* Corresponding author.

International Journal of Non-Linear Mechanics 37 (2002) 135}151

Post-critical behavior of Beck's column with a tip mass
Steen Brahe Andersen, Jon Juel Thomsen*
Department of Solid Mechanics, Building 404, Technical University of Denmark, DK-2800 Lyngby, Denmark
Received 1 May 2000; accepted 2 October 2000
Abstract
This study examines how a tip mass with rotary inertia a!ects the stability of a follower-loaded cantilevered column.
Using nonlinear modeling and perturbation analysis, expressions are set up for determining the stability of the straight
column and the amplitude of post-critical #utter oscillations. Bifurcation diagrams are given, showing how the vibration
amplitude changes with follower load and other parameters. These results agree closely with numerical simulation. It is
found that su$ciently large values of tip mass rotary inertia can change the primary bifurcation from supercritical into
subcritical. This can imply very large motions for follower loads just beyond critical, contrasting the "nite amplitude
motions accompanying supercritical bifurcations. Also, the straight column may be destabilized by a su$ciently strong
disturbance at loads far below the value of critical load predicted by linear theory. A similar change in bifurcation is
found to occur with increased external (as compared to internal) damping, and with a shortening in column length. These
e!ects are not revealed by linear modeling and analysis, which may consequently fail to predict even qualitatively the real
critical load for a column with tip mass. 2001 Elsevier Science Ltd. All rights reserved.
1. Introduction
We examine how a tip mass with rotary inertia
a!ects the stability of a follower-loaded canti-
levered column. If there is no tip mass this system
corresponds to the well-known Beck's column [1],
which is known to exhibit soft #utter as the magni-
tude of the follower load exceeds a critical value
[2]. We show how this behavior is qualitatively
changed by the rotary inertia of a tip mass, which
can imply that the straight equilibrium of the col-
umn can be destabilized at a forcing lower than
critical by a "nite-sized disturbance.
This work was motivated by theoretical and ex-
perimental studies by Sugiyama et al. [3] on a col-
umn subjected to follower forcing produced by
a solid rocket motor.
Beck's column has a load at the tip that remains
tangent to the deformed column axis. The signi"-
cance of this system is its capability to model and
illustrate, in a simple manner, the basic properties
of systems that can display #utter vibrations due to
follower loading. Such loads are inherently
non-conservative, acting along lines that move
with the deformation or displacement of a struc-
ture. They occur with many real structures,
e.g., rocket driven aerospace structures, aircraft
wings, wind turbine blades, and pipes carrying
#uid #ow. The present study sheds further light
on possible behavior of follower loaded structures
for cases where the follower load is delivered by
a device whose mass and rotary inertia cannot be
ignored.
0020-7462/01/$- see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 0 - 7 4 6 2 ( 0 0 ) 0 0 1 0 2 - 5
Beck's initial work [1] on determining the criti-
cal force of the follower-loaded column has been
followed by numerous studies of di!erent aspects
and variants of the system. For example, Herrmann
and Jong [4] and Bolotin [5] described the seem-
ingly paradoxical destabilizing e!ect of internal
damping (the #utter-critical force in the presence of
vanishing internal damping being about half of the
value for zero damping), as contrasted by the
stabilizing e!ect of external damping [6]. P#uK ger
[7] seems to be the "rst to examine the stability of
a Beck's column with a point mass, whereas more
recently Sugiyama et al. [3] paid consideration to
the e!ects of damping and of a tip mass with rotary
inertia in their combined theoretical and experi-
mental study of a clamped column subjected to
follower forcing from a rocket motor; see also Lan-
gthjem and Sugiyama [8].
Most studies of Beck's column concerns the de-
termination of critical loads. This can be accomp-
lished using only linear modeling and analysis.
However, for studying post-critical behavior one
needs to incorporate in the model those nonlinear
e!ects that are thought to be responsible for limit-
ing the initially exponentially growing motions at
post-critical loads. To the authors knowledge only
Kolkka [2] and Chen [9] have employed nonlinear
modeling for Beck's column, showing that the bi-
furcation describing stationary oscillation ampli-
tudes at near-critical loads is supercritical, that is:
for post-critical loads the column performs stable
oscillations at a constant "nite amplitude (so-called
`softa #utter).
The primary motivation for the present work
was recent laboratory experiments conducted by
Sugiyama and Langthjem [10], indicating a critical
load much lower than the one predicted by linear
modeling. One possible explanation for this could
be a change of the type of bifurcation at the (lin-
early predicted) critical force from supercritical into
subcritical, perhaps due to the inertial properties of
the rocket motor. As will be explained further be-
low the existence of such a subcritical bifurcation
implies that the straight con"guration of the col-
umn can lose stability at loads lower than (linearly)
critical, if the disturbance is su$ciently large. We
therefore investigate how the rigid body properties
of a tip mass (mass, rotary inertia, center of gravity)
a!ects the stability properties and types of bifurca-
tions for Beck's column.
The paper is organized as follows: Section 2 de-
scribes the physical model, and the setup of the
di!erential equation governing transverse motions
with consideration to geometrical nonlinearities.
Using modal expansion, this equation is discretised
into a set of nonlinear ordinary di!erential equa-
tions for use in subsequent analysis. Section 3 de-
scribes how these equations are solved numerically
and also shows numerical solutions illustrating
typical system behavior. Section 4 presents a per-
turbation analysis of the nonlinear equations of
motions, leading to analytical expressions for the
stationary solutions and their stability. In this anal-
ysis, nonlinear terms up to "fth order are retained,
in order to be able to predict possible secondary
bifurcations associated with a primary subcritical
bifurcation. If a secondary bifurcation exists, it
means that stable oscillations could exist at loads
greater than critical. In this case, the third-order
solution would be insu$cient to describe the dy-
namic of the system. Section 5 then employs these
expressions to present bifurcation diagrams, show-
ing how the solutions and their stability change
with the magnitude of the follower load and para-
meters describing damping and inertial properties
of the tip mass. The perturbation results presented
here are supported by two other types or levels of
analysis: Numerical time integration of a "nite ele-
ment discretisation of the underlying physical
model, and numerical time integration of the mode
shape discretisation. Section 6 "nally sums up the
"ndings and the main results: the rotary inertia of
a tip mass on Beck's column can change the type
of bifurcation at the critical load from supercritical
to subcritical, and increased external damping as
compared to internal damping can have the same
e!ect.
2. Model system and equation of motion
In this section a geometric nonlinear model is set
up, which describes motion of the column for loads
larger than the critical force. The model can be used
to examine the nonlinear e!ects from the mass, its
rotary inertia and the location of the center of
136 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
Fig. 1. System model.
gravity on the oscillations of the column. If the
magnitude of the follower load exceeds the critical
force at which #utter is initiated, linear stability
theory predicts that the oscillation amplitude will
grow unbounded. In reality, nonlinear e!ects will
either stabilize the #utter motion into a periodic
oscillation with constant amplitude (a limit cycle)
or it will result in amplitudes growing even more
violently than predicted by the linear model (ex-
plosive #utter).
Fig. 1 shows the model. A rocket motor is moun-
ted at the tip of the cantilevered column. The rocket
motor produces a thrust P, which is the follower
load applied to the column. The motor is assumed
a rigid body, with mass m and rotary inertia J. The
center of gravity is located at the distance a from
the tip of the column. The column is assumed to be
viscoelastic and uniform, and has mass per unit
length jA, length , moment of inertia I, and
Young's modulus E.
A curvilinear coordinate system is used, where
y"y(s, t) denotes the transversal deformation of
the column, t denotes time, and s denotes the cur-
vilinear position on the column axis. The inextensi-
bility condition for an in"nitesimal column element
gives the relation
(x')`#(y')`"1, (1)
where primes denote di!erentiation with respect to
position s.
Hamilton's extended principle is employed for
setting up the equation of motion. Hamilton's prin-
ciple (e.g. [11]), which holds only for conservative
systems, states that
oH"0 where H,

