Sie sind auf Seite 1von 11

A consistent theory of finite stretches and finite rotations,

in space-curved beams of arbitrary cross-section


S. N. Atluri, M. Iura, S. Vasudevan
Abstract Attention is focused in this paper on the devel-
opment of a consistent nite deformation beam theory,
and its mixed variational formulation. The shearing
deformation, as well as cross-sectional warping displace-
ment, are taken into account in this formulation. Beginning
with the equilibrium equations of 3-D continuum body, we
obtain the linear momentum balance (LMB), angular mo-
mentum balance (AMB) and director momentum balance
(DMB) conditions of the beam. The conjugate relationships
between the strain and stress measures are obtained
through the stress power, in which the AMB condition
plays an important role. The use of the strain measures
proposed herein, leads to the strain energy function which
is invariant under a rigid-body motion. The present for-
mulation is shown to be objective by using a numerical
example. On the basis of Atluri's variational principle, we
develop a mixed type variational functional for a space-
curved beam, undergoing arbitrarily large rotations and
arbitrarily large stretches. A choice of a proper nite ro-
tation vector, and unsymmetric curvature strains, makes it
possible for constructing a consistent variational principle.
The use of the present functional always leads to a sym-
metric tangent stiffness. The mixed variational functional
developed herein leads to a powerful tool for obtaining
accurate numerical results of 3-D space-curved beams,
undergoing arbitrarily large stretches and rotations.
1
Introduction
This paper deals with the development of a consistent
theory of nite stretches and nite rotations, in space-
curved beams of an arbitrary cross-section. The nite ro-
tation vector plays an important role for describing the
kinematics of the beam. The shearing deformation, as well
as cross-sectional warping displacement, are taken into
account in this formulation. Using the kinematic and ki-
netic variables for the present theory of a curved beam in
Atluri's variational principle (Atluri, 1979, 1980, 1983;
Atluri and Cazzani, 1995), a consistent multi-eld varia-
tional theorem for a space-curved beam undergoing ar-
bitrarily large stretches and rotations, under bending and
torsion, is developed.
Beginning with the equilibrium equations of a 3-D
continuum body, we obtain the linear momentum balance
(LMB), angular momentum balance (AMB) and director
momentum balance (DMB) equations of the beam. The
balance equation for the bimoment is also obtained. With
the use of the AMB and DMB equations, the conventional
moment equilibrium equation is derived. The AMB
equation plays an important role in the expression of
stress power of the beam, in which the deformation gra-
dient tensor and the rst PiolaKirchhoff stress tensor are
used. In addition to the conventional strain measures, we
introduce new strain measures for constructing a consis-
tent beam theory. The conjugate relationships between the
strain and stress measures are obtained through the stress
power.
The invariance condition of strain energy function is
discussed, and the important role of the AMB equation is
emphasized. The AMB equation used herein is derived by
using the kinematic and stress-state assumptions. The use
of the new strain measures gives a strain energy function
which is invariant under a rigid-body motion, provided
that the AMB equation is satised.
Criseld and Jelenic (1999) have discussed the nite
element formulation for 3-D beams, and concluded that
the existing formulations lead to a non-objectivity of the
interpolated strain measures. We show herein that the
objectivity condition obtained by Criseld and Jelenic
(1999) is misleading. In contrast to their conclusion, the
formulation developed by Iura and Atluri (1988, 1989) and
the present one are shown to be objective by using a nu-
merical example.
On the basis of Atluri's variational principle, we develop
a mixed type variational functional for a space-curved
beam, undergoing arbitrarily large rotations and arbi-
trarily large stretches. A choice of a proper nite rotation
vector, and unsymmetric curvature strains, makes it pos-
sible for constructing a consistent variational principle. As
a rotational variation, we employ the variation of nite
Computational Mechanics 27 (2001) 271281 Springer-Verlag 2001
271
Received 22 November 2000
S. N. Atluri
Center for Aerospace Research & Education,
7704 Boelter Hall, School of Engineering & Applied Science,
UCLA, Los Angeles, CA 90095-1600, USA
M. Iura (&)
Department of Civil and Environmental Engineering,
Tokyo Denki University,
Hatoyama, Hiki, Saitama, Japan
e-mail: iura@g.dendai.ac.jp
S. Vasudevan
Development Engineer,
Dowell-Schlumberger,
Methods, Models and Mechanics Group,
Software Engineering Products
rotation vector, which is the Lagrangian rotational varia-
tion. In the existing literature, the Eulerian rotational
variation has always been employed. As a result, the re-
sulting tangent stiffness does not become symmetric. The
use of the present functional always leads to a symmetric
tangent stiffness. The Euler equations of the present
functional give the constitutive equations, the compati-
bility conditions, the balance equations, and the boundary
conditions. The mixed variational functional developed
herein leads to a powerful tool for obtaining accurate
numerical results for 3-D space-curved beams, and 3-D
spatial structures made up of such beams, undergoing -
nite rotations and nite strains. The proposed method has
been extended to thin-walled composite beams commonly
used in aerospace structures and has been demonstrated to
be computationally very efcient (see Vasudevan et al.
2000).
2
Preliminaries
The fundamental hypotheses for deriving the present
beam theory are itemized as follows:
(1) There are no deformations along the in-plane coor-
dinates in any cross-section. However, the cross-sec-
tion may undergo warping deformation due to twist.
(2) The cross-sectional shape is arbitrary, but invariant
along the beam axis. The beam axis remains a smooth
space curve throughout the deformation.
The above assumptions are associated with the kinematics
of the beam. In Sect. 4.4, an additional assumption for
stress states will be introduced.
Let X
j
(j = 1; 2; 3) denote the Cartesian coordinates xed
in the space with the base vectors I
j
. Let Y
j
denote con-
vected orthogonal curvilinear coordinates with the base
vectors
^
E
j
in the undeformed conguration. The coordi-
nates Y
a
(a = 1; 2) are taken in the cross-section of the
beam while Y
3
is taken along the beam axis, as shown in
Fig. 1. The position vector at the beam axis is given by
X = X
j
(Y
3
)I
j
: (1)
The tangent base vector
^
E
3
in the undeformed congura-
tion is dene by
^
E
3
= dX=dY
3
: (2)
In general, the tangent base vector is not a unit vector. For
later convenience, we introduce the unit base vector as-
sociated with Y
3
dened by
E
3
=
^
E
3
=
^
E
3

