STUDY OF THE DYNAMICS OF A ROTATING FLEXIBLE BEAM AND

A FLEXIBLE PENDULUM
Fernando S. Buezas
a,c
, Marta B. Rosales
b,c
, Rubens Sampaio
d
and Carlos P. Filipich
e
a
Department of Physics, Universidad Nacional del Sur, Baha Blanca, Argentina, fbuezas@[Link]
b
Department of Engineering, Universidad Nacional del Sur, Baha Blanca, Argentina,
mrosales@[Link]
c
CONICET
d
Department of Mechanical Engineering, Ponticia Universidade Catolica do Rio de Janeiro, Rio de
Janeiro, Brazil, rsampaio@[Link]
e
Mechanical Systems Analysis Group, FRBB, Universidad Tecnolgica Nacional, Baha Blanca,
Argentina,clipich@[Link]
Keywords: pendulum, rotating exible beam, gravitational effect, stiffening effect.
Abstract. The dynamic of a exible beam forced by an imposed rotating around an axis perpendicular
to its plane is addressed. Three approaches are dealt with, two related with simplied theories, belonging
to the Strength of Materials and the third one using Finite Elasticity. Within the Strength of Materials
approaches, the governing equations are derived by superposing the deformations and the rigid motion in
the rst model, and by stating the stationarity of the Lagrangian (including rst and second order effects
in order to capture the stiffening due to the centrifugal forces) through Hamiltons principle in the second
one. Two actions are considered: gravity forces (pendulum) and prescribed rotation. The stiffening effect
due to the centrifugal forces is considered only for the beam rotating at high speeds. Comparison of this
too models with the equation of Finite Elasticity is carried out. Energy analysis are performed in order
to obtain information about the quality of the numerical solution.
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (artculo completo)
Eduardo Dvorkin, Marcela Goldschmit, Mario Storti (Eds.)
Buenos Aires, Argentina, 15-18 Noviembre 2010
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
1 INTRODUCTION
In the last decade the problem of a plane rotation of a beam has been studied by several
authors due to the importance of the problem.
Simo and Vu-Quoc showed that the use of linear beam theory results in a spurious loss of
stiffness due to the partial transfer of centrifugal force action to the bending equation. Hence,
one should account for geometric stiffening when the imposed rotation is large. To the best of
the authors knowledge, the rst study of vibration of rotating beams was published by Schil-
hansl Schilhans (1958) who analyzed the bending vibration, assuming steady-state revolution
and negligible Coriolis force. He derived a formula relating the fundamental bending eigenfre-
quency with the angular velocity of revolution.
J.M. Mayo (2004) address the issue comparing different formulations. El-Absy and Shabana
(1997) study the effect of geometric stiffness forces on the stability of elastic and rigid body
modes. A simple rotating beam model is used to demonstrate the effect of axial forces and dy-
namic coupling between the modes and the rigid body motion. The effect of higher order terms
in the inertia forces as the result of including longitudinal displacement caused by bending de-
formation is examined, in that paper, using several models. In J.Y. Liu (2004) the equations
of motion are derived taking into account the foreshortening deformation term, which have ge-
ometric stiffening effect on the rigid-exible coupling dynamics of the system. An inuence
ratio is employed as a criterion to clarify the application range of the conventional linear mod-
eling method, in which the stiffening effect is neglected in Al-Qaisia and Al-Bedoor (2005),
the approach used in that work takes into account the rotating speed and the effect of vibration
amplitude. The free bending vibration of rotating tapered beams is investigated by using the
dynamic stiffness method in J.R. Banerjee (2006). A clamped-free rotating exible robotic arm
is studied in the paper of Fung and Yau (1999). Hamiltons principle is used to derive the equa-
tion of motion of the arm together with the associated boundary conditions and then the power
series method is used to solve the differential problem. The paper J.M. Mayo (2004) reviews
different formulations to account for the stress stiffening or geometric stiffening effect arising
from deections large enough to cause signicant changes in the conguration of the system.
In other paper, El-Absy and Shabana (1997)proposes a simple model which demonstrates that
there are applications in which the reference motion can have strong dependence on the elastic