R`
R
Ldt, L,T!V, (2)
where H de"nes the Hamiltonian of the system,
Lis the Lagrangian, Tand Vare the kinetic and
potential energy, t

and t
`
are arbitrary instants of
time, and oHis the variation of H.
Hamilton's extended principle states that the vir-
tual work oW done by non-conservative forces
added to the variation of the Hamiltonian should
vanish (e.g. [12,13])
oH#

R`
R
oWdt"0. (3)
The kinetic and potential energy, T

and V

, for
a column element is
T

"
`
jA(y `#x `), (4)
V

"
1
2
EI`"
1
2
EI
y"`
1!y'`
, (5)
where dots denote di!erentiation with respect to
time t, and where the non-linear curvature measure
for the curvilinear coordinate s has been used. (This
di!ers from the curvature in rectangular coordi-
nates, see e.g. [14, p. 46]). The Lagrangian L and
the Lagrangian density E are introduced as
L"

*
"
(T

!V

) ds"

*
"
Eds. (6)
The Hamiltonian can be divided in two:
H"H

#H

, where H

is the Hamiltonian for

the column and H

is the Hamiltonian for the

mass. The variation of the Hamiltonian H

for the
column is
oH

"

R`
R

*
"

cE
cy
oy #
cE
cy'
oy'
#
cE
cy"
oy"#
cE
cx
ox

ds dt. (7)
By using the inextensibility condition (1) and integ-
ration by parts, (7) can be written as
oH

"!

R`
R

*
"

c
ct
E
cy
#
c
cs
E
cy'
!
c`
cs`
E
cy"
!y'

1#
1
2
y'`

c
ct
cE
cx
#y"

1#
3
2
y'`

*
Q
c
ct
cE
cx
ds

oy ds dt. (8)
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 137
The kinetic energy T

of the mass has contribu-

tions from plane rotation and translations, i.e.
T

"T
G
#T
V
#T
W
. By using Dirac's delta-
function oK, the energies per unit length can be
written as
T
G
"
`
JoK(s!s
"
)0

Q `"
`
JoK(s!s
"
)(1#y'`)y '`, (9)
T
V
"
`
moK(s!s
"
)

d
dt
(x#a cos(0

))

`
"
`
moK(s!s
"
)(x #ax ')`, (10)
T
W
"
`
moK(s!s
"
)

d
dt
(y#a sin(0

))

`
"
`
moK(s!s
"
)(y #ay ')`, (11)
where s
"
is the location of the mass on the column-
axis and 0

denotes the rotation of the column. The

center of gravity of the mass is located at the dis-
tance a from s
"
in tangential direction. Later s
"
is
set equal to . The variation of the Hamiltonian
H

for the mass can now be written as

oH

"

R`
R

*
"
oT

ds dt
"

R`
R

*
"

oT
G
oy'
oy'#
oT
G
oy '
oy '#
oT
V
ox '
ox '
#
oT
V
ox
ox #
oT
W
oy '
oy '#
oT
W
oy
oy

ds dt.
(12)
By using the inextensibility condition (1) and integ-
ration by parts this can be transformed to
oH

"!

R`
R

*
"

c
ct
T
W
cy
#
c
cs
T
G
cy'
!
c`
cs ct
T
W
#T
G
cy '
!y'

1#
1
2
y'`

!c`
ct cs
cT
V
cx'
#
c
ct
cT
V
cx
#y"

1#
3
2
y'`

*
Q

!c`
ct cs
cT
V
cx'
#
c
ct
cT
V
cx
ds

oy ds dt.
(13)
The follower force P has x- and y-components
P
V
, P
W
:
P"(P
V
, P
W
)"(!Pcos( 0

()),!Psin(0

())), (14)
where 0

() is given by 0

()"arcsin(y'()) which
gives that
P
V
"!P(1!y'`()
+!P(1!
`
y'`()!
`
y'"()), (15)
P
W
"!Psin(arcsin(y'()))"!Py'(). (16)
The virtual work of the non-conservative follower
force can be divided in two: oW"oW
V
#oW
W
,
where oW
V
and oW
W
are the virtual work in the x-
and y-directions, respectively. By using (15) and (16)
these can be written as
oW
V
"P
V
ou"P(1!
`
y'`()!
`
y'"())ou, (17)
oW
W
"P
W
oy()"!Py'()oy(). (18)
Here u denotes the tip displacement in the negative
x-direction. By considering an in"nitesimal column
element and integrating over the column length
, u can be found as
du"