: (3)
The base vectors E
a
associated with Y
a
are chosen to be
unit vectors without loss of generality. The FrenetSerret
formulae is written as
E
j;3
= K E
j
; K = K
j
E
j
; (4)
where ()
;3
= d()=dL and dL =
^
E
3

dY
3
; K
a
are the com-
ponents of initial curvatures and K
3
is the initial twist. The
position vector at an arbitrary material point before the
deformation is given by
P = X Y
a
E
a
: (5)
The base vectors at an arbitrary material point in the
undeformed conguration are dened by
A
j
= P
;j
; A
a
= E
a
; A
3
= E
3
Y
a
K E
a
; (6)
where ()
;a
= d()=dY
a
.
Let e
m
be the maps of the base vectors E
m
after a purely
rigid nite rotation, represented by rotation tensor R.
Then we have e
m
= RE
m
. In general, due to the shearing
deformation and the warping deformation, the unit tan-
gent vector to the beam axis does not coincide with the
vector e
3
. Using the kinematic assumptions, we obtain the
base vectors at an arbitrary material point after the de-
formation, dened by
a
j
= (X u Y
a
e
a
w)
;j
(7)
where u is the displacement vector at the beam axis and w
the cross-sectional warping displacement vector. Equation
(7) can be rewritten in the following form:
a
a
= e
a
w
;a
; a
3
= t Y
a
e
a;3
w
;3
(8)
where
t = (X u)
;3
; (9)
In the case of linear problem, the warping displacement
vector is assumed to take the form as w = k(Y
3
)/(Y
a
)E
3
where / is the warping function of SaintVenant's torsion
and k the unknown parameter. It has often been assumed
for the warping torsion that k is equal to the rate of twist
(see Kollbrunner and Basler, 1969). The warping dis-
placement vector for the nonlinear problem will be
discussed later.
Next, we briey describe the nite rotation vector
(see Atluri and Cazzani, 1995, for more detailed
discussion). Let x be a unit vector satisfying Rx = x
and h the magnitude of rotation about the axis of
rotation. Then the nite rotation vector h is dened
by h = hx. With the use of the nite rotation vector,
we obtain the relationships between R and h expressed
as
R = I sinh(xI) (1 cos h)x(xI) ; (10) Fig. 1. Space curved beam before and after the deformation
272
where I is the identity tensor. Since RR
t
= I, dRR
t
is the
skewsymmetric tensor. Then there exists a variation d/
satisfying the following relation:
d/ I = dRR
t
: (11)
The variation d/ has often been used to construct the
weak form of variational principle since d/ is conjugate
with the moment equilibrium equation. The disadvantage
of using d/ is to obtain the unsymmetric tangent stiffness.
Iura and Atluri (1988, 1989) have shown that the use of dh
always leads to the symmetric tangent stiffness. The rela-
tion between d/ and dh is given by (see Atluri and Caz-
zani, 1995)
d/ = Xdh; X = I
1 cos h
h
2
(h I)