deformation is used to examine the coupling between the rigid body and elastic modes. All of
this papers model the rotating beam with a theory of Strength of Materials (EulerBernoulli or
Timoshenko equation), taking into account only the transverse strain.
The dynamic of a slender body that is subjected to large displacements can be dealt with
the Strength of Materials (SM) Theory (EulerBernoulli and Timoshenko equation), or directly
with a Theory of Elasticity (Hunter (1983); Truesdell and Noll (1965); Truesdell and Toupin
(1960); Fung (1968)). Almost exclusively, the study of rotational dynamics of a beam is stud-
ied from the viewpoint of EulerBernoulli or Timoshenko equation (SM) for the transverse
displacement and deformation. In this work we go beyond this approach. First we add the lon-
gitudinal deformation of a rotating-beam model according to the theory SM of EulerBernoulli
and, second, we propose a model based on the nite elasticity theory of the rotating beam, which
allows more complex and realistic phenomena such as dry friction and the nonlinear effects due
to large deformations.
In this work, two types of actions applied to the exible beam are considered. Firstly the
beam is only subjected to gravity forces (i.e. the well-known "pendulum", an object that is
attached to a pivot point about which it can swing freely) and secondly a rotation of a section
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4154
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
of the exible beam is imposed at the extreme point (a exible beam subjected to a prescribed
rotation means that the speed of rotation will be imposed at a section, in this case the end from
which it hangs).
Both cases will be tackled by the following three approaches: a) SM with a oating frame,
b) SM via Hamiltons principle including the stiffening contributions and c) Finite Elasticity
(in 2D). In the last case, the constitutive equations issue is dealt with in Truesdell and Noll
(1965)and Fung (1968), using the Piola-Kirchhoff stress tensor that lead to strongly non-linear
equations. The differential equations of motion are developed in Section 2. In Section 3,
these equations are presented in weak form and discretized through nite element via Galerkin
method. The boundary conditions are also discussed since the equations are stated in a La-
grangian conguration in the case of Finite Elasticity. The inuence of the stiffening effect in
the rotating beam motion, in the tree models are compared. Also an analysis of the energy con-
servation is included ([Link] (1993)) which permits the control of the numerical convergence.
We take as a reference NLTE model. In order to compare these results with the literature, we
include a section which compares the motion of a 2D pendulum with the paper of Y. Vetyukov
and H. Irschik (2004). Section 4 concludes the article.
2 STATEMENT OF THE EQUATIONS OF MOTION
In this section the equations of plane motion of a rotating beam are stated and discussed.
Three approaches are presented and compared. The rst two are Strength of Materials (SM)
model for a one dimension continuum in plane motion and the third one, belong to the 2D non-
linear theory of Elasticity (NLTE) that is the beam is a two-dimensions body in planar motion.
Whereas the stiffening effect is included naturally in the NLTE approach, in the SM theory a
second order effect should be considered to take into account the stiffening effect.
2.1 Linear one-dimensional model (SM)
Two models are stated within the SM theory. One of them is constructed by superimposing
a rigid body motion of the beam with small deformation around the rigid conguration J.Y. Liu
(2004); Hunter (1983); J.R. Banerjee (2006). The other, stating the stationarity of the action
(including the second order effects in order to capture the phenomenon of stiffening) through
Hamiltons principle. The effect superposition and the Hamiltons principle models will be
named Model SM1 and Model SM2.
2.1.1 Model SM1
The governing equations of a beam undergoing plane rotation are stated superposing the
equations governing the small deformations of the beam to the ones of the rigid motion J.Y. Liu
(2004); Hunter (1983); J.R. Banerjee (2006). That is, let us suppose that the body motion is
given by the displacement vector u = u
r
+u
d
where u
r
(t) = (u
r
(t), v
r
(t))
T
is the rigid part of
the motion and u
d
(X, t) = (u
d
(X, t), v
d
(X, t))
T
is the part related with the beam deformation.
Then the governing equations are in the oating frame are
EA

