European Journal of Computational Mechanics

Vol. 21, Nos. 3–6, June–December 2012, 184–194

Buckling analysis of tape springs using a rod model with flexible

cross-sections
Stéphane Bourgeoisa,b*, Bruno Cochelina,b, François Guinotc and Elia Picaulta
a
Laboratoire de Mécanique et d’Acoustique – UPR CNRS 7051, 31 chemin Joseph-Aiguier, F-13402
Marseille Cedex 20, France; bEcole Centrale Marseille, Technopôle de Château-Gombert, 38 rue Joliot
Curie, F-13451 Marseille Cedex 20, France; cThales Alenia Space, 100 bd du midi – B.P. 99, F-06156
Cannes La Bocca Cedex, France

This work is devoted to the study of tape springs behaviour. We focus on the instabilities
that lead to the creation of localised folds, due to the flattening of the cross-section.
Depending on the kind of loading, the folds can move along the tape, split or disappear.
Using an extended four parameters rod model with flexible cross-sections, we study several
modes of instabilities (number and localisation of folds) for such structures submitted to an
overall compression or bending.
On s’intéresse à la modélisation du comportement des mètres rubans, qui ont la particularité
de développer des pliages localisés dus à un aplatissement par flambage de la section. Sui-
vant les scénarios de chargement, les zones de pliages peuvent se déplacer le long du
ruban, se dupliquer ou disparaître. S’appuyant sur un modèle original de poutre à section
flexible à quatre paramètres cinématiques, on étudie les différents modes d’instabilités
(nombre et localisation des pliages) de ces structures soumises à des sollicitations de com-
pression et de flexion dans le plan.
Keywords: tape springs; folding; buckling; rod model
Mots-clés: mètres rubans; pliage; flambement; modèle de poutre

1. Introduction
A crucial point for satellite manufacturers is to develop ingenious compact systems that can
have an autonomous deployment during transfer orbit. In this framework, deployable struc-
tures composed of folded tape springs can offer an interesting alternative to articulated rigid
mechanisms (Santer & Pellegrino, 2008; Seffen & Pellegrino, 1999). They may lead to more
compact systems and may increase fiability by suppressing hinges, bolts and motorisations.
However, the deployment needs to be perfectly controlled and the modelling of tape springs
behaviour becomes a challenging issue.
In its free configuration, a tape spring can be assimilated to a straight rod with an open
and thin-walled cross-section of constant transverse curvature. As shown in Figure 1, when
an overall bending is applied, the tape behaves at first as a classical beam (with a more or
less uniform bending curvature) but there is suddenly the creation of a localised fold (Seffen,
2001; Seffen & Pellegrino, 1999), indicating snap-through buckling. These folds occur, thanks

*Corresponding author. Email: [email protected]

ISSN 1779-7179 print/ISSN 1958-5829 online

Ó 2012 Taylor & Francis
http://dx.doi.org/10.1080/17797179.2012.714848
http://www.tandfonline.com
 European Journal of Computational Mechanics 185

Figure 1. Folding of a measuring tape.