1
h
2
1
sinh
h
_ _
h (h I) :
(12)
3
Linear theory of straight beams
with arbitrary cross sections
In this section, we briey describe a linear theory of
straight beams with an arbitrary cross section. Since the
beam is straight before the deformation, we have K = 0.
On the basis of balance equations for 3-D dimensional
continuum, we derive the balance equations of the beam.
The AMB and DMB equations are expressed in terms of the
director moments. The conventional moment equilibrium
equation is derived from the AMB and DMB equations.
The kinematics of the beam is given with the use of the
innitesimal rotation vector. We introduce the curvature
strains which are conjugate with the director moments.
Finally, the constitutive equations of the beam are given.
3.1
Balance laws
The LMB and AMB equations for a 3-D continuum body
are written as (see Atluri, 1983)
r
1
;1
r
2
;2
r
3
;3
q
0
B = 0;
A
1
r
1
A
2
r
2
A
3
r
3
= 0 ;
(13)
where ()
;j
is the partial differentiation with respect to the
reference coordinates, r
j
the stress vector, B the body force
vector and q
0
the density in the undeformed state.
Multiplying Eq. (13a) by dA, integrating it in the cross-
sectional area of the beam and using the divergence the-
orem, we obtain the LMB equation of the beam expressed
as
_
(r
1
;1
r
2
;2
r
3
;3
q
0
B)dA = n
;3
b = 0 ; (14)
where
n =
_
r
3
dA;
b =
_
q
0
BdA
_
r
1
m
1
ds
_
r
2
m
2
ds ;
(15)
m
a
being the direction cosines of the unit normal drawn
outwards on the beam surface.
Multiplying Eq. (13b) by dA, integrating it and using
the divergence theorem, we obtain the AMB equation as:
_
(E
j
r
j
Y
1
E
1;3
r
3
Y
2
E
2;3
r
3
) dA
= E
a
n
a
E
3
n E
a;3
h
a
= 0 ; (16)
where
h
a
=
_
r
3
Y
a
dA; n
a
=
_
t
a
dA : (17)
The relations between the director moment vector h
a
and
the conventional moment vector will be dened later.
Multiplying Eq. (13a) by Y
a
dA, integrating over the
cross-section and using the divergence theorem, we obtain
the DMB equations of the beam expressed as
_
(r
1
;1
r
2
;2
r
3
;3
q
0
B) Y
a
dA
= n
a
h
a
;3
p
a
= 0 ; (18)
where p
a
=
_
q
0
BY
a
dA
_
r
1
Y
a
m
1
ds
_
r
2
Y
a
m
2
ds:
Eliminating n
a
from Eqs. (16) and (18), we obtain the
moment equilibrium equation expressed as
m
;3
E
3
n q = 0 ; (19)
where the internal and external moment vectors are
dened by
m = E
a
h
a
=
_
Y
a
E
a
r
3
dA; q = E
a
p
a
:
(20)
Taking the scalar product of Eq. (13a) with /E
3
dA, inte-
grating over the cross-section and using the divergence
theorem, we obtain the balance equation for the bimoment
expressed as
_
(r
1
;1
r
2
;2
r
3
;3
q
0
B) /E
3
dA
= n
w
h
w;3
d = 0 ; (21)
where
n
w
=
_
r
a3
/
;a
dA; h
w
=
_
r
33
/dA;
d =
_
q
0
B /E
3
dA
_
r
a3
/m
a
ds : (22)
3.2
Kinematics: infinitesimal deformation
Let e
j
be the maps of the base vectors E
j
after a purely rigid
rotation. Using the innitesimal rotation vector h(= h
j
E
j
),
the relationships between e
j
and E
j
are written as
[e
j
= h E
j
[:
e
1
= E
1
h
3
E
2
h
2
E
3
;
e
2
= h
3
E
1
E
2
h
1
E
3
; (23)
e
3
= h
2
E
1
h
1
E
2
E
3
:
273
Using Eq. (23), we obtain
e
1;3
e
2;3
e
3;3
_
_
_
_
_
_ =
0 j
3
j
2
j
3
0 j
1
j
2
j
1
0
_
_
_
_
_
_
E
1
E
2
E
3
_
_
_
_
_
_
or e
j;3
= j E
j
; j = j
j
E
j
;
(24)
where j
j
= h
j
;3
.
Let u(= u
j
E
j
) be the displacement vector at the beam
axis. The displacement vector at an arbitrary material
point is dened by
U = u Y
1
(e
1
E
1
) Y
2
(e
2
E
2
) k/E
3
: (25)
Substituting Eq. (23) into Eq. (25), we obtain
U = u Y
1
(h
3
E
2
h
2
E
3
)
Y
2
(h
3
E
1
h
1
E
3
) k/E
3
: (26)
With the use of Eq. (26), we obtain the displacement
components U
j
(U = U
j
E
j
), expressed as
U
1
= u
1
Y
2
h
3
; U
2
= u
2
Y
1
h
3
;
U
3
= u
3
Y
1
h
2
Y
2
h
1
k/ :
(27)
The nonzero strains of the beams are dened by
e
33
= U
3
;3
= e
0
Y
1
j
2
Y
2
j
1
k
;3
/;
e
31
= U
3
;1
U
1
;3
= c
1
Y
2
j
3
k/
;1
;
e
32
= U
3
;2
U
2
;3
= c
2
Y
1
j
3
k/
;2
;
(28)
where
e
0
= u
3
;3
; c
1
= h
2
u
1
;3
; c
2
= h
1
u
2
;3
;
j
1
= h
1
;3
; j
2
= h
2
;3
; j
3
= h
3
;3
:
(29)
In view of Eqs. (8) and (9), we obtain the base vector a
3
,
dened by
a
3
= t w
;3
Y
a
e
a;3
= u
a
;3
E
a
(1 u
3
;3
k
;3
/)E
3
Y
a
j E
a
(30)
With the use of Eqs. (29) and (30), we have
a
3
= (c
1
h
2
)E
1
(c
2
h
1
)E
2
(1 e
0
k
;3
/)E
3
Y
a
j E
a
: (31)
Since e
3
= h
2
E
1
h
1
E
2
E
3
(see Eq. (23)), Eq. (31) can be
rewritten as
a
3
= s e
3
k
;3
/E
3
Y
a
j E
a
; (32)
where s = c
1
E
1
c
2
E
2
e
0
E
3
. Comparing Eq. (30) with
Eq. (32) leads to
t = s e
3
(33)
For later use, we introduce the following curvature strains:
e
a;3
= C
a
; C
a
= C
aj
E
j
(34)
Using Eqs. (24) and (34), the components of C
a
are given
by
C
11
= 0; C
12
= j
3
; C
13
= j
2
;
C
21
= j
3
; C
22
= 0; C
23
= j
1
(35)
Note that the curvature strains C
aj
are not symmetric.
3.3
Constitutive equations
In the case of beam theory, the following stress and strain
relationships has often been used:
r
33
= Ee
33
; r
31
= Ge
31
; r
32
= Ge
32
: (36)
For the sectorial properties, the following equations are
assumed to hold by appropriately choosing the origin of
the coordinate system (Y
1
; Y
2
; Y
3
):
_
Y
a
dA = 0;
_
Y
1
Y
2
dA = 0;
_
/dA = 0;
_
/Y
a
dA = 0 :
(37)
The component forms of n and h
a
are written as
n = n
j
E
j
; h
a
= h
aj
E
j
: (38)
Using the denition of stress resultants and moments, and
Eq. (36), we obtain the constitutive equations for the
present problem expressed as
where k
0
is the shearing coefcient (see Cowper, 1966) and
A =
_
dA; S
a
=
_
/
;a
dA;
S
ab
= e
ab
_
Y
a
/
;
b
dA;
S
0
=
_
(/
;1
)
2
(/
;2
)
2
dA;
I
a
=
_
(Y
a
)
2
dA; I
w
=
_
(/)
2
dA ;
(40)
n
1
n
2
n
3
h
12
h
13
h
21
h
23
n
w
h
w
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
GAk
0
0 0 0 0 0 0 GS
1
0
0 GAk
0
0 0 0 0 0 GS
2
0
0 0 EA 0 0 0 0 0 0
0 0 0 GI
1
0 0 0 GS
12
0
0 0 0 0 EI
1
0 0 0 0
0 0 0 0 0 GI
2
0 GS
21
0
0 0 0 0 0 0 EI
2
0 0
GS
1
GS
2
0 GS
12
0 GS
21
0 GS
0
0
0 0 0 0 0 0 0 0 EI
w
_