2
u
d
X
2
A
_

2
u
r
t
2
+

2
u
d
t
2
_
= f
1
(1)
EI

4
v
d
X
4
+ A
_

2
v
r
t
2
+

2
v
d
t
2
_
= f
2
(2)
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4155
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Figure 1: Geometry of the pendulum. a) Displacement vectors (SM); b) Pendulum scheme in the non-deformed
conguration (NLTE).
where X is the material coordinate xed in the frame, t is the time, E is the modulus of elastic-
ity, A is the cross-sectional area of the beam, I is the second area moment, is the volumetric
density, u = (u, v)
T
is the displacement vector, u and v are the longitudinal and transverse dis-
placement components, and f
1
and f
2
are the normal and transversely applied forces (gravity
components). Now u
r
should be obtained from the equations governing the rigid motion to be
replaced in Eq. 2 in order to solve the problem. For instance, in the case of a pendulum of
length L and gravity g (see Figure 1), one obtains

+
3
2L
g sin = 0, this equation gives (t),
and then, u
r
= X(sin 1) and v
r
= X cos can be computed. For example, if Pis a point
of rigid at X then
P = X (sin , cos )
T
= X

i
then

P = X
_

i +

j
_
were

i and

j are the versors that generate the mobile frame. Then

2
u
r
t
2
= X

2
;

2
v
r
t
2
= X

2.1.2 Model SM2

This model is derived by stating the action. The kinematic transformation equations (Figure
1 a) are
x(X, t) = X sin + u(X, t) sin + v(X, t) cos (3)
y(X, t) = X cos u(X, t) cos + v(X, t) sin (4)
The following four energies contributions are introduced:
2W
1
= EA
_
L
0
_
u
X
_
2
dX + EI
_
1
0
_

2
v
X
2
_
2
dX; (5)
2W
2
=
_
(A)
_
1
0

_
v
X
_
2
dX dA (6)
2K = A
_
1
0
_
x
2
+ y
2
_
dX ; P = g
_
(A)
_
1
0
y dX dA (7)
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4156
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
where u, v, x, y are functions of (X, t), W
1
is the strain energy due to axial and bending
deformations, K is the kinetic energy, P is the gravitational potential energy and W
2
is the
internal work done by the axial stress is that arise from the centrifugal effect and the change of
length due to the bending deformation. The stress due to the centrifugal effect is =
2
(L
2

X
2
)/2 . This contribution is well-known within the classical Strength of Materials is a second
order effect, or P effect (yet linear). Bleich and Ramsey (1952) name this contribution as
potential energy of the axial loads. Consequently the Lagrangian is L =K (W
1
+ W
2
+ P)
and with this, the Hamiltons principle
_
t
2
t
1
Ldt = 0 gives the equations of motion
k
L

2
u
X
2

_
u
2
u 2 v
_
= F(X, t) (8)
k
v

4
v
X
4
+
_
v
2
v 2 u
_
+
2
_
X
2
L
2
2

2
v
X
2
+ X
v
X
_
= G(X, t) (9)
where =

, k
L
= E/, k
v
= (EI) / (A), F(X, t) = (
2
X + g cos t), and G(X, t) =
g sin(t).
2.2 Two dimensional model with nite deformations (NLTE)
2.2.1 Equations of motion
In this Section the equations of the elastic body, in two dimensions for nite displacements
and deformations, are stated within the framework of Continuum Mechanics in the Lagrangian
form, or material representation. Some advantages over the Eulerian form, or spatial represen-
tation, result in this problem. Now, if the problem is given in the Lagrangian form, the only
vectorial equation of motion to be solved is
P+
0
b =
0
A (10)
where P is the rst Piola-Kirchhoff stress tensor and P is the divergence of P calculated in
material coordinatesTruesdell and Noll (1965),
0
is the mass density in the reference (already
known), b are the body forces and A = V/t =
2
x/t
2
where x = (x
1
, x
2
)
T
is the position
vector (spatial coordinate of material point X) and A is the acceleration eld that is simply
the partial derivative of the velocity eld V). The boundary conditions are imposed over the
reference (always known), which together with the initial conditions and the equation of motion,
yield a determined problem. All the non-linearities are transferred to the non-symmetric P
tensor. The next relationship relates P and (the Cauchy stress tensorspatial description)
P =det(F) F
T
were F is the deformation gradient F
ij
= x
i
/X
j
2.2.2 Constitutive equation
The following constitutive law between the second Piola-Kirchhoff stress tensor S (symmet-
ric) (P = FS)) and nite strain tensor
E =
1
2
_
F
T
F I
_
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4157
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
is proposed Fung (1968)
S(E) = tr(E)I + 2E (11)
in which and are Lames-type constants, = E

/(1+)(12) , = E

/2(1+) and E

and are constants. Eq. (11) is also known as St. Venant-Kirchhoff material model Truesdell
and Noll (1965). Additional alternatives of possible constitutive equations are discussed in
Filipich and RosalesFilipich and Rosales (2000).
As we can see, E is a function of the derivatives of x, Then by Eq.(11) S is also function of
the derivatives of x. And the same for the tensor P, Then Eq.(10) is
P(x(X, t)) +
0
b =
0

