Вы находитесь на странице: 1из 16

Tensegrity Structures

Prestressability
Investigation
Cornel Sultan* and Robert Skelton**
*Harvard University, Boston, MA 02115, U.S.A.
**University of California, La Jolla, CA 92093-0411, U.S.A.
(Received 10th October 2001, revised version received 12th March 2002)
ABSTRACT: A methodology for the investigation of tensegrity structures
prestressability is presented. The methodology is aimed at identifying
particular prestressable configurations for which the prestressability conditions
can be expressed in a simple form. Examples of the application of the
methodology to a class of tensegrity structures, called tensegrity towers, are
presented.

INTRODUCTION
Tensegrity structures represent a class of space
structures composed of a set of soft members and a set
of hard ones. The difference between the soft and hard
members is that the soft ones must carry tension forces
in order to avoid entanglement (the best example
being that of elastic tendons or membranes). These
members cannot carry compressive forces (for
example a tendon cannot be compressed). Because of
this property we shall refer to these members as tensile
members. On the other hand the hard members are
characterised by the fact that they can carry any type
of force or moment. The representative example is that
of bars, which can carry tension, compression forces,
bending moments etc.
A structure composed of soft and hard elements as
described above is a tensegrity structure if it has the
property of prestressability. This consists of the
structures ability to maintain an equilibrium shape
with all tensile members in tension and in the absence
of external forces or torques. Tensegrity structures
integrity is guaranteed by the tensile members in
tension, hence their denomination, tensegrity, an
acronym of tension-integrity coined by R. B. Fuller. A
perspective view of a tensegrity tower, composed of
33 elastic tendons (soft members), 9 rigid bars, a rigid
base, and a rigid top (hard members) is given in Fig. 1.

International Journal of Space Structures Vol. 18 No. 1 2003

Figure 1.Three stage tensegrity tower

From the technological point of view, tensegrity


structures offer several important advantages. First, as
mentioned by Emmerich (1996), many tensegrity
structures do not have bar to bar connections. Thus
complicated joints are eliminated. Second, tensegrity

15

Tensegritty Structures Prestressability Investigation

structures offer excellent opportunities for physically


integrated structure and controller design since their
members can serve simultaneously as sensors,
actuators, and load carrying elements. Having
incorporated sensors and actuators, tensegrity
structures have considerable promise as smart
structures (see Skelton, 1997, Sultan, 1997, 1998).
Third, most tensegrity structures are lightweight
structures, making them amenable for various space
applications such as deployable structures (see Sultan,
1998), space telescopes (see Sultan, 1999), robotic
arms etc. Fourth, tensegrity structures are capable of
large displacement, belonging to the class of flexible
structures.
As
mentioned
before,
the fundamental
characteristic of tensegrity structures is their
prestressability, defined as the ability to yield
equilibrium configurations under no external forces or
torques and such that all tensile members are in
tension. The corresponding conditions are called
prestressability conditions, and the equilibrium
configurations prestressable configurations. A
derivation of the prestressability conditions for certain
types of tensegrity structures similar to the one in Fig.
1, under very general modelling assumptions, has
been presented in Sultan (2001). Numerical methods
to investigate the prestressability problem have been
proposed over the years. Research of the prestressable
configurations has been developed simultaneously
with infinitesimal mechanisms determination (see
Pellegrino, 1986, 1990, whose work is based on
equilibrium matrix analysis). Vassart (1999)
developed a method allowing a multiparametered
form-finding. The method exploits the force density
method, and its application to tensegrity structures is
shown to provide the designer with an efficient way to
achieve interesting new prestressable configurations.
The dynamic relaxation method has also been applied
successfully to form-finding problems of tensegrity
prisms (see Motro, 1994). Hanaor (1988) developed
an algorithm based on the flexibility method of
structural analysis for the analysis and optimum
prestress design of prestressable pin-jointed
assemblies, including tensegrity structures. Hanaor
(1992) used the stiffness method in obtaining
prestressable configurations of double layer tensegrity
domes. The previously mentioned approaches to
prestressability problem solution are numerical ones,
and, as with all numerical methods, they might suffer
from computational inneficiency. On the analytical
solutions front, Tarnai (1980), extending some pin-

16

jointed space trusses results, presented geometries for


which the equilibrium matrix is singular (which, in
this case of square equilibrium matrix, is a necessary
but not sufficient condition for prestressable
configurations to occur). Kenner (1976) had
previously arrived at the same results for tensegrity
prisms based on the assumption that in a prestressable
configuration cable lengths are a minimum. Sultan
(2001) reported the discovery of classes of
prestressable configurations for several tensegrity
structures for which complete analytical solutions of
the prestressability problem are given.
In this paper we present a methodology for the
investigation of the prestressability conditions. The
methodology, which combines symbolic and
numerical computation, is aimed at identifying
particular prestressable configurations for which the
prestressability conditions are expressed in a simpler
form. This simpler form allows for further analytical
investigations or efficient and accurate numerical
solutions. Hence the advantage of the method is
computational efficiency in configurations possessing
high regularity. Examples of the application of the
proposed methodology are included.

