See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/259097006

Stiffness matrix based form-finding method of

tensegrity structures

Article in Engineering Structures · January 2014

DOI: 10.1016/j.engstruct.2013.10.014

CITATIONS READS

28 437

4 authors:

Li-Yuan Zhang Yue Li

University of Science and Technology Beijing Tsinghua University
16 PUBLICATIONS 78 CITATIONS 16 PUBLICATIONS 230 CITATIONS

SEE PROFILE SEE PROFILE

Yanping Cao Xi-Qiao Feng

Tsinghua University Tsinghua University
108 PUBLICATIONS 2,111 CITATIONS 417 PUBLICATIONS 8,356 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimization of ballistic performance of new hybrid plate containing stainless steel with nanograined
layer for helmets View project

Mechanics of Biological Materials View project

All content following this page was uploaded by Li-Yuan Zhang on 14 November 2017.

The user has requested enhancement of the downloaded file.

 Engineering Structures 58 (2014) 36–48

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Stiffness matrix based form-finding method of tensegrity structures

Li-Yuan Zhang a, Yue Li b, Yan-Ping Cao a, Xi-Qiao Feng a,⇑
a
CNMM & AML, Department of Engineering Mechanics, Tsinghua University, Beijing 100084, China
b
Institute of Nuclear and New Energy Technology, Tsinghua University, Beijing 100084, China

a r t i c l e i n f o a b s t r a c t

Article history: A highly efficient form-finding method of tensegrity is presented on the basis of the structural stiffness
Received 24 February 2013 matrix, which is defined as the derivative of the out-of-balance force vector with respect to the nodal
Revised 7 October 2013 coordinate vector. The stiffness matrix and the total potential energy of the structure are utilized to direct
Accepted 9 October 2013
the rapid convergence of the structural configuration to the self-equilibrated and stable state. In the pro-
grammed procedure, we employ the stochastic selecting algorithm to exclude rigid-body motions, the
restricted step algorithm to guarantee the positive definiteness of the structural stiffness matrix, and
Keywords:
the line search algorithm to minimize the total potential energy. This form-finding method allows us
Tensegrity
Form-finding
to easily determine the self-equilibrated and stable configuration of a tensegrity from an arbitrary initial
Structural stiffness matrix state. A number of representative examples are given to demonstrate its accuracy and efficacy for both
Numerical method regular and irregular tensegrity structures of large scale.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction polyhedral tensegrities consisting of Z-shaped and rhombic ele-

mentary cells [19,20].
As a type of form-active structures, tensegrities need a form- For most tensegrities, the form-finding analysis can be made
finding process to determine their self-equilibrated configurations only by numerical methods. The form-finding of a tensegrity can
in the absence of external loads [1]. In the form-finding analysis, a be transformed into a nonlinear programming problem, in which
tensegrity is usually modelled as a series of axially loaded elements the total length of bars (or strings) is maximized (minimized) un-
(bars and strings) jointed together by spherical hinges (nodes) [2]. der the constraint condition that the total length of strings (or bars)
The bars (or ‘struts’) can withstand both compression and tension, is specified [21,22]. However, the nonlinear programming method
while the strings (or ‘cables’) can only be tensioned [3]. In the past is time-consuming and not viable for irregular tensegrities of large
decades, the form-finding of tensegrity structures has attracted scale [2,7,23]. As another widely used form-finding method, the
considerable attention of researchers from different disciplines, dynamic relaxation method solves a pseudo-dynamic process and
e.g. mathematics, biomechanics, civil and aerospace engineering controls the structural equilibrium by integrating a fictitious dy-
[4–6], and various form-finding methods have been proposed (for namic equation [24–26]. In the dynamic relaxation method, both
reviews, see, e.g. [1,2,7]). the mass and damping parameters are artificial, and one needs to
The existing form-finding methods can be divided into two set the sizes of integration time steps based on previous experi-
broad classes, analytical and numerical. In general, the analytical ences in order to rapidly produce a convergent solution [1,3]. By
form-finding methods are used for regular symmetric structures, introducing the concept of element force density, the nonlinear
e.g. prismatic (or cylindrical) tensegrities [8–10] and truncated equilibrium equations can be linearized and a self-equilibrated
regular polyhedral tensegrities [11–13]. According to the symme- tensegrity is determined when the force density matrix satisfies
try-adapted formulation based on the group representation theory, the rank deficiency condition [27–30]. However, the force density
Zhang et al. [14,15] and Zhang and Ohsaki [16] systematically ana- method cannot directly control the geometric parameters of a
lyzed the self-equilibrium and stability properties of tensegrities structure, and several nodal coordinates or element lengths must
with dihedral and tetrahedral symmetries. Recently, by solving be further specified to determine its final configuration [1,5,7]. As
the self-equilibrium conditions in terms of the force densities of a powerful tool for structural analysis, the finite element method
elements, Zhang et al. [17,18] derived the unified closed-form solu- has also been introduced for the form-finding of tensegrities
tions for the self-equilibrated and super-stable truncated regular [31,32], in which one generally needs to assume an initial shape
close to the final solution. Recently, a few bioinspired form-finding
algorithms have been proposed, including the evolutionary meth-
⇑ Corresponding author. Tel.: +86 10 6277 2934; fax: +86 10 6278 1824. od and the genetic method [23,33–36]. In addition, Micheletti
E-mail address: [email protected] (X.-Q. Feng).
and Williams [37] developed a marching algorithm by altering

0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.10.014
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 37

Nomenclature

e(ij) element e connecting nodes i and j N serial number of an iteration step

ge(ij) axial stiffness of element e(ij) ne(ij) unit direction vector of element e(ij)
i, j serial number of nodes Oi vector consisting of the direction indices for the ele-
I serial number of the elements containing node i ments listed in Mi
J, L row and column numbers of a matrix or vector pi coordinate vector of node i
ke(ij) stiffness matrix of element e(ij) p coordinate vector of all nodes in the structure
K structural stiffness matrix of the tensegrity Dp increment of the coordinate vector p in an iteration step
0
leðijÞ natural length of element e(ij) qe(ij) force density of element e(ij)
le(ij) current length of element e(ij) te(ij) internal force of element e(ij)
Mi vector consisting of the serial numbers of the elements U total potential energy of the structure
containing node i li total number of the elements containing node i
nnode total number of nodes ei out-of-balance force vector of node i
nelem total number of elements e out-of-balance force vector of the structure