2
x(X,t)
t
2
and the goal of this problem is to nd the position vector x (or displacement vector u = xX)
for all X and t subject to the boundary conditions discussed in the next section.
3 NUMERICAL IMPLEMENTATION
The following subsections describe the numerical scheme implemented to solve the equa-
tions of motion.
3.1 Variational formulation and Galerkin approximation of model SM1
Let
j
=
_

1
j
,
2
j
_
T
be a nite element basis (admissible functions) i.e. Adm is a set of trial
functions dened as
Adm =
_

j
|
j
(0) = 0 and
_
L
0
_

1
j
X
_
2
dX < ,
_
L
0
_

2
j
X
2
_
2
dX <
_
Multiplying Eq.(1) by
1
j
and integrating by parts, we obtain
EA
u
d
X

1
j

L
0

_
L
0
_
EA
u
d
X

1
j
X
+ Ap
1

1
j
_
dX = 0 (12)
were p
1
=

2
u
r
t
2
+

2
u
d
t
2

f
1
A
. In the same way for Eq.(2)
EI
_

3
v
d
X
3

2
j

L
0


2
v
d
X
2

2
j
X

L
0
_
+
_
L
0
_
EI

2
v
d
X
2

2
j
X
2
+ Ap
2

2
j
_
dX = 0 (13)
were p
2
=

2
v
r
t
2
+

2
v
d
t
2

f
2
A
.
In the case of a rotating beam, in the pivot point, X = 0, the function
j
=
X

j
= 0, j,
and in the free end, X = L, the functions
u
d
X
=

3
v
d
X
3
=

2
v
d
X
2
= 0, then
EA
u
d
X

1
j

L
0
= 0
EI

3
v
d
X
3

2
j

L
0
= 0
EI

2
v
d
X
2

2
j
X

L
0
= 0
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4158
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Then, the goal of this problem is to nd u
d
= (u
d
, v
d
)
T
such that
_

_
u
d
, v
d
Adm

1
j
Adm,
_
L
0
_
EA
u
d
X

1
j
X
+ Ap
1

1
j
_
dX = 0

2
j
Adm,
_
L
0
_
EI

2
v
d
X
2

2
j
X
2
+ Ap
2

2
j
_
dX = 0
Now, expanding u and v in terms of
i
u =
N

i1

1
i
(X)C
i1
(t); v =
N

i1

2
i
(X)C
i2
(t) (14)
and replacing in Eq.(12) and Eq.(13) we nd the matrix form
M

C
1
+K
1
C
1
= Q
1
M

C
2
+K
2
C
2
= Q
2
where
M
ij
= A
_

j
dX; = 1, 2
K
1ij
= EA
_

1
i
X

1
j
X
dX; K
2ij
= EI
_

2

2
i
X
2

2
j
X
2
dX
Q
1j
= EA
u
d
X

1
j

L
0
+
_ _
f
1
A

2
u
r
t
2
_

1
j
dX
Q
2j
= EI
_

3
v
d
X
3

2
j

L
0
+

2
v
d
X
2

2
j
X

L
0
_
+
_ _
f
2
A

2
v
r
t
2
_

2
j
dX
3.2 Galerkin approximation of model SM2
As before, let
j
=
_

1
j
,
2
j
_
T
be a nite element basis with the same properties and using
Eq.(14). Multiplying Eq.(8) by
1
j
and Eq.(9)
2
j
and then integrating by parts, or directly from

_
t
2
t
1
Ldt = 0, we obtain the matrix form
M
_

C
1

2
C
1
_
2A

C
2
+K
1
C
1
= Q
1
M
_

C
2

2
C
2
_
2A

C
1
+(K
2
+K

) C
2
= Q
2
when
M
ij
=
_

j
dX; = 1, 2
A
ij
=
_

1
i

2
j
dX
K
1ij
= k
L
_

1
i
X

1
j
X
dX; K
2ij
=
2
_
_
L
2
X
2
_

2
i
X

2
j
X
dX
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4159
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
K

ij
= k
v
_

2

2
i
X
2

2
j
X
2
dX
Q
1j
= k
L
u
X

1
j

L
0

_
F
1
j
dX
Q
2j
=
_

_
k
v

3
v
X
3
+
_
L
2
X
2
_
v
X
_

2
j

L
0
+ k
v

2
v
X
2

2
j
X

L
0
_
+
_ _
GA

2
v
r
t
2
_

2
j
dX
One more time, in the case of a rotating beam, in the pivot point X = 0 the function
j
=

