Technical Report 2002:04
ISSN 0348467X
ISRN KTH/MEK/TR02/04SE
Deployable Tensegrity Structures
for Space Applications
Gunnar Tibert
Royal Institute of Technology
Department of Mechanics
Doctoral Thesis
Stockholm, 2002
Deployable Tensegrity Structures for Space Applications
Doctoral Thesis by Gunnar Tibert
ERRATA
11 July 2002
The indices of the cable and strut lengths in Equations (3.2) and (3.3) on pages 35
and 36 should be changed from (i) t to c and (ii) c to s so that they read:
l
2
c
= 2R
2
(1 −cos θ) + H
2
, (3.2a)
l
2
s
= 2R
2
1 −cos
θ +
2πi
v
+ H
2
, (3.2b)
and
l
2
s
= 4R
2
sin
θ +
πi
v
sin
πi
v
+ l
2
c
. (3.3)
Equation (4.13) on page 83 should read:
1
P
cr
= 4
C
M
l
(4.13)
As a result of the error in (4.13), the beginning of the second paragraph on page 84
should read:
...which yields P
cr
= 1315 N using (4.13). Substituting the buckling load, ∆ = 3.65
mm and M = 13 Nm into (4.15) yielded P = 960 N.
Furthermore, the end of the third sentence on page 85 should read:
...P
Eu
= 5.28 kN or about 5.5P.
Finally, the end of line 13 on page 87 should read:
The buckling load of the struts, 960 N,...
1
Thanks are due to Dr. Zhong You for pointing out this error to the author.
Deployable Tensegrity Structures
for Space Applications
by
Gunnar Tibert
April 2002
Technical Reports from
Royal Institute of Technology
Department of Mechanics
SE100 44 Stockholm, Sweden
Akademisk avhandling som med tillst˚and av Kungliga Tekniska H¨ ogskolan i Stock
holm framl¨ agges till oﬀentlig granskning f¨ or avl¨ aggande av teknologie doktorsexa
men m˚andagen den 15:e april kl 13.00 i Kollegiesalen, Administrationsbyggnaden,
Kungliga Tekniska H¨ ogskolan, Valhallav¨ agen 79, Stockholm.
c (Gunnar Tibert 2002
Preface
The work presented in this thesis was carried out at the Department of Engineering
at the University of Cambridge, the Department of Structural Engineering and the
Department of Mechanics at the Royal Institute of Technology between December
1999 and April 2002.
First, I thank Prof. Sergio Pellegrino for giving me the opportunity to work at
his excellent Deployable Structures Laboratory (DSL). His guidance, enthusiasm,
encouragement and vast knowledge of structures have been invaluable throughout
the course of this research.
I thank my supervisor Prof. Anders Eriksson for his constant encouragement and
guidance.
A debt of gratitude is owed to Peter Knott, Roger Denston, Alistair Ross and the
technicians at the workshop of the Engineering Department in Cambridge and to
Olle L¨ ath and Daniel Hissing, at the laboratory of the Department of Structural
Engineering in Stockholm, for their technical assistance.
Thanks are due to Prof. Christopher R. Calladine of the University of Cambridge
for constructive comments on a draft paper and to Prof. William O. Williams of
Carnegie Mellon University in Pittsburgh, USA and Andrea Micheletti of Universit` a
di Roma “Tor Vergata” in Rome for valuable discussions on tensegrity.
Many grateful thanks go to Eric Pak for excellent teamwork during the design and
construction of the mast models, to Gerard James for interesting discussions and
painstaking proofreading of the manuscript, and to Dr. Anders Ansell and Dr. Karin
Forsell for meticulous proofreading of the manuscript.
I thank Tim Reynolds of the Defence Evaluation and Research Agency in Farnbor
ough, UK for providing the reﬂector antenna design task and Ian Stern of Harris
Corporation in Melbourne, USA for the data on their deployable reﬂector antennas
and for the interesting discussions on deployable tensegrities.
A thank goes to colleagues and former colleagues of the Department of Structural
Engineering and Department of Mechanics; in particular, Peter Andr´en, Anders
Ahlstr¨ om, Dr. JeanMarc Battini, Abraham Getachew, Rickard Johnson, Anders
Olsson, Dr. Jonatan PaulssonTralla, Tom Th¨ oyr¨ a, Roger Ullmann, Anders Salw´en,
Anders Wiberg and those already mentioned above. I also thank the members
of DSL and the Structures Research Group for many helpful discussions and a
iii
great time in Cambridge. Special thanks go to Dr. Marcus Aberle, Lars Ekstr¨ om,
Dr. Annette Fischer, Guido Morgenthal, Dr. Elizbar ‘Buba’ Kebadze, Sangarapillai
Kukathasan, Dr. Jason Lai, Andrew Lennon, Hannes Schmidt, Lin Tze Tan, Alan
Watt and Wesley Wong.
Finally, I would like to thank my parents for their encouragement, continuous love
and support.
Financial support from the Royal Institute of Technology and The Royal Swedish
Academy of Sciences is gratefully acknowledged.
Stockholm, April 2002
Gunnar Tibert
iv
Abstract
This thesis deals with the development of deployable structures, based on the tenseg
rity concept, for applications in space.
A stateoftheart review of deployable masts and reﬂector antennas for space appli
cations is presented. A comparison is made between the various reﬂector antennas
in terms of deployed and stowed sizes, mass and accuracy.
The key step in the design of tensegrity structures is the formﬁnding analysis.
Several methods proposed for this step are scrutinised and classiﬁed into two groups,
kinematic and static methods, and the advantages and disadvantages of each method
are investigated. Two of the statical methods seems to be identical. It is concluded
that several formﬁnding methods are available, but no single method is suitable for
general tensegrities. The force method, for the analysis of the kinematic and static
properties of large bar frameworks, is presented.
The analysis and design of deployable tensegrity masts, with three struts per stage,
is described. A routine for the manufacturing of physical models is proposed and
evaluated. Diﬀerent schemes for deployment are investigated. A way to deploy
the struts using selfdeployable hinges is introduced and demonstrated by four and
eightstage mast models. Finally, the tensegrity mast is compared with an existing
deployable mast with respect to stiﬀness. The mast is relatively stiﬀ in the axial
direction but very weak in bending.
The requirements for a deployable reﬂector antenna used on small satellites are
formulated. A concept, which uses a triangulated cable network to approximate
the reﬂecting surface, is adopted. The kinematically determinate triangulated ca
ble network is thoroughly analysed. The achievable surface accuracy of the net,
both to systematic errors arising from the triangular approximation of the surface
and random manufacturing errors, is evaluated. The underlying principles and the
statical and kinematical properties of the new concept are presented. A physical
model is built to analyse the feasibility of the concept and to test various deploy
ment schemes. The scheme using telescopic struts are identiﬁed as the most suitable
and a preliminary design an antenna, with a diameter of three metres, for a future
space mission is performed. Numerical computations show that the antenna is stiﬀ
and extremely light.
Keywords: deployable structures, tensegrity, formﬁnding, cable net, analysis, de
sign, spacecraft, mast, reﬂector antenna.
v
Contents
Preface iii
Abstract v
List of Symbols and Abbreviations xx
1 Introduction 1
1.1 Deployable Structures . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Tensegrity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Mechanics of Bar Frameworks . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Scope and Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Outline of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Deployable Space Structures 9
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Deployable Masts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 ThinWalled Tubular Booms . . . . . . . . . . . . . . . . . . . 11
2.2.2 Telescopic Masts . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Coilable Masts . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.4 Articulated Trusses . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Deployable Reﬂector Antennas . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Mesh Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 Solid Surface Antennas . . . . . . . . . . . . . . . . . . . . . . 26
2.3.3 Inﬂatable Antennas . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.4 Antenna Comparison . . . . . . . . . . . . . . . . . . . . . . . 30
vii
3 Analysis Methods for Tensegrity Structures 33
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Kinematic FormFinding Methods . . . . . . . . . . . . . . . . . . . . 35
3.2.1 Analytical Solutions . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.2 NonLinear Programming . . . . . . . . . . . . . . . . . . . . 36
3.2.3 Dynamic Relaxation . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Static FormFinding Methods . . . . . . . . . . . . . . . . . . . . . . 39
3.3.1 Analytical Solutions . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.2 Force Density Method . . . . . . . . . . . . . . . . . . . . . . 40
3.3.3 Energy Method . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.4 Reduced Coordinates . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Implementation of the Force Density Method . . . . . . . . . . . . . . 47
3.4.1 A TwoDimensional Example . . . . . . . . . . . . . . . . . . 47
3.4.2 Tensegrity Prisms . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.4.3 Spherical Tensegrities . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 The Force Method for Analysis of Bar Frameworks . . . . . . . . . . 53
3.5.1 Equilibrium and Compatibility Matrices . . . . . . . . . . . . 54
3.5.2 Static and Kinematic Properties . . . . . . . . . . . . . . . . . 54
3.5.3 RigidBody Mechanisms . . . . . . . . . . . . . . . . . . . . . 56
3.5.4 Internal Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 57
3.5.5 Structural Computations . . . . . . . . . . . . . . . . . . . . . 59
3.5.6 Example: Hanging Triangular Net . . . . . . . . . . . . . . . . 60
3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4 Deployable Tensegrity Masts 65
4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Static and Kinematic Properties . . . . . . . . . . . . . . . . . . . . . 65
4.3 FormFinding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.1 TwoStage Tensegrity Mast . . . . . . . . . . . . . . . . . . . 67
4.3.2 MultiStage Tensegrity Masts . . . . . . . . . . . . . . . . . . 68
viii
4.4 Manufacturing Technique . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.5 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.5.1 Strut Deployment . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.5.2 Demonstrator Masts . . . . . . . . . . . . . . . . . . . . . . . 75
4.6 Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.6.1 Initial Equilibrium Element Forces . . . . . . . . . . . . . . . 79
4.6.2 Preliminary Design of Struts and Cables . . . . . . . . . . . . 83
4.6.3 Vibration Analysis . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6.4 Static Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5 Design Prerequisites for a Deployable Reﬂector Antenna 95
5.1 Small Satellites and Deployable Structures . . . . . . . . . . . . . . . 95
5.1.1 MicroSatellites Astrid1 and 2 . . . . . . . . . . . . . . . . . 96
5.1.2 Small Satellite Odin . . . . . . . . . . . . . . . . . . . . . . . 97
5.1.3 Space Technology Research Vehicles . . . . . . . . . . . . . . . 98
5.1.4 Future STRV Missions . . . . . . . . . . . . . . . . . . . . . . 100
5.2 New Reﬂector Concept for Small Satellites . . . . . . . . . . . . . . . 101
5.2.1 Existing Concepts with Passive Structure . . . . . . . . . . . . 102
5.2.2 Tensegrity Reﬂector Concept . . . . . . . . . . . . . . . . . . 102
5.3 Geometry of Parabolic Reﬂector Antennas . . . . . . . . . . . . . . . 104
5.3.1 AxiSymmetric Reﬂector . . . . . . . . . . . . . . . . . . . . . 104
5.3.2 Oﬀset Reﬂector . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.3.3 Values for F/D and X
A
. . . . . . . . . . . . . . . . . . . . . 109
5.4 Reﬂector Antenna Theory . . . . . . . . . . . . . . . . . . . . . . . . 110
5.4.1 Antenna Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.4.2 Eﬀects of Random Surface Errors . . . . . . . . . . . . . . . . 112
5.4.3 Systematic Surface Error of Faceted Paraboloids . . . . . . . . 113
5.4.4 Allowable Surface Error of Reﬂector Antennas . . . . . . . . . 115
5.4.5 Ground Resolution . . . . . . . . . . . . . . . . . . . . . . . . 117
ix
5.4.6 Accuracy Goals for the Present Antenna . . . . . . . . . . . . 118
5.5 Selection of Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.5.1 Materials for the Antenna Structure . . . . . . . . . . . . . . . 118
5.5.2 Materials for the RF Reﬂective Surface . . . . . . . . . . . . . 119
6 Analysis of Tension Trusses 121
6.1 Static and Kinematic Properties . . . . . . . . . . . . . . . . . . . . . 122
6.2 AxiSymmetric Conﬁgurations . . . . . . . . . . . . . . . . . . . . . . 123
6.2.1 SagtoSpan Ratio . . . . . . . . . . . . . . . . . . . . . . . . 123
6.2.2 Position of Additional Nodes . . . . . . . . . . . . . . . . . . . 126
6.2.3 Tension Tie Force Distribution . . . . . . . . . . . . . . . . . . 126
6.3 Oﬀset Conﬁgurations . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.3.1 Focal Length and Oﬀset Distance . . . . . . . . . . . . . . . . 133
6.3.2 Two Symmetric Oﬀset Conﬁgurations . . . . . . . . . . . . . . 134
6.4 Eﬀects of Manufacturing Errors on the Reﬂector Accuracy . . . . . . 136
6.4.1 Monte Carlo Technique for the Tension Truss . . . . . . . . . 137
6.4.2 BestFit Paraboloid Analysis . . . . . . . . . . . . . . . . . . . 137
6.4.3 Extracting the Random Surface Deviations . . . . . . . . . . . 139
6.4.4 Systematic Facet Surface Deviation . . . . . . . . . . . . . . . 140
6.4.5 Inﬂuence of Tension Tie Loading . . . . . . . . . . . . . . . . 140
6.4.6 Statistical Considerations . . . . . . . . . . . . . . . . . . . . 142
6.4.7 Inﬂuence of Member Length Imperfections . . . . . . . . . . . 142
6.4.8 Inﬂuence of Tension Tie Load Variation . . . . . . . . . . . . . 143
6.4.9 Inﬂuence of Ring Structure Distortion . . . . . . . . . . . . . 145
6.4.10 Combined Inﬂuence of Manufacturing Imperfections . . . . . . 146
6.4.11 Inﬂuence of Thermal Strains . . . . . . . . . . . . . . . . . . . 147
6.4.12 Achievable Reﬂector Accuracy . . . . . . . . . . . . . . . . . . 150
7 Tensegrity Reﬂector Antennas 153
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
x
7.2 The AstroMesh Concept . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.2.1 Net Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.2.2 Deployable Ring Structure . . . . . . . . . . . . . . . . . . . . 154
7.2.3 Static and Kinematic Properties . . . . . . . . . . . . . . . . . 154
7.3 New Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.3.1 Stiﬀened Hexagonal Module . . . . . . . . . . . . . . . . . . . 157
7.3.2 Hexagonal Tensegrity Module and Tension Trusses . . . . . . 158
7.3.3 Minimum Separation between Front and Rear Nets . . . . . . 160
7.3.4 ThreeRing AxiSymmetric Reﬂector . . . . . . . . . . . . . . 161
7.4 Demonstration Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.5 Deployment Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
7.5.1 Hinged Struts . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.5.2 Telescopic Struts . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.6 Preliminary Design of 3 m Reﬂectors . . . . . . . . . . . . . . . . . . 170
7.6.1 Design Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.6.2 AxiSymmetric Reﬂectors . . . . . . . . . . . . . . . . . . . . 175
7.6.3 Oﬀset Reﬂectors . . . . . . . . . . . . . . . . . . . . . . . . . 175
7.6.4 Stowage Considerations . . . . . . . . . . . . . . . . . . . . . . 179
7.7 Vibration Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8 Conclusions 187
8.1 Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8.2 Deployable Masts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
8.3 Deployable Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
8.4 Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Bibliography 191
A Overlap Values for Tensegrity Masts 207
xi
B Flat Cable Net for Constructing Tensegrity Masts 213
C Mesh Generation Procedure for the Tension Truss 217
xii
List of Symbols and Abbreviations
Roman Symbols
A
a
Aperture area, p. 111
A
b
Projected blockage area, p. 111
a Side length of equilateral triangle, p. 46
A A
ij
= ∂l
j
/∂g
i
, p. 45
b Number of pinjointed elements (bars), p. 4
b
an
Number of additional bars for kinematic determinacy, p. 123
c Number of kinematic constraints, p. 4
C Compatibility matrix (b 3j −c), p. 54
C
f
Incidence matrix of constrained coordinates, p. 40
C
s
Incidence matrix, p. 40
C
u
Incidence matrix of free coordinates, p. 40
D Aperture diameter, p. 30
D
a
Diameter of oﬀset reﬂector, p. 106
D
p
Diameter of parent paraboloid, p. 105
d Diameter of stowed antenna, p. 30
D Matrix of internal mechanisms (3j −c m), p. 56
D
rb
Matrix of rigidbody mechanisms (m
rb
3j −c), p. 57
D Force density matrix
d Vector of nodal displacements (3j −c), p. 37
˙
d Vector of nodal velocities (3j −c), p. 37
¨
d Vector of nodal accelerations (3j −c), p. 37
xiii
d
e
Vector of nodal displacements compatible with e (3j −c), p. 60
T Dimension of space, p. 41
e Element elongation, p. 38
e Vector of bar elongations (b), p. 54
e
0
Vector of initial bar elongations (b), p. 59
F Focal length, p. 104
f
1
Lowest natural frequency, p. 30
f
op
Operating frequency, p. 32
f Vector of nodal loads (3j −c), p. 37
G Gain of reﬂector antenna, p. 111
g Generalised coordinate, p. 44
g Gravity (9.80665 m/s
2
), p. 174
G
i
Matrix of geometric loads (3j −c m), p. 58
g Vector of generalised coordinates, p. 44
H Height, p. 35
H
0
Height of axisymmetric reﬂector, p. 104
H
a
Height of oﬀset reﬂector, p. 106
h Stage overlap, p. 46
H Equilibrium matrix (3j −c b), p. 54
i An integer < v, p. 35
j Number of frictionless joints (nodes), p. 3
j
an
Number of additional joints for kinematic determinacy, p. 123
¯
k Fictitious stiﬀness, p. 38
K Stiﬀness matrix (3j −c 3j −c), p. 37
l Length, p. 5
l
e
Buckling length, p. 173
Side length of facet, p. 114
m Number of internal mechanisms, p. 4
xiv
m
rb
Number of of rigidbody mechanisms, p. 57
M Mass matrix (3j −c 3j −c), p. 37
n Number of nodes, p. 40
n Number of mast stages, p. 65
n
max
Maximum number of net rings for a given sagtospan ratio, p. 126
N Damping matrix (3j −c 3j −c), p. 37
n Translation vector (3), p. 43
^ Dimension of nullspace, p. 43
P External force, p. 5
p Tension in RF reﬂective mesh, p. 114
p
i
Nodal coordinates, (x
i
y
i
z
i
)
T
, p. 35
q Force density, p. 39
Q Diagonal matrix of force densities, p. 40
R Radius, p. 35
R
a
Radius of oﬀset reﬂector, p. 105
r Distance from centre of reﬂector, p. 112
r
g
Radius of gyration, p. 173
r
H
Rank of H, p. 55
r
R
Rank of R, p. 57
R Matrix of kinematic constraints (c 6), p. 57
r Rotation vector (3), p. 57
s Number of selfstress states, p. 4
S Matrix of states of selfstress (b s), p. 56
T Temperature (K), p. 147
t Axial element force, p. 5
t Vector of bar axial forces (b), p. 54
t
f
Vector of bar forces in equilibrium with f (b), p. 59
U Left singular matrix (3j −c 3j −c), p. 55
xv
v Number of vertices of regular polygon (or struts per stage), p. 35
W Right singular matrix (b b), p. 55
X, Y, Z Coordinates in global system (parent paraboloid), p. 105
X
A
Oﬀset distance, p. 106
X
O
, Y
O
, Z
O
Coordinates of origin (oﬀset reﬂector), p. 105
x, y, z Coordinates in global system, p. 104
x
, y
, z
Coordinates in local system, p. 105
x Vector of xcoordinates, p. 40
x
f
Vector of constrained xcoordinates, p. 41
x
u
Vector of free xcoordinates, p. 41
Greek Symbols
α Azimuth angle, p. 45
α
T
Coeﬃcient of thermal expansion, p. 149
∆F Diﬀerence between ideal and bestﬁt focal lengths, p. 138
∆H Separation of front and rear nets, p. 158
∆n Surface deviation in normal direction, p. 112
∆℘ Phase error of reﬂected rays, p. 112
∆Z
P
Translation of paraboloid apex, p. 138
∆z Surface deviation in axial direction, p. 112
δ General displacement, p. 5
δ Colatitude, p. 45
δ
rms
Radiometric rms surface deviation, p. 112
δ
rms,z
Axial rms surface deviation, p. 112
ε Strain, p. 137
η Overlap ratio of mast stages, p. 65
η
a
Total antenna eﬃciency factor, p. 111
η
b
Aperture blockage eﬃciency factor, p. 111
η
r
Random error eﬃciency factor, p. 112
xvi
η
t
Aperture taper eﬃciency factor, p. 111
θ Rotation angle, p. 35
θ
0
Angle between zaxis and rim (axisymmetric reﬂector), p. 104
θ
∗
Rotation angle of interior stages, p. 69
ϑ Temperature (
◦
C), p. 149
λ Wavelength of operating frequency f
op
, p. 111
ν Relative error, p. 38
ξ Ellipticity of oﬀset reﬂector rim, p. 106
ρ Density
ρ Sagtospan ratio, p. 125
Areal density, p. 32
σ Singular value, p. 55
σ
ε
Standard deviation of length imperfections, p. 136
σ
ρ
Standard deviation of ring structure distortion, p. 145
σ
τ
Standard deviation of tension tie loads, p. 143
Σ Matrix of singular values (3j −c b), p. 55
φ
a
Angle between x
axis and xaxis (oﬀset reﬂector), p. 105
Φ Matrix of bar ﬂexibilities (b b), p. 59
ω Element stress (equivalent to q), p. 42
Ω Matrix of element stresses ω, p. 42
ω Vector of element stresses (equivalent to q), p. 42
Abbreviations
ACeS Asia Cellular Satellite, p. 20
ADAM Able Deployable Articulated Mast, p. 16
AFRA Advanced Folding Rib Antenna, p. 20
ARISE Advanced Radio Interferometry between Space and Earth, p. 30
ASAP Ariane Structure for Auxiliary Payloads, p. 96
ATS Applications Technology Satellite, p. 18
xvii
BRC Bistable Reeled Composite, p. 11
CFRP Carbon Fibre Reinforced Plastic, p. 11
CM Coilable Mast, p. 13
CRTS Collapsible RibTensioned Surface, p. 101
CSPDA CableStiﬀened Pantographic Deployable Antenna, p. 25
CTE Coeﬃcient of Thermal Expansion, p. 118
CTM Collapsible Tubular Mast, p. 11
DAISY Deployable Antenna Integral System, p. 27
DERA Defence Evaluation and Research Agency, p. 96
DLR Deutschen Zentrum f¨ ur Luft und Raumfahrt, p. 11
DSL Deployable Structures Laboratory, p. iii
EGS EnergiaGPISpace, p. 21
ESA European Space Agency, p. 9
ETS Engineering Test Satellite, p. 23
FAST Folding Articulated Square Truss, p. 14
FE Finite Element, p. 136
FEM Finite Element Method, p. 58
GEO Geosynchronous Orbit, p. 96
GPS Global Positioning System, p. 100
GTO Geosynchronous Transfer Orbit, p. 99
HALCA Highly Advanced Laboratory for Communications and Astronomy,
p. 23
HCA Hoop/Column Antenna, p. 20
HRA HingedRib Antenna, p. 20
HST Hubble Space Telescope, p. 9
IAE Inﬂatable Antenna Experiment, p. 29
IR Infrared, p. 147
ISAS Institute of Space and Astronautical Science, p. 23
xviii
ISRS Inﬂatable Space Rigidised Structure, p. 29
ISS International Space Station, p. 15
JPL NASA Jet Propulsion Laboratory, p. 18
LaRC NASA Langley Research Center, p. 20
LDR Large Deployable Reﬂector, p. 102
LEO Low Earth Orbit, p. 95
MDF Medium Density Fibreboard
NASA National Aeronautics and Space Administration, p. 9
NASDA National Space Development Agency of Japan, p. 23
NGST Next Generation Space Telescope, p. 1
RF Radio Frequency, p. 17
rms rootmeansquare, p. 112
RRA RigidRib Antenna, p. 17
rss rootsumofsquares, p. 117
SAR Synthetic Aperture Radar, p. 100
SBA SpringBack Antenna, p. 21
SSC Swedish Space Corporation, p. 96
SSDA Solid Surface Deployable Antenna, p. 28
STEM Storable Tubular Extendible Member, p. 11
STRM Shuttle Radar Topography Mission, p. 16
STRV Space Technology Research Vehicle, p. 96
SVD Singular Value Decomposition, p. 54
TDRS Tracking and Data Relay Satellite, p. 17
TSR TapeSpring Rolling
UV Ultraviolet, p. 29
WRA WrapRib Antenna, p. 18
IEEE Frequency Bands
HF 3–30 MHz
xix
VHF 30–300 MHz
UHF 300–1000 MHz
Lband 1–2 GHz
Sband 2–4 GHz
Cband 4–8 GHz
Xband 8–12 GHz
K
u
band 12–18 GHz
Kband 18–27 GHz
K
a
band 27–40 GHz
Millimetre wave band 40–300 GHz
xx
Chapter 1
Introduction
1.1 Deployable Structures
Deployable structures are structures capable of large conﬁguration changes in an
autonomous way. Most common is that the conﬁguration changes from a packaged,
compact state to a deployed, large state. Usually, these structures are used for easy
storage and transportation. When required, they are deployed into their service
conﬁguration. A well known example is the umbrella. Deployable structures are
sometimes known under other names like expandable, extendible, developable and
unfurlable structures.
Deployable structures have many potential applications both on Earth and in space.
In civil engineering, temporary or emergency structures have been used for a long
time. A more recent application is retractable roofs of large sports stadia. In space,
deployable structures have been used since the former Soviet Union launched its
ﬁrst satellite Sputnik on October 4, 1957. In the beginning, all spacecraft were
small by virtue of the limited capacity of the launch vehicles. As the spacecraft
grew bigger so did the launch vehicles, but not at the same rate. In the foresee
able future, the capacity of launch vehicles will remain unchanged while the need
to launch very large spacecraft, e.g. the 25 m diameter Next Generation Space Tele
scope (NGST) [82], constantly is growing. Considerable research has been conducted
over the last decade on erectable structures, which are assembled in space by astro
nauts or robots, cf. [82, 95]. Erectable structures are versatile and can be compactly
stowed, but possess the disadvantage of requiring risky inspace construction. There
fore, deployable structures are the only practical way to construct large, lightweight
structures for remote locations in space. Obvious advantages of deployable struc
tures are savings in mass and volume. Another, not easily recognised, beneﬁt is
that the structure can better withstand the launch loads in the stowed conﬁgura
tion. In its deployed conﬁguration, the structure is only subjected to the orbital
loads, which are considerably lower. An important issue in the design of all deploy
able space structures is the tradeoﬀ between the size of the packaged structure and
its precision in the deployed state. Both aspects are usually critical to the mission
performance, but are sometimes conﬂicting requirements.
1
CHAPTER 1. INTRODUCTION
1.2 Tensegrity
The word tensegrity, which is a contraction of tensile integrity, was coined by R.
B. Fuller in his patent from 1962 [44]. The meaning of the word is vague and dif
ferent interpretations are possible. Fuller [44] describes a tensegrity structure as
“an assemblage of tension and compression components arranged in a discontinu
ous compression system...” Referring to the work by Fuller, Pugh [132] deﬁnes a
tensegrity system as: “A tensegrity system is established when a set of discontinu
ous compressive components interacts with a set of continuous tensile components
to deﬁne a stable volume in space.” Hanaor [54] describes tensegrity structures as
“internally prestressed, freestanding pinjointed networks, in which the cables or
tendons are tensioned against a system of bars or struts”. A broader interpretation
by Miura and Pellegrino [105] is “that a tensegrity structure is any structure realised
from cables and struts, to which a state of prestress is imposed that imparts tension
to all cables.” A narrower interpretation, also by Miura and Pellegrino, adds to the
above deﬁnition the notion that “as well as imparting tension to all cables, the state
of prestress serves the purpose of stabilising the structure, thus providing ﬁrstorder
stiﬀness to its inﬁnitesimal mechanisms.”
To explain the mechanical principle of tensegrity structures, Pugh [132] uses a bal
loon analogy. If the enclosed air is at higher pressure than the surrounding air it
pushes outwards against the inwardspulling balloon skin. If the air pressure inside
the balloon is increased, the stresses in the skin become greater and the balloon will
be harder to deform. In a tensegrity structure the struts have the role of the air
and the cables that of the skin. Increasing the forces in the elements of a tensegrity
structure will increase its strength and load bearing capacity.
The origin of tensegrity structures can be pinpointed to 1921 and a structure called
Study in Balance made by the Russian constructivist K. Ioganson, Figure 1.1(a).
MoholyNagy [109] explains “that if the string was pulled the composition would
change to another position while maintaining its equilibrium.” According to the
deﬁnitions above, this structure was not a true tensegrity structure, but it bore
a close resemblance. Today, it is generally regarded that K. D. Snelson’s XPiece
structure, Figure 1.1(b), constructed in 1948, represents the birth of the tensegrity
concept, cf. [113]. Snelson, who is a sculptor, has since built numerous tensegrity
structures, mainly for art exhibitions. While Snelson was the inventor, Fuller [44]
was the ﬁrst to look upon tensegrity structures from an engineering point of view.
A more extensive investigation into the origin of tensegrity is given in reference [83].
Even though the concept of tensegrity is more than ﬁfty years old, few applications
exist, e.g. Geiger’s cable domes [46, 123]. In recent years, the concept has received
new attention from mathematicians, engineers and biologists. Ingber [64] argues that
tensegrity is the fundamental architecture of life. Tensegritylike structures in cells
have been observed in cell biology experiments. From the deployable structures point
of view, tensegrity structures are very interesting since the compressive elements
are disjointed. This provides the possibility to fold these members and hence the
structure can be compactly stowed.
2
1.3. MECHANICS OF BAR FRAMEWORKS
(a) (b)
Figure 1.1: Early tensegritylike structures: (a) Karl Ioganson’s Study in Balance
and (b) Kenneth Snelson’s XPiece.
1.3 Mechanics of Bar Frameworks
When properly prestressed, tensegrity structures can be treated in a similar manner
as frameworks with pinjointed bars. Central to an understanding of the structural
mechanics of any framework is the concepts of static and kinematic determinacy.
M¨obius [108] was the ﬁrst to show that a general plane framework consisting of j
frictionless joints has to have at least 2j −3 bars in order to be rigid, while a space
framework needs 3j −6. M¨obius is aware of exceptions to this rule and he observes
that the determinant of the equilibrium equations of the nodes vanishes in those
cases. M¨obius also observes that the removal of a bar from a framework with the
minimum number of bars, according to the rule, transforms the framework into a
ﬁnite mechanism. But he further observes that no additional degree of mobility is
introduced if the length of the removed bar is either maximum or minimum [121].
About three decades later, Maxwell [90] rediscovers M¨obius’ rule. Maxwell also an
ticipates exceptions to the rule by stating [13, 90]: “In those cases where stiﬀness
can be produced with a smaller number of lines, certain conditions must be fulﬁlled,
rendering the case one of a maximum or minimum value of one or more of its lines.
The stiﬀness of the frame is of an inferior order, as a small disturbing force may pro
duce a displacement inﬁnite in comparison with itself.” Although it was introduced
3
CHAPTER 1. INTRODUCTION
(b)
(a)
l/2 l/2
l/2 l/2
δ
P
Figure 1.2: A twobar framework in two conﬁgurations.
by M¨obius, the rule for the construction of rigid threedimensional frameworks,
3j −b −c = 0, (1.1)
where j is the number of joints, b the number of bars and c the number of kinematic
constraints (c ≥ 6 in three dimensions) is now widely known as Maxwell’s rule.
In 1912, 75 years after M¨ obius’ ﬁndings, K¨otter [73] presents the ﬁrst study entirely
concerned with the static and kinematic indeterminacy of pinjointed frameworks.
K¨otter introduces an analytical way of evaluating whether or not a plane assembly
with b < 2j −3 is rigid [121]. The method is rather cumbersome and does not give
any general statements.
In 1978, Calladine [13] went back to the paper by Maxwell [90] for an explanation of
the mechanics of bar frameworks. Calladine studies a physical model of a tensegrity
structure with 12 joints and 24 bars, which should be loose with 6 degrees of freedom,
according to Maxwell’s rule; yet, it is stiﬀ. The key result from his study is the new
version of Maxwell’s rule, which includes all possible special cases:
3j −b −c = m−s, (1.2)
where m is the number of internal mechanisms and s the number of states of self
stress. Equation (1.2), which hereafter will be referred to as the extended Maxwell’s
rule, does not by itself solve m and s of a general bar framework, but it introduces
a clear explanation of the fundamental mechanics of bar frameworks. The values
of m and s depend not only on the numbers of bars and joints, nor even on the
topology of the connections, but on the complete speciﬁcation of the geometry of
the framework [127].
