0% found this document useful (0 votes)

17 views41 pages

IJNME1

Uploaded by

John Manaois

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

17 views41 pages

IJNME1

Uploaded by

John Manaois

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 41

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226437472

Total and Updated Lagrangian Geometrically Exact Beam Elements

Chapter · June 2008

DOI: 10.1007/1-4020-5370-3_658

CITATIONS READS

3 2,881

2 authors:

Jari Mäkinen Heikki Marjamäki

Tampere University Tampere University
38 PUBLICATIONS 522 CITATIONS 11 PUBLICATIONS 102 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jari Mäkinen on 12 March 2015.

The user has requested enhancement of the downloaded file.

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
Int. J. Numer. Meth. Engng 2007; 70:1009–1048
Published online 18 October 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nme.1892

Total Lagrangian Reissner’s geometrically exact beam

element without singularities

Jari Mäkinen∗, †
Institute of Applied Mechanics and Optimization, Tampere University of Technology, P.O. Box 589,
Tampere, FIN-33101, Finland

SUMMARY
In this paper, we introduce a new Reissner’s geometrically exact beam element, which is based on a total
Lagrangian updating procedure. The element has the rotation vector as the dependent variable and the
singularity problems at the rotation angle 2 and its multiples are passed by the change of parametrization
on the rotation manifold. The beam formulation has several benefits such as all the unknown vectors belong
to the same tangential vector space, no need for secondary storage variables, the path-independence in the
static case, any standard time-integration algorithm may be used, and the symmetric stiffness. Copyright
q 2006 John Wiley & Sons, Ltd.

Received 26 October 2005; Revised 18 April 2006; Accepted 30 August 2006

KEY WORDS: non-linear dynamics; geometrically exact beam; finite rotations; rotation manifold

1. INTRODUCTION

A spatial corotational procedure for dynamic cases has been given in [1] where an explicit time-
integration scheme is used without the need for tangential matrices. A spatial corotational beam
element, with the consistent stiffness matrix in static cases, has been introduced in [2] where
the derivative of transformation matrices has been correctly accounted, but the element is limited
to small increments and small local rotations. A fully consistent corotational beam with large
increments and rotations in static cases has been presented in [3] whose consistent mass matrix is
given in [4, Chapter 24.19]. Corotational technique has an advantage, namely, a relatively simple

∗ Correspondence to: Jari Mäkinen, Institute of Applied Mechanics and Optimization, Tampere University of
Technology, P.O. Box 589, Tampere, FIN-33101, Finland.
†
E-mail: [email protected]

Contract/grant sponsor: The Academy of Finland; contract/grant number: 206020

Copyright q 2006 John Wiley & Sons, Ltd.

1010 J. MÄKINEN

form of the consistent stiffness matrix, giving an effective formulation especially for static cases.
As disadvantages, we can mention the highly coupled kinetic energy with the relative displacement
and the motion of corotational frame. In addition, applying a corotational technique leads to the
additional kinematic assumptions when a different choice of a corotational frame produces slightly
divergent results.
An approach extended for the spatial beams has been given in [5] where a nine-parameter
representation of rotation is given. These parameters are orthonormal base vectors of a moving
basis, hence six orthonormal constraint conditions per node, in addition to an absolute motion
representation, is applied. In this beam formulation, the mass matrix is constant but singular that
is equivalent to the hiding of constraint equations. The number of displacement variables is 12 per
node with six constraint equations, which considerably diminishes the efficiency of the formulation.
A rotation description for a static spatial beam is derived in [6] where two base vectors in
a cross-section are used for the parametrization of rotation. These base vectors are allowed to
change their lengths and their relative angle, yielding the additional elastic relations. The number
of displacement variables is nine per node if no warping parameters are applied. This formulation
gives a simple handle of vectorial rotation parameters, but the classical Hookean law may yield
erroneous results since a high sensitivity with the extension of base vectors comparing with the
other deformations, a large number of variables per node, and in addition, the moment load is not
easy to apply.
A finite-strain beam element has been proposed in various contents. The assumptions of the
beam kinematics can be divided into two types: the Timoshenko–Reissner-hypothesis and the
Euler–Bernoulli hypothesis. In an Euler–Bernoulli beam theory, the normals of a central line
remain normals in a deformed state with no in-plane or out-of-plane warping deformations of a
cross-section, i.e. the cross-section remains in-plane and one’s shape, and its normal and the tangent
of the central line are parallel in a deformed state. This means that the translational displacement
and rotation fields are kinematically coupled. Therefore, Euler–Bernoulli beam elements are very
rare in a geometrically exact beam theory, but are commonly used in a corotational technique.
In the Timoshenko–Reissner beam hypothesis, transversal shears are allowed, such that a cross-
section remains in-plane and one’s shape in a deformed state, but the normal of a cross-section
is not necessarily parallel with the tangent of the central line. In addition, all warping effects
are excluded in the Timoshenko–Reissner beam hypothesis. A finite-strain beam theory has been
introduced by Reissner for planar beams in [7] and for spatial beams in [8] where the author
derived the beam equations from the classical curve theory.
Here we consider a geometrically exact beam theory if no other kinematic simplifications during
derivation are applied than the basic kinematic hypothesis, and a continuum-consistent beam theory,
which is geometrically exact, and in addition, all the in-plane and out-of plane warping effects are
included. Hence, a beam theory gives equivalent solutions with a corresponding continuum theory
[9]. In general, an in-plane warping is due to an axial extension and an out-of-plane warping is
due to a bending, torsion, and transversal shear.
In modern contents, the geometrically exact spatial beam theory with finite element implementa-
tions has been mainly developed by Simo and Vu-Quoc, Cardona and Géradin and Ibrahimbegović
et al. In the paper [10], the author gives a dynamic formulation for Reissner’s beam, and its finite
element implementation, in static cases, is given in [11]. In that paper, a spin rotation vector is
used as a dependent variable, and a placement is updated with the aid of a rotation tensor and
an exponential mapping, where memory requirements are reduced using quaternion parameters.
Here we use the phrase spin rotation vector that is a vector on a tangent space of a manifold,

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1011

and two successive spin rotation vectors belong to different tangent spaces, thus they must not be
added by the parallelogram rule. The main drawbacks of this formulation are that the consistent
stiffness tensor is an unsymmetrical tensor away from an equilibrium, the need for secondary
storage variables (quaternions) and their manipulations, and the spin rotation field has to be inter-
polated an inconsistent way, and moreover, a solution has a path-dependent property even when
a conservative loading is applied. A finite element implementation in the dynamic case by the
authors is given in [12] where spatial and material quantities are exploited leading to the large
number of secondary storage variables and calculations between them. This dynamic finite element
implementation is quite different from the static implementation; moreover, the dynamic formula-
tion has a simplification in their Newmark time-stepping method, see proof in the paper [13]. An
updating procedure where a spin rotational vector is used as a dependent variable is also called an
Eulerian formulation.
In an important paper [14], the authors give another finite element implementation for Reissner-
beam element with a different updating procedure. They named the formulations as Eulerian, total
Lagrangian, and updated Lagrangian and gave a finite element implementation for an updated
Lagrangian formulation with the rotation vector as a dependent variable. The updated Lagrangian
formulation can bypass the singularity problem of the total Lagrangian formulation, which is
singular at the rotation angle 2 and its multiples. The updated Lagrangian formulation has
additional benefits such as a fully symmetrical stiffness tensor when applying a conservative
loading and any single-step time-integration algorithm can be used since, in this formulation,
the changes of the rotation vector belong to the same tangent space of a manifold. The updated
Lagrangian formulation requires some secondary storage variables for the curvature and rotation
vector, at every spatial integration point. The authors made some simplifications in the tangent
operator of the inertial force vector by neglecting centrifugal and gyroscopic tensors; and in
addition, the tangent stiffness tensor was simplified. The authors have also written the text book
[15] where a finite element approach for flexible multibody dynamics is given.
A total Lagrangian formulation in static cases with the consistent stiffness tensor is given
in [16]. The consistent stiffness tensor, which is a symmetrical tensor, has the same form in the
total and updated Lagrangian formulation and is considerably more complicated than the consistent
stiffness tensor in an Eulerian formulation, which leads to an unsymmetrical stiffness tensor away
from an equilibrium. This symmetry issue of tangential operations has been studied thoughtfully
in [17] with the perspective of differential geometry.
A total Lagrangian formulation in a static case with a conservative loading has an important
property that is path-independence, whereas an updated Lagrangian formulation is path-dependent.
Lagrangian formulations have a consistent interpolation, while in Eulerian formulations the in-
terpolation has to apply within an approximate, inconsistent, way. As we have noted earlier,
total Lagrangian formulations have singularity at the rotation angle 2 and its multiples that is a
remarkable restriction, especially in dynamic cases.
Different updated Lagrangian and Eulerian formulations in dynamic cases are introduced in [18],
wherein only spatial quantities are present. These formulations have been developed from [12]
and they have the same simplification in the Newmark time-stepping scheme that yields a reduced
form of the force vector and tangent tensors, as pointed out in [13]. We suppose these simplifica-
tions having only an insignificant effect on numerical solutions. Given formulation requires some
secondary storage variables, as updated Lagrangian and Eulerian formulations generally need.
Finally, we can summarize that a total Lagrangian geometrically exact finite element formu-
lation is competitive if its major drawback, singularity, could be bypassed. The total Lagrangian

formulation has several benefits such as all dependent variable vectors belong to the same
tangential vector space, no need for secondary storage variables, the path-independence property
(in static cases), any standard time-integration algorithm may be used, the symmetric stiffness
tensor, a simple form of the kinetic energy and all non-linear effects are included. Hence, we will
develop a singularity-free, geometrically exact, finite element formulation with a total Lagrangian
updating procedure and we will explicitly give the consistent tangential tensors.
This paper is organized as follows: the rotation manifold including its tangential spaces is
carefully studied in Section 2, the total Lagrangian formulation for the geometrically exact
Reissner’s beam element is introduced in Section 3, the objectivity of the strain measures is
proven in Section 3.1.1, and finally, the numerical results are shown in Section 4.

2. ROTATION MANIFOLD

In this section, we introduce mathematical preliminaries that are required to understand the rotation,
angular velocity, and angular acceleration vectors that are vectors on the rotation manifold. Some
elementary knowledge of differential geometry is necessary to understand the rotation vector that
is a vector of a tangential vector space of a manifold, where the manifold is a Lie group of special
orthogonal tensors. Especially, we note that an interpolation can be consistently accomplished if
the rotational vectors of each node belong to the same tangential vector space.
Figure 1 shows a basic idea of differentiable manifold. A differentiable manifold can be mapped
from a chart in a parameter space into a chart of manifold in an embedding space. The change of
parametrization is differentiable for differentiable manifolds. Let u() be a parametrized curve in
the manifold M through the base point x ∈ M such that u( = 0) = x. The tangent of curve u()

Figure 1. Geometric interpretation for parametrization of the manifold when n = 3 and d = 2.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1013

at = 0 to the manifold M is defined as

u() − u(0)
ẋ = lim , where u(0) = x, u() ∈ M (1)
→0
The tangent vector x belongs to a tangent space of the manifold, namely ẋ ∈ Tx M. The tangent
(vector) space Tx M is a set of tangent vectors at x ∈ M. A tangent bundle T M is defined by a
union of the tangent spaces on the manifold M at its every point

T M := (x, Tx M) (2)
x∈M

The dimension of the tangent bundle is twice as the dimension of the manifold M. Especially, the
pair of state vectors, the placement x(t) ∈ M and its velocity vector v(t) ∈ Tx M, belongs to the
tangent bundle, (x, v)(t) ∈ T M.
The special non-commutative Lie group of the proper orthogonal linear transformations is
defined as
SO(3) := {R:E 3 → E 3 linear | RT R = I, det R = 1} (3)
where E3indicates a three-dimensional Euclidean vector space, see Lie groups in [19]. The rotation
group is a three-dimensional manifold with differentiable structure, i.e. Lie group. Euler angles
are frequently used presentations in the literature of analytical dynamics. However, a simpler and
more useful parametrization can be obtained if the parameters are canonical, i.e. the rotation vector
parametrization.
Definition 2.1
A rotation operator transforms linearly and isometrically a vector into another vector in a rotation
motion that is represented by a rotation vector. The rotation operator R ∈ SO(3) is defined with
the aid of the rotation vector W ∈ E 3 by
sin 1 − cos 2
R := I + W+ W = exp(W), = W
2
where W is the skew-symmetric tensor of its axial vector W. The rotation operator and the rotation
vector are related by an exponential mapping [19, Chapter 9.2]
Definition 2.1 gives a canonical parametrization of the rotation manifold SO(3). The parametriza-
tion can represent a rotation operator only locally, and there exists no parametrization that is global
as well as non-singular. Note that a parametrization is a mapping from an open set of Euclidean
space into some open set of the manifold. The rotation vector parametrization is singular at the
rotation angle equal to 2 and its multiples. It is clear that the singularity naturally appears in
dynamical analysis with large rotations, and cannot be omitted. Singularity should be considered
as a non-differentiable hole that must not be omitted by skipping. The singularity is due to fact
that the rotation manifold is compact, and there exists not a single continuous parametrization
from an open set of the Euclidean space E 3 onto this compact manifold, see details in [20].
with respect to the parameter t at t = 0, according to (1),
Differentiating the expression exp(t W)
gives the tangent of the rotation operator at the identity, yielding

d exp(t W)
=W (4)
dt
t=0

Thus, the skew-symmetric tensor W belongs to the tangential space of the rotation manifold.
∈ TI SO(3), where the identity I ∈ SO(3), represents a base point of the rotation
The notation W
manifold. A base point is a point of the manifold where its tangent space is induced. The set of
skew-symmetric tensors so(3) is defined as
so(3) := {
A:E 3 → E 3 linear |
AT = −
A} (5)

Definition 2.2
Let us consider a rotation vector W with a rotation angle larger than zero and less than 2,
i.e. 0<<2, then its complement rotation vector WC is defined by
2
WC := W − W, := W

Then the rotation angle of the complement rotation vector is C = 2 − and the rotation axis is
nC = − n.