the lengths of elements in the form-finding process, which needs j, where i ¼ 1; 2; . . . ; nnode ; j ¼ 1; 2; . . . ; nnode ; and e ¼ 1; 2; . . . ; nelem .
to start from a known stable configuration. Inspired by the molec- Its length can be calculated by
ular dynamics method, Li et al. [3] proposed a Monte Carlo form- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
finding method to search the stable configuration of a tensegrity leðijÞ ¼ kpi  pj k ¼ ðxi  xj Þ2 þ ðyi  yj Þ2 þ ðzi  zj Þ2 ; ð1Þ
by employing the stochastic scheme of nodal displacements. By
integrating the finite element and mathematical programming where kk denotes the norm of a vector, pi ¼ ½xi ; yi ; zi T 2 R3 is the
methods, Klinka et al. [38] established an energy optimization coordinate vector of node i, and (xi, yi, zi) are the nodal coordinates.
based form-finding method, in which a novel type of elements hav- Here and in the sequel, the superscript T denotes the transpose of a
ing constant stresses was introduced. With the combination of the vector or matrix.
equilibrium and geometrical compatibility equations, Koohestani Then the unit direction vector of element e(ij) is defined as
and Guest [39] proposed a form-finding method using the Carte- pi  pj
sian components of element vectors as variables. Their method neðijÞ ¼ ; ð2Þ
leðijÞ
can be considered as a natural counterpart of the force density
method. where i and j are the first and second nodes of the element. Note
Despite the above development of form-finding analysis, most that the nodal sequences in an element can be selected arbitrarily,
previous methods are efficient only for tensegrity structures of rel- but once determined, they will be fixed throughout the form-find-
atively small scale. Therefore, it is still of great importance to de- ing process.
velop a straightforward and efficient numerical form-finding In the following analysis, for simplicity, all elements are as-
method suitable for irregular tensegrities consisting of a large sumed linear elastic. Thus the constitutive relation of element
number of nodes and elements. To this end, we here propose a e(ij) is written as
stiffness matrix based form-finding (SMFF) method for tensegrity 8
<0 0
structures. Considering the relation between the total elastic strain for a string with leðijÞ 6 leðijÞ ;
teðijÞ ¼ ð3Þ
energy and the stiffness matrix of the structure, a self-adaptive :g 0
leðijÞ Þ otherwise;
eðijÞ ðleðijÞ 
technique is used to direct the rapid convergence of structural con-
0
figuration to a self-equilibrated and stable state even when the ini- where te(ij) denotes the internal force in the element, leðijÞ is the nat-
tial configuration assumed in the calculation is greatly different ural length of the element, and ge(ij) is a constant and referred to as
from it. The present paper is organized as follows. Section 2 defines the axial stiffness. The force density of the element is then defined
the parameters needed to evaluate the structural stiffness matrix as qe(ij) = te(ij)/le(ij). Eq. (3) implies that during the form-finding iter-
of a tensegrity. Section 3 describes the proposed SMFF method in ation, the strings are allowed to be slack and the bars are allowed to
detail. Section 4 shows the applications of the SMFF method to a be tensioned.
number of representative tensegrities. Section 5 further discusses Correspondingly, the elastic strain energy of element e(ij) can be
the features of this method. Finally, Section 6 gives the concluding computed by
remarks. 8
0
<0 for a string with leðijÞ 6 leðijÞ ;
U eðijÞ ¼ 2 ð4Þ
:1g 0
 leðijÞ Þ otherwise:
2 eðijÞ ðleðijÞ
2. Structural stiffness matrix of a tensegrity

In this section, we define some geometric and mechanical The total potential energy of the structure is the sum of the
parameters of a tensegrity, which will be used to calculate its energies in all elements, that is,
structural stiffness matrix.
nX
elem

U¼ U eðijÞ : ð5Þ
e¼1
2.1. Basic parameters

Consider a tensegrity with a pre-specified topology. Refer to a 2.2. Structural stiffness matrix
Cartesian coordinate system ðO  xyzÞ. Let nnode and nelem denote
the total numbers of nodes and elements in the structure, respec- The structural stiffness matrix K 2 R3nnode 3nnode of a tensegrity is
tively. We use e(ij) to denote the element e connecting nodes i and given by [40]
38 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

K ¼ S þ A  ðG  Q Þ  AT ; ð6Þ compute the initial potential energy 0U, the structural stiff-
ness matrix 0K, and the out-of-balance forces 0e (see Sec-
where S 2 R3nnode 3nnode is the stress matrix calculated from the force
tion 3.2). Number the iteration step as N = 1. Here and in
densities of all elements [41], A 2 R3nnode nelem is the equilibrium ma-
the sequel, we use the preceding superscripts 0 and N to
trix consisting of all element direction vectors ne(ij) [30], and
denote the parameters in the initial configuration and in
Q 2 Rnelem nelem and G 2 Rnelem nelem are the diagonal matrix of element
the configuration obtained at iteration step N, respectively.
force densities and the diagonal axial stiffness matrix defined as
Q ¼ diagð. . . ; qeðijÞ ; . . .Þ and G ¼ diagð. . . ; g eðijÞ ; . . .Þ. Since all ele-
Iteration:
ments are linear elastic, Eq. (6) can be simplified as [27,29]
0
!
leðijÞ (ii) Select appropriate constraint conditions to exclude the
K ¼ S þ A  diag . . . ; g eðijÞ ; . . .  AT : ð7Þ structural rigid-body motions according to the stochastic
leðijÞ
selecting algorithm in Section 3.3. Judge whether the stiff-
ness matrix with the applied constraints is positive definite
For the numerical procedure, the structural stiffness matrix is or not. If not, use the restricted step algorithm in Section 3.4
calculated by: to modify the stiffness matrix such that it is positive definite.
(iii) Solve the nodal displacements NDp by using the stiffness
8 matrix obtained from Step (ii), and then employ the line
> keðijÞ if i–j and i is connected with j by an element;
>
< search algorithm in Section 3.5 to produce the updated
0 if i–j and i does not connect with j;
Kij ¼ X ð8Þ nodal coordinates Np.
>
>
:  Kik if i ¼ j; (iv) Calculate the out-of-balance forces Ne, and judge whether
k–j
the current configuration is self-equilibrated or not accord-
where the subscript k ¼ 1; 2; . . . ; nnode , and Kij 2 R33 is a submatrix ing to the criterion of equilibrium in Section 3.2. If not, com-
of K spanning the rows from 3i  2 to 3i and columns from 3j  2 to pute the structural stiffness matrix NK, and then go to Step
3j. keðijÞ 2 R33 is referred to as the element stiffness matrix and ex- (ii) and set N = N + 1.
pressed as [42]
Termination:
0
leðijÞ
keðijÞ ¼ qeðijÞ I3 þ g eðijÞ neðijÞ  nTeðijÞ ; ð9Þ
leðijÞ (v) Terminate the iteration, when an equilibrated configuration
33
has been obtained.
where I3 ¼ diagð1; 1; 1Þ 2 R is the three-dimensional identity
matrix.
3.2. Out-of-balance forces and criterion of equilibrium