The extended Maxwell’s rule can be illustrated by a simple example. Consider the
twodimensional framework in Figure 1.2 with two bars and three joints of which
two are fully ﬁxed (c = 4). In the ﬁrst conﬁguration, Figure 1.2(a), the nodes do
not lie along the same line and it is easily seen that this framework is rigid. In the
second conﬁguration, Figure 1.2(b), all nodes lie along the same line. Both bars can
now be prestressed to the same magnitude, s = 1, and the assembly has no stiﬀness
against vertical loads, m = 1. Resolving the vertical equilibrium at the node yields
P ≈ 4t
0
δ
l
+ 8AE
δ
3
l
3
, (1.3)
4
1.4. SCOPE AND AIMS
where t
0
is the prestressing force and AE the axial stiﬀness. Thus, in the absence of
prestress, the framework has zero vertical stiﬀness in the initial conﬁguration. For
a small deﬂection δ the stiﬀness is proportional to δ
2
. However, if the framework
is prestressed the stiﬀness is proportional to the level of prestress. Thus, the mech
anism is inﬁnitesimal; as soon as the node is displaced, the elongation of the bars
will stiﬀen the structure. This illustrates what Maxwell means by “stiﬀness of ...
an inferior order”. In the case of tensegrity structures, which often have several in
ﬁnitesimal mechanisms, the prestress stabilises the structure by providing additional
stiﬀness.
1.4 Scope and Aims
Conventional satellite technology has for several decades been focused on a small
number of large, complex spacecraft. In many of these, some kind of deployable
structure was needed to make them ﬁt into the launch vehicle compartment. In
most missions, every attempt is made to circumvent the use of deployables because
of their perceived high risk and high cost. Many of the deployable structure concepts
available are very complex, but those concepts that have ﬂown are actually very
simple and use simple deployment mechanisms. A new era has now begun where
the spacecraft production for commercial applications exceeds those of the military.
This community requires structural systems that are cheaper, but with the same
reliability as their predecessors. Commercial satellites are in many cases quite small,
but an important aspect, that is frequently overlooked, is that deployable reﬂectors
are also beneﬁcial for these satellites. Deployables can be used for increased power,
aperture or to position sensitive instruments away from the interference caused by
the satellite.
Most deployable space structures are aimed for large satellites and, hence, devel
oped for a certain scale. However, the size constraints for a small satellite may be
completely diﬀerent. It is, therefore, not just a matter of scaling down an exist
ing structure. Typically, the volume of the stowed structure must be very small
to ﬁt the spacecraft bus. The basis for the present study is the assumption that
the stowed volume of deployable structures can be reduced by using the tensegrity
concept. Tensegrity structures have, for a long time, been considered suitable for
space applications, but no study has considered the entire process from initial idea
to working prototype.
The primary aim of this work is to develop deployable tensegrity structures for
space applications. To achieve this goal requires the mastering of analysis methods
and construction techniques not encountered in normal design work. The ﬁrst step
in the design of a tensegrity structure is to ﬁnd a prestressed conﬁguration. This
step is called formﬁnding. Several formﬁnding methods for tensegrity structures
have been proposed, but the advantages and limitations of each method have not
been clearly described. The aim is therefore to scrutinise and classify the exist
ing formﬁnding methods for tensegrity structures. The most challenging phase in
the development of new deployable structures is certainly the study of the deploy
5
CHAPTER 1. INTRODUCTION
ment process. Another aim is therefore to develop simple and reliable deployment
mechanisms for tensegrities. This requires studies of suitable hardware and also ex
periments with physical models. In order to thoroughly evaluate proposed ideas and
concepts, two speciﬁc deployable space structures, a mast and a reﬂector antenna,
are to be designed.
The present research is concerned with structural aspects, i.e. geometry, strength,
stiﬀness, loading, dynamics, materials, construction, etc., related to the secondary
structures of a spacecraft, e.g. deployables. No attention is put to the loadcarrying
primary structure of the spacecraft or to other issues normally related to spacecraft
design, i.e. orbit type, guidance, propulsion, electric power, communication links,
etc. Unless they can be translated into structural terms which directly aﬀect the
design of the secondary spacecraft structures considered here.
1.5 Outline of Thesis
In Chapter 2, the current stateoftheart of deployable structures for space appli
cations is reviewed. The ﬁrst part of the chapter is concerned with various systems
for deployable masts. Deployable reﬂector antennas from three groups—mesh, solid
surface and inﬂatable antennas—are studied in the second part. The antenna sys
tems are compared with respect to the deployed and stowed size, weight, surface
accuracy, natural frequencies, etc.
In Chapter 3, analysis methods for tensegrity structures are presented. Various
methods used to ﬁnd the initial equilibrium conﬁguration of tensegrity structures
are described. These methods are classiﬁed into two groups, kinematic and static,
and the advantages and disadvantages of each method are investigated. Two of
the static methods are linked to each other. In the second part of the chapter,
an eﬃcient method for structural analysis of large bar frameworks is presented.
The method is linear and usually applicable only to assemblies which undergo small
displacements. The parts of this chapter concerned with formﬁnding are reproduced
from reference [170] with some minor modiﬁcations.
In Chapter 4, the analysis and design of deployable tensegrity masts are described.
The chosen mast conﬁguration is that of Snelson [151] with three struts per stage.
Two diﬀerent conﬁgurations of the mast are analysed: one with equal strut lengths
and one with uniform element forces. A routine for the manufacturing of physical
models is proposed and evaluated. Diﬀerent schemes for deployment are investigated
and a way to deploy the struts, using selfdeployable hinges, is introduced. The
deployment approach is demonstrated by four and eightstage mast models. Finally,
the tensegrity mast is compared with an existing deployable mast with respect to
stiﬀness.
In Chapter 5, the requirements for a deployable reﬂector antenna aimed for small
satellites are formulated. A concept, which uses a triangulated cable network to
approximate the reﬂecting surface, is proposed. The geometries of axisymmetric
6
1.5. OUTLINE OF THESIS
and oﬀset reﬂector antennas are given. A good performance of the antenna requires
high accuracy of the surface. The eﬀects of diﬀerent types of errors on the antenna
performance are reviewed along with allowable surface accuracies of existing deploy
able reﬂector antennas. Another crucial aspect for the performance is the choice
of suitable material for the diﬀerent parts of the antenna. Low weight and high
stiﬀness are important properties.
In Chapter 6, the triangulated cable network is thoroughly analysed. First, a method
for making the networks kinematically determinate is described. Then, conﬁgura
tion details for adequate prestress distributions in the nets are determined. Finally,
the achievable surface accuracy of the net, both to systematic errors arising from
the triangular approximation of the surface and random manufacturing errors, is in
vestigated. This gives a clear indication on how accurate the manufacturing process
must be for a satisfactory performance of the antenna.
In Chapter 7, a detailed description of the new antenna concept is given. First, the
features of the current stateoftheart deployable mesh antenna, on which the new
concept partly is based, are highlighted. Subsequently, the underlying principles and
the static and kinematic properties of the new concept are presented. A physical
model is built to analyse the feasibility of the concept and to test various deployment
schemes. The scheme using telescopic struts is identiﬁed as being the most suitable.
A preliminary design of a three metre diameter antenna is performed.
Chapter 8 concludes the study and gives some suggestions for further research.
In Appendix A, overlap values for two to 50stage tensegrity masts with equal
length struts are given.
In Appendix B, the various conﬁgurations of the twodimensional cable net used in
the construction of tensegrity masts are presented.
In Appendix C, the routine for generating the triangular network for the reﬂector
surface is described.
7
Chapter 2
Deployable Space Structures
2.1 Introduction
Many space missions have been completed, or are currently being planned, that in
volve spacecraft which are signiﬁcantly larger than the volume capacity on available
launch vehicles. The cargo compartment of the National Aeronautics and Space
Administration’s (NASA) Space Shuttle has a diameter of 4.6 m and a length of
18.3 m and the Ariane5 launcher of Arianespace and the European Space Agency
(ESA) has a diameter of 4.56 m and is about 11 m in length [77]. As these dimen
sions are not likely to change very soon, some components of the spacecraft have
to be folded to ﬁt into the cargo bay of the launcher and then deployed once the
spacecraft is in orbit. Another limiting factor is the maximum allowable spacecraft
weight. This depends on the type of launch vehicle and destination of the satellite
and are, contrary to the volume constraint, generally negotiable.
Over the past three decades, a signiﬁcant amount of research has been carried out
in the ﬁeld of deployable space structures. There are some diﬀerences between
the existing concepts: some structures can be retracted again after they have been
deployed, others rely on stored strain energy for deployment and some structures are
stiﬀ during deployment. Retraction is not a necessity in space, but may be required
in some cases, cf. [125]. Structures not depending on stored energy are deployed by
external means, e.g. a motor. Most of the structures do not obtain full stiﬀness until
fully deployed while others can immediately sustain loads.
Despite the amount of research into deployable structures, several highproﬁle fail
ures have occurred in the last two decades [187]. Failures in space are very expensive
and extremely diﬃcult to correct. The most wellknown example is the Hubble Space
Telescope (HST) which, beside its lens problems, experienced unexpected levels of
mechanical vibration caused by the thermal loading of its solar arrays. One reason
for these failures is an incomplete understanding of the behaviour of the structure.
Another reason, probably more rare, is that the concept itself is poor. Miura [101],
inventor of several deployable structures, emphasise that “creating a rational struc
tural concept should be the ﬁrst step in the process of designing a structure and it
9
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
precedes the practical design step.” The robustness of the concept has to be ﬁrst
proven by analytical and physical models. The actual design procedure that follows
involves just a tradeoﬀ between packing eﬃciency, structural stiﬀness and precision
of the deployed structure. The ﬁnal steps in the construction of a space structure
involve extensive ground and, possibly, ﬂight testing, which are extremely costly but
unavoidable to ensure mission success.
In the aerospace industry there are three main types of deployable structures:
• Masts,
• Antennas, and
• Solar panels.
Masts are typically used for separating electronic instruments to reduce interfer
ence [95] or for supporting other structures such as solar arrays [190]. A vast number
of mast concepts exist, cf. [95], whereof only a few will be presented in this chapter.
All satellites need to communicate with Earth and therefore need some type of an
tenna. Amongst the many antenna types available, the parabolic reﬂector antenna
is the most common one mainly due to its high gain, which enables high data rate
transmission at low power [158]. Antennas are used not only for communication
but also for Earth observation and astronomical studies. The current ﬂexible solar
cell technology can produce 223 W/m
2
[136], which means that the solar arrays
have to be quite large to produce enough power for the ever increasing number of
instruments aboard a satellite. One of the most sophisticated spacecraft is the HST,
which requires 4.7 kW for its instruments. Four 2.39 6.06 m
2
solar arrays provide
this power [77]. Deployable concepts for solar panels are not within the scope of this
thesis and will therefore not be reviewed. For information on solar arrays see [125].
The remaining of this chapter is concerned with the most important concepts avail
able for deployable masts and reﬂector antennas.
2.2 Deployable Masts
Deployable masts can be divided into the following four groups [95]:
• Thinwalled tubular booms,
• Telescopic masts,
• Coilable masts, and
• Articulated trusses.
A few reviews on deployable masts exist: Mikulas and Thompson [95] present struc
tures developed in the U.S.A., Pellegrino [125] covers concepts that also are re
tractable, and Jensen and Pellegrino [68] present the most recent and extensive one.
10
2.2. DEPLOYABLE MASTS
2.2.1 ThinWalled Tubular Booms
Thinwalled tubular booms are probably the earliest types of deployable and re
tractable structures. They make use of the elastic deformability of thinwalled shells.
Typical materials for thinwalled booms are stainless steel, CopperBeryllium and
Carbon Fibre Reinforced Plastic (CFRP) [68, 125].
Storable Tubular Extendable Member
The Storable Tubular Extendible Member (STEM) was invented in Canada in the
early 1960s [137]. It is an extension of the principle used for coilable selfstraightening
steel tape measure. While the tape measure only subtends a small angle of the com
plete cylinder formed by its radius of curvature, the STEM covers more than 360
◦
,
Figure 2.1(a). The choice of overlap depends on several factors but is at least
50
◦
[125]. A STEM is quite stiﬀ axially and in bending, but because of the open
tubular section it has low torsional stiﬀness. Increased torsional stiﬀness can be
provided if there is suﬃcient friction in the overlap region. The STEM is rolled up
and ﬂattened onto a drum within a cassette for stowage. It is deployed by rotating
the drum whereby the stored elastic energy automatically brings it back into the
tubular conﬁguration. The ploy region, where it goes from ﬂat to tubular cross
section, is contained with the stowage cassette. An extension of the STEM is the
biSTEM with two identical strips placed one inside the other, Figure 2.1(b). A
biSTEM has higher bending stiﬀness and better mechanical damping behaviour.
Other advantages of a biSTEM over a STEM of similar stiﬀness and length are
shorter drums and ploy length, which give a more compact stowage cassette [125].
One version of the biSTEM has interlocking STEMs, Figure 2.1(c), which increase
the torsional stiﬀness. A recently developed tubular boom is the Bistable Reeled
Composite (BRC) [65], which looks exactly like a STEM but is stable in both the
stowed and deployed conﬁguration. This means that the stowage cassette can be
made smaller and lighter. In addition, the retraction mechanism can be simpliﬁed.
Collapsible Tubular Mast
The Collapsible Tubular Mast (CTM) is made from two STEMs bonded at the
edges, which create a boom with higher torsional stiﬀness than the STEM. CTMs
are rolled up in a similar way as STEMs. In the deployed conﬁguration, the tube
is unstressed and has a lenticular crosssection [125]. A 14 m long CTM to be used
for solar sails was recently developed by the German Aerospace Centre (DLR). It is
made of CFRP and weighs only 0.1 kg/m [33].
2.2.2 Telescopic Masts
Telescopic masts normally consist of a series of concentric, thinwalled cylindrical
tubes that are nested inside one another. Limiting factors in terms of length are
11
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
(a)
(b)
(c)
Figure 2.1: Tubular booms: (a) STEM, (b) biSTEM, and (c) interlocking biSTEM.
Figure 2.2: Collapsible tubular mast (Courtesy of DLR).
tube thickness and overlap length. Typical thickness of aluminium alloy and CFRP
is 0.5 mm. Telescopic masts can be deployed either sequentially using a spindle
andnut technique, Figure 2.3(a), or synchronously, Figure 2.3(b), by combining
the previous technique with cables and pulleys. A 40 m long telescopic mast of
CFRP has been developed by the company Dornier [77]. The mast consists of 18
segments of about 0.5 m in diameter and has a stowed length of 3 m. A 2.4 m long
synchronously deployed telescopic mast consisting of seven tubes has been developed
for the Tethered Satellite [125].
12
2.2. DEPLOYABLE MASTS
Figure 2.3: Deployment of a telescopic mast: (a) sequential and (b) synchronous
(Reproduced from [125]).
2.2.3 Coilable Masts
In 1967, the Coilable Mast (CM) was invented by H. R. Mauch of the Astro Re
search Corporation [182]. A coilable mast normally consists of three longitudinal
elements braced at regular intervals. The bracing consists of members perpendicu
lar to the longitudinal ones and diagonal members. The mast is stowed by coiling
the longerons, Figure 2.4. It can be deployed in two diﬀerent ways: self or motor
driven extension. The ﬁrst method relies on the stored elastic energy in the coiled
longerons and the rate is controlled by a lanyard, i.e. an axial cable attached to
the tip of the mast, which pays out to control the deployment and reels in to re
tract the mast. As a result, the tip of the mast rotates during deployment. As the
stiﬀness of the mast is lower during deployment than in the deployed conﬁguration,
this method is only suitable for shorter masts, typically less than 3 m [125]. For
longer masts, the motor driven method is used whereby the mast is stowed inside a
special canister, Figure 2.4. The canister, which is about two mast diameters higher
than the retracted mast length, contains a motor driven rotating nut. The transi
tion zone, where a section of the mast goes from stowed to deployed conﬁguration,
is contained within the canister so the part leaving the canister has full strength.
Another advantage with the canister is that it is the nut rather than the tip of the
mast that rotates. A CM is very eﬃcient from a stowage viewpoint; the retracted
length is about 2–3% of the deployed length. Masts with diameters up to 0.75 m
have been constructed and the practical limit is considered to be about 1 m [95].
2.2.4 Articulated Trusses
Articulated trusses are widely used for space applications and are available in many
diﬀerent conﬁgurations. They are capable of higher stiﬀness, structural eﬃciency
and precision than the previous concepts. One thing that popular truss masts have
13
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Figure 2.4: Deployment of a coilable mast from a canister by a rotating nut (Cour
tesy of AECAble Engineering Company, Inc.).
in common is a constant diameter during deployment. A large number of other inge
nious truss concepts have been invented, e.g. the Variable Geometry Truss [103] and
the CableStiﬀened Pantographic Mast [77, 186], but it is the author’s opinion that
most of them have fallen into disfavour partly because of their changing diameters.
Folding Articulated Square Truss Mast
The Folding Articulated Square Truss (FAST) mast was developed by AECAble
Engineering Company. The mast has revolute hinges along the longerons with axes
parallel to the sides of the square bays and two pairs of diagonal bracing cables on
each face of the bays [105], Figure 2.5(a). The cables are prestressed by four lateral
bows. During folding, half of the bracing cables become slack as the bows bend,
Figure 2.5(b) and (c). The strain energy stored in the bows actuates the deployment
of one bay of the mast. Hence, each deployed bay has full stiﬀness. As for CMs, the
14
2.2. DEPLOYABLE MASTS
Force to
retract
Deployed
Transition
Folded
(a)
(b) (c)
Figure 2.5: Principle of the FAST mast (Courtesy of S. Pellegrino).
Figure 2.6: FAST mast for the ISS (Courtesy of AECAble Engineering Company,
Inc.).
retracted mast and transition zone are enclosed by a canister. Eight FAST masts,
each 1.09 m in diameter and 34.75 m in length, support the solar arrays on the
International Space Station (ISS), Figure 2.6. The canister length is 2.3 m or about
6.6% of the deployed length [1].
15
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
(a) (b)
Figure 2.7: ADAM deployment sequence (Courtesy of AECAble Engineering Com
pany, Inc.).
Able Deployable Articulated Mast
AECAble Engineering Company recently developed the Able Deployable Articu
lated Mast (ADAM) for applications requiring very long and stiﬀ masts. Compared
to the FAST mast, ADAM uses no bows and only one pair of cross bracing cables
on each face of a bay, Figure 2.7. Spherical hinges are ﬁtted at the ends of the
longitudinal members and the rigid lateral square rotates almost 90
◦
during deploy
ment, Figure 2.7(a). Special latches on the diagonal cables stop the deployment and
stiﬀen each bay, Figure 2.7(b). Like in CMs, a canister/nut technique is used for
deployment. A 60 m long ADAM is used in the Shuttle Radar Topography Mission
(STRM), which maps the Earth. The STRM ADAM, whose prime function is to
separate two radar antennas, has a diameter of 1.12 m and consists of 87 bays. Its
stowed height is 1.42 m and the length of the canister is 2.92 m or 4.8% of full mast
length [1].
2.3 Deployable Reﬂector Antennas
Several reviews of deployable reﬂector antennas are available: Freeland [40], Roed
erer and RahmatSamii [138], Rogers et al. [139], Mikulas and Thomson [95], Pel
legrino [125], and Hachkowski and Peterson [52]. The reviews by Freeland and
Mikulas and Thomson cover only concepts developed in USA, while Roederer and
RahmatSamii also include European ones. Pellegrino focuses on concepts, from
16
2.3. DEPLOYABLE REFLECTOR ANTENNAS
USA, Europe and Japan, which are retractable. The latest one by Hachkowski and
Peterson is concerned with the precision of the antennas and includes 50 structures,
not only reﬂectors. In recent years a small number of new, very important con
cepts have been developed which are not included, or only mentioned brieﬂy, in the
previous reviews. A new review, including only the most important structures, is
therefore necessary for a more complete background.
Basically, there are three diﬀerent types of deployable antennas:
• Mesh antennas,
• Solid surface antennas, and
• Inﬂatable antennas.
In each category, several diﬀerent concepts have been proposed but only a limited
number have proven to be viable and even fewer have actually ﬂown. In the following
three sections, the most important concepts from each category will be presented,
some of which have ﬂown while others never left the ground. Finally, the antenna
concepts will be summarised and compared in terms of packaging eﬃciency, areal
mass, surface accuracy and natural frequency.
2.3.1 Mesh Antennas
The most common type of deployable antennas is the mesh antenna with a reﬂec
tive surface composed of a knitted lightweight metallic mesh. Although the mesh is
discontinuous, it can reﬂect radio frequency (RF) waves up to about 40 GHz [95].
Deployable mesh antennas are available in many conﬁgurations which diﬀer in the
way the mesh is supported. The most common antenna design is an inverted um
brella with curved ribs emanating from a hub and the mesh suspended between the
ribs. Umbrellatype designs are still prevalent but several new concepts, which can
achieve higher surface accuracy, are now being developed. In the following, the most
important deployable mesh antenna concepts are considered.
RigidRib Antenna
Harris Corporation developed the RigidRib Antenna (RRA) for the NASA Tracking
and Data Relay Satellite
1
(TDRS) and the NASA Galileo mission to Jupiter. The
RRA is an umbrellatype antenna with 16 parabolic, tubular CFRP ribs attached to
a central hub and an RF reﬂective mesh between the ribs, Figure 2.8(a). The ribs are
hinged only at the hub and they simply fold towards the feed structure, which gives
a stowed height about the same as the antenna radius. The antennas constructed
for the TDRS and Galileo are almost identical with a diameter of 5 m. In its
1
The rigid rib antennas by Harris were used on TDRSA through G and then replaced by the
springback antenna for TDRSH through J.
17
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Goldplated
molybdenum wire
1.2 mil diameter
10 openings/inch
Equipment
compartment
Hub
Struts
18 GRP ribs
Subreflector
(a) (b)
Figure 2.8: The 5 m diameter rigid rib antenna for the Galileo mission (Courtesy of
Harris Corporation).
stowed conﬁguration, the diameter and height is 0.9 and 2.7 m, respectively
2
. The
entire antenna structure, including the ribs, reﬂector surface, feed and deployment
mechanisms that fold and unfold the structure, weighs 24 kg. The RRA antenna
onboard the Galileo spacecraft, which was launched on October 18, 1989, failed to
deploy as commanded on April 11, 1991. The mission objectives could, however,
be accomplished using an antenna with lower gain and various enhancements to
the communication link. The failure is believed to be caused by very high friction
between restraint pins and their receptacles.
WrapRib Antenna
NASA Jet Propulsion Laboratory (JPL) and Lockheed Missiles and Space Company
developed the WrapRib Antenna (WRA) in the 1970s. This is an umbrellatype
concept with a central hub, parabolic CFRP ribs of lenticular crosssection, and an
RF reﬂective mesh. In the stowed position, the ribs are rotated on vertical hinge pins
and then tangentially wrapped around the hub. The ribs are deployed by cutting
a restraining cable placed around the hub. Springloaded doors are opened and
the stored energy in the ribs causes them to unwrap into their original position.
In vacuum, this sequence takes about two seconds. The deployment mechanism
is shown in Figure 2.9(b). Note that it is also possible to retract the antenna in
orbit by reversing the drive. A 9.1 m diameter WRA was launched in 1974 with
the Applications Technology Satellite 6 (ATS6), Figure 2.9(b). The ATS6 antenna
consisted of 48 ribs, which were stowed in a 2.0 m diameter hub with a height of
0.45 m. The whole antenna weighed about 60 kg [87, 138, 146].
2
From measurement in ﬁgure on page 415 of [40].
18
2.3. DEPLOYABLE REFLECTOR ANTENNAS
(a) The 9.1 m diameter ATS6 wraprib antenna (Courtesy of
Lockheed Missiles and Space Company).
RF reflective mesh
Unfurl./refurl. drive
Furled wrap ribs/
RF reflective mesh
pack
Wrap ribs
Unfurl./refurl.
drive
Tape
Tape spool
R
i
b
1
R
ib
1
R
ib
1
(b) Wraprib deployment mechanism.
Figure 2.9: The wraprib reﬂector antenna.
19
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Figure 2.10: The 12 m diameter hingedrib antenna for the ACeS system (Courtesy
of Harris Corporation).
HingedRib Antenna
Harris Corporation developed the HingedRib Antenna (HRA) for the Asia Cellular
Satellite (ACeS) system. In this concept, the ribs are hinged along their length,
which produce a smaller height of the stowed package. In order to provide a smaller
package diameter, the ribs are straight and the parabolic mesh surface is shaped
by standoﬀ elements along the ribs, Figure 2.10. Two 12 m diameter HRAs are
mounted on the ACeS Garuda1 satellite, launched February 13, 2000. The stowed
diameter and height of each antenna are 0.86 and 4.5 m, respectively. The total
weight of the antenna and boom is 127 kg. Harris is currently working on a reﬁned
version of the HRA, called Advanced Folding Rib Antenna (AFRA), with the same
deployed diameter as the ACeS HRA, but with lower mass and smaller stowed
diameter and height [154].
Hoop/Column Antenna
NASA Langley Research Center (LaRC) and Harris Corporation developed the
Hoop/Column Antenna (HCA) around 1980, to demonstrate the feasibility of a large
space antenna. The concept is a simple tension and compression preloaded struc
ture. A central column and a largediameter hoop are the compression members,
which maintain pretension in a cable network. In addition to its structural contri
butions, the cable network supports and shapes the RF reﬂective mesh surface [40].
The antenna deploys in three principal steps: the column deploys simultaneously
from both top and bottom, Figure 2.11(a) and (b); the hoop deploys by means of
motors mounted in eight of the 24 hoop joints, Figure 2.11(c)–(e). Finally, an out
ward extending preload segment is deployed at the bottom of the column, which
pretensions all cables and the mesh. A 15 m diameter HCA, with a height of 9.5 m,
has been built and tested on ground. In the stowed conﬁguration, the antenna ﬁts
inside a cylinder of 2.7 m in height and 0.9 m in diameter. The total weight of the
20
2.3. DEPLOYABLE REFLECTOR ANTENNAS
(a)
(b) (c) (d)
(e)
Figure 2.11: Deployment sequence of the hoop/column antenna. Reproduced from
[146].
antenna structure alone is 291 kg. The total weight, including the instrumentation
and feed, is 410 kg [9].
EGS Antenna
The RussianGeorgian company EnergiaGPISpace (EGS) has developed a new
concept aimed at deployable reﬂector antennas in the diameter range 5–25 m. The
EGS antenna consists of a circular pantograph ring and radial tensioned membrane
ribs connecting to a central hub. An elliptical reﬂector antenna with dimensions 5.6
m by 6.4 m was tested in space on the Russian orbital station MIR, Figure 2.12.
The stowed diameter and height of the antenna are 0.6 and 1.0 m, respectively.
The weight of the antenna, including both the mechanical and electrical systems,
is 35 kg. A 13.5 m diameter antenna has also been built, but only tested on the
ground [4, 91].
SpringBack Antenna
A totally diﬀerent antenna concept, without moveable connections, was developed
by the Hughes Space and Communication Company
3
. The SpringBack Antenna
(SBA) consists of a thin graphite mesh dish with an integral lattice of ribs and a
stiﬀening hoop along the rim. The antenna is elastically folded like a taco shell and
3
Now Boeing Satellite Systems, Inc.
21
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Figure 2.12: A 5.6 m diameter EGS antenna in space. Photo taken from the orbital
station MIR on July 28, 1999 (Courtesy of EnergiaGPISpace, Ltd.).
held in that conﬁgurations by tie cables along the rim, Figure 2.13. The stowed
volume of this antenna is quite large compared to other antennas and the stowed
height is about the same as the reﬂector diameter. In orbit, the reﬂector deploys by
cutting the tie cables, which releases its stored elastic strain energy. The MSAT1
satellite, launched April 20, 1996, was the ﬁrst to use the SBA. Each of the two
SBAs had a 6.8 m by 5.25 m elliptical shape and weighed 20 kg. In the stowed
conﬁguration, the two SBAs are rolled together into a 4.9 m high truncated cone on
top of the spacecraft. The top and bottom diameters of the cone are 1.5 and 3.0 m,
respectively. In a more recent application, NASA’s next generation TDRS, the 5 m
diameter RRAs are replaced by 4.6 m diameter SBAs [10, 105].
Tension Truss Antenna
The Tension Truss concept was developed by Miura in 1986 [104] to meet demands
for high precision large deployable reﬂectors. In this concept, the reﬂector is divided
into triangular facets instead of gores as in an umbrellatype reﬂector. The idea
of approximating bowlshaped surfaces by polygons was not new, cf. the geodesic
dome, and had been previously applied in space antenna applications. However, the
novelty of the tension truss concept was to use ﬂexible members for the sides of the
triangles so that the whole assembly could be easily folded. Outofplane forces, e.g.
22
2.3. DEPLOYABLE REFLECTOR ANTENNAS
Figure 2.13: Folded springback antenna (Courtesy of Hughes Space and Commu
nications Company).
springs, would be used to prestress the triangular network. A main advantage of the
tension truss antenna over the umbrellatypes is that the surface accuracy can be
increased, e.g. by decreasing the size of the triangles, without increasing the number
of supports. The triangular prestressed network is the support for the reﬂective
mesh. In Japan, two diﬀerent types of antennas using the tension truss concept
have been developed. The main diﬀerence between them is the supporting structure
for the triangular network. A reﬂector with an eﬀective aperture diameter of 8 m
was developed by the Institute of Space and Astronautical Science (ISAS) and ﬂown
on the Highly Advanced Laboratory for Communications and Astronomy (HALCA)
satellite, Figure 2.14. From the main bus, the tension truss is deployed by six truss
masts
4
. Maximum diameter of the antenna is 10 m and the total mass is 246 kg;
the mass of the mesh and cables is about 11 kg and that of the six deployable masts
is about 100 kg [63]. The National Space Development Agency of Japan (NASDA)
developed another antenna for their Engineering Test Satellite VIII (ETSVIII). The
ETSVIII antenna is based on a modular approach with 4.8 m diameter hexagonal
tension truss modules. A 14module antenna has an aperture diameter of 13 m, but
the total dimension of the antenna is 19.2 16.7 m
2
. The stowed size is 1 m in
diameter and 4 m in height. The total weight of a 14module antenna, including
supporting booms and deployment mechanics, is about 170 kg [92]. A tension truss
4
The exact mast type is not known to the author, but according to K. Miura, [102], CMs could
not be used because of the high compressive forces.
23
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Figure 2.14: Tension truss antenna for HALCA (Courtesy of ISAS).
antenna developed in USA will be described next.
AstroMesh Antenna
In 1990, Astro Aerospace Corporation
5
started developing what is now the current
stateoftheart deployable reﬂector antenna—the AstroMesh. As mentioned in the
previous section, the AstroMesh is based on the tension truss concept and its compo
nents are shown in Figure 2.15(a). Two identical paraboloidal triangular networks
are attached to a deployable ring truss. This assembly is prestressed by tension
ties connected to mirroring nodes of the two networks. The RF reﬂective mesh is
attached to the backside of the front net. The antenna is deployed by shortening a
cable which continuously runs through the telescopic diagonal members of the ring
truss. Deployment synchronisation is achieved through special joints at the truss
connections where only three members meet [166], which can be clearly seen in
Figure 2.15(c). The latest application of the AstroMesh is onboard the telecommu
nication satellite Thuraya, which was launched on December 5, 2000. This antenna,
which is shown in Figures 2.15(b) and (c), has a diameter of 12.25 m and weighs
55 kg. In the stowed conﬁguration, the diameter and height are 1.3 and 3.8 m,
respectively [172].
5
Now TRW Astro Aerospace.
24
2.3. DEPLOYABLE REFLECTOR ANTENNAS
Rear Net
Tension
Ties
Truss
Mesh
Front Net
(a) (b)
(c)
Figure 2.15: The AstroMesh deployable reﬂector antenna: (a) concept, (b) and (c)
the 12.25 m diameter reﬂector for Thuraya (Courtesy of M. W. Thom
son and TRW Astro Aerospace).
CableStiﬀened Pantographic Antenna
DSL developed the CableStiﬀened Pantographic Deployable Antenna (CSPDA).
The deployable ring structure consists of three diﬀerent pairs of rods connected
by scissor joints. The pairs of rods are connected at their end points to form a
circular pantographic structure that can be folded. Crucial to a successful folding
is accurate positioning and manufacturing of the scissor and end joints. A 3.5 m
diameter model has twelve sides and is composed of 48 pantograph elements, Fig
ure 2.16. A double layer cable network, which supports the RF reﬂective mesh, is
25
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
(a) (b)
(c)
Figure 2.16: Deployment sequence of the 3.5 m diameter model of the cablestiﬀened
pantographic deployable antenna (Courtesy of Z. You and S. Pelle
grino).
attached to the ring structure. The layout of the network is chosen such that the
total structure is statically determinate, or indeterminate to only a small degree.
Like the AstroMesh, an active cable is used to deploy the ring structure. In the
stowed conﬁguration, the diameter and height are 0.6 and 1.2 m, respectively [188].
2.3.2 Solid Surface Antennas
Antenna applications operating at frequencies over about 40 GHz require high sur
face accuracy. Solid material is usually chosen for their reﬂective surface [95]. Most
of the solid surface deployable reﬂector concepts consist of a central hub with rigid
curved panels, often CFRP face sheets over an aluminium honeycomb core, arranged
as radial petals. The concepts diﬀer by the manner in which the petals fold [51].
Because of mechanical complexity and launch vehicle size constraints, deployable
high precision reﬂectors are limited to approximately 10 m in diameter [95].
26
2.3. DEPLOYABLE REFLECTOR ANTENNAS
Sunﬂower
The ﬁrst concept with rigid panels was the Sunﬂower by TRW, Figure 2.17. It
folds using simple revolute joints between the panels, but does not achieve a great
reduction in size. A 4.9 m diameter model was built, which has a stowed diameter
and height of 2.15 and 1.8 m, respectively [51, 52]. An extended version of the
Sunﬂower, with a diameter of 15 m, stows into a cylinder of 4.4 m in diameter
and 6.6 m in height [60]. This version is, however, more complex that the original
Sunﬂower.