After substituting the complement rotation vector into Definition 2.1, we notice that the rotation
vector and its complement represent the same rotation operator, i.e. R(WC ) = R(W). Definition 2.2
is a change of parametrization in the parameter space E 3 , see Figures 1 and 2. This change of
parametrization is a continuously differentiable mapping on the open domain 0<<2, giving
a smooth construction of the rotation manifold SO(3) at this domain. The complement of a
complement rotation vector is a rotation vector itself, i.e. (WC )C = W. We could represent the
rotation manifold globally with these two parametrization charts. When a rotation angle exceeds
straight angle (>), we accomplish the change of parametrization according to Definition 2.2,
giving a new rotation angle less than the straight angle. Thus, we never get into trouble with

Figure 2. The change of parametrization in the parameter space E 3 for the

canonical representation of rotation manifold.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1015

the singularity at = 2. As it is illustrated in Figure 2, the change of parametrization maps

the rotation angle > from the outside of straight angle into the inside of straight angle. Note
that there exists no other canonical parametrizations with a rotation less than perigon such as the
parametrizations given in Definition 2.2.
A rotation operator is an element of Lie group that is a differentiable manifold as well as a
non-commutative group. A compound of successive rotations is also a rotation itself and induces
a Lie group structure with underlying Lie algebra. Compound rotation can be defined by two
different, however, equivalent ways: the material description and the spatial description.

Definition 2.3
We define the material description of a compound rotation by the left translation map LeftR :SO(3)
→ SO(3) as

inc := RRinc = R exp(HR ),
LeftR Rmat inc , R ∈ SO(3)
mat
Rmat

inc is an incremental material rotation operator and HR is an incremental material rotation

where Rmat
vector with respect to the base point R ∈ SO(3).

Definition 2.4
We define the spatial description of a compound rotation by the right translation map RightR :SO(3)
→ SO(3) as
RightR Rinc := Rinc R = exp(
spat spat spat
hR )R, Rinc , R ∈ SO(3)
spat
where Rinc is an incremental spatial rotation operator, and hR is an incremental spatial rotation
vector with respect to the base point R ∈ SO(3).

We use majuscules for material vectors and minuscules for spatial vectors. Material and spatial
rotation incremental tensors and their rotation vectors are related by
spat R RT
inc = R Rinc R,
Rmat hR = RH and hR = RHR
T
(6)
where the first relation is called inner automorphism that is an isomorphism onto itself, the second
relation is a Lie algebra so(3) adjoint transformation AdR H R RT , and the last relation is
R = RH
another Lie algebra adjoint transformation in the Euclidean space with the vector cross-product as
the Lie algebra (E 3 , · × ·).

Definition 2.5 (material tangent space)

Differentiating the material expression of the compound rotation R exp(H) with respect to the
parameter and setting = 0 yields the material tangent space at the base point R ∈ SO(3). This
material tangent space on the rotation manifold SO(3) at any base point R is defined as

mat TR SO(3) := {HR := (R, H) |
R ∈ SO(3), H
with RH; ∈ so(3)}

where an element of the material tangent space is H R ∈ mat TR SO(3) that is a skew-symmetric
R ∈ so(3). The notation (R, H),
tensor, i.e. H the pair of the rotation operator R and the skew-
symmetric tensor H, represents the material skew-symmetric tensor at the base point R ∈ SO(3),
see Figure 3. Hence, we may express that H R is a skew-symmetric tensor, or a tangent tensor, at
the point R on the manifold SO(3). For simplicity, we could omit the base point R by denoting
∈ mat TR SO(3) if there is no danger of confusion.
H

Figure 3. Geometric representation of the material tangent space (on the left) and the spatial tangent space
(on the right) on the rotation manifold SO(3).

Definition 2.6
The spatial tangent space on the rotation manifold SO(3) at any base point R is defined

spat TR SO(3) := {hR := (R, h) | with
hR; R ∈ SO(3),
h ∈ so(3)}

where an element of the spatial tangent space is hR ∈ spat TR SO(3) that is a skew-symmetric tensor,

i.e. hR ∈ so(3). The notation (R, h), the pair of the rotation operator R and the skew-symmetric
tensor h, represents a spatial skew-symmetric tensor at the base point R, see Figure 3. Again, we
could omit the base point R, i.e. h ∈ spat TR SO(3) if there is no danger of confusion.

Let us consider the material form of compound rotation, given in Definition 2.3, with the aid
of -parametrized exponential mappings
+ W)
exp(W = exp(W) R)
exp(H (7)
where we are finding an incremental rotation tensor, the virtual rotation tensor W, such that it

belongs to the same tangential space as the rotation tensor W, i.e. such that W, W ∈ mat TI SO(3).
and H
Note that R = exp(W) R ∈ mat TR SO(3). We point out that the skew-symmetric tensors W and
=
HR do not belong to the same tangential space, as it can be verified numerically, exp(W) exp(H)

exp(W + H). The associated rotation vector W for the skew-symmetric tensor W is called the total
material rotation vector whose base point is the identity. Taking the derivatives of expression (7)
with respect to the parameter at = 0, we get, see, e.g. [16]
HR =T · W
sin 1 − cos − sin
T= I− W+ W⊗W (8)
2 3
=W, R = exp(W), lim T(W) = I
W→0

where the material tangential transformation T = T(W) is a linear mapping between virtual material
tangent spaces mat TI SO(3) → mat TR SO(3). Now, we could make another verification that the virtual
rotation vector HR and the virtual total rotation vector W belong to different vector spaces on

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1017

the manifold. This is because the tangential transformation T is equal to the identity only at W = 0.
Note that the transformation T has an effect on the base points, changing the base point I into R.
By examining the tangential transformation T in (8), we found that the transformation is non-
singular when the rotation angle is less than perigon, i.e. <2. It is worth noting that the
transformation tensor T(W), the rotation operator R(W) and the skew-symmetric rotation tensor
(as well as their adjoint operators) have the same eigenvectors. Hence, T(W), R(W), and W
W are
commutative, see [16].

Definition 2.7
For convenience, we define the material vector space on the rotation manifold at any point R as

mat TR := {HR ∈
R ∈ mat TR SO(3)}
E3 | H
where an element of the material vector space is HR ∈ mat TR , which is an affine space with the
rotation vector W as the base point and the incremental rotation vector H as a tangent vector.
Hence, the tangential transformation T is a mapping T:mat TI → mat TR . Note that the elements
of this material vector space can be added by the parallelogram rule only if they occupy the same
affine space, i.e. if their associated skew-symmetric tensors belong to the same tangential space
of the rotation manifold. Definition 2.7 for the material vector space should be considered as a
useful and simple notation with equivalence relation with the material tensor space defined in
Definition 2.5.
Similarly, we could determine the spatial tangential transformation, yielding
hR = TT · w, T = T(w), R = exp(
w), = w (= W) (9)
where T is the same linear operator as in the material form (8).

Definition 2.8
We define the spatial vector space on the rotation manifold at any point R as

spat TR := {hR ∈ E 3 |
hR ∈ spat TR SO(3)} (10)
An element of the spatial vector space is hR ∈ spat TR .

Hence, the spatial tangential transformation TT is a mapping between vector spaces on the
rotation manifold spat TI → spat TR . The spatial and the material vector spaces are related by
the rotation operator as given in Equation (6)c. The identity I maps between the vector spaces
mat TI → spat TI . Now, the relation between the spatial and the material vectors can be given as
(w, h) = (IW, RH) where w and W represent the base vectors in the spatial and material vector
spaces, respectively. This relation can be written more compactly as hR = RHR , called a push-
forward operation, where the rotation operator should be considered as a mapping between the
material and spatial vector spaces of rotation, R:mat TR → spat TR , see Figure 4. A push-forward
operator maps a material vector space into a spatial vector space (one-to-one and onto). It makes
sense since a rotation operator is a two-point tensor.

2.1. Angular velocities, accelerations and curvatures

In this section, we give definitions for material as well as spatial angular velocities, acceleration
and curvatures.

Figure 4. Commutative diagram of variations of material and spatial rotation vectors on the rotation
manifold (on the left), and their corresponding vector spaces (on the right).

Definition 2.9
The material angular velocity (skew-symmetric) tensor is defined with the aid of rotation operator
R ∈ SO(3) and its time derivative by
R := RT Ṙ
X
where the dot represents the time derivative. See justification in [19, Chapters 8.6 and 15.2].

The rotation tensor is a mapping R:mat TR → spat TR between the material and spatial vector
spaces. Then the material angular tensor is a mapping X R :mat TR → mat TR . Thus, the material
angular velocity tensor is a true material tensor. The skew-symmetry can be obtained by taking
derivative for the equation RT R = I.
If the rotation operator is expressed with the aid of exponential mapping by Rnew = R(I + HR +
2
O(HR )), where the fixed rotation R is superimposed by an infinitesimal rotation (I + HR ) plus
R →
higher-order terms and substituting this into Definition 2.9 yields after the limit process H 0
=H
X ˙
⇔ X = Ḣ (11)
R R R R

This states that the angular velocity vector is the time derivative of the incremental rotation
vector HR , moreover (if the base point is omitted) H, Ḣ, X ∈ mat TR , which is the material rotation
vector space on the rotation manifold. The result in (11) is often given as definition for the angular
velocity vector in elementary textbooks.
Similar expression and derivation can be accomplished for the spatial angular velocity tensor
and vector, yielding

xR :=ṘRT
(12)
x =
R h˙ ⇔ x = ḣ
R R R

where the spatial incremental rotation vector hR , its time derivative vector ḣR and the spatial
angular vector xR belong to the same spatial vector space on the manifold, h, ḣ, x ∈ spat TR ; the
base point R is omitted.

Definition 2.10
The material and spatial angular acceleration tensor and the corresponding vector is defined as the
time derivative of the angular velocity terms, giving

AR :=X˙ , AR ∈ mat TR SO(3), AR = ẊR , AR ∈ mat TR
R
˙ R,
aR :=
x
aR ∈ spat TR SO(3), aR = ẋR , aR ∈ spat TR

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1019

where AR is the material angular acceleration vector and aR is the spatial angular acceleration
vector at the base point R.

Note that the incremental material rotation vector HR , the material angular velocity vector
XR and the material angular acceleration vector AR (majuscule of alpha-letter) belong to the
same material vector space on the rotation manifold, i.e. HR , XR , AR ∈ mat TR with the base point
R = exp(WI ). At separate moments, these vectors, however, occupy different vector spaces because
the rotation operator depends on time, namely R = R(t). The base point is moving as time travels.
Vector quantities of this kind may be called spin vectors. Spin vectors are rather tricky in numerical
sense as they always occupy a distinct vector space on a manifold. Correspondingly, the spatial
spin vectors are hR , xR , aR ∈ spat TR .
Angular velocity vectors and the time derivative of total rotation vector are related by, see (8)
and (9)

XR =T(WI ) · ẆI where XR ∈ mat TR , ẆI , WI ∈ mat TI for material description

(13)
xR =TT (wI ) · ẇI where xR ∈ spat TR , ẇI , wI ∈ spat TI for spatial description

where the tangential transformation depends on the total rotation vector, and the rotation operator
is R = exp(W I ) = exp(
wI ). Similar expression for the angular acceleration vector can be obtained
by differentiating the above formulas, giving

AR =T · ẄI + Ṫ · ẆI where AR ∈ mat TR , WI , ẆI , ẄI ∈ mat TI for material description
(14)
aR =TT · ẅI + ṪT · ẇI where aR ∈ spat TR , wI , ẇI , ẅI ∈ spat TI for spatial desrcripton

Note that the tangential transformations T, Ṫ ∈ L(mat TI , mat TR ) and TT , ṪT ∈ L(spat TI , spat TR )
operate with different base points.

Definition 2.11
R and
The material and spatial curvature tensors K jR of a parametrized line (beam) are defined
as

R := RT dR = RT R ,
K jR := R RT

ds
where the prime denotes the derivative with respect to the length parameter s. The axial vectors
KR and jR are called the material and spatial curvature vector, respectively.

Material and spatial curvature tensors and corresponding curvature vectors are related by

jR = RKR RT , jR = RKR (15)

The relation between curvature vectors and total rotation vectors comes from the analogy of angular
vector, see (13)

KR =T(WI ) · WI where KR ∈ mat TR , WI ∈ mat TI , for material description

(16)
jR =TT (wI ) · wI where jR ∈ spat TR , wI ∈ spat TI , for spatial description

3. GEOMETRICALLY EXACT BEAM THEORY

In this section, we apply the principle of virtual work for deriving the resulting equations of
Reissner’s beam theory. The same resulting equations of motion may be derived by very different
principles and approaches. One of the eldest approaches is Newton’s second law of motion that
is a fundamental principle. In Newtonian mechanics, forces are of two kinds, internal forces
and external forces with respect to a corresponding mechanical system. Moreover, Newtonian
mechanics is rather involved in vectorial representation that is often named as vectorial mechanics.
As we have shown in the previous section, vectors, tensor, and corresponding spaces are profound
objects of differential geometry. However, virtual displacements and rotations are not included in
Newtonian mechanics rather in Lagrangian mechanics.