j
= 0j and in the free end (X = L) the functions
u
X
= k
v

3
v
z
3
+(L
2
X
2
)
v
z
=

2
v
X
2
= 0
then the boundary conditions are
u
X

1
j

L
0
= 0

_
k
v

3
v
z
3
+
_
L
2
X
2
_
v
z
_

2
j

L
0
= 0

2
v
z
2

2
j
z

L
0
= 0
3.3 Variational formulation of the problem NLTE
It is very simple to get the variational formulation of equations of motion. Let W be a
test vector eld (admissible functions) of variables referred to the body in its non deformed
conguration (Lagrangian description). Once again, multiplying the equations of motion by
W and integrating over V
0
we get:
_
( P+
0
b
0
A) W dV
0
=0 (15)
_
V
(t
0
W) dA
0
+
_
[
0
(b A) WP W] dV
0
= 0 (16)
where (16) is obtained after integrating (15) by parts (using Greens formula), t
0
is the stress
vector. Here V is the boundary of volume V . The surface integral is divided in two parts
(V
1
, V
2
). Suppose than the displacement u, and consequently x is prescribed in a part of
the boundarys surface (V
1
) (essential boundary conditions) and the stress is given in the other
part (V
2
). In order to incorporate the boundary conditions
x = x on V
1
(17)
t
0
=

t
0
on V
2
(18)
In the case of non homogeneous essential boundary conditions, the solution x(X) must satisfy
the Eq.(17) on V
1
but the test function W must satisfy the homogeneous essential boundary
condition. Then, in the variational problem (16), the admissible test functions Ware dened as
W(X) Adm
1
=
_
W | W = 0 on V
1
and
_
(W)
2
dV
0
<
_
and the solution x(X)
x(X,t) Adm
2
=
_
x | x = x on V
1
and
_
(P)
2
dV
0
<
_
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4160
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
and the natural boundary conditions are automatically imposed (16). Then the surface integral
in Eq. (16) reduces to
_
V
(t
0
W) dA
0
=
_
V
2

t
0
W dA
0
were

t
0
is the value of the tension t
0
in the boundary.
In the case of a pendulum (beam rotating under gravity) the stress are null on the external
body surface with exception to the pivot point p. On the other hand the problem with prescribed
motion (beam with prescribed rotation with constant velocity), the stress are null on the external
body surface with exception to the the clamped boundary. At these points essential conditions
are imposed. In the Lagrangian description the stress is given by t
0
= P N, where t
0
is the
stress vector of PiolaKirchhoff and N is the normal vector of the surface in the reference
conguration(Figure 1b).
Finally, the variational problem consists to nd the vector x(X, t) implicit in P such that
_
W Adm
1
, nd x Adm
2
, that satises
_
V
2
(

t
0
W) dA
0
=
_
V
[
0
(b x) WP(x) W] dV
0
and the initial conditions
x(X,t
0
)=x
0
(X) x(X,t
0
) = V
0
(X) (19)
3.3.1 Galerkin method and discretization in nite elements
By Adm
1
a base of a subspace of a Hilbert space. In this paper
i
are a shape vector
function. Let the function x(X,t
0
) be expanded in a series of this vectorial functions
i
(X)
x(X, t)
N

i=1

i
(X) c
i
(t) (20)
here c
i
(t) are functions only of time. The admissible vector functions are

i
(X) =[
x
1
i
(X),
x
2
i
(X)]
T
Replacing the Eq.(20) in (10) and integrating on the whole domain:
_
_
P(x) +
0
b
0
_
N

i=1

i
(X) c
i
(t)
__

j
(X) dV
0
= 0 (21)
for j from 1 to n. P(x) means than Piola - Kirchhoff stress tensor is calculated from x(X, t)
through constitutive relations (11).
At last, integrating by parts in the general sense, that is using the theorems of vectorial
calculus (Greens formula) we get
_
V
2
j
_
t
0
(x)
j
_
dA
0
+
_
V
j
_