DAISY and MEA
The second concept is the Deployable Antenna Integral System (DAISY) by ESA
and Dornier, Figure 2.18. Here, each panel is connected to the central hub by
a revolute joint, which allows the panels to fold by nesting in front of the hub.
The position and orientation of the hinges are determined by extensive deployment
simulations to achieve good packaging and avoid interference between the panels
during deployment. The truss structure on the back of each panel, seen in Figure
2.18, provides additional stiﬀness for better surface accuracy. An 8 m diameter
engineering model has been built, with a stowed diameter and height of 2.9 and
4.1 m, respectively [51, 105].
The third concept, also by Dornier and ESA, is MEA. The folding conﬁguration is
similar to the DAISY, with panels nested in front of the hub. Each panel is connected
to the hub by a joint that allows rotation about two axes and to neighbouring panels
by rods with spherical joints. The paneltopanel connecting rods compensate for
the increased kinematic freedom introduced by the twoaxis joints and synchronise
the motion of the panels. During folding, each panel folds towards the hub and
twists at the same time. A 4.7 m diameter MEA model has a stowed diameter and
height of 1.7 and 2.4 m, respectively [51, 186].
Figure 2.17: Sunﬂower antenna: (a) folded; (b) deployed, plan view (Courtesy of
TRW).
27
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
(a) (b)
Figure 2.18: DAISY antenna: (a) folded; (b) deployed (Courtesy of Dornier GmbH).
(a) (b)
(c) (d)
Figure 2.19: Deployment sequence of the solid surface deployable antenna: (a)–(d)
(Courtesy of S. D. Guest).
Solid Surface Deployable Antenna
The fourth concept, the Solid Surface Deployable Antenna (SSDA) by DSL, Fig
ure 2.19, is quite diﬀerent from the others. In this concept, the surface is split
into wings rather than radial petals. Each wing is further subdivided into panels
which are connected by revolute joints. This concept is extendable to any number
of wings and panels and will package eﬃciently as the curved panels nest inside one
another. A 1.5 m diameter SSDA model with six wings and ﬁve panels per wing
had a stowed diameter and height of 0.56 and 0.81 m, respectively. Estimates show
that increasing the number of panels from ﬁve to seven, yield 0.36 and 0.75 m as
stowed dimensions [51].
28
2.3. DEPLOYABLE REFLECTOR ANTENNAS
2.3.3 Inﬂatable Antennas
An inﬂatable antenna would give the smallest package size and potentially the lowest
mass. It is constructed from a thin ﬂexible material which is folded prior to launch
and then deployed by inﬂation. The reﬂector structure is like a circular paraboloidal
cushion with a transparent front side and a reﬂective rear side. It is stiﬀened by an
inﬂatable torus along the edge. The structure can be made more rigid by impreg
nating the membrane material with a resin, which slowly cures at high temperature
or by the Sun’s ultraviolet (UV) rays. The main disadvantage of inﬂatable anten
nas is the diﬃculty of achieving high shape accuracy for the reﬂective surface [186].
Inﬂatable structures have a high deployment reliability because of their simplicity
and a low number of failure modes. For testing the structures on the ground, a zero
gravity environment can be simulated through the use of Helium [40].
Inﬂatable Space Rigidised Reﬂector
ESA and the company Contraves developed an antenna using Inﬂatable Space Rigidised
Structure (ISRS) technology for a joint ESA/NASA programme requiring a large
antenna [135]. The antenna structure is made of Kevlar
6
membranes, impregnated
with an epoxytype resin. When in orbit, the inﬂated antenna is positioned so that
its faces the Sun. The resin hardens in six hours at a temperature of 110
◦
C, which
is easily achieved when the structure is exposed to direct sunlight. After hardening,
the inﬂation gas is expelled [135]. Three engineering models of the antenna, with
diameters 3.5, 6 and 12 m, have been built. Figure 2.20 shows the latest model,
the 12 m diameter oﬀset reﬂector. However, no ﬂight test was performed and the
programme has now been cancelled [28]. An early design study showed that a 20 m
diameter reﬂector would have a mass of 134 kg [135].
Inﬂatable Antenna Experiment
To demonstrate the potential of inﬂatable structure technology for large antennas,
JPL initiated the Inﬂatable Antenna Experiment (IAE), Figure 2.21. Together with
the company L’Garde they developed and manufactured a 14 m diameter inﬂatable
antenna with a low pressure canopy structure, high pressure torus and three high
pressure struts, which support the feed. As the antenna would only be used for a
short time, no attention was paid to space rigidised material. The canopies were
constructed of 6.5 µm Mylar
7
ﬁlm, the front canopy was left transparent while the
back was aluminised for reﬂectivity. The torus and struts were made of 0.3 mm
thick Kevlar. A smooth reﬂective surface could be obtained at a very low pressure
of 2.1 Pa. The torus and struts were both pressurised to 6.9 kPa. The total weight
of the inﬂatable structure was 60 kg. The size of the box container, which housed
the stowed inﬂatable structure, was 2.0 1.1 0.46 m
3
. Deployment starts with
6
Kevlar is a registered trademark of E. I. du Pont de Nemours and Company.
7
Mylar is a registered trademark of E. I. du Pont de Nemours and Company.
29
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
Figure 2.20: The 12 m diameter model of the inﬂatable space rigidised reﬂector
antenna (Courtesy of A. G. Roederer, ESA).
an ejection of the package from its container. Then, the struts are inﬂated followed
by the torus and ﬁnally, the canopy. The IAE was launched on May 29, 1996
and thereby provided the ﬁrst demonstration of the potential of inﬂatable precision
structures [41, 43]. Currently, JPL and L’Garde are working on a project called
Advanced Radio Interferometry between Space and Earth (ARISE), which includes
the development of a 25 m diameter inﬂatable antenna pressurised to 2.8 Pa [20].
2.3.4 Antenna Comparison
A comparison between the presented antenna concepts is given in Table 2.1. For
mesh antennas, the stowed diameter d varies between 6 and 22% of the deployed
diameter D and the stowed height h varies between 5 and 93%. The solid surface
antennas have, as expected, worse packaging ratios than their mesh counterparts.
The IAE has, as expected, the best packaging ratio with ratios of 6 and 14% in
diameter and height. Another crucial parameter is the mass and surface density.
Low mass is necessary because of the extremely high launch costs in the order of
10,000 USD/kg [185]. In that respect, it should be noted that the AstroMesh has an
areal density signiﬁcantly lower than other mesh antennas and comparable to those
of inﬂatable antennas. Low mass combined with a high structural stiﬀness give a
high lowest natural frequency of the structure, which is desirable to separate the
structural and attitude control system frequencies [55]. The large mesh antennas in
Table 2.1 have a fundamental frequency f
1
lower than 1 Hz. Unfortunately, no value
30
2.3. DEPLOYABLE REFLECTOR ANTENNAS
Figure 2.21: The 14 m diameter IAE reﬂector antenna in space (Courtesy of
L’Garde, Inc.).
for the deployed frequency of the 12.25 m diameter AstroMesh reﬂector is given in
references [165, 166]. A 6 m diameter AstroMesh is reported to have a fundamental
frequency of 2.0 Hz [166], which can be compared to 17 Hz of the EGS antenna
of about the same size. The last parameters, the surface accuracy and operating
frequency, will be discussed in Chapter 5 in connection to the development of the
tensegrity reﬂector antenna.
31
T
a
b
l
e
2
.
1
:
C
o
m
p
a
r
i
s
o
n
o
f
d
i
ﬀ
e
r
e
n
t
d
e
p
l
o
y
a
b
l
e
r
e
ﬂ
e
c
t
o
r
a
n
t
e
n
n
a
c
o
n
c
e
p
t
s
(
s
o
r
t
e
d
b
y
t
y
p
e
a
n
d
a
p
e
r
t
u
r
e
s
i
z
e
)
.
T
y
p
e
a
S
t
r
u
c
t
u
r
e
S
t
a
t
u
s
b
D
d
/
D
h
/
D
M
a
s
s
c
f
1
δ
r
m
s
f
o
p
R
e
f
.
C
o
m
m
e
n
t
(
m
)
(
k
g
)
(
k
g
/
m
2
)
(
H
z
)
(
m
m
)
(
G
H
z
)
M
H
C
A
E
M
1
5
0
.
0
6
0
.
1
8
2
9
1
1
.
6
5
0
.
0
6
8
1
.
5
2
1
1
.
6
[
9
,
1
7
]
Q
u
a
d

a
p
e
r
t
u
r
e
c
o
n
ﬁ
g
u
r
a
t
i
o
n
T
T
E
T
S

V
I
I
I
E
M
1
3
0
.
0
8
0
.
3
1
1
7
0
0
.
8
1
0
.
1
4
2
.
4
4
[
9
2
]
1
4
m
o
d
u
l
e
s
A
s
t
r
o
M
e
s
h
O
1
2
.
2
5
0
.
1
0
0
.
3
1
5
5
0
.
3
6
—
—
2
[
1
6
6
,
1
7
2
]
D
m
i
n
=
1
2
.
2
5
m
;
D
m
a
x
=
1
6
m
H
R
A
A
C
e
S
O
1
2
0
.
0
7
0
.
3
8
1
2
7
1
.
1
2
0
.
1
3
—
—
[
1
5
4
]
A
F
R
A
—
1
2
0
.
0
7
0
.
3
5
1
2
0
1
.
0
6
0
.
3
5
—
—
[
1
5
4
]
S
t
a
t
u
s
n
o
t
k
n
o
w
n
W
R
A
O
9
.
1
0
.
2
2
0
.
0
5
6
0
0
.
9
2
—
0
.
8
8
.
2
5
[
8
7
,
1
4
6
]
T
T
H
A
L
C
A
O
8
—
—
2
3
0
4
.
5
8
—
0
.
6
2
2
[
6
3
]
E
G
S
O
5
.
6
0
.
1
1
0
.
1
8
3
5
1
.
2
4
1
7
—
—
[
9
1
]
D
m
i
n
=
5
.
6
m
;
D
m
a
x
=
6
.
4
m
S
B
A
O
5
.
2
5
—
0
.
9
3
2
0
0
.
7
1
—
—
2
[
1
0
]
D
m
i
n
=
5
.
2
5
m
;
D
m
a
x
=
6
.
8
m
R
R
A
T
D
R
S
O
5
0
.
1
8
0
.
5
4
2
4
1
.
2
2
—
0
.
5
6
1
5
[
4
0
,
5
2
]
T
T
E
T
S

V
I
I
I
E
M
4
.
8
0
.
0
6
0
.
7
3
1
1
0
.
7
3
—
—
—
[
9
8
]
O
n
e
m
o
d
u
l
e
,
1
5
m
2
C
S
P
D
A
E
M
3
.
5
0
.
1
7
0
.
3
4
—
—
—
—
—
[
1
8
8
]
E
x
t
.
S
u
n
ﬂ
o
w
e
r
F
S
1
5
0
.
2
9
0
.
4
3
—
—
—
—
—
[
6
0
]
D
A
I
S
Y
E
M
8
0
.
3
6
0
.
5
1
—
6
—
0
.
0
0
8
3
0
0
0
[
5
1
,
1
3
4
]
S
S
u
n
ﬂ
o
w
e
r
E
M
4
.
9
0
.
4
4
0
.
3
7
3
1
1
.
6
4
—
0
.
0
5
1
6
0
[
5
1
,
5
2
]
M
E
A
E
M
4
.
7
0
.
3
6
0
.
5
1
9
4
5
.
4
2
—
0
.
2
3
0
[
5
2
,
1
5
3
]
S
S
D
A
E
M
1
.
5
0
.
3
7
0
.
5
4
—
—
—
—
—
[
5
1
]
A
R
I
S
E
F
S
2
5
—
—
1
9
2
0
.
3
9
0
.
3
0
.
5
8
6
[
2
0
]
T
a
r
g
e
t
s
u
r
f
a
c
e
a
c
c
u
r
a
c
y
I
I
S
R
S
E
M
2
0
—
—
1
3
4
0
.
4
2
3
.
2
0
.
9
8
2
2
[
1
3
5
]
I
A
E
O
1
4
0
.
0
6
0
.
1
4
6
0
0
.
3
9
4
1
.
0
0
—
[
4
3
]
d
/
D
=
4
1
.
1
0
.
4
6
/
π
/
1
4
≈
0
.
0
6
a
M
:
m
e
s
h
,
S
:
s
o
l
i
d
a
n
d
I
:
i
n
ﬂ
a
t
a
b
l
e
.
b
O
:
O
r
b
i
t
,
E
M
:
E
n
g
i
n
e
e
r
i
n
g
M
o
d
e
l
,
F
S
:
F
e
a
s
i
b
i
l
i
t
y
S
t
u
d
y
c
S
u
r
f
a
c
e
a
r
e
a
:
π
D
2
/
4
o
r
π
D
m
a
x
D
m
i
n
/
4
i
f
e
l
l
i
p
t
i
c
.
Chapter 3
Analysis Methods for Tensegrity
Structures
3.1 Introduction
A major obstacle in the analysis and design of tensegrity structures is the deter
mination of their equilibrium conﬁguration. This key step in the design procedure
is usually known as formﬁnding. For other tension structures, such as membrane
and cable nets, eﬃcient formﬁnding methods have been available for a long time,
cf. [5, 147]. For general tensegrity structures, however, the formﬁnding process has
proven to be more complicated.
Early studies, by Fuller [44, 88], Snelson [151] and Emmerich [37] into the form of
tensegrity structures, use mainly regular, convex polyhedra as the basis for ﬁnding
new conﬁgurations. This purely geometric research has resulted in a large number of
conﬁgurations which are classiﬁed by Pugh [132] into three pattern types: diamond,
circuit and zigzag. A large number of diﬀerent tensegrities, with detailed schemes
and advice on how to build them, are found in reference [132].
However, physical models of these structures show that the shape of the tensegrity,
corresponding to a particular polyhedron, is diﬀerent from that of the polyhedron.
This happens, for example, both for the truncated tetrahedron, Figure 3.1, and the
expandable octahedron (icosahedron), Figure 3.2. Hence, the selfstressed shape of
a tensegrity is not identical to that of the polyhedron and, therefore, proper form
ﬁnding methods are needed to ﬁnd the equilibrium conﬁguration of even the simplest
tensegrity structure [112].
Formﬁnding methods for tensegrity structures have been investigated by many au
thors, and recently by Connelly and Terrell [26], Vassart and Motro [177], and Sul
tan et al. [160]. Diﬀerent approaches are proposed by these authors, but the various
methods have not previously been classiﬁed or linked. In the ﬁrst two sections of
this chapter the existing methods were classiﬁed into two broad families—kinematic
and static methods—and advantages and limitations of each method were identi
ﬁed. Closer scrutiny of the seemingly diﬀerent approaches in references [26, 160, 177]
33
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
(a) Side view (b) Top view
Figure 3.1: Comparison of truncated tensegrity tetrahedron and the polyhedron
from which it originates (drawn with dashed lines). Note the distor
tion of the hexagonal faces.
(a) Side view (b) Top view
Figure 3.2: Comparison of expandable octahedron and icosahedron (drawn with
dashed lines).
revealed some links: indeed Vassart and Motro’s force density method was linked
directly to the more abstract energy approach by Connelly.
Once the formﬁnding step is completed, the static and kinematic properties, i.e.
the states of selfstress and internal mechanisms, must be found. A suitable method
34
3.2. KINEMATIC FORMFINDING METHODS
for this task is the Force Method developed by Pellegrino and Calladine [127]. As
this method will be used frequently in the remainder of the thesis, it is described in
the last part of this chapter.
3.2 Kinematic FormFinding Methods
The characteristic of these method is that the lengths of the cables are kept constant
while the strut lengths are increased until a maximum is reached. Alternatively, the
strut lengths may be kept constant while the cable lengths are decreased until they
reach a minimum. This approach mimics the way in which tensegrity structures are
built in practice, without explicitly requiring that the cables be put in a state of
pretension.
3.2.1 Analytical Solutions
Consider a simple structure consisting of cables arranged along the edges of a regular
prism, plus a number of struts connecting the v vertices of the bottom polygon to
vertices of the upper polygon. Depending on the value of v and the oﬀset between
vertices connected by a strut, there is a special rotation angle θ between the plane,
regular top and bottom polygons for which a tensegrity structure is obtained.
A compact description of the geometry of this problem, taking advantage of its
symmetry, was introduced by Connelly and Terrell [26], as follows. Figure 3.3 shows
the elements connected to one of the nodes of the bottom polygon. In the starting
conﬁguration the lateral cable, 12, is vertical and the angle between the ends of the
strut is 2πi/v, where i is an integer smaller than v, Figure 3.3.
The coordinates of nodes 1–5 are:
p
1
=
R 0 0
T
, (3.1a)
p
2
=
Rcos θ Rsin θ H
T
, (3.1b)
p
3
=
Rcos
θ +
2πi
v
Rsin
θ +
2πi
v
H
T
, (3.1c)
p
4
=
Rcos
2π
v
−Rsin
2π
v
0
T
, (3.1d)
p
5
=
Rcos
2π
v
Rsin
2π
v
0
T
. (3.1e)
The kinematic formﬁnding proceeds as follows, by considering the square of the
lengths of the lateral cable, 12, and strut, 13,
l
2
t
= 2R
2
(1 −cos θ) + H
2
, (3.2a)
l
2
c
= 2R
2
¸
1 −cos
θ +
2πi
v
+ H
2
, (3.2b)
35
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
R
H
θ
2πi
v
x
y
z
1
2
3
4
5
Figure 3.3: Elements meeting at node 1 of structure with vfold symmetry, radius
R and height H.
where the subscripts c and s denote cable and strut, respectively. Equation (3.2b)
can be rewritten as:
l
2
c
= 4R
2
sin
θ +
πi
v
sin
πi
v
+ l
2
t
. (3.3)
For a given cable length l
c
the length of the strut l
s
is maximised for
θ = π
1
2
−
i
v
. (3.4)
The simplicity of the kinematic method for structures with vfold symmetry is mir
rored by the static method, see section 3.3, but for other, nonsymmetric, cases the
present formulation becomes infeasible due to the large number of variables required
to describe a general conﬁguration.
3.2.2 NonLinear Programming
This general method, proposed by Pellegrino [121], turns the formﬁnding of any
tensegrity structure into a constrained minimisation problem. Starting from a sys
tem for which the element connectivity and nodal coordinates are known, one or
more struts are elongated, maintaining ﬁxed length ratios, until a conﬁguration is
reached in which their lengths are maximised. The general constrained minimisation
problem has the form:
Minimise f(x, y, z)
subject to g
i
(x, y, z) = 0 for i = 1, ..., n,
(3.5)
where the objective function f(x, y, z) is, for example, the negative length of one
of the struts and constraint equations g
i
(x, y, z) are the ﬁxed lengths of the cables.
Pellegrino [121] applies this method to two tensegrities: the triangular prism and
the truncated tetrahedron.
36
3.2. KINEMATIC FORMFINDING METHODS
A triangular prism, v = 3 and i = 1, has nine cables of length l
c
= 1 and three
struts of equal length. One of the base triangles is ﬁxed, hence three of its six nodes
are ﬁxed in space. In Cartesian coordinates, the constrained minimisation problem
has the form:
Minimise −l
2
s1
subject to
l
2
c1
−1 = 0
l
2
c2
−1 = 0
.
.
.
l
2
c6
−1 = 0
l
2
s2
−l
2
s1
= 0
l
2
s3
−l
2
s1
= 0
(3.6)
where c1, c2, ..., c6 denote the six remaining cables and s1, s2 and s3 the struts.
This problem can be solved, for example, using the constrained optimisation function
fmincon in Matlab [89]. The ﬁnal length of the struts is 1.468, compared to the
theoretical value of
1 + 2/
√
3 ≈ 1.4679 obtained from (3.3) and (3.4).
Similarly, for the case of the truncated tetrahedron there are six struts and 18 cables,
Figure 3.1. The objective function to be minimised, equal to the negative length
of one of the struts, has to satisfy 20 constraint equations, 15 on the cable lengths
plus 5 on the struts. The ﬁnal length of the struts is 2.2507. Note that the strut
length obtained from a purely geometric analysis of the ideal truncated tetrahedron
is
√
5 ≈ 2.2361; hence the slight warping of the hexagonal faces leads to an increase
of the strut length by 0.7%.
An advantage of the nonlinear programming approach is that it makes use of general
purpose, standard software. However, the number of constraint equations increases
with the number of elements so this approach is not feasible for larger systems. Also,
although diﬀerent geometric conﬁgurations of structures with the same topology can
be found by specifying diﬀerent relationships between the lengths of the struts, there
is no direct way of controlling the corresponding variation in the state of prestress.
3.2.3 Dynamic Relaxation
The method of dynamic relaxation, that had already been successfully used for
membrane and cable net structures [6, 31], was put forward by Motro [110] and
Belkacem [8] as a general formﬁnding method for tensegrity structures.
For a structure in a given initial conﬁguration and subject to given general forces
the equilibrium conﬁguration can be computed by integrating the following ﬁctitious
dynamic equations
M
¨
d +N
˙
d +Kd = f , (3.7)
where K is a stiﬀness matrix, M a mass matrix, N a damping matrix, f the vector
of external loads, and
¨
d,
˙
d and d, are the vectors of acceleration, velocity and
37
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
displacement from an initial conﬁguration, respectively. Both Mand Nare assumed
to be diagonal, for simplicity, and the velocities and displacements are initially set
to zero.
There are several ways of carrying out a formﬁnding analysis, for example by pre
scribing for each element of the structure a constitutive relationship of the type
t = t
0
+
¯
ke, (3.8)
where t is the axial force and e the extension, measured from the initial conﬁguration,
of the element; t
0
is the desired prestress and
¯
k a ﬁctitious, small axial stiﬀness. In
any current conﬁguration of the structure, nodal equations of equilibrium are used to
compute outofbalance forces from which the current acceleration can be obtained
through (3.7). The resulting system of uncoupled equilibrium equations can then
be integrated using a centred ﬁnite diﬀerence scheme.
The coeﬃcients of the damping matrix are usually all assigned the same value,
chosen such as to maximise the speed of convergence to the equilibrium conﬁgura
tion. Alternatively, a technique called kinetic damping can be used, whereby the
undamped motion of the structure is traced. When a local peak in the total kinetic
energy of the structure is detected, all velocity components are set to zero. The
process is then repeated, starting from the current conﬁguration, until the peak
kinetic energy becomes suﬃciently small [6]. The latter technique usually gives a
faster convergence.
Motro [110] applies the dynamic relaxation method to the formﬁnding of the trian
gular tensegrity prism. The lengths of the cables were held constant while the struts
were gradually elongated, until a state of prestress was set up in the structure. This
analysis converged to l
s
/l
c
= 1.468, as above.
Belkacem [8] analyses the triangular and square tensegrity prisms, and also the
expandable octahedron. The results for the tensegrity prisms are compared with
the theoretical values obtained from (3.3) and those for the expandable octahedron
to the results of a static method. The relative error ν in the nodal coordinates is,
for node i, computed as
ν
i
=
(x
i
− ˜ x
i
)
2
+ (y
i
− ˜ y
i
)
2
+ (z
i
− ˜ z
i
)
2
x
2
i
+ y
2
i
+ z
2
i
, (3.9)
where˜denotes the approximate values. This error is computed for each node and
the largest error is taken to represent the structure. The relative errors in the nodal
coordinates for the three structures are 0.2, 4 and 2%, respectively. For the tensegrity
prisms, errors in the rotation angle θ of 1 and 8%, respectively, are obtained. An
analysis of the truncated tetrahedron [111] yields a ratio l
s
/l
c
slightly greater than
2.24, which is close to the value determined by Pellegrino [121].
Motro et al. [114] later conclude that the dynamic relaxation method has good con
vergence properties for structures with only a few nodes but is not eﬀective when
the number of nodes increases. Also, the method becomes rather cumbersome if
several diﬀerent ratios between strut lengths and cable lengths are desired, which
38
3.3. STATIC FORMFINDING METHODS
restricts its applicability to less regular structural forms. However, the same restric
tion applies to kinematic methods in general.
3.3 Static FormFinding Methods
The general characteristic of these methods is that a relationship is set up between
equilibrium conﬁgurations of a structure with given topology and the forces in its
members. This relationship is then analysed by various methods.
3.3.1 Analytical Solutions
Kenner [70] uses node equilibrium and symmetry arguments to ﬁnd the conﬁguration
of the expandable octahedron, Figure 3.2, whose six identical struts are divided into
three pairs of struts which are mutually perpendicular. The distance between the
struts in each pair is exactly half the strut length. Other, more complex, spherical
tensegrities with polyhedral geometries, i.e. the cuboctahedron and the icosidodec
ahedron, were also analysed using the same approach.
Connelly and Terrell [26] use an equilibrium approach to ﬁnd the prestress stable
form of vfold symmetric tensegrity prisms. To set up a system of linear equilibrium
equations, they use force density
1
, i.e. force divided by length, as variable for each
element.
Denoting by q
ij
the force density in element ij—note that q
14
= q
15
due to symme
try—the equilibrium of node 1 in the z and ydirection can be written as
q
12
H + q
13
H = 0 (3.10)
and
q
12
Rsin θ + q
13
Rsin
θ +
2πi
v
= 0, (3.11)
respectively. Equations (3.10) and (3.11) give
q
12
¸
sin θ −sin
θ +
2πi
v
= 0. (3.12)
The only solution of (3.12) for which the cables are in tension is [26]
θ = π
1
2
−
i
v
. (3.13)
At the rotation given by (3.13) the resultant force from cable 12 and strut 13 is
radial. The values of θ for tensegrity prisms with v going from 3 to 6 are given in
Table 3.1. Note that (3.13) is identical to (3.4), as expected.
Nishimura [117] uses the force method, described in section 3.5 and Group Theory
to obtain closedform expressions for the equilibrium geometry of highly symmetric
spherical tensegrities. This is discussed further in section 3.4.3.
1
Also called tension coeﬃcient, cf. [152].
39
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
Table 3.1: Values of θ (
◦
) for v = 3, 4, 5, 6.
v
i
1 2 3 4 5
3 30 −30 — — —
4 45 0 −45 — —
5 54 18 −18 −54 —
6 60 30 0 −30 −60
3.3.2 Force Density Method
The force density method for cable structures, ﬁrst proposed by Linkwitz and Schek
in 1971 [85, 86, 147], uses a simple mathematical trick to transform the nonlinear
equilibrium equations of the nodes into a set of linear equations. For example, the
equilibrium equation in the xdirection for node i is
¸
j
t
ij
l
ij
(x
i
−x
j
) = f
ix
, (3.14)
where node i is connected to j nodes by cables or struts and t
ij
is the tension
in element ij. Although this may appear to be a linear equation in the nodal
coordinates, it is actually nonlinear because the lengths l
ij
in the denominator are
also functions of the coordinates. These equations can be linearised by introducing
for each element the force density
q
ij
= t
ij
/l
ij
, (3.15)
whose value needs to be known at the start of the formﬁnding process.
For a general structure with b elements and n nodes the equilibrium equations in
the xdirection can be written as
C
T
s
QC
s
x = f
x
, (3.16)
where C
s
is the incidence matrix, see below, Qa diagonal matrix containing the force
densities, x a column vector of xcoordinates, and f
x
a column vector of external
nodal forces in xdirection. Equations identical in form to (3.16) can be written also
in terms of the y and zcoordinates.
The incidence matrix C
s
, of size b n, describes the connectivity of the structure;
if an element connects nodes i and j, then the corresponding row of C
s
has +1 in
column i and −1 in column j. If the coordinates of some of the nodes are given, e.g.
these nodes are attached to a foundation, C
s
can be partitioned as
C
s
= [C
u
C
f
] , (3.17)
where the restrained nodes have been put at the end of the numbering sequence.
Equation (3.16) can now be written as
C
T
u
QC
u
x
u
= f
x
−C
T
u
QC
f
x
f
, (3.18)
40
3.3. STATIC FORMFINDING METHODS
where x
u
and x
f
are the column vectors of unknown and given xcoordinates, re
spectively. Equation (3.18), together with analogous equations for the y and z
directions, can be solved to ﬁnd the nodal coordinates. Usually, the external loads
are zero during formﬁnding.
In a structure consisting of cables only, e.g. cable net, all tension coeﬃcients are
positive, i.e. q
ij
> 0, and hence C
T
u
QC
u
is positive deﬁnite and, thus, invertible.
Therefore, there is always a unique solution to the formﬁnding problem. The same
approach can be extended to the formﬁnding of membrane structures by converting
the stresses in the membrane into forces in a virtual cable net [5, 81].
A similar formulation can be applied to the formﬁnding of tensegrity structures,
but as these structures are selfstressed, there are usually no foundation nodes or
external loads. Hence, (3.16) becomes
Dx = 0, (3.19)
where D = C
T
s
QC
s
. Analogous equations hold in the y and zdirections.
The force density matrix D can be written directly [177], without going through C
s
and Q, following the scheme
D
ij
=
−q
ij
if i = j,
¸
k=i
q
ik
if i = j,
0 if i and j are not connected.
(3.20)
Note that the n n matrix D is always singular, with a nullity, i.e. the dimension
of the nullspace, of at least 1 since the row and column sums are zero, by (3.20).
Unlike the matrix C
T
u
QC
u
for a cable net attached to foundation nodes, which is
positive deﬁnite, see page 120 of reference [147], the D matrix for a tensegrity is
semideﬁnite and, due to the presence of compression elements, with q
ij
< 0, several
complications arise during formﬁnding. A practical procedure for ﬁnding a set
of force densities that yield a matrix D with the required rank, was presented by
Vassart [175]. Further details will be given in section 3.4.
3.3.3 Energy Method
In the following, some key main ﬁndings by Connelly [22] will be summarised using
as far as possible the original terminology.
A conﬁguration of n ordered points in Tdimensional space is denoted by
P =
p
1
p
2
. . . p
n
. (3.21)
A tensegrity framework G(P) is the graph on P where each edge is designated as
either a cable, a strut or a bar; cables cannot increase in length, struts cannot
decrease in length and bars cannot change length. A stress state ω for G(P) is a
selfstress if the following condition holds at each node i:
¸
j
ω
ij
(p
j
−p
i
) = 0, (3.22)
41
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
where ω
ij
≥ 0 for cables, ω
ij
≤ 0 for struts, and no condition is stipulated for the
bars. Comparing (3.22) with the equilibrium equations for the same node written in
terms of force densities, it is obvious that the stresses ω
ij
are identical to the force
densities q
ij
.
Satisfying the above equilibrium condition is a necessary, but not suﬃcient, condition
for the tensegrity framework to be in a stable equilibrium conﬁguration. A basic
principle in the analysis of the stability of structures is that the total potential energy
functional should be at a local minimum for a given conﬁguration to be stable. In
analogy with the total potential, reference [21] deﬁnes the following energy form
associated with the stress ω:
E(P) =
1
2
¸
ij
ω
ij
p
j
−p
i

2
. (3.23)
The idea is that when the end points of an element are displaced, energy builds
up as a function of the square of the extension. The function in (3.23) is set up
to have an absolute minimum corresponding to the rest length of the element [22].
All members are assumed to behave as linear elastic springs. Mathematically, the
cables, which only take tension, have a rest length of zero while the struts, which
only take compression, have an inﬁnite rest length.
Let
p =
¸
x
y
z
(3.24)
be a column vector, of length Tn, containing the xcoordinates of P, followed by
the ycoordinates, etc. Then, (3.23) can be written as the quadratic form:
E(P) =
1
2
p
T
Ω
Ω
Ω
¸
¸
p, (3.25)
where the elements of Ω are given by
Ω
ij
=
−ω
ij
if i = j,
¸
k=i
ω
ik
if i = j,
0 if there is no connection between i and j.
(3.26)
Note that Ω is identical to D, hence the above formulation provides a deeper insight
into the characteristics of the force density method and how it can be used to ﬁnd
stable equilibrium conﬁgurations of tensegrity structures. The link between the force
density method and the energy minimisation was ﬁrst pointed out by K¨ otter [73]
and later by Schek [147].
A necessary condition for the tensegrity framework to be prestress stable in conﬁg
uration P is that the quadratic form E(P) has a local minimum at P. The positive
deﬁniteness of E(P) is directly related to that of Ω, but expecting positive deﬁ
niteness is unrealistic, because—as already noted above for D—the nullspace of Ω
contains at least the nontrivial vector (1 1 . . . 1)
T
.
42
3.3. STATIC FORMFINDING METHODS
1 2
3 4
Figure 3.4: Snelson’s Xframe
The strongest type of prestress stability, named super stability by Connelly [23],
requires prestress stability with the additional condition that Ω is positive semi
deﬁnite with maximal rank. The maximal rank of Ωfor a structure in Tdimensional
space that does not in fact lie in a subspace of smaller dimension, see examples in
section 3.4.1, is n − T − 1. Hence, to design a super stable tensegrity framework
one has to ﬁnd a set of force densities such that the nullity ^ of Ω is T + 1. A
further condition for super stability is that there are no aﬃne inﬁnitesimal ﬂexes
of the tensegrity framework G(P). Aﬃne inﬁnitesimal ﬂex is another name for a
linear inﬁnitesimal mechanism. Such a mechanism can be described by the linear
map d
i
= Ap
i
+ n, where A is a T T matrix and n a translation vector [22].