3.1. Virtual work forms

The virtual work may be decomposed into three terms: external, internal and inertial virtual
works as

W = Wext −Wint + Winert (17)

where the subscripts ‘ext’, ‘int’ and ‘inert’ correspond to ‘external’, ‘internal’ and ‘inertial’,
respectively. In the virtual work of internal forces, the minus sign indicates that internal forces
work against the virtual displacements. In addition, the inertial virtual work Winert includes a
minus sign in its form. It is convenient to avoid additional minus signs by introducing the virtual
work of acceleration forces by the formula Wacc := −Winert .
Next, we give the form of virtual work and we show that the principle of virtual work satisfies
the equations of motions. Let a virtual work be stated as

W = x · b dV + x · T dA − R F:P dV − x · ẍ0 dV = 0 (18)
B0 *B0 B0 B0

where b, 0 , T are the body force vector, the density of the material body B0 , and the given
traction vector on the boundary of the material body. The first two terms on the right-hand side
of Equation (18) correspond to the external virtual work Wext , the third term the internal virtual
work Wint and the last term the accelerational virtual work Wacc . Deformation gradient is defined
by F := *x/*X (= 0 ∇x). The Lie derivative of deformation gradient R F is equal to F− hF. The
first Piola–Kirchhoff stress tensor is defined by P = Ti ⊗ Ei , where the stress vector Ti acts on
the faces of the deformed element of volume. Note that the term of virtual work of internal forces
R F:P can be simplified into

R F:P = F:P − (

hF):P

= (0 ∇x):P − 12 ((
hF):P − (
T
h F):P)

= 0 ∇ · (PT x)−x · (0 ∇ · P) − 12
h:(PFT − FPT ) (19)

h = −
where we have used the skew-symmetry of
T
h and relations with gradient and divergence
operators.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1021

By substituting (19) into the virtual work we get after the use of the divergence theorem

W = ( 12 (PFT −FPT ):
h+x · (0 ∇ · P+b−0 ẍ)) dV + x · (T −P · N) dA = 0 (20)
B0 *B0

Now examining the above equation, we found that the principle of virtual work (18) satisfies all
the equations of motion with a kinematically admissible virtual displacement, see [14, 21]. The
kinematically admissible virtual displacement field x fulfils the essential boundary conditions.
We note that the term R F:P induces the balance equation of moment of momentum, but the term
F:P does not. Moreover, the virtual rotation tensor field
h depends on the virtual displacement
field x in continuum, but if the beam kinematic assumption is utilized, the field
h is an unknown
variable itself.

3.1.1. Beam kinematics. Let {Oe , e1 , e2 , e3 } be a spatially fixed Cartesian frame and let {OE , E1 ,
E2 , E3 } be a material Cartesian frame, where the base vector E1 coincides with the tangent of
the centerline of the beam. We will assume that the plane perpendicular to the central line of the
beam in an undeformed placement remains undeformed in a deformed state, i.e. the Timoshenko–
Reissner beam hypothesis is applied. We note that transversal shears are allowed but no warping
effects in this beam hypothesis.
Let {x1 , x2 , x3 } be co-ordinates of a spatially fixed frame and let {X 1 , X 2 , X 3 } be co-ordinates
of a material frame then the material point X in the deformed placement field can be given with
the aid of the spatial (floating) frame {Ot , t1 , t2 , t3 } as
x(t, s) = xc (t, s) + X 2 t2 (t, s) + X 3 t3 (t, s), (s, X 2 , X 3 ) ∈ [0, L] × A (21)
where s is the beam length parameter and xc determines the central line of cross-section. The
spatial frame {Ot , t1 , t2 , t3 } coincides with the material frame at an initial placement, i.e. ti = Ei ,
i =1, 2, 3. Due to shear affects, the tangent of central line dxc /ds does not necessarily coincide
with the cross-section normal vector t1 .
Definition 3.1
The material placement of a Timoshenko–Reissner beam at time moment t = 0 can be written with
the aid of the initial rotation operator R0 (s) as
B0 := {x ∈ E 3 | xc (s) + X 2 R0 (s)e2 + X 3 R0 (s)e3 ; (s, X 2 , X 3 ) ∈ [0, L] × A}
where L , A are the beam length and its cross-section area, respectively. The initial placement B0
is a constraint manifold with a vector bundle at fixed time moment t = 0. The global base vectors
and the material base vectors are related by Ei (s) = R0 (s)ei , i = 1, 2, 3. Its tangent space is
TX B0 := {x ∈ E 3 | x1 E1 + x2 E2 + x3 E3 , x1 , x2 , x3 ∈ R}
Definition 3.2
Correspondingly, the spatial placement of the beam may be defined as
B := {x ∈ E 3 | xc (t, s) + X 2 R(t, s)E2 (s) + X 3 R(t, s)E3 (s); (s, X 2 , X 3 ) ∈ [0, L] × A}
where the rotation operator R depends on the rotation vector W ∈ mat TR SO(3). The spatial and
material base vectors are related by ti = REi . Its tangent space is given by
Tx B := {x ∈ E 3 | xc + RE, x ∈ B, kinematically admissible variation}

R for the material rep-

where we have denoted E := X 2 E2 + X 3 E3 ∈ TX B0 . Note that R = RH
resentation and R =
hR R for the spatial representation.

3.1.2. Virtual work of Reissner’s beam. In this section, we derive the virtual work for Reissner’s
beam when the kinematic assumptions are applied. We could substitute the virtual displacement
field x, Definition 3.2, and the acceleration field ẍ of (21) into the virtual work of acceleration
forces Wacc , which is the last term in (18), giving [22]

Wacc = · (ẍc + RX
(xc + RHE) XE
+ R AE)0 dV
B0

= xc · (A0 ẍc ) ds + · (JA +
J) ds (22)
L L

where we have used the central line condition

E dA = (X 2 E3 ) dA =
E2 + X 3 0 (23)
A A

Note that JA, XJX ∈ mat TR (= TX B0 ) are elements of the material vector space, see Definition 2.7.
Moreover, we have denoted the inertial tensor J ∈ TX B0 ⊗ TX B0 as
J=Ji j Ei ⊗ E j
⎡ 2 ⎤
X 2 + X 32 0 0
⎢ ⎥ (24)
[Ji j ]= ⎢ ⎣ 0 X 32 −X 2 X 3 ⎥
⎦ 0 dA
A
0 −X 2 X 3 X 22

If the principal axes of the inertial tensor J are parallel to the base vectors {E2 , E3 }, then the
matrix of the inertial tensor [Ji j ] is diagonal. In the following, we will use this assumption for
simplicity, having for the matrix [Ji j ] = diag(J1 , J2 , J3 ), J1 = J2 + J3 .
As for the virtual work of acceleration forces, we could write the virtual work of external forces,
first two terms in (18), by substituting the virtual displacement of the beam x, Definition 3.2,
giving

Wext = xc · n ds + HR · MR ds (25)
L L

where the external force vector n and the external material moment vector MR are denoted by

n:= b(X) dA + T (X) dr
A *A
(26)
MR := ERT b(X) dA + ERT T (X) dr
A *A

where b and T are the external body force vector and the external stress vector. The external
moment vector MR can be viewed as an element of the material vector space of rotation mat TR ,

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1023

see Definition 2.7. This is because the external moment vector MR is the work conjugate of the
virtual incremental rotation HR .
Next, we derive the virtual work of internal forces Wint , the third term in (18), using the
beam assumptions. The deformation gradient F states, after substituting the beam kinematics
assumptions (21)
R E − t 1 ) ⊗ E1 + R
F = (xc (s) + RK (27)
R := RT R , Definition 2.11. Here we can use the definition for
where the material curvature tensor K
Lie variation R · := R(R ·), which generates an objective variation, see more common definitions
T

for Lie derivative in [22, 23]. The Lie variation of the deformation gradient is, correspondingly
T x + RK
R F = (xc − RHR R E) ⊗ E1 ∈ Tx B ⊗ TX B0 (28)
c

Now the term R F:P (the virtual work of internal forces) in (18) can be written with the aid of
(28), yielding

T T
Wint = (R xc −HR xc ) · N ds + KR · MR ds (29)
L L
where L corresponds to the initial length of the beam. The material internal force vector N and
the material internal moment vector MR are denoted by the formulas

N := RT T1 dA ∈ TX B0 , MR :=
ERT T1 dA ∈ TX B0 (30)
A A
where T1 is the stress vector acting on the cross-section face. The work conjugate of the material
vector N in (29) is the variation of the material strain vector C, defined by the formula and its
variation
C:=RT xc − E1 ∈ TX B0
(31)
T x ∈ TX B0
C = RT xc −HR c

The components of the material strain vector C(s) = i Ei , at any point of the beam s, are shown
in Figure 5 in plane case.
Now, we can give the internal virtual work in its material representation in addition to its spatial
form

Wint = (C · N + KR · MR ) ds = (R c · n + R jR · mR ) ds (32)
L L

where R corresponds to the Lie variation with respect to the rotation operator R. Material and
spatial vectors are related by a push-forward, the rotation operator R ∈ L(TX B0 , Tx B), and are
defined by
c:=RC ∈ Tx B, jR := RKR ∈ Tx B = spat TR
(33)
n:=RN ∈ Tx B, mR := RMR ∈ Tx B = spat TR
The components of the spatial strain vector c(s) = i ei , at any point of the beam s, are shown in
Figure 5 in the plane case. The material and spatial strain components are connected by relation
i ti = i ei .

E e
1

1 E1 , e1

Figure 5. Material strain components and spatial strain components, plane case.

At this point, we should realize that we are searching for the solution of the beam central
line displacement field xc (s) and its rotation field R(s) that corresponds to the resultant form of
the internal virtual work (32). This placement manifold of Reissner’s beam may be identified by
(xc (s), R(s)) ∈ E 3 × SO(3) at any point of the beam. A stress tensor, which gives equal internal
force and moment vectors, is a valid stress tensor. Hence, the equations of motion for continuum
are realized only by their resultant form in one dimension. This also yields the constitutive relation
in terms of the resultant quantities.

3.2. Constitutive relations

We introduce the constitutive relations of Reissner’s beam, and we show how these relations are
connected to continuum mechanics and the strain energy function.
Let Wstr be a strain energy function per unit volume of a hyperelastic material. The first Piola–
Kirchhoff stress tensor P may also be defined from the strain energy function Wstr (F) by

*Wstr (F)
P := ∈ Tx B ⊗ TX B0 (34)
*F
We also assume that the strain energy function is frame-indifferent under an orthogonal transfor-
mation F+ = QF by obeying the identity Wstr (F+ ) = Wstr (F).
This means that the strain energy function Wstr is invariant under rigid body rotation. If we set
Q = RT , then we have via the pull-back operator the strain energy function in the material domain,
and it makes it possible to define a different stress quantity.
The Lie variation of the strain energy function Wstr (F) with respect to the rotation operator R
can be written by using the above relation as

*Wstr (F)
R Wstr = :R F = P:R F ∈ R (35)
*F

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1025

that is equal to the virtual work of internal forces Wint , see (18). We get the same result for
Wstr (RT F). The first Piola–Kirchhoff stress tensor P is a two-point tensor defined on material
and spatial placement manifolds, in our case these manifolds are TX B and Tx B0 . We use the Lie
variation, push-forward and pull-back relations to derive a one-point stress tensor that corresponds
(35), yielding
R Wstr = P:(R(RT F)) = (RT P):(RT F − I) (36)
The Lie variation of internal energy function (36) introduces new material stress and strain tensors
defined by
R:=RT P ∈ TX B0 ⊗ TX B0
(37)
H:=RT F − I ∈ TX B0 ⊗ TX B0
The material stress tensor R = i j Ei ⊗ E j , as well as, its work conjugate strain tensor
H =Hi j Ei ⊗ E j are both unsymmetrical tensors and not named in continuum mechanics. The
material strain tensor H for Reissner’s beam can be written after substituting (27) and (31), giving
R E) ⊗ E1
H = (C + K (38)
We note that the rotation operator R is not, in general, equal to the rotation tensor of the
polarization decomposition. However if the shear components 2 and 3 , as well as, the tor-
sion curvature K 1 vanish, then RT F is equal to the right (material) stretch tensor U, which is
a symmetric tensor. Thus, in this particular case (K 1 = 2 = 3 = 0), the material strain tensor
H may be identified with the right-stretch strain tensor, and the symmetric part of the stress tensor
R may be identified with the Biot stress tensor, see [21].
Let us consider a following linear constitutive relation between the stress components of the
tensor R and the strain components of the tensor H given by R = C:H, where the material elasticity
tensor C ∈ TX B0 ⊗ TX B0 ⊗ TX B0 ⊗ TX B0 is a fourth-order material tensor. At a cross-section of
Reissner’s beam, we introduce simple constitutive relations in component forms
11 = E H11 , 21 = G H21 , 31 = G H31 (39)
where E denotes Young’s modulus and G the shear modulus. The constitutive relations (39)
R E reads
are similar to engineering strains and stresses. We note that the vector Hi1 Ei = C + K
according to (38), and thus we may express the material strain vector i1 Ei as
R E)
i1 Ei = (EE1 ⊗ E1 + GE2 ⊗ E2 + GE3 ⊗ E3 ) · (C + K (40)
If we compare the equation for the first Piola–Kirchhoff stress tensor P = Ti ⊗ Ei and the material
stress tensor R (37)a, we get the stress vector at a cross-section
i1 Ei = RT T1 (41)
Now we could substitute the above equation into the formula of internal force vector (30)a that
yields after integrating over a cross-section area and using constitutive relation (40)

N= i1 Ei dA = (E AE1 ⊗ E1 + G AE2 ⊗ E2 + G AE3 ⊗ E3 ) · C (42)
A
where we have used the beam kinematics that satisfies the central line conditions, see details
in [22].