_
_

_
c
1
c
2
e
0
C
12
C
13
C
21
C
23
k
k
;3
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
(39)
274
in which e
ab
is the permutation symbol. According to
Eq. (20), the moment vector is dened by
m = E
a
h
a
= m
j
E
j
; (41)
where m
a
and m
3
are the bending and torsional moments,
respectively, and given by
m
1
= h
12
= EI
2
j
1
;
m
2
= h
13
= EI
1
j
2
;
m
3
= h
12
h
21
= G(I
1
I
2
)j
3
G(S
12
S
21
)k :
(42)
4
Non-linear theory of space-curved beams, with arbitrary
cross-sections, undergoing arbitrary stretches and
rotations, under bending and torsion
4.1
Balance laws
The LMB and AMB equations for a 3-D continuum body
are written as (see Atluri, 1983)
r
1
;1
r
2
;2
r
3
;3
q
0
B = 0;
a
1
r
1
a
2
r
2
a
3
r
3
= 0 ;
(43)
where ()
;m
is the partial differentiation with respect to the
reference coordinates, r
m
the rst PiolaKirchhoff stress
vector, B the body force vector and q
0
the density in the
undeformed state.
In the case of nonlinear problem, the warping dis-
placement vector is assumed to take the form such that
w = /(Y
a
)s(Y
3
). The details of the warping displacement
will be described later.
Multiplying Eq. (43a) by dA, integrating it and using the
divergence theorem, we obtain the LMB equation of the
beam expressed as
_
(r
1
;1
r
2
;2
r
3
;3
q
0
B)dA = 0
= (LMB) : n
;3
b = 0 ;
(44)
where
n =
_
r
3
dA;
b =
_
q
0
BdA
_
r
1
m
1
ds
_
r
2
m
2
ds :
(45)
Multiplying Eq. (43b) by dA, integrating it and using the
divergence theorem, we obtain the AMB equation ex-
pressed as
_
(e
a
r
a
t r
3
Y
a
e
a
;
3
r
3
s /;
a
r
a
s;
3
/r
3
)dA = 0;
= (AMB) : e
a
n
a
t n e
a
;
3
h
a
s n
w
s;
3
h
w
= 0 ; (46)
where
n
a
=
_
r
a
dA; h
a
=
_
r
3
Y
a
dA;
n
w
=
_
/;
a
r
a
dA; h
w
=
_
/r
3
dA :
(47)
Multiplying Eq. (43a) by Y
a
dA, integrating over the cross
section and using the divergence theorem, we obtain the
DMB equation of the beam expressed as
_
(r
1
;
1
r
2
;
2
r
3
;
3
q
0
B)Y
a
dA = 0
= (DMB) : n
a
h
a
;
3
p
a
= 0 ;
(48)
where p
a
=
_
q
0
BY
a
dA
_
r
1
Y
a
m
1
ds
_
r
2
Y
a
m
2
ds.
Eliminating n
a
from Eqs. (46) and (48), we obtain the
moment equilibrium equation expressed as
m;
3
t n s n
w
s;
3
h
w
q = 0 ; (49)
where
m = e
a
h
a
; q = e
a
p
a
: (50)
Multiplying Eq. (43a) by / dA, integrating over the cross-
section and using the divergence theorem, we obtain the
balance equation for the bimoment (BEB) expressed as
_
(r
1
;
1
r
2
;
2
r
3
;
3
q
0
B)/dA = 0;
= (BEB) : n
w
h
w
;
3
d = 0 ;
(51)
where d =
_
q
0
B/dA
_
r
a
/m
a
ds.
Since the rst PiolaKirchhoff stress tensor is the two
point tensor (Atluri, 1983), the stress-resultant vectors
dened above are associated with the deformed base
vectors. In the following, we introduce the stress-
resultant vectors associated with the undeformed base
vectors:
N
a
= R
t
n
a
; N = R
t
n; H
a
= R
t
h
a
; M = R
t
m;
N
w
= R
t
n
w
; H
w
= R
t
h
w
:
(52)
The stress resultant vectors expressed by small letters
are dened by using the 1st PiolaKirchhoff stress ten-
sor. The relation between the 1st PiolaKirchhoff stress
tensor and the BiotLure stress tensor (see Atluri, 1983)
shows that the stress resultant vectors dened by
Eq. (52) correspond to the BiotLure type stress
resultants.
4.2
Kinematics
The deformed base vectors are dened in Eq. (8). In a
manner similar to that of the linear beam theory, the base
vector a
3
can be written as
a
3
= t w;
3
Y
a
e
a
;
3
= s e
3
w;
3
Y
a
j e
a
(53)
The strain vector s and the curvature vector j are dened
as
s = (X u);
3
RE
3
;
j =
1
2
e
ijk
RE
i
(RE
j
);
3
(RE
k
)
(54)
where e
ijk
is the permutation tensor. Since the beam has
the initial curvature, the curvature vector due to the
deformation is written as
^ j = j RK : (55)
275
Note that the strain vectors dened above are associated
with the deformed base vectors.
In the following, we introduce the strain vectors S and C
associated with the undeformed base vectors:
t = R(S E
3
); S = S
j
E
j
;
e
j;3
= R(C K) E
j
; C = C
j
E
j
:
(56)
It is simply shown that
s = RS; ^ j = RC : (57)
For later use, we introduce the strain vectors T, C
j
, U and
W dened by
t = RT; e
j;3
= RC
j
; s = RU; s;
3
= RW ; (58)
where T = T
j
E
j
and C
j
= C
jk
E
k
. The component forms for
U and W will be discussed later. The strain vector T has
been used by Reissner (1981). The relationship between T
and S is given, with the use of Eqs. (56) and (58), by
T = S E
3
: (59)
In the case of linear problem, T is written as
T = c
1
E
1
c
2
E
2
(1 e
0
)E
3
: (60)
Next, we consider the kinematic meaning of C
j
. Let k be
the skewsymmetric tensor dened by
k =
0 j
3
j
2
j
3
0 j
1
j
2
j
1
0
_
_
_
_
; e
j;3
= ke
j
: (61)
Then, it follows from Eq. (61) that
e
j;3
= ke
j
= RR
t
kRR
t
e
j
= RWE
j
; (62)
where W = R
t
kR is the curvature tensor in the material
description (see Sect. 5 for objective strain measures).
Comparing Eq. (58) with Eq. (62) leads to
C
j
= WE
j
: (63)
This equation shows that C
j
is the curvature vector in the
material description.
Let us consider the frame-invariance of the strain
components in the spatial description. In view of Eq. (54),
the strain components are dened by
s
i
= (t e
3
) e
i
;
j
i
=
1
2
e
ijk
e
j;3
e
k
:
(64)
Let (