When describing a rigid body motion, A is skew symmetric, i.e. A = −A
T
.
For example, consider the twodimensional (T = 2) tensegrity structure in Figure 3.4
where the outside edges are cables and the diagonals are struts. A stress equal to 1
in the cables and −1 in the struts is a selfstress for this structure. The stress, i.e.
force density, matrix is [21]:
Ω =
1 −1 1 −1
−1 1 −1 1
1 −1 1 −1
−1 1 −1 1
¸
¸
¸
¸
, (3.27)
which is positive semideﬁnite with nullity 3. Hence the tensegrity structure in
Figure 3.4 is super stable [23].
Connelly and Back [24] analyse tensegrity structures with diﬀerent types of symme
try using this method. Their initial assumption is that there is a symmetric state of
selfstress with a force density of 1 in each cable and −ω
s
in each strut. A further
assumption is that there are two types of cables but only one type of strut, arranged
such that satisfying equilibrium at only one node of the structure implies, by sym
metry, that it is satisﬁed also at all other nodes. The force density in the strut is
chosen such that the structure is super stable [23, 24].
A complete catalogue of all the tensegrity structures that are possible for each sym
metry group is produced, using group theory. Although some of the structures in the
catalogue have struts that go through each other, and therefore are of limited prac
tical interest, the catalogue contains many solutions that were previously unknown.
43
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
(a) (b)
Figure 3.5: Symmetric tensegrities from Connelly and Back’s catalogue.
See Figure 3.5 for some structures from their catalogue.
3.3.4 Reduced Coordinates
This method was introduced by Sultan et al. [160]. Consider a tensegrity structure
whose b elements consist of M cables and O struts. The struts are considered as a set
of bilateral constraints acting on the cable structure. Hence, a set of N independent,
generalised coordinates g = (g
1
g
2
. . . g
N
)
T
is deﬁned, which deﬁne the position
and orientation of these struts.
2
Consider a state of selfstress for the structure and let t
j
be the axial force in a
generic cable element j; the cable forces t = (t
1
t
2
. . . t
M
)
T
are in equilibrium
with appropriate forces in the struts and zero external loads. A set of equilibrium
equations relating the forces in the cables, but without showing explicitly the forces
in the struts, can be obtained from virtual work.
Consider a virtual displacement δg of the structure that involves no extension of the
struts. The change of length of cable j is
δl
j
=
N
¸
i=1
∂l
j
∂g
i
δg
i
. (3.28)
Considering all cables, (3.28) gives
δl = A
T
δg, (3.29)
2
If T = 2 three generalised coordinates are required for each struts, hence N = 3 O; if T = 3
then N = 5 O.
44
3.3. STATIC FORMFINDING METHODS
where the elements of the N M matrix A are
A
ij
=
∂l
j
∂g
i
. (3.30)
Because the extensions of the struts are zero, the virtual work in the struts is also
zero and so the total internal work, from the cables only, is
t
T
δl = (At)
T
δg. (3.31)
For the structure to be in equilibrium, this must be zero for any virtual displacement
δg. This gives the following reduced equilibrium equations
At = 0. (3.32)
For this equation to have a nontrivial solution it is required that
rank A < M, (3.33)
where only solutions that are entirely positive are of interest, i.e.
t
j
> 0 for j = 1, 2, ..., M. (3.34)
General analytical conditions that govern the form of a tensegrity structure of given
topology can be obtained by analysing (3.33) and (3.34).
Sultan [159] applies this method to a tensegrity mast of which Figure 3.6 shows a
simple twostage example. The same structure had been previously considered by
Snelson [151]. The mast consists of three struts per stage, held in place by three
sets of cables—saddle, vertical, and diagonal—between two rigid triangular plates
at the top and bottom; in Figure 3.6 note the deﬁnition of the overlap h. Having
shown that a structure in which the rigid plates have been replaced by cables, has
the same equilibrium conﬁguration as the original structure, but involves a smaller
number of unknown cable forces, Sultan analyses this simpler problem.
The ﬁrst step in the formﬁnding process is to identify a set of generalised coordinates
which describes the conﬁguration of this structure. The 18 coordinates chosen by
Sultan [159] for the twostage mast are
• for each strut, the azimuth angle α
j
, i.e. the angle between the vertical plane
containing the strut and the x–z plane, and the colatitude δ
j
, i.e. the angle
between the strut and the zaxis, and
• three translation and three rotation parameters deﬁning the position and ori
entation of the rigid plate at the top with respect to the bottom plate.
By using symbolic manipulation software, e.g. Maple or Mathematica, the length of
each cable can be expressed in terms of the 18 coordinates and then diﬀerentiated to
obtain the 18 18 matrix A, in symbolic form. At this stage, the ﬁnal shape of the
structure is still unknown and the existence of a prestressable conﬁguration is de
pendent on ﬁnding a suitable set of strut lengths. Sultan [159] reduces the number of
45
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
α
x
x
y
y
z
Base
D
i
a
g
o
n
a
l
V
e
r
t
i
c
a
l
S
a
d
d
le
S
t
r
u
t
h
(a)
(b)
(c)
Figure 3.6: Sultan and Skelton’s two stage tensegrity tower: (a) threedimensional
view, (b) side view, and (c) top view.
independent generalised coordinates by considering only symmetric conﬁgurations,
with the same azimuth, α, and colatitude, δ, and by considering a ﬁxed position of
the top plate. Then, by assuming a special symmetry in t, the problem could be
reduced even further, to 3 3 with the forces in the diagonal, saddle and vertical
cables remaining as the only unknowns. Finally, applying to this reduced matrix
the condition for the existence of nontrivial solutions, rank A = 2 equivalent to
det A = 0, (3.35)
yields a quadratic equation that could be solved for the overlap h:
h =
1
2 tan δ cos
α +
π
6
a
2
3
−3l
2
sin
2
δ cos
2
α +
π
6
−
a
√
3
+ l sin δ cos
α +
π
6
if α =
π
3
,
l cos δ
2
if α =
π
3
.
(3.36)
Here, a is the side length of the equilateral triangles at the top and bottom of the
mast, and l the length of each strut. A particular symmetric conﬁguration, in which
all the nodes lie on the surface of a cylinder, is deﬁned by the following relationship
between δ and α:
δ = arcsin
¸
2a
l
√
3
sin
α +
π
3
. (3.37)
46
3.4. IMPLEMENTATION OF THE FORCE DENSITY METHOD
1
2 3
4
5 6
1
2
3
4
5
6
7
8
9
Figure 3.7: Twodimensional hexagonal tensegrity.
For masts with more than two stages, A was derived using symbolic software, but
(3.35) was then solved numerically. Sultan [159] successfully applied this form
ﬁnding method to masts with up to nine stages.
3.4 Implementation of the Force Density Method
The force density method has been outlined in section 3.3.2. This section will deal
with procedures to actually ﬁnd superstable tensegrities, i.e. with positive semi
deﬁnite matrix D with nullity ^ = T + 1. Vassart and Motro [177] list three
techniques for ﬁnding a set of force densities that achieve the required nullity: (i)
intuitive, (ii) iterative and (iii) analytical.
Of these three techniques, the ﬁrst is suitable for systems with only a few members,
and will be illustrated in section 3.4.1; the second technique is based on a trialand
error, or more reﬁned search for a set of force densities that yield the required nullity.
The third technique is the most eﬀective; D is analysed in symbolic or semisymbolic
form, in the case of systems with a large number of elements [177]. The following
examples show how this is done in practice, for structures of increasing complexity.
3.4.1 A TwoDimensional Example
Consider the hexagonal tensegrity shown in Figure 3.7. For it to be super stable, the
nullity of D has to be three, but it is interesting to consider also the cases ^ = 1, 2
to better understand why in section 3.3.3 it was stated that one must look for sets
of force densities that make rank D = n −T −1.
Case ^ = 1
Most sets of force densities yield a D matrix with nullity one. For example, the
47
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
arbitrary set q
1
= (1 2 3 4 5 6 −7 −8 −9)
T
produces
D
1
=
0 −1 0 7 0 −6
−1 −5 −2 0 8 0
0 −2 −4 −3 0 9
7 0 −3 0 −4 0
0 8 0 −4 1 −5
−6 0 9 0 −5 2
¸
¸
¸
¸
¸
¸
¸
¸
. (3.38)
The nullspace of D
1
is spanned by (1 1 1 1 1 1)
T
—see reference [156] for further
details on how to compute a basis for a nullspace—which, through (3.19) and the
analogous equation in the ycoordinates, give the conﬁguration x = (x
1
. . . x
6
)
T
=
(α α α α α α)
T
and y = (y
1
. . . y
6
)
T
= (β β β β β β)
T
. Here, α and β can
take arbitrary values. This solution corresponds to conﬁgurations of the structure
where all the nodes coincide and so the whole structure is reduced to a single point,
which is of little practical interest.
Case ^ = 2
Next, uniform force densities in all cable elements and in two of the struts force
densities of half those in the cables were prescribed; the force density in the third
cable was arbitrary. For example, for q
2
= (2 2 2 2 2 2 −1 −1 −3)
T
D
2
=
3 −2 0 1 0 −2
−2 3 −2 0 1 0
0 −2 1 −2 0 3
1 0 −2 3 −2 0
0 1 0 −2 3 −2
−2 0 3 0 −2 1
¸
¸
¸
¸
¸
¸
¸
¸
. (3.39)
It can be readily veriﬁed that columns ﬁve and six are dependent, and hence that
the nullspace of D
2
is spanned by (−1 −1 0 1 1 0)
T
and (2 2 1 0 0 1)
T
. Hence,
denoting by α, β the xcoordinates of nodes 5, 6 respectively, and by γ, δ their y
coordinates, the conﬁguration of the system is described by x = (−α + 2β −α + 2β
β α α β)
T
and y = (−γ +2δ −γ +2δ δ γ γ δ)
T
. This conﬁguration corresponds
to all nodes lying on a straight line, as shown in Figure 3.8(a), and is again of little
practical interest.
Case ^ = 3
Finally, uniform force densities both in the cable elements and in the struts, in a
ratio of two to one, were prescribed. For example, q
3
= (2 2 2 2 2 2 −1 −1 −1)
T
yields
D
3
=
3 −2 0 1 0 −2
−2 3 −2 0 1 0
0 −2 3 −2 0 1
1 0 −2 3 −2 0
0 1 0 −2 3 −2
−2 0 1 0 −2 3
¸
¸
¸
¸
¸
¸
¸
¸
. (3.40)
Here, columns four, ﬁve and six are dependent, hence the nullspace of D
3
is spanned
by (1 2 2 1 0 0)
T
, (−2 − 3 − 2 0 1 0)
T
, and (2 2 1 0 0 1)
T
. Denoting
48
3.4. IMPLEMENTATION OF THE FORCE DENSITY METHOD
1 2
3
4 5
6
,
,
,
(a) ^ = 2
1
2
3
4
5
6
(b) ^ = 3
1
2 3
4
5 6
(c) ^ = 3
Figure 3.8: Conﬁgurations of twodimensional hexagonal tensegrities.
by α, β, γ the free xcoordinates of nodes 4, 5, 6, respectively, and δ, , ζ their
ycoordinates, the system conﬁguration is given by x = (α − 2β + 2γ 2α − 3β +
2γ 2α−2β +γ α β γ)
T
and y = (δ −2 +2ζ 2δ −3 +2ζ 2δ −2 +ζ δ ζ)
T
.
The original solution in Figure 3.7 is reobtained for α = −1, β = −1/2, γ =
1/2, δ = 0, = −
√
3/2, and ζ = −
√
3/2. However, note that, despite the force
densities q
3
being symmetric, this solution also produces the conﬁguration shown
in Figure 3.8(b), which has only twofold symmetry, for α = −1, β = −1/2, γ = 1,
δ = 0, = −
√
3/2, and ζ = −1/2. The reason why it is possible to ﬁnd less
symmetric or even asymmetric conﬁgurations for a given, symmetric state of force
densities is because the element lengths are not explicitly set in the force density
formulation.
In concluding, it is noted that the particular q
3
considered above was obtained after
noticing that in the conﬁgurations shown in Figure 3.7 the force densities must have
a particular distribution, to satisfy nodal equilibrium. However, by carrying out
a symbolic analysis of the force density matrix other solutions were subsequently
found. For example, an alternative choice was q
4
= (1 2 1 2 1 2 −2/3 −2/3
−2/3)
T
, for which a particular conﬁguration (with α, β, etc. as in the original
conﬁguration) is that shown in Figure 3.8(c).
3.4.2 Tensegrity Prisms
Consider a structure with the topology shown in Figure 3.9. A set of force densities
with threefold symmetry is prescribed as follows. The force densities in the cable
forming the bottom and top triangles are q
b
and q
t
, respectively; they are q
l
in the
lateral cables and q
s
in the struts. Assuming that the top and bottom triangles are
parallel, equilibrium perpendicular to the planes of the triangles gives q
s
= −q
l
, cf.
(3.10).
49
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
1
2
3
4
5
6
2
3
4
6
8
7
9
10
12
11
5
1
Figure 3.9: Tensegrity prism.
Hence, the D matrix can be set up in terms of only three force densities
D =
2q
t
−q
t
−q
t
−q
l
0 q
l
−q
t
2q
t
−q
t
q
l
−q
l
0
−q
t
−q
t
2q
t
0 q
l
−q
l
−q
l
q
l
0 2q
b
−q
b
−q
b
0 −q
l
q
l
−q
b
2q
b
−q
b
q
l
0 −q
l
−q
b
−q
b
2q
b
¸
¸
¸
¸
¸
¸
¸
¸
. (3.41)
By Gaussian elimination D is reduced to the upper echelon form [156]
U =
q
l
0 −q
l
−q
b
−q
b
2q
b
0 q
l
−q
l
q
b
−2q
b
q
b
0 0 0 −q
∗
0 q
∗
0 0 0 0 q
∗
−q
∗
0 0 0 0 0 0
0 0 0 0 0 0
¸
¸
¸
¸
¸
¸
¸
¸
, (3.42)
where q
∗
= (q
2
l
− 3q
b
q
t
)/q
l
. Since q
l
= 0, rank D is either four, if q
∗
= 0, or two, if
q
∗
= 0. If super stability is required, then ^ = 4, i.e. rank D = 2 and so q
∗
= 0.
Any set of positive cable force densities that satisﬁes the condition
q
2
l
−3q
b
q
t
= 0 (3.43)
is possible, but Vassart [175] presents two interesting cases: (i) q
t
= q
b
and (ii)
q
t
= q
l
. In both cases, the last four coordinates can take arbitrary values; denoting
those xcoordinates, for example, by α, β, γ, δ, in case (i) the conﬁguration of the
structure is described by x = (α+(β+γ−2δ)/
√
3 α+(−β+2γ−δ)/
√
3 α β γ δ)
T
,
similarly for y and z. In case (ii) the conﬁguration of the structure is described by
50
3.4. IMPLEMENTATION OF THE FORCE DENSITY METHOD
1
2
3
4
5
6
(a) q
t
= q
b
, q
l
=
√
3q
b
, and q
s
= −q
l
1
2
3
4
5
6
(b) q
t
= q
l
, q
b
= q
l
/3, and q
s
= −q
l
Figure 3.10: Top views of two diﬀerent rotationally symmetric tensegrity prisms
x = (α + (β + γ − 2δ)/3 α + (−β + 2γ − δ)/3 α β γ δ)
T
. The rotationally
symmetric conﬁgurations, obtained by giving appropriate values to α, etc. are shown
in Figure 3.10.
Changing the relationship between q
t
and q
b
while keeping q
l
ﬁxed, changes the
relative sizes of the top and bottom triangles. Again, many geometrically non
symmetric conﬁgurations may be found by appropriate choices of the free nodes.
3.4.3 Spherical Tensegrities
The earlier part of this section has shown applications of the force density method
to the formﬁnding of some relatively straightforward tensegrity structures; several
symmetric conﬁgurations that had already been found by other methods were thus
reobtained. Further applications of the same method, to slightly more complex
systems will be presented next.
Expandable octahedron
For the expandable octahedron, Figure 3.2, there is only one type of cable and one
type of strut. Earlier analysis, cf. [70], has shown that the distance between parallel
struts is half the length of the strut, equilibrium in the strut direction yields q
s
=
−3q
c
/2. An analysis of D produces two possible solutions for ^ = 4: q
s
= −3q
c
/2
and q
s
= −2q
c
, but D is positive semideﬁnite only for the ﬁrst one. In addition to
the symmetric conﬁguration of Figure 3.2, many asymmetric conﬁgurations can be
found.
Truncated tetrahedron
An equilibrium conﬁguration for the truncated tetrahedron, Figure 3.1, was found
51
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
in reference [121]. For that conﬁguration, the single state of selfstress was com
puted by the force method described in section 3.5. For a force of 1 in the cable
elements forming the triangle faces, there is a force of 1.3795 in the remaining ca
bles, and −1.5016 in the struts. Since the strut and cable lengths are 2.2507 and 1,
respectively, the corresponding force densities are 1, 1.3795 and −0.6672.
As shown in Figure 3.1, the nodes of the truncated tetrahedron lie in four diﬀer
ent horizontal planes. An independent formﬁnding study showed that, unlike the
structures analysed so far, no relationships between the force densities could be
obtained from equilibrium statements without considering the geometric symmetry
conditions. Hence, the general state of selfstress is characterised by three diﬀerent
force density values: q
t
for cables forming the triangles, q
l
for the other cables, and
q
s
for the struts.
By setting q
l
= γq
t
, and after carrying out a Gaussian elimination on the matrix D,
the condition for ^ = 4 was found to be
2 (1 + γ)
q
s
q
t
2
+ [3 + 2γ (3 + γ)]
q
s
q
t
+ γ (3 + 2γ) = 0. (3.44)
To reobtain the earlier results γ = 1.3795 and thus q
s
/q
t
= −0.6672, and also
a second solution q
s
/q
t
= −2.5022. However, for that solution the D matrix has
negative eigenvalues, and hence the corresponding conﬁgurations are unstable.
For diﬀerent values of γ, e.g. γ = 1 as in reference [24], diﬀerent conﬁgurations, with
unequal cable lengths, of the truncated tetrahedron were obtained. However, it was
diﬃcult to specify the ratio γ to ﬁnd a conﬁguration with a particular ratio between
the lengths of the cables. Therefore, it is concluded that the force density method is
an excellent method for ﬁnding the conﬁguration of new tensegrities, but less than
ideal for structures with some known, or desired element lengths.
To complete the formﬁnding of spherical tensegrities it should be mentioned that
Nishimura [117] uses the force method together with group theory to ﬁnd the initial
equilibrium conﬁgurations of such structures. This method resembles the approach
used by Connelly and Back [24]. With Nishimura’s contribution, the formﬁnding
of tensegrities corresponding to the truncated versions of the following regular poly
hedra: tetrahedron, octahedron, cube, icosahedron, and dodecahedron, is ﬁnally
solved analytically. In the case of the truncated tetrahedron, the condition for a
nontrivial selfstress is [117]
4κ
2
cos α + 4 cos α
1 + cos
2
α
+ 4
√
3 sin α(cos α −1)
−12κ
cos
2
α + cos α
√
3 sin α + 2
−2
√
3 sin α
+ 9
cos α −
√
3 sin α
= 0,
(3.45)
where κ is the ratio of the side length of the truncating tetrahedron and the side
length of the original tetrahedron, a, and α is the angle of rotation of the triangles
with respect to an inertial system xyz. Nishimura [117] notes that as κ → 0,
52
3.5. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS
α →π/6, and as κ →1/2, α →0. The lengths of the cables forming the truncating
triangles, the cables connecting the triangles, and the struts are
l
t
= κa, (3.46)
l
l
=
a
3
9 + 24κ(κ −1) −4κ[(3 −4κ) cos α + κcos 2α], (3.47)
l
s
=
a
√
18
9 +
3 −4κ −2κ
cos α −
√
3 sin α
2
, (3.48)
respectively. The speciﬁc case where the lengths of the cables are equal is considered.
Simultaneously solving (3.46)–(3.48) for a, κ, and α subject to l
l
= l
t
= 1 yields
a = 2.9873, κ = 0.3348, and α = 0.1127. Substituting these value into (3.48)
yields the strut length l
s
= 2.256274. This value is very close to 2.2507, found by
Pellegrino [121]. The state selfstress of the truncated tetrahedron, expressed in
terms of force densities, is
q
t
= −q
s
3 −4κ(1 −cos α)
6κcos α
, (3.49)
q
l
= −q
s
cos α +
√
3 sin α
[3 −4κ(1 −cos α)]
cos α
6 −4κ
2 + cos α +
√
3 sin α
. (3.50)
Inserting the present values into (3.49) and (3.50) gives q
l
/q
t
= 1.379421 and q
s
/q
t
=
−0.667142, again in accordance with the previous analysis.
3.5 The Force Method for Analysis of Bar Frame
works
Once the prestress stable form of the tensegrity structure has been found its kine
matic and static properties are sought. As described in Chapter 1, a framework of
pinjointed bars, with three degrees of freedom at each joint, can be completely char
acterised by the extended Maxwell’s rule, (1.2). Three of the ﬁve parameters, the
number of bars, joints and kinematic constraints, are prescribed but the other two,
the number of selfstress states and internal mechanisms, depend on the geometrical
conﬁguration of the framework.
A method to investigate the properties of a general framework of pinjointed bars
is developed by Pellegrino and Calladine in a series of articles [14–16, 122, 124, 127].
The method, which is now known as the force method, is ﬁrst described in [127].
Kuznetsov [74] found that the method, as presented, was incomplete as it failed to
correctly analyse relatively simple frameworks. It included the matrix rank condi
tions that are necessary, but not suﬃcient, for prestress stability; the basic positive
deﬁnitiveness conditions, however, were missing. Following the criticism, a reﬁned
version of the method was presented in [15]. Kuznetsov [75] came up with what
looked as a counterexample to the improved method, but the questions around that
were ﬁnally resolved in [16]. In reference [127], the computational scheme to ﬁnd the
53
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
number of selfstress states and mechanisms was based on Gaussian elimination, but
such a scheme will have problems with illconditioned equilibrium matrices. Pelle
grino [124] presents a scheme based on the Singular Value Decomposition (SVD) of
the equilibrium matrix. The SVD scheme is computationally more expensive than
the Gaussian elimination scheme, but has several advantages which will be explained
later.
Another method of analysing the properties of bar frameworks is presented by
Kuznetsov in [76], based on a series of articles by him over the past three decades. In
comparison, the method does not lend itself to eﬃcient numerical analysis as well as
the previous one. Therefore, the previous method seems more advantageous, because
what really is needed in the static and kinematic analysis of very large frameworks,
involving several hundred elements, is a robust and eﬃcient numerical method. In
the next ﬁve sections the method by Pellegrino and Calladine [14–16, 122, 124, 127]
is described.
3.5.1 Equilibrium and Compatibility Matrices
Consider a threedimensional pinjointed bar framework with j joints and b bars
acted upon by external forces at the joints. The framework is in its deployed state
and restrained in the threedimensional space by c kinematic constraints. If the
structural deformations under the external loads are small, a linear analysis is suf
ﬁcient. The set of equilibrium equations for the framework is written as
Ht = f , (3.51)
where H is the 3j − c b equilibrium matrix containing the direction cosines, in
the x, y and z directions, of each element, t is the internal force vector of length b,
and f is the external force vector of length 3j − c. The external forces give rise to
displacements which must be compatible with the elongations of the bars. The set
of linear compatibility equations is
Cd = e (3.52)
where C is the b 3j − c compatibility matrix, d is the joint displacement vector
of length 3j − c, and e is the bar elongation vector of length b. The work done by
the external loads is
1
2
f
T
d and the strain energy stored in the framework is
1
2
t
T
e.
Equating the external work and strain energy yields
H
T
= C. (3.53)
Equation (3.53) expresses the static–kinematic duality of bar frameworks [124, 155].
3.5.2 Static and Kinematic Properties
The equilibrium and compatibility matrices contain essential information about the
framework. Wellknown in the linearalgebraic treatment of matrices are the four
54
3.5. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS
Dim. Equilibrium H Compatibility C
r
H
Row space:
bar tensions in
equilibrium with the
loads in the column
space.
=
Column space:
compatibility bar
elongations
Bar space
1
b
⊥
⊥
s
Nullspace:
states of selfstress.
(Solutions of Ht = 0)
=
Left nullspace:
incompatible bar
elongations.
r
H
Column space: loads
which can be
equilibrated in the
initial conﬁguration.
=
Row space:
extensional
displacements.
Joint space
1
3j−c
⊥
⊥
m
Left nullspace: loads
which cannot be
equilibrated in the
initial conﬁguration
=
Nullspace:
inextensional
displacements.
(Solutions of Cd = 0)
Figure 3.11: The four fundamental subspaces associated with H and C [122].
vector subspaces. The physical signiﬁcance of the four vector subspaces associated
with H and C are brieﬂy explained in Figure 3.11; for further information on sub
spaces see [156]. The standard way to ﬁnd the bases for the four subspaces is through
Gaussian elimination. Another approach, suggested by Pellegrino [124], is the SVD,
in which the equilibrium matrix H is factorised as
H = UΣW
T
, (3.54)
where U is a 3j −c 3j −c orthogonal matrix, W is a b b orthogonal matrix, and
Σ is a 3j −c b matrix with r
H
positive elements σ
ii
(i = 1, ..., r
H
) on the leading
diagonal and all other elements zero. The coeﬃcients of Σ are the singular values of
H, which are the square roots of the eigenvalues of HH
T
and H
T
H. The rank r
H
of the equilibrium matrix H is the number of nonzero diagonal elements in Σ, cf.
the number of pivots after a Gaussian elimination of H. In practice, the diagonal of
55
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
Table 3.2: Classiﬁcation of structural assemblies.
Assembly Static and kinematic properties
type
I s = 0 Statically determinate and
m = 0 kinematically determinate
II s = 0 Statically determinate and
m > 0 kinematically indeterminate
III s > 0 Statically indeterminate and
m = 0 kinematically determinate
IV s > 0 Statically indeterminate and
m > 0 kinematically indeterminate
Σ often contains up to min(3j −c, b) singular values of decreasing magnitude; none
of these values are actually equal to zero, but some are much smaller than others.
Values smaller than a set tolerance have to be treated as zero. Pellegrino [124] sets
the tolerance to 10
−3
σ
11
for most structural assemblies, but notes that a lower
tolerance may be required near critical points where the rank of the equilibrium
matrix drops. A wide gap in an otherwise continuous range of singular values is
another sign to look for. Once r
H
has been decided the number of selfstress states
and mechanisms are
s = b −r
H
(3.55)
and
m = 3j −c −r
H
. (3.56)
A basis S for the states of selfstress is given by the last s columns of W,
S =
w
r
H
+1
. . . w
b
. (3.57)
Similarly, a basis D for the mechanisms is given by the last m columns of U,
D =
u
r
H
+1
. . . u
3j−c
. (3.58)
Pellegrino [122] classify structural assemblies into four groups depending on their
degree of static and kinematic indeterminacy, Table 3.2. What matters in the clas
siﬁcation is only if s or m is zero or not.
3.5.3 RigidBody Mechanisms
The mechanisms in D can either be internal mechanisms or rigidbody mechanisms,
arising from inadequate kinematic restraints of the structure. These two types of
mechanisms are fundamentally diﬀerent but the SVD algorithm makes no distinction
between them. In a case when the structure is not fully restrained, D generally
contains combinations of the two mechanism types. It is therefore not possible to
just pick out the rigidbody mechanisms from D. A scheme to separate the internal
56
3.5. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS
mechanisms from the rigidbody ones was proposed by Pellegrino [121, 127]. It can
cope with up to six rigidbody mechanisms.
Any rigidbody displacement in threedimensional space may be described by a
translation n and a rotation r, where
n =
n
x
n
y
n
z
T
, (3.59a)
r =
r
x
r
y
r
z
T
. (3.59b)
The displacement d
i
of a point i in such a rigidbody motion is given by
d
i
= n +r p
i
, (3.60)
where p
i
is the position vector of point i in the original conﬁguration. If the structure
has a total of c constrained degrees of freedom, the system of c equations in six
unknowns is
R
n
r
= 0, (3.61)
where R has size c 6. The rank r
R
of R counts how many of the c kinematic
constraints that suppress the rigidbody degrees of freedom. Thus, the number of
rigidbody mechanisms is
m
rb
= 6 −r
R
. (3.62)
A basis for the nullspace of R is an independent set of m
rb
rigidbody motions
in terms of n and r. The corresponding set of rigidbody mechanisms to those
motions is found by (3.60). A matrix D
rb
of size m
rb
3j − c containing the
rigidbody mechanisms is formed. Now each of the rigidbody mechanisms of D
rb
has to be removed from each column of D. This is done by the GramSchmidt
orthogonalisation procedure [156]: for each mechanism d
j
(j = 1, ..., m) each
independent rigidbody mechanisms d
rb,k
(k = 1, ..., m
rb
) is removed by use of the
formula
d
j
:= d
j
−
m
rb
¸
k=1
d
T
rb,k
d
j
d
T
rb,k
d
rb,k
d
rb,k
. (3.63)
D is now transformed into a set of internal mechanisms, but only m−m
rb
of them
are independent. A subsequent Gaussian elimination step can be used to remove
those dependent on others [121].
When dealing with complex frameworks it is important to be able to diﬀerentiate
between rigidbody mechanisms and internal mechanisms. However, the main aim
here is to eﬃciently restrain any structure so that D does not contain any rigidbody
motions.
3.5.4 Internal Mechanisms
Once the structure has been adequately ﬁxed, the remaining mechanisms are only the
internal ones. Physically, there are two types of internal mechanisms: inﬁnitesimal
and ﬁnite. In a ﬁnite mechanism the joints can move with no change in lengths of
57
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
the bars, while in an inﬁnitesimal mechanism there are some small length changes
of the bars when the joints move. In terms of displacements, this change in length
is, generally, of second or higher order. Hence, inﬁnitesimal mechanisms tighten
up when activated, as shown in the simple example in section 1.3. Also shown in
that example was that a state of selfstress gave the assembly a positive ﬁrstorder
stiﬀness. If a state of selfstress can impart positive ﬁrstorder stiﬀness to every
mechanism of a structure, then the mechanisms are ﬁrstorder inﬁnitesimal. Such
inﬁnitesimal mechanisms are associated with secondorder changes of bar lengths.
However, mechanisms which cannot be stabilised by a state of selfstress are of
second or higher order or are ﬁnite. Thus, if an assembly has some mechanisms they
must be ﬁrstorder inﬁnitesimal to ensure the stability of the structure.
Despite the stability issues related to second and higherorder mechanisms, they
have been given some attention; Connelly and Servatius [25] discuss the various
deﬁnitions of order of rigidity and mechanisms in literature and redeﬁne higher
order rigidity. Vassart et al. [176] present an analytical method in two steps, i.e.
geometry and energy characterisation, which can determine the order of the mech
anisms and identify ﬁnite mechanisms. Now if the only objective is to determine if
the mechanisms are ﬁrstorder inﬁnitesimal or not there is a better approach pro
posed by Calladine and Pellegrino [15, 122]. For each state of selfstress t
i
and each
mechanism d
j
a vector of geometric loads g
ij
can be computed as
g
ij
= B
d
j
t
i
, (3.64)
where B
d
j
is similar to the equilibrium matrix H but with coeﬃcients of type (d
j
kx
−
d
j
lx
)/l
kl
instead of (x
k
− x
l
)/l
kl
, i.e. displacements of mechanism j instead of nodal
coordinates. The m geometric loads for selfstress state i form the columns of the
geometric load matrix G
i
of size 3j − c m. Each state of selfstress has its own
matrix G. For assemblies with s = 1 and m > 0, the test for ﬁrstorder mechanisms
is to check that the symmetric matrix G
T
D is positive or negative deﬁnite. A
negative deﬁnite G
T
D corresponds to a selfstress −t. For assemblies with s > 1,
the test becomes slightly more complicated as a linear combination of all matrices
G
T
D, each corresponding to a selfstress state, which is positive deﬁnite has to be
found. For simple problems this can be done symbolically, but for larger problems,
an automatic numerical routine is required. Such a routine is presented by Calladine
and Pellegrino [15]. However, they later found a certain framework with two states
of selfstress and two internal mechanisms, one of ﬁrstorder and one of higherorder.
This framework is classiﬁed as a ﬁrstorder inﬁnitesimal mechanisms but still cannot
be stabilised by a state of selfstress [16].
At this point, the tools to completely classify a pinjointed bar framework have been
provided. What is left is to analyse the response of the framework to external loads.
This is normally done by the Finite Element Method (FEM) but can, for certain
frameworks, be done by the present method. In many cases this method gives more
insight into the problem.
58
3.5. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS
3.5.5 Structural Computations
Pellegrino [122,124] presents a method for computing the forces and displacements of
a pinjointed bar framework due to external loads. The assumptions of the method,
as it is presented below, are that geometrical and material nonlinearities can be
neglected. Hence, the method is linear and relates to the initial conﬁguration of
the assembly, which in the case of a deployable structure is equal to the deployed
one. This method has been found to be very accurate for assemblies of type II,
Table 3.2. Assemblies of type IV, however, undergo signiﬁcant increases of prestress
during loading and deform less than predicted by the linear method. Nevertheless,
the linear method is presented below with respect to a type IV assembly.