2 PRESTRESSABILITY
2.1 Modelling A ssumptions
Consider a tensegrity structure composed of E elastic
tendons (the tensile members) and R rigid bodies (the
hard members). We assume that all the joints of the
structure are affected, at most, by kinetic friction. This
means that the friction forces/torques acting at a joint
are zero if the relative linear/angular velocities
between the elements in contact at the joint are zero.
Also, the tendons are affected, at most, by kinetic
damping. This means that the damping force
introduced by a tendon is zero if its elongation rate is
zero. Hence the system is not affected by static
(Coulomb) friction. We also assume that all
constraints on the system are holonomic. This implies
that we can find a set of independent generalised
coordinates, qi, i = 1, . . ., N, to describe the
mechanical motion of the system. Additionally, we
assume that all constraints are scleronomic and
bilateral. In other words, they are not time dependent
and they are not mathematically expressed as
inequalities. Lastly, we assume that the external
constraint forces are workless, which means that they
do no work in a virtual displacement consistent with
the geometric constraints. We neglect the forces

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

exerted on the structure by external force fields (e.g.


gravitational).

2.2 Prestressability Conditions


Sultan (2001) proved that under these general
modelling assumptions the prestressability conditions
can be expressed as a set of nonlinear equations and
inequalities:
A(q)T = 0, Ti > 0, i = 1,2,..., E

(1)

where the elements of the matrix A(q), called the


equilibrium matrix, are given by
Aij(q) =

Lj
i = 1,2,...,N, j = 1,2,...,E
qi

(2)

and T is the vector of tendon tensions. Here Lj is the


length of tendon j, q is a vector of independent
generalised coordinates used to describe the
configuration of the system, and N is the number of
independent generalised coordinates.

3 METHODOLOGY
In the following we present a methodology which
allows for the identification of particular prestressable
configurations and of the corresponding, simpler,
prestressability
conditions.
The
proposed
methodology consists of several steps as follows.
1.

Identify an appropriate set of independent


generalised coordinates qi, i = 1, 2, . . ., N.

2.

Symbolically compute the equilibrium matrix


A(q), Aij(q) =

3.

Lj
= 1,2,...,N, j = 1,2,...,E.
qi

Define a class of configurations of interest and


parameterize its geometry using fewer
This
independent coordinates, pi, i = 1,2, . . ., N.
results in q = q (p).

4.

Substitute q= q (p) in A(q), resulting in (p) =


A(q(p)).

5.

Numerically solve the prestressability conditions


(p)T = 0 for one solution, (p p, Tp).

6.

Based on the previous numerical solution,


identify the structure of T and symbolically
substitute T into (p)T = 0. This is done as
follows: T p has a certain structure (for

International Journal of Space Structures Vol. 18 No. 1 2003

example Tp1 = Tp3 = ...). Then the vector T is


assumed to have the corresponding structure,
T = [Ta Tb T a . . .]T, and this expression of T is
symbolically substituted into (p)T = 0. Next,
eliminate the redundant equations in (p)T = 0.
The resulting set of equations and inequalities on
the tensions (e.g. T a > 0, Tb > 0 etc.) will be
called the essential prestressability conditions.
Steps 2, 4, and 6 can be carried out using symbolic
computational software (Maple, Mathematica, etc.).
The structure of T is indicated just by one numerical
solution. The essential prestressability conditions
characte rize all the elements of the class of
prestressable configurations of interest for which the
structure of T is the same as the one indicated by the
numerical solution. If T p does not have any particular
structure then no simplification of the prestressability
conditions is possible.
In order for T to have a special structure it is
recommended that the prestressable configurations of
interest have some geometrical regularity because it is
expected that this regularity would result in some
symmetry in the vector of tendon tensions, T.
The main advantage this methodology presents
over the classical numerical methods is that families
of prestressable configurations can be identified for
which the prestressability conditions can be more
easily solved numerically or analytically. Hence, not a
single solution of the prestressability conditions is
generated but a whole set, parameterized by certain
geometrical characteristics of the structure under
investigation.

4 TENSEGRITY TOWERS
In the following we illustrate the above methodology
through its application to a class of tensegrity
structures called tensegrity towers.

4.1 Description
A tensegrity tower is composed of n stages (where n 3),
each stage having 3 bars. The endpoints of a bar are
labeled Aij and Bij, i =1,2,3, j =1,2,...,n, where the A
point denotes the bottom of the bar and the B point its
top. Bar Aij Bij will be referred to as bar ij. The indices
of bar ij respect the following rule: the second index,
j, denotes the number of the stage the bar belongs to,
while the first index, i, indicates the number of the bar
in a stage. Stage number j is composed of bars
ij, i =1,2,3. The bars of the first stage are attached via

17

Tensegritty Structures Prestressability Investigation

ball and socket joints to a base, and the bars of the


n-th stage are attached via ball and socket joints to a
top. We will refer to the base by the points of the first
stage bars which are attached to it, that is A1l A2l A3l.
Similarly, the top will be referred to as Bln B2n B3n. The
end points of the bars are connected through a total of
15n12 tendons. These are defined as follows.

Tendons associated with the first stage: A11A12,


A11B21, A21A22, A21B31, A31A32, A31B11

Tendons associated with stage j, j even, j < n:


A1jB3j, A1jA3j+1, A2jB1j, A2jA1j+1, A3j B2j, A3j A2j+1,
B1j-1B3j, B2j-1B1j, B3j-1B2j.