1!
dx
ds
ds"(1!(1!y'`) ds, (19)
u"

*
"
(1!(1!y'`) ds"

*
"
f (y') ds. (20)
The variation ou become
ou"

*
"
of(y') ds"

*
"
cf(y')
cy'
oy' ds
K

*
"

y'#
1
2
y'`#
3
8
y'`

oy' ds. (21)

Inserting (21) into (17), the virtual work in the
x-direction becomes
oW
V
"

*
"
P

1!
1
2
y'`()!
1
8
y'"()

y'#
1
2
y'`#
3
8
y'`

oy' ds. (22)

138 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
Integrating (22) by parts and applying Dirac's
delta-function to (18), the total virtual work from
the force P can be written as
oW"

*
"

!Py'oK(s!)
!P

y"#y"

3
2
y'`!
1
2
y'()`

#y"

15
8
y'"!
3
4
y'()`y'`
!
1
8
y'()"

oy ds. (23)
By inserting (8), (13) and (23) in (3) and requiring
that (3) is ful"lled for arbitrary admissible variation
in oy, one obtains in dimensionless form
pK #p""#p"#oK(!1)p'
#oo(!
"
)(pK #pK ')!v(oK(!
"
)pK ')'
!o(oK(!
"
)(pK #pK '))'
#R
#'`
#R
#'`
#R
`
#R
`
#R
.`
#R
.`
#R
K`
#R
K`
"0, (24)
where the nondimensional quantities are given by
p"
y

, "
x

, t"
,
EI
jA"
t, "
P`
EI
,
v"
J
jA`
, o"
m
jA
, "
a

(25)
and the dots and primes denote di!erentiation with
respect to dimensionless time t and position . This
equation describes the transverse displacements
p(, t) of the column. The "rst and second term in
(24) represent linear inertia and sti!ness from the
column, the third and fourth term represent a linear
sti!ening e!ect from the follower force and the "fth
to the seventh term represent linear inertia from the
tip mass, its rotary inertia and the location of the
center of gravity. The functions R
#'`
, R
#'`
, R
`
,
R
`
, R
.`
, R
.`
, R
K`
and R
K`
represent the nonlinear
e!ects and are given in the appendix. They repres-
ent terms of third and "fth order and describe,
respectively, sti!ness of the column R
#'`
, R
#'`
, axial
inertia of the column R
`
, R
`
, e!ects from the
follower force R
.`
, R
.`
, and e!ects from the mass
R
K`
, R
K`
.
A mode shape expansion is now applied to trans-
form the partial di!erential equation (24) into an
approximating set of ordinary di!erential equa-
tions. The expansion
p(, t)"
,

H
()+
H
(t) (26)
is used, where
H
are the mode shapes of a canti-
levered column with a tip mass, and +
H
are the
unknown time varying modal coe$cients. By use of
Galerkin's method (e.g. [15]) one obtains
M$ #CQ #S#g(,Q ,$ )#g` (,Q ,$ )"0
(27)
where M is the mass matrix, C is the damping
matrix, and S is the sti!ness matrix. The vectors
g and g` contain nonlinear terms of third and "fth
order, respectively:
g"
,

HIJ
((
HIJ
#
HIJ
)+
H
+
I
+
J
)
#
,

HIJ
((
HIJ
#
HIJ
)+
H
(+
I
+$
J
#+Q
I
+Q
J
)), (28)
g` "
,

HIJKL
((I
HIJKL
#I
HIJKL
)+
H
+
I
+
J
+
K
+
L
)
#
,

HIJKL
((I?AA
HIJKL
#I?AA
HIJKL
)+
H
+
I
+
J
+
K
+$
L
)
#
,

HIJKL
((ITCJ
HIJKL
#ITCJ
HIJKL
)+
H
+
I
+
J
+Q
K
+Q
L
).
(29)
The constants
HIJ
,
HIJ
,
HIJ
,
HIJ
, I
HIJKL
, I
HIJKL
,
I?AA
HIJKL
, I?AA
HIJKL
, ITCJ
HIJKL
and ITCJ
HIJKL
are computed from
the expansion functions and can be found in the
appendix.
The constants
HIJ
and
HIJ
, which correspond to
the nonlinear e!ects of third order from the sti!ness
and the axial inertia of the column, respectively,
have also been found and used by e.g. Semler et al.
[12], who set up the equation of motion for a pipe
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 139
Fig. 2. Tip displacement and frequency spectrum for a column with no tip mass for a force "13.53 which is greater than the critical
force
"
"12.93. Parameters: o"0, v"0, "0, ["0.1, j"0.01. (a): Tip displacement as function of time for four di!erent initial
conditions, where the tip amplitude is given and the velocity is zero; (b): Frequency spectrum. &*' nonlinear analytic model, &- - -' Finite
element model.
conveying #uid. To the author's knowledge, the set
up of an equation of motion including "fth-order
terms by use of Galerkins method for a column
with a tip mass subjected to a follower force is not
reported in the literature. The remaining constants
other than
HIJ
and
HIJ
are therefore new results.
Damping has been added to (27) in the form of
internal and external viscous damping, where the
elements in C are given by
c
GH
"[

"

H
d#j

"
"
G
"
H
d. (30)
Here [ is the external damping factor and j is the
internal damping factor. The elements in M and
S are given by
m
GH
"

"

H
d, (31)
s
GH
"