First, the force and displacement systems have to be connected: assuming linear
elastic material, the element forces t are related to the bar elongations e by the
ﬂexibility matrix Φ as
e = e
0
+Φt, (3.65)
where e
0
is the vector of initial bar elongations. The b b diagonal ﬂexibility matrix
Φ has φ
i
= l
i
/A
i
E
i
as its entry of position (i, i).
Before starting the computations of the internal bar forces t, the bar elongations
e, and the displacements d one has to check that the framework can carry the
load f ; the external load f must be zero in the subspace of loads which cannot be
equilibrated, cf. Figure 3.11:
D
T
f = 0. (3.66)
If the framework passes this test, the general solution to (3.51) is
t = t
f
+Sα, (3.67)
where t
f
contains the internal forces in equilibrium with the load f and α is a vector
of s free parameters. t
f
is eﬃciently computed as
t
f
=
r
H
¸
i=1
u
T
i
f
σ
ii
w
i
. (3.68)
The value of α is determined by the following compatibility condition
S
T
e = 0, (3.69)
stating that the elongation e must vanish in the subspace of incompatible strains.
Substituting (3.65) and (3.67) into (3.69) and solving for α yields
α = −(S
T
ΦS)
−1
S
T
(e
0
+Φt
f
) (3.70)
Finally, substituting α into (3.67) gives the resulting internal forces t.
In analogy with (3.67) the general solution to (3.52) is
d = d
e
+Dβ, (3.71)
59
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
l
l
l
l
l
l
z
x
y
W
W
W
1
2
3
1
2
3
4
5
6
Figure 3.12: Hanging triangular net
where d
e
is any vector of displacements compatible with e and β is a vector of m
free parameters. Similarly to t
f
, d
e
is computed as
d
e
=
r
H
¸
i=1
w
T
i
e
σ
ii
u
i
. (3.72)
Note that the term Dβ in (3.71) represents a general inextensional displacement
satisfying Cd = 0. β is determined by the orthogonality between the matrix of
geometric loads G and d,
G
T
d = 0, (3.73)
which is obtained from virtual work considerations. Substituting (3.71) into (3.73)
and solving for β yields
β = −(G
T
D)
−1
G
T
d
e
. (3.74)
Substituting β back into (3.71) gives the resulting joint displacements d.
3.5.6 Example: Hanging Triangular Net
Consider the hanging triangular net in Figure 3.12 with six pinjointed bars. Three
nodes are fully ﬁxed in space. The horizontal projections of the members all have
equal length l and the middle triangle lies a distance h = l/4 beneath the plane of
the supports. A weight W hangs in each of the unconstrained nodes. The set of
equilibrium equations for the hanging net is written as:
60
3.6. DISCUSSION
√
3/2 0
√
3/2 −4/
√
17 0 0
−1/2 0 1/2 0 0 0
0 0 0 −1/
√
17 0 0
−
√
3/2 0 0 0 2/
√
17 0
1/2 1 0 0 −2
√
3/
√
17 0
0 0 0 0 −1/
√
17 0
0 0 −
√
3/2 0 0 2/
√
17
0 −1 −1/2 0 0 2
√
3/
√
17
0 0 0 0 0 −1/
√
17
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
t
1
t
2
t
3
t
4
t
5
t
6
=
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
0
0
−W
0
0
−W
0
0
−W
,
(3.75)
where t
i
is the tension force in bar i and the right hand side the external loads. An
SVD of the equilibrium matrix gives s = 0 and m = 3, hence assembly type II.
The mechanisms in D are orthogonal but not necessarily symmetric with respect
to the geometry. To ﬁnd the symmetric mechanisms a special approach is needed,
cf. [69], but this will not be used here. The computation of the internal forces are
straightforward as S = 0; (3.68) gives
t = W
2.31 2.31 2.31 4.12 4.12 4.12
T
. (3.76)
Using AE/W = 10
3
, (3.74) gives β ≈ 0 in the present numerical precision. Thus,
the displacements of the assembly are due to extensional deformation only: the
displacements are
d =
l
10
3
1.33 0 −22.9 −0.667 1.15 −22.9 −0.667 −1.15 −22.9
T
.
(3.77)
If the level of the triangular middle platform is set to zero, i.e. all bars lie in one plane,
the numbers of selfstress states and mechanisms change to s = 1 and m = 4 using
the recommend tolerance 10
−3
σ
11
for the small singular values. Figure 3.13 shows
the variation of the two lowest singular values with the height h of the triangular
platform. Theoretically, the assembly has a state of selfstress only for h = 0 but with
the present tolerance for the singular values the range of prestressability is about
−0.01 < h/l < 0.01. It appear that a lower tolerance must be used if the geometry
has to be accurately determined. This need to be considered when analysing the
tensegrities in the following chapters.
3.6 Discussion
Seven formﬁnding methods for tensegrity structures have been reviewed and clas
siﬁed into two categories. The ﬁrst category contains kinematic methods, which
determine the conﬁguration of either maximal length of the struts or minimal length
of the cable elements, while the length of the other type of element is not allowed
to vary. The second category contains static methods, which search for equilib
rium conﬁgurations that permit the existence of a state of prestress in the structure
61
CHAPTER 3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES
0.25 0.2 0.15 0.1 0.05 0 0.05 0.1 0.15 0.2 0.25
10
18
10
16
10
14
10
12
10
10
10
8
10
6
10
4
10
2
10
0
σ
i
i
/
σ
1
1
σ
55
/σ
11
σ
66
/σ
11
h/l
Figure 3.13: The variation of the two lowest singular values, σ
55
and σ
66
, with the
height h.
with certain required characteristics. Each category includes an analytical method,
suitable only for simple or very symmetric structures.
The nonlinear optimisation approach and the method of pseudo dynamical relax
ation have both successfully been used to determine conﬁguration details, i.e. nodal
coordinates, of structures that were already essentially well known. However, nei
ther of the two can be applied to problems that are not completely deﬁned, e.g.
when the lengths of all cables are not known in the formulation where the lengths
of the struts are maximised.
The three remaining static methods, sections 3.3.2–3.3.4, are in fact only two, since
the force density method and the energy method are equivalent. The main strength
of the force density method is that it is well suited to situations where the lengths of
the elements of the structure are not speciﬁed at the start. Thus, new conﬁgurations
can be easily produced, but it is diﬃcult to control the variation in the lengths of
the elements, as the set of force densities is varied. The reduced coordinates method
oﬀers a greater control on the shape of the structure, but involves more extensive
symbolic computations.
The linear force method, for analysis of the structure subsequent to formﬁnding,
is an eﬃcient numerical method which makes use of the advantages of the SVD. It
should be emphasised that the method, as it is presented here, is useful only for
62
3.6. DISCUSSION
structures that behave linearly. As will be seen later, this is suﬃcient for many
problems, but not all. In cases where the structures undergo large displacements, a
geometrically nonlinear FEM must be used. A FEM, suitable for cable structures
where cable slackening can occur, is presented in reference [171]. This will be used
in some of the following analyses.
Now that the necessary tools to analyse a general tensegrity structure have been
provided it is time to move on to the development of deployable tensegrity structures
for space applications.
63
Chapter 4
Deployable Tensegrity Masts
4.1 Background
In Snelson’s US patent “Continuous tension, discontinuous compression structures”
from 1965, [151], the construction of highly complex tensegrity structures by simple
modules is described. One of these structures is a mast with three struts per stage.
This mast is created by assembling triangular prisms on top of each other. The di
rection of rotation of the prisms vary so that every second prism is rotated clockwise
and every other counterclockwise. This procedure is illustrated in Figure 4.1 for
a threestage mast. They merge into a mast by substituting their individual base
cables by the saddle cables. The height of each module is H, but the height of the
mast is lower than 3H due to the overlap h of the saddle cables. It can be shown
that for a cylindrical mast this overlap can be expressed as a ratio of the module
height, η = h/H, cf. [117].
An important characteristic of Snelson’s tensegrity structures is that they have a
single state of selfstress. Hence, the length of only one element has to be adjusted
to prestress the structure. This is a key property to the practical implementation of
tensegrity structures, as has been shown by Snelson.
Fuller presents a tetrahedral mast in his patent from 1962, [44]. This mast has struts
that are connected to each other and is, therefore, less interesting from the point of
deployability.
4.2 Static and Kinematic Properties
The ﬁrst step in the analysis of any bar framework is the determination of its static
and kinematic properties. Consider an nstage tensegrity mast with v struts per
stage, constructed according to the scheme in Figure 4.1. Counting the number of
joints and bars in the mast yields
j = 2vn (4.1)
65
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
(a)
(b) (c)
(d)
Top cables
Base cables
S
t
r
u
t
s
V
e
r
t
i
c
a
l
c
a
b
l
e
s
S
a
d
d
le
c
a
b
le
s
D
i
a
g
o
n
a
l
c
a
b
l
e
s
Stage 1
Stage 2
Stage 3
z
x
y
Diameter D
θ
z = 0
z = H(1 − η)
z = H
z = H(2 − 2η)
z = H(2 − η)
z = H(3 − 2η)
Figure 4.1: Assembling a threestage tensegrity tower with three struts per stage
from basic tensegrity modules: (a) three modules are (b) assembled by
replacing the cables of the bases with saddle cables and ﬁnally (c) adding
diagonal cables to prestress the structure. The top base of the upper
and lower modules is rotated an angle θ w.r.t. the bottom base. The
middle module is rotated counterclockwise the same angle.
and
b = 2v(3n −1). (4.2)
Substituting (4.1), (4.2) and c = 6 into the extended Maxwell’s rule, (1.2), yields
m−s = 2v −6, (4.3)
which is independent of the number of stages n. Hence, under the assumption of
only one state of selfstress, s = 1, the number of mechanisms is
m = 2v −5. (4.4)
For most applications the stiﬀness is important, hence the masts with three struts
per stage are preferable as they have the lowest number of internal mechanisms.
66
4.3. FORMFINDING
Another aspect is to keep the total number of struts low as they will comprise the
major part of the mass of the mast. Therefore, one might think that the masts with
twostruts per stage would be better. However, closer examination of the masts with
v = 2 yielded j = 4n, in accordance with (4.1), but b = 6(2n − 1) as the bottom
and top bases now only consist of a single cable each. The extended Maxwell’s rule
for this mast yielded m − s = 0, hence the same number of mechanisms as the
masts with three struts per stage. From a practical viewpoint it would be more
diﬃcult to provide adequate restraints at the base, for this case with two struts per
stage. The masts with three struts per stage have triangular bases which can easily
be constrained. In the following, the focus will, therefore, be on masts with three
struts per stage.
4.3 FormFinding
Snelson [151] has built highly complex tensegrity masts for several decades. The
mathematical conditions for the existence of a prestressable conﬁguration have, for
the majority of these masts, been unknown. Sultan [159] presents the ﬁrst math
ematical treatment of Snelson’s multistage tensegrity mast with three struts per
stage. Shortly thereafter, Nishimura [117] presents closedform solutions for the
equilibrium conﬁguration of multistage tensegrity masts with v struts per stage.
4.3.1 TwoStage Tensegrity Mast
Sultan, [159], shows that the initial equilibrium solution of a cylindrical twostage
tensegrity mast with three struts per stage and equal stage height, reduces to a
quadratic equation in η with θ as the only remaining variable. Following the pio
neering work by Sultan, Nishimura [117], derives the following general equation for
the overlap ratio η of a twostage tensegrity mast with vstruts per stage:
R
1
R
1
cos
π
v
+ θ
=
η
1 −(1 −η) cos
π
v
2
1 −cos
π
v
(1 −η) + η
2
, (4.5)
where R
1
and R
1
are the base and top radii of the ﬁrst stage, respectively, and θ the
angle of relative rotation. Symmetry implies R
1
= R
2
and R
1
= R
2
. Some special
cases of (4.5) are noted: (i) for θ = 0 and R
1
= R
1
, the overlap is
η =
2 cos
π
v
1 + 2 cos
π
v
, (4.6)
and (ii) for v = 3 and R
1
= R
1
, (4.5) simpliﬁes to
η
2
cos
π
3
+ θ
−
1
2
−η
cos
π
3
+ θ
+
1
2
+ cos
π
3
+ θ
= 0. (4.7)
67
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
Equation (4.6) indicates that the overlap span of the twostage mast is quite small:
η = 1/2 for v = 3, and η →2/3 as v →∞.
4.3.2 MultiStage Tensegrity Masts
Sultan [159] uses a symbolic approach to ﬁnd the overlap of multistage masts.
As the number of stages increases, the computations becomes too complex for the
symbolic mathematical softwares, even though symmetry conditions are used to
reduce the size of the matrices; masts with up to nine stages are analysed in [159].
In order to analyse masts with a very large number of stages, say 100, a numerical
approach based on the force method is suggested here. The initial equilibrium
conﬁguration of the tensegrity mast is the solution to (3.51) in the absence of external
forces, f = 0. Hence, the task is to ﬁnd the overlap η which renders the equilibrium
matrix singular, i.e. det H(η) = 0. This approach, which was implemented by the
author and H. Y. E. Pak, starts with an initial value for η. The mast conﬁguration
was then generated by the scheme in [159]. The determinant of the 6(3n−1)6(3n−
1) equilibrium matrix H was then computed (N.B. v = 3). The Matlab [89] function
fsolve, which solves a system of nonlinear equations by the method of least squares,
10 5 0 5 10 15 20 25 30
5
10
15
20
25
30
35
40
45
50
0
.
0
5
0
.
0
5
0
.
0
5
0
.
1
0
.
1
0
.
1
0
.
1
5
0
.
1
5
0
.
1
5
0
.
2
0
.
2
0
.
2
0
.
2
5
0
.
2
5
0
.
2
5
0
.
3
0
.
3
0.3
0
.
3
5
0
.
3
5
0
.
3
5
0
.
4
0
.
4
0
.
4
0 4
0.45
0.5
2
N
u
m
b
e
r
o
f
s
t
a
g
e
s
θ(
◦
)
Figure 4.2: Contour lines of the dimensionless overlap η for multistage tensegrity
masts with three struts per stage.
68
4.3. FORMFINDING
was used to solve det H(η) = 0 for η. With the fsolve function, masts with up to 15
stages could be solved, but the termination tolerance TolFun had to be set as low as
10
−40
to achieve convergence. To analyse masts with more stages, another approach
was needed. Instead of implementing a sophisticated linesearch routine, which also
may fail, a more rudimentary routine, based on successive interval bisectioning, was
implemented, cf. [29]. The convergence was slow but stable. Several overlap values
may exist, which give a zero determinant, but only one with the correct prestress.
Care must therefore be taken in choosing the initial interval. This approach may
not be the most elegant way of ﬁnding the initial equilibrium conﬁgurations, but it
nevertheless yields the correct solution. The equilibrium conﬁgurations for masts
with up to 50 stages were computed by this improved numerical method. The
overlap values of these masts are given in Appendix A and Figure 4.2 shows a
graphical representation of these values. It is observed that the overlap diminishes
with the number of stages. The state of selfstress also changes with n and θ. The
fact that the overlap decreases with the number of stages may not be ideal from a
manufacturing point of view as new stages cannot be added without changing the
geometry of the complete mast.
Investigating multistage tensegrity masts, Nishimura [117] ﬁnds that even though
all symmetry groups are used to simplify the prestressability condition, no analytical
solution could be obtained for masts with more than four stages using current sym
bolic software. Nishimura then investigates a class of tensegrity masts with the same
selfstress for the interior stages independent of the number of stages. An interior
stage is deﬁned as any stage but the ﬁrst and the last stage in a mast [117]. Recall
that for a given relative rotation θ, the overlap ratio η decreased as the number
of stages increased, Figure 4.2. By having diﬀerent rotation angles for the interior
and the ﬁrst and last stages, Nishimura shows that it is possible to keep a constant
overlap ratio for any number of stages. The geometry of a cylindrical multistage
tensegrity mast with the same selfstress for interior stages and constant stage height
(H = H
∗
) is thus described by three parameters [117]: the rotation angle of the ﬁrst
and last stages θ, the rotation angle of the interior stages θ
∗
, and the overlap ratio
η (= η
∗
). Note that the rotation angle of the second and second to last stages is the
same as that of the interior stages. For the interior stages, the relationship between
the overlap ratio η
∗
and relative rotation θ
∗
is [117]
η
∗2
cos
π
v
−cos
π
v
+ θ
∗
+ η
∗
1 −cos
π
v
1 + cos
π
v
+ θ
∗
−cos
π
v
+ θ
∗
1 −cos
π
v
= 0.
(4.8)
If θ
∗
= 0,
η
∗
=
cos
π
v
1 + cos
π
v
, (4.9)
while if η
∗
= 0,
θ
∗
= π
1
2
−
1
v
, (4.10)
which is identical to the relative rotation of a tensegrity prism with i = 1, (3.4).
69
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
After ﬁnding the interior overlap ratio η
∗
by (4.8), the relative rotation of the ﬁrst
and last stages is subsequently determined by solving, symbolically or numerically,
the characteristic equation of a 7 7 matrix (Equation (41) in [116]). However,
Micheletti [93, 94] shows that the rotation of the end stages is actually found using
the prestressability condition for the twostage mast, (4.5). To ﬁnd the geometry of
a cylindrical tensegrity mast with uniform interior selfstress, one speciﬁes θ
∗
to ﬁnd
η
∗
by (4.8). Then θ is computed by (4.5) with R
1
= R
1
. Note that it is possible to
design multistage masts with θ = θ
∗
and constant η, but the nodes of these masts
do not all lie on the surface of a cylinder; the base and top radii have to change
according to (4.5), i.e., R
1
/R
1
= 1. For example, θ = θ
∗
= 0 gives
R
1
R
1
=
2
3 −cos
π
v
. (4.11)
The multistage tensegrity masts by Nishimura are better suited for applications as
the forces are relatively uniform through the mast.
4.4 Manufacturing Technique
The practical implementation of tensegrity masts requires an eﬃcient and accu
rate manufacturing technique. Conceptually, the structure is easy to make since
the compression members do not touch each other, which otherwise would require
complicated joint designs. The present manufacturing technique was inspired by
Pugh’s [132] illustration of the diamond pattern system, Figure 4.3. However, be
cause of the intrinsic diﬀerence between the cables and struts in terms of material
properties, it was decided to separate the construction of the tension and compres
sion members. This is in contrast to the conventional way of construction, where,
e.g., a mast is built from node to node and stage by stage. An obvious disadvantage
of the conventional method is the diﬃculty in retaining the required precision when
the whole structure has to be constructed in three dimensional space, which in the
case with ﬂexible structures, is especially awkward. A direct analogy to the pro
posed manufacturing method is the separation of tensile and compression members
in the construction of tension roof structures, cf. [171].
The separation of the tension and compression members would result in more free
dom in the manipulation of the ﬂexible cables. The ﬁrst step in the manufacturing
procedure is to map the threedimensional net of cables onto a twodimensional
plane, as done by Pugh, without changing any cable lengths. Since the cable net is
composed mainly of triangles connected to each other in a special way, the number
of net conﬁgurations is restricted. It was found that only two cable net conﬁgu
rations, which preserve all cable lengths, exist. These conﬁgurations are shown in
Figures B.2(a) and (b) for a twostage mast and in Figures B.3(a) and (b) for a three
stage mast. At ﬁrst, net 1 and 2 for the twostage mast seemed equal. However,
close scrutiny of net 1 showed that the diagonal cables had changed place with the
vertical cables, which means that the saddle cables had been inverted. Net 2 agrees
70
4.4. MANUFACTURING TECHNIQUE
more with the layout of threedimensional version of the cable net, Figure B.1(a).
The two key aspects of the manufacturing method required to obtain good precision
are to ensure accurate element lengths and to make certain that the angles in the
twodimensional net do not diﬀer too much from the corresponding angles in three
dimensions. Net 1 may satisfy the ﬁrst requirement but certainly not the latter one.
In this respect net 2 is better, but still not ideal. The horizontal distance between
the nodes along the saddles is identical to the length of the base cables, which pro
duces an overlap in two dimensions that is larger than that in three dimensions.
For a threestage mast, this layout has overlapping triangles, Figure B.3(b), which
is highly undesirable. By relaxing the length preservation condition slightly, the
saddle overlap in the twodimensional cable net was set equal to ηH, Figures B.2(c)
and B.3(c). This setting yielded that the distance between the nodes at the bases
was too long and, thus, could not be constructed along with the rest of the net.
However, the angles between the members in net 3 should agree better with those
in three dimensions. Considering accuracy issues, net 3 is the preferred layout. It
was identiﬁed as sensible to have as few openends of the cables as possible, so an
attempt was made to ﬁnd a way of completing the entire tracing of the net with
only one or two separate cables. However, it was soon realised that because there
existed more than two nodes with an odd number of cables connected to them, by
simple Graph Theory, it was impossible to use one cable only.
The construction of three tensegrity masts by Pak and the author is described
in reference [118]. It is reported that these masts suﬀer from large cable length
inaccuracies, despite accurate cable nets. These nets were constructed by ﬁrst gluing
the plotted cable net to a wooden board. Holes were then drilled at all intersection
points and small threaded rods were positioned in the holes. Stiﬀ Kevlar cords
were strung between the intersection points and fastened by looping the cords once
around each threaded rod. Subsequent to the tracing of the net, the connections
were secured by epoxy resin. According to [118], the length errors were caused by
uneven lengthening or shortening of individual cables. The changes in length seemed
to depend on the way the cables were looped around the nodes. In hindsight, the
looping direction decided if the cable lengthened or shortened, but it was not the
heart of the problem. The direct cause was the use of net 1, which forces the
nodes to rotate to adopt to the threedimensional conﬁguration. Unfortunately, the
deﬁciencies of net 1 was not recognised until all three masts had been constructed.
D
C
B
A
joins to A
to B
to C
to D
D
Figure 4.3: Illustration of the Pugh’s diamond pattern systems for a threestage,
three struts per stage mast, [132].
71
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
To examine the feasibility of net 3, an eightstage mast was constructed by the au
thor, using almost the same method as the previous masts. However, the experience
from the previous constructions indicated that the connections must be manufac
tured to a higher precision. Preferably, the cables of the twodimensional cable net
should lie in the same plane and go through the centre of the nodes; looping of the
cables around the nodes must be avoided. The net for the eightstage mast was plot
ted on an A0 paper (1189841 mm
2
), which was fastened on a 40 mm thick Medium
Density Fibreboard (MDF). Holes of 4 mm in diameter, approximately normal to
the board, were drilled through the board at the cable net intersections. Threaded
4 mm rods (M4), 44 mm in length, were positioned in the holes. To the parts of
the rods sticking up from the board, 14 mm long aluminium tubes of 6 mm outer
and 4 mm inner diameter were glued. The aluminium tubes had been predrilled
with eight 1 mm coplanar holes, in the direction of the cables through the centre of
the tube. Thus, all the cables in the ﬁnal net would lie in the same plane and each
cable would go through the centre of the node. The cables were made of a 0.45 mm
diameter stainless steel cable composed of 49 individual wires for maximum ﬂexi
bility. After the tracing of the cables, the connections were secured by ﬁlling the
aluminium cylinders with highstrength epoxy resin, which eﬃciently bonded to the
thin steel cable.
The struts were made of the same aluminium tubes as the connections. Two nuts
were placed on the threaded rods to enable length adjustment. One of the initial
aims was to make a mast with the minimum number of adjustment mechanisms,
since they are usually large, diﬃcult to design and will easily fail. However, from
the experience from earlier models, it was considered necessary to include some
adjustment possibility. It was immediately recognised that the eightstage mast was
far more accurate than the previous models. Nevertheless, surprisingly large length
adjustments were needed to prestress the nets. One reason for the inaccuracies is
shown in Figure 4.4. In the twodimensional net, the cables connected to a node
lie in the same plane and go through the same point, Figure 4.4(a). However, in
the threedimensional net they do not necessarily intersect, Figure 4.4(b). Thus,
the geometry of the mast will change to a conﬁguration where they intersect, i.e.
where the forces can be in equilibrium. To eliminate this error source it is necessary
to manufacture joints, preferably spherical, with holes drilled in the directions of
the cables in the threedimensional net conﬁguration. By adjusting the distance
between the nodes in the twodimensional net according to the layout of the holes in
the spherical joint it would still be possible to construct the net in two dimensions,
but the cables will no longer lie in the same plane
1
. No net was, however, constructed
according to this principle.
To summarise, the basic idea of the construction scheme, which was to separate
the construction of the cable net and that of the bars, worked well after some early
mistakes. The modiﬁcation of the struts to enable mast folding and deployment will
be presented next.
1
This solution was pointed out to the author by R. E. Skelton at “Colloquium Lagrangianum—
Strutture Tensegrity: Analisi e Progetti” (Tensegrity Structures: Analysis and Design) in Rome,
6–8 May 2001.
72
4.5. DEPLOYMENT
(a) (b)
Figure 4.4: Cables which coincide in two dimensions (a) may not coincide in three
dimensions (b).
4.5 Deployment
An nstage tensegrity mast with three struts per stage has 5−2/n times more cables
than struts. In addition, the struts are the only stiﬀ members and also the longest
ones. Concerning deployment and packaging eﬃciency, the focus would therefore be
on the struts. In the few studies on foldable or deployable tensegrities, this has not
always been the case.
Bouderbala and Motro [11] analyse diﬀerent approaches to tensegrity folding: (i)
strut mode, (ii) cable mode and (iii) mixed mode. In the ﬁrst mode, only the
lengths of the strut are changed and in the second one, only the cable lengths. In
the third mode both lengths are changed. The folding of expandable octahedron
assemblies are studied. The cable mode is found to be less complex than the strut
mode, although the latter one produces a more compact package.
Furuya [45] analyses three approaches for the deployment of a tensegrity mast with
connected struts: (i) ﬁxed lateral cable length, (ii) ﬁxed strut length and (iii) ﬁxed
base cable length. For modes (i) and (ii), the size of the triangular bases changes
simultaneously during deployment. Only in mode (iii) is it possible to deploy the
mast sequentially, which usually is more convenient for space applications. Tele
scopic struts are suggested for realising the mechanisms of this mode.
In the deployment approach by Skelton and Sultan [159, 162], the lengths of the
cables are changed in a way so that the structure at every step is in an initial
equilibrium conﬁguration. Hence, the structure is in stable equilibrium throughout
the deployment.
There are, however, a few disadvantages with this approach. First, using a cable
activated deployment means that the diameter of the mast will increase if the struts
are stacked horizontally on top of each other in the stowed conﬁguration. As ob
served in section 2.2.4, the popular masts for space applications have a constant
diameter during deployment. If the struts are stacked vertically the diameter need
not change, but such a scheme will probably lead to problems as it is no longer
obvious how the struts should be stacked for a troublefree deployment. Second, a
large number of devices is needed to control the lengths of the cables; each strut
73
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
has one electric motor to control the length of the saddle cables and three reels
on which the vertical and diagonal cables are stored to prevent entanglement [149].
The motors and the reels produce a very reliable cableactivated deployment, but
they add complexity to the mast. As described in reference [149], this deployment
procedure is neither synchronous nor sequential but rather a combination of both.
In the search for a deployment procedure that is simpler than the cableactivated
approach, but hopefully equally reliable, a procedure based on strut deployment will
be proposed.
4.5.1 Strut Deployment
If the strut length is equal to the diameter and folded by a midpoint hinge, the
struts can ﬁt horizontally beside each other in a canister with a diameter equal to
that of the mast. If the struts are longer, some parts of them will lie on top of
each other and, thereby, increase the stack height. This is a drawback, although
not too serious. Telescopic struts can be collapsed into shorter lengths. A telescopic
alternative, however, would require a motor on each of the 3n struts in the mast.
Having so many motors presents too large a risk. The failure of only one motor
would end the deployment. Thus, a hinge with selfdeployable characteristics would
be preferable.
TapeSpring Rolling (TSR) Hinge
The use of carpenter tapes, or tapesprings, i.e. the curved metal tape found in
tape measures has been considered for a long time in the design of selfdeploying
hinges [128]. In order to fold these tapes, a large moment is required to initially
buckle the tape. Subsequent to buckling, a much lower moment is required to con
tinue folding the tape. During deployment, the tapes provide a small but constant
restoring moment to eventually lock into the straight position. A key property of
the tapespring is that it can be signiﬁcantly deformed several times without per
manent damage. To increase the buckling moment, and also the restoring moment,
two tapesprings can be placed a certain distance apart with the concave sides fac
ing each other. The resulting hinge will be quite stiﬀ in bending but weaker in
torsion. A tapespring hinge, which also is torsionally stiﬀ, has been developed
at DSL, Figure 4.5. Its basic components are two tapesprings and two Rolamite
hinges
2
, one on each side of the tapesprings. The Rolamite hinges are made of
steel cables and Delrin
3
, a space qualiﬁed acetal resin, with approximately half the
density of aluminium. The overall weight of the hinge depicted in Figure 4.5 is
0.105 kg [181]. Including attachments to the struts the TSR hinge has a total mass
of 0.2 kg. Other advantages of this hinge is simple assembling and low friction, the
2
The Rolamite, or the rollingband, concept was invented by D. F. Wilkes of the Atomic Energy
Commission’s Sandia Laboratory, Albuquerque, NM, USA. In its basic form, the device consists
of a metal band looped around two rollers, [19]
3
Delrin is a registered trademark of E. I. du Pont de Nemours and Company.
74
4.5. DEPLOYMENT
(a) (b)
Figure 4.5: The tape spring rolling hinge: (a) deployed and (b) folded (Courtesy of
A. M. Watt).
latter eliminates the need for lubrication. More information about the TSR hinge
is found in [180, 181].
BiStable Tube Hinge
A wellknown property of tapesprings is that they are stable only in the straight
position. Once folded into a bent conﬁguration, they need external restraints to
keep them in that position, e.g. the cassette containing the rolledup carpenter tape.
Recently, the BRC tube, which is stable in both the bent and straight positions, was
invented by DatonLovett [30, 65]. The bistability is obtained by a special layout
of the ﬁbres in the composite material. In the straight conﬁguration the tube is
unstrained, while it is strained in the rolledup conﬁguration. To move the tube
from the rolledup position to the straight position requires only a small quantity of
energy.
A strut with a hinge made of a bistable tube is shown in Figure 4.6. A fourstage
tensegrity mast, with bistable hinges, was built by H. Y. E. Pak and the author.
Another, more distant, alternative to bistable hinges is to manufacture the whole
strut as a bistable tube. The torsional stiﬀness can be increased by closing the open
crosssection with velcro [30].
4.5.2 Demonstrator Masts
As mentioned earlier, three tensegrity masts, two with three stages and one with
four stages, were built at DSL, cf. [118]. The ﬁrst threestage mast was not foldable
and was only made to gain some experience with the manufacturing procedure.
The mast was prestressed by two adjustable plastic ties, one on the top base cable
and the other on the bottom cable. The second threestage mast had struts made
of aluminium rods and singleblade tapesprings as hinges. When the mast was
75
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
Figure 4.6: Strut with a bistable hinge in straight and bent conﬁgurations.
prestressed, by shortening the top and bottom cables, several struts buckled in
torsion. Closer examination revealed that the blades had been ﬂattened by the pop
rivets which connected the blades with the rods. This ﬂattening signiﬁcantly reduced
the stiﬀness of the curved tape and yielded the mast useless. For the following masts,
a heavier boltandnut alternative was used instead of poprivets.
FourStage Mast
The fourstage mast, which was the third tensegrity mast built at DSL [118], was
the ﬁrst one that could be folded in a satisfactory way. The struts were made of
solid 6.35 mm diameter aluminium rods and bistable tubes were used as hinges,
Figure 4.6. The torsional buckling problems with the previous model was eliminated
by the bistable tubes. In order to achieve increased axial stiﬀness, the rods were
attached to the outside of the tubes. The struts, which had a length accuracy of ±1
mm, had turnbuckles at the ends that enabled length adjustment. The deployed
mast is shown in Figure 4.7. The mast was reasonably straight considering the ac
curacy problems of the cable net discussed earlier. However, high local stresses were
induced at the connections between bistable composite tubes and the aluminium
rods, which led to local buckling of the tubes. This signiﬁcantly reduced the com
pressive strength of the struts and the mast could not be prestressed to the desired
level. Compared to the mast with tapespring hinges, the mast with the bistable
struts was much heavier which led to cable slackening. The mast could easily be
folded by hand, thanks to the bistability of the hinges. The folded mast is shown
in Figure 4.8. The bistable tubes are disproportionately large in comparison to the
complete mast. However, no bistable tubes of smaller diameter, that functioned
well, were available at the time.
76
4.5. DEPLOYMENT
(a) (b)
Figure 4.7: The deployed fourstage mast with bistable struts.
(a) (b)
Figure 4.8: The stowed fourstage mast with bistable struts.
EightStage Mast
To fully evaluate the concept of selfdeployable struts, the eightstage mast was
equipped with tapespring hinged struts. Two tapesprings (19 mm wide) were
77
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
connected to two aluminium tubes, with the concave sides of the tapesprings facing
each other. Compared with the tapespring hinges of the second threestage model,
the new hinges were much stiﬀer in both torsion and bending. The mast could thus
be given the necessary prestress to reduce the eﬀects of gravity. One immediate
problem was how to keep the struts in the folded conﬁguration. This was solved by
enclosing the stowed package by a canister, Figure 4.9, similar to that used for the
CM, Figure 2.4. Another problem was how to control the deployment of the mast.