to the flexibility of the cross-section, which becomes flat in some localised areas to decrease
the whole strain energy. It leads to concentrate the deformation in the region of the fold,
whereas the rest of the tape recovers its undeformed shape. Playing with a measuring tape,
one can easily experience the creation of one or several folds, the migration of a fold along
the tape, the splitting of a single fold into two or the merging of two folds into one.
To simulate these buckling phenomena, the first approach consists of using a non-linear
elastic shell model in the framework of large displacements and large rotations. This can be
performed by using a commercial finite element software package (Hoffait, Bruls, Granville,
Cugnon, & Kerschen, 2009; Seffen, You, & Pellegrino, 2000; Soykasap, 2007; Walker &
Aglietti, 2007). This approach, which can be used for any kind of loading, provides accurate
results but leads to heavy and hard to drive computations. A more suitable approach consists
of taking into account the specificities of the tape springs behaviour. For instance, some
authors (Seffen & Pellegrino, 1999) consider the tape in its folded configuration as an assem-
bly of two rigid bars and a non-linear spiral spring accounting for the bending stiffness of the
fold. The rigid bars are of variable length to allow the migration of the fold along the tape. It
leads to a non-linear system with two degrees of freedom, which is successfully used to
model some experiments of dynamic deployment. Unfortunately, this method requires the
introduction of the folds (number and localisation) ab initio and is not able to account for the
creation of new folds.
Recently, an intermediate approach has been proposed in which the tape spring is consid-
ered as a rod with highly flexible cross-sections (Guinot, Bourgeois, Cochelin, & Blanchard,
2012). This approach appears to be more general than the above mentioned discrete model
and easier to handle than the full non-linear shell model. Many works have been published
on beam models with deformable cross-sections (see e.g. Gonçalves, Ritto-Corrêa, & Camo-
tim, 2010; Pimenta & Campello, 2003; Simo & Vu-Quoc, 1991; Vlassov, 1962; Zivkovic,
Kojic, Slavkovic, & Grujovic, 2001). The main idea is always to incorporate shape functions
to describe the deformation of the cross-section. The method proposed in Guinot et al. (2012)
follows this idea. Starting from a non-linear shell model in the framework of large displace-
ments, large rotations and dynamics, the main novelty resides in the use of an elastica kine-
matics to account for the large changes of the cross-section shape with very few parameters.
At last, it leads to a planar rod model with only four kinematic parameters.
The present work aims at showing that this quite simple model is able to handle complex
scenarios of buckling, taking into account the creation and disappearance of folds depending
on the loading path. The key points of the general model presented in Guinot et al. (2012)
and its implementation using an energetic approach are first recalled for the restricted case of
tape springs with a circular cross-section. In a second part, several applications examples
illustrate the ability of the model to face with complex scenarios of buckling, involving the
creation, the duplication and the migration of folds along the tape.
186 S. Bourgeois et al.

2. The rod model

2.1. Kinematic description and basic assumptions
Starting from a shell model, a tape spring is regarded as a rod described in its initial free
configuration by a straight rod line and a planar circular cross-section curve, as shown in
Figure 2. The fixed orthonormal frame (O, e1, e2, e3) is chosen such that the axis (O, e1)
contains the rod line and that the plane (O, e1, e3) is the plane of symmetry of the cross-
section curve.
We introduce a curvilinear coordinate system (s1, s2) 2 [0, L]  [a/2, a/2] to map the
geometry of the tape, with L the initial length of the rod line and a the initial length of the
cross-section curve. In the deformed configuration, the position of a material point M is given
by:

OMðs1 ; s2 Þ ¼ OGðs1 Þ þ GMðs1 ; s2 Þ; ð1Þ

where G is the point of the cross-section on the rod line. The rod model kinematics presented
in Guinot et al. (2012) relies on four assumptions:

(i) the cross-section curve remains in a plane after deformation;

(ii) the cross-section plane is orthogonal to the tangent vector of the rod line in the
deformed configuration;
(iii) the cross-section curve is considered inextensible and
(iv) the shape of the tape remains symmetric with respect to the plane (O, e1, e3).

The two first assumptions are the classical hypotheses used in the Euler-Bernoulli beam
theory. In accordance with the assumption (iv), the motion of the rod line is restrained to the
plane (O, e1, e3): the displacement of a point G on the rod line is then given by the two com-
ponents u1 (s1) and u3 (s1), and the rotation of the cross-section plane hðs1 Þ is around the axis
e2. We introduce the rotated frame ðG; er1 ; er2 ; er3 Þ, following the cross-section plane, in which
the coordinates of the material point M are denoted by y (s1, s2) and z (s1, s2). We then
obtain:

OG ¼ ðs1 þ u1 ðs1 ÞÞ e1 þ u3 ðs1 Þ e3 ;
ð2Þ
GM ¼ yðs1 ; s2 Þ er2 þ zðs1 ; s2 Þ er3 :

Figure 2. Description of the tape spring.