"
("
G
"
H
!'
G
'
H
) d#
G
(1)'
H
(1). (32)
Eq. (27) along with (28)}(32), describes a set of
coupled nonlinear ordinary di!erential equations
where the time-dependent modal coe$cients +
H
are
the unknowns. The equation is now ready to be
solved, and this is done numerically and by a per-
turbation method in the following two sections.
When the modal coe$cients are found the behavior
of the column in time can be described by using (26)
to calculate the transverse de#ection p of the col-
umn.
3. Numerical simulation
Numerical simulations of (27) are performed
with the purpose to verify pertubation solutions,
and to verify the model itself by comparing the
numerical solutions with solutions obtained by
a "nite element model. Furthermore, the transient
behavior and a frequency spectrum can be ob-
tained.
Eq. (27) is solved numerically using the trapez
method (e.g. [16]), which is a "nite-di!erence
method. The solution gives the displacement of the
column during time. Fig. 2(a) shows the tip dis-
placement as function of time for four di!erent
initial tip displacements, for a situation where the
load is greater than critical for #utter. As can be
seen the post-transient vibration amplitude is the
same for the four simulations. A frequency spec-
trum for the simulations in Fig. 2(a) is shown in Fig.
2(b), along with a frequency spectrum from a "nite
element simulation. It appears that the column
vibrates at its "rst eigenfrequency and a higher
harmonic. The "nite element simulation gives both
140 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
Fig. 3. Finite tip vibration amplitude p

as function of number
of shape functions. Critical force
"
"12.93, o"0, v"0, "0,
["0.1, j"0.01.
qualitative and quantitative the same results as the
model presented here.
Fig. 3 shows the dependency of the number N of
expansion functions
H
on the post transient vibra-
tion amplitude p

. It is necessary to use from four

to nine expansion functions to ensure that the solu-
tion is independent of the discretisation. It is neces-
sary to use even more expansion functions near
critical loads.
The geometric nonlinear dynamic "nite element
model, used above for comparison was set up using
64 2D beam elements with two nodes and six de-
grees of freedoms each. The nonlinearities have
been applied by using a method of updating the
coordinates [17]. Lin and Tsai [18] have used this
method in a nonlinear dynamic analysis of a canti-
levered column conveying #uid. The method here is
used similarly, to simulate the dynamics of a canti-
levered column with a tip mass and a follower force
at the tip. The follower force and the e!ects from
the tip mass have been applied as external forces at
the "nite element node corresponding to the col-
umn tip.
Numerical di$culties (lack of convergence) have
been encountered with the "nite element model in
cases where the tip mass is large or rotary inertia is
present. Implementation of another numerical
method, which takes into account the external for-
ces in the tangential sti!ness matrix might solve
this problem.
The numerical solutions of (27) and the "nite
element solutions described above are used for
verifying the pertubation solutions to be described
next.
4. Perturbation solutions
At loads greater than the critical, #utter occurs,
which can evolve into periodic oscillations about
the previously stable equilibrium. This change can
be described as a supercritical Hopf-bifurcation. In
this case, the nonlinear e!ects are stabilizing. An-
other possibility is that the oscillation amplitude
grow unbounded, which can be described as a sub-
critical Hopf-bifurcation. In this case, the non-lin-
ear e!ects are destabilizing on the oscillations.
To obtain insight into the system behavior at
near-critical loads, and to analyze the e!ect of
changing system variables, a perturbation analysis
is performed. A multiple scales method is employed
for this (e.g. [19,20]). The procedure presented here
is equivalent to the one used by Thomsen [21] and
Jensen [22].
Eq. (27) can be written as a system of "rst-order
equations
x "A()x#cf(, x, x )#cfI(, x, x ), (33)
where x" Q ' and the parameter c is introduc-
ed as a bookkeeping parameter to indicate the
smallness of non-linearities. The matrix A and the
vectors f and fI are given by
A"
_
0 I
!MS !MC
, f"

0
!Mg
,
fI"

0
!Mg`

, (34)
where I is the identity matrix. Two time scales are
introduced: a fast scale
"
"t for describing
motions at frequencies comparable to the natural
frequencies, and a slow scale

"ct describing
slow modulation of oscillation amplitudes.
Solutions x to (33) are approximated through
uniformly valid expansions of the form
x"x
"
(
"
,

)#cx

(
"
,

)#O(c`), (35)
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 141
where x
"
and x

are the functions to be determined.

The follower force is considered to be given
a small perturbation c

beyond the critical value

"
, and can thus be written as
"
"
#c

. (36)
Now f, fI and A are expanded in terms of c in the
vicinity of
"
:
f(, x, x )"f(
"
, x
"
, x
"
)#O(c),
fI(,x,x )"fI(
"
, x
"
, x
"
)#O(c), (37)
A()"A
"
#c

#O(c`), (38)
where
A
"
"A(
"
)
and
A

"
cA
c
(("
"
_
0 0
!M/
S
/(

("
0
. (39)
The expansions above and the solution to be found
are good approximations only when the terms con-
taining c, are small. This holds true when the load
increment

and the nonlinear terms are small.

By substituting (35), (37) and (38) into (33) and
equating coe$cients of like powers of c one obtains,
to order c"
cx
"
c
"
!A
"
x
"
"0 (40)
and to order c
cx

c
"
!A
"
x

"!
cx
"
c

x
"
#L
`,

HIJ
(b
HIJ
x
"H
x
"I
x
"J
#c
HIJ
x
"H
(x
"I
xK
"J
#x
"I
x
"J
))
#L
`,

HIJKL
(d
HIJKL
x
"H
x
"I
x
"J
x
"K
x
"L
#e
HIJKL
x
"H
x
"I
x
"J
x
"K
x
"L
#f
HIJKL
x
"H
x
"I
x
"J
x
"K
xK
"L
), (41)
where x
"H
is the jth component of x
"
and the non-
zero elements b
GHIJ
, c
GHIJ
, d
GHIJKL
, e
GHIJKL
og f
GHIJKL
of the
vectors b
HIJ
, c
HIJ
, d
HIJKL
, e
HIJKL
og e
HIJKL
are given as
b
G>,H>,I>,J>,
"!(I
GHIJ
#A
GHIJ
)
for i"1,2,
2
, N, (42)
c
G>,H>,I>,J>,
"!(D
GHIJ
#B
GHIJ
)
for i"1,2,
2
, N, (43)
d
G>,H>,I>,J>,K>,L>,
"!(II
GHIJKL
#AI
GHIJKL
)
for i"1,2,
2
, N, (44)
e
G>,H>,I>,J>,K>,L>,
"!(DITCJ
GHIJKL
#BITCJ
GHIJKL
)
for i"1,2,
2
, N, (45)
f
G>,H>,I>,J>,K>,L>,
"!(DI?AA
GHIJKL
#BI?AA
GHIJKL
)
for i"1,2,
2
, N (46)
and the matrix L is given by
L"
_
0 0
0 M
. (47)
The general solution to (40) is
x
"
"a(

)u exp(ic
"

"
)#a (

)u exp(!ic
"