An approach suggested by Pellegrino [126] was to use an inﬂatable tube inside the
mast. The tube would activate the deployment and stabilise the unstressed mast
during deployment. For the present mast, the inﬂatable tube was replaced by an
aluminium rod through the base of the canister and connected to the three nodes at
the top of the mast. When the rod was pushed from underneath the base, it deployed
the mast sequentially. The complete deployment of the eightstage mast is shown
in Figures 4.10 and 4.11. The mast was not fully prestressed until the bottom stage
was fully deployed. However, the slender aluminium rod could not provide enough
force and the tapespring hinges could not provide suﬃcient moment to deploy the
bottom stage, i.e. to prestress the whole structure. A ﬁnal prestressing of the mast
can be obtained by allowing one of the base cables to be a little longer during
deployment. All the tapespring hinges can then be deployed since they do not have
to prestress the mast. Finally, the longer base cable is shortened by a motorised
turnbuckle and the mast is prestressed. Hence, only two motors would be needed:
one that shortens the base cable and one that deploys the structure by actuating
and controlling the mast deployment. Twice the deployment had to be stopped to
resolve some entanglement problems. This must be regarded as very satisfactory
considering no precautions had been taken to avoid entanglement.
4.6 Structural Analysis
Several studies, e.g. [45, 117, 159], talk about the potential of the tensegrity masts as
lightweight, deployable structures, but none do a comparison with existing deploy
able masts. To satisfactorily prove the applicability of tensegrity masts as eﬃcient
deployable structures, they must be compared to a realised and successful mast. The
current stateoftheart mast is the primary deployable structure for the STRM—
the 60 m long ADAM. The ADAM supports a 360 kg antenna at its tip and carries
200 kg of electric and ﬁbreoptic cables and a gas line along its length [1]. Data
for the STRM ADAM is given in Table 4.1. Recall that the tensegrity masts with
three struts per stage are kinematically indeterminate to one degree. Their stiﬀness
is therefore dependent on the prestress level. The stiﬀness of the ADAM is presum
ably higher than that of the tensegrity mast, as it uses stiﬀ square frames between
each bay, Figure 2.7. However, it is possible to remove the inﬁnitesimal mechanisms
of the tensegrity mast by adding cables in such a way that completely triangulates
the bottom stage. This stiﬀening approach will be tested here.
78
4.6. STRUCTURAL ANALYSIS
(a) (b)
Figure 4.9: Eightstage mast stowed in a canister: (a) side view and (b) top view.
In order to facilitate an accurate comparison, the tensegrity mast should have the
same diameter, length and number of bays as the ADAM. First, a tensegrity mast
with struts of equal lengths was studied. Then, a mast with uniform interior forces
was analysed.
In the present mast generation routine, cf. [118], one of the input values is the strut
length. To generate a mast with a given bay length, the relationship between the
bay and strut length must be found. For an nstage mast with three struts per stage,
the length of the struts, l
s
, for given values of the mast diameter D, bay length H
bay
,
rotation angle θ and overlap η, is computed as:
l
s
=
¸
¸
¸
¸
D
2
4
2 + cos θ
∗
+
√
3 sin θ
∗
+
H
2
bay
¸
1 −η
1 −
1
n
2
1/2
. (4.12)
The length of the struts at the end stages for a mast with uniform interior forces is
found by setting θ
∗
:= θ.
4.6.1 Initial Equilibrium Element Forces
An appropriate initial equilibrium mast conﬁguration should have fairly uniform
internal forces.
For D = 1.12 m and H
bay
= 0.6975 m, the maximum and minimum cable and strut
forces versus the number of stages, for a mast with equallength struts, are shown
79
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
F
i
g
u
r
e
4
.
1
0
:
R
o
d

c
o
n
t
r
o
l
l
e
d
d
e
p
l
o
y
m
e
n
t
o
f
t
h
e
e
i
g
h
t

s
t
a
g
e
m
a
s
t
:
ﬁ
r
s
t
f
o
u
r
s
t
a
g
e
s
.
80
4.6. STRUCTURAL ANALYSIS
F
i
g
u
r
e
4
.
1
1
:
R
o
d

c
o
n
t
r
o
l
l
e
d
d
e
p
l
o
y
m
e
n
t
o
f
t
h
e
e
i
g
h
t

s
t
a
g
e
m
a
s
t
:
l
a
s
t
f
o
u
r
s
t
a
g
e
s
.
81
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
Table 4.1: Data for the 60 m long STRM ADAM, [1].
Geometry Stiﬀness
Length (m) 60.68 Bending stiﬀness (MNm
2
) 13
Diameter (m) 1.12 Torsional stiﬀness (MNm
2
) 0.15
Bay length (m) 0.6975 Shear stiﬀness (MN) 0.49
Number of bays 87 First bending mode (Hz) 0.10
Mass (kg) 290 First torsion mode (Hz) 0.17
Mechanical stability Strength
Bending (
◦
/N) 0.0059 Bending strength (Nm) 8140
Twist (
◦
/Nm) 0.0228 Torsional strength (Nm) 305
Axial (µm/N) 0.7 Shear strength (N) 400
in Figure 4.12. The overlap values were taken from Appendix A. The maximal and
minimal forces were normalised by the force in the base cables. The maximum cable
and strut force increased linearly with the number of stages while the minimum
cable force decreased only slightly. Already for seven stages, the maximum strut
force was ten times the force in the base cables at θ
∗
= −10
◦
. Hence, in a multi
stage mast with equallength struts, the relationships between element forces are
undesirably large. This is a serious disadvantage and the only conclusion must be
that long masts with equallength struts are unfeasible for applications.
0 5 10 15 20 25 30 35 40 45 50
40
30
20
10
0
10
20
Number of stages
θ = −10
◦
θ = 0
◦
θ = 10
◦
θ = 20
◦
t
i
t
base
Figure 4.12: Maximum and minimum forces in the cables and struts in a multistage
mast with struts of equal length.
82
4.6. STRUCTURAL ANALYSIS
The mast conﬁguration by Nishimura [117], which has uniform forces for the interior
stages, is probably more suitable for practical uses. In the following analysis of the
Nishimura mast it was assumed that each stage had the same radius, R
i
= R
i
,
height, H
∗
= H and overlap ratio, η
∗
= η. For a given value of the interior stage
rotation θ
∗
, η was found by (4.8). Subsequently, the rotation of the end stages, θ,
was computed by (4.5). As (4.5) and (4.8) are valid for any number of stages, the
element forces in an nstage mast (n ≥ 5) was generated by two sets of element
forces; the ﬁrst set corresponds to stages 1 and n and the second set corresponds
to the intermediate stages. The total number of diﬀerent force values is ten. Set
1 is the struts and base, vertical, diagonal, and saddle cables (N.B. two values) at
the ﬁrst (or last) stage and set 2 the struts and vertical, diagonal and saddle cables
at the interior stages. Hence, the forces in an nstage mast were determined by
analysing a ﬁvestage mast. This remarkable characteristic signiﬁcantly simpliﬁed
the analysis of multistage masts. The variations of the normalised forces with θ
∗
are shown in Figure 4.13. It was immediately noted that the forces were lower than
those of the equallength strut masts. As before, a suitable conﬁguration should
have uniform element forces. Therefore, the magnitudes of the forces in the mast
were restricted: (i) no cable force was allowed to be lower than that in the base
cables and (ii) the maximum compressive force in the struts could not be larger
than ﬁve times the tension force in the base cables. These requirements restricted
the interval of possible solutions to θ
∗
= 0–10
◦
.
4.6.2 Preliminary Design of Struts and Cables
The struts of the mast with uniform interior forces were subjected to the largest
forces; their design is therefore of most importance. As the struts most likely would
be slender, their load bearing capacity is governed by buckling. Consider a perfectly
straight column of two rigid bars and a rotational spring at the connecting hinge,
Figure 4.14(a). The buckling load of the column is [36]
P
cr
=
C
M
l
, (4.13)
where C
M
is the spring stiﬀness and l the column length. Now consider that the
column has an initial imperfection, ψ
ini
, with the corresponding lateral midpoint
displacement ∆ ≈ ψ
ini
l/2. The midpoint moment M
l/2
in the imperfect column
due to the compressive load P is
M
l/2
= P∆
1
1 −
P
P
cr
, (4.14)
where P
cr
the buckling load according to (4.13). Rearranging (4.14) yields
P =
1
∆
M
l/2
+
1
P
cr
. (4.15)
83
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
30 20 10 0 10 20 30
6
5
4
3
2
1
0
1
2
3
4
vertical interior
v
e
r
t
ic
a
l
1
d
i
a
g
o
n
a
l
1
d
ia
g
o
n
a
l in
t
e
r
io
r
sa
d
d
le
1 sa
d
d
le in
terio
r
strut length
stru
ts 1
s
t
r
u
t
s
i
n
t
e
r
i
o
r
t
i
t
base
θ
∗
(
◦
)
Figure 4.13: Cable and strut forces in the ﬁrst (and last) and interior stages in a
multistage mast with uniform interior forces. D = 1.12 m, H
bay
=
0.6975 m and n = 87 in (4.12).
In reference [181], the buckling moment, M
cr
, and rotational stiﬀness, C
M
, of the
TSR hinge are measured to 13 Nm and 480 kNmm/rad, respectively. The test
for the rotational stiﬀness was done with a hinge of slightly diﬀerent geometry
than that which was tested for the buckling moment. Formulae for predicting the
stiﬀness indicate that the latter hinge has higher stiﬀness than the former one.
Nevertheless, in the following calculation of the load bearing capacity of the hinged
strut, the measured values were used. In order to obtain a high load capacity, the
manufactured strut must be as close to straight as possible. While the two bars of
the strut can be made nearly straight, the error arising from the alignment of the
hinge attachments is probably more severe. Considering the extremely high cost of
launching a spacecraft, high manufacturing tolerances are justiﬁed; the aim for the
strut crookedness was ∆ ≤ l/400.
At θ
∗
= 0
◦
, the strut length at the ﬁrst stage is l
s,1
= 1.46 m, which yields P
cr
=
328 N using (4.13). Substituting the buckling load, ∆ = 3.65 mm and M = 13 Nm
into (4.15) yielded P = 300 N. As this value is very low, no safety factor against
buckling was considered in the following analyses and sometimes the compression
load was allowed to be higher than the buckling load in order to study the eﬀects
of increased prestress. The tubes connected to the hinge were made of CFRP with
E = 210 GPa and ρ = 1660 kg/m
3
[96]. The minimum wall thickness of the tubes
84
4.6. STRUCTURAL ANALYSIS
(a) (b) (c)
l
2
l
2
l
2
l
2
P
P
P
C
M
δ
ψ
ψ
ini
Figure 4.14: Buckling of a twolink column: (a) initial state, (b) buckled state and
(c) initial, imperfect state.
was set to 1 mm to ensure adequate toughness for handling and assembly [96]. With
regard to length precision requirements, the struts cannot be too slender. A strut
diameter of 25 mm yielded, for a pinended strut of constant tubular crosssection
and length l
s,1
, P
Eu
= 5.28 kN or about 17P. The eﬀective axial stiﬀness of the
strut was computed as
AE
l
eﬀ
=
AE
l
TSR
AE
l
CFRP
AE
l
TSR
+
AE
l
CFRP
, (4.16)
where (AE/l)
TSR
is the stiﬀness of the TSR hinge and (AE/l)
CFRP
the stiﬀness of
the CFRP tubes. The measured TSR hinge stiﬀness is 7223 N/mm, with l = 88 mm,
[181], and the computed stiﬀness of the CFRP tubes, with l = 1460−88 = 1372 mm,
is 11540 N/mm. By (4.16) the eﬀective axial stiﬀness of the strut was 4442 N/mm
or 6.485 MN for a length of 1460 mm.
The cables were assumed to be made of thin CFRP tape 5 mm wide and 0.5 mm
thick. Its properties were the same as for the tubes above. The allowable stress was
set to 200 MPa with safety factor between 5 and 10, which gave an allowable force
of 500 N. The axial stiﬀness of the cables was 0.525 MN.
85
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
4.6.3 Vibration Analysis
A few earlier studies have been concerned with the vibrational characteristics of
tensegrity masts. Furuya [45] analyses a ﬁvestage mast with contacting struts
made of simple triangular prisms. An nstage mast of this type has s = m = n. The
prestress, therefore, plays an important part in providing the mast with geometric
stiﬀness. Not surprisingly, the frequencies of the zero energy modes increase with
the level of prestress. Murakami [115] performs a modal analysis of a sixstage
mast with v = 3. The fundamental mode is the one corresponding to the internal
mechanism, which is characterised by adjacent countertwisting of each stage, i.e. a
global axial mode. The second mode is a bending mode and its frequency is much
higher than the ﬁrst one. While the frequency of the ﬁrst mode can be increased by
increasing the prestress, the frequency of the ﬂexural mode, which has a nonzero
elastic energy, does not change signiﬁcantly with the prestress level.
The natural frequencies are computed by the classical eigenvalue problem of the
homogeneous linear systems
Kd = ω
2
Md, (4.17)
where K is the tangent stiﬀness matrix, M the mass matrix, d the displacement
vector and ω the angular frequency. For problems with a small number of degrees of
freedom, e.g. tenstage mast, the Matlab function eig was used. For larger problems,
such as the 87stage mast, the function eigs was used instead. The eigs function
computes the N largest eigenvalues and their corresponding eigenvectors, where N
is a userdeﬁned value. Since the lowest eigenvalues were sought for the present
problem, (4.17) was rewritten as [47]:
¯
K
˜
d =
˜
d, (4.18)
where
¯
K = L
T
K
−1
L, (4.19)
˜
d = L
T
d, (4.20)
and
=
1
ω
2
. (4.21)
The lower triangular matrix L was obtained by a Choleski factorisation of M:
M = LL
T
. (4.22)
Although not optimised for eﬃciency, this approach required signiﬁcantly less com
puter time when analysing larger problems.
In the present study, the mast and element dimensions were those determined in
the previous section. For simplicity, the mass of each hinge, 0.2 kg, was evenly dis
tributed along the length of each strut. The joints between the struts and the cables
were assumed to be 25 mm diameter aluminium spheres, each weighing 0.025 kg.
The total mass of these joints was evenly distributed along the total length of the
cables and the struts.
86
4.6. STRUCTURAL ANALYSIS
To gain some understanding of the vibrational characteristics of the tensegrity masts,
a tenstage mast was ﬁrst analysed. This mast had the same diameter and bay
length as the former one, but slightly shorter struts by (4.12). First, the mast was
analysed for diﬀerent values of θ
∗
between 0 and 10
◦
and a constant prestress of 50 N
in the base cables. This analysis yielded, as anticipated, that only the frequency of
the mode corresponding to the “axial” internal mechanism, Figure 4.15(a), varied
with the prestress in the structure (which changes with θ
∗
). The frequencies of the
ﬂexural modes, Figures 4.15(b) and (c), were unaﬀected by a change of θ
∗
. These
observations agrees with the ones by Murakami [115]. One conﬁguration, θ
∗
= 10
◦
,
was chosen for further studies on the eﬀects of the prestress level. Note that this
particular conﬁguration did not have the highest frequencies but had the lowest
internal forces and lowest total mass. The tenstage mast was analysed for four
levels of prestress: 50, 100, 200 and 500 N. The buckling load of the struts, 300 N,
was disregarded in this analysis. As before, the ﬂexural modes were not aﬀected by
the increased prestress. However, a continuous increase of the prestress eventually
resulted in that the frequency of the fundamental, axial mode becoming higher than
those of the two ﬁrst ﬂexural modes. This shift in fundamental mode can be seen in
Table 4.2 when going from 200 to 500 N. From the opposite point of view, the axial
mode was the fundamental mode up to 24 stages for t
base
= 50 N. Increasing t
base
to 100 N gave a fundamental mode shift at 16 stages. Thus, for very long masts the
fundamental mode is a bending mode and its frequency is independent of the level
of prestress.
The results for the 87stage mast are shown in Table 4.3. The ﬁrst two modes were
ﬂexural modes with frequencies of 0.037 Hz, which can be compared with 0.10 Hz
for the STRM ADAM. The frequency of the axial mode was 0.13 Hz for the lowest
prestress. For a cantilever beam, the lowest bending frequency is [120]
f
1,cb
≈
3.516
2π
EI
ml
4
. (4.23)
Substituting the data for the STRM ADAM, f
1,cb
= 0.10 Hz, m = 4.80 kg/m and
l = 60.68 m, into (4.23) and solving for EI yielded EI = 2.1 MNm
2
, which is about
6 times lower than the value given in Table 4.1. However, the frequency given in
Table 4.2: Data and results from the modal analysis of the tenstage mast. Axial
mode frequencies in bold face.
n θ
∗
η Length Mass m
c
m
s
¸
l
c
¸
l
s
(
◦
) (m) (kg) (kg/m) (kg/m) (m) (m)
10 10 0.2412 6.975 13.1 0.0136 0.2819 118.2 40.7
t
base
f
1
f
2
f
3
f
4
f
5
t
max
c
t
min
c
t
max
s
t
min
s
50 1.1787 2.8214 2.8468 4.1371 12.0952 99 48 −194 −156
100 1.6442 2.8224 2.8478 5.2359 12.6260 199 97 −388 −312
200 2.3006 2.8245 2.8499 6.8581 13.5714 397 194 −776 −623
500 2.8307 2.8562 3.6006 9.9618 15.6138 994 484 −1939 −1559
87
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
(a) (b) (c)
Figure 4.15: First three eigenmodes of a tenstage tensegrity mast (t
base
= 50 N
and θ
∗
= 10
◦
): (a) f
1
= 1.179 Hz, (b) f
2
= 2.821 Hz and (c) f
3
= 2.847 Hz.
Table 4.3: Data and results from the modal analysis of the 87stage mast. Axial
mode frequencies in bold face.
n θ
∗
η Length Mass m
c
m
s
¸
l
c
¸
l
s
(
◦
) (m) (kg) (kg/m) (kg/m) (m) (m)
87 10 0.2412 60.683 114.2 0.0134 0.2805 1047.9 357
t
base
f
1
f
2
f
3
f
4
f
5
t
max
c
t
min
c
t
max
s
t
min
s
50 0.0374 0.0374 0.1324 0.2339 0.2340 100 49 −196 −158
100 0.0374 0.0374 0.1872 0.2340 0.2341 199 98 −393 −315
200 0.0374 0.0374 0.2342 0.2342 0.2646 399 196 −785 −631
500 0.0375 0.0375 0.2347 0.2348 0.4174 997 490 −1963 −1577
Table 4.1 may include the 360 kg antenna attached to the tip. By Raleigh’s method
the ﬁrst frequency of a cantilever beam with a point mass M at its tip can be written
as [120]:
f
1,cb,pm
≈
1
2π
3EI
l
3
M +
33
140
ml
. (4.24)
Substituting M = 360 kg into (4.24) yielded EI = 12.6 MNm
2
, which is much closer
to the stated value. Inserting the values of the 87stage (or the tenstage) tensegrity
mast in (4.23) produced EI = 0.11 MNm
2
. Hence, the STRM ADAM is more than
100 times stiﬀer than the tensegrity mast in bending.
88
4.6. STRUCTURAL ANALYSIS
4.6.4 Static Analysis
For a complete analysis of the tensegrity mast, its static properties must be investi
gated. Static analyses of tensegrity masts with three struts per stage have previously
been performed by several researchers. Skelton and He [149] compute the axial stiﬀ
ness and the buckling load of a 186 m long, sixstage tensegrity mast for a deep sea
project. Sultan and Skelton [161] analyse the axial and torsional stiﬀness of a two
stage mast. As anticipated, the stiﬀness increases with the prestress level. A more
thorough study of the twostage mast stiﬀness is presented by Skelton et al. [150].
The axial stiﬀness increases as the struts become more vertical, i.e. decreasing co
latitude, and the prestress increases the axial stiﬀness in the case of small external
forces; as the external forces increase, the eﬀects of prestress can be neglected. The
bending stiﬀness is constant until one or more cables go slack. After a cable has
gone slack, the bending stiﬀness drops and becomes a nonlinear monotonically de
creasing function of the applied loading. As the colatitude increases, i.e. the struts
become more horizontal, the bending stiﬀness increases. Their ﬁnal observation is
that prestress does not aﬀect the bending stiﬀness of the mast provided the cables
are taut, but does delay the onset of cable slackening. Pak [118] analyses the axial,
bending and torsional stiﬀness of a fourstage mast. Again it is found that the axial
stiﬀness is almost equal in tension and compression and increases with the exter
nal loading. The bending stiﬀness of the four stage mast is constant until a cable
goes slack. The torsional stiﬀness is identical in the clockwise and counterclockwise
directions.
The studies above clearly show that the strength and stiﬀness in bending are the
critical properties of tensegrity masts. Therefore, the focus of this study was on
the bending properties of the ten and 87stage masts, but the axial stiﬀness was
also computed. As this analysis was not intended to be a parametric study, only
conﬁgurations with θ
∗
= 10
◦
were considered. Following the approach in reference
[150], four load cases were considered, Figure 4.16.
The masts were analysed by a geometrically nonlinear FEM, with the struts mod
elled by twonode bar elements and the cables by nocompression catenary elements,
cf. [171]. The special catenary element is advantageous in problems where cable
(a) (b) (c) (d)
F/3
F/3
F/3
F/3
F/3
F/3
F/3
F/3
F/3
F/3
F/3
F/3
Figure 4.16: Load cases for the tensegrity mast: (a) tension, (b) compression, (c)
bending in direction B1 and (d) bending in direction B2.
89
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
1.38
0.98
1.99
3.88
1.93
1
3.12
0.97
1.99
1.91
0.98
1.14
1.19
0.85
1.72
1
0.43
1.72
3.37
0.85
2.82
1.55
0.42
2.03
0.99
(a) (b)
Figure 4.17: Normalised forces in the two lowest stages of tenstage tensegrity masts:
(a) unstiﬀened and (b) stiﬀened.
slackening may occur.
Since the internal mechanism of the masts makes them weak in the axial direc
tion, a way of stiﬀening the mast was investigated. Three additional cables were
added to the ﬁrst stage so that it became fully triangulated, Figure 4.17. In the
original conﬁguration, these cable were unstressed, which cannot be accepted. One
way to prestress the cables was to further rotate the ﬁrst stage so that the struts
became longer. The resulting mast now had three independent states of selfstress,
s = 3, and no internal mechanism. Among these states of selfstress a rotationally
symmetric combination s
sym
was sought,
s
sym
= α
1
s
1
+ α
2
s
2
+ α
3
s
3
. (4.25)
The α’s were found by equating the element forces in the base cables to unity. The
element forces in the mast now changed. For the tenstage mast the element forces
in the ﬁrst two stages are shown in Figure 4.17, both for a normal and a stiﬀened
mast with an additional 15
◦
rotation of the ﬁrst stage. The results for the tenstage
masts are shown in Figures 4.18 and 4.19.
The behaviour of the unstiﬀened mast under axial loading was about the same
in tension and compression. In compliance with the ﬁndings by others, the mast
got stiﬀer as the load increased. The initial stiﬀness of the unstiﬀened mast was
275 kN/m. The response of the stiﬀened mast diﬀered slightly in tension and com
pression. Generally, the stiﬀened mast was stiﬀer in tension although the initial
stiﬀness was about the same, 420 kN/m. However, at a loading of 16.3 N three
diagonal cables on stage 2 went slack (marked as a discontinuity in the curve) and
the stiﬀness decreased. Under compression no cable became slack.
In bending, the diagonal cable on the compressed side at the second stage became
90
4.7. DISCUSSION
80 60 40 20 0 20 40 60 80
50
40
30
20
10
0
10
20
30
40
50
A
x
i
a
l
f
o
r
c
e
F
(
N
)
Axial displacement (mm)
2
7
5
N
/
m
m
4
2
0
N
/
m
m
s
t
i
f
f
e
n
e
d
Figure 4.18: Load–displacement curves for an unstiﬀened and a stiﬀened tenstage
tensegrity mast subjected to axial loading.
slack at F = 11.2 N for load direction B1. For direction B2, the same cable went
slack at 11.0 N. The bending stiﬀness was about the same in each direction before
and after cable slackening. The bending stiﬀness of the mast with no slack cables
was 110 kNm
2
, as determined by the vibration analysis. After a cable went slack, the
bending stiﬀness immediately dropped to 22 kNm
2
. The case of the stiﬀened mast
was a bit diﬀerent. For direction B1, one of the additional cables on the compressed
side of the mast already went slack at 3.9 N; for direction B2 it happened at 4.2 N.
For B1 the next cable to go slack was a diagonal cable at stage 3 and this occurred
at 9.8 N. For B2 it was a diagonal cable at stage 1 at the load of 8.6 N. A further
increase of the load resulted in more and more cables becoming slack and the bending
stiﬀness dropped suddenly at every occurrence as noted in Figure 4.19.
As regards the low bending stiﬀness of the tenstage mast, it was not necessary
to perform a complete analysis of the 87stage mast. For example, a lateral load
F = 1 N would give a tip displacement of 677 mm, hence the mast is much too
ﬂexible.
4.7 Discussion
Using the remarkably simple relationship by Nishimura [117], the formﬁnding of
straight, multistage tensegrity masts with an adequate prestress distribution is no
91
CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS
0 5 10 15 20 25 30
0
2
4
6
8
10
12
14
16
18
20
B1s
B2s
B1
B2
1
1
0
k
N
m
2
Lateral displacement at top (mm)
L
a
t
e
r
a
l
f
o
r
c
e
F
a
t
t
o
p
(
N
)
2
2
k
N
m
2
Figure 4.19: Load–displacement curves for an unstiﬀened and a stiﬀened tenstage
tensegrity mast subjected to lateral loading. Each discontinuity indi
cates that yet another cable has become slack.
longer an obstacle in design.
The proposed manufacturing scheme eventually turned out to be satisfactory when
all the early mistakes had been eliminated. The use of joints, with holes drilled
to the correct threedimensional angles, should eliminate the last obvious source of
inaccuracy.
The suggested strut deployment approach worked well. One inherent problem with
tensegrity masts are that they do not reach full stiﬀness until the last stage is
deployed. This seriously limits the applicability of these masts. However, it might
not be that diﬃcult to provide the deployed portion of the mast full stiﬀness. Since
most of the stages are equal, with nodes lying on a regular hexagon, it must be
possible to give the saddle cables the correct prestress and the saddle nodes the
proper restraints in order to prestress the deployed part of the mast. These restraints
must take on the role of the struts and the cables of the stage that is next in line to
be deployed.
One approach to reduce the risk of the cables getting tangled up in the struts is to
use tapelike instead of cordlike cables. Tapes have a natural folding direction and
it is presumably attainable to arrange the cables such that they do not get caught
by struts during deployment. This needs to be studied further.
92
4.7. DISCUSSION
The structural behaviour of the multistage mast concur with the predictions of
studies by other researchers; namely, the masts are relatively stiﬀ axially but very
weak in bending. The removal of the internal mechanism did not improve the
behaviour signiﬁcantly; the initial axial stiﬀness increased by about 50%, but the
bending stiﬀness cannot be said to have increased at all as the ﬁrst cable went slack
at a very low load level.
93
Chapter 5
Design Prerequisites for a
Deployable Reﬂector Antenna
This and the following two chapters will provide an indepth description of the de
velopment of a deployable reﬂector antenna aimed for small satellites. In the present
chapter, the background and motivation behind the new concept are described along
with examples of other small satellites using deployable structures. Stateoftheart
antenna concepts, which may be suitable for the present application, are studied.
Then, the new antenna structure is introduced.
A parabolic reﬂector antenna can be given two diﬀerent conﬁgurations: axisymmetric
or oﬀset; the geometries of these conﬁgurations will be presented. One particular
aspect, the required smoothness of the reﬂector surface, will signiﬁcantly aﬀect the
choice of structural concept. The eﬀects of the surface accuracy on the antenna
performance, accuracy characteristics of diﬀerent concepts, and surface accuracies
of existing deployable antennas will be presented. Finally, suitable materials for the
antenna structure will be selected.
This chapter can be seen as an introduction to reﬂector antennas with a require
ment speciﬁcation for the particular antenna, which will be developed in Chapters
6 and 7. This background information is necessary to fully appreciate decisions and
judgments made later on in the thesis.
5.1 Small Satellites and Deployable Structures
Conventional satellite technology has for several decades been focused on a small
number of large, complex spacecraft. Recently these satellites have been com
plimented by systems which use several smaller satellites in Low Earth Orbit
1
(LEO) [39]. Microsatellites, i.e., with a mass less than 50 kg, have been used
for technology tests and amateur radio for almost two decades, but they were not
1
A LEO often has an altitude below 1,000 km, where the radiation from the Van Allen belt is
low (p. 181 in [183]).
95
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
viewed as useful for space science research because of their small size [48]. In the
1980s, microsatellites in LEO were recognised due to their advantage in digital com
munication. Users send a message to the satellite during its passage and the message
could then be delivered throughout the world, thereby providing global mail service,
which even a large satellite in a Geosynchronous Orbit
2
(GEO) cannot provide [39].
Today, small satellites are developed for shorttime missions and quickly put to
gether by small teams. The use of fewer and more contemporary components also
helps to reduce the cost of the spacecraft [39]. Small satellites are therefore an
aﬀordable way to space for the military, universities and industry. Modern small
satellites are launched at low cost as secondary payloads, or socalled piggyback pas
sengers, along with a large spacecraft payload. NASA’s Space Shuttle has several
ways of accommodating payloads in its payload bay. Two of the carrier systems
are the Get Away Special and Hitchhiker, using cylindrical containers of diﬀerent
sizes mounted on the bay wall [39]. In Europe, Arianespace developed the Ariane
Structure for Auxiliary Payloads (ASAP) to accommodate secondary payloads on
the Ariane rockets. Mounted on the Ariane4 rocket the ASAP can carry six small
satellites, each with a mass of up to 50 kg. The ASAP on the Ariane5 rocket,
Figure 5.1, can accommodate eight satellites, each weighing less than 120 kg [3, 39].
Launch resources for secondary payloads often have a single price for payloads under
a certain mass limit, but, in general, they are volume rather than mass constrained.
However, it should be kept in mind that these secondary payloads do not have a
guaranteed launch date [39].
Deployable structures have been used on small satellites since the beginning of the
space programme. In recent years they have fallen into disfavour due to their higher
risk, complexity, and several failures [136], e.g. the highgain antenna on Galileo. In
order to facilitate the design of reliable mechanisms, it is important to recognise the
risk associated with such structures from the onset of the mission programme [136].
A few examples of small and microsatellites with deployable structures are given
below.
5.1.1 MicroSatellites Astrid1 and 2
Sweden’s ﬁrst microsatellite, Astrid1, Figure 5.2(a), was launched on January
24, 1995 from the Plesetsk Cosmodrome in Russia as a piggyback passenger on
a Kosmos3M launch vehicle to an altitude of about 1,000 km. Astrid1 was de
signed and developed by Swedish Space Corporation (SSC). The total mass of
Astrid1 was 27 kg. With solar panels stowed, the dimensions was approximately
0.45 0.45 0.29 m
3
, and with solar panels deployed 1.1 1.1 0.46 m
3
(including
antennas). The mass and dimensions were chosen because they represented about
half the maximum permitted values of a microsatellite on the Ariane4 ASAP [48].
In total, the four solar panels, each 0.39 0.29 m
2
, produced 42 W.
2
The period of a GEO is 23 hours 56 minutes and 4 seconds, matching the Earth’s rotational
motion.
96
5.1. SMALL SATELLITES AND DEPLOYABLE STRUCTURES
ASAP
STRV
Figure 5.1: DERA’s small satellites STRV1 C and D on Ariane5 ASAP for launcher
ﬁt check (Courtesy of DERA).
The second Swedish microsatellite, Astrid2, Figure 5.2(b), also designed by SSC,
was launched on December 8, 1998 from the same site, on a similar launch vehicle,
and to a similar altitude as Astrid1. The total mass of Astrid2 was 29 kg. With
solar panels stowed, the size was 0.950.450.30 m
3
, and with solar panels deployed
1.7 1.1 0.3 m
3
. The total power of the six solar panels, each 0.39 0.29 m
2
, was
80 W. The design lifetime for Astrid2 was one year, but on July 24, 1999 contact
with the satellite was lost and never reestablished despite several attempts [163].
On Astrid1, the solar panels are the only deployable structures. The design is simple
and reliable; one pair of hinges with a common axis of rotation on each solar panel.
On Astrid2, a similar deployment technique was used for four of its six solar panels,
while the remaining two panels were nondeployable. Astrid2 also had a twohinged
deployable boom, oriented along its spin axis, which carried a device for attitude
determination plus one of the scientiﬁc payloads. It appears from Figure 5.2(b) that
the boom unfolds in a plane and that a cable is used to stop the unfolding and to
correctly position the boom.
5.1.2 Small Satellite Odin
Another satellite developed by the SSC is the small satellite Odin. The primary
payload on Odin is a 1.1 metre nondeployable, solid surface, oﬀset reﬂector antenna.
The mass of the main reﬂector is 5.5 kg, and the total antenna mass, including
the subreﬂector and supporting structure, is 9.9 kg. Odin has a total mass of
250 kg, a height of 2.0 m, and a base area of 1.1 1.1 m
2
with stowed solar panels,
97
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
(a)
(b)
Figure 5.2: The Swedish microsatellites (a) Astrid1 and (b) Astrid2 with solar
panels deployed (Courtesy of Swedish Space Corporation).