3. SMFF method For a given structure without externally applied forces, the
out-of-balance forces are defined as the resultants of the internal
In this section, we formulate the SMFF method for three-dimen- forces at all nodes. Let ei 2 R3 denote the out-of-balance force
sional tensegrity structures. All nodes and elements are assumed to vector at node i, which can be solved by using the neighbor lists
have zero mass and damping. li, Mi , and Oi [3]. Here, li is the total number of elements con-
taining node i, and Mi 2 Rli records the serial numbers of these
3.1. Form-finding procedure elements. Oi 2 Rli consists of the direction indices for the ele-
ments listed in Mi, and all its components have the value of 1
To establish a form-finding method of high efficiency for var- or 1. Corresponding to the element direction vector in Eq. (2),
ious types of tensegrities, several difficult problems need to be we specify Oi as
solved. Firstly, as any rigid-body motion of a structure will make (
the structural stiffness matrix singular, constraint conditions 1 if i is the first node of element MiI ;
OiI ¼ ð10Þ
must be applied properly to exclude all possible rigid-body mo- 1 if i is the second node of element M iI ;
tions. Secondly, the stiffness matrix of the structure may lose
its property of positive definiteness in the presence of large and where the subscript I = 1, 2, . . ., li denotes a local serial number of
highly nonlinear responses. A negative eigenvalue of the stiffness node i.
matrix indicates that the structure is unstable, whereas a zero- For a configuration p at an iteration step, the out-of-balance
eigenvalue will cause the divergence of the numerical procedure. force vector at node i is calculated by [3]
To guarantee the positive definiteness of the stiffness matrix,
li  
X  
therefore, one should eliminate all eigenvalues with either nega-
ei ¼ OiI tMi nMi ; ð11Þ
tive or zero value. Thirdly, an appropriate optimization objective I¼1
I I

should be defined to ensure that the calculations can approach

the targeted self-equilibrated configuration. In the SFMM method, where M iI denotes the Ith element containing node i, and the direc-
we will invoke the stochastic selecting algorithm, the restricted tion vector nMi and the internal force t Mi of element M iI are com-
I I

step algorithm and the line search algorithm to solve these prob- puted from Eqs. (2) and (3), respectively.
lems, respectively. Once the out-of-balance force vectors of all nodes has been
As shown in the flow chart in Fig. 1, the SMFF procedure con- determined from Eq. (11), the out-of-balance force vector of the
sists of the following five main steps. whole structure e 2 R3nnode can be obtained by
Input: T
e ¼ ½. . . ; eTi ; . . . : ð12Þ
(i) Specify the basic data of a tensegrity, including the total
numbers of nodes and elements, the structural topology, A tensegrity will be in self-equilibrium when all out-of-balance
and the natural lengths and axial stiffnesses of all elements. forces are zero. In our method, a sufficiently small value, eth , is set
Give an arbitrary configuration as its initial state 0p, and as the tolerance to judge the structural equilibrium. Define the
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 39

Fig. 1. Flow chart of the SMFF method.

maximum norm among the out-of-balance force vectors at all dth = 20%. This method has been demonstrated to be highly effi-
nodes by cient for various types of tensegrities.
As constraint conditions have been applied to exclude rigid-
e ¼ max jjei jj: ð13Þ
body motions, the structural stiffness matrix will change, from
i¼1;...;nnode
the original K to K, which can be determined from

When the criterion 8

< K JJ ¼ 1
> if the Jth DOF is constrained;
e 6 eth ð14Þ K JL ¼ K LJ ¼ 0 if the Jth DOF ðJ–LÞ is constrained; ð15Þ
>
:
is satisfied, the structure is regard to have reached an equilibrated K JL ¼ K JL otherwise;
state.
where the subscripts J and L are in the range from 1 to 3nnode.
3.3. Stochastic selecting algorithm
3.4. Restricted step algorithm
As aforementioned, in the form-finding analysis of a tensegrity,
all rigid-body motions should be excluded to avoid the singularity In the SMFF method, we start the form-finding calculation of a
of the structural stiffness matrix. For a three-dimensional tenseg- tensegrity from an arbitrary initial configuration, which may be
rity, therefore, six constraint conditions, corresponding to the six greatly different from the final self-equilibrated state. In this case,
degrees of freedom (DOFs), should be properly given. As the basic large deformations and strong nonlinearity may be involved in the
rule of specifying constraint conditions, the resulted configuration form-finding process [40,43]. As a result, the stiffness matrix of the
of the tensegrity should be self-equilibrated when all external sup- structure with all possible rigid-body motions having been ex-
ports and constraints have been removed. Our simulations show cluded may still have one or more eigenvalues with negative or
that the convergence of iteration can be significantly accelerated zero values. Thus in each iteration step, the positive definiteness
by appropriate selection and adjustment of constraint conditions of the structural stiffness matrix K should be examined. To this
in the form-finding process. In addition, it is found that the calcu- end, we introduce the restricted step algorithm [44] into the SMFF
lation will be time-consuming if we simply specify the same group form-finding process, which can avoid the divergence of numerical
of constrained DOFs throughout the iteration or randomly change calculations and ensure the achievement of the stable configura-
the constrained DOFs in each iteration step. Instead, a combination tion of the considered structure.
of the two methods for specifying constraint conditions will help For the configuration p at an iteration step, if the corresponding
save calculation time. stiffness matrix K is positive definite, one can proceed the form-
Based on the above considerations, we invoke a stochastic finding process. Otherwise, we use a diagonal matrix consisting
selecting algorithm to apply six constraint conditions that can ex- of sufficiently large components to modify K such that it becomes
clude all possible rigid-body motions of the studied tensegrity. In positive definite. Thus, the modified stiffness matrix, K, e is ex-
the first iteration step (N = 1), six constraints are randomly given pressed as
to six different DOFs. In the iteration step N P 2, constraints will
be applied according to the relative decrease of the out-of-balance (
e ¼ K if K is positive definite;
forces evaluated in step N  1, i.e., N1d = [(N2e)  (N1e)]/(N2e). If K ð16Þ
N1
d is larger than the threshold dth, the DOFs under constraints in K þ njkK;min jI otherwise;
step N will be the same as those in step N  1; otherwise, they will
be randomly selected again. Through the simulations of a large where I ¼ diagð. . . ; 1; . . .Þ 2 R3nnode 3nnode is the identity matrix, kK;min
number of tensegrities, we suggest setting the threshold is the minimum eigenvalue of K with jkK;min j denoting its absolute
40 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

(a) (b) qs
qs
q
b
q
b

(c) (d) qs1

qs1 qs2
qb1 qb

qb2 qs2

Fig. 2. Self-equilibrated configurations of two-dimensional tensegrities obtained by the SMFF method.