) denote the value associated with the conguration


obtained by superposing an arbitrary constant rigid-body
motion. Since Eq. (64) holds after the rigid-body motion,
we have
s
i
= (

t e
3
) e
i
;
j
i
=
1
2
e
ijk
e
j;3
e
k
:
(65)
Let K
R
and u
R
denote the constant rigid-body rotation and
displacement vector, respectively. Then we have

t = K
R
(X u u
R
)
;3
= K
R
(X u)
;3
= K
R
t
e
i
= K
R
e
i
:
(66)
Substituting Eq. (66) into Eq. (65) leads to
s
i
= (tK
t
R
e
3
K
t
R
)K
R
e
i
= (t e
3
) e
i
= s
i
;
j
i
=
1
2
e
ijk
e
j;3
K
t
R
K
R
e
k
=
1
2
e
ijk
e
j;3
e
k
= j
i
:
(67)
Since we obtain s
i
= s
i
and j
i
= j
i
, the frame-invariance
of the strain components are satised. The frame-invari-
ance of strain components associated with the warping
displacement is shown in the same manner as discussed
before. The frame-invariance of variables in the material
description is automatically satised since these strain
vectors are associated with the undeformed base vectors
(see Ogden, 1984). Therefore, the strain vectors H; C; T;
and C
j
are frame-invariant.
4.3
Stress power
Stress power (S.P.) for 3-D body is dened by (see Atluri,
1983)
S.P. =
_
r :
_
F
t
dV =
_
A
j
r_ a
j
dV ; (68)
where r is the rst PiolaKirchhoff stress tensor, F the
deformation gradient tensor and (
v
) the time derivative.
First, we consider the stress power expressed in terms of
the values in the spatial description. Using Eqs. (8) and
(9), we obtain the time derivatives of a
a;
t and e
a;3
, which
are expressed as
_ a
a
= _ e
a
/
;a
_ s = W e
a
/
;a
_ s;
_
t = _ s _ e
3
= _ s W s W t
_ e
a;3
= _ j e
a
j _ e
a
= _ j e
a
(W j) e
a
W (j e
a
) ;
(69)
where
_
RR
t
= W I. Then, the stress power of the beam is
written as
S.P. =
_
[n s

m j

n
w
s

h
w
s

;3
W (e
a
n
a
t ne
a;3
h
a
s n
w
s
;3
h
w
)[dL
(70)
in which
(v

) = o(v)=ot W (v) : (71)


It should be noted that the underlined term in the right
hand side of Eq. (70) takes the same form as that of AMB
in Eq. (46). When the AMB equation is satised in a weak
sense, the expression for the stress power leads to the
conjugate relationships between the stress resultants and
moments and the strain measures.
When the strain vectors S and C in the material de-
scription are used to construct the S.P., the time deriva-
tives of t and e
a;3
are given by
_
t =
_
R(S E
3
) R
_
S = W t R
_
S;
_ e
a;3
= R(C K) E
a
R(
_
C E
a
)
= W e
a;3
(R
_
C) e
a
;
(72)
where the following formulae is used:
276
R(A B) = RA RB : (73)
The time derivative of a
a
in the material description takes
the same form as that of Eq. (69). Using Eq. (72), we
obtain the following S.P. in a material description:
S.P. =
_
[N
_
SM
_
CN
w

_
UH
w

_
W
W (e
a
n
a
t ne
a;3
h
a
sn
w
s
;3
h
w
)[dL :
(74)
Equation (74) shows that N, M, N
w
and H
w
are conjugate
with S, C, U and W, respectively, when the AMB equation
is satised in a weak sense.
When the strain vectors T and C
a
are used instead of S
and C, the time derivatives of t and e
a;3
are given by
_
t =
_
RT R
_
T = W t R
_
T;
_ e
a;3
=
_
RC
a
R
_
C
a
= W e
a;3
R
_
C
a
:
(75)
Then the S.P. is expressed as
S.P. =
_
[N
_
TH
a

_
C
a
N
w

_
UH
w

_
W
W (e
a
n
a
t ne
a;3
h
a
s n
w
s
;3
h
w
)[ dL
(76)
It follows from Eq. (76) that N and H
a
are conjugate with T
and C
a
, respectively, when the AMB equation is satised in
a weak sense.
4.4
Invariance of strain energy function
Before the discussion of invariance of strain energy func-
tion, we consider the balance equations under an additional
assumption. In the Sect. 4.1, the kinematic assumptions are
used to derive the balance equations. Here, we introduce an
additional assumption for the stress states. In the existing
beam theory, the stress resultant vectors n
a
do not appear
explicitly. The vectors n
a
are not related with the conjugate
strains through the constitutive equations. It might be
appropriate, therefore, to assume that n
a
= 0. In this sec-
tion, in addition to the kinematic assumptions, we employ
the above assumption. Under these assumptions, the
balance equations for AMB and DMB are written as
(AMB) : t n e
a;
3
h
a
s n
w
s
;3
h
w
= 0;
(DMB) : h
a
;3
p
a
= 0 :
(77)
Using the above AMB and DMB equations, we obtain the
following moment equilibrium equation:
m
;3
t n s n
w
s
;3
h
w
q = 0 : (78)
The above equation coincides with Eq. (49). Note that the
conventional moment equilibrium equation is recovered
even though we assume that n
a
= 0.
As shown before, we have derived the conjugate rela-
tions between the strain and stress measures. On the basis
of the present results, the following strain energy functions
are proposed:
W = W(s; j; s; s
;3
); W = W(S; C; U; W)
and
W = W(T; C
a
; U; W) : (79)
We consider, herein, the invariance condition of strain
energy function. Since the Lagrangian formulation has
always been used in solid mechanics, we pay attention to
the strain energy function in a material description. Let a
frame F be in a relative rigid motion with respect to
another frame F
+
. After the rigid rotation, the vectors in
the original frame F will appear in the frame F
+
as
S
+
= KS; C
+
= KC; T
+
= KT; C
+
a
= KC
a
;
U
+
= KU; W
+
= KW; E
+
j
= KE
j
;
(80)
where K is the proper orthogonal tensor.
First, we consider the strain energy function expressed
as W = W(S; C; U; W). After the rigid motion, the
strain energy function may be expressed as W
+
=
W
+
(S
+
; C
+
; U
+
; W
+
). The variation of W
+
is given by
dW
+
=
oW
+
oS
+
dS
+

oW
+
oC
+
dC
+

oW
+
oU
+
U
+

oW
+
oW
+
dW
+
=
oW
oS
K
t
(dKS KdS)
oW
oC
K
t
(dKC KdC)

oW
oU
K
t
(dKUKdU)
oW
oW
K
t
(dKWKdW)
=
oW
oS
dS
oW
oC
dC
oW
oU
dU
oW
oW
dW
N (dw S) M (dw C)
N
w
(dw U) M
w
(dw W)
= dW dw R
t
s n RC m s n
w
s
;3
h
w
; (81)
where dW I = K
t
K and Eq. (73) is used. Since the un-
derlined term in Eq. (81) does not vanish in general, we
have dW
+
,= dW. The invariance of strain energy function
asserts that dW
+
= dW. As shown above, when the strain
vectors S and C are used to construct the strain energy
function, the invariance condition of strain energy func-
tion is not satised.
Second, we consider the strain energy function ex-
pressed as W = W(T; C
a
; U; W). After the rigid motion,
the strain energy function may be expressed as W
+
=
W
+
(T
+
; C
+
a
; U
+
; W
+
). The variation of W
+
is given by
dW
+
=
oW
+
oT
+
dT
+