Tendons associated with stage j, j odd, 1 < j < n:


A3jB1j, B1j-1B1j, A1jB2j, B2j-1B2j, A2jB3j, B3j-1B3j,
A1jA1j+1, A2jA2j+1, A3jA3j+1.

Tendons associated with the n-th stage, n 4


even: B2n-1B1n, A2nB1n, B3n-1B2n , A3nB2n , B1n-1B3n,
A1nB3n.

Tendons associated with the n-th stage, n 3


odd: B1n-1B1n, A3nB1n, B2n-1 B2n, A1nB2n, B3n-1B3n,
A2nB3n.

Tendons separating stages j and j + 1 (where 1


j n-1): B1jA1j+1, B2jA1j+1, B2jA2j+1, B3jA2j+1,
B3jA3j+1, B1jA3j+1.

The spatial polygon composed of the 6 tendons


separating stages j and j + 1 will be referred to as the
j-th saddle.
We ascertain that there are three different types of
tendons: the ones labeled A*B * and called vertical
tendons, the ones labeled A*A* and B *B*, called
diagonal tendons, and the ones labeled B*A* and
called saddle tendons.
We introduce the vector of tendon lengths L bj
associated with stage j; its components are the lengths
of the tendons associated with stage j, j = 1, 2, . . .,n, in
the order they are listed above. We also introduce the
vector Ls of tendon lengths associated with the j-th
j
saddle, j = 1, 2, . . . ,n-1, defined in the same manner
as L b . The vector L of tendon lengths is defined by
j
assembling these vectors as follows:
L = [L Tb LTs L Tb ... L Ts
1

n-1

LTb ]T.
n

T = [T Tb TTs T Tb ... TTs


1

n-1

TTb ]T.
n

(4)

For further mathematical modelling and analysis


we assume that the tendons are massless and linear
elastic, the base and the top are rigid, the bars are
rigid, axially symmetric, and for each bar the
rotational degree of freedom around the longitudinal
axis of symmetry is neglected. The base is assumed to
be inertially fixed. We neglect the forces exerted upon
the structure by external force fields (e.g.
gravitational). The system is affected, at most, by
kinetic friction/damping.
We introduce an inertial frame of reference, b 1, b 2,
b 3, as a dextral (right-handed) set of unit vectors, whose
origin coincides with the geometric center of the
triangle A11A21A31. Axis b 3 is orthogonal to A11A21A31
pointing upward while b 1 is parallel to A11A31, pointing
towards A31. We introduce another reference frame, t1,
t2, t3, called the top reference frame, which is fixed in
the top rigid body. Its origin coincides with the
geometric center of the top triangle, B1nB2nB3n, t3 is
orthogonal to B1nB2nB3n and points upward, while t1 is
parallel to B1nB3n, pointing towards B3n.
In the following we shall relate the analysis to the
methodology previously described.

4.2 Step 1: Generalised Coordinates


The independent generalised coordinates necessary to
describe the configuration of this system are:
d ij, a ij, the declination and the azimuth of bar ij,

i = 1,2,3, j = 1,2, . . .,n, defined as follows: d ij is

the angle between Aij Bij and b 3 and a ij is the

angle between b 1 and the projection of Aij Bij


onto plane (b 1, b 2) (see Fig. 2).

(3)

Analogously we introduce the tension vector T b


j
associated with the j-th stage, j = 1, 2, . . .,n, and the

18

tension vector Tsj associated with the j-th saddle, j = 1,


2, . . .,n-1. The vector of all tensions is:

Figure 2.declination and azimuth of bar ij

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

x ij,yij,zij, the inertial Cartesian coordinates of the


mass center of bar ij, i = 1,2,3, j = 2,3, . . . ,n-1.

c ,f ,u , the Euler angles for a 3-1-2 sequence to


characterize the orientation of the top reference
frame with respect to the inertial frame.

X, Y, Z, the inertial Cartesian coordinates of the


geometric center of BlnB2nB3n.
We introduce the following vectors:
q~ 1 = [d

q~ j = [d

1j

11

11

1j x 1j y 1j z1j

2j

21

21

31

2j x2j y2j z2j

3j

3j x3j y3j z3j]

1n

1n

2n

2n

3n

3n]

, q~ T = [c f

q = [q~ T1 q~ T2 ... q~ Tn q~ TT ]T.

(6)

The number of independent generalised coordinates


is equal to the number of tendons: E = N = 15n-12.

4 .3 Step 2: Geomet ry and Equilib rium


Matrix
The inertial Cartesian coordinates of the nodal
points can be determined as follows.

i1)

cos(a

yBi1 = yAi1 + li1 sin(d

i1)

sin(a

xAij = xij sin(a

sin( d ij) cos(a

ij), zAij= zij-

xBij = xij +
sin(a

lij

lij
2

lij
2

lij
2

zBi1 = li1 cos(d

yAij = yij -

lij

i1),

ij),

(7)

sin(d ij)

cos(d ij),

sin(d ij) cos(a

ij), zBij= zij+

ij),

in)

Z - lin cos (d

in)

in)

in

(10)

[ ][] [ ]

xB
X
in
RBin = yBin = Y
zB
Z

+ C312

in

XBin
YB
in
0

(11)

Here XBin and YBin are the Cartesian coordinates of


Bin with respect to the top reference frame and

C312 =

c(c ) c(u ) - s(c ) s(f ) s(u )


s(c ) c(u ) + c(c ) s(f ) s(u )
- c(f ) s(u )

c(c ) s(u ) + s(c ) s(f ) c(u )


s(c ) s(u ) - c(c ) s(f ) c(u )
c(f ) c(u )

- s(c ) c(f )
c(c ) c(f )
s(f )

(12)

where c(f ) = cos(f ), s(f ) = sin(f ).

i1),
i1),

XB

u X Y Z] T.