(a) (b) 1
6
Lateral
Bar
string
4 5
Bottom
string
Top
4 string

5 6
3 2 2 3
1

Fig. 3. Definition of elements in a prismatic tensegrity. (a) A triangular prism, and (b) the topology graph of the corresponding 3-prismatic tensegrity.

value, and n is a coefficient. From Eq. (16), one can readily verify
e will be positive definite provided that n > 1.0.
1+ v + a that the matrix K
It should be mentioned that the positive definiteness of the ma-
1 + v r1 qs1 2+v trix Ke defined in Eq. (16) is crucial to ensure the convergence of the
form-finding calculation. As aforementioned, the structural config-
qb uration in an iteration step may be far from the final self-equili-
brated solution. In this case, the corresponding stiffness matrix K
qs3 may be not positive definite. A negative eigenvalue of K indicates
that the structure is unstable, whereas a zero-eigenvalue may
cause the divergence of the numerical procedure. In order to pro-
O ceed the form-finding calculation, some ‘artificial stiffnesses’ can
1 be added at some DOFs to modulate K into a positive definite ma-
x qs2 r2 2 e Here the artificial stiffnesses are determined based on the
trix K.
α y minimal eigenvalue of K. After a few iteration steps, K will become
positive definite as the obtained configuration approaches the final
Fig. 4. A part of nodes and elements in a v-prismatic tensegrity. stable state. Then no modification will be further needed and
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 41

(a) v = 3, a = 1, r2 r1 = 1.0000000 (b) v = 6, a = 2, r2 r1 = 1.5031169

(c) v = 12, a = 3, r2 r1 = 1.7756715 (d) v = 24, a = 4, r2 r1 = 2.2295597

(e) v = 48, a = 5, r2 r1 = 2.0564734 (f) v = 96, a = 6, r2 r1 = 1.3767790

Fig. 5. Self-equilibrated configurations of prismatic tensegrities obtained by the SMFF method.

Table 1
Comparison between analytical and numerical form-finding solutions of prismatic tensegrities.

v a 0 0
ls2 =ls1
0 0
ls3 =ls1
0 0
lb =ls1 r2/r1 Figure a/ p qs1 =qb qs2 =qb qs3 =qb
Analyticala Numericalb Analytical Numerical Analytical Numerical Analytical Numerical
3 1 1.0 1.5 2.0 1.0000000 Fig. 5a 1/6 0.166667 0.577350 0.577350 0.577350 0.577350 1 1.000000
6 2 1.75 2.0 4.0 1.5031169 Fig. 5b 1/6 0.166667 2.603475 2.603475 1.152306 1.152306 1 1.000000
12 3 2.2 5.0 8.0 1.7756715 Fig. 5c 1/4 0.250000 9.371846 9.371846 2.972350 2.972350 1 1.000000
24 4 2.85 21.0 24.0 2.2295597 Fig. 5d 1/3 0.333333 32.716269 32.716270 6.581510 6.581510 1 1.000000
48 5 3.0 32.0 38.0 2.0564734 Fig. 5e 19/48 0.395833 77.267212 77.267212 18.270442 18.270442 1 1.000000
96 6 10.75 750.0 800.0 1.3767790 Fig. 5f 7/16 0.437500 125.449146 125.449146 66.181906 66.181905 1 1.000000
a
The analytical solutions are given in Eqs. (20)–(23).
b
The numerical results are obtained from the SMFF method.

e ¼ K. Therefore, the introduction of artificial stiffness in some

K 3.5. Line search algorithm
intermediate steps does not interfere with the final form-finding
results. Another key issue in the form-finding analysis is how to rapidly
In addition, attention should be paid to the specification the approach the self-equilibrated and stable configuration of the
coefficient n in Eq. (16). If n is very close to 1.0, Ke may become structure. In the SMFF method, we use the line search algorithm
ill-conditioned and make the numerical calculation to diverge. [44] in each iteration step to minimize the total potential energy
e will be greatly different from K, lower-
Whereas if n is too large, K of the system.
ing the computational efficiency of the form-finding process. Our In step N, we update the nodal coordinate vector Np from that in
numerical experiments show that setting n = 1.01 can achieve a step N  1 by
balance between the convergence and efficiency of the form-find- N
p ¼ N1 p þ ðN gÞðN DpÞ; ð17Þ
ing computations for most tensegrity structures.
42 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

energy. Then we set g = gmax in the calculation of step N. The effec-

(a) tiveness of the above line search algorithm has been demonstrated
by a number of examples, as described below.

4. Examples

In this section, we provide some representative examples to

demonstrate the high accuracy and efficacy of the proposed SMFF
method for both regular and irregular tensegrity structures of large
scale. In these examples, the axial stiffnesses of all elements are
specified as unity, and their force densities are tuned by varying
their natural lengths.

4.1. Two-dimensional tensegrities

(b) Remaining string

First, we employ the SMFF method to investigate four represen-
tative two-dimensional tensegrities, as shown in Fig. 2, whose
Truncating topologies are given in [7,28,39]. In all these structures, the natural
string
lengths of all strings and all bars are specified as 1.0 and 5.0,
respectively. Then an arbitrary initial configuration is assumed to
start the form-finding analysis.
The first example we here consider is a X-frame (Fig. 2a), which
Bar
is the simplest two-dimensional tensegrity. The SMFF method
gives the force densities of the strings and bars as qs = 0.5590
and qb = 0.5590, respectively, in consistency with the analytical
result in [7].
The second structure has a hexagonal shape, as illustrated in
Fig. 2b. The force densities of the strings and bars are determined
as qs = 0.5000 and qb = 0.2500, respectively, satisfying the analyt-
ical result qs + 2qb = 0 [28].
(c) Type-2 string
The third structure also has a hexagonal shape (Fig. 2c), in
Type-1
string which the elements are divided into four kinds, including inclined
strings, vertical strings, horizontal bars, and vertical bars. Their
force densities are calculated as qs1 ¼ 0:5539, qs2 ¼ 0:3770,
qb1 ¼ 0:2770, and qb2 ¼ 0:3193, respectively, satisfying the ana-
lytical results qs1 þ 2qb1 ¼ 0 and 2qs1 qs2 þ ðqs1 þ 2qs2 Þqb2 ¼ 0 [39].
In the octagonal tensegrity in Fig. 2d, the elements include ver-
Bar
tical (or horizontal) strings, inclined strings, and bars. Their force
densities are determined as qs1 ¼ 0:3926, qs2 ¼ 0:4776, and
qb ¼ 0:1485, respectively, which agree with the analytical result
qs1 qs2 þ ð2qs1 þ qs2 Þqb ¼ 0 obtained using the nodal static equilib-
rium analysis [7].