oW
+
oC
+
a
dC
a
+
oW
+
oU
+
U
+

oW
+
oW
+
dW
+
=
oW
oT
K
t
(dKT KdT)
oW
oC
a
K
t
(dKC
a
KdC
a
)

oW
oU
K
t
(dKUKdU)
oW
oW
K
t
(dKWKdW)
=
oW
oT
dT
oW
oC
a
dC
a

oW
oU
dU
oW
oW
dW
N (dw T) H
a
(dw C
a
)
N
w
(dw U) H
w
(dw W)
= dW dw R
t
(t n e
a;3
h
a
s n
w
s
;3
h
w
) ; (82)
277
It should be noted that the underlined term in Eq. (82)
denote the AMB condition under the kinematic and stress
state assumptions, which is shown in Eq. (77a). According
to Eq. (82), we have dW
+
= dW as long as the AMB
equation (77) is satised. Therefore, when the strain vec-
tors T; C
a
; U; W are used to construct the strain energy
function, the invariance condition of strain energy func-
tion is satised.
Suetake, Iura and Atluri (1999) have derived the
objective strain energy function for a thick shell theory.
They have introduced the strain measures which are
conjugate with the director moment vector. In the beam
theory developed herein, we have used the strain
measures which are not conjugate with the conven-
tional moment vector but with the director moment
vector.
4.5
Warping displacement
According to the kinematic hypothesis (1), we have
a
1
a
1
= Constant; a
2
a
2
= Constant : (83)
We assume, herein, that the shear strains due to shear
stresses in equilibrium with the changes of normal
stresses are small and can be neglected. Furthermore,
the bending curvatures are also neglected in the
equilibrium equation. Then, for isotropic materials, we
have
(a
1
a
3
);
1
(a
2
a
3
);
2
= 0 : (84)
Using Eq. (83) and the relations such that a
3;a
= a
a;3
, the
Eq. (84) can be rewritten as
(a
1;1
a
2;2
) a
3
= 0 : (85)
Since a
a
= e
a
w;
a
(see Eq. (8)), it follows from Eq. (85)
that
(w;
11
w;
22
) a
3
= 0 : (86)
With the use of the separation of variables, we may
write that w = /(Y
a
)s(Y
3
) where / is the warping
function. Then, using Eq. (86) and noting that s a
3
,= 0,
we have
/;
11
/;
22
= 0 : (87)
The above equation shows that the warping function for
the present problem takes the same form as that of Saint-
Venant's torsion.
In the case of the linear problem, s is assumed to be
s = kE
3
. It is difcult, for the nonlinear problem, to dene
the direction of warping displacement. The extension of
the linear problem to the nonlinear one may lead to
s = ke
3
; the direction of warping displacement is normal
to the deformed cross-section. This might be accepted
when the shearing deformation is neglected. In this case, U
and W are written as
U = kE
3
; W = j
2
kE
1
j
1
kE
2
k
;3
E
3
(88)
Once the shearing deformation can not be neglected, it
might be proper to assume that s = kt= t | |; the direction
of warping displacement is not normal to the deformed
cross-section.
4.6
Constitutive equations
Since the use of strain measures T and C
a
leads to the
objective strain energy function, we consider the con-
stitutive equations expressed in terms of the conjugate
stress measures N and H
a
. In what follows, attention is
restricted to the elastic and isotropic cases. The warping
displacement is assumed to be w = /k e
3
. For present
purposes, the constitutive equations for the beam are
given by
where
^
C
ij
= C
ij
K
ij
; K
ij
= e
ijk
K
k
: (90)
The moment vector M is related with H
a
through Eqs. (50)
and (52), and expressed as
M = R
t
m = R
t
(e
a
h
a
) = E
a
H
a
= H
a j
e
a jk
E
k
:
(91)
The components of moment vector are given by
M
1
= H
23
; M
2
= H
13
; M
3
= H
12
H
21
: (92)
According to the denition of the permutation tensor, the
components H
11
and H
22
do not appear in Eq. (91). This
fact coincides with that C
11
= C
22
= 0.
5
Objectivity of strain measures
Criseld and Jelenic (1999) have discussed the objec-
tivity of strain measures in the geometrically exact 3-D
beam theory. They have concluded that all of the
N
1
N
2
N
3
H
12
H
13
H
21
H
23
N
w
H
w
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
=
GAk
0
0 0 0 0 0 0 GS
1
0
0 GAk
0
0 0 0 0 0 GS
2
0
0 0 EA 0 0 0 0 0 0
0 0 0 GI
1
0 0 0 GS
12
0
0 0 0 0 EI
1
0 0 0 0
0 0 0 0 0 GI
2
0 GS
21
0
0 0 0 0 0 0 EI
2
0 0
GS
1
GS
2
0 GS
12
0 GS
21
0 GS
0
0
0 0 0 0 0 0 0 0 EI
w
_

_
_

_
T
1
T
2
T
3
1
^
C
12
^
C
13
^
C
21
^
C
23
k
k ;
3
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
(89)
278
available nite element implementation of the theory are
non-objective. We discuss herein the objectivity or
invariance of the curvature strains under rigid-body
rotations. First, we will present the requirement for the
objectivity in the geometrically exact beam theory.
Second, with the use of a simple example, we will show
that Criseld and Jelenic (1999) have misled the
requirement for the objectivity. A numerical example
proposed by them is used to show that the nite ele-
ment implementation developed by Iura and Atluri
(1988, 1989) is objective.
Let (v

) denote the value associated with the congura-


tion obtained by superposing an arbitrary constant rigid-
body motion. Let

E
i
be maps of the base vectors E
i
after a
rigid rotation: that is,

E
i
= K
R
E
i
where K
R
is the rigid-
body rotation tensor. Let H denote the axial vector of the
nite rotation tensor R, which follows the transformation
rule expressed as

H = K
R
H : (93)
The components of curvature vector and axial vector are
expressed as
C = C
i
E
i
; H = H
i
E
i
: (94)
After the rigid-body rotation, the components of these
vectors may be expressed as