"
).
(48)
Here overbars denote complex conjugation, a(

)
is an unknown function, and u and ic
"
are, respec-
tively, an eigenvector and eigenvalue for the eigen-
value problem
(A
"
!ic
"
I)u"0. (49)
Substituting the zero-order solution (48) into the
"rst-order problem (40), one obtains
cx

c
"
!A
"
x

"q

exp(ic
"

"
)#q
`
exp(i3c
"

"
)
#q
`
exp(i5c
"

"
)#c.c. (50)
where c.c. denotes the complex conjugated terms,
and where
q

"!
ca
c

u#

aA

u
#a`a L
`,

HIJ
(b
HIJ
(u
H
u
I
u
J
#u
H
u
I
u
J
#u
H
u
I
u
J
)
!2c`
"
c
HIJ
u
H
u
I
u
J
)#a`a `L
`,

HIJKL
142 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
;

(d
HIJKL
!c`
"
f
HIJKL
)
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L

!c`
"
e
HIJKL
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L

, (51)
q
`
"a`L
`,

HIJ
(b
HIJ
u
H
u
I
u
J
!2c`
"
c
HIJ
u
H
u
I
u
J
)
#a"a L
`,

HIJKL
(52)
;

(d
HIJKL
!c`
"
f
HIJKL
)
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L

!c`
"
e
HIJKL
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L

q
`
"a`L
`,

HIJKL
(d
HIJKL
u
H
u
I
u
J
!c`
"
(e
HIJKL
#f
HIJKL
)u
H
u
I
u
J
u
K
u
L
), (53)
where u
H
is the jth element of u. Since ic
"
is an
eigenvalue of A
"
, the solution of (50) will contain
secular terms proportional to
"
exp(ic
"

"
) unless
q

is orthogonal to the left eigenvector v of A

"
(e.g.
[23]), which means that v'q

"0 where v is given by

(A'
"
!ic
"
)v"0. (54)
Substituting q

of (51) into v'q

"0 yields the

solvability condition
da
d

!a!a`a !,a`a `"0, (55)

where , , and , are given by
"
v'A

u
v'u
,
"
#i
'
(56)
"
v'L`,
HIJ
(b
HIJ
(u
H
u
I
u
J
#u
H
u
I
u
J
#u
H
u
I
u
J
)
v'u
!
2c`
"
c
HIJ
u
H
u
I
u
J
)
v'u
,
"
#i
'
(57)
,"
v'L
v'u
`,

HIJKL
;

(d
HIJKL
!c`
"
f
HIJKL
)
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L

S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 143
!c`
"
e
HIJKL
_
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
!
u
H
u
I
u
J
u
K
u
L
#
u
H
u
I
u
J
u
K
u
L

,,
"
#i,
'
. (58)
Here (
"
,
'
), (
"
,
'
) and (,
"
, ,
'
) denotes the real and
imaginary parts of , and ,. The function a(

)
can be expressed in terms of two real-valued func-
tions A(

) and (

):
a(

)"A(

) exp(i(

)), A, 3R. (59)

Inserting (59) in the solvability condition (55), this
can be written as two modulation equations, de-
scribing evolutions of the slowly changing oscilla-
tion amplitude A and frequency :
dA
d

"
"
A#
"
A`#,
"
A`, (60)
d
d

A"
'
A#
'
A`#,
'
A`. (61)
Solutions corresponding to oscillations with con-
stant amplitude can be found by letting dA/d

"
0. Below, one solution is found for the third- and
"fth-order expansion, respectively. It is possible
that the solution for the "fth-order expansion gives
a secondary bifurcation that does not show up
using only the third-order expansion [22].
4.1. Third-order solution
For this approximation terms of "fth order in the
modulation equations are ignored. Eq. (60) then
has two singular points: A"0 which corresponds
to the zero-solution, and A"(!
"
/
"
which
corresponds to a stationary periodic solution for
!
"
/
"
'0 (a limit cycle). Combining Eqs. (35),
(48), (59) and (61) and setting c"1, the limit cycle
approximation becomes
x"x
"
#cx

#O(c`)
"2 Re[Aexp iu exp(ic
"

"
)]#O(c)
"2
,
!

"

"
ucos((c
"
#
'
!(
'
/
"
)
"
)t)#O(c).
(62)
The post-transient tip amplitude p
^
for the column
can now be found by inserting the amplitudes
max(+
G
)"max(x
G
), i)N of the modal coordi-
nates into the modal expansion (26):
p
^
"
,

G
(1) max(x
G
)"2
,
!

"

"
,

G
(1)u
G
,
(63)
where
G
is the ith eigenmode for the cantilevered
column with a tip mass and u
G
is the ith component
of the eigenvector u from (49).
4.2. Fifth-order solution
Retaining the "fth-order terms in (60), the non-
zero singular points become
A"
,
!
"
$(`
"
!4,
"

"
2,
"
.
(64)
The tip amplitude p
^
can now be found as above by
inserting (62) into (26)
p
^
"2
,
!
"
$(`
"
!4,
"