4.2. Prismatic tensegrities

Fig. 6. Truncated regular polyhedral tensegrities: (a) a truncated cube, and the
In the second example, we consider the class of prismatic
corresponding (b) Z-based and (c) rhombic TRP tensegrities.
tensegrities. A v-prismatic tensegrity structure has two parallel
regular polygons on its top and bottom, each consisting of v nodes
where NDp denotes the increment of the nodal displacement vector and v strings. The two polygons are connected into a tensegrity
corresponding to the out-of-balance force vector N1 e
 after the iter- prism by v bars and v strings. Taking the triangular prism in
ation step N  1. It is calculated by Fig. 3a as an example, the topology graph of the corresponding
3-prismatic tensegrity is illustrated in Fig. 3b.
1 
N e
Dp ¼ ðN1 KÞ  ðN1 eÞ; ð18Þ By employing the SMFF method, we can readily solve the geo-
metric and mechanical parameters of the self-equilibrated pris-
where e
 is modified from e according to the constraint conditions matic tensegrities. In such a structure, there exist four kinds of
determined in Section 3.3. It is written as elements, including top strings, bottom strings, lateral strings,
 and bars, as shown in Fig. 3b. Assume that each kind of elements
0 if the Jth DOF is constrained;
eJ ¼ have the same value of force density. In this case, the form-finding
eJ otherwise:
problem of prismatic tensegrities can be analytically solved by
The coefficient g in Eq. (17) has a value in the range of 0 < g 6 1. In using the nodal static equilibrium analysis method [7]. A part of
this study, we determine the value of g as follows. First, we assign nodes and elements in a self-equilibrated v-prismatic tensegrity
g = 1.0 to update the nodal coordinates and the structural energy. If are shown in Fig. 4, where a is an integer in the range of
the energy decreases, we will use g = 1.0 in the line search algo- 1 < = a <v, and r1 and r2 are the radii of the top and bottom
rithm; otherwise, the bisection method [44] is employed to search circumcircles, respectively. Let a denote the relative twisting angle
the maximum value 0 < gmax < 1.0 that can decrease the structural between the top and bottom polygons, and qs1, qs2, qs3 and qb the
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 43

(a) 0.0 (b) 0.0

Analytical solution Analytical solution
-0.2 Numerical solution -0.2 Numerical solution

-0.4 -0.4
qZb qZs1

qZb qZs1
-0.6 -0.6

-0.8 -0.8

-1.0 -1.0
0.0 0.6 1.2 1.8 2.4 0.0 0.6 1.2 1.8 2.4
qZs2 qZs1 qZs2 qZs1

(c) 0.0 (d) 0.0

Analytical solution Analytical solution
-0.2 Numerical solution -0.2 Numerical solution

-0.4 -0.4

qZb qZs1
qZb qZs1

-0.6 -0.6

-0.8 -0.8

-1.0 -1.0
0.0 0.6 1.2 1.8 2.4 0.0 0.6 1.2 1.8 2.4
qZs2 qZs1 qZs2 qZs1

(e) 0.0
Analytical solution
-0.2 Numerical solution

-0.4
qZb qZs1

-0.6

-0.8

-1.0
0.0 0.6 1.2 1.8 2.4
qZs2 qZs1

Fig. 7. Form-finding solutions of Z-based truncated (a) tetrahedral, (b) cubic, (c) octahedral, (d) dodecahedral, and (e) icosahedral tensegrities. The lines denote the analytical
solutions of Zhang et al. [17], while the circles are our numerical results.

0 0 0 0
force densities of the top strings, bottom strings, lateral strings, and ls1 , ls2 , ls3 and lb denote the natural lengths of each top string, bot-
bars, respectively. Then the analytical solution of such a v-pris- tom string, lateral string, and bar, respectively. For some represen-
matic tensegrity is written as [8,9] tative examples shown in Fig. 5a–f with the parameters listed in
Table 1, the numerical results are compared with the above analyt-
a 1 a
¼  ; ð20Þ ical solution. It is seen that the twisting angles a, the force densities
p 2 v qs1 =qb , qs2 =qb , and qs3 =qb in the elements determined from the
SMFF method are all in perfect agreement with the analytical solu-
qs1 r 2 sin pva
 ¼  ; ð21Þ tion in Eqs. (20)–(23).
qb r 1 1  cos 2vp

qs2 r 1 sin pva

 ¼  ; ð22Þ 4.3. Z-based truncated regular polyhedral tensegrities
qb r 2 1  cos 2vp
Now we investigate Z-based truncated regular polyhedral (TRP)
qs3
 ¼ 1: ð23Þ tensegrities, including truncated tetrahedral, cubic, octahedral,
qb
dodecahedral, and icosahedral tensegrities [17]. In these struc-
tures, the connecting relations of strings correspond to the edges
The SMFF method proposed in this paper is then used to solve a of a truncated regular polyhedron, and the bars are added accord-
number of prismatic tensegrities with various values of v, a, and ing to the rules of Z-shaped elementary cells [20]. For illustration,
r2/r1. Each kind of elements are assumed to have the same natural Fig. 6a and b give a truncated cube and the corresponding Z-based
length and the same force density in our numerical simulation. Let TRP tensegrity.
44 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

(a) 2.5 (a)

Analytical solution
Numerical solution
2.0
−qRs2 qRb

1.5

qRs1 < qRs2

1.0

qRs1 > qRs2

0.5
(b) z
y
x
0.5 1.0 1.5 2.0 2.5
−qRs1 qRb

(b) 2.5
Analytical solution
Numerical solution
2.0
−qRs2 qRb

1.5
(c)
qRs1 < qRs2
1.0

qRs1 > qRs2

0.5
0.5 1.0 1.5 2.0 2.5
−qRs1 qRb

(c) 2.5
Analytical solution Fig. 9. Construction of a cubic grid tensegrity by using the polyhedral truncation
Numerical solution scheme: (a) a polyhedron consisting of seven regular cubes, (b) the edges of its
corresponding truncated polyhedron, and (c) the obtained tensegrity.
2.0
−qRs2 qRb

Fig. 6b, respectively. Assume each kind of elements have an iden-

1.5 tical natural length. By varying the natural length ratios between
the different kinds of elements, we can obtain the self-equilibrated
qRs1 < qRs2 and stable Z-based TRP tensegrities with various combinations of
1.0 force densities in the elements. The qZb =qZs1 versus qZs2 =qZs1 curves
are shown in Fig. 7, where the circles and lines represent the
qRs1 > qRs2
numerical and analytical solutions, respectively. It can be seen
from Fig. 7a–e that for truncated tetrahedral, cubic, octahedral,
0.5
dodecahedral, and icosahedral tensegrities, the SMFF results all
0.5 1.0 1.5 2.0 2.5
agree well with the analytical solution in Ref. [17]. In addition, it
−qRs1 qRb is noticed that all numerical results are located on the super-stabil-
ity curve theoretically predicted by Zhang et al. [17], as shown in
Fig. 8. Form-finding solutions of rhombic truncated (a) tetrahedral, (b) cubic/
octahedral, and (c) dodecahedral/icosahedral tensegrities. The lines denote the
Fig. 7, indicating that all configurations obtained from our numer-
analytical solutions of Zhang et al. [18], while the circles are our numerical results. ical method are super-stable.