C =

C
i

E
i
=

C
#
i
E
i
;

H =

H
i

E
i
=

H
#
i
E
i
; (95)
where (v)
#
denote the value referred to the undeformed
base vectors E
i
. Since

H = K
R
H = K
R
H
i
E
i
and

H =

H
i

E
i
,
we have H
i
=

H
i
. The components of curvature vector are
determined from the components of axial vector. Since
H
i
=

H
i
holds, we have
C
i
=

C
i
=

C
#
j
(E
j


E
i
) : (96)
This is the requirement for the objectivity of components
of curvature vector. The requirement for the objectivity of
curvature vector is written as

C =

K
R
C : (97)
The objectivity of the curvature tensor is written as

C = K
R
CK
t
R
: (98)
Let us consider a simple example for understanding the
objectivity of strain measures. Figure 2 shows a beam
subjected to a torque. Stage-A shows the undeformed
beam with the base vectors E
i
= I
i
. Stage-B shows the
deformed beam subjected to a torque. Since the torque is
applied around the beam axis, we have e
3
= E
3
. Stage-C
shows the conguration obtained by superposing a rigid-
body rotation K
R
onto the stage-B so that e
3
= E
2
. The
value (v

) is associated with the stage-C. In the above


manipulation, a rigid-translation is neglected for simplic-
ity. In this example, a non-zero curvature is a twist de-
noted by s. It is clear that the curvature vector in the stage
B is expressed as as C = sE
3
. Since no relative motions
occur during the rigid-body motion, the curvature vector
in the stage-C is expressed as

C = sE
2
. It is clear, therefore,
that
C ,=

C or sE
3
,= E
2
: (99)
According to Criseld and Jelenic (1999), the objecti-
vity of strain tensor in the material description is
dened by (see Appendix B of Criseld and Jelenic,
1999)
C =

C : (100)
The simple example described above shows that
Eq. (100) does not hold; Criseld and Jelenic (1999) have
misled the requirement for the objectivity of curvature
tensor.
Let us consider a numerical example proposed by
Criseld and Jelenic (1999). The rotation vectors at both
end-nodes of nite element w
1
and w
2
and the rigid rota-
tion vector w
R
are given by
w
1
=
1
0:5
0:25
_

_
_

_
; w
2
=
0:4
0:7
0:1
_

_
_

_
;
w
R
=
0:2
1:2
0:5
_

_
_

_
; (101)
where these vectors are referred to the undeformed base
vectors. The components of curvature vector associated
with E
i
are given by
C
i
= 1:2746 1:2676 0:4035
t
: (102)
The above result is the sama as that of Criseld and Jelenic
(1999).
Following the objectivity rule and Eq. (95), we obtain
the rotation vectors associated with E
j
expressed as

w
#
1
=
0:2442
0:8031
0:7797
_
_
_
_
_
_
;

w
#
2
=
0:3060
0:6773
0:3280
_
_
_
_
_
_
: (103)
The mean value and difference are obtained as
Fig. 2. Beam subjected to pure torsion
279

w
#
a
=
1
2
(

w
#
1


w
#
2
) =
0:2751
0:0629
0:2259
_

_
_

_
;
1
L

w
#
d
=
1
L
(

w
#
2


w
#
1
) =
0:0618
1:4804
1:1077
_

_
_

_
:
(104)
Using the denition of curvature tensor, we obtain the
components of curvature tensor associated with E
i
written
as

C
#
i
= 0:0853 1:6094 0:8926
t
: (105)
In view of Eq. (96), we obtain the components of curvature
vector associated with

E
i
expressed as

C
i
= 1:2746 1:2676 0:4035
t
: (106)
Comparing Eq. (102) with Eq. (106), we obtain C
i
=

C
i
;
the requirement of objectivity is satised. This numeri-
cal example shows that the formulation developed by
Iura and Atluri (1988, 1989) and the present one are
objective.
6
Variational principle for the beam
On the basis of variational principle developed by Atluri
(1979, 1980, 1983) and Atluri and Cazzani (1995), we
derive a generalized functional for the beam. Special at-
tention should be paid to the rotational parameters. Since
the variation d/ is conjugate with moment equilibrium
equation, this Eulerian rotational variation has often been
used for the nite element analysis. However, the vector
/ does not exist so that the functional can not be ex-
pressed in terms of this rotational vector. Iura and Atluri
(1988, 1989) have employed the variation of nite rota-
tion vector, which is the Lagrangian rotational variation.
As a result, the resulting tangent stiffness is always
symmetric.
On the basis of the present results, Atluri's variational
functional is applied to the beam problem and expressed
as
F
0
=
_
[W
0
(T; C
a
; U; W) ^ n
a
(e
a
RE
a
)
^ n (X u)
;3
RT
^
h
a
(e
a;3
RC
a
)
^ n
w
(s RU)
^
h
w
(s
;3
RW)
b u p
a
(e
a
E
a
) d s[dL
[^ n (u ~ u)
^
h
a
(e
a
~e
a
)
^
h
w
(s ~s)[
s
u
[~ n u
~
h
a
(e
a
E
a
)
~
h
w
s[
s
r
; (107)
where W
0
is the strain energy function of the beam and
^
(
v
)
the Lagranege multiplier. Note that the nite rotation
tensor R is a function of nite rotation vector h. The rst
variation of F
0
is given by
dF
0
=
_ _
oW
0
oT
R
t
^ n
_ _
dT

oW
0
oC
a
R
t
^
h
a
_ _
dC
a

oW
0
oU
R
t
^ n
w
_ _
dU

oW
0
oW
R
t
^
h
w
_ _
dWd^ n
a
(e
a
RE
a
)
d^ n (Xu)
;3
RT
d
^
h
a
(e
a;3
RC
a
) d^ n
w
(s RU)
d
^
h
w
(s
;3
RW)
(^ n
;3
b) du (n
a
h
a
;3
p
a
) de
a
(h
w
;
3
n
w
d) ds
(RE
a
^ n
a
RT ^ n RC
a

^
h
a
RUn
w
RWh
w
) Xdh[dL
[d^ n (u ~ u) d
^
h
a
(e
a
~e
a
) d
^
h
w
(s ~s)[
s
u
[(~ n ^ n) du (
~
h
a