 European Journal of Computational Mechanics 187

According to the assumption (iii), the inextensible cross-section curve can be regarded as
an elastica. Introducing the angle bðs1 ; s2 Þ between the tangent to the cross-section curve and
vector er2, we have:


y;2 ¼ cosðbÞ;
ð3Þ
z;2 ¼ sinðbÞ:

Moreover, the cross-section curve is assumed to remain circular. The angle b is then lin-
ear with respect to s2 and the cross-section curve is entirely defined by the angle be ðs1 Þ
between the tangent at s2 = a/2 and the vector er2 (see Figure 2):
8 
> a s2 
>
< yðs1 ; s2 Þ ¼ 2be ðs Þ sin 2b ðs1 Þ a ;
e
1
a  s2  ð4Þ
>
>
: zðs1 ; s2 Þ ¼ e 1  cos ð2be ðs1 Þ Þ :
2b ðs1 Þ a

Finally, the tape spring kinematics is described by only four kinematic parameters
attached to the rod line:

• the translations u1 (s1) and u3 (s1) of the cross-section;

• the rotation hðs1 Þ of the cross-section around e2 and
• and the angle be ðs1 Þ characterising the shape of the cross-section.

We must notice here that the displacements u1 ðs1 Þ and u3 ðs1 Þ and the rotation hðs1 Þ are
linked by the assumption (iv), that leads to:

OG;1 1 þ u1;1 u3;1

C ¼ er3  ¼ sinðhÞ þ cosðhÞ r ¼ 0 with
kOG;1 k jr j
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5Þ
2 2
jr ¼ kOG;1 k ¼ ð1 þ u1;1 Þ þ ðu3;1 Þ :

The novelty of the model resides in the inextensibility assumption of the cross-section
curve. It allows to use an elastica kinematics and leads to describe the large change of the
cross-section shape with very few parameters, only one here. The idea underlying this
assumption is that the tape spring wants to decrease the strain energy by creating some folds,
where the cross-section becomes flat to decrease the bending inertia. Due to the large relative
displacements inside the cross-section, an inextensibility assumption is sufficient to obtain the
overall shape of the cross-section and to efficiently estimate the bending inertia.

2.2. Strains measures and strain energy

The tape spring is first considered as a thin shell undergoing membrane and bending strains
eab and kab defined, respectively, by the Green–Lagrange measure and the difference between
the initial and actual curvature tensors. The tape spring is submitted to large displacements
but the membrane strains remain small in practice; since the thickness is very small compared
to the two other dimensions a and L of the shell, local and global buckling will prevent large
strains to occur. As shown in Guinot et al. (2012), the small membrane strains assumption
leads to following simplified expressions:
188 S. Bourgeois et al.
8 r
> e ¼ u1;1 þ 12ðu21;1 þ u23;1 Þ;
8 >
>
¼ er þ zk r þ es ; >
> k r ¼ h;1 ;
>
>
e11 >
>
>
<k >
< es ¼ 1ðy2 þ z2 Þ;
11 ¼ k r cos b þ k11
s
; 2 ;1 ;1
with ð6Þ
>
> k22 ¼ k22 ;
s >
> k11 ¼ z;11 cos b  y;11 sin b;
s
>
: >
>
>
>
¼ k12 ; > 22 ¼ b;2  b0;2 ;
s s
k12 > k
: s
k12 ¼ b;1 ;

where b0 is the value of the angle field b in the undeformed initial configuration. This Equa-
tion (6) enlightens the strains induced by the global rod kinematics (variables with the super-
script r) and those induced by the deformation of the cross-section curve (superscript s). We
recognise in er and kr, the classical expressions of the usual tensile strain and the bending cur-
vature of a rod in the framework of large displacements and large rotations. We can remark
that the strains es and kab
s
only depend on the angle b and vanish if the cross-section remains
undeformed. The strains eab and kab are associated with the membrane stresses and bending
moments Nab and Mab in the strain energy. Since the shell width a is small compared to the
tape spring length L, we suppose that N22 ¼ N12 ¼ 0 according to classical beam theory
assumptions and we have:

Z L Z a=2
1
Ue ¼ ðe11 N11 þ kab Mab Þ ds2 ds1 : ð7Þ
0 a=2 2

Moreover, the shell is considered elastic and isotropic. The constitutive equations are then
expressed by:
8
< M11 ¼ D11 k11 þ D12 k22 ;
N11 ¼ Ae11 and M ¼ D12 k11 þ D22 k22 ; ð8Þ
: 22
M12 ¼ D33 ð2 k12 Þ;

with

Eh3 Eh3
A ¼ Eh; D11 ¼ D22 ¼ ; D12 ¼ mD11 ; D33 ¼ ; ð9Þ
12ð1  m Þ
2 24ð1 þ mÞ

where h is the shell thickness. E and m are the Young’s modulus and the Poisson’s ratio,
respectively. Using the constitutive equations (8) and the expressions (6) for the strains, the
strain energy (7) can be written, after integration over the cross-section as:

Z L
Ue ¼ ðure þ use þ urs
e Þ ds1 ; ð10Þ
0

in which ure , use and urs

e are the three terms that define the strain energy density of the rod
model:
 European Journal of Computational Mechanics 189
8  
>
> u r
¼ 1
Aaðe r 2
Þ þ ðAz 2 þ D cos2 ðbÞÞðk r Þ2 þ 2Azer k r ;
>
> e 2 11
>
> 
>
< us ¼ 1 Aðes Þ þ D11 ðk s Þ þ D22 ðk s Þ2
2 2
e 11 22
2
 ð11Þ
>
>
>
> þ2D12 k11 s 2
k22 þ 4D33 ðk12
s s
Þ ;
>
>
>
: rs
ue ¼ Aer es þ Ak r zes  k r ðD11 cosðbÞk11
s
þ D12 cosðbÞk22 s
Þ;

R a=2
where the overline denotes an integration over s2: X ðs1 Þ ¼ a=2 X ðs1 ; s2 Þ ds2 . With the kine-
matics presented in the previous section, all these integrals can be analytically derived (Gui-
not et al., 2012), leading to a closed-form expression of the strain energy density depending
on the kinematic parameters and the initial characteristics of the tape spring (geometrical and
material properties). The first term ure corresponds to the classical strain energy of a rod with
a coupling between axial stretching and bending which appears because the rod line does not
pass through the cross-section centroid. The second term use only depends on the variable b
and represents the strain energy due to the variation of the cross-section shape, independently
of the overall rod behaviour. The last term urs e induces a coupling between the overall rod
behaviour and the deformation of the cross-section.

3. Numerical implementation
The weak form of the problem is needed to use the finite element method. This weak formu-
lation can be obtained by the differentiation of the potential energy with respect to the kine-
matic parameters u1 ðs1 Þ, u3 ðs1 Þ, hðs1 Þ and be ðs1 Þ. Let us remind that the first three parameters
are constrained by Equation (5) that ensures the orthogonality of the cross-section plane with
respect to the rod line. Introducing a Lagrange multiplier kðs1 Þ associated with the constraint
Cðs1 Þ ¼ 0, we must proceed to the differentiation of the following functional:
Z L
Pðu1 ; u3 ; h; b ; kÞ ¼ U  Wext þ
e e
kC ds1 ; ð12Þ
0

where Wext is the work of external forces, depending on the loading conditions. The model
has been implemented in the finite element software COMSOL (COMSOL AB, 2008) that
handles the formal differentiation of the potential energy. The aim here is to study the rele-
vance of the model, and not to develop a robust and fast numerical tool. In this way, the
finite element software COMSOL allows to overcome the required numerical developments
because the expressions (11) can be used as such to define the strain energy density of the
rod.
The loading is only applied at the end sections where the following boundary conditions
(or their dual force) can be prescribed:

• the translations u1 and u3 of the cross-section;

• the rotation h of the cross-section;
• the angle be that defines the transverse curvature of the cross-section and
• the first derivative be;1 of this angle that prescribes the local rotation at each point of the
cross-section curve (clamped end section for example).
190 S. Bourgeois et al.

In the following, the loading is defined in terms of prescribed displacements or rotation at

ends. The default time-dependent solver (backward differentiation formula: BDF solver) of
COMSOL is used because it allows to introduce small viscous terms in the weak formulation
to face with the snap-through phenomena. Finally, the discretisation of the rod line is per-
formed with one-dimensional Hermite quintic finite elements to take into account the high-
order derivatives that appear in the model for the opening angle be .