Similarly we may derive for the internal moment vector (30)b, yielding [22].
MR = (G I11 E1 ⊗ E1 + E I E ⊗ E ) · KR , , ∈ {2, 3} (43)

The component matrix of the moment tensor of inertia of the cross-section [Ii j ] can be
computed by
⎡ 2 ⎤
X 2 + X 32 0 0
⎢ ⎥
⎢ X 32 −X 2 X 3 ⎥
[Ii j ] = ⎢ 0 ⎥ dA (44)
A⎣ ⎦
0 −X 2 X 3 X 22

If the principal axes of the inertia moment tensor are parallel to the base vector {E2 , E3 }, then the
matrix [Ii j ] is diagonal. In the following, we will use this assumption for simplicity.
We introduce linear constitutive relations in the material representation between the material
strain and internal force vectors and the material curvature and moment vectors by writing

N = Cn C
MR = Cm KR (45)

where the constitutive tensors in the material frame are

Cn =Cinj Ei ⊗ E j , Cinj = diag(E A G A2 G A3 )
(46)
Cm =Cimj Ei ⊗ E j , Cimj = diag(G J E I2 E I3 )

Here, E A is the axial stiffness, G A2 and G A3 are the shear stiffnesses with respect to the cross-
section in principal directions, G J is the torsion stiffness and E I2 and E I3 are the principal
bending stiffnesses. Coefficients, which depend on the shear modulus G, may be adjusted such
that they include some secondary warping effects. The cross-section areas A2 and A3 include shear
coefficient factors, and the torsional square moment J is the square moment of the Saint-Venant
torsion theory, instead of the polar moment IP = I2 + I3 .

3.3. Lagrangian formulations

Next, we consider a total Lagrangian updating formulation for a material rotation vector and
compare it with an updated Lagrangian formulation. We also discuss some difficulties about an
Eulerian formulation. We exploit only the material representation but the spatial version may be
considered similarly.
Updating formulations of the material rotation vector can be divided into three classes [14]

(1) The total Lagrangian formulation where the total rotation vector W ∈ mat TI is the dependent
variable.
(2) The updated Lagrangian formulation where the updated rotation vector WIref ∈ mat TIref is the
dependent variable.
(3) The Eulerian formulation where the incremental rotation vector HR ∈ mat TR is the dependent
variable.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1027

In the total Lagrangian formulation, we have updating procedure similar to displacement vector

W(i+1) = W(i) + W(i) ∈ mat TI (47)

where the rotation vector change W(i) always occupies the vector space mat TI since the base point
I ∈ SO(3) remains fixed. The superscripts in brackets correspond to iterative steps.
We may consider the total Lagrangian updating procedure as a linear space updating procedure.
This formulation preserves the path-independent property (in static case) and it can be regarded
as a consistent updating formulation. The major drawback is singularities at the rotation angle
= 2 and its multiples but we overcome this difficulty by the complement rotation vector with
the change of parametrization.
The updated Lagrangian formulation is closely related to the total Lagrangian but the base point
Rref ∈ SO(3), the reference placement, is updated occasionally. Usually, we update the base point
at every incremental step in static analysis, and at every time step in transient dynamic analysis.
The updated Lagrangian procedure reads
(i+1) (i)
Wref =Wref + DW(i) ∈ mat TIref at every iteration
(48)
(n) (0)
ref =Rref exp(Wref ),
Rnew Wref = 0
old
when updating the base point
where the second formula corresponds to updating the base point, which we execute occasionally.
The updated Lagrangian formulation does not preserve the path-independent property (in static
(n)
case). This is because the new base point Rnew ref depends on the previously computed solution Wref
and then the error of the updated rotation vector W(n) ref causes a cumulative error to the base point
Rnew
ref . Here we have used the notation I ref in the subindex of the vector space instead of Rref since
Iref indicates a new reference placement, and it can be identified with the identity.
The Eulerian updating formulation for material rotation is somehow a more complicated pro-
cedure although it may be written with rather a simple formula
(i)
) ∈ SO(3) at every iteration
R(i+1) = R(i) exp(H R (49)

The above updating formula is a non-linear procedure and uses the property of Lie group and its
Lie algebra. In static analysis, we could employ the Eulerian formulation where there exists no
angular velocity and angular acceleration vectors, and therefore no time-integration procedure. The
algebra with rotation vectors, angular velocity and acceleration vectors or corresponding skew-
symmetric tensors are more involved, and may be properly computed by the Lie group methods
of differential equations, see e.g. [24].
The Lagrangian formulations (total and updated) are geometrical integration methods where the
constraint manifold is parametrized. The total rotation vector, likewise the updated rotation vector,
forms a local parametrization for the rotation manifold SO(3) via the exponential mapping, but the
incremental rotation vector H(i) (i)
R ∈ mat TR at each iteration step (i) belongs to a different vector
space. We called quantities like incremental rotation vectors as spin vectors. Spin vectors have a
vector character at a point but not in local sense, contrary to total and updated rotation vectors
that are vectors in local sense, i.e. a local parametrization of the manifold.
We note that the direct application of the incremental rotation vector HR with standard time-
integration methods yields serious trouble where we are adding quantities, which belong to different
tangential spaces. It is well known that the spatial form of Newmark time-integration scheme is

improper for time-integration purposes but the material form of (Newmark) time-integration has
the same problems, see [13]. This produces a new concept of consistency, which is defined in the
following.

Definition 3.3
Manifold consistency is defined as property where we distinguish all the quantities on manifolds
in their tangential spaces, and we execute algebraic calculations if they are properly suited from
manifold point of view.

For example, Newmark integration methods given in [12] can be considered as a manifold-
inconsistent time-integration method. The problem arises from knowledge that the incremental
material tensor H R occupies the tensor space mat TR SO(3) and not the tensor space mat TI SO(3) as
it is explicitly assumed in that paper. However, the Newmark integration method may give accurate
results when using a small time step size.
An integration method with manifold consistency is given in the paper [14] where the authors
utilize the Newmark integration scheme although any standard time-integration schemes can be
exploited. The time-integration method is based on the updated Lagrangian formulation giving a
manifold-consistent vector addition to the dependent variable, as in (48)a. Only a few researches
have followed with this manifold consistent approach. This is mainly because the exact tangential
tensor is rather complicated.
Next, we will derive the virtual work form in terms of the total material rotation vector W and its
virtual displacement W in order to give a total Lagrangian formulation. We will choose the material
description since the linearization procedure is somehow simpler than in spatial description, no
emergence of Lie derivative. We need to express spin vector like the virtual incremental rotation
vector HR , the angular velocity vector XR , the angular acceleration vector AR and the curvature
vector KR in terms of the total rotation vector W and W, giving
HR =TW, XR = TẆ
(50)
AR =TẄ + ṪẆ, KR = TW

where the tangential transformation is given in (8). We note that W, W, Ẇ, Ẅ ∈ mat TI whereas the
spin vectors HR , XR , AR ∈ mat TR .
Now we could write the virtual work of acceleration forces Wacc , (22), in terms of W, W, Ẇ, Ẅ

Wacc = xc · (A0 ẍc ) ds + W · (TT JTẄ + TT JṪẆ + TT (TẆ)∼ JTẆ) ds (51)
L L

that can be decomposed into the acceleration-dependent part WaccA and the velocity-dependent
part WaccB

WaccA = xc · (A0 ẍc ) ds + W · (TT JTẄ) ds
L L
(52)
WaccB = W · (TT JṪẆ + TT (TẆ)∼ JTẆ) ds
L

Note that the rotational part of the virtual work of acceleration forces Wacc does not depend
on the central line displacement vector xc and its derivatives, and the translational part does not

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1029

depend on the rotation vector W and its derivatives. This is due to the kinematic assumption of
Reissner’s beam (21), where the displacement and rotation fields are independent variables.
The virtual work of external forces reads in mat TI -formulation

Wext = xc · n ds + W · TT MR ds (53)
L L

where the external material moment vector MR ∈ mat TR , so it is a spin vector. According to the
above formula, we may define a new external material moment vector by setting

MI := TT MR ∈ mat TI (54)

where mat TI is the initial vector space of rotation in the material description.
Finally, the virtual work of internal forces, (32), can be written in the total Lagrangian
formulation

Wint = C · N ds + KR · MR ds (55)
L L

where the variation of material strain and curvature vectors is according to (31)b, (50)d and (A1)

C = RT xc + (RT xc )∼ TW

and
KR = T · W + TW

= C1 (W , W) · W + TW (56)

Substituting the above equation into the form (55) yields

Wint = (xc · RN) ds + (W · (−TT (RT xc )∼ N + CT1 (W , W)MR ) + W · TT MR ) ds (57)
L L

where RN can be viewed as a push-forward of the material force vector N ∈ TX B0 . It is clear

that the virtual work of internal forces is more coupled than the other virtual work forms. This is
mainly because of the choice of dependent variable, the rotation vector.

3.3.1. On objectivity for strain vector and curvature tensor. In this section, we show that the
strain and curvature vectors in the material and spatial descriptions are objective quantities. This
objectivity may be called observer frame-indifference, see [21, Chapter 2].
The material strain vector field and the curvature tensor field for Reissner’s beam have been de-
fined in (31)a and in Definition 2.11. Let the observer transformation to the rotation
operator R and the central line placement xc be

R+ = QR, x+
c = Q(xc + c) (58)

where the orthogonal operator Q ∈ SO(3) corresponds to the rigid body rotation and the vector
c ∈ E 3 corresponds to the rigid body translation. Substituting the observer transformations (58)

into the material strain vector and the curvature tensor fields, we obtain the transformed vectors
C+ (s) = (QR)T (Q(xc + c)) − E1 = RT xc − E1 (= C(s))

K R (s)) ∀s ∈ [0, L]
+ (s) = (QR)T (QR) = RT R (= K (59)
R

where we have used the properties of observer transformation Q (s) = O, c (s) = 0 and QQT = I.
By comparing Equations (59), we found that the material strain vector and the material curvature
tensor are both frame-indifferent. This yields also a conclusion that the internal virtual work (55)
with the constitutive relation (45) is frame-indifferent under the observer transformation. Hence,
the internal virtual work is an objective quantity.
We observe that objectivity is also achieved if any interpolation is applied for the material
strain vector and curvature tensor with the aid of the total rotation vector W ∈ mat TI , or via
the updated rotation vector Wref ∈ mat TIref . Therefore, the total Lagrangian formulation, where

the rotation operator R = exp(W(s)), as well as the updated Lagrangian formulation, where the
(i)
rotation operator R = Rref exp(Wref (s)), are both frame-indifferent formulations. We note that an
interpolation for the incremental rotation vector HR (s) ∈ mat TR may not be directly applied, since
at every point the incremental rotation vector occupies a different vector space of rotation mat TR .
That is because the base point R = R(s) depends on the length parameter s.
Correspondingly, the spatial strain vector field c and the spatial curvature tensor field
jR have
been defined in (33)a and in Definition 2.11. By substituting the observer transformations (58)
into the above spatial quantities, we get
c+ (s) = Q(xc − RE1 ) = Qc(s)
(60)
j+
T T
R (s) = QR R Q = Q
jR (s)QT ∀s ∈ [0, L]
Hence, the spatial strain field c and the curvature field
jR are spatial objective quantities. Note that
the spatial curvature tensor field
jR transforms like a second-order tensor in the spatial description.
Its associated vector field jR varies in the observer transformation like a spatial vector, giving
j+
R (s) = QjR (s) ∀s ∈ [0, L] (61)
We note that spatial objectivity is also achieved if any interpolation is applied for spatial strain
vector and curvature tensor with the aid of a total rotation vector w(s) ∈ spat TI , or via an updated
rotation vector wref ∈ spat TIref .
Let W1 , W2 ∈ mat TI be nodal vectors of total material rotation for element, which has a
linear interpolation, and let WQ be a total rotation vector for an objective transformation operator
Q = exp(W Q ), i.e. for rigid body rotation. These rotation vectors have the following component
values [25] with respect to the global frame {O, e1 , e2 , e3 }
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 −0.4 0.2
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
[1i ] = ⎣−0.5⎦ , [2i ] = ⎣ 0.7 ⎦ , [Qi ] = ⎣ 1.2 ⎦ (62)
0.25 0.1 −0.5
hence, e.g. W1 = 1i ei with conventional summation. Linear interpolation functions for the nodal
rotation vectors W1 and W2 read
s s
N1 (s) = 1 − , N2 (s) = , s ∈ [0, L] (63)
L L

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1031

Interpolated values for the total rotation vector are therefore W(s) = Ni (s)Wi ∈ mat TI . This interpo-
lation is clearly acceptable because the nodal rotation vectors belong to the same tangential vector
space of rotation. The observer transformation to the rotation operator R is given in (58)a, yielding
+
1 ) = Q exp(W
1 ), +
exp(W exp(W2 ) = Q exp(W2 ) (64)
The transformed nodal rotation vector Wi+ ∈ mat TI can be obtained by extracting it from the
rotation operator via Spurrier’s algorithm. We note that the original rotation vectors W1 and W2
in the observer transformation (64) occupy the tangential vector space mat TQ . Although we may
extract the transformed nodal rotation vectors Wi+ such that they satisfy the observer transformation
relations (64), the linear interpolation is not preserved, yielding
+
i ) = Q exp(Ni (s)W
i ) ∀s ∈ [0, L],
exp(Ni (s)W Q ∈ SO(3) (65)
This arises from the fact that the rotation manifold SO(3) has a curved character. Indeed, a linear
vector valued function in the tangential vector space mat TQ is not linear in the different tangential
vector space mat TI . Results in Figure 6 show that the linear interpolation does not preserve the
observer transformation, settings after (62). Note that we never use Equation (65) in our beam
formulation, but in paper [25] this equation is frequently exploited yielding incorrect results about
the objectivity of geometrically exact formulations. Hence, the rotation interpolation has rather an
extraordinary property, which arises from the curvature feature of the rotation manifold SO(3).
Therefore, we may pronounce that extracting the nodal rotation vectors from the corresponding
rotation operations may not be interpolated. There are infinite possibilities for the interpolation
since the transformation operator Q ∈ SO(3) is arbitrary.
Despite Equation (65) states that the interpolation is not preserved in the observer transformation,
this does not mean that the beam formulation is non-objective. Indeed, this property is never
required for being an objective formulation and we never interpolate the extracted rotation vectors.
It is sufficient that the beam formulation satisfies conditions (59). Generally speaking, a global

2 2

1.9
Components of rotation vector

1.5
Norm of rotation vector

1.8
1 1.7

0.5 1.6

1.5
0
1.4
−0.5
1.3

−1 1.2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Length parameter s/L Length parameter s/L