Here q~ j , j = 1,2, . . .,n, is the vector of independent


generalised coordinates associated with the j-th stage
and q~ T is associated with the rigid top. The vector of
independent generalised coordinates is

xBi1 = xAi1 + li1 sin(d

X - lin sin (d in) cos (a

= Y - lin sin (d in) sin (a

+ C312 YBin

(5)

j = 2,3,...,n-1
q~ n = [d

[ ][
[ ]

xAin
RAin = yAin
zA
in

T
31 ] ,

where i = 1,2,3, j = 2,3,...,n-1. Here xAij, yAij, zAij


i = 1,2,3, j = 1,2,...,n-1, are the inertial coordinates of
Aij, xBij, yBij, zBij, i = 1,2,3, j = 1,2,...,n-1, are the inertial
coordinates of Bij, and lij, i = 1,2,3, j = 1,2, . . ., n-1, is
the length of bar ij. The inertial Cartesian coordinates
of the nodal points of the n-th stage, Ain and Bin,
i = 1,2,3, are expressed using the transformation
matrix for a 3-1-2 sequence, C312 :

elements of the equilibrium matrix, Aji =


(8)

yBij = yij +

lij
2

Once the nodal points coordinates are expressed in


terms of the independent generalised coordinates,
symbolic computation can be used to determine the
lengths of the tendons Li(q), i = 1, 2, . . .,N, and the
Li
i = 1, 2,
qj

. . .,N, j = 1,2,...,N.

4.4 Step 3: Particular Configurations


sin(d ij)

cos(d ij),

International Journal of Space Structures Vol. 18 No. 1 2003

(9)

We shall now define a set of particular


configurations called the class of cylindrical
symmetrical configurations. This set is defined as
follows.

19

Tensegritty Structures Prestressability Investigation

Figure 3. Top view of a tensegrity tower in a cylindrical symmetrical configuration

Triangles A 11 A 21 A 31 and B 1n B 2n B 3n are


congruent, equilateral triangles of side length b.

Planes A1j A2j A3j and A1j+1 A2j+1 A3j+1, j = 1,2,...,


n -1, are parallel.

All bars have equal length, l.

All bars have the same declination, d .

Bars 11 and [i]i, i = 2,3,...,n, are parallel; bars 21


and [1 + i]i, i = 2,3,...,n, are parallel; bars 31 and
[2 + i]i, i = 2,3, . . ., n, are parallel. In these
formulas [x] is a periodic function such that [x] =
[x-3] if x 4 and [x] = x if x = 1, 2, 3; hence bar
[4]4 is actually bar 14.

The distance between A1j+1 A2j+1 A3j+1 and B1j B2j


B3j is the same for all j = 1,2, . . ., n - 1, and it is
called the overlap, h, being positive if B1j B2j B3j
is closer to A11 A21 A31 than A1j+1 A2j+1 A3j+1.

20

All nodal points, Aij, Bij, i = 1, 2, 3, j = 1, 2, . . .,n,


lie on the surface of a rectangular cylinder.
The projections onto the base of the j-th saddle
points, A3j+1, B1j, B3j, A2j+1, B2j, A1j+1, make a
regular hexagon.

A top view of such a configuration is given in Fig. 3.


The class of cylindrical symmetrical configurations
can be parameterized using two independent
parameters, a and h. The corresponding values of the
independent generalised coordinates are:
d

ij

= d , where i = 1,2,3, j = 1,2,...,n,

zij =

2j - 1
l cos(d ) - (j-1)h,
2

where i = 1,2,3, j = 2,3,...,n-1,

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

[i]i

=a ,a

[i+1]i

=a +

4p
3

,a

[i+2]i

=a +

2p
,
3

5p
3

where k is a natural number. Here a and d are


related by the constraint that all nodal points lie on the
surface of a cylinder:

where i = 1,2,...,n,
x[2p-1]2p-1 = -

b
l
+
2
2

sin(d ) cos(a ), y[2p-1]2p-1

3
l
=b+
sin(d ) sin(a ),
6
2
l
x[2p]2p-1 =
sin(d ) cos(a
2

21 ),

3
l
b+
sin(d ) sin(a
3
2

x[2p+1]2p-1 =

b
l
+
2
2

21 ),

3
l
b+
sin(d ) sin(a
6
2

=-

y[2p]2p-1

31 ),

b - l
2
2

x[2p]2p =
=

b - l
2
2

3
l
bsin(d ) sin(a
6
2

x[2p+1]2p = -

l
2

12 ),

y[2p-1]2p

22 ),

Cylindrical symmetrical configurations which are


prestressable configurations will be called
symmetrical cylindrical prestressable configurations
(SCPC). In order to analyze them we substitute (13)
into the prestressability equations A(q)T = 0 yielding

y[2p]2p

32 ), y [2p+1]2p

(16)

Since ( a ,h) is a square matrix, a necessary


condition for prestressability is

32 ),

det((a , h)) = 0.