Figure 5.3(a), and 3.8 3.8 m
2
with deployed solar panels, Figure 5.3(b). Odin was
launched on February 20, 2001 from Svobodny, Russia, as the only payload on a
Start1 launch vehicle to an altitude of 600 km. A piggyback launch would have
been a more economically viable alternative, but that would have compromised the
scientiﬁc objectives and led to a more complex satellite design [164]. The four solar
panels, each approximately 1.4 0.75 m
2
, provided 340 W of power. The design
lifetime for Odin is two years.
The Odin satellite is signiﬁcantly larger than the Astrid satellites in order to ac
commodate the reﬂector antenna, solar reﬂectors and other scientiﬁc instruments.
Further, given the high operating frequency of the antenna (580 GHz), only a solid
reﬂector surface can give the required surface accuracy (10 µm). However, a deploy
able solid surface antenna could have been used instead, but at a higher mission
risk.
5.1.3 Space Technology Research Vehicles
In 1989, the Defence Evaluation and Research Agency (DERA) at Farnborough
3
in
United Kingdom started the Space Technology Research Vehicle (STRV) programme
as a single microsatellite project. The project objectives were to show that small,
low cost satellites could oﬀer aﬀordable spacebased research to civil government,
industry, and academic organisations, and also provide support to military defence
research objectives [32].
3
DERA Farnborough was renamed QinetiQ on July 2, 2001, but will in this thesis still be
referred to as DERA.
98
5.1. SMALL SATELLITES AND DEPLOYABLE STRUCTURES
(a) (b)
Figure 5.3: The Swedish small satellite Odin with solar panels (a) stowed, and (b)
deployed (Courtesy of Swedish Space Corporation).
The ﬁrst two microsatellites of the programme, STRV1 A and B, were launched on
June 17, 1994 from the Guiana Space Centre, Kourou in French Guiana. They were
piggyback passengers on an Ariane4 launch vehicle to a Geosynchronous Transfer
Orbit (GTO) [32]. The dimensions of both STRV1 A and B were 0.45 0.45
0.47 m
3
, while the mass was 50 and 53 kg, respectively, due to diﬀerent scientiﬁc
payloads. The satellites had solar panels on four of its sides, each approximately
0.40 0.45 m
2
, providing a total power of 31 W [32]. Unlike the Astrid satellites,
STRV1 A and B had no deployable structures, e.g. solar panels. The design lifetime
of STRV1 A and B was one year, but they continued conducting orbital operations
until September 1998 when they were donated to the University of Colorado at
Boulder, USA, allowing students to operate the satellites [32].
Following the success of STRV1 A and B, DERA designed two new satellites, STRV
1 C and D, which could accommodate more scientiﬁc payloads and thereby widen
the scope of mission objectives [32]. The new payload required a larger spacecraft
structure, capable of generating 80 W, which is more than twice the power available
on STRV1 A and B. The dimensions of STRV1 C and D were 0.7 0.6 0.7 m
3
,
Figure 5.4. Each satellite weighed about 100 kg. STRV1 C and D were launched
on November 15, 2000 from the same site as their precursors, again as auxiliary
payloads, but this time on an Ariane5 launch vehicle to a GTO. However, only
two weeks after launch, the mission was brought to a premature ending when the
receivers of both satellites were cut from their power source because of a design mis
take. This means that the satellites cannot receive anything sent from the ground,
including recovery commands [145].
99
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
5.1.4 Future STRV Missions
In the spring of 2000, DERA started the plans for future STRV missions. The main
areas of interest include a space based Global Positioning System (GPS) for attitude
and orbit determination, and miniature remote sensing [32]. These tasks would
necessitate the use of deployable structures and outline requirements are formulated
for a range of such structures by DERA in reference [136]. These requirements can
be divided into three categories:
1. Booms for space based GPS applications,
2. Synthetic Aperture Radar (SAR) antenna for remote sensing, and
3. Solar arrays of 1.5–2 kW, needed to run the SAR antenna.
In remote sensing, a high image resolution is the primary goal. Generally, a higher
resolution is obtained using a larger antenna, but placing a large antenna in space
is extremely expensive. By using the motion of the spacecraft and advanced signal
processing techniques, a larger antenna can be simulated. This is, in very simple
terms, the principle behind the SAR technology. The performance of a SAR antenna
is aﬀected by a number of parameters but one especially important is its size. Hence,
one of the biggest limitations in performing a SAR mission using a small satellite,
is the problem of stowing a large antenna in a small volume. Various SAR antenna
options are available [136]:
• Parabolic reﬂector,
• Planar array, or
• Reﬂect array.
In DERA’s requirements, no speciﬁc mission had been identiﬁed, so the stowage
constraints for the deployable structures were assumed to be deﬁned by the envelope
around the STRV1 C and D satellites on the Ariane5 ASAP [136], Figure 5.4.
DSL was given the task of designing deployable structures to meet the requirements
by DERA. As most of the existing technologies of deployable structures are aimed
for large spacecraft applications, it was soon found that new technology was needed
to meet the requirements by DERA. In response to these requirements, three novel
deployable structures, related to categories 2 and 3 above, were developed. These
are presented in reference [130]. In that study, the author was involved in the
development of a 3 m diameter parabolic reﬂector SAR. The operating frequency
for this antenna is assumed to be in the Xband (9.65 GHz) as most recent systems
use this frequency. The mass goal for the main reﬂector alone is 20 kg. Including
electronics, feed, support, etc. the total payload mass goal is 150 kg [136].
The preliminary study of the SAR reﬂector antenna, performed by Pellegrino and the
author, is presented in references [130] and [169]. This study will now be signiﬁcantly
100
5.2. NEW REFLECTOR CONCEPT FOR SMALL SATELLITES
8
0
0
900
700
ASAP
(100) (100)
∅ 140
∅ 200
3
4
9
9
1
8
5
W
A
V
E
R
8
5
W
A
V
E
R
8
0
0
9
1
ASAP
∅ 140
∅ 200
(a) Front view (b) Side view
Figure 5.4: Envelope of the STRV on ASAP ring (Courtesy of DERA).
expanded. The aim of the following and the next two chapters is to provide a
thorough feasibility study of the new reﬂector concept.
5.2 New Reﬂector Concept for Small Satellites
The main diﬃculty in meeting the requirements by DERA is the very stringent
constraints on the dimensions of the packaged envelope, 800 200 100 mm
3
. A
concept which would easily conform to these dimensions is the inﬂatable reﬂector.
Much work is currently being done on inﬂatable reﬂectors, mainly in the USA,
cf. [42, 67], but still they cannot be regarded as a mature technology. An inﬂatable
antenna is therefore not considered as an alternative for the present application.
In addition, the required stowed size immediately rule out all umbrellatype mesh
antennas as the hubs of these are too large. The stowed size of these antennas
also tend to be more cubic than oblong that is needed here, e.g. the Collapsible
RibTensioned Surface (CRTS) reﬂector [78]. Many umbrellatype antennas also
rely on active control of the surface shape which complicates the overall design, e.g.
the HRA. A better design would be one that passively, i.e. without active control,
provides the required surface accuracy. This naturally leads us to the concept of the
tension truss, brieﬂy described in section 2.3.1.
101
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
5.2.1 Existing Concepts with Passive Structure
The research on antennas which passively, i.e. without active control, achieve the
required surface accuracy started in the early 1980s with NASA’s Large Deployable
Reﬂector (LDR) programme, cf. [96].
The current stateoftheart of high precision antennas with a passive structure is
the AstroMesh antenna. The stowed height of the AstroMesh is given by the length
of one vertical plus one horizontal strut. For a given aperture and number of bays it
is not possible to change the length of the horizontal struts. It is, however, possible
to shorten the vertical struts by making the rear net shallower, say half the depth of
the front net, if larger forces in the ring are accepted. For example, a 3 m diameter
reﬂector with F/D = 0.4 and a ring truss divided into 18 bays would, with identical
front and rear nets and no separation between the nets, have a stowed height of
1.46 m. If the height of the rear net is halved, the stowed height is 1.22 m. In
general, an AstroMesh reﬂector with small F/D requires a high ring truss with too
high stowed height for the present application. An alternative ring structure based
on a pantograph concept, cf. [188], with, again, 18 bays would have a packaged
height of about 0.9 m. However, this requires a large number of joints which add a
signiﬁcant mass to the structure. The tension truss antenna for the HALCA satellite,
Figure 2.14, has, like the umbrellatype antennas, a large hub and is therefore not
suitable for the present application.
Although none of the existing designs are directly applicable, it is considered that
an adaptation of the AstroMesh concept oﬀers the greatest potential for meeting
the requirements with a lowcost system. The main parts of the new concept is
presented below.
5.2.2 Tensegrity Reﬂector Concept
The new reﬂector antenna concept, Figure 5.5, is based on the tensegrity and tension
truss concepts. Like the AstroMesh antenna, it is composed of three main parts:
• a deployable ring structure, Figure 5.5(b),
• two identical cable nets (front and rear nets), Figure 5.5(b), connected by
tension ties, Figure 5.5(c), and
• the reﬂecting mesh, attached to the front net, Figure 5.5(a).
The complete antenna structure is composed of a large number of tension elements,
i.e. the net cables and constanttension springs, and only six compression elements,
i.e. the struts. The deployable ring structure is a onestage tensegrity module with
six struts which has one state of selfstress and seven internal mechanisms of in
extensional deformation. The tensegrity module alone is very ﬂexible, but as will
be shown in the next chapters, the complete antenna structure is quite stiﬀ.
102
5.2. NEW REFLECTOR CONCEPT FOR SMALL SATELLITES
(
a
)
(
b
)
(
c
)
(
d
)
F
i
g
u
r
e
5
.
5
:
T
h
e
n
e
w
t
e
n
s
e
g
r
i
t
y
r
e
ﬂ
e
c
t
o
r
c
o
n
c
e
p
t
f
o
r
s
m
a
l
l
s
a
t
e
l
l
i
t
e
s
:
(
a
)
t
h
e
c
o
m
p
l
e
t
e
s
t
r
u
c
t
u
r
e
w
i
t
h
R
F
r
e
ﬂ
e
c
t
i
v
e
m
e
s
h
a
t
t
a
c
h
e
d
t
o
t
h
e
f
r
o
n
t
n
e
t
,
(
b
)
t
h
e
t
e
n
s
e
g
r
i
t
y
r
i
n
g
s
t
r
u
c
t
u
r
e
,
(
c
)
t
h
e
f
r
o
n
t
a
n
d
r
e
a
r
n
e
t
s
,
a
n
d
(
d
)
t
h
e
t
e
n
s
i
o
n
t
i
e
s
c
o
n
n
e
c
t
i
n
g
t
h
e
n
e
t
s
.
103
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
Although the concept above is new, it consists of several known substructures put
together to meet the requirements of stowage size. The main diﬀerence compared to
previous concepts must be the deployable ring structure. The present ring structure
was inspired by new ideas of using tensegrity for deployable space antennas, cf. [71].
5.3 Geometry of Parabolic Reﬂector Antennas
Reﬂector systems are the basic types for satellites because of their low mass, low
complexity and cost, and design maturity. Two important properties make the
parabolic reﬂector, where the reﬂecting surface is a paraboloid of revolution, espe
cially useful as an antenna [158]:
• Incident rays parallel to the reﬂector axis converge to a spot known as the focal
point. Conversely, all rays leaving the focal point are parallel to the reﬂector
axis after reﬂection from the parabolic surface.
• All path lengths from the focal point to the reﬂector and onto the aperture
plane are the same and equal to 2F, where F is the focal length, i.e. the
distance from the apex of the paraboloid to the focal point, Figure 5.6.
Basically, a reﬂector antenna consists of two components: a reﬂecting surface that is
large relative to the operating wavelength and a much smaller subreﬂector, placed
close to the focal point of the main reﬂector. The role of the subreﬂector is to
redirect the rays reﬂected in the main reﬂector to the feed.
5.3.1 AxiSymmetric Reﬂector
The simplest antenna has an axisymmetric parabolic reﬂecting surface. In a Carte
sian coordinate system, xyz, this surface is described by
z =
x
2
+ y
2
4F
. (5.1)
It is most common to specify the reﬂector in terms of the diameter D and focal
lengthtodiameter ratio F/D, which give the size and curvature, respectively [158].
The angle θ
0
from the Zaxis to the reﬂector rim is
θ
0
= 2 arctan
D
4F
. (5.2)
As F/D →∞, the reﬂector becomes planar and when F/D = 0.25, the focal point
lies in the plane passing through the reﬂector rim. It is sometimes useful to know
the height of the reﬂector H
0
, which is computed as:
H
0
=
D
2
16F
. (5.3)
104
5.3. GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS
0.5D 0 0.5D
0
0.5D
F/D = 0.3
0.4
0.5
θ
0
Figure 5.6: Parabola shapes for F/D = 0.3, 0.4, 0.5.
A disadvantage of axisymmetric reﬂectors is that the feed antenna and its support
structure in many cases are bulky and block the incident rays, resulting in reﬂec
tor performance degradation. To eliminate or reduce aperture blocking reﬂector
antennas with oﬀset feeds are used [158]. An oﬀset parabolic reﬂector may also sim
plify the satellite design as the feed can be contained within the satellite structure,
without the need of a support structure [183].
5.3.2 Oﬀset Reﬂector
The oﬀset parabolic antenna is constructed by considering a parent paraboloid with
diameter D
p
, Figure 5.7. Then a cylinder of radius R
a
, whose axis is parallel to the
Zaxis, is deﬁned in the ﬁrst and fourth quadrants of the XY plane. The intersection
of this cylinder and the parent paraboloid is a plane ellipse which deﬁnes the rim
of the oﬀset reﬂector. In Figure 5.7, the surface of the oﬀset reﬂector is visualised
by the intersections of several concentric cylinders and the parent paraboloid. Also
shown are the coordinate systems xyz and x
y
z
, both with origins at O in the
centre of the oﬀset reﬂector. Referring to Figure 5.8, the following relations are
derived [78]:
φ
a
= arctan
X
O
2F
, (5.4)
x
z
=
X
Z
−
X
O
X
2
O
/4F
, (5.5)
and
x
z
=
¸
cos φ
a
sin φ
a
−sin φ
a
cos φ
a
x
z
. (5.6)
Note that Y = y = y
. For a given point (x
, y
) the z
coordinate is found by
substituting (5.5) and (5.6) into (5.1). Simplifying gives the following quadratic
105
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
F
Z
Y
X
x
z
y, y
P
A
O
B x′
z′
′
Parent paraboloid
Offset reflector
Figure 5.7: Threedimensional view of the oﬀset reﬂector antenna on parent
paraboloid.
equation [78]:
sin
2
φ
a
. .. .
A
(z
)
2
−[4F cos φ
a
+ 2 sin φ
a
(X
O
+ x
cos φ
a
)]
. .. .
B
z
+
(y
)
2
+ x
2X
O
cos φ
a
+ x
cos
2
φ
a
−4F sin φ
a
. .. .
C
= 0,
(5.7)
with the solutions
z
1,2
=
−B ±
√
B
2
−4AC
2A
. (5.8)
Only the “−” solution is of interest as the other, much larger value, corresponds to
the intersection of the z
axis and the part of the parent paraboloid in the second
and third quadrants, i.e., X < 0. The depth of the oﬀset reﬂector, H
a
, can be found
by (5.5) and (5.6) as:
H
a
=
D
2
a
16ξF
, (5.9)
where ξ is the ellipticity, i.e., the ratio of the major and minor axis of the reﬂector
aperture, calculated as:
ξ =
1 +
D
a
+ 2X
A
4F
2
. (5.10)
Note that the height of the axisymmetric reﬂector, (5.3), is reobtained for X
A
=
−D
a
/2.
106
5.3. GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS
F
Z
O
X
O
O
X
A
X
B
A
B
X
Z
x
z
z ′
x ′
Parent paraboloid
Offset reflector
P
φ
a
Figure 5.8: Oﬀset reﬂector geometry deﬁnitions.
Standard Conﬁguration
The normal choice for O is X
O
= (X
B
+ X
A
)/2 = X
A
+ R
a
. This reﬂector conﬁg
uration is therefore called the standard conﬁguration [78]. In the xyz system, the
coordinates of a point C on the reﬂector rim, Figure 5.9, are
x
C
= R
a
cos ϕ, (5.11a)
y
C
= R
a
sin ϕ, (5.11b)
z
C
=
R
a
4F
(R
a
+ 2X
O
cos ϕ) . (5.11c)
The coordinates of point C in the x
y
z
system are obtained by (5.6). Two views of
the standard conﬁguration are shown in Figure 5.9. The ﬁrst view, Figure 5.9(a), is
along the zaxis; the second view, Figure 5.9(b), is along the z
axis. Note that in
Figure 5.9(b) the reﬂector centre O does not coincide with the centre of the elliptic
rim. For the CRTS reﬂector, Lai [78] observed that the standard conﬁguration is
not ideal and therefore proposed a new conﬁguration, the central hub conﬁguration,
with two axis of symmetry.
Central Hub Conﬁguration
For this conﬁguration point O is in the centre of the elliptic reﬂector rim. The posi
tion of the central hub, X
O
, is determined by the following relation, Figure 5.9(b):
x
A
+ x
B
= 0, (5.12)
107
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
(a) (b)
y
x x ′
y ′
A B
C
A B
C
R
a
ϕ ϕ
Figure 5.9: Standard oﬀset conﬁguration projected onto (a) xy plane and (b) x
y
plane.
where
x
A
= (X
A
−X
O
) cos
arctan
X
O
2F
+
X
2
A
−X
2
O
4F
sin
arctan
X
O
2F
(5.13)
and
x
B
= (2R
a
+ X
A
−X
O
) cos
arctan
X
O
2F
+
(2R
a
+ X
A
)
2
−X
2
O
4F
sin
arctan
X
O
2F
,
(5.14)
see [78] for further details. Equation (5.12) is solved numerically for X
O
using e.g.
the function fzero in Matlab [89]. In the x
y
z
coordinate, system the x
and y
coordinates of a point on the elliptic reﬂector rim are
x
= x
A
cos ϕ
, (5.15)
y
= R
a
sin ϕ
, (5.16)
while the z
coordinate is found by (5.7). As anticipated by Lai [78], the central
hub conﬁguration produces a better membrane prestress distribution in the CRTS
reﬂector than the standard conﬁguration. Note that for the central hub conﬁgura
tion, the z
axis is not normal to the elliptic aperture plane which is the case for
the standard conﬁguration. To make it normal, the local system must be rotated
an angle ∆φ
a
, which is the diﬀerence between φ
a
for the standard conﬁguration and
φ
a
for the central hub conﬁguration.
108
5.3. GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS
(a) (b)
y y ′
x x ′
A B
C
A B
C
R
a
ϕ ϕ
Figure 5.10: Central hub oﬀset conﬁguration projected onto (a) xy plane and (b)
x
y
plane.
Circular Conﬁguration
Lai [78] also considered a third conﬁguration where the aperture projection on the
x
y
plane is a circle instead of an ellipse. This conﬁguration, called the circular
conﬁguration, was chosen for the CRTS reﬂector as it produces a better membrane
prestress distribution and more uniform loading of the ribs, than the two previous
conﬁgurations. For the circular conﬁguration, the reﬂector rim is not coplanar as
in the other conﬁgurations. While this aspect was not a problem for the CRTS
umbrellatype reﬂector, it is probably undesirable for a deployable reﬂector with a
supporting ring structure. The circular conﬁguration will therefore not be further
considered in this study.
5.3.3 Values for F/D and X
A
The values for the antenna diameter, focal length and oﬀset distance are determined
by taking a number of aspects, both electrical and structural, into account. It is
always desirable to have as large a diameter as possible for increased gain. When it
comes to the F/D ratio and oﬀset distance X
A
, it seems that these are determined by
special optimisation routines not immediately accessible to structural engineers [12,
84]. From a structural viewpoint, a large F/D ratio means a bulky supporting
structure for the subreﬂector which unnecessary adds weight to the whole assembly.
Regarding the oﬀset distance, X
A
, it should be large enough to eliminate the feed
blockage. However, a larger oﬀset means a larger ellipticity factor which may have
some implications for the present concept concerning the prestressability of the
tension truss. The simplest way to get an idea of the appropriate values for these
antenna parameters is to look at values used for other deployable space antennas.
109
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
Table 5.1: Geometrical data for some axisymmetric and oﬀset deployable reﬂector
antennas.
Type
a
Structure/satellite D F X
A
F/D
b
ξ Ref.
(m) (m) (m)
M
Aerospatiale 5.00 1.90 0.70 0.167 1.307 [141]
AstroMesh
c
antenna 23.56 15.70 3.92 0.286 1.118 [62]
Fan rib antenna 3.40 2.50 0.04 0.363 1.059 [66]
BAe/Surrey antenna 5.00 3.00
d
0.50 0.273 1.120 [141]
S
ODIN radiometer 1.10 0.61 0.05 0.265 1.115 [148]
SSDA 1.48 0.62 — 0.423 — [51]
I
ARISE antenna 25.00 11.55 0.00 0.231 1.137 [20]
Space rigidised antenna 5.53 2.13 — 0.385 — [174]
a
M: mesh, S: solid, and I: inﬂatable.
b
For oﬀset reﬂectors the diameter of the parent reﬂector, D
p
= 2(D +X
A
), is used.
c
Values for the AstroMesh onboard the Thuraya satellite is not known, so the values
for the antenna from the ﬁrst study, [62], is used instead.
d
The value of 1.5 m given in Table 3 in [141] is corrected to 3.0 m stated elsewhere.
This was, however, not such a simple task as in many cases only parts of the sought
information were provided. It is likely that such information is classiﬁed and cannot
be made public. The antennas for which all parameters were available are listed in
Table 5.1. The F/D ratio varies between 0.167 and 0.363 for the parent paraboloid,
which means that they are quite deep. The ellipticity interval is 1.056–1.307, with
most values around 1.1. X
A
/D
a
varies from 0, which is the absolute minimum, to
0.166.
5.4 Reﬂector Antenna Theory
Despite the title, this section will not aim at describing the complete theory of
reﬂector antennas, as it is well beyond the scope of this thesis. Nevertheless, within
that theory there are some aspects directly related to the structural design of the
antenna. These aspects will be presented here.
5.4.1 Antenna Gain
In communication systems, an increase in aperture size means an increase in the
amount of information that can be sent or received. In sensing systems, however,
it means that a ﬁner ground resolution can be obtained. The factor describing
the performance of the antenna is its gain. High gain antennas allow for high
transmission data rates at low power and improve the signal to noise ratio [158].
110
5.4. REFLECTOR ANTENNA THEORY
For a reﬂector antenna, the theoretical maximum gain is
G
th
=
πD
λ
2
. (5.17)
where λ is the wavelength of the operating frequency. This maximum gain is reduced
by a number of factors, e.g. aperture blockage and phase errors, to
G
0
= η
a
G
th
(5.18)
where η
a
is the product of all antenna eﬃciency factors η
i
, 0 < η
i
≤ 1.
The eﬃciency factor due to aperture blockage, η
b
, can be estimated as [158]:
η
b
=
1 −
1
η
t
A
b
A
a
2
(5.19)
where A
b
is the blockage area projected onto the aperture area A
a
, and η
t
the aper
ture taper eﬃciency factor. This takes into account that the central part of the
aperture usually is more aﬀected by the blockage. For optimum antenna perfor
mance, it is shown that η
t
≈ 0.89 [158].
Phase errors arise due to surface deviations from the ideal paraboloid. Considering
their impact on the antenna radiation pattern, Pontoppidan [131] divides the surface
deviations into the following three groups:
1. Deviations which change the desired paraboloid into another bestﬁt paraboloid.
Deviations due to slowly varying distortions, like thermal distortions, and their
eﬀect on the performance are completely described by a defocus. The bestﬁt
paraboloid, i.e. a paraboloid which in a leastsquare sense best approximates
to the ideal one, may deviate from the ideal one up to about one wavelength
without seriously aﬀecting the performance of the antenna.
2. Other systematic errors.
Systematic surface deviations are inherent in the construction of the antenna
and can be predicted. In general, systematic deviations cannot be considered
similar to random errors, but have to be treated separately.
3. Other random errors.
Random errors are caused by the fabrication tolerances in the manufacturing
process and are unpredictable within given statistical limits.
According to Stutzman and Thiele [158], many applications have η
a
≈ 0.65. Values
of η
a
between 0.43 and 0.72, depending on the frequency, are reported by van ’t
Klooster et al. [174]. The inﬂuence of random errors on the antenna performance is
treated in the following section.
111
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
5.4.2 Eﬀects of Random Surface Errors
An approximate method to compute the eﬀects of random surface deviations on the
antenna gain is presented by Ruze [142]. He assumed that the surface deviation at
any point is a random sample from a single Gaussian distribution with zero mean and
a standard deviation equal to the rootmeansquare (rms) of the surface deviation
of the reﬂector. A further assumption is that the surface deviations are correlated
in small regions. Under these assumptions, the random error eﬃciency factor, η
r
,
can be written as:
η
r
= exp
¸
−
4πδ
rms
λ
2
¸
, (5.20)
where δ
rms
is the radiometric rms surface deviation [50, 142]. The radiometric rms
surface deviation is deﬁned as [50]:
δ
rms
=
1
A
a
A
a
∆℘
2
2
dA
¸
¸
1/2
, (5.21)
where ∆℘ is the phase error, i.e. pathlength diﬀerence of a reﬂected ray due to
the surface imperfections, and A
a
the aperture area. Usually, the surface deviations
are measured in a direction axial or normal to the reﬂector surface for reasons of
simplicity. The relation between the deviation in the axial direction, ∆z, and ∆℘
is [50, 142, 184]
∆℘ =
2∆z
1 + (r/2F)
2
, (5.22)
while for the deviation in the normal direction, ∆n, it is
∆℘ =
2∆n
1 + (r/2F)
2
, (5.23)
where r is the distance from the centre of the axisymmetric reﬂector to the point of
measure. For shallow reﬂector surfaces, the factor 1 +(r/2F)
2
is close to unity and
even for deeper ones it is small, e.g., F/D = 0.4 and r = D/2 gives 1 + (r/2F)
2
=
1.39. Assuming that the denominator can be set to unity, (5.21) simpliﬁes to
δ
rms,z
=
1
A
a
A
a
(∆z)
2
dA
¸
¸
1/2
. (5.24)
Equation (5.24) is simpler than (5.21) and frequently used to predict the performance
of a reﬂector antenna, cf. [2,50]. Using δ
rms,z
in (5.20), will overestimate the gain loss
as δ
rms,z
≥ δ
rms
. However, later studies [184, 189] show that the assumptions related
to (5.20) in general underestimate the gain and therefore can be considered to be
a worst case. Although (5.20) relates only to random surface deviations, it is with
good approximation found to be valid for any type of deviation measured relative
to the best ﬁt paraboloid [131]. Thus, (5.24) is likely to give a good approximation
112
5.4. REFLECTOR ANTENNA THEORY
of the radiometric rms surface deviation and antenna gain loss and can therefore
safely be used.
An important result from (5.20) is that if a given reﬂector operates at increasing
frequency, the gain, at ﬁrst, increases as the square of the frequency until the surface
deviation eﬀect takes over and then a gain deterioration occurs. Maximum gain is
obtained at the wavelength of
λ
max
= 4πδ
rms
, (5.25)
Substituting λ
max
into (5.20) gives the maximum gain
G
max
≈
η
a
43
D
δ
rms
2
, (5.26)
which is proportional to the square of the manufacturing accuracy D/δ
rms
[142].
5.4.3 Systematic Surface Error of Faceted Paraboloids
Many reﬂector antennas, e.g. AstroMesh [165] and ETSVIII [107], divide the re
ﬂector surface into ﬂat facets rather than gores as is the case for umbrellatypes.
Figure 5.11 shows a shallow spherical cap approximated by triangular and hexago
nal ﬂat facets. Faceting introduces a systematic deviation of the actual surface from
the desired surface which degrade the performance of the reﬂector. As shown in the
previous section, the parameter governing the performance degradation is the rms
surface deviation. Thus, the maximum facet size required to meet a speciﬁc surface
accuracy is sought.
In general, it is not possible to map regular polygons onto a curved surface. To
ﬁnd an expression for the relation between the surface error and the facet size some
assumptions have to be made [2, 58]. For a shallow reﬂector, with focal length F,
a six bay triangular division, Figure 5.11(a), results in nearly equilateral triangles.
If the reﬂector is shallow, it can be closely approximated by a sphere with radius
2F. Thus, the δ
rms,z
calculation for an equilateral triangle on a spherical surface is
assumed to be a good approximation for the actual geometry. This assumption was
checked by Agrawal et al. [2] and found to be valid.
Hedgepeth [56,58] provides the ﬁrst and most thorough study of rms errors of faceted
mesh antennas. Beside the surface deviation due to the facet size, this study also
includes eﬀects of mesh saddling. With no lateral loading an isotropic uniformly
tensioned membrane must have zero Gaussian curvature, i.e. saddle shape. At the
intersection between adjacent facets on the paraboloid, the mesh tension changes
direction. The resulting lateral loading tends to curve the supporting element in
wards, Figure 5.12. The corresponding normalised rms surface deviation δ
rms,z
of
the bestﬁt facet and a sphere of radius 2F is, [56],
δ
rms,z
D
= 0.01614
(/D)
2
F/D
1 + 0.33
p
t
, (5.27)
113
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
(a) (b)
Figure 5.11: Approximating a shallow spherical dish with ﬂat facets of (a) triangular
and (b) hexagonal shape.
p
(a) (b)
Figure 5.12: Mesh saddling of triangular facets on a synclastic surface.
where is the side length of the triangle, p the mesh tension, and t the force in the
supporting element. To minimise the eﬀects of mesh saddling, i.e., allow the largest
facet size, the tendon force must be large compared to the mesh tension.
The rms deviation of facets including mesh saddling was recently reanalysed by
Tanizawa et al., as reported in [106]. By applying a Fourier series solution, instead
of a polynomial as Hedgepeth [56], for the deformation of the supporting tension
element, Figure 5.12, the rms deviation of the equilateral triangle was found to be
δ
rms,z
D
= 0.01614
(/D)
2
F/D
1 + 0.660
p
t
+ 0.133
p
t
2
(5.28)
for p/t < 1. A Taylor series expansion of the square root part of (5.28) around
p/t = 0 gives 1 + 0.33p/t +O[(p/t)
2
], which is similar to the corresponding part
of (5.27). For practical values of p/t, the diﬀerence between (5.28) and (5.27) is
negligible.
Agrawal et al. [2] compute the rms surface deviation between bestﬁt ﬂat facets of
triangular, square, and hexagonal shape and a sphere. For equilateral triangles it is
δ
rms,z
D
=
1
16
√
15
(/D)
2
F/D
, (5.29)
i.e., identical to (5.27) as p/t → 0. For a given value of δ
rms,z
the required side
length is
D
= 7.872
F
D
δ
rms,z
D
. (5.30)
114
5.4. REFLECTOR ANTENNA THEORY
For squares and regular hexagons, the value 7.872 is substituted by 6.160 and 4.046,
respectively. Hence, the triangles have the largest side length for a given rms de
viation tolerance. Agrawal et al. [2] also analyse the electrical performance of a
proposed large reﬂector antenna (D = 660 m and F = 573 m) with an operation
frequency of 1 GHz and an allowable surface error, δ
rms,z
, of 12 mm (λ/25). It is
found that hexagonal facets provide a slightly better radiation pattern than triangu
lar facets. A reﬂector surface made out of square facets with identical δ
rms,z
yield a
worse radiation pattern, which is explained by the variation of the surface deviation
over the reﬂector. Although hexagonal facets yield a better radiation pattern, tri
angular facets are structurally desirable and require the fewest number of structural
members to meet a given surface tolerance. Considering all factors, their conclu
sion is that triangles appear to be the best option for approximating a paraboloidal
reﬂector surface.
Referring to the research by Agrawal et al. [2], Fichter [38] proposes a method of
changing the boundary shape of the facets so that the total rms error could be
improved. For congruent facets with a boundary radius of 2F, it is found that
δ
rms,z
= 0.0082
2
/F, which is about 49% below the ﬂat facet rms error. The mini
mum rms deviation was found to be δ
rms,z
= 0.0069
2
/F for a boundary radius of
1.4F. This optimum solution can also be reobtained by minimising (5.28) with
respect to p/t, which yields p/t = −2.474. Hence, to realise this optimum facet
as well as the congruent one, the boundary members must be in compression which
requires elastic rods rather than cables along the facet boundaries. The optimum
solution for a cable supported facet is, therefore, the ﬂat facet.
Hedgepeth [56] also studies the rms surface deviation of umbrellatype reﬂectors
having gores rather than facets. Including the mesh saddling between the parabolic
ribs, the rms surface deviation is
δ
rms,z
D
= 0.01076
(/D)
2
F/D
1 +
p
1
p
2
, (5.31)
where is the gore width at the reﬂector rim, and p
1
and p
2
the membrane tensions
in the radial and circumferential directions, respectively. For isotropic mesh tension,
p
1
= p
2
, the required gore width is
D
= 6.817
F
D
δ
rms,z
D
. (5.32)
A very large number of ribs is needed to obtain a high surface accuracy.