"
2,
"
,

G
(1)u
G
.
(65)
4.3. Stability of stationary solutions
The stability of the singular point A is evaluated
by inserting it into the gradient of (60)
A(A)"
"
#3
"
A`#5,
"
A". (66)
If this value is negative, then A is stable, otherwise
A is unstable.
Now, the stationary solutions and their stability
are calculated for particular parameters of the
column. These solutions are valid for loads in a
vicinity of the critical force. This means that
144 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
Fig. 4. Bifurcation diagrams. (a) No tip mass o"0.0; (b) With tip mass o"0.2. Damping: ["0.1, j"0.01. * Finite element;
numerical solution to (27); &*' stable, and &- - -' unstable 3rd and 5th order pertubation solutions.
a bifurcation diagram can be made from the solu-
tions and their stabilities, and hereby the e!ects
from the nonlinearities can be evaluated.
5. Bifurcations
The nonlinear oscillations of the column at loads
near the critical force will now be examined. The
nonlinear e!ects from the sti!ness of the column,
the tip mass, its rotary inertia, and the location of
the center of gravity will be examined separately.
For loads greater than critical the nonlinear ef-
fects can either stabilize the #utter into oscillations
with constant amplitude, or they can destabilize
even further so the #utter becomes explosive. This
change in response as the load exceeds the critical
value corresponds to a supercritical or a subcritical
bifurcation, respectively.
If the bifurcation is supercritical only the stable
zero-solution exists at loads smaller than the criti-
cal force. The stability limit of the zero-solution can
therefore be found by linear theory. If the bifurca-
tion is subcritical, however, the nonlinear e!ects
can destabilize the zero-solution if the column ex-
periences a disturbance of su$cient strength. In
this case, the behavior of the system is therefore
radically changed and linear stability theory is in-
su$cient to determine the real critical force for
disturbances of "nite magnitude. It is therefore of
great importance to know which kind of bifurca-
tion there will exist at particular values of para-
meters.
5.1. Column with no tip mass
Fig. 4(a) shows the bifurcation diagram for a col-
umn with no tip mass. The bifurcation is supercriti-
cal, which means that the nonlinear e!ects stabilize
the #utter into a limit cycle. This is in agreement
with Kolkka [2], who shows mathematically that
the bifurcation is supercritical for Beck's column.
As appears, the numerical simulations of Eq. (27)
match very well with the "nite element simulations
and the approximative perturbation solution. The
perturbation solutions of third and "fth order show
no signi"cant di!erences.
5.2. Column with a tip mass
A tip mass with a mass of 20% of the column
mass has been applied at the tip-end of the column.
The bifurcation diagram in Fig. 4(b) shows that the
bifurcation is still supercritical. The critical force is
smaller than for the column with no tip mass. The
shape of the bifurcation is comparable with the
one for the column with no tip mass, indicating that
the mass does not qualitatively change the dynam-
ics of the column.
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 145
Fig. 5. Bifurcation diagrams for a column having rotary inertia v at the tip. (a) v"0.002; (b) v"0.2. Damping: ["0.1, j"0.01.
Numerical solution to (27); &*' stable, and &- - -' unstable 3rd and 5th order pertubation solutions.
5.3. Column with rotary inertia at the tip
To shed light on the isolated e!ect of rotary
inertia, we consider here the (unrealizable) case
of a zero tip mass having nonzero rotary inertia.
Fig. 5(a) and (b) shows the bifurcations for a col-
umn with two di!erent values of inertia at the
tip-end. The bifurcation is still supercritical as long
as the rotary inertia is very small. The shape of the
branching bifurcation becomes steeper, indicating
that the nonlinear e!ects from the rotary inertia are
destabilizing. This indication is supported by the
bifurcation in Fig. 5(b), where the rotary inertia is
even larger. First, it should be noted that the larger
rotary inertia decreases the critical load signi"-
cantly. Secondly the bifurcation has changed and
has become subcritical, meaning that the system
behaves in a completely di!erent way. At loads
smaller than critical, it is possible to destabilize the
zero solution by disturbing the column su$ciently
so the amplitude becomes greater than the ampli-
tude of the unstable limit circle. The "fth-order
perturbation solution shows the same as the solu-
tion of third order: there are no stable solutions at
loads greater than the critical. At loads greater than
the critical, motions are therefore not limited by the
nonlinearities included in the present model; for
a real model other nonlinearities such as plastic
deformation or fracture will limit the response. In
numerical simulations explosive growths in ampli-
tudes were observed until numerical instability oc-
curred at very large deformations.
5.4. Column with a tip mass having rotary inertia
Fig. 6(a) shows the bifurcation for a column with
a tip mass having rotary inertia and Fig. 6(b) shows
the bifurcation for a case where the center of gravity
of the mass is located at a distance from the tip. The
"gures show that the shape of the bifurcations are
steeper compared to the bifurcation for the column
with no tip mass. This is due to the rotary inertia
and the location of the center of gravity, which
gives a resulting moment of inertia at the tip. The
numerical simulations match very well the solu-
tions found by the third- and "fth-order perturba-
tion solutions. If the rotary inertia is su$ciently
large, the bifurcation will become subcritical, and
the #utter will be explosive. This was con"rmed by
numerical simulations.
5.5. Change of bifurcation
For a given tip mass having rotary inertia, Fig. 7
shows if the bifurcation at the critical load is sub- or
supercritical. As appears, when the rotary inertia is
su$ciently large the type of bifurcation changes
from super- to subcritical, we call this value the
critical rotary inertia. When the tip mass is about
the same as the mass of the column or larger (o'1)
146 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
Fig. 6. Bifurcation diagrams for a column with a tip mass o having rotary inertia v, the center of gravity beeing located a distance from
the tip. (a) o"2.0, v"0.2, "0.0; (b) o"2.0, v"0.2, "0.1. Damping: ["0.1, j"0.01. Numerical solution to (27); &*' stable,
and &- - -' unstable 3rd and 5th order pertubation solutions.
Fig. 7. Type of bifurcation as function of tip mass o, tip mass
rotary inertia v, and position of center of gravity . Damping:
["0.1, j"0.01. Solid lines mark the boundaries between
supercritical and subcritical bifurcations.
there is an approximative linear relationship be-
tween the mass and the critical rotary inertia
(o/v"m`/J"constant). As also appears a sub-
critical bifurcation can occur only when the
rotary inertia is non-zero. Finally, it should
be observed, that for a given tip mass having
rotary inertia, a shortening of column length can
cause a change in bifurcation from super- to sub-
critical.
5.6. Dependency of damping
Fig. 8 shows the signi"cants of the ratio of ex-
ternal to internal damping [/j on the type of bifur-
cation. As appears the bifurcation type is highly
dependent on this ratio. The damping coe$cients
[ and j can be very di$cult to estimate correctly
for real systems, and thus it may be di$cult to
estimate which kind of bifurcation will actually
occur. It should also be noted that if the energy
dissipation is dominated by external damping
([/j<1) then subcritical bifurcations may exist
even for small amount of rotary inertia (v/o;1).
6. Summary and conclusions
Using nonlinear modeling and analysis, we have
established critical loads and investigated the post-
critical behavior of a Beck's column with a tip mass
having rotary inertia.
Numerical simulations showed the column to
vibrate at its fundamental natural frequency with
a higher harmonic, in a form dominated by the
fundamental and the second mode shape. These
results were con"rmed by numerical solutions for
a "nite element model of the column system.
Using a multiple scales method a perturbation
analysis including terms to "fth order was performed
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 147
Fig. 8. Type of bifurcation as function of the ratio of tip mass to
tip mass rotary inertia, o/v, and the ratio of external to internal
damping [/j.
for near-critical loads. This lead to analytical ex-
pressions for the post-critical solutions and their
stability. The results were subsequently employed
to establish bifurcation diagrams, showing how the
stationary vibration amplitude depends on the
magnitude of the follower load and other para-
meters. Again, results were veri"ed by numerical
simulation.
The main result of this work is the observation
that the rotary inertia of a tip mass on a Beck's
column can change the type of bifurcation at the
critical load. At su$ciently large values of rotary
inertia, this bifurcation will change from supercriti-
cal to subcritical. This implies two phenomena that
are not revealed by linear analysis: First, at su$-
ciently large values of rotary inertia there will be no
stable small-amplitude motions for follower loads
just beyond the critical; only motions with very
large amplitude are then possible. Secondly, the
straight con"guration of the column may be de-
stabilized by a su$ciently strong disturbance at
loads below the critical load predicted by linear
theory. Thus, the presence of large rotary inertia
may make stability analysis based only on linear
modeling insu$cient. The same observation holds
true even for a tip mass with negligible rotary
inertia, if its center of gravity is displaced rigidly
beyond the column tip, or if a column with a given
tip mass and rotary inertia is shortened.
Increasing the external damping of the system (as
compared to internal damping) also has this e!ect
of changing the primary bifurcation from super-
critical to subcritical. Since with real systems the
character and magnitude of damping e!ects are
rarely known to any great precision, this implies
that predictions based only on linear theory might
fail simply due to a wrong estimate of damping
coe$cients.
The above theoretical results will need experi-
mental veri"cation. It would be relevant to system-
atically examine critical loads and post-critical
behavior for a follower-loaded clamped column as
a function of changing rotary inertia of a tip mass.
For example, tests similar to those described in
Sugiyama et al. [3] and Langthjem and Sugiyama
[8] could be performed with two additional masses
at the rocket motor, mounted on a rigid beam
perpendicular to the column and with variable sep-
aration.
Acknowledgements
The authors wish to thank Mikael Langthjemfor
suggesting this work, and for many useful com-
ments and suggestions.
Appendix A
A.1. Dimensionless functionals
R
#'`
"p"`#4p'p"p"'#p'p'p"" (A.1)
R
#'`
"6p'`p"`#8p'`p"p"'#p'"p"", (A.2)
R
`
"p'