(17)

So far the first four steps of the methodology were


applied. In order to go to the next step, involving
numerical computation of one solution of the
prestressability conditions, we need to fix the number
of stages.
Consider first the three stage tensegrity tower (n = 3)
in Fig. 1 for which E = N = 33.

= u = X = Y = 0, Z = nl cos( d ) - (n-1)h,
2a if n = 3 + 6k

c =

(15)

4.5 Step 4 : Prestressability Conditions for


Particular Configurations

(a , h)T = 0.

sin(d ) cos( a

respectively.

where p = 1,2,..., and 2p < n,


f

22 ),

3
l
=bsin(d ) sin(a
3
2

b +h ,
WWWW
3

WWWWWW
2 WW WWWW WWW
D = l2 + h2 - 2 b cos (2a ) - 2lh cos (d ) ,
3

12 ),

sin(d ) cos( a

(14)

In this geometry all tendons of a given type (saddle,


vertical, and diagonal) have the same lengths given by

31 ),

sin(d ) cos(a

3
l
=
bsin(d ) sin(a
6
2

p
2b sin(a + _3)
.
W
3l

WWWWWWWW
WW
b2WW
V = l2 - b2 cos(2 a ) - W
sin (2 a ) ,
3

y[2p+1]2p-1

where p = 2,3,..., and 2p - 1 < n


x[2p-1]2p = -

sin(d ) =

S=

sin(d ) cos( a

if n = 8 + 6k

if n = 4+ 6k
2a p

2p
3

if n = 5 + 6k
(13)

if n = 6 + 6k

2a +

2p
3

if n = 7 + 6k

International Journal of Space Structures Vol. 18 No. 1 2003

4.6 T hree Stage Towers; Step 5: Numerical


Solution

A numerical solution of ( a ,h)T = 0, T i > 0,


i = 1, 2,...,33, is found as follows. First we fix
l = 0.4 m, b = 0.27 m, a = 5 deg

(18)

21

Tensegritty Structures Prestressability Investigation

and numerically solve det((h,a )) = 0 for h using a


line search combined with a Newton-Raphson
p
method. This results in h = 0.0875 m. We next
p
p
determine the kernel of (h , a ). Every vector T in
this kernel can be expressed as:
p

T = P[1 0.74 1 0.74 1 0.74 1.37 1.11 1.37 1.11

investigations and can be very easily numerically


solved allowing for the following facts to be
established.

The
essential
prestressability
equation
(det(A e) = 0) yields at most a cubic equation in h.
For a = 0 this equation reduces to a linear one,
whose solution is

1.37 1.11 0.96 0.64 0.96 0.64 0.96 0.64 0.64


0.64 0.64 1.37 1.11 1.37 1.11 1.37 1.11 1 0.74

(19)

1 0.74 1 0.74]T
p

where P is an arbitrary scalar (the rank of (h , a )


p
is 32). Clearly for P > 0 all elements of T are positive.

4 .7 Step 6: Essent ial Prest ressabilit y


Conditions
Based on this solution we assume the following
structure for T:
T = [TD T V TD TV T D T V TS TS TS TS TS TS
1

TV TD T V T D T V T D TD TD T D
2

Tri > 0,

i = 1,2, . . .,6

Tr = [T S TS TV T V T D T D ]T.
1

(22)

The last equation in AeTr = 0 is


(TS + TS )
1

h
h - l cos( d )
+ (T D + T D )
= 0 (23)
1
2
S
D

which, since all tensions must be strictly positive,


results in
0 < h < , cos(d ).

(24)

Thus the overlap must be positive and less than the


height of a stage.
The much simpler essential prestressability
conditions are further amenable to analytic

22

(25)

Numerical experiments showed that at a


symmetrical
cylindrical
prestressable
configuration for three stage towers the rank of
Ae is always 5, thus the general solution for
tensions is:

where Ae is a function of h and a , whose nonzero


elements are given in the Appendix, and

Tr = PT = P[TS

(21)

WWW
l -b

Numerical experiments showed that, for fixed a ,


the essential prestressability equation can yield
several real solutions for the overlap h, but only
one might satisfy the inequalities on the tensions
(T ri > 0, i = 1,. . .,6).

Substitution of (13), (14), and (20) into (1) leads to


a system of 33 equations and 6 inequalities. Of the 33
equations only 6 are independent, the ones associated
with the rows of A whose numbers are 1, 2, 7, 9, 28,
and 33 (see Sultan, 1999, for details). They yield the
essential prestressability conditions:

5
12

(20)

TS1 TS1 T S1 TS1 T S1 TS 1 T D1 TV1 TD1 T V1 T D1 TV1]T.

A e Tr = 0

h=

10

TS

10

TV

10

TV

20

TD

10

T D ]T
20

(26)

where P is an arbitrary positive scalar (the


pretension coefficient) and T0 is a base of Ker(A e).

Equilibrium equations of the nodal points yield


the forces in the bars. All bars of the first and
third stage experience the same forces, equal to
C r , while the forces in all bars of the second
1
stage are equal to C r2. These forces can be
expressed as C ri = PC 0i, i= 1,2.