0
_
b
N

i=1

i
c
i
_

j
(X) P(x)
j
_
dV
j
= 0 (22)
where V
j
is the volume of the j-th element.
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4161
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
If the problem has non homogeneous essential boundary conditions, the approximation are
x(X, t)
N

i=1

i
(X) c
i
(t) +
0
(X)
where
0
is known and
0
= x on V
1
, and that is why x Adm
2
.
3.4 Simulation and results
In this subsection the following numerical examples will illustrate the three approaches, as
listed next:
1. A exible beam rotating in a plane, under gravity action is studied using Model SM1, i.e.
a Strength of Materials approach with superposition of the beam vibrations of the rigid
overall motion.
2. A exible beam subjected to a prescribed rotation is solved using Model SM2, i.e. a
Strength of Materials approach using Hamiltons principle. Since the simulations are
done for high speed rotations, the consideration of the stiffening is essential. It was
introduced by means of a second order term in the governing functional (Eq. 6).
3. A exible beam subjected to both a prescribed rotation and gravity is addressed with
Model NLTE.
Since the rotating bar is exible, a exible beam subjected to a prescribed rotation means that
the speed of rotation will be imposed at a point (in this case the end of which hangs).
When dealing with the linear SM models, elements with a sectional cubic nite element basis
were employed. Instead, a sectional quadratic polynomial basis was employed to discretized the
spatial domain in all the NLTE simulations. Temporal integration was performed using the Gear
method (second-order implicit Backward Difference Formula).
3.4.1 The Pendulum
The rst example deals with a beam with L = 5 m, a square cross-sectional area A = 0.01
m
2
, Youngs-type modulus E = 4 10
7
N m
2
, Poisson-type coefcient = 0.3 and mass
density = 7850 kg/m
3
(Figure 1). The beam is released from a horizontal position with null
velocity and restricted to plane motion under a gravitational eld. Figure 2a shows the beam
motion through the eleven instantaneous congurations during the rst second of the motion,
corresponding to the Model SM1 and Model NLTE with a 2D discretization. Also the energy
variation is depicted in Figure 2b. It is seen that the total energy remains constant, a necessary
condition for the numerical solution since we are dealing with a conservative system. The total
energy E
t
is the sum of the kinetic, elastic strain and potential energies, i.e. E
t
= T + U
e
+ U
g
with T =
1
2
_

0
V V dV
0
, and U
g
= g
_

0
x
2
dV
0
. It can be proved that for the constitutive
law (11), it takes the form of the elastic energy as
U
e
=
_ _

2
tr(E)
2
+ tr(E E)
_
dV
0
The number of nite elements and the time step were adjusted after an error study. For this
purpose the error was dened as follows
error =
100
L
_
V
_
(x
m
1
x
m

1
)
2
+ (x
m
2
x
m

2
)
2
)
_
dV
0
(23)
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4162
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Figure 2: Example 1. Conguration of the bar at different times. Plots at time intervals t = 0.1 s. a) Model
SM1 (dark line) and Model NLTE (2D) (clear line). b) Energy in joules (J) as function of time. (1) Total energy,
(2) kinetic energy, (3) strain energy and (4) gravitational potential energy.
Simulated time (s) CPU time (s)
1D (SM2) 2D (NLTE) Ratio 2D/1D
1 10 158 15.8
2 19 304 16
Table 1: Duration of numerical experiments with 1D and 2D simulations. Time in seconds.
Superscript m denote the present numerical experiment and superscript m