4.4. Rhombic truncated regular polyhedral tensegrities

By solving the self-equilibrium conditions in terms of the force
densities of elements, Zhang et al. [17] derived a unified analytical We proceed to study all rhombic truncated regular polyhedral
solution for the self-equilibrated and super-stable Z-based TRP tensegrities, including rhombic truncated tetrahedral, cubic/octa-
tensegrities. In this subsection, these tensegrities are solved by hedral, and dodecahedral/icosahedral tensegrities [18]. A rhombic
using the SMFF method, and the results are compared with the tensegrity can be constructed by replacing all Z-shaped cells in a
analytical solution in Zhang et al. [17]. Z-based tensegrity with rhombic cells [19]. Corresponding to the
Let qZs1 , qZs2 , and qZb denote the force densities of the truncating Z-based truncated cubic tensegrity in Fig. 6b, for instance, a rhom-
strings, remaining strings, and bars in the Z-based TRP tensegrity in bic truncated cubic tensegrity is built, as shown in Fig. 6c.
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 45

polyhedral truncation scheme [20]. For illustration, Fig. 9a shows

a polyhedron assembled by seven regular cubes, one of which is lo-
cated at the center and the others are attached to its six surfaces.
By truncating all vertices of the cubes, a closed cubic grid is ob-
tained, as shown in Fig. 9b. Then a cubic grid tensegrity can be de-
rived by adding strings and bars according to the rules of Z-based
tensegrities [19].
For comparison, the SMFF method proposed in this paper and
the Monte Carlo form-finding method developed by Li et al. [3]
are employed to investigate the self-equilibrated configurations
of cubic grid tensegrities. In the calculations, we assume that all
strings have identical properties, and so do the bars. It is found that
the results obtained from the two form-finding methods agree very
well for all examples, demonstrating the efficacy and accuracy of
the SMFF method. However, the Monte Carlo form-finding method
is much more time-consuming, as will be discussed in Section 5.
The three-dimensional and planar projection images of an
equilibrated cubic grid tensegrity obtained from the form-finding
analysis are shown in Fig. 9c, which has the same orthogonality
as the polyhedron in Fig. 9a. Just like the polyhedron in Fig. 9a
which can be assembled into a three-dimensional network, cubic
grid tensegrities can also be expanded in three dimensions to con-
struct reticulated tensegrities of larger scale and various shapes.
Such an example is provided in Fig. 10, which consists of eight
tensegrities given in Fig. 9c. This kind of expandable tensegrities
hold promise for significant applications in a diversity of fields,
e.g. aerospace engineering.

4.6. Irregular tensegrities

Finally, we apply the SMFF method to some representative

irregular structures. Three examples are provided here. The first
two have the same topology as the prismatic tensegrity with
v = 50 and a = 1. The natural lengths of each top string, bottom
string, and lateral string are set as 0.02, 0.02, and 1.40, respectively.
In the first structure, the natural lengths of the fifty bars along the
circumferential direction constitute an arithmetic progression
from 1.50 to 1.99, with a step difference of 0.01, while in the sec-
ond, their natural lengths are randomly specified between 1.00
and 2.00.
In the third example, a tensegrity consisting of 138 strings, 69
Fig. 10. A large-scale tensegrity assembled by eight cubic grid tensegrities in bars, and 207 nodes is constructed by invoking the polyhedral
Fig. 9c. truncation method [3]. The natural lengths of the strings and bars
are set as 1.0 and 10.0, respectively.
The self-equilibrated configurations of the three examples
solved by the SMFF method are shown in Fig. 11a–c. We also make
By employing the SMFF method, we can easily determine the
the form-finding analysis by applying the Monte Carlo form-find-
self-equilibrated configurations of rhombic TRP tensegrities. Let
ing method developed by Li et al. [3]. Again, the two methods have
qRs1 , qRs2 , and qRb denote the force densities of the type-1 strings,
a good agreement.
type-2 strings, and bars (see Fig. 6c), respectively. Assume each
kind of elements have the same natural length. Various numerical
solutions are obtained in the SMFF simulation for different ratios 5. Discussions
between the natural lengths of type-1 strings, type-2 strings, and
bars. Fig. 8 gives the normalized force density qRs2 =qRb as a func- The examples in Section 4 have demonstrated the accuracy and
tion of qRs1 =qRb , where the circles and lines represent the numer- efficacy of the SMFF method proposed in this paper for both regu-
ical and analytical solutions, respectively. For the rhombic lar and irregular tensegrity structures of large scale. In this section,
truncated tetrahedral, cubic/octahedral, and dodecahedral/icosa- we will further discuss some other features of this form-finding
hedral tensegrities, the numerical results of the SMFF method have method.
an excellent agreement with the analytical solution derived by Firstly, we test the convergence of the SMFF method to self-
Zhang et al. [18]. equilibrated solution starting from an arbitrary initial configura-
tion. By modifying the stiffness matrix to be positive definite by
4.5. Cubic grid tensegrities using Eq. (16) and introducing the total elastic strain energy as
the optimization objective function, the SMFF method can easily
To demonstrate the efficacy of SMFF method for large scale render a self-equilibrated configuration of tensegrity even when
structures, we further study cubic grid tensegrities in this subsec- the input configuration at the beginning of calculation is greatly
tion. Such tensegrity structures can be constructed by using the different from the stable state. For illustration, taking a 6-prismatic
46 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

Fig. 11. Three self-equilibrated irregular tensegrities determined by the SMFF method.

