motion for
thin walled and open cross-section axially loaded beams are pre-
sented. The axial load can be applied in an arbitrary point of the
cross-section. A beam element is developed and the element mass,
elastic stiffness and geometrical stiffness matrices of unsymmetri-
cal cross-section beams are presented.
Several numerical examples are analyzed concluding that the
numerical results are in good agreement with those of referenced
work. The developed beam element is proved to be an efcient -
nite element reducing the computational cost and could be easily
incorporated to any commercial or developing software. It must
be emphasized that, although only free vibration examples are pre-
sented, the presented formulation (see Eq. (15)) allows the consid-
eration of external dynamic loading.
Acknowledgements
The nancial support of the CONICET and the Universidad Nac-
ional de Cuyo is gratefully acknowledged. Special acknowledge-
ments are extended to the reviewers of the rst version of the
paper, since their useful suggestions led to substantial improve-
ments of the nal version of the paper.
References
[1] Pestel E, Leckie F. Matrix methods in elastomechanics. New York: Mc Graw-
Hill; 1963.
[2] Ebner A, Billington D. Steady state vibrations of damped Timoshenko beams. J
Struct Div ASCE 1968;94:73760.
[3] Vlasov V. Thin-walled elastic beams. 2nd ed. Jerusalem: Israel Program for
Scientic Translations; 1963.
[4] Ambrosini D, Riera JD, Danesi R. Dynamic analysis of thin-walled and variable
open section beams with shear exibility. Int J Numer Meth Eng
1995;38:286785.
[5] Ambrosini D, Riera JD, Danesi R. A modied Vlasov theory for dynamic analysis
of thin-walled and variable open section beams. Eng Struct 2000;22:890900.
[6] Tanaka M, Bercin A. Free vibration solution for uniform beams of
nonsymmetrical cross-section using mathematica. Comput Struct
1999;71:18.
[7] Bishop R, Cannon S, Miao S. On coupled bending and torsional vibration of
uniform beams. J Sound Vib 1989;131:45764.
[8] Prokic A. On vefold coupled vibrations of Timoshenko thin-walled beams. Eng
Struct 2006;28:5462.
[9] Kim N, Kim M. Exact dynamic/static stiffness matrices of nonsymetric thin-
walled beams considering coupled shear deformation effects. Thin Walled
Struct 2005;43:70134.
[10] Rafezy B, Howson H. Exact dynamic stiffness matrix for a thin-walled beam of
doubly asymmetric cross-section lled with shear sensitive material. Int J
Numer Meth Eng 2007;69:275879.
[11] Yoon K, Kang Y, Choi Y, Park N. Free vibration analysis of horizontally curved
steel I-girder bridges. Thin Walled Struct 2005;43:67999.
[12] Piovan M, Cort nez V. Mechanics of thin-walled curved beams made of
composite materials, allowing for shear deformability. Thin Walled Struct
2007;45:75989.
[13] Kim N, Yun H, Kim M. Exact static element stiffness matrices of non symmetric
thin walled curved beam. Int J Numer Meth Eng 2004;61:274302.
[14] Hashemi S, Richard M. Free vibrational analysis of axially loaded bending-
torsion coupled beams: a dynamic nite element. Comput Struct
2000;77:71124.
[15] Jun L, Wanyou L, Rongying S, Hongxing H. Coupled bending and torsional
vibration of nonsymmetrical axially loaded thin-walled BernoulliEuler
beams. Int J Mech Sci 2004;31:697711.
[16] Arpaci A, Bozdag E. On free vibration analysis of thin-walled beams with non
symmetrical open cross-sections. Comput Struct 2002;80:6915.
[17] Kim M, Yun H, Kim N. Exact dynamic and static element stiffness matrices of
nonsymmetric thin-walled beam-columns. Comput Struct 2003;81:142548.
[18] Viola E, Ricci P, Aliabadi M. Free vibration analysis of axially loaded cracked
Timoshenko beam structures using the dynamic stiffness method. J Sound Vib
2007;304:12453.
[19] Chen H, Hsiao K. Quadruply coupled linear free vibrations of thin walled
beams with a generic open section. Eng Struct 2008;30:131934.
[20] Back S, Will K. Shear-exible thin-walled element for composite I-beams. Eng
Struct 2008;30:144758.
[21] Sapountzakis E, Mokos V. A BEM solution to transverse shear loading of beams.
Comput Mech 2005;36(5):38497.
[22] Sapountzakis E, Mokos V. 3-d beam element of composite cross section
including warping and shear deformation effects. Comput Struct
2007;85:10216.
[23] Sapountzakis E, Dourakopoulos J. Shear deformation effect in exural
torsional vibrations of beams by BEM. Acta Mech 2009;203(3):197221.
[24] Erkmen R, Mohareb M. Buckling analysis of thin walled open members: a nite
element formulation. Thin Walled Struct 2008;46(6):61836.
Table 4
Natural frequencies of a non symmetric beam axially loaded, Example 4.
Mode Frequencies (Hz)
Clamped Cantilever
P = 0 N P = 2000 N Diff (%) P = 0 N P = 2000 Diff (%)
1 Flexural vertical 87.52 83.29 4.83 18.67 8.56 54.15
2 Flexural lateral 182.51 180.61 1.04 34.19 30.24 11.55
3 Flexural vertical 233.72 227.96 2.46 84.69 80.49 4.96
4 Flexural vertical 449.21 443.07 1.37 93.44 85.61 8.38
F. de Borbn et al. / Computers and Structures 89 (2011) 14061416 1415
[25] Ambrosini, R. Dynamic analysis of thin walled and open section in the
frequency domain. Masters Thesis, Universidad Nacional de Tucuman,
Argentina, 1991. (in Spanish).
[26] Becker EB, Carey GF, Oden JT. Finite elements an introduction, vol. 1. New
Jersey: Prentice Hall; 1981.
[27] Zienkiewicz O, Taylor R. The nite element method. McGraw-Hill; 1994. 4th
ed.
[28] Clough R, Penzien J. Dynamics of structures. Singapore: McGraw- Hill; 1975.
[29] Fish J, Belytschko T. A rst course in nite elements. England: John Wiley and
Sons Ltd; 2007.
[30] Ambrosini D. On free vibration of nonsymmetrical thin-walled beams. Thin
Walled Struct 2009;47:62936.
[31] SAP2000. Integrated nite element analysis and design of structures, v. 7.42.
Computers and Structures Inc.; 2001.
1416 F. de Borbn et al. / Computers and Structures 89 (2011) 14061416