a reference solution.
The reference case was performed with a hundred times more elements and a time step 1/100
smaller than the m case. The aim was to yield results with errors less that 1% in t = 1 s. The
computational, or CPU, time of the reference experiment to simulate 1 s was approximately
12500 s. The duration of other experiments are depicted in Table 1.
3.4.2 Beam subjected to prescribed rotations
The numerical simulation of the dynamics of a beam subjected to prescribed rotations is
now presented. The beam length is L = 1 m, the cross-sectional area is 0.01 m
2
, the prescribed
angular velocity is = 3000 rad/s, the Youngs modulus is E = 2.1 10
11
Nm
2
, the mass
density
0
= 7850 kg/m
3
and Poissons coefcient = 0.3. In particular, a large value of
was assumed in order to obtain large deformations. Thus, the differences between the two
approaches could be made apparent. The beam starts its motion from a zero reference (null
displacement) and its initial velocity is the one resulting from the angular speed. That is, the
beam begins its movement from a horizontal position (x = X) with a speed given by V
1
=
0 and V
2
= X
1
. The results and comparisons are depicted in Figures 3 and 4. The temporal
variation of the coordinate x at the free end of the beam during the motion is plotted in Figure
3 in which the values were found with Model SM1, Model SM2 and Model NLTE (2D). The
curves are qualitatively similar though the response found with SM exhibit larger peaks than
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4163
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Figure 3: Example 2. Comparison of Models SM1, SM2 and NLTE (2D). Temporal variation of coordinate x at
beam tip.
the NLTE model results. It can be observed that the NLTE leads to a stiffer response. Probably
this could be explained due to the choice of a linear Lagrangian constitutive law S(E). If
the second Piola-Kichhoff stress tensor S is transformed into the Cauchy stress tensor , a
stiffening behavior with respect to the SM model would arise. This peculiarity is then, not due
to the rotation event but to the chosen constitutive model. Figure 4 depicts the variation of the
vibration frequency (nondimensionalized with respect to the corresponding frequency at = 0)
when the rotating velocity is increased. The Model SM1 results are shown in dashed lines, the
SM2 are shown in dashed-dot line and the NLTE model in full lines. The results are evidently
very close in the cases SM2 and NLTE. The case SM1 is indifferent to the rotation speed as
follows directly from Eq. (2). Unlike the case SM2 plotted in Fig. 4 , the velocities were
assumed small so as to show that the stiffening is not only consistent with the observed physical
behavior but that it is also consistent with NLTE model, in the range of small deformations. For
example, the difference in the frequencies between SM2 and NLTE when = 50 rad/s are 5%
for the rst mode, 7% for the second one, 11% for the third one, 14% for the fourth one and
11% for the fth one. Similar differences are found when = 0.
3.4.3 Comparison of NLTE model with data published for the case of a pendulum
Results found with the NLTE model of the present work are contrasted with results reported
in Y. Vetyukov and H. Irschik (2004) in which the oating frame of reference approach is used
for the analysis of the in-plane oscillations of a suspended rectangular plate. The approach
and the computations are different in reference Y. Vetyukov and H. Irschik (2004) and in the
present work although the same formal theory is used. In Y. Vetyukov and H. Irschik (2004) the
rotation of the frame attached to the body is represented with one rotational degree of freedom.
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4164
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Figure 4: Comparison of Models SM1, SM2 and NLTE (2D). Variation of the rst ve frequencies with .
Besides, a set of continuous polynomial shape functions form the Ritz approximation of the
deformation of the body. In the present study, no oating frame is employed and a quadratic
nite element basis is chosen for the Galerkin discretization. The plate (or beam) dimensions
are L = 4 m and a cross sectional area of 1 m
2
, mass density
0
=7800 kg/m
3
, = 0.3,
E = 4 10
7
N/m
2
. The beam was released from the horizontal undeformed position and then
freely oscillated in the gravitational eld. An interesting comparison was made regarding with
the rotation angle. Given the approach used in Y. Vetyukov and H. Irschik (2004) the angle that
rotates the beam is an independent variable and then its value is solved for each time instant. In
the present study this angle is not direct result of the calculations. To compare withY. Vetyukov
and H. Irschik (2004) the angle was measured at each instant as the slope of the straight line that
joins the midpoints of both end cross-sections. Figure 5a shows the angle vs. time reported in
Y. Vetyukov and H. Irschik (2004), Fig. 5b plots the resulting angle from the present study and
Fig 5c , the superposition of both results. As can be observed, and notwithstanding the diverse
methodologies, the excellent agreement yields in Figure 5.
4 FINAL COMMENTS AND CONCLUSIONS
This work has presented a brief review of the open literature on the geometric stiffening of
rotating exible beams. Some of the several methodologies proposed in the literature to account
for the stiffening effect in the dynamics equations were analyzed. The dynamics of a exible