Figure 6. The components of the rotation vectors W(s) and W+ (s) interpolated through the length of beam
(left), and the norm W(s) and W+ (s) (right). The solid line indicates the initial rotation vector W(s)
and the broken line the transformed rotation vector W+ (s). Solid lines correspond to the proper values.

interpolation on a non-flat manifold never preserves in an observer transformation. This is because

of the non-linear character of the manifold; a parametrization mapping is a non-linear mapping
for a non-flat manifold.
We have assumed above that the observer transformed rotation interpolation W+ ∈ C(mat TI )
keeps the base point I fixed. However, it also makes sense and can be assumed that the base point
transforms under the observer transformation (I → Q) giving W+ ∈ C(mat TQ ). Then we could
denote W+ (s) = W(s) and an interpolation is preserved under an observer transformation. This is
an important issue and clarifies the frame-indifference of geometrically exact beam formulations.
Next, we show why the updated Lagrangian formulation has the path-dependent property. Let
(n)
ei be the interpolation error of the exact rotation vector Wref , and let Rerr,i be the error of the
exact rotation operator Rref at the ith incremental step. The error of the rotation operators at the
first incremental step and at the following steps up to the incremental step r , reads as
(n)
ref,1 +
Rref,1 + Rerr,1 =I exp(W e1 )
(n)
Rref,2 + Rerr,2 =(Rref,1 + Rerr,1 ) exp(W ref,2 +
e2 )
(66)
..
.
(n)
ref,r +
Rref,r + Rerr,r =(Rref,r −1 + Rerr,r −1 ) exp(W er )

Thus, in the updated Lagrangian formulation the error of the rotational operator is cumulative yield-
ing the statement that the updated Lagrangian formulation does not preserve the path-independent
property of the mechanical model. Although the mechanical model may be path-independent only
in some static cases, it is a desired quality also for dynamic cases where one source of error is
totally absent. Computational models that preserve the path-independent property may be called as
consistent in model type. We note that the total Lagrangian formulation has this kind of property
since the base point remains fixed. Hence, the rotation vector belongs to the fixed vector space of
rotation, i.e. W(i) (s) ∈ mat TI ∀i.

3.4. Consistent tangential tensors of Reissner’s beam

In this section, we will derive consistent tangential tensors for geometrically exact Reissner’s
beam element. We will not introduce a finite element approximation, giving infinite-dimensional
tangential tensors. Tangential tensors arise from the linearization of the virtual work forms. We
call the tangent of internal force vector the stiffness tensor, the tangent of external force vector the
loading tensor, and the tangents of acceleration force vector the mass, gyroscopic, and centrifugal
tensors.
The total and updated Lagrangian formulations yield identical tangential tensors. However, the
updated Lagrangian formulation requires secondary storage variables like the curvature vector
and the rotation operator at the previously converged solution, see [14]. In the total Lagrangian
formulation, secondary storage variables are completely avoided since the reference placement is
always the initial placement. We note that consistent gyroscopic, centrifugal and loading tensors
for the total and updated Lagrangian formulations have not been presented anywhere by author’s
knowledge. The consistent stiffness tensor of Reissner’s beam with the total Lagrangian formulation
has been firstly introduced in [16]. This formulation suffers singularity at the rotation angle = 2
and its multiples, and is therefore restricted with the rotation angle <2.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1033

The linearization of the virtual work W (q, q̇, q̈; q) around the state point (q0 , q̇0 ) ∈ T C at
the time moment t = t0 can be written as
Lin(W (q, q̇, q̈; q)) = W0 (q0 , q̇0 , q̈; q) + Dq W · q + Dq̇ W · q̇ (67)
where Dq denotes Fréchet partial derivative with respect to q that denotes the generalized displace-
ment field including the translational displacement and rotational quantities, i.e. q := (d, W) ∈ C.
We have denoted the change of the base point by (q, q̇). The virtual work of acceleration forces
WaccA (52)a can be written via the bilinear form
⎛ ⎞
A0 I O
WaccA (q0 , q̈; q) = M(W0 ):(q ⊗ q̈) ds, M(W) := ⎝ ⎠ (68)
L O TT JT

where M(W0 ) is the generalized mass tensor. Now the generalized displacement field C, at any
point of it, can be viewed as the vector space E 3 × mat TI that is a flat manifold because the base
point I ∈ SO(3) remains fixed. Hence, in this case, the tangent bundle T C is a flat manifold.
This flatness arises for natural reasons: we have parametrized the rotational manifold SO(3) and
this parametrization chart occupies the three-dimensional Euclidean space. In the cases where the
tangent bundle T C is not a flat manifold, the generalized virtual displacement field q ∈ Tq C
depends on the base point q ∈ C. Thus, this dependency has to be taken into account when
linearizing the virtual work forms.
Next, we give the linearization for the virtual work forms Wacc , Wext and Wint , separately.
The virtual work of acceleration forces WaccA is already in a linear form with respect to the
generalized acceleration field q̈. The generalized mass tensor given in (68), however, depends
on the generalized displacement field q. In addition, the form WaccB does not depend on the
translational displacement field d, which simplifies the linearization procedure considerably. We
denote the rotation field by C(mat TI ) and the translational displacement field by C(E 3 ), hence the
displacement manifold C is equal to C(E 3 × mat TI ).
The linearization of the virtual work of acceleration forces Wacc (22), at the state point
(q0 , q̇0 ) ∈ T C yields
Lin(Wacc (q, q̇, q̈; q)) = Wacc (q0 , q̇0 , q̈; q) + DW Wacc · W + DẆ Wacc · Ẇ (69)
where the linear virtual forms DW Wacc · W and DẆ Wacc · Ẇ may be given by

DW Wacc · W= Kcent :(W ⊗ W) ds
L
(70)
DẆ Wacc · Ẇ= Cgyro :(W ⊗ Ẇ) ds
L
and the centrifugal and gyroscopic tensors Kcent and Cgyro are denoted as
0 JX0 + JA0 , W0 ) + TT · (XJ
Kcent :=C2 (X − (JX)∼ )C1 (Ẇ0 , W0 )

+ TT J · (C5 (Ẇ0 , W0 ) + C1 (Ẅ0 , W0 )) (71)

0 J − (JX0 )∼ )T + JC4 (Ẇ0 , W0 )]
Cgyro :=TT [(X
Here we have utilized formulas given in Appendix A. The angular velocity field X0 ∈ C(mat TR )
and the angular acceleration vector A0 ∈ C(mat TR ) are computed by (50) via the total material

rotation vector and its time derivatives W0 , Ẇ0 , Ẅ0 ∈ C(mat TI ) at the fixed time moment t = t0 .
The centrifugal and gyroscopic tensors are, in general, non-symmetrical tensors. This non-symmetry
arises from the kinetic energy that depends on the generalized displacement field q as well as the
generalized velocity field q̇.
Correspondingly, the linearization of the virtual work of external forces Wext at the point q0 ∈ C
yields the loading tensor Kload

Dq Wext · q = Kload :(q ⊗ q) ds (72)
L

We have assumed above that the external forces and moments do not depend on the generalized
velocities q̇. External force or moment fields are called conservative if they are obtained via a
generalized displacement-dependent potential functional. On a linear space (i.e. on a flat manifold),
the conservative loading and the symmetry of the loading stiffness tensor are equivalent issues.
This symmetry of the loading stiffness tensor is due to the local parametrization of the rotation
operator, see details in [17, 22].
Next, we show that under gravitation the corresponding loading stiffness tensor is symmetric.
It is clear that the external force field n defined in (26)a is constant in this case, hence we will
only study an external moment vector due to gravitation. The rotation-dependent part of the virtual
work of external forces Wextrot and the external moment field M ∈ C(
I mat TI ) under gravitation reads,
according to (53)

Wext (W; W) =
rot
W · MI ds, MI := T T
ERT b dA (73)
L A

The linearization of (73) at the point 0 ∈ C(mat TI ) gives the loading stiffness tensor Kload

Dq Wext · q = Kload :(q ⊗ q) ds
L

where
Kload :=C2 (MR , W0 ) + TT bcross T

(74)
MR :=
ERT b dA0 , bcross :=
E(RT b)∼ dA0
A0 A0

where the body force vector b := b(vt (E)) ∈ Tx B and the cross-section placement vector
E :=X 2 E2 + X 3 E3 ∈ TX B0 . The tensor C2 is defined in Appendix A. The external moment vector
MR ∈ C(mat TR ) generates the work conjugate pair with the virtual incremental rotation field
HR ∈ C(mat TR ). We note that usually a loading stiffness tensor and the corresponding exter-
nal moment vector vanish due to the central line conditions.
It is a straightforward, but rather lengthy, calculation to show that the loading stiffness tensor
(74) is indeed a symmetric tensor. In general, a tensor
C2 ( a
ab, W) + TTbT (75)
is symmetrical for all a, b ∈ E 3 . We note that taking the linearization of the moment field
MR ∈ C(mat TR ), (26)b and (25), does not produce symmetric loading stiffness tensor, since
C(mat TR ) is a tangent bundle on the non-flat manifold SO(3) a contrast to the tangent bundle

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1035

C(mat TI ), which may be identified with C(E 3 ) via an isomorphism. We may consider the external
moment vector MI (s) ∈ C(mat TI ) as a conservative loading since its tangential tensor is symmetric,
and hence may be integrated into a potential functional, see [22].
Next, we linearize the virtual work form of internal forces Wint given in (57). Finally,

Wint = q̂ · BT Fint ds (76)
L

where the generalized virtual displacement q̂, the kinematic operator B and the generalized
internal force field Fint are defined by
⎛ ⎞
xc T
⎜ ⎟ R O (RT xc )∼ T N
⎜ ⎟
q̂ := ⎝W ⎠ , B := , Fint := (77)
O T C1 (W , W) MR
W

and where the tensor C1 is given in Appendix A. Moreover, the field Fint can be given in terms of
the material strain and curvature vectors with the aid of the linear constitutive relations (45) and
(46)

C Cn O
Fint = Cnm , Cnm := (78)
KR O Cm

Now, we could linearize the internal work form Wint (76) at the point q0 = (d0 , W0 ) ∈ C in the
vector direction q̂ = (xc , W , W), giving

Lin Wint (q; q̂) = Wint (q0 ; q̂) + Dq Wint (q0 , q̂) · q̂ (79)
where the linear virtual form Dq Wint · q is denoted with the aid of the material stiffness tensor
Kmat and the geometric stiffness tensor K

Dq Wint · q = (Kmat + K ):(q̂ ⊗ q̂) ds
L

Kmat := BT0 Cnm B0

⎛ ⎞
O O −RNT
⎜ ⎟
K := ⎜⎝ O O C2 (MR , W0 ) ⎟
⎠
TT NRT xc , W) + TT
NRT CT2 (MR , W0 ) C3 (MR , W0 , W0 ) + C2 ( N(RT xc )∼ T
(80)

where the material stiffness tensor Kmat arises from the linearization of the vector Fint , and the
geometric stiffness tensor K arises from the linearization of the kinematic operator B. The tensors
C2 and C3 are given in Appendix A.
We note that the material stiffness tensor Kmat is clearly symmetric because of the symmetry
of the elasticity tensor Cnm . In addition, the geometric stiffness K is also a symmetric tensor
because the tensor C3 and the tensors in the form of (75) are both symmetrical in the lower-right

corner of the K . This symmetry of the stiffness tensor is due to the local parametrization of the
rotation operator, see details in [17].

3.4.1. Switching beam element. In the total Lagrangian representation, we could avoid singularities
at the rotation angle equal to 2 and its multiples by switching the rotation vector W ∈ mat TI into
its complement rotation vector WC ∈ mat TIC , that reads, see Definition 2.2
2
WC := W − W, := W (81)

when the rotation angle exceeds straight angle, i.e. >. Thus, the complement rotation angle
C = 2 − will be less than the straight angle, i.e. C <, and singularity at 2 is avoided,
see Figure 2. When the rotation vector W ∈ mat TI is switched into its complement rotation vector
WC ∈ mat TIC , we also need the time derivatives of the complement rotation vector, giving
C
Ẇ =B · Ẇ ∈ mat TIC
(82)
C
Ẅ =B · Ẅ + Ḃ · Ẇ ∈ mat TIC

where the symmetric kinematic tensor B := DW WC ∈ L(mat TI , mat TIC ) and its derivatives Ḃ and
B̈ are

2 2
B= 1 − I+ n⊗n

2
Ḃ= [(n · Ẇ)I + (Ẇ ⊗ n + n ⊗ Ẇ) − 3(n · Ẇ)n ⊗ n]
2
(83)
2
B̈= [((n · Ẅ) + (ṅ · Ẇ))I + Ẅ ⊗ n + n ⊗ Ẅ + Ẇ ⊗ ṅ + ṅ ⊗ Ẇ]
2
6 2
− [((n · Ẅ) + (ṅ · Ẇ))n ⊗ n + (n · Ẇ)(ṅ ⊗ n + n ⊗ ṅ)] − (n · Ẇ)Ḃ
2

and where n = W/. The linearized forms of the complement rotation vector WC ∈ mat TIC with
respect to the rotation vector W ∈ mat TI are correspondingly

DW WC · W=B · W
C C
DW Ẇ · W=Ḃ · W, DẆ Ẇ · Ẇ = B · W (84)
C C
DW Ẅ · W=B̈ · W, DẆ Ẅ · Ẇ = 2B · W
Here, the method for resolving singularity problems is similar to that of Cardona and Géradin
[14], however, they have not solved the problem where the rotation vector at one node of the
element has been switched but the rotation vector at another node of the same element remains
unaltered. In large rotational analysis, there often occurs a situation where the rotation vector has
been switched into its complement rotation vector at the one end of the beam element, and the
rotation vector has remained the same at the other end, since the rotation angle is less than the

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1037

Figure 7. A typical situation when switching beam element is needed. The switching beam is constructed
from an ordinary beam element in addition to the parametrization change at node 2.