For the form-finding of all aforementioned tensegrities, the time

consumed by the SMFF method is about 95% faster than that of
the Monte Carlo method. The time saving is insensitive to the
structural irregularity but gradually decreases with the increase
of the total node number. This is because matrix operations are in-
volved in the SMFF method but not in the Monte Carlo form-find-
ing method. In addition, in the dynamic relaxation method, the
artificial mass and damping parameters need to be empirically
specified to accelerate the calculation [1,3]. In the SMFF method,
however, the artificial stiffnesses which ensure the positive defi-
Fig. 12. Form-finding of a 6-prismatic tensegrity. (a) An initial state which is niteness of the structural stiffness matrix can be easily determined
generated randomly, and (b) the final state obtained by the SMFF method.
in a programmed manner following Eq. (16). Thus the SMFF meth-
od is more efficient than the dynamic relaxation method.
Thirdly, the SMFF method can easily yield a stable configuration
tensegrity as an example, we randomly generate an initial config- of tensegrity. A tensegrity will be stable when the total potential
uration and then perform the form-finding calculation. The initial energy attains a local minimum value, ensuring that the structural
and the final configurations are shown in Fig. 12a and b, respec- stiffness matrix is positive definite after excluding rigid-body mo-
tively. It is found that the SMFF is always convergent and insensi- tions [14,17]. As described in Section 3, the SMFF method guaran-
tive to the input initial configurations. This is a distinct advantage tees the structural stability in a final state by introducing the
of the present method. artificial stiffness and optimizing the total potential energy of sys-
Secondly, we evaluate the efficiency of the SMFF method by tem. In this sense, the SMFF method is in the category of the well-
comparing its calculation speed with the Monte Carlo form-finding known energy optimization approach, as those in Refs. [31,38]. In
method proposed by Li et al. [3], which employs stochastic nodal comparison with the equilibrium methods (e.g. [18,39]), the SMFF
displacements to search the self-equilibrated and stable tensegrit- method can automatically exclude equilibrated but unstable states
ies. We compare the times needed by the two methods to complete and always leads to equilibrated and stable configurations. For
the form-finding analysis of all tensegrities in Section 4. Because illustration, an example is given in Fig. 13, which consists of 30
stochastic schemes are used in both the form-finding methods, bars, 90 strings, and 60 nodes. We proceed the form-finding anal-
we repeat each example for more than 20 times by each method ysis by choosing an equilibrated but unstable state shown in
and then average the consumed calculation times. These examples Fig. 13a, whose stiffness matrix has negative eigenvalues, as the
show that the SMFF method is much more efficient than the latter. initial configuration. Our procedure can still rapidly converge to a
 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48 47

Fig. 13. (a) An equilibrated but unstable planar configuration, which is specified as the initial state of the form-finding analysis, and (b) the obtained stable 30-prismatic
tensegrity.

Fig. 14. Another equilibrated solution obtained from the same topology of
elements as the Z-based truncated dodecahedral tensegrity in Fig. 7d.

stable configuration after only a few iteration steps, as shown in

Fig. 13b.
Finally, the SMFF method can also find different stable configu-
rations of the structures with the same topology of elements. The
existence of two or multiple stable states in a tensegrity is of inter- Fig. 15. Two other equilibrated configurations obtained from the same topology of
elements as the cubic grid tensegrity in Fig. 9c.
est for the design of, for instance, deployable/foldable robots and
sensor/actuator devices [3,45–47]. Two groups of such configura-
tions are illustrated as follows. By inputting the same file as a
In summary, the SMFF method has the following advantages:
Z-based truncated dodecahedral tensegrity shown in Fig. 7d, the
SMFF procedure sometimes produces another dome like structure,
(i) It is robust and highly efficient for both regular and irregular
as illustrated in Fig. 14, with a potential energy higher than that of
tensegrities of large scale.
the corresponding Z-based tensegrity. For the regular cubic grid
(ii) It is accurate and has good convergence performance regard-
tensegrity shown in Fig. 9c, two more configurations with less
less of the initial configuration.
symmetry are obtained from the form-finding process, as illus-
(iii) The structural configurations determined by the SMFF
trated in Fig. 15a and b. Among them, the structures in Figs. 9c
method are not only equilibrated but also stable.
and 15b possess the lowest and highest potential energies, respec-
(iv) This method can find multiple stable configurations with the
tively. The examination of the eigenvalues of their stiffness ma-
same topology.
trixes confirms that the structures in Figs. 14 and 15 are all stable.
48 L.-Y. Zhang et al. / Engineering Structures 58 (2014) 36–48