We give in Figs. 4-6 the solution of the essential


prestressability conditions plotting the overlap as a
function of a , the variation of the corresponding
height of the structure, the declination of the bars, and
the basis tensions (T 0) and compressive forces (C 0) for
a three stage tensegrity tower with l = 0.4 m and
b = 0.27 m (in these Figures SCPC stands for
symmetrical cylindrical prestressable configurations).
We note that T 0 has been normalized such that Ts
Euclidean norm is 1.
Finally we note that numerical experiments with the
prestressability conditions, A(a , h)T = 0, Ti > 0,
i = 1,...,33, did not identify other particular structures
for the vector of tensions, T.

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

Figure 4. Prestressable configuration parameters for a 3 stage tensegrity tower

4 .8 Four Stage Towers; Step 5: Numerical Solution


As a second example we consider a four stage tensegrity tower. In this case the number of tendons (and
independent generalised coordinates) is 48. For the numerical investigation step we assign l, b, and a the same
values as in (18) and solve the prestressability conditions for hP and TP. We identify the following solution:
hP = 0.08038 m,
T P= P [1 0.69 1 0.69 1 0.69 1.41 1.07 1.41
1.07 1.41 1.07 0.97 0.88 0.97 0.88 0.97
0.88 0.52 0.52 0.52 1.45 1.45 1.45 1.45
1.45 1.45 0.97 0.88 0.97 0.88 0.97 0.88
0.52 0.52 0.52 1.07 1.41 1.07 1.41 1.07
1.41 1 0.69 1 0.69 1 0.69] T.

(27)

where P > 0 is an arbitrary scalar.

International Journal of Space Structures Vol. 18 No. 1 2003

23

Tensegritty Structures Prestressability Investigation

Figure 5. Basis tensions for a 3 stage tensegrity tower

Figure 6. Basis compressions for a 3 stage tensegrity tower

24

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

the essential prestressability equation can yield


several real solutions, but only one might satisfy
the inequalities on the tensions (Tri > 0, i = 1,...,8).

4 .9 Step 6: Essential Prest ressabilit y


Conditions
Based on the numerical solution the following
structure of T is assumed:
T = [TD1 TV1 TD1 TV1 TD1 TV1 TS1 TS1 TS1 TS1 TS1 TS1

TV2 TD2 TV 2 TD2 TV2 TD2 TD2 TD2 TD2


2

Tr = PT0

(28)

TS TS TS TS TS TS
2

TS1 TS1 TS1 TS1 TS1 TS1 TD1 TV1 TD1 TV1 TD1 TV1]T.

Proceeding in the same way as for the three stage


tensegrity towers, we obtain a system of 48 equations,
of which only 8 are independent (the ones associated
with rows number 1, 2, 7, 8, 9, 10, 11, 48), and 8
inequalities on the tendon tensions (see Sultan, 1999,
for details). These are the essential prestressability
conditions, expressed as
AeTr = 0,

T ri > 0 for i = 1,..,8

(29)

where the structure of Ae is

Ae =

*
*
*
*
*
*
*
*

*
*
*
*
*
*
*
*

0
0
*
*
*
*
*
0

*
*
0
0
0
0
0
0

0
0
*
*
*
*
0
0

0
0
*
*
*
*
*
*

0
0
*
*
*
*
*
0

*
*
*
*
*
*
*
*

Figs. 7-9 show the solution of the essential


prestressability conditions (the overlap), the
declination of bars, height of the structure as well as
the variation of the basis tensions T0 and compressions
C0 with a . As for the three stage tower, T0 has been
normalized such that Ts Euclidean norm is 1.
Numerical experiments with the prestressability
conditions, A(a ,h)T = 0, T i > 0, i = 1,...,48, did not
identify other structures for the vector of tensions, T,
except for the one indicated by (28).
The methodology has been successfully applied to
tensegrity towers with 3 to 10 stages. The same
class of symmetrical cylindrical prestressable
configurations has been investigated. Considerable
reduction in the size of the prestressability conditions
was obtained.
Our numerical experiments allowed for some
important conclusions as follows.

The last line of Ae yields the condition that the


overlap must be positive and less than the height
of a stage:
0 < h < l cos(d ).

The forces in all bars of the first and fourth stage


are equal to C r while the forces in the second
1
and third stage bars are equal to C r and can be
2
expressed as C ri = PC 0i, i = 1,2.

5 GENERALIZATION

The interested reader is refered to Sultan


(1999) for the nonzero elements of Ae.
Further investigations led to the following
conclusions.

(33)

where P is an arbitrary positive scalar (the


pretension coefficient) and T 0 is a base of
Ker(Ae).

TV TD TV TD TV TD TD TD TD
2

Numerical experiments showed that at a


symmetrical
cylindrical
prestressable
configuration of four stage towers the rank of Ae is
always 7, thus the general solution for tensions is:

(31)

The
essential
prestressability
equation
(det(A e) = 0) yields at most a polynomial
equation of degree 4 in h. For a = 0 this equation
reduces to a quadratic one:
2
12h215h lWW
- b2 + 4(, 2 - b2) = 0.
(32)

For fixed ,, b, and a only one solution for h of


the general prestressability conditions has been
identified. In general, the corresponding overlap,
h, decreases with the number of stages, as shown
in the next table. The data in this table were
obtained using
l = 0.4 m, b = 0.27 m, a = 5 deg.