straight angle. In Figure 7 we give an example of this situation: at node 1, the rotation vector
WC1 ∈ mat TIC has been switched, and at node 2 the rotation vector W2 ∈ mat TI is still less than that
of straight angle, hence it is non-switched. The virtual work at node 2 reads in the spaces mat TIC
and mat TI
C C
2 · (f2 (W2 , Ẇ2 , q, q̇, q̈) − M2 Ẅ2 ) = W2 · (f2 (W2 , Ẇ2 , q, q̇, q̈) − M2 Ẅ2 )
WC C C C
(85)

2 := f2
where fC Cext − fCint − fCaccB corresponds to the nodal force vector at node 2 including
2 2
the external, internal and velocity-dependent forces, MC 2 is the mass matrix of node 2, and
q, q̇, q̈ are the displacement, velocity and acceleration vectors for the other nodes than node 2.
Note that the element matrices and the nodal force vectors for the ordinary element 2, in
Figure 7, are known and these nodal vectors and matrices are given in terms of the rotation
vectors W2 , Ẇ2 , Ẅ2 ∈ mat TI . Substituting WC
2 = B · W2 and (82)b for (85), we have the force
vector and the mass matrix at the node for all W2
C
2 (W2 , Ẇ2 , q, q̇, q̈) − M2 ḂẆ2 ) ∈ mat TI
f2 =BT (fC C C
(86)
M2 =BT MC
2 B ∈ mat TI ⊗ mat TI

Linearized form of the force vector f2 reads after using the chain rule and relations (84)
C
2 B + C2 Ḃ + M2 B̈) + K2 (−f2 + M2 Ẅ2 ) ∈ mat TI ⊗ mat TI
K2 :=−DW2 f2 = BT (KC C C C C
(87)
2 B + 2M2 Ḃ) ∈ mat TI ⊗ mat TI
C2 :=−DẆ2 f2 = BT (CC C

2 := −DWC f2 , C2 := −D
where KC C C C
Cf , and the geometric stiffness matrix K2 yields
2 Ẇ2 2

2
K2 (f) = [(f · n)I + f ⊗ n + n ⊗ f − 3(f · n)n ⊗ n] (88)
2
We have only considered the virtual work at node 2. Of course, the other nodes of the same element
depend on the rotation WC
2 and its time derivatives. When linearizing W2 and its time derivative
C
C
Ẇ2 , we may just substitute following formulas:

WC
2 =B · W2
C
(89)
Ẇ2 =B · Ẇ2 + Ḃ · W2
C
and the linearizations of the acceleration Ẅ (84)d, e.
The interpolation in the switching beam element, Figure 7, is carried out as follows:

WC (s) = N1 (s)WC
1 + N2 (s)W2 ∈ mat TIC
C
(90)

where linear shape functions Ni (s) are used. We note that interpolation (90) is fully defined and it
is a genuine formula where the nodal rotation vectors WiC and the interpolated value WC (s) belong
to the same vector space of rotation, namely mat TIC .
Finally, we note that the complement of complement rotation vector is equal to the rotation
vector itself, i.e. WCC = W, hence the switching beam element in Figure 7 is equivalent to the
case where at node 2 the rotation vector WC 2 has been switched, and at node 1 the rotation vector
W1 (= WCC 1 ) is still less than that of straight angle hence it is non-switched.

4. NUMERICAL RESULTS

In this section, we present several numerical examples to illustrate the performance of the proposed
total Lagrangian beam element. We choose the simplest finite element approximations based on
the 2-node isoparametric beam element with linear interpolation functions. The rotation field is
interpolated in the vector spaces of rotation mat TIC or mat TI , namely

W(s)=N1 (s)W1 + N2 (s)W2 ∈ mat TI

(91)
WC (s)=N1 (s)WC
1 + N2 (s)W2 ∈ mat TIC
C

The nodal rotation vector Wi is altered according to Definition 2.2 into the complement nodal
rotation vector WiC when the rotation angle exceeds straight angle, i.e. >. Correspondingly,
the complement nodal rotation vector WiC is altered into the vector Wi with the aid of the same
formula when the complement rotation angle exceeds straight angle, i.e. C >. In the elements
where one node has the complement rotation vector and another has the (primary) rotation vector
we use the switching beam element described in Section 3.4.1.
The two-point Gaussian quadrature is used in the numerical computations of the acceleration
vectors and their tangential matrices, and to avoid the locking phenomenon the one-point quadrature
is used for the internal vectors and their tangential matrices.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1039

4.1. Cantilever 45-degree bend

The first example presents a geometrically non-linear analysis of a cantilever 45◦ bend (with the
radius R = 100, yielding the length of the cantilever beam L = 25) placed in the horizontal plane
with a vertical static force applied at the free end, Figure 8. This example was introduced by
Bathe and Bolourchi [26] and it is one of the most frequently used example. The length of the
rotation vector remains smaller than 2, so that a single rotation parametrization chart is used during
computations. The cantilever beam is divided into eight 2-node beam elements with cross-section
properties given in Figure 8.
The free end displacements are given in Table I compared with the other author’s results. Com-
paring results in Table I, we found a minor difference from each other that is due to the difference of
constitutive relations, interpolation functions, rotation parametrization or even cross-section prop-
erties. However, our model and the model presented in [16] give practically the same response,
because of the identical rotation parametrization, i.e. total Lagrangian updating formulation. The
final solution is obtained by applying the load in six equal load steps with the full Newton–Raphson
iteration method. We also observed the identical quadratic convergence rates as in [16].

4.2. Helical beam

This example was introduced by Ibrahimbegović [28] where the straight cantilever beam is at the
free end under the force vector f and the moment vector m ∈ spat TR load both in the direction
e3 , Figure 9. Purpose of the example is to show how our switching beam element described in
Section 3.4.1, works in situation when the free end rotation angle exceeds 2. If acting on the
moment load only, the free end rotation angle is equal to 20 that corresponds to 10 revolutions.

10 7 10
7 10 10
1

Figure 8. Cantilever 45◦ bend beam under free end force.

Table I. Cantilever free end displacement components (u i is the component

in the direction ei ).
Reference Displacement u 1 Displacement u 2 Displacement u 3

Present −23.696 53.497 13.668

[26] −23.5 53.4 13.4
[11] −23.48 53.37 13.5
[14] −23.67 53.50 13.73
[3] −23.87 53.71 13.63
[27] −23.746 53.407 13.601
[16] −23.697 53.498 13.668

Figure 9. Straight cantilever beam under the force f and the moment m ∈spat TR load.

0.9 100 elements

200 elements
0.8

0.7

0.6
Loading

0.5

0.4

0.3

0.2

0.1

0
−1 −0.5 0 0.5 1 1.5 2 2.5 3 3.5
Displacement

Figure 10. Free end out-of-plane displacement curve versus loading.

During computation the spatial moment vector m ∈ spat TR , which is a spin vector, is mapped
into the initial vector space of rotation mat TI via (54), (33)d and RT = TT [16], giving MI = T · m.
This mapping is necessary in order to get the external nodal load vector of total Lagrangian beam
element where the rotation vector W(s) ∈ mat TI .
At the free end, the out-of-plane displacement component (in the direction e3 ), when loading
is increased to its final value, is shown in Figure 10. We have similar but not identical results
compared with [28], where out-of-plane displacement is almost symmetrical around zero. We
observe the decrease in interpolation error when the cantilever beam is divided into 200 elements
compared with the model of 100 elements. In addition, we also observed similar results as in
Figure 10 when the updated Lagrangian formulation described in [14] was used.

4.3. Fast symmetrical top

Next, we examine how our beam element acts as a rigid body in a dynamical case. We consider
in this example the motion of the fast symmetrical top with one point fixed originally given

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1041

in [29], Figure 11. Newmark time-integration scheme is used with scheme parameters = 0.25
and = 0.5, i.e. the trapezoidal rule.
We obtain the acceleration force vector and its tangential tensors, in addition to, the mass tensor
of rigid body element in total Lagrangian formulation from Equations (52), (68), and (71) by
keeping the dependent variables constant and integrating over the length of the beam. The top is
under gravitation load −mge3 , Figure 11, that causes the moment load in the rigid body element
and the force load in Reissner’s beam element. The numerical constants, initial condition, and the
external moment MI in the material frame {E1 , E2 , E3 } are for the rigid body element as follows:
J=5 · E1 ⊗ E1 + 5 · E2 ⊗ E2 + 1 · E3 ⊗ E3
MI =−mglTT (W) · (E3 × (RT E3 )) ∈ mat TI (92)
W0 =0.3 · E1 , X0 = 50 · E3 , Ẇ0 = T−1 (W0 ) · X0
where m, g, l are the mass of the top, the gravity constant, and the distance of the top gravity
centre with respect to the origin O. Initial conditions mean that angular velocity of the top is about
eight revolutions per time unit.

Figure 11. Fast symmetrical tops at the initial position modelled by a rigid body and a Reissner’s beam.

0.335 0.6
Rigid body Rigid body
0.33 Stiff beam Stiff beam
Angle of precession [rad]

0.5
Angle of nutation [rad]

Flexible beam Flexible beam

0.325
0.4
0.32
0.3
0.315

0.31 0.2

0.305 0.1

0.3 0
0 0.5 1 1.5 0 0.5 1 1.5
Time t [s] Time t [s]

Figure 12. Time history for the nutation and the precession angle.

The numerical results of time history for the nutation and precession angles, for the top modelled
by a rigid body and by a single beam element are shown in Figure 12 with the constant time step
h = 0.001. Two beam models are exploited: ‘Stiff beam’ where constant is given in Figure 12 and
‘Flexible beam’ where stiffness constants are reduced to the 100th part. Figure 12 indicates a very
nearby behaviour for the ‘Rigid body’ and the ‘Stiff beam model’, as expected. The results are
identical with computation given in [13] when a different rotation updating scheme, an updated
Lagrangian scheme, is used.

4.4. Right-angle cantilever beam

This test example is adopted from [12]. An L-shaped cantilever beam, with material properties
shown in Figure 13, is subjected to an out-of-plane concentrated load applied at the elbow.
The magnitude of this load follows the pattern of a hat function, as shown in Figure 13. The
computations are carried out by the Newmark scheme with the constant time step h = 0.2. Two
finite element models are built: one model has in total four linear beam elements and another has in

Figure 13. Right-angle cantilever beam.

15 15
20 elements 20 elements
10 4 elements 10 4 elements
Elbow displacement

Tip displacement

5 5

0 0

−5 −5

−10 −10
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time t [s] Time t [s]

Figure 14. Right-angle cantilever beam. The elbow and the tip out-of-plane displacement time history.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1043

total 20 elements. The cantilever undergoes a finite free vibration with combined bending and
torsion after the removal of the applied load. In this example, the length of the rotation vector
remains less than .

Figure 15. Two-component robot arm: problem data.

0 12

−2 40 elements 40 elements
8 elements 10 8 elements
Tip horizontal displacement

Tip vertical displacement

−4
8
−6

−8 6

−10
4
−12
2
−14

−16 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Time t [s] Time t [s]

9 20
40 elements 40 elements
8 18
8 elements 8 elements
Tip out−of−plane displacement

7 16
Tip displacement norm

6 14
5
12
4
10
3
8
2
6
1
0 4
−1 2
−2 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Time t [s] Time t [s]

Figure 16. Robot arm: time history of tip displacement components and
its length for 8- and 40-element mesh.

The computed time history response for the elbow and the tip out-of-plane displacement is
shown in Figure 14. These results correspond to the results given in [12, 18].

4.5. Two-component robot arm

This example, introduced in [18], considers a simple multibody system composed of two flexible
beams connected by a spherical joint. The system is brought in motion by imposing simultaneously
the forced displacement in the out-of-plane direction e3 and a rotation about this axis, see Figure 15.
Two finite element models are built: one model has in total eight equal linear beam elements and
another has in total 40 elements. The computations are carried out by the Newmark scheme with
the constant time step h = 0.01.
The computed time history response for the tip horizontal, vertical, out-of-plane displacement
components and the length of the tip displacement vector is shown in Figure 16. These results
correspond to the results given in [18, 30].

5. CONCLUSIONS

We have introduced the new Reissner’s geometrically exact beam element that is based on the total
Lagrangian updating procedure. The element has the total rotation vector as the dependent variable
and the singularity problems at rotation angle 2 and its multiples are handled by the change of
parametrization on the rotation manifold SO(3). This change of parametrization is applied via the
complement rotation vector. The kinematic assumptions of the beam element are based on the
Timoshenko–Reissner hypothesis. The beam formulation has several benefits like all the unknown
vectors belong to the same tangential vector space, no need for secondary storage variables,
the path-independence property in static case, any standard time-integration algorithm may be
used, the symmetric stiffness tensor, the simple form of the kinetic energy and all the non-linear
effects are included. Several numerical examples illustrate a good efficiency of the introduced
element.
The consistent stiffness, gyroscopic, centrifugal and loading tensors of the total Lagrangian
formulation are also given explicitly. We have shown that the work conjugate pair for the total
rotation vector, the total moment vector (54), has a conservative nature contrary to the moment
vector (26). In addition, we have clarified what the material stress and strain tensors (37), as the
work conjugate pair, are used in Reissner’s beam theory. Nevertheless, a further study for modelling
constraint equations in the total Lagrangian formulation is still needed.