[15] Zhang JY, Guest SD, Connelly R, Ohsaki M. Dihedral ‘star’ tensegrity structures.
6. Conclusions Int J Solids Struct 2010;47:1–9.
[16] Zhang JY, Ohsaki M. Self-equilibrium and stability of regular truncated
A novel SMFF method has been developed for the form-finding tetrahedral tensegrity structures. J Mech Phys Solids 2012;60:1757–70.
[17] Zhang LY, Li Y, Cao YP, Feng XQ, Gao HJ. Self-equilibrium and super-stability of
of tensegrity structures. The structural stiffness matrix is truncated regular polyhedral tensegrity structures: a unified analytical
introduced to guide the calculation to rapidly converge to a self- solution. Proc R Soc A 2012;468:3323–47.
equilibrated and stable configuration, even when the initial config- [18] Zhang LY, Li Y, Cao YP, Feng XQ. A unified solution for self-equilibrium and
super-stability of rhombic truncated regular polyhedral tensegrities. Int J
uration assumed at the beginning of calculation is greatly different Solids Struct 2013;50:234–45.
from its finally obtained state. This method has been demonstrated [19] Feng XQ, Li Y, Cao YP, Yu SW, Gu YT. Design methods of rhombic tensegrity
by a number of examples to be highly efficient for various types of structures. Acta Mech Sin 2010;26:559–65.
[20] Li Y, Feng XQ, Cao YP, Gao HJ. Constructing tensegrity structures from one-bar
tensegrities. This work may help the design and construction of elementary cells. Proc R Soc A 2010;466:45–61.
various tensegrity structures for practical applications. [21] Pellegrino S. Mechanics of kinematically indeterminate structures. UK:
In the SMFF method, in order to avoid complicated matrix oper- University of Cambridge; 1986.
[22] Burkhardt R. The application of nonlinear programming to the design and
ations and to improve the computational efficiency, we employ the
validation of tensegrity structures with special attention to skew prisms. J Int
stochastic selecting algorithm to exclude rigid-body motions of a Assoc Shell Spatial Struct 2006;47:3–15.
configuration. This purpose can also be reached by using some [23] Koohestani K. Form-finding of tensegrity structures via genetic algorithm. Int J
Solids Struct 2012;49:739–47.
other methods, e.g. the eigenvalue analysis, singular value decom-
[24] Baudriller H, Maurin B, Canadas P, Montcourrier P, Parmeggiani A, Bettache N.
position, and generalized inverse approaches. Finally, it is noted Form-finding of complex tensegrity structures: application to cell cytoskeleton
that though we have only addressed the form-finding of self-equil- modelling. C R Mec 2006;334:662–8.
ibrated tensegrities in this paper, the SMFF method can also be [25] Zhang L, Maurin B, Motro R. Form-finding of nonregular tensegrity systems. J
Struct Eng ASCE 2006;132:1435–40.
used to investigate structures under specified external and internal [26] Ali NBH, Rhode-Barbarigos L, Smith IFC. Analysis of clustered tensegrity
loads and shape constraints as in Ref. [40]. structures using a modified dynamic relaxation algorithm. Int J Solids Struct
2011;48:637–47.
[27] Masic M, Skelton RE, Gill PE. Algebraic tensegrity form-finding. Int J Solids
Acknowledgments Struct 2005;42:4833–58.
[28] Estrada GG, Bungartz HJ, Mohrdieck C. Numerical form-finding of tensegrity
structures. Int J Solids Struct 2006;43:6855–68.
Supports from the National Natural Science Foundation of China [29] Zhang JY, Ohsaki M. Adaptive force density method for form-finding problem
(Grant Nos. 10972121 and 31270989), Tsinghua University of tensegrity structures. Int J Solids Struct 2006;43:5658–73.
[30] Tran HC, Lee J. Advanced form-finding of tensegrity structures. Comput Struct
(2009THZ02122 and 20121087991) and the 973 Program of MOST 2010;88:237–46.
(2010CB631005) are acknowledged. [31] Pagitz M, Tur JMM. Finite element based form-finding algorithm for tensegrity
structures. Int J Solids Struct 2009;46:3235–40.
[32] Gasparini D, Klinka KK, Arcaro VF. A finite element for form-finding and static
References analysis of tensegrity structures. J Mech Mater Struct 2011;6:1239–53.
[33] Rieffel J, Valero-Cuevas F, Lipson H. Automated discovery and optimization of
large irregular tensegrity structures. Comput Struct 2009;87:368–79.
[1] Veenendaal D, Block P. An overview and comparison of structural form finding
[34] Lobo D, Vico FJ. Evolutionary development of tensegrity structures. Biosystems
methods for general networks. Int J Solids Struct 2012;49:3741–53.
2010;101:167–76.
[2] Juan SH, Tur JMM. Tensegrity frameworks: static analysis review. Mech Mach
[35] Xu X, Luo YZ. Form-finding of nonregular tensegrities using a genetic
Theory 2008;43:859–81.
algorithm. Mech Res Commun 2010;37:85–91.
[3] Li Y, Feng XQ, Cao YP, Gao HJ. A Monte Carlo form-finding method for large
[36] Yamamoto M, Gan BS, Fujita K, Kurokawa J. A genetic algorithm based form-
scale regular and irregular tensegrity structures. Int J Solids Struct
finding for tensegrity structure. Proc Eng 2011;14:2949–56.
2010;47:1888–98.
[37] Micheletti A, Williams WO. A marching procedure for form-finding for
[4] Skelton RE, de Oliveira MC. Tensegrity systems. Dordrecht: Springer; 2009.
tensegrity structures. J Mech Mater Struct 2007;2:857–82.
[5] Sultan C. Tensegrity: 60 years of art, science, and engineering. In: Aref H, van
[38] Klinka KK, Arcaro VF, Gasparini D. Form finding of tensegrity structures using
der Giessen E, editors. Advances in applied mechanics. San Diego: Elsevier
finite elements and mathematical programming. J Mech Mater Struct
Academic Press Inc.; 2009. p. 69–145.
2012;7:899–907.
[6] Rhode-Barbarigos L, Ali NBH, Motro R, Smith IFC. Designing tensegrity modules
[39] Koohestani K, Guest SD. A new approach to the analytical and numerical form-
for pedestrian bridges. Eng Struct 2010;32:1158–67.
finding of tensegrity structures. Int J Solids Struct 2013;50:2995–3007.
[7] Tibert AG, Pellegrino S. Review of form-finding methods for tensegrity
[40] Zhang LY, Li Y, Cao YP, Feng XQ, Gao HJ. A numerical method for simulating
structures. Int J Space Struct 2003;18:209–23.
nonlinear mechanical responses of tensegrity structures under large
[8] Connelly R, Terrell M. Globally rigid symmetric tensegrities. Struct Topol
deformations. J Appl Mech Trans ASME 2013;80:061018.
1995;21:59–79.
[41] Guest SD. The stiffness of tensegrity structures. IMA J Appl Math 2011;76:
[9] Murakami H. Static and dynamic analyses of tensegrity structures. Part II.
57–66.
Quasi-static analysis. Int J Solids Struct 2001;38:3615–29.
[42] Guest SD. The stiffness of prestressed frameworks: a unifying approach. Int J
[10] Crane CD, Duffy J, Correa JC. Static analysis of tensegrity structures. J Mech Des
Solids Struct 2006;43:842–54.
2005;127:257–68.
[43] Tran HC, Lee J. Geometric and material nonlinear analysis of tensegrity
[11] Murakami H, Nishimura Y. Static and dynamic characterization of regular
structures. Acta Mech Sin 2011;27:938–49.
truncated icosahedral and dodecahedral tensegrity modules. Int J Solids Struct
[44] Smyth GK. Optimization and nonlinear equations. In: Armitage P, Colton T,
2001;38:9359–81.
editors. Encyclopedia of biostatistics. London, UK: John Wiley & Sons, Ltd.;
[12] Murakami H, Nishimura Y. Infinitesimal mechanism modes of tensegrity
2005. p. 3857–63.
modules. In: Watanabe K, Ziegler F, editors. IUTAM symposium on dynamics of
[45] Korkmaz S, Ali NBH, Smith IFC. Determining control strategies for damage
advanced materials and smart structures. Dordrecht, The
tolerance of an active tensegrity structure. Eng Struct 2011;33:1930–9.
Netherlands: Springer; 2003. p. 273–84.
[46] Xu X, Luo YZ. Multistable tensegrity structures. J Struct Eng ASCE 2011;137:
[13] Pandia Raj R, Guest SD. Using symmetry for tensegrity form-finding. J Int Assoc
117–23.
Shell Spatial Struct 2006;47:245–52.
[47] Benjeddou O, Limam O, Ben Ouezdou M. Experimental and theoretical study of
[14] Zhang JY, Guest SD, Ohsaki M. Symmetric prismatic tensegrity structures: Part
a foldable composite beam. Eng Struct 2012;44:312–21.
I. configuration and stability. Int J Solids Struct 2009;46:1–14.

View publication stats