(34)

Table 1: Overlap (h) vs Number of Stages (n)


n

h (mm) 87 80 77 75 73 72 71

Numerical experiments showed that, for fixed a ,

International Journal of Space Structures Vol. 18 No. 1 2003

25

Tensegritty Structures Prestressability Investigation

Figure 7. Prestressable configurations parameters for a 4 stage tensegrity tower

At a solution of the prestressability conditions


the rank of A is N - 1, thus T = PT0 with P > 0
arbitrary (here T 0i > 0, i = 1,2,...,N).

The tension vector, T, has a structure which


implies symmetry of tensions with respect to the
middle stage(s). Mathematically this is expressed
as follows:

1. If n is odd, n = 2p + 1, Ts structure respects the


following rules:
(a)

The first and the last (n-th) stage tension vectors are
T b1 = Tbn = [TD1 T V1 TD1 T V1 TD1 T V1]T.

(b) The middle stage, m = p + 1, tension vector is


T b = [TV TD T V T D TV T D TD T D TD ] T.
m

26

(c)

The tension vectors of the i-th stage, i = 2, 3,...,


n-1, i m, is
Tb = [T V TD TV TD T V T D T D TD TD ] T
i

and the tension vector of stage number n + 1 - i is


Tbn + 1-i = [TVi TD i T Vi T Di T Vi TDi TDi TDi TDi] T
(d) The tension vector of the i-th saddle,
i = 1,2,..., n-1, is equal to the tension vector of the
n-i-th saddle:
Ts = Ts
i

n-i

= [TS T S T S TS TS TS ] T .
i

2. If n is even, n = 2p:
(a) The first and last stages tension vectors are
Tb = T b = [T D T V T D T V T D T V ]T.
1

(b) The tension vectors of all other stages have the


following structure:
International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

Figure 8. Basis tensions for a 4 stage tensegrity tower

Figure 9. Basis compressions for a 4 stage tensegrity tower

International Journal of Space Structures Vol. 18 No. 1 2003

27

Tensegritty Structures Prestressability Investigation

T b i = [T Vi T D i T Vi T D i T Vi T D i T D i T D i T D i] T ,
i = 2,3,..., n-1.
(c)

The tension vector of the middle saddle, m = p, is


T sm = [TSm T Sm TSm T Sm TSm T Sm]T.

(d) The tension vector of the i-th saddle, i = 1,2,...,n-1,


i m, is
Tsi = [T Si TSi TSi TSi TSi TSi]

and the tension vector of the n-i-th saddle is:


T s = [T S T S TS T S TS TS ]T.
i

After identification of the structure of the tension


vector, T, and symbolical substitution of the
expressions for the generalised coordinates and for T
in the general prestressability conditions, we
ascertained the following facts:

The corresponding essential prestressability


conditions result in a polynomial equation in h
and inequalities on the tendon tensions. There are
5n-3
2

inequalities if n is odd and

5n-4
2

if n is even.

Numerical experience showed that, for fixed ,, b


and a there is always, at most, one solution of
this polynomial equation satisfying the
inequalities on the tendons, thus only one
solution of the essential prestressability
conditions.

Numerical experience showed that, at a


prestressable configuration, the rank of the
essential prestressability conditions matrix (Ae) is
always equal to the number of columns of Ae
minus one.

It can be easily proved, in the same manner as for


the three and four stage tensegrity towers, that at
a cylindrical symmetrical prestressable
configuration the overlap is less than the height
of a stage and positive:
0 < h < l cos(d ).

(35)

An important issue is the stability of these


prestressable equilibria. It is worth mentioning that
our numerical experience indicated that these
equilibria are stable. This has been ascertained by
computing the stiffness matrix of the structures - the
Hessian of the potential energy - at these prestressable
configurations. The stiffness matrices turned out to be
positive definite indicating stability of these
equilibria.

6 CONCLUSIONS
Symbolic and numerical computation have been
combined in a methodology aimed at simplifying the
prestressability conditions for tensegrity structures.
The proposed methodology, has been successfully
applied to some complicated tensegrity structures,
termed tensegrity towers, for the investigation of
certain prestressable configurations. These studies
revealed that, through the application of the proposed
methodology, the reduction in the size of the
prestressability conditions is considerable. For
example, for three stage tensegrity towers, the
prestressability conditions consist of 33 nonlinear
equations and 33 inequalities. For a certain class of
prestressable configurations this system reduces to 6
equations and 6 inequalities whose numerical solution
is much simpler.

7 APPENDIX
The structure of Ae for three stage tensegrity tower is:

Ae =

28

*
*
*
*
*
*

*
*
*
*
*
*

0 0 *
0 0 *
* * *
* * *
* 0 * *
0 0 * *
*
*
0
0

International Journal of Space Structures Vol. 18 No. 1 2003

C. Sultan and R Skelton

and its nonzero elements are given by:

Ae =

l
(-2 3b cos(d ) sin(a ) + 3l cos(d ) sin(d ) - 6h sin(d )),
6S

Ae

l
p
(2 3b cos(d ) sin(a - ) + 3 sin(d )(lcos( d ) - 2h)),
6S
3

11

13

21

p
l
l
b cos(d ) cos(a ), Ae =
(-3h sin(d ) + 3, sin(2d ) + 3b cos( d ) sin(a - )),
16
3
V
3D