beam under gravity (pendulum) and with a prescribed rotation were addressed with models
of Strength of Material (SM) and the nite Elasticity (Model NLTE). The SM approach was
performed with two models, superposition of motions (Model SM1) and Hamiltons principle
(Model SM2). The latter included the stiffening effect. In the case of the superposition model,
the equations resulted partially coupled. That is, only the deformation equations are coupled
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4165
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
Figure 5: Comparison of Models NLTE (present study) and reference [12]. Temporal variation of rotation angle.
a) reference [12]; b) present study; c) superposition of a) and b).
with the rigid motion but the rigid motion equations are uncoupled with the deformations. This
derived from the type of construction. The non-linearity is only present in the rigid body mo-
tion. On the other hand, when applying Hamiltons principle, fully coupled equations arose.
When the beam is subjected only to gravity, model NLTE yielded similar results to the ones
obtained with SM theory (Model SM1). However, in the second example, since a very exi-
ble beam with high rotational velocity was studied, the resulting deformations were not small,
and consequently the response was not identical. Obviously, as more exible is the body more
differences are found between the linear and nonlinear models.
A exible beam undergoing low speed prescribed rotation was also studied. The stiffening
effect in Model SM2 makes it possible to nd almost coincident values of frequencies found
via nite deformation Model NLTE. As is known the vibration frequencies of a rotating beam
increase as the angular velocity does which is associated with the stiffening effect introduced
by the rotation. In the linear one-dimensional theory (Model SM2) the stiffening effect is due
to the contribution of the second order work done by the axial stress caused by the centrifugal
force over the bending deformation. For the general case of the dynamics of the elastic body
considering nite deformation theory of elasticity, it is not necessary to introduce additional
terms in the equation of motion. That is, in the SM2 theory could not only model the effect
of stiffening due to centrifugal force, but when compared with the analogue given by NLTE
can be concluded that the, stiffening law is the correct. In the conservative pendulum case,
the total energy composed of the gravitational, strain and kinetic parts remains constant in
time. This is useful to check that the integration scheme introduces neither numerical damping
nor instabilities in the solutions. The justication for using the full NLTE model, is that it
includes all effects. Obviously the CPU times are larger. Besides it allows to tackle large
deformations that can occur with very exible beams and high rotational speeds. Furthermore
other complexities such a nite dimensional pivot can be tackled by this approach. This case
also includes the phenomenon of friction in the pivot responsible for the so-called stick and
F. BUEZAS, M. ROSALES, R. SAMPAIO, C. FILIPICH 4166
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]
slip. The energy review allows, in this latter case, understanding how energy is dissipated in a
exible pendulum body.
REFERENCES
Al-Qaisia A. and Al-Bedoor B. Evaluation of different methods of the consideration of the
effect of rotation on the stiffening. Journal of Sound and Vibration, 280:531553, 2005.
Bleich F. and Ramsey L. Buckling Strength of Metal Structures. New York,McGraw-Hill, 1952.
El-Absy H. and Shabana A.A. Geometric stiffness and stability of rigid body modes. Journal
of Sound and Vibration, 207(4):465496, 1997.
Filipich C. and Rosales M. A further study on the post-buckling of extensible elastic rods.
International Journal of Nonlinear Mechanics, 35:9971022, 2000.
Fung E.H.K. and Yau D.T.W. Effects of centrifugal stiffening on the vibration frequencies of a
constrained exible arm. Journal of Sound and Vibration, 224(5):809841, 1999.
Fung Y. Foundations of solid mechanics. Prentice-Hall, 1968.
Hunter S.C. Mechanics of continuous media. Ellis Horwood, 1983.
J.M. Mayo D. Garcia-Vallejo J.D. Study of the geometric stiffening effect: Comparison of
different formulations. Multibody System Dynamics, 11:321341, 2004.
J.R. Banerjee H. Su D.J. Free vibration of rotating tapered beams using the dynamic stiffness
method. Journal of Sound and Vibration, 298:10341054, 2006.
J.Y. Liu J.H. Geometric stiffening effect on rigid-exible coupling dynamics of an elastic beam.
Journal of Sound and Vibration, 278:11471162, 2004.
Schilhans M. Bending frequency of a rotating cantilever beam. Journal of Applied Mechanics,
25:2830, 1958.
Truesdell C. and Noll W. The non-linear eld theories of mechanics. Encyclopaedia of Physics,
Vol. III/3,Springer Verlag, 1965.
Truesdell C. and Toupin R. The Classical Field Theory. Encyclopaedia of Physics Vol.
III/1,Springer Verlag, 1960.
[Link] D. Rubin E.K. Introduction to continuum mechanics. Oxford, New York, Butterworth
Heinemann, 1993.
Y. Vetyukov J.G. and H. Irschik ". ... The comparative analysis of the fully nonlinear, the linear
elastic and consistently linearized equation of motion of the 2d elastic pendulum. Computers
and Structures, 82:863870, 2004.
Mecnica Computacional Vol XXIX, pgs. 4153-4167 (2010) 4167
Copyright 2010 Asociacin Argentina de Mecnica Computacional [Link]