K
"
(p '`#p'pK ')d!p"

K
"
(p '`
#p'pK ') d d, (A.3)
R
`
"p'

K
"

3
2
p'`p '`#
1
2
p'`pK '

d
#
1
2
p'`

K
"
(p '`#p'pK ') d
!p"

K
"

3
2
p'`p '`#
1
2
p'`pK '

d d
148 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
!
3
2
p"p'`

K
"
(p '`#p'pK ') d, (A.4)
R
.`
"p"

3
2
p'`!
1
2
p'`(1)

, (A.5)
R
.`
"p"

15
8
p'"!
3
4
p'(1)`p'`!
1
8
p'(1)"

, (A.6)
R
K`
"!v(oK(!
"
)(p'`pK '#p'p '`))'
!op'

oK(!
"
)

K
"
(p '`#p'pK ') d
#(p '`#p'pK ')

'
!oK(!
"
)

K
"
(p '`
#p'pK ') d#(p '`#p'pK ')

!op"

K"
"
(p '`#p'pK ') d#(p '`#p'pK ')

,
(A.7)
R
K`
"!v(oK(!
"
)(p'"pK '#2p'`p '`))'
!op'

oK(!
"
)

K
"

3
2
p'`p '`#
1
2
p'`pK '

d
#

3
2
p'`p '`#
1
2
p'`pK '

'
!oK(!
"
)

K
"

3
2
p'`p '`#
1
2
p'`pK '

d
#

3
2
p'`p '`#
1
2
p'`pK '

!
o
2
p'`

oK(!
"
)

K
"
(p '`
#p'pK ') d#(p '`#p'pK ')

'
!oK(!
"
)

K
"
(p '`#p'pK ') d
#(p '`#p'pK ')

!op"

K"
"

3
2
p'`p '`#
1
2
p'`pK '

d
#

3
2
p'(
"
)`p '(
"
)`
#
1
2
p'(
"
)`pK '(
"
)

(A.8)
!
3
2
op"p'`

K
"
(p '`#p'pK ') d
#(p '`#p'pK ')

. (A.9)
A.2. Non-linear coezcients
I
GHIJ
"

3
2

"

G
"
H
'
I
'
J
d
!
1
2

"

G
"
H
d'
I
(1)'
J
(1)

, (A.10)
A
GHIJ
"

"

G
"
H
"
I
"
J
d#4

"

G
'
H
"
I
"'
J
d
#

"

G
'
H
'
I
""
J
d, (A.11)
D
GHIJ
"

"

G
'
H

K
"
'
I
'
J
d d
!

"

G
"
H

K
"
'
I
'
J
d d d, (A.12)
B
GHIJ
"v'
G
(1)'
H
(1)'
I
(1)'
J
(1)
#o(
G
(1)'
H
(1)#('
G
(1)'
H
(1)#
G
(1)"
H
(1)))
;

"
'
I
'
J
d#'
I
(1)'
J
(1)

!o

"

G
"
H
d

"
'
I
'
J
d#'
I
(1)'
J
(1)

,
(A.13)
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 149
II
GHIJKL
"
15
8

"

G
'
H
'
I
'
J
'
K
"
L
d
!
3
4
'
H
(1)'
I
(1)

"

G
'
J
'
K
"
L
d (A.14)
!
1
8
'
H
(1)'
I
(1)'
J
(1)'
K
(1)

"

G
"
L
d,
(A.15)
AI
GHIJKL
"6

"

G
'
H
'
I
"
J
"
K
"
L
d
#8

"

G
'
H
'
I
'
J
"
K
"'
L
d, (A.16)
#

"

G
'
H
'
I
'
J
'
K
""
L
d, (A.17)
DITCJ
GHIJKL
"
3
2

"

G
'
H

K
"
'
I
'
J
'
K
'
L
d d
#
1
2

"

G
'
H
'
I
'
J

K
"
'
K
'
L
d d
!
3
2

"

G
"
H

K
"
'
I
'
J
'
K
'
L
d d d
!
3
2

"

G
"
H
'
I
'
J

K
"
'
K
'
L
d d d,
(A.18)
BITCJ
GHIJKL
"2v'
G
(1)'
H
(1)'
I
(1)'
J
(1)'
K
(1)'
L
(1)
#
3
2
o

G
(1)'
H
(1)#('
G
(1)'
H
(1)
#
G
(1)"
H
(1))!