Ae = -

Ae

12

p
l W
3
l W
3
(2b sin(d ) cos( a ) - 3l sin2(d )), Ae =
(2b sin(d ) sin(a + ) - 3l sin2(d ),
22
6
6S
6S

=-

Ae23 =

Ae

Ae

34

31

l
lb
p
b sin(d ) sin(a ), Ae26 =
sin(d ) sin(a + ),
V
D W
3
6

l
p
l
bcos( d )
3b cos(d ) sin(a ) - 3h sin(d )), Ae =
(-h sin(d ) +
cos(a + )),
32
S
3S
W
6
3

l
p
lcos(d )
(-3l sin(d ) - 2b cos(a + )), Ae =
(9l cos(d ) sin(d ) + 2 3b cos(d ) sin(a ) - 6h sin(d )),
35
6D
3
4V

l
(-3b cos(d ) cos(a ) + 6h sin(d ) + 3b cos(d ) sin(a )),
6D

Ae =
36

Ae

41

=-

Ae42 =

1
(- 3b sin(2a ) 6b cos2( a ) - 3b + 12l sin(d ) cos(a )),
6S

Ae

Ae

45

51

44

1
(3l sin(d ) cos(a ) - 3b + 3l sin(d ) sin(a )),
2V

1
(-6b sin2(a ) + 3l sin(d ) cos(a ) + 3 3l sin(d ) sin(a ) - 3b sin(2a )),
6D
Ae46 =

Ae

1
( 3b sin(2a ) + 6b sin2 (a ) + 6l sin(d ) cos(a ) - 6 3l sin(d ) cos(a )),
6S

=-

1
( - W
3b sin(2a ) - 6b cos 2(a ) - 3b + 12l sin(d ) cos(a )),
6D

b
(-b sin(2a ) - 3l sin(d ) sin(a ) + lW
3 sin(d ) cos(a ) - bW
3 cos(2 a )),
2S

Ae

52

=-

b
(2b sin(2a ) + 3l sin(d ) cos( a ) - 3l sin(d ) sin(a )),
2S

International Journal of Space Structures Vol. 18 No. 1 2003

29

Tensegritty Structures Prestressability Investigation

Ae53 =

Ae

56

3lb sin(d ) sin(a )


V

, Ae55 =

b2
(sin(2 a ) + W
3 cos(2 a )),
2D

3h
3
b sin(a ) (3l sin(d ) - 2b cos( a ))
, Ae = Ae = , Ae = Ae =
(l cos(d ) - h)
61
62
65
66
S
D
D

where
d = arcsin(

2b sin(a + p 3 )
).
lW
3

REFERENCES
1.

2.

3.

4.

5.

6.

7.

30

Emmerich, D. G., 1996. Emmerich on self-tensioning


structures, International Journal of Space Structures, 11
(1 and 2), 29-36.
Skelton, R. E., and Sultan, C., 1997. Controllable
tensegrity, a new class of smart structures, Mathematics
and Control in Smart Structures , Proc. SPIE 4th
Symposium on Smart Structures and Materials, 3039,
166-177.
Sultan, C., and Skelton, R. E., 1997. Integrated design of
controllable structures, Adaptive Structures and Material
Systems, Proc. ASME Intl. Congress and Exposition, 54,
27-37.
Sultan, C., and Skelton, R. E., 1998. Force and torque
smart tensegrity sensor, Mathematics and Control in
Smart Structures, Proc. SPIE 5th Symposium on Smart
Structures and Materials, 3323, 357-368.
Sultan, C., 1999. Modelling, design, and control of
tensegrity structure s with applicatio ns, Ph.D.
dissertation, Purdue University, School of Aeronautics
and Astronautics, 200 pages.
Sultan, C., Corless, M., and Skelton, R. E., 2001. The
prestressability problem of tensegrity structures. Some
analytical solutions, International Journal of Solids and
Structures, 38(30-31), 5223-5252
Pellegrino, S., and Calladine, C. R., 1986. Matrix

8.

9.

10.

11.

12.

13.

14.

analysis of statically and kinematically indetermined


framew orks. International Journal of Solids and
Structures, 22(4), 409-428.
Pellegrino, S., 1990. Analysis of prestresse d
mechanisms. Internatio nal Journal of Solids and
Structures, 26(12), 1329-1350.
Vassart, N., and Motro, R., 1999. Multiparametered
formfinding method: application to tensegrity systems.
International Journal of Spaee Structures, 14(2), 147154.
Motro, R., 1994. Form-finding numerical methods for
tensegrity systems. Spatial, Lattice and Tension
Structures. Proceedings of the IASS-ASCE Intl.
Symposium, 704 713.
Hanaor, A., 1988. Prestressed pinjointed structures flexibility analysis and prestress design. International
Journal of Solids and Structures, 28(6), 757-769.
Hanaor, A., 1992. Aspects of design of double layer
tensegrity domes. International Journal of Spaee
Structures, 7(2), 101-113.
Tarnai, T., 1980. Simultaneous static and kinematic
indeterminacy of space trusses with cyclic symmetry.
International Journal of Solids and Structures, 16(12),
347-359.
Kenner, H., 1976. Geodesic math and how to use it.
University of California Press, Berkeley.

International Journal of Space Structures Vol. 18 No. 1 2003

Вам также может понравиться