4. Numerical applications
4.1. Introduction
The geometrical and material properties of the tape spring under consideration are given in
Table 1. In the following, several scenarios of buckling are studied with several kinds of
boundary conditions. It is shown that it always leads to the creation of one or more folds that
can move along the tape, disappear, merge or split. In the following figures, the 3D deformed
shapes of the tape spring are reconstructed, thanks to the results of the 1D rod model and the
kinematics exposed in Section 2.1. The colour plot of b are superimposed on the deformed
shapes to give information on the local z-coordinate in the local cross-section frame. The dark
blue regions correspond to areas where the cross-section is completely flattened; z increases
from dark blue (z = 0) to red. The static analyses of buckling are performed under the assump-
tion that the tape spring remains symmetric with respect to the plane (O, e1, e3). An extension
of the model to 3D motions and out-of-plane buckling analysis constitutes an ongoing work.

4.2. Buckling of the cross-section under pure overall bending

A pure overall bending is obtained with prescribed opposite rotations h at the end sections.
The other boundary conditions are given by: u1 ¼ u3 ¼ 0 and be ¼ be0 at one end and
u3 ¼ 0, be ¼ be0 and u1 free at the other end. Figure 4 (left) shows that at the beginning,
for small values of the prescribed angle hðLÞ, the tape spring behaves as a classical beam
(CBT) with non-deformable cross-sections; the response is linear. We then observe
ð0:04\hðLÞ\0:12Þ rad) a non-linearity, due to a non-uniform flattening of the cross-sections
all along the tape. The maximum flattening is obtained in the middle s1 ¼ L2 but there is no
localisation (deformed shape 2 in Figure 3). The bending moment reaches a maximum and
decreases before a snap-through that occurs for a prescribed rotation of hðLÞ ¼ 0:12 rad. The
tape switches from a first equilibrium branch to another one that corresponds to the creation
of a localised fold (deformed shape 3 in Figure 3). In the fold region, the cross-section is
completely flat and the bending inertia is much smaller that the initial one. It leads to an
important loss of the tape bending strength. When the prescribed rotation is decreased, we
observe other snap-through phenomena. Figure 4 (right) shows the evolution of the strain
energy density ue ¼ ure þ use þ urs
e along the tape with respect to the prescribed rotation hðLÞ.
This plot clearly illustrates the sudden localisation of the strain energy density. These results
are in agreement with the results obtained in Seffen et al. (2000).

Table 1. Geometrical and material properties of the tape spring.

Length L Width a Thickness h Initial angle be0 Young’s modulus E Poisson’s ratio m
1170 mm 60 mm 0.15 mm 0.6 rad 210,000 MPa 0.3
 European Journal of Computational Mechanics 191

1 2 3 4

Figure 3. Creation of a fold under pure overall bending. Deformed shapes with superimposed colour
plots of the angle b, indicating the level of flattening of the cross-section.

Figure 4. Reaction moment with respect to the prescribed rotation (left). Strain energy density along
the tape with respect to the prescribed rotation (right).

4.3. Buckling under compression

A pure compressive load is prescribed at the end sections of the tape. When this load is
applied at the centroid of the initial cross-section, the tape remains straight, corresponding to
the trivial solution (u3 ðs1 Þ ¼ hðs1 Þ ¼ be ðs1 Þ ¼ 0). On this trivial solution branch, the first
bifurcation branch is asymmetric. In the following, loading imperfections are introduced to
investigate the buckling and post-buckling behaviour under an overall compression. In partic-
ular, we study the compressive reaction force with respect to the prescribed compressive axial
displacement. Boundary conditions are known to play an important role in the buckling
behaviour of beams (Bazant & Cedolin, 1991; Nguyen, 2000; Timoshenko & Gere, 1961).
For tape springs, it is shown hereafter that they have a direct impact on the number and local-
isation of the folds that can appear.