APPENDIX A

We give some useful and practical formulas that we have accomplished in the main text. Espe-
cially the derivatives of the tangential transformation T, (8), have essential rule in a linearization
procedure. We define the tensor C1 with the aid of directional derivative of the vector T · V in
direction W
C1 (V, W) · W:=DW (T · V) · W ∀V ∈ E 3
(A1)
⊗ W+c3 (W · V)W ⊗ W−c4
C1 (V, W) = c1 V ⊗ W−c2 (WV) V+c5 ((W · V)I+W ⊗ V)

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1045

where coefficients ci are given by

cos − sin sin + 2 cos − 2
c1 := 3
, c2 :=
4
(A2)
3 sin − 2 − cos cos − 1 − sin
c3 := , c4 := , c5 :=
5 2 3
The limit process gives

lim C1 (V, W) = 12
V (A3)
W→0

because of the coefficient c4 → − 12 as → 0. The tensor C1 can be viewed as an operator

C1 ∈ L(mat TI , mat TR ) just like the material transformation operator T itself.
A very similar expression (in the spatial description) comes from the directional derivative of
the vector TT · V in the direction W where the tensor C2 is defined via relation

C2 (V, W) · W:=DW (TT · V) · W ∀V ∈ E 3

(A4)
⊗ W+c3 (W · V)W ⊗ W+c4
C2 (V, W) = c1 V ⊗ W+c2 (WV) V + c5 ((W · V)I+W ⊗ V)

where coefficients are given in (A2). The limit process gives

lim C2 (V, W) = − 12
V (A5)
W→0

We also note that tensors C1 (V, W) and C2 (V, W) depend linearly on the vector V but nonlinearly
on the total rotation vector W.
We also need the time derivative of the transformation T, giving

Ṫ(Ẇ, W) = c1 (W · Ẇ)I − c2 (W · Ẇ)W ˙ + c (Ẇ ⊗ W + W ⊗ Ẇ)

+ c3 (W · Ẇ)W ⊗ W + c4 W (A6)
5

where the coefficients are given in (A2). The limit value of the tensor Ṫ is
˙

lim Ṫ(Ẇ, W) = − 12 W (A7)
W→0

The directional derivative of the term CT1 (W , W) · V can be written as

C3 (V, W , W) · W := DW (CT1 (W , W) · V) · W ∀V ∈ E 3

VW ) + c3 (W · W )(V · W))I + c2 (W ⊗
C3 (V, W , W) = (c1 (W · V) + c2 (W · VW )
S

1
+ c3 (W · W )(W ⊗ V) + (c (W · V) + c2 (W ·
VW )
S 1
+ c3 (W · W )(V · W))(W ⊗ W)
+ c3 (V · W)(W ⊗ W ) + c5 (W ⊗ V) (A8)
S S

where ⊗ denotes the symmetric tensor product, and ci represents the derivatives of the coefficients
S
(A2), given by

(a ⊗ b):=(a ⊗ b) + (b ⊗ a) ∀a, b ∈ E 3
S

3 sin − 2 sin − 3 cos 2 cos − 5 sin − 8 cos + 8

c1 = , c2 = (A9)
4 5
7 cos + 8 + 2 sin − 15 sin
c3 =
6
The limit process appears that

lim C3 (V, W , W) = − 13 (W · V)I + 16 (W ⊗ V) (A10)

W→0 S

We note that the tensor C3 is symmetrical since it is the second derivative of the scalar term
V · TW .
The variation of the angular rotation vector XR reads in terms of the total rotation vector

XR = T · Ẇ + C1 (Ẇ, W) · W (A11)

where we have used (13)a and (A1). The variation of material angular acceleration tensor AR ,
(14a), reads in terms of the total rotation vector

AR = T · Ẅ + C4 (Ẇ, W) · Ẇ + (C1 (Ẅ, W) + C5 (Ẇ, W)) · W (A12)

where C1 is given in (A1) and the tensors C4 and C5 are defined by the following derivative
formulas:

C4 (Ẇ, W) · Ẇ := DẆ (ṪẆ) · Ẇ and C5 (Ẇ, W) · W: = DW (ṪẆ) · W

C4 (Ẇ, W) = (c1 + c5 )(W · Ẇ)I + 2c3 (W · Ẇ)(W ⊗ W) + 2c5 (W ⊗ Ẇ)

+ (c1 + c5 )(Ẇ ⊗ W) − c2 (W Ẇ) ⊗ W − c2 (W · Ẇ)W
(A13)

c
C5 (Ẇ, W) = (c3 (W · Ẇ)2 + c5 (Ẇ · Ẇ))I + 3 (W · Ẇ)2 + c3 (Ẇ · Ẇ) (W ⊗ W)

c1 c
+ 2c3 (W · Ẇ)(W ⊗ Ẇ) + Ẇ) ⊗ W
+ c3 (W · Ẇ)(Ẇ ⊗ W) − 2 (W · Ẇ)(W

− c2 ( W ˙ + (c + c )(Ẇ ⊗ Ẇ)
Ẇ) ⊗ Ẇ + c2 (W · Ẇ)W
1 5

The limit process gives

lim C4 (Ẇ, W) = O and lim C5 (Ẇ, W) = − 16 (Ẇ ⊗ Ẇ) + 16 (Ẇ · Ẇ)I (A14)
W→0 W→0

The tensors C4 and C5 are non-symmetrical tensors in general.

Copyright q 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 70:1009–1048
DOI: 10.1002/nme
TOTAL LAGRANGIAN GEOMETRICALLY EXACT BEAM ELEMENT 1047

REFERENCES
1. Belytschko T, Schwer L. Large displacement, transient analysis of space frames. International Journal for
Numerical Methods in Engineering 1977; 11:65–84.
2. Oran C. Tangent stiffness in space frames. Journal of the Structural Division (ASCE) 1973; 99:987–1001.
3. Crisfield MA. A consistent co-rotational formulation for non-linear, three-dimensional, beam elements. Computer
Methods in Applied Mechanics and Engineering 1990; 81:131–150.
4. Crisfield MA. Non-Linear Finite Element Analysis of Solids and Structures, Volume 2: Advanced Topics. Wiley:
Chichester, 1997.
5. Avello A, Garcia de Jalon JG. Dynamics of flexible multibody systems using cartesian co-ordinates and large
displacement theory. International Journal for Numerical Methods in Engineering 1991; 32:1543–1563.
6. Rhim J, Lee SW. A vectorial approach to computational modelling of beams undergoing finite rotations.
International Journal for Numerical Methods in Engineering 1998; 41:527–540.
7. Reissner E. On one-dimensional finite-strain beam theory: the plane problem. Journal of Applied Mathematics
and Physics (ZAMP) 1972; 23:795–804.
8. Reissner E. On one-dimensional large-displacement finite-strain beam theory. Studies in Applied Mathematics
1973; 52:87–95.
9. Petrov E, Géradin M. Finite element theory for curved and twisted beams based on exact solutions for three
dimensional solids, part I: Beam concept and geometrically exact nonlinear formulations. Computer Methods in
Applied Mechanics and Engineering 1998; 165:43–92.
10. Simo JC. A finite strain beam formulation. The three-dimensional dynamic problem. Part I. Computer Methods
in Applied Mechanics and Engineering 1985; 49:55–70.
11. Simo JC, Vu-Quoc L. A three-dimensional finite rod model. Part II: computational aspects. Computer Methods
in Applied Mechanics and Engineering 1986; 58:79–116.
12. Simo JC, Vu-Quoc L. On the dynamics in space of rods undergoing large motion—a geometrically exact approach.
Computer Methods in Applied Mechanics and Engineering 1988; 66:125–161.
13. Mäkinen J. Critical study of newmark-scheme on manifold of finite rotations. Computer Methods in Applied
Mechanics and Engineering 2001; 191:817–828.
14. Cardona A, Géradin M. A beam finite element non-linear theory with finite rotations. International Journal for
Numerical Methods in Engineering 1988; 26:2403–2438.
15. Géradin M, Cardona A. Flexible Multibody Dynamics: A Finite Element Approach. Wiley: Chichester, 2001.
16. Ibrahimbegović A, Frey F, Kozar I. Computational aspects of vector-like parametrization of three-dimensional
finite rotations. International Journal for Numerical Methods in Engineering 1995; 38:3653–3673.
17. Makowski J, Stumpf H. On the ‘symmetry’ of tangent operators in nonlinear mechanics. Journal of Applied
Mathematics and Mechanics (ZAMM) 1995; 75:189–198.
18. Ibrahimbegović A, Al-Mikdad M. Finite rotations in dynamics of beams and implicit time-stepping schemes.
International Journal for Numerical Methods in Engineering 1998; 41:781–814.
19. Marsden JE, Ratiu TS. Introduction to Mechanics and Symmetry: A Basic Exposition of Classical Mechanical
Systems. Springer: New York, 1999.
20. Stuelpnagel J. On the parametrization of the three-dimensional rotation group. SIAM Review 1964; 6:422–430.
21. Ogden RW. Non-Linear Elastic Deformations. Ellis Horwood: Chichester, 1984.
22. Mäkinen J. A formulation for flexible multibody mechanics—Lagrangian geometrically exact beam elements
using constraint manifold parametrization. TUT Applied Mechanics and Optimization, Research Report, vol. 3,
2004; 89. http://www.tut.fi/∼jmamakin/vk.pdf
23. Stumpf H, Hoppe U. The application of tensor algebra on manifolds to nonlinear continuum mechanics—invited
survey article. Journal of Applied Mathematics and Mechanics (ZAMM) 1997; 77:327–339.
24. Iserles A, Nørsett SP. On the solution of linear differential equations in lie groups. Philosophical Transactions of
the Royal Society of London, Series A—Mathematical Physical and Engineering Sciences 1999; 357:983–1019.
25. Crisfield MA, Jelenic G. Objectivity of strain measures in the geometrically exact three-dimensional beam theory
and its finite element implementation. Proceedings of the Royal Society of London, Series A—Mathematical
Physical and Engineering Sciences 1999; 455:1125–1147.
26. Bathe K-J, Bolourchi S. Large displacement analysis of three-dimensional beam structures. International Journal
for Numerical Methods in Engineering 1979; 14:961–986.
27. Ibrahimbegović A. A finite element implementation of geometrically nonlinear Reissner’s beam theory: three-
dimensional curved beam elements. Computer Methods in Applied Mechanics and Engineering 1995; 122:11–26.

28. Ibrahimbegović A. On the choice of finite rotation parameters. Computer Methods in Applied Mechanics and
Engineering 1997; 149:49–71.
29. Simo JC, Wong KK. Unconditionally stable algorithms for rigid body dynamics that exactly preserve energy and
momentum. International Journal for Numerical Methods in Engineering 1991; 31:19–52.
30. Ibrahimbegović A, Mamouri S. On the rigid components and joint constraints in nonlinear dynamics of flexible
multibody systems employing 3D geometrically exact beam model. Computer Methods in Applied Mechanics
and Engineering 2000; 188:805–831