^
h
a
) de
a
(
~
h
w

^
h
w
) ds
_
s
r
(108)
The Euler equations obtained from Eq. (108) are written as
(Constitutive equations):
oW
0
oT
R
t
^ n = 0;
oW
0
oC
a
R
t
^
h
a
= 0;
oW
0
oU
R
t
^ n
w
= 0;
oW
0
oW
R
t
^
h
w
= 0 : (109)
(Compatibility conditions):
e
a
RE
a
= 0; (X u)
;3
RT = 0; e
a;3
RC
a
= 0
s RU = 0; s
;3
RW = 0 : (110)
(Balance equations):
^ n
;3
b = 0;
^ n
a

^
h
a
;3
p
a
= 0 ;
RE
a
^ n
a
RT ^ n RC
a

^
h
a
RU ^ n
w
RW
^
h
w
= 0:
h
w;3
n
w
d = 0 (111)
(Boundary conditions):
u ~ u = 0; e
a
~e
a
= 0; s ~s = 0 on S
u
;
^ n ~ n = 0;
^
h
a

~
h
a
= 0;
^
h
w

~
h
w
= 0 on S
r
: (112)
It should be noted that e
a
= RE
a
is recovered from the
compatibility condition since the Lagrange multipliers ^ n
a
are introduced. The stress resultant vectors n
a
are not
related with the conjugate strain measures. Therefore,
there are no constitutive equations for n
a
. The magnitude
of n
a
might be obtained from the AMB equations. As
shown in Sect. 4.4, the stress vectors n
a
can be neglected to
derive the moment equilibrium equation. Moreover, the
280
invariance condition of strain energy function is satised
under the assumption that n
a
= 0.
When the equation e
a
= RE
a
is taken as subsidiary
condition, we obtain the Hu-Washizu type functional ex-
pressed as
F
HW
=
_
[W
0
(T; C
a
; U; W) ^ n (X u)
;3
RT

^
h
a
(e
a;3
RC
a
) ^ n
w
(s RU)
^
h
w
(s
;3
RW) b u p
a
(e
a
E
a
) d s[dL
[^ n (u ~ u)
^
h
a
(e
a
~e
a
)
^
h
w
(s ~s)[
s
u
[~ n u
~
h
a
(e
a
E
a
) h
w
s[
s
r
; (113)
The rst variation is given by
dF
HW
=
_ _
oW
0
oT
R
t
^ n
_ _
dT
oW
0
oC
a
R
t
^
h
a
_ _
dC
a

oW
0
oT
R
t
^ n
_ _
dT
oW
0
oC
a
R
t
^
h
a
_ _
dC
a
d^ n (Xu)
;3
RT d
^
h
a
(e
a;3
RC
a
)
d^ n
w
(s RU) d
^
h
w
(s
;3
RW)
(^ n
;3
b) du (
^
h
w;3
^ n
w
d) ds
(RT ^ n RC
a

^
h
a
e
a

^
h
a
;3
RU
n
w
RWh
w
e
a
p
a
) Xdh[dL
[d^ n (u ~ u) d
^
h
a
(e
a
~e
a
) d
^
h
w
(s ~s)[
s
u
[(~ n ^ n) du e
a
(
~
h
a

^
h
a
) Xdh
(
~
h
w

^
h
w
) ds
_
s
r
(114)
It should be pointed out that the AMB equation can not be
recovered from the above functional since the equation
e
a
= RE
a
is taken as subsidiary condition. When the
constitutive equations and the compatibility conditions
are satised, the Euler equation associated with dh is
written as
t n m
;3
s n
w
s
;3
h
w
q = 0 (115)
Note that the moment equilibrium equation is obtained
directly from the functional F
HW
.
7
Concluding remarks
A consistent beam theory has been developed in this
paper. The AMB equation plays an important role in the
stress power and also in the invariance condition of
strain energy function. New strain measures are intro-
duced so that the strain energy function is invariant
under a rigid-body motion. An important role of the
stress-states assumptions is emphasized especially when
the invariance condition of strain energy function is
discussed.
On the basis of present results, Atluri's variational
principle is applied to the beam problem. A choice of a
proper nite rotation vector, and unsymmetric curvature
strains, makes it possible for constructing a consistent
variational principle. A generalized variational functional
presented herein leads to the LMB, AMB and DMB equa-
tions as the Euler equations.
References
Atluri SN (1979) On Rate Principle For Finite Strain Analysis of
Elastic and Inelastic Nonlinear Solids. Recent Research on
Mechanical Behavior of Solids. pp. 79107. University of
Tokyo Press
Atluri SN (1980) On some new general and complementary en-
ergy theorems for the rate problems in nite strain, classical
elastoplasticity. J. Struct. Mech. 8: 6192
Atluri SN (1983) Alternate stress and conjugate strain measures,
and mixed variational formulations involving rigid rotations,
for computational analysis of nitely deformed solids, with
application to plates and shellI Theory. Comput. Struct.
18(1): 93116
Atluri SN, Cazzani A (1995) Rotations in computational solid
mechanics. Arch. Compu. Meth. Eng. 2(1): 49138
Cowper GR (1966) The shear coefcient in Timoshenko's beam
theory. J. Appl. Mech. 33: 335340
Criseld MA, Jelenic G (1999) Objectivity of strain measures in
the geometrically exact three-dimensional beam theory and
its nite-element implementation. Proc. R. Soc. Lond. A 455:
11251147
Kollbrunner CF, Basler K (1969) Torsion in Structures, Springer-
Verlag
Iura M, Atluri SN (1988) Dynamic analysis of nitely stretched
and rotated three-dimensional space-curved beams. Comput.
Struct. 29(5): 875889
Iura M, Atluri SN (1989) On a consistent theory, and variational
formulation of nitely stretched and rotated 3-D space-
curved beams. Comput. Mech. 4: 7388
Ogden RW (1984) Non-linear Elastic Deformations, Ellis
Horwood Limited
Suetake Y, Iura M, Atluri SN (1999) Shell theories with drilling
degrees of freedom and geometrical and material assump-
tions. Comput. Mod. Simu. Eng. 4(1): 4249
Reissner E (1981) On nite deformations of space-curved beams.
J. Appl. Math. Phy. (ZAMP) 32: 734744
Vasudevan S, Atluri SN, Iura M (2000) Space-curved beams
undergoing large strains and rotations: Part 2: Complemen-
tary energy based nite elements for isotropic and aniso-
tropic beams, (to appear)
281

Das könnte Ihnen auch gefallen