"

G
"
H
d

"
'
I
'
J
'
K
'
L
d
#'
I
(1)'
J
(1)'
K
(1)'
L
(1)) (A.19)
#
1
2
o(
G
(1)'
H
(1)'
I
(1)'
J
(1)
#('
G
(1)'
H
(1)'
I
(1)'
J
(1))'
!3

"

G
"
H
'
I
'
J
d

"
'
K
'
L
d#'
K
(1)'
L
(1)

, (A.20)
DI?AA
GHIJKL
"
1
2

"

G
'
H

K
"
'
I
'
J
'
K
'
L
d d
#
1
2

"

G
'
H
'
I
'
J

K
"
'
K
'
L
d d
!
1
2

"

G
"
H

K
"
'
I
'
J
'
K
'
L
d d d
!
3
2

"

G
"
H
'
I
'
J

K
"
'
K
'
L
d d d
(A.21)
BI?AA
GHIJKL
"v'
G
(1)'
H
(1)'
I
(1)'
J
(1)'
K
(1)'
L
(1)
#
1
2
o

G
(1)'
H
(1)#('
G
(1)'
H
(1)
#
G
(1)"
H
(1))!

"

G
"
H
d

"
'
I
'
J
'
K
'
L
d
#'
I
(1)'
J
(1)'
K
(1)'
L
(1)) (A.22)
#
1
2
o

G
(1)'
H
(1)'
I
(1)'
J
(1)
#('
G
(1)'
H
(1)'
I
(1)'
J
(1))'
!3

"

G
"
H
'
I
'
J
d

"
'
K
'
L
d#'
K
(1)'
L
(1)

. (A.23)
References
[1] M. Beck, Die Knicklast des einseitig eingespannten tan-
gential gedruK ckten Stabes, Zeitscrift fuK r Angewandte
Mathematik und Physik 3 (1952) 225}228.
[2] R.W. Kolkka, On the non-linear Beck's problem with
external damping, Int. J. Non-Linear Mech. 14 (6) (1984)
497}505.
150 S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151
[3] Y. Sugiyama, K. Katayama, S. Kinoi, Flutter of canti-
levered column under rocket thrust, J. Aerospace Engng.
8 (1) (1995) 9}15.
[4] G. Herrmann, I.C. Jong, On the destabilizing e!ect of
damping in nonconservative elastic systems, J. Appl.
Mech. 32 (1965) 592}597.
[5] V.V. Bolotin, Nonconservative Problems of the Theory of
Elastic Stability, Pergamon Press Ltd., Oxford, 1963.
(Translated by T.K. Lusher).
[6] R.H. Plaut, E.F. Infante, The e!ect of external damping on
the stability of Beck's column, Int. J. Solids Struct. 6 (1970)
491}496.
[7] A. P#uK ger, Zur StabilitaK t des tangential gedruK ckten Stabes,
Zeitscrift fuK r Angewandte Mathematik und Mechanik 35
(5) (1955) 191.
[8] M.A. Langthjem, Y. Sugiyama, Optimum shape design
against #utter of a cantilevered column with an end-mass
of "nite size subjected to a non-conservative load, J. Sound
Vib. 226 (1) (1999) 1}23.
[9] M. Chen, Hopf bifurcation in Beck's problem, Nonlinear
Anal. Theory, Methods Appl. 11 (9) (1987) 1061}1073.
[10] M.A. Langthjem, Personal communication, November
(1998).
[11] C. Lanczos, The Variational Principles of Mechanics, 2nd
Edition, University of Toronto Press, Toronto, 1962.
[12] C. Semler, G. Li, M.P. PamKdoussis, The non-linear equa-
tions of motion of pipes conveying #uid, J. Sound Vib. 169
(1994) 577}599.
[13] J.S. Jensen, Fluid transport due to nonlinear #uid-
structure interaction, J. Fluids Struct. 11 (1997)
327}344.
[14] V.V. Bolotin, The Dynamic Stability of Elastic Systems,
Holden-Day, San Francisco, CA, 1964.
[15] L. Meirovitch, Analytical Methods in Vibrations, Macmil-
lan, New York, 1967.
[16] R.D. Cook, D.S. Malkus, M.E. Plesha, Concepts and Ap-
plications of Finite Element Analysis, Wiley, New York,
1989.
[17] P.C. Kohnke, Large de#ection analysis of frame structures
by "ctitious forces, Int. J. Numer. Methods Engng. 12
(1978) 1279}1294.
[18] Y.H. Lin, Y.K. Tsai, Nonlinear vibrations of Timoshenko
pipes conveying #uid, Int. J. Solids Struct. 34 (23) (1996)
2945}2956.
[19] A.H. Nayfeh, D.T. Mook, Nonlinear Oscillations, Wiley-
Interscience, New York, 1979.
[20] J.J. Thomsen, Vibration and Stability, Order and Chaos,
McGraw-Hill, London, 1997.
[21] J.J. Thomsen, Chaotic dynamics of the partially follower-
loaded elastic double pendulum, J. Sound Vib. 188 (3)
(1994) 385}405.
[22] J.S. Jensen, Articulated pipes conveying #uid pulsating
with high frequency, Nonlinear Dyn. 19 (1999) 171}191.
[23] A.H. Nayfeh, B. Balachandran, Applied nonlinear dynam-
ics: Analytical, Computational, and Experimental Meth-
ods, Wiley, New York, 1995.
S.B. Andersen, J.J. Thomsen / International Journal of Non-Linear Mechanics 37 (2002) 135}151 151