View publication stats

Computers and Structures: C. Martín Saravia, Sebastián P. Machado, Víctor H. Cortínez
No ratings yet
Computers and Structures: C. Martín Saravia, Sebastián P. Machado, Víctor H. Cortínez
15 pages
2006 Nanakorn Vu Large Displ Beam FINEL
No ratings yet
2006 Nanakorn Vu Large Displ Beam FINEL
9 pages
Determination of Natural Frequency of Euler's Beams Using Analytical and Finite Element Method
No ratings yet
Determination of Natural Frequency of Euler's Beams Using Analytical and Finite Element Method
75 pages
Geometrically Exact 3D Beam Theory: Implementation of A Strain-Invariant Finite Element For Statics and Dynamics
No ratings yet
Geometrically Exact 3D Beam Theory: Implementation of A Strain-Invariant Finite Element For Statics and Dynamics
31 pages
1 s2.0 S0045782522008040 Main
No ratings yet
1 s2.0 S0045782522008040 Main
36 pages
Conference
No ratings yet
Conference
10 pages
Transverse Vibration of EulerBeam-iitg
100% (1)
Transverse Vibration of EulerBeam-iitg
20 pages
A Total Lagrangian Timoshenko Beam Formulation For Geometrically Nonlinear Isogeometric Analysis of Spatial Beam Structures
No ratings yet
A Total Lagrangian Timoshenko Beam Formulation For Geometrically Nonlinear Isogeometric Analysis of Spatial Beam Structures
29 pages
On A Non-Linear Formulation For Curved Timoshenko Beam
No ratings yet
On A Non-Linear Formulation For Curved Timoshenko Beam
17 pages
Nonlinear Analysis of Curved Beams
No ratings yet
Nonlinear Analysis of Curved Beams
17 pages
Finite Strain Beam Model for Space Frames
No ratings yet
Finite Strain Beam Model for Space Frames
31 pages
130 Dynamic
No ratings yet
130 Dynamic
6 pages
1 s2.0 0045794989902551 Main
No ratings yet
1 s2.0 0045794989902551 Main
20 pages
Dissertation Remo
No ratings yet
Dissertation Remo
222 pages
Numerical Meth Engineering - 2021 - Jirásek - Efficient Finite Difference Formulation of A Geometrically Nonlinear Beam
No ratings yet
Numerical Meth Engineering - 2021 - Jirásek - Efficient Finite Difference Formulation of A Geometrically Nonlinear Beam
41 pages
Zhang 2016
No ratings yet
Zhang 2016
11 pages
Nonlinear Beam Dynamics Analysis
No ratings yet
Nonlinear Beam Dynamics Analysis
18 pages
386-Dynamic Analysis of Finitely Stretched and Rotated Three-Dimensional Space-Curved Beams
No ratings yet
386-Dynamic Analysis of Finitely Stretched and Rotated Three-Dimensional Space-Curved Beams
15 pages
273-An Accurate and Robust Geometrically Exact Curved Beam Formulation For Multibody Dynamic Analysis
No ratings yet
273-An Accurate and Robust Geometrically Exact Curved Beam Formulation For Multibody Dynamic Analysis
42 pages
275-Consistent Quaternion Interpolation For Objective Finite Element Approximation of Geometrically Exact Beam
No ratings yet
275-Consistent Quaternion Interpolation For Objective Finite Element Approximation of Geometrically Exact Beam
17 pages
(2018) ASM International Volume 141 Issue 5 Linguagem Inglês
No ratings yet
(2018) ASM International Volume 141 Issue 5 Linguagem Inglês
9 pages
111-Large Displacement Analysis of Three-Dimensional Beam Structures
No ratings yet
111-Large Displacement Analysis of Three-Dimensional Beam Structures
26 pages
Bathe 79-Libre PDF
No ratings yet
Bathe 79-Libre PDF
26 pages
Paper 043
No ratings yet
Paper 043
7 pages
Numerical Vibration Analysis of Rectangular Beams For Different End Conditions
No ratings yet
Numerical Vibration Analysis of Rectangular Beams For Different End Conditions
4 pages
Lecture Notes - Finite Element Method: January 2020
No ratings yet
Lecture Notes - Finite Element Method: January 2020
6 pages
EXACT DYNAMIC STIFFNESS MATRIX OF A Bending Torsional Coupled Beam
No ratings yet
EXACT DYNAMIC STIFFNESS MATRIX OF A Bending Torsional Coupled Beam
9 pages
Vibrations of Non-Uniform Continuous Beams Under Moving Loads
No ratings yet
Vibrations of Non-Uniform Continuous Beams Under Moving Loads
16 pages
CourseNotes FEM SHZ
No ratings yet
CourseNotes FEM SHZ
6 pages
1 s2.0 S0045782520306204 Main
No ratings yet
1 s2.0 S0045782520306204 Main
28 pages
430-Nonlinear Elastic Frame Analysis by Finite Element（积分平均化）
No ratings yet
430-Nonlinear Elastic Frame Analysis by Finite Element（积分平均化）
20 pages
Nonlinear Buckling Behaviours of Thin-Walled Functionally Graded Open Section Beams
No ratings yet
Nonlinear Buckling Behaviours of Thin-Walled Functionally Graded Open Section Beams
11 pages
Science PDF
No ratings yet
Science PDF
10 pages
324-Quadratically Convergent Direct Calculation of Critical Points For 3d Structures Undergoing Finite Rotations-9
No ratings yet
324-Quadratically Convergent Direct Calculation of Critical Points For 3d Structures Undergoing Finite Rotations-9
14 pages
Vibration 07 00050
No ratings yet
Vibration 07 00050
21 pages
397-An Accurate Singularity-Free Formulation of A Three-Dimensional Curved Euler-Bernoulli Beam For Flexible Multibody Dynamic Analysis
No ratings yet
397-An Accurate Singularity-Free Formulation of A Three-Dimensional Curved Euler-Bernoulli Beam For Flexible Multibody Dynamic Analysis
10 pages
Exact Shape Functions For Timoshenko Beam Element: IOSR Journal of Computer Engineering May 2017
No ratings yet
Exact Shape Functions For Timoshenko Beam Element: IOSR Journal of Computer Engineering May 2017
10 pages
Engineers' Guide to Rotational Analysis
No ratings yet
Engineers' Guide to Rotational Analysis
51 pages
Nonlinear Vibration of 2D FG Beams
No ratings yet
Nonlinear Vibration of 2D FG Beams
13 pages
430-A Co-Rotational, Updated Lagrangian Formulation For
No ratings yet
430-A Co-Rotational, Updated Lagrangian Formulation For
12 pages
Free Vibration of Functionally Graded Beams and Frameworks Using The Dynamic Stiffness Method
No ratings yet
Free Vibration of Functionally Graded Beams and Frameworks Using The Dynamic Stiffness Method
24 pages
Nonlinear Finite Element Methods
No ratings yet
Nonlinear Finite Element Methods
8 pages
On A Geometrically Exact Beam Model and Its Finite Element Approximation
No ratings yet
On A Geometrically Exact Beam Model and Its Finite Element Approximation
11 pages
Geometrical Stiffness of 3D Beam Elements
No ratings yet
Geometrical Stiffness of 3D Beam Elements
27 pages
Rose School: European School of Advanced Studies in Reduction of Seismic Risk
No ratings yet
Rose School: European School of Advanced Studies in Reduction of Seismic Risk
49 pages
JVC PJP 2019
No ratings yet
JVC PJP 2019
7 pages
Application of Geometrically Exact Beam Formulation For Modeling Advanced Geometry Rotor Blades - M.Tech Thesis
No ratings yet
Application of Geometrically Exact Beam Formulation For Modeling Advanced Geometry Rotor Blades - M.Tech Thesis
66 pages
10 1 1 492 2300 PDF
No ratings yet
10 1 1 492 2300 PDF
93 pages
Problems in Engineering: Fifth Edition
No ratings yet
Problems in Engineering: Fifth Edition
5 pages
Nguyen
No ratings yet
Nguyen
8 pages
Vo-Nanakorn2020 Article ATotalLagrangianTimoshenkoBeam
No ratings yet
Vo-Nanakorn2020 Article ATotalLagrangianTimoshenkoBeam
22 pages
B04510717 PDF
No ratings yet
B04510717 PDF
11 pages
Vibration Analysis of Beam Subjected To Moving Loads Using Finite Element Method
No ratings yet
Vibration Analysis of Beam Subjected To Moving Loads Using Finite Element Method
11 pages
Tu 2020 6122300053 13251 13279
No ratings yet
Tu 2020 6122300053 13251 13279
116 pages
Chen and Huang - 2000 - Dynamic Stiffness of Infinite Timoshenko Beam On Viscoelastic Foundation in
No ratings yet
Chen and Huang - 2000 - Dynamic Stiffness of Infinite Timoshenko Beam On Viscoelastic Foundation in
18 pages
Exact Stiffness Matrix for Nonprismatic Beams
No ratings yet
Exact Stiffness Matrix for Nonprismatic Beams
14 pages
3D Timoshenko Beam Element Design
No ratings yet
3D Timoshenko Beam Element Design
12 pages
Seismic Assessment of Bridges Through Structural H
No ratings yet
Seismic Assessment of Bridges Through Structural H
50 pages
Non-Destructive Testing of Metal-Based Additively
No ratings yet
Non-Destructive Testing of Metal-Based Additively
28 pages
Nanomaterials 13 02741
No ratings yet
Nanomaterials 13 02741
26 pages
Total and Updated Lagrangian Rev 3
No ratings yet
Total and Updated Lagrangian Rev 3
3 pages
2009 Level 6-8 Paper 1
No ratings yet
2009 Level 6-8 Paper 1
28 pages
Bra Ket Notation
No ratings yet
Bra Ket Notation
3 pages
Advanced Trigonometry Problems and Solutions
No ratings yet
Advanced Trigonometry Problems and Solutions
3 pages
7th, 8th Prelim Round
No ratings yet
7th, 8th Prelim Round
48 pages
Catch Up - Math
No ratings yet
Catch Up - Math
4 pages
Zsigmondy's Theorem
No ratings yet
Zsigmondy's Theorem
6 pages
Linear Algebra for Engineering Students
No ratings yet
Linear Algebra for Engineering Students
17 pages
CSEC Math Question - Transformation
100% (2)
CSEC Math Question - Transformation
18 pages
Functions
No ratings yet
Functions
8 pages
Introduction to Eigenvalues
No ratings yet
Introduction to Eigenvalues
12 pages
Newton Methods for Optimization
No ratings yet
Newton Methods for Optimization
5 pages
Form 4 Mathematics Marking Scheme
No ratings yet
Form 4 Mathematics Marking Scheme
14 pages
MMDSC-1.3: I Semester M.Sc. Degree Examination, June/July - 2024 (CBCS, SLM Scheme) Mathematics
No ratings yet
MMDSC-1.3: I Semester M.Sc. Degree Examination, June/July - 2024 (CBCS, SLM Scheme) Mathematics
3 pages
1969 04erdos
No ratings yet
1969 04erdos
9 pages
Unit Hydrograph and Routing
No ratings yet
Unit Hydrograph and Routing
31 pages
Piecewise & Absolute Value Functions
No ratings yet
Piecewise & Absolute Value Functions
35 pages
Unit 4 - Partial Differential Equations
No ratings yet
Unit 4 - Partial Differential Equations
49 pages
Electrical Math Evaluation Exam
No ratings yet
Electrical Math Evaluation Exam
6 pages
Quantum Mechanics Variational Problems
No ratings yet
Quantum Mechanics Variational Problems
16 pages
Gate Question - ECE
No ratings yet
Gate Question - ECE
8 pages
Maths Class 11 Formulas
No ratings yet
Maths Class 11 Formulas
15 pages
CPP Functions PDF
No ratings yet
CPP Functions PDF
12 pages
工數傅立葉級數
No ratings yet
工數傅立葉級數
35 pages
CBSE Class 12 Maths Question Paper Solution 2019
No ratings yet
CBSE Class 12 Maths Question Paper Solution 2019
16 pages
C1 Sequences and Series - Questions PDF
100% (1)
C1 Sequences and Series - Questions PDF
8 pages
4th Grade Math Objectives
No ratings yet
4th Grade Math Objectives
24 pages
Syllabus BAP MATHS
No ratings yet
Syllabus BAP MATHS
11 pages
CBSE Test Paper 02 CH-03 Trigonometric Functions
No ratings yet
CBSE Test Paper 02 CH-03 Trigonometric Functions
6 pages
Degeneracy & Special Cases in LPP
No ratings yet
Degeneracy & Special Cases in LPP
7 pages

IJNME1

Uploaded by

IJNME1

Uploaded by

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Total and Updated Lagrangian Geometrically Exact Beam Elements

Chapter · June 2008

Jari Mäkinen Heikki Marjamäki

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Total Lagrangian Reissner’s geometrically exact beam

Received 26 October 2005; Revised 18 April 2006; Accepted 30 August 2006

Contract/grant sponsor: The Academy of Finland; contract/grant number: 206020

Copyright q 2006 John Wiley & Sons, Ltd.

Figure 1. Geometric interpretation for parametrization of the manifold when n = 3 and d = 2.

at = 0 to the manifold M is defined as

Figure 2. The change of parametrization in the parameter space E 3 for the

the singularity at = 2. As it is illustrated in Figure 2, the change of parametrization maps

inc is an incremental material rotation operator and HR is an incremental material rotation

Definition 2.5 (material tangent space)

2.1. Angular velocities, accelerations and curvatures

XR =T(WI ) · ẆI where XR ∈ mat TR , ẆI , WI ∈ mat TI for material description

KR =T(WI ) · WI where KR ∈ mat TR , WI ∈ mat TI , for material description

3. GEOMETRICALLY EXACT BEAM THEORY

3.1. Virtual work forms

W = Wext −Wint + Winert (17)

R F:P = F:P − (

R for the material rep-

where we have used the central line condition

3.2. Constitutive relations

where the constitutive tensors in the material frame are

3.3. Lagrangian formulations

W(i+1) = W(i) + W(i) ∈ mat TI (47)

C = RT xc + (RT xc )∼ TW

= C1 (W , W) · W + TW (56)

Substituting the above equation into the form (55) yields

where RN can be viewed as a push-forward of the material force vector N ∈ TX B0 . It is clear

interpolation on a non-flat manifold never preserves in an observer transformation. This is because

3.4. Consistent tangential tensors of Reissner’s beam

+ TT J · (C5 (Ẇ0 , W0 ) + C1 (Ẅ0 , W0 )) (71)

Kmat := BT0 Cnm B0

W(s)=N1 (s)W1 + N2 (s)W2 ∈ mat TI

4.1. Cantilever 45-degree bend

4.2. Helical beam

Figure 8. Cantilever 45◦ bend beam under free end force.

Table I. Cantilever free end displacement components (u i is the component

Present −23.696 53.497 13.668

0.9 100 elements

Figure 10. Free end out-of-plane displacement curve versus loading.

4.3. Fast symmetrical top

Flexible beam Flexible beam

4.4. Right-angle cantilever beam

Figure 13. Right-angle cantilever beam.

Figure 15. Two-component robot arm: problem data.

Tip vertical displacement

4.5. Two-component robot arm

where coefficients ci are given by

because of the coefficient c4 → − 12 as → 0. The tensor C1 can be viewed as an operator

C2 (V, W) · W:=DW (TT · V) · W ∀V ∈ E 3

where coefficients are given in (A2). The limit process gives

Ṫ(Ẇ, W) = c1 (W · Ẇ)I − c2 (W · Ẇ)W ˙ + c (Ẇ ⊗ W + W ⊗ Ẇ)

The directional derivative of the term CT1 (W , W) · V can be written as

C3 (V, W , W) · W := DW (CT1 (W , W) · V) · W ∀V ∈ E 3

3 sin − 2 sin − 3 cos 2 cos − 5 sin − 8 cos + 8

lim C3 (V, W , W) = − 13 (W · V)I + 16 (W ⊗ V) (A10)

XR = T · Ẇ + C1 (Ẇ, W) · W (A11)

AR = T · Ẅ + C4 (Ẇ, W) · Ẇ + (C1 (Ẅ, W) + C5 (Ẇ, W)) · W (A12)

C4 (Ẇ, W) · Ẇ := DẆ (ṪẆ) · Ẇ and C5 (Ẇ, W) · W: = DW (ṪẆ) · W

C4 (Ẇ, W) = (c1 + c5 )(W · Ẇ)I + 2c3 (W · Ẇ)(W ⊗ W) + 2c5 (W ⊗ Ẇ)

The limit process gives

The tensors C4 and C5 are non-symmetrical tensors in general.

View publication stats

You might also like

KR =T(WI ) · WI where KR ∈ mat TR , WI ∈ mat TI , for material description

W(i+1) = W(i) + W(i) ∈ mat TI (47)

C = RT xc + (RT xc )∼ TW

= C1 (W , W) · W + TW (56)

C2 (V, W) · W:=DW (TT · V) · W ∀V ∈ E 3

The directional derivative of the term CT1 (W , W) · V can be written as

C3 (V, W , W) · W := DW (CT1 (W , W) · V) · W ∀V ∈ E 3

lim C3 (V, W , W) = − 13 (W · V)I + 16 (W ⊗ V) (A10)

C4 (Ẇ, W) · Ẇ := DẆ (ṪẆ) · Ẇ and C5 (Ẇ, W) · W: = DW (ṪẆ) · W