4.3.1. Free rotations at ends

In this first case, the rotation h is free and the flexural displacement u3 is fixed at each end.
The compressive load is applied by imposing a displacement u1 at one end (s1 = L) whereas
the other is fixed. Moreover, the cross-section shape is fixed at each end (be ¼ be0 ). The
imperfection consists of a small flexural force F3 ¼ 104 N applied at the middle of the tape,
leading to a small opposite bending that preserves the symmetry with respect to the plane (O,
e1, e3). Figure 5 illustrates the overall behaviour of the tape spring. In the first part, the
response curve follows (up to the state 1) the usual scenario of a perturbed bifurcation
branch, getting closed to the trivial perfect compressive solution first and then turning sharply
to follow the bifurcated branch with a decreasing strength. We observe then a snap-through to
another branch, accompanied by the creation of a fold. This branch is characterised by a con-
192 S. Bourgeois et al.

Figure 5. Buckling under compression with free rotations at ends. The deformed shapes with the
superimposed colour plot of b illustrate the creation of the fold (right). The plot of the compressive
reaction force with respect to the axial displacement at end shows the sudden snap-through (left).

Figure 6. Buckling under compression with fixed rotations at ends. The deformed shapes with the
superimposed colour plot of b illustrate the creation of the folds (right). The plot of the compressive
reaction force with respect to the axial displacement at end shows the snap-throughs (left).

stant reaction force that does not depend on the prescribed displacement at the end section.
This snap-through is similar in nature to the case presented in Section 4.2. This is again the
overall bending that leads to the flattening of the cross-section in a localised region of the
tape. But in the present case, the overall bending occurs during the post-buckling.

4.3.2. Fixed rotations at ends

Contrary to the previous case, the rotation h is fixed at each end of the tape. Since the cross-
section is not allowed to rotate, we observe at the very beginning a snap-through with the
creation of two localised regions of flattening near the ends, where the overall bending
moment generated by the flexural displacement in the middle is maximum (see deformed
shape 2 in Figure 6). These partial foldings are followed by an overall bending with a distrib-
uted flattening all along the tape with a maximum that takes place in the middle (see the
evolution of colour plot of b on the deformed shape 3). This flattening increases, leading to a
second snap-through to another equilibrium branch for which a significant fold takes place in
the middle and the flattening near the ends are somewhat released (deformed shape 4).
Furthermore, we observe some oscillations on the colour plot of b in this deformed state,
indicating that local buckling occurs due to localised compressive loads at the ends of the
cross-section curve. Increasing the prescribed displacement at end removes these oscillations,
 European Journal of Computational Mechanics 193

Figure 7. Splitting of a fold into two.

restores the folds near the ends and increases the folding region in the middle (deformed
shapes 4 and 5).
The results obtained in these two cases (free and fixed rotations at ends) show that such a
structure is highly non-linear and sensitive to imperfections. A linearised buckling analysis is
then unsuitable for predicting the buckling behaviour of tape springs.

4.4. Splitting of a fold into two

The starting state is the deformed shape 1 shown in Figure 7. This state is obtained with the
same boundary conditions than those used in Section 4.2, except that the bending is in the
opposite sense. The boundary conditions are then modified: the displacements u1 and u3 are
fixed at each end in the deformed shape 1 and opposite increments of rotation Dh are pre-
scribed to spread the two straight parts of the tape (from 0 to p=2 at one end and 0 to p=2
at the other one). At the beginning, the rotation of the straight parts of the tape leads to an
extension of the region of flattening in the middle (see deformed shapes 1 and 2 in Figure 7).
We also observe some waves on the colour plot of deformed shape 2, indicating localised
buckling due to local compression. This intermediate state is unstable and a snap-through
occurs between the states 2 and 3 that leads to the duplication of the fold which relaxes the
local compression. In the region between the two new folds, the cross-section recovers its ini-
tial shape. The two new folds have the same properties than the initial single one: the fold
regions are cylindrical with a completely flat cross-section and a longitudinal curvature equal
to the initial transverse curvature of the cross-section (Guinot et al., 2012). Increasing the
increment of rotation moves the two folds away from each other.

5. Conclusions
In this work, we have recalled the basics of the rod model with flexible cross-sections pro-
posed in Guinot et al. (2012). This 1D planar rod model involves only four parameters and is
suitable to simulate the behaviour of tape springs. It is shown that this approach is able to
handle complex scenarios of buckling, leading to the creation, duplication and migration of
folds along the tape. Compared to shell models, this extended rod model is easier to handle
and naturally involves some mechanical quantities of interest such as the bending inertia, con-
194 S. Bourgeois et al.

trolled by the kinematic parameter be that governs the cross-section shape. This approach
enables to describe but also to understand the main phenomena that control the behaviour of
tape springs. The COMSOL software and its time dependent solver (BDF solver) was conve-
nient here for testing and evaluating the model without writing lines of code. However, it
does not allow to get the full bifurcation diagram and this may limit the understanding of the
complete behaviour. We are currently working on the use of a path following algorithm
(asymptotic numerical method) to obtain the full equilibrium diagram with stable and unstable
branches of solutions, together with the definition of a suitable 1-D finite element, for an effi-
cient and reliable numerical implementation of the proposed tape-spring model.

References
Bazant, Z.P., & Cedolin, L. (1991). Stability of structures: Elastic, inelastic, fracture, and damage theo-
ries. New York, NY: Oxford University Press.
Comsol (2008). Comsol Multiphysics User’s Guide, Comsol AB, Stockholm, http://www.comsol.com.
Gonçalves, R., Ritto-Corrêa, M., & Camotim, D. (2010). A large displacement and finite rotation thin-
walled beam formulation including cross-section deformation. Computer Methods in Applied
Mechanics and Engineering, 199(23–24), 1627–1643.
Guinot, F., Bourgeois, S., Cochelin, B., & Blanchard, L. (2012). A planar rod model with flexible thin-
walled cross-sections: Application to the folding of tape springs. International Journal of Solids and
Structures, 49(1), 73–86.
Hoffait, S., Bruls, O., Granville, D., Cugnon, F., & Kerschen, G. (2009). Dynamic analysis of the self-
locking phenomenon in tape-spring hinges. Acta Astronautica, 66(7–8), 1125–1132.
Nguyen, Q.S. (2000). Stability and nonlinear solid mechanics. Chichester: Willey.
Pimenta, P., & Campello, E. (2003). A fully nonlinear multi-parameter rod model incorporating general
cross-sectional in-plane changes and out-of-plane warping. Latin American Journal of Solids and
Structures, 1(1), 119.
Santer, M., & Pellegrino, S. (2008). Compliant multistable structural elements. International Journal of
Solids and Structures, 45(24), 6190–6204.
Seffen, K. (2001). On the behavior of folded tape-springs. Journal of Applied Mechanics – Transactions
of the ASME, 68(3), 369–375.
Seffen, K., & Pellegrino, S. (1999). Deployment dynamics of tape springs. Proceedings of the Royal
Society London A, 455, 1003–1048.
Seffen, K., You, Z., & Pellegrino, S. (2000). Folding and deployment of curved tape springs. Interna-
tional Journal of Mechanical Sciences, 42(10), 2055–2073.
Simo, J., & Vu-Quoc, L. (1991). A geometrically-exact rod model incorporating shear and torsion-war-
ping deformation. International Journal of Solids and Structures, 27(3), 371–393.
Soykasap, O. (2007). Analysis of tape spring hinges. International Journal of Mechanical Sciences, 49
(7), 853–860.
Timoshenko, S.P., & Gere, J.M. (1961). Theory of elastic stability. New-York, NY: Mc Graw-Hill.
Vlassov, B. (1962). Pièces longues en voiles minces (Long thin-walled beams). Paris: Eyrolles.
Walker, S., & Aglietti, G. (2007). A study of tape spring fold curvature for space deployable structures.
Proceedings of the Institution of Mechanical Engineers Part G – Journal of Aerospace Engineering,
221(G3), 313–325.
Zivkovic, M., Kojic, M., Slavkovic, R., & Grujovic, N. (2001). A general beam finite element with
deformable cross-section. Computer Methods in Applied Mechanics and Engineering, 190(20–21),
2651–2680.