5.4.4 Allowable Surface Error of Reﬂector Antennas
The main design points for deployable reﬂector antennas are deployment feasibility
and high surface accuracy. Only the surface accuracy can be considered at this point.
In this section, an acceptable value of rms surface distortion will be identiﬁed. The
requirement report by DERA [136] does not directly specify the required surface
115
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
accuracy for the present reﬂector but states that an rms accuracy of at least λ/20–
λ/10 is necessary.
The allowable surface accuracy is somewhat dependent on the antenna type; for
symmetrical reﬂectors the peak gain is the most important parameter [138]. From
(5.18) and (5.20) it is evident the there can be a tradeoﬀ in the choice of reﬂector
diameter and rms surface accuracy. A larger, less accurate reﬂector can provide the
same peak gain as a smaller more accurate one. For oﬀset reﬂectors, the choice of re
ﬂector diameter is also constrained by aspects related to the radiation pattern [138].
It is generally not suﬃcient to specify a total rms surface accuracy. Diﬀerent sur
face error distributions with the same total rms surface accuracy can give diﬀerent
radiation patterns. A study of these radiation patterns is, however, well beyond the
scope of this thesis.
Roederer and RahmatSamii [138] observe that rms surface accuracy goals lie typ
ically in the interval λ/100–λ/25. In the extensive and more recent review of the
precision of deployable antennas by Hachkowski and Peterson [52], the rms surface
deviation ranges from
4
λ/353 to λ/6. Both of these studies include values from
antennas at diﬀerent stages of development: in orbit, ﬂight models, engineering
models, and engineering feasibility studies.
The tension truss concept was earlier identiﬁed as the most viable solution for pas
sively achieving a reﬂector surface with high accuracy. A closer look at the practical
surface accuracy of structures based on this concept is therefore of great value in
determining an allowable rms surface accuracy for the present reﬂector.
Miyasaki et al. [107] uses λ/50 (2.4 mm at Sband) as the allowable surface deviation
for the 13 m ETSVIII antenna. Analysis and test of a 3module assembly result in
an rms surface accuracy of 0.4 mm and 0.5 mm, respectively. Using these results,
the rms surface deviation for the entire reﬂector is estimated to less than 1.3 mm.
In [62], where the concept of the AstroMesh reﬂector is introduced, the design of
a 23.56 m diameter oﬀset reﬂector antenna (F/D
p
= 0.286) is described. The
allowable rms surface deviation is taken as λ/50, or 1 mm at Cband. The ﬁrst
contribution to the total surface error comes from the faceting of the reﬂector surface.
The acceptable rms deviation error due to faceting is set to λ/200 or 0.25 mm
for the operating frequency. The required side length of the triangles, (5.29),
is 0.5 m, requiring a tension truss with 27 rings. Hedgepeth et al. [62] also note
that an umbrellatype antenna would require 170 ribs to yield the same accuracy,
(5.32), again showing the advantage of the tension truss concept. Other systematic
contributions to the total surface error are distortions due to
• Gravity during ground testing,
• Centrifugal forces due to satellite spinning,
• Loading of the tension ties, and
4
Values computed from 24 of the 50 listed antenna structures for which both λ and δ
rms,z
are
given.
116
5.4. REFLECTOR ANTENNA THEORY
Table 5.2: Surface distortions for the 6 m AstroMesh reﬂector [166].
Error source
δ
rms,z
(mm) (%)
a
Surface faceting 0.33 32.7
Manufacturing 0.41 50.4
Mesh saddling < 0.10 3.0
Deployment repeatability < 0.08 1.9
Thermal extremes 0.20 12.0
Total rootsumsquare < 0.6 100.0
a
Computed as δ
2
rms,i
/
¸
i
δ
2
rms,i
• Large strains arising due to systematic error in net fabrication, from a large
temperature change, or from pretensioning.
The computed rms error estimates for these sources are: 0.4 mm if the rim is
supported during ground testing, 0.12 mm at a spinning rate of 3.2 rpm, 0.6 mm
with a tension tie force of 4.5 N, and around 0.2 mm for a strain of 254.4 µm/m
in each net element. The rms surface error due to random member length errors in
fabrication and thermal expansion coeﬃcient variation are evaluated by the Monte
Carlo technique and found to be less than 0.5 mm. Random errors due to uncertainty
in the tension tie loading yield an rms error of 0.2 mm. A total rms surface error
estimate is not given for the reﬂector but this extensive study clearly shows the
inﬂuences of diﬀerent error sources.
More recent results on the surface accuracy are available for the 6 m AstroMesh
reﬂector [165, 166]. The rms surface distortion due to gravity during ground testing
is smaller than the resolution of the photogrammetry measuring system, which is
0.07 mm. The rms surface error due to mesh saddling, section 5.4.4, is less than
λ/500, thus having very little inﬂuence on the total error. The total rootsumof
squares (rss) error from all sources is less than 0.6 mm or 1.0 10
−4
D, Table 5.2.
A total surface distortion of 2.5 10
−5
D, is believed to be achievable with existing
materials and manufacturing technology. For a 12 m reﬂector operating at 40 GHz,
that would give an rms surface accuracy equal to λ/20.
5.4.5 Ground Resolution
The gain is not the only factor deciding the antenna diameter. For Earth observ
ing systems the ground resolution is of signiﬁcant importance [18]. The ground
resolution is expressed as the size of an object that can be distinguished from the
background. For a satellite at altitude, h, the ground resolution, χ, at nadir
5
is
χ = 2.44
hλ
D
(5.33)
5
Nadir is the direction from the spacecraft to the centre of the Earth.
117
CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA
where h can be replaced by the slant range to determine the ground resolution
away from nadir. It should be noted, however, that a SAR can provide ground
resolutions similar to visual systems, λ = 0.5 µm, independent of the range and
wavelength by synthesising the required aperture. For the present antenna, D = 3 m,
f = 9.65 GHz, and h = 550 km, the ground resolution would be χ = 13.9 km and
with the ground resolution goal of χ = 1.5 m [136], the SAR antenna will be able
to simulate an aperture with D = 27.8 km.
5.4.6 Accuracy Goals for the Present Antenna
This review of surface accuracies of deployable mesh antennas shows that a high, but
realistic, surface accuracy goal for the present application would be a total rms error
of λ/50 with a budget of λ/100 for the faceting rms surface deviation. However, since
the present reﬂector will use SAR technology, which uses signal processing to achieve
a high resolution, a lower surface accuracy can be accepted [4, 183]. Therefore, a
lower accuracy goal of λ/25, with a budget of λ/50 for the facet error, will also be
considered.
5.5 Selection of Materials
The space environment is extremely harsh with very low vacuum levels, high doses
of thermal and particle radiation, micrometeoroids and debris, magnetic and grav
itational ﬁelds, and temperatures as low as 4 K [144, 183]. The selection of ma
terials is of cardinal importance to the survivability of a spacecraft in this envi
ronment. Spacecraft structures typically contain both metallic and nonmetallic
materials. Most metals are isotropic while nonmetals, i.e. composites, normally are
anisotropic. The selection of materials is based on several factors [183]: strength,
stiﬀness, density, thermal expansion, cost, etc. For precision structures, such as
reﬂector antennas, the relevant material properties are high modulus of elasticity
and low Coeﬃcient of Thermal Expansion (CTE).
5.5.1 Materials for the Antenna Structure
Aluminium alloy is the most commonly used metal for spacecraft structures. Ad
vantages of aluminium are low density, high strengthtoweight ratio, availability,
low cost, and machinability. The main disadvantage of aluminium is its high CTE,
22µ/
◦
C [183]. Other metals, e.g. titanium or magnesium, have lower CTE but pro
vide other disadvantages.
CFRP oﬀer better material properties than metals and can be tailored for high
strength, high stiﬀness, and extremely low CTE. By controlling the direction of the
graphite ﬁbres, very high stiﬀnesstoweight ratios are obtained. Material with a
modulus of 115–124 GPa is readily available [62, 141]. According to Hedgepeth et
118
5.5. SELECTION OF MATERIALS
al. [62], CFRP with a modulus of 227.5 GPa can be produced at extra cost which
for highprecision antenna applications may be justiﬁed. For example, using the less
stiﬀ CFRP for the 23.5 m reﬂector in reference [62] would have resulted in doubled
rms surface errors due to ground testing and pretensioning, and lower vibration
frequencies. Typical tensile strength for highmodulus CFRP is about 1500 MPa,
but higher strengths are available [133]. As CFRP is a very brittle material, it can be
assumed that the proportional limit is above 90% of the ultimate strength. CFRP is
also very light; densities between 1660 and 1740 kg/m
3
are available [62, 96]. These
densities are a little higher than raw CFRP, to allow for an impermeable coating of
the struts [96]. CFRP with a CTE less than 1.0µ/
◦
C can easily be procured.
Although the main antenna structure will be composed of composite material, details
such as end ﬁttings are made of metal. These details are preferably attached by
bonding as bolting would give rise to undesirable stress concentrations. Other more
complex moving details, e.g. hinges, can be made of several diﬀerent materials where
each material aﬀect the overall performance of the detail. For example, the TSR
hinge, section 4.5.1, contains tapes of spring steel, thin steel cables, aluminium alloy
spacers, steel bolts, and a main body of Delrin.
5.5.2 Materials for the RF Reﬂective Surface
First of all, the material used for the RF reﬂective surface must be easy to compactly
fold and require a low density. The most common surface material for space reﬂectors
of moderate precision is a mesh knitted from metallic or synthetic ﬁbres that have
been plated with RF reﬂective material. A mesh provides the compliance necessary
to conform to the doubly curved surface without wrinkling. A high compliance also
means that the mesh can be connected directly to the net structure without any
special interface [61]. Because of its openings, the mesh is limited to frequencies
up to 40 GHz. At higher frequencies, the losses through the mesh are too great.
Solid membrane surfaces have been developed for higher frequency applications, but
without success. The very low inplane compliance of membrane means that it will
not easily conform to a doubly curved surface without developing wrinkles, which
degrade the antenna performance [95]. For the present antenna, it is assumed that
the mesh is knitted from goldplated molybdenum wires, which are readily available.
This mesh has a surface density of 25 g/m
2
[62].
To assure good electrical conductivity between the mesh wires, the tension in the
mesh must be fairly isotropic and uniform. Mesh tensions of 2.0–2.5 N/m were
reported in earlier studies [61,62]. Recently, values of 10–11 N/m have been used [34,
165]. A higher mesh tension will more eﬀectively smooth out the creases formed
in the mesh during folding and, thereby, give a better antenna performance. In
addition, it has been shown, [61, 62], that the strength of a structure designed for
a mesh tension value of 2.5 N/m is suﬃcient to withstand lateral accelerations
many times greater than those experienced in orbit without severe distortion of the
reﬂector surface.
119
Chapter 6
Analysis of Tension Trusses
In the previous chapter it was found that a new reﬂector antenna concept is needed
in order to meet the stowage requirements of a future STRV mission. Ideally, the
required accuracy of the reﬂector surface should be achieved with a passive structure,
i.e. without active control. Miura’s tension truss [104] was identiﬁed as the best
candidate for this task as its surface accuracy can be altered without major changes
to the supporting structure. As the performance of the antenna is a function of
the condition of the reﬂecting surface, it was considered necessary to analyse the
behaviour of tension trusses separately from that of the antenna structure.
A tension truss is basically a geodesic dome with elements that are ﬂexible rather
than stiﬀ. In that way, it can easily be folded. External forces are applied at the
nodes to provide a state of tensile forces in the assembly, Figure 6.1. In practice,
the external forces are provided by springs, called tension ties. The key feature
of the tension truss is that its shape is more or less predetermined by the lengths
and arrangement of its elements. The elastic deformation eﬀects on the shape are
of secondary importance. For this to hold, the assembly must be kinematically
determinate.
Despite the clear advantages of the tension truss for lightweight, highprecision
deployable structures, only a few studies consider its fundamental characteristics,
cf. [100,104]. Other studies are mainly concerned with the implementation of the ten
sion truss in various deployable structures, cf. [99, 106, 165]. Recently, basic research
on the characteristics and applicability of the tension truss with diﬀerent types of
support conditions has been undertaken at the DSL, cf. [34, 79, 80, 169]. These stud
ies provide new insight into the prestressability of the tension truss. In this chapter,
the preliminary study of references [168, 169] is signiﬁcantly expanded. First, a way
of making an nring tension truss kinematically determinate is presented. Then,
studies on the prestressability of axisymmetric and oﬀset net conﬁgurations are
undertaken. Finally, the eﬀects of systematic and random manufacturing errors on
the antenna performance parameters are investigated.
121
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Figure 6.1: Tension truss prestressed by external loads acting approximately normal
to the net surface.
6.1 Static and Kinematic Properties
Consider an axisymmetric tension truss generated according to the scheme in Ap
pendix C. The number of elements b and joints j in the assembly are
b = v
n(1 + 3n)
2
, (6.1)
j = 1 + v
n(1 + n)
2
, (6.2)
respectively. Substituting (6.1) and (6.2) into the extended Maxwell’s rule, (1.2),
yields the following number of kinematic constraints required for static determinacy:
c = 3 + vn. (6.3)
Along the boundary there are 3vn degrees of freedom, which is more than suﬃcient.
However, in both the HALCA antenna, Figure 2.14, and the present one, Figure 5.5,
supports are provided only at the v outermost vertices, Figure 6.1. Inserting c = 3v
into (1.2) yields
m−s = v(n −3) + 3. (6.4)
As a state of selfstress generally cannot exist in a net of synclastic shape, i.e. bowl
shape, s = 0. Thus, the number of internal mechanisms in the tension truss is
m = v(n −3) + 3. (6.5)
122
6.2. AXISYMMETRIC CONFIGURATIONS
Note that for n < 3, several states of selfstress can exist in the synclastic net, thus
disproving the statement s = 0 above. However, none of the selfstress combinations
produced tension in all elements. It is therefore correct to say that no feasible state
of selfstress can exist in a synclastic net.
With a limited number of supports, the only way to remove the internal mechanisms
is to add more bars, and possibly more nodes, to the assembly. Preferably, the
additional nodes and bars should be located along the boundary and connected in a
manner that preserves the vfold symmetry of the original tension truss. Denoting
the number of additional nodes and bars per bay by j
an
and b
an
, respectively, (1.2)
yields for kinematic determinacy, assuming s = 0,
s = v(n −3) + 3 + v(3j
an
−b
an
). (6.6)
Solving (6.6) for b
an
gives
b
an
=
3 + s
v
+ 3(j
an
−1) + n. (6.7)
For v > 3, it is not possible to construct a statically determinate assembly. For the
case v = 6 and s = 3, the minimum number of extra bars is b
an
= n−2. However, a
quick study showed that the resulting conﬁgurations are not ideal. Setting j
an
= 1
gives b
an
= n+1. This alternative is far better considering that there are n+1 nodes
along the boundary of each bay to which the added bars can be connected directly. A
further increase of j
an
produces more complicated conﬁgurations involving crossing
elements. The above analysis provide an indication of the feasibility of the new
conﬁguration. If it can be shown that all of the mechanisms can be removed, the
tension truss is a type III assembly, cf. Table 3.2, which can be accurately analysed
by the linear force method described in section 3.5.
6.2 AxiSymmetric Conﬁgurations
The main advantage of the tension truss is its ability to conform to a multitude
of shapes. Here, it approximates an axisymmetric reﬂector. For a given diameter
D and curvature F/D, the ﬁrst task for a designer is to specify the remaining
parameters in the mesh generation routine. Most important of these is the number
of rings n, which is given by the required surface accuracy. Other parameters are the
sag of the boundaries and the position of the additional nodes for the best internal
force distribution.
6.2.1 SagtoSpan Ratio
The sagtospan ratio of the edge cables is the ﬁrst property to be determined. A
small ratio will give rise to undesirably large forces in the edge elements, while a
large ratio cuts away too much of the reﬂector surface and distorts the boundary
triangles. In tensile roof applications, a sagtospan ratio of 8–12% is common,
123
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
2 1.4 1.4
2.1 2.1
5.2
0.7
1.5
1.5
0.7
5.2
3.8
5.2
1.7
1.6
1.9
1.5
1.5
2 2
4.4
0.1
1.6
1.6
0.1
9.7
8.4
9.7
1.7
1.6
1.8
1.5
1.5
1.9 1.9
3.6
0.5
1.7
1.7
0.5
14.3
13
14.3
1.6
1.6
1.7
1.6
1.6
1.8 1.8
2.8
1.1
1.8
1.8
1.1
18.8
17.7
18.8
1.6
1.6
(a) (b)
(c) (d)
Figure 6.2: Forces in a 3 m diameter threering net with 5% sagtospan ratio (only
one sixth of the complete net is shown). Loads on the inner nodes are
1 N and loads on the edge nodes are: (a) 1 N, (b) 2 N, (c) 3 N and (d)
4 N.
cf. [171]. The analysis of the edge sag will be done with a threering tension truss
(D = 3 m, F/D = 0.4) were the additional nodes and elements have been taken out.
This is done to remove the inﬂuence of the additional nodes on the net forces. The
resulting net is kinematically indeterminate with m = 3. However, it can be shown
that each of the three mechanisms is orthogonal to the particular sets of tension tie
forces used in the analysis, cf. (3.66). The threering tension truss was analysed for
three sagtospan ratios: 5, 10 and 15%. For each ratio, the initial setting of the
tension tie forces was 1 N throughout, which is most practical as identical constant
tension springs would be used in all of the tension ties. If the force pattern in the
net is irregular or, worse, some elements are in compression, the tension tie forces
have to be adjusted. The results for a 5% sagtospan ratio are shown in Figure 6.2.
For the case where the tension tie forces are all equal to 1 N, Figure 6.2(a), some
members are in compression. By increasing the edge forces, the compression forces
gradually become smaller and then tensile, Figures 6.2(b)–(d). An almost uniform
force distribution was obtained for edge forces of 4 N. However, the largest force
along the edge was over 15 N. When the sagtospan ratio was increased to 10%
there was still compression for tension tie forces of 1 N, Figure 6.3(a). As the force
124
6.2. AXISYMMETRIC CONFIGURATIONS
1.9
1.5
1.5
2.1 2.1
4.5
0.2
1.7
1.7
0.2
5
3.5
5
1.7
1.6
1.7
1.6
1.6
1.9 1.9
3.2
0.9
1.8
1.8
0.9
9.2
8
9.2
1.6
1.6
(a) (b)
Figure 6.3: Forces in a 3 m diameter threering net with 10% sagtospan ratio.
Loads on inner nodes: 1 N; loads on edge nodes: (a) 1 N and (b) 2 N.
1.8
1.5
1.5
2 2
4
0.3
1.9
1.9
0.3
4.8
3.2
4.8
1.6
1.6
1.6
1.7
1.7
1.7 1.7
2.5
1.8
2.1
2.1
1.8
8.8
7.6
8.8
1.6
1.6
(a) (b)
Figure 6.4: Forces in a 3 m diameter three ring net with 15% sagtospan ratio.
Loads on inner nodes: 1 N; loads on edge nodes: (a) 1 N and (b) 2 N.
in the edge ties was increased to 2 N an acceptable distribution of net forces was
obtained and the edge forces were smaller than for the 5% ratio, Figure 6.3(b). A
further increase of the sagtospan ratio to 15% yielded no compressed elements
even for the case of uniform 1 N tension tie loads, Figure 6.4(a). When increased to
2 N, the forces in the edge ties gave an almost uniform force pattern, in the range
1.6–2.5 N, and the forces in the edge elements were slightly smaller than for 10%.
Although a sagtospan ratio of 15% produced a better force pattern than the 10%
ratio, the further loss of reﬂecting area was not justiﬁed. Note also that with the
present mesh generation routine there is a limit to the maximum number of rings
n, for a certain value of the sagtospan ratio ρ. According to Appendix C, the
maximum sagtospan ratio is
ρ
max
=
1
2(n −1) tan(π/v)
. (6.8)
This can be rewritten as
n = 1 +
1
2ρ tan(π/v)
. (6.9)
125
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
δ
sag
δ
an
Figure 6.5: The position of the extra node in the radial direction.
The maximum number of rings n
max
for a given sagtospan ratio is
n
max
=
¸
1 +
1
2ρ tan(π/v)
¸
, (6.10)
where the ﬂoor function x yields the largest integer ≤ x. For v = 6, ρ = 0.15
yields n
max
= 6 and ρ = 0.10 yields n
max
= 9. Hence, a sagtospan ratio of 10% was
selected for the tension truss in the analyses that follow. The outer diameter D of
the tension truss was ﬁxed to 3 m, so with ρ = 0.10, the eﬀective aperture diameter
was 2.32 m using (C.7).
6.2.2 Position of Additional Nodes
Another beneﬁt of a larger sagtospan ratio, is that it leaves more room for the
additional elements along the boundary. Vertically, the additional nodes lay on
the paraboloid. Tangentially, they were positioned in the middle of the supports.
Radially the position varied; theoretically, any value between 0 and δ
sag
from the
edge cable was possible, Figure 6.5. To determine the best position of the additional
nodes, threering trusses with diﬀerent focal lengths were analysed. The tie forces
were equal to 1 N on interior nodes and 2 N on the edge and additional nodes. The
ratios between the maximum and minimum forces in the interior and all elements,
respectively, were used to identify the best conﬁguration, Figure 6.6. The aim was
an interior force ratio close to unity and a fairly low ratio for all forces. It appears
that for all the three F/D ratios, the best conﬁgurations were obtained for δ
an
/δ
sag
between 0.4 and 0.5.
6.2.3 Tension Tie Force Distribution
It was shown for the threering tension truss that the tension ties on the edges had to
provide twice the force of the interior ties to give a desirable internal force pattern.
As the number of rings increases, it is likely that more than two diﬀerent tie forces
are needed to obtain an adequate prestress. For practical purposes, however, it is
necessary to restrict the number of diﬀerent ties and their diﬀerences in force. If
possible, only four diﬀerent types of ties should be used within a structure and the
minimal diﬀerence in force was set to 0.5 N. Also, the position of the additional
nodes should not be too close to the edge cable in order to avoid problems that
126
6.2. AXISYMMETRIC CONFIGURATIONS
may arise when connecting many members to a single node; a minimum value of 0.4
of the total sag seemed appropriate. Hence, the problem of prestressing an nring
tension truss has been reduced to ﬁnding (i) a tension tie force combination on the
edge and additional nodes and (ii) a position for the additional nodes. Throughout
this analysis, D = 3 m and the sagtospan ratio was ﬁxed to 10%. The tie force
combinations on the edge and additional nodes were, for each number of rings, found
by a trialanderror procedure. After a few tests on a tension truss with four rings
it became clear which edge ties aﬀected certain elements. The results from the trial
analysis are shown in Figures 6.7–6.10. The corresponding element forces and their
ratios for three diﬀerent positions of the additional nodes are shown in Table 6.1. In
all cases, the best ratios were obtained for δ
an
/δ
sag
= 0.4. The maximum number of
rings in the analysed tension trusses was seven. The prestressing problems became
signiﬁcantly severe for more than seven rings, so this was taken as a limit. Another
issue that will limit the number of rings is that the struts of the ring structure in
the antenna concept, Figure 5.5, pass through the tension tie forest.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0
1
2
3
4
5
6
7
8
F/D = 0.4
F/D = 0.6
F/D = 0.8
Interior elements
All elements
t
max
t
min
δ
an
/δ
sag
Figure 6.6: Ratios of the maximum and minimum forces in the tension truss. Two
cases: interior elements and all elements.
127
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Table 6.1: Tension truss element forces and their ratios (axisymmetric conﬁgura
tion). The best ratios are in bold face.
n F F
D
p
H
a
δ
an
δ
sag
t
int
min
t
int
max
t
int
max
t
int
min
t
all
min
t
all
max
t
all
max
t
all
min
(m) (m) (N) (N) (N) (N)
0.4 1.59 2.31 1.45 1.59 9.62 6.05
1.2 0.4 0.469 0.5 1.45 2.45 1.59 1.55 9.38 6.07
0.6 1.42 2.60 1.83 1.42 9.26 6.51
0.4 2.39 3.25 1.36 2.39 14.04 5.88
3 1.8 0.6 0.313 0.5 2.28 3.46 1.51 2.28 13.70 5.99
0.6 2.10 3.66 1.74 2.10 13.52 6.43
0.4 3.16 4.23 1.34 3.16 18.53 5.86
2.4 0.8 0.234 0.5 3.03 4.50 1.49 3.03 18.08 5.97
0.6 2.79 4.77 1.71 2.79 17.85 6.41
0.4 1.35 4.24 3.14 1.30 15.81 12.18
1.2 0.4 0.469 0.5 1.30 4.33 3.34 0.92 15.83 17.25
0.6 1.24 4.43 3.58 0.56 15.86 28.18
0.4 2.00 5.94 2.97 1.84 22.93 12.49
4 1.8 0.6 0.313 0.5 1.92 6.07 3.16 1.30 22.96 17.69
0.6 1.83 6.21 3.39 0.80 23.01 28.91
0.4 2.65 7.72 2.91 2.40 30.20 12.61
2.4 0.8 0.234 0.5 2.55 7.88 3.09 1.69 30.24 17.86
0.6 2.43 8.06 3.32 1.04 30.30 29.21
0.4 1.92 5.26 2.74 1.92 27.20 14.17
1.2 0.4 0.469 0.5 1.91 5.32 2.78 1.57 27.21 17.33
0.6 1.70 5.40 3.17 0.78 27.23 34.92
0.4 2.84 7.34 2.58 2.84 39.30 13.82
5 1.8 0.6 0.313 0.5 2.83 7.43 2.63 2.34 39.30 16.80
0.6 2.51 7.53 3.00 1.16 39.34 33.97
0.4 3.77 9.52 2.52 3.77 51.67 13.70
2.4 0.8 0.234 0.5 3.75 9.63 2.57 3.11 51.67 16.60
0.6 3.33 9.77 2.93 1.54 51.72 33.62
0.4 2.69 5.79 2.16 1.48 43.21 29.11
1.2 0.4 0.469 0.5 2.64 5.89 2.23 0.75 43.28 57.69
0.6 2.59 6.00 2.32 0.10 43.36 417.97
0.4 3.98 8.07 2.03 2.10 62.24 29.63
6 1.8 0.6 0.313 0.5 3.92 8.21 2.10 1.06 62.34 58.75
0.6 3.83 8.36 2.18 0.15 62.46 425.90
0.4 5.28 10.46 1.98 2.74 81.75 29.84
2.4 0.8 0.234 0.5 5.20 10.64 2.05 1.38 81.88 59.18
0.6 5.09 10.83 2.13 0.19 82.04 429.13
0.4 2.49 7.76 3.12 1.45 57.00 39.24
1.2 0.4 0.469 0.5 2.44 7.86 3.22 1.38 57.09 41.40
0.6 2.39 7.97 3.34 1.00 57.19 57.37
0.4 3.68 10.80 2.93 2.17 81.94 37.78
7 1.8 0.6 0.313 0.5 3.62 10.93 3.02 2.06 82.07 39.82
0.6 3.54 11.08 3.13 1.47 82.22 56.07
0.4 4.89 13.98 2.86 2.89 107.55 37.24
2.4 0.8 0.234 0.5 4.80 14.16 2.95 2.74 107.71 39.25
0.6 4.70 14.35 3.06 1.94 107.91 55.58
128
6.2. AXISYMMETRIC CONFIGURATIONS
1 1 1
1 1
1.1
1 1 1 1
1.1 1.2 1.1
2
0.6 1.4 1.1 1.1 1.4
0.6
7.4
5.2 5.2
7.4
1
1
3.2
1.2
0.6
1.2
3.2
2.5
2
2.5
2.5
(b)
(a)
Figure 6.7: Normalised prestress distribution (a) under tension tie forces (b). Mul
tiplication factor for element forces was 2.1. D = 3 m, F/D = 0.4,
ρ = 0.10 and δ
an
/δ
sag
= 0.4.
129
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
1 1 1
1 1
1.1 1 1 1 1
1 1.1 1
1.3 0.9 1.1 0.9 0.9 1.1 0.9
1 1.3 1.3 1
2.3
0.7 1.7
0.8 1.2 1.2
0.8 1.7
0.7
10.2
7.2
5.8
7.2
10.2
1
1
3.1
1.5
1 1
1.5
3.1
(a)
(b)
3.5 3.5 3.5
2.5 2.5
Figure 6.8: Normalised prestress distribution (a) under tension tie forces (b). Mul
tiplication factor for element forces was 2.7. D = 3 m, F/D = 0.4,
ρ = 0.10 and δ
an
/δ
sag
= 0.4.
130
6.2. AXISYMMETRIC CONFIGURATIONS
1 1 1
1 1
1 1 1 1 1
1 1 1
1.1
1 1.1 1 1
1.1 1
1 1 1 1
1.2 1 1.1 1 1 1 1 1.1 1
1.1 1.1 1.1 1.1 1.1
1.8
0.8
1.6
1.1 1.6
1.4 1.4
1.6 1.1
1.6
0.8
13.5
11.5
9.4 9.4
11.5
13.5
1
1
3.4
0.6
1.3
0.5
1.3
0.6
3.7
(a)
(b)
4.5 4.5
4 4
3.5
4.5
Figure 6.9: Normalised prestress distribution (a) under tension tie forces (b). Mul
tiplication factor for element forces was 3.2. D = 3 m, F/D = 0.4,
ρ = 0.10 and δ
an
/δ
sag
= 0.4.
131
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
1 1 1
1 1
1 1 1 1 1
1 1 1
1.1 1 1 1 1 1 1
1 1 1 1
1.2 0.9 1.1 1 1 1 1 1.1 0.9
1 1.1 1.1 1.1 1
1.4 0.9 1.3 0.9 1.1 0.9 0.9 1.1 0.9 1.3 0.9
1 1.2 1.1 1.1 1.2 1
2.1
0.7 1.8
1 1.7
1.5
1.8 1.8
1.5
1.7
1 1.8
0.7
15.3
13.1
11.2
10.1
11.2
13.1
15.3
1
1
3.6
0.4
0.7
0.8 0.8
0.7
0.4
3.6
(a)
(b)
5 5
5
4.5 4.5
4 4
Figure 6.10: Normalised prestress distribution (a) under tension tie forces (b). Mul
tiplication factor for element forces was 3.7. D = 3 m, F/D = 0.4,
ρ = 0.10 and δ
an
/δ
sag
= 0.4.
132
6.3. OFFSET CONFIGURATIONS
6.3 Oﬀset Conﬁgurations
The oﬀset version of the tension truss is created by simply mapping the positions
of the joints of the circular aperture onto an elliptic one. Thus, in the present
case where the minor axis of the ellipse is equal to D, only the xcoordinate needs
changing, x := ξx, where ξ is the ellipticity, (5.10). The vertical joint positions are
recalculated using (5.8).
6.3.1 Focal Length and Oﬀset Distance
For the oﬀset antennas in Table 5.1, F/D
p
varies between 0.167 and 0.363 and
X
A
/D
a
varies between 0 and 0.17. From a blockage viewpoint, a larger value of X
A
is better. Other issues, e.g. the size of the spacecraft bus and the antenna support
structure, may prescribe both the focal length and oﬀset distance. To determine
the sensitivity of the force distribution in the tension truss to the ellipticity, ξ, three
diﬀerent realistic values of X
A
/D
a
were used: 0, 0.1 and 0.2. Three diﬀerent focal
length were also used: 1.5, 1.8 and 2.1 m. A preliminary analysis of a threering
oﬀset tension truss showed that a focal length less than 1.5 m was more diﬃcult
to prestress. Initially, the tension tie forces applied to the edge and additional
nodes of the oﬀset tension trusses were identical to those of the axisymmetric cases,
Figures 6.7–6.10. If, however, some of the elements ended up in compression, due
to the change in geometry, the tension tie forces were changed.
(a) (b)
Figure 6.11: Two symmetric oﬀset conﬁgurations for the tension truss: (a) conﬁgu
ration 1 and (b) conﬁguration 2.
133
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
6.3.2 Two Symmetric Oﬀset Conﬁgurations
There are only two ways of arranging the tension truss in the oﬀset conﬁguration
which preserve a certain symmetry, Figure 6.11. The ﬁrst alternative, Figure 6.11(a),
is to let two opposite support nodes lie along the major axis of the elliptic aperture.
The second alternative, Figure 6.11(b), diﬀers from the ﬁrst by a 90
◦
rotation so
that the opposite nodes now lie along the minor axis. A preliminary analysis with
a threering net showed that, in terms of prestressability, no conﬁguration could be
said to be better than the other. From a more practical reasoning, conﬁguration 1
may be easier to attach to the spacecraft with nodes along the major ellipse axis,
but this is only an assumption. Since there was hardly any diﬀerence in the force
distribution between the two conﬁgurations, only conﬁguration 1 will be used in the
following analyses, based on the assumed simpler attachment. The results for the
oﬀset tension trusses are given in Table 6.2. The tension tie forces on the edge and
additional nodes had to be changed slightly to give a satisfactory force pattern in
the truss. The new values for three to seven rings are given in Figure 6.12. The
value X
A
= 0 gives the lowest ratio for the internal forces, but not necessarily the
lowest for all forces.
2
2.5
2
2.5
2
2
2.5
2
2
A B
(a)
2.5
3
2
2.5
3
2.5
2
2.5
3
2.5
2
2.5
(b)
4.5
3
2.5
3.5
3.5
4.5
3 2.5
3.5
3.5
4.5
3
2.5
3.5
3.5
(c)
6
4.5
3.5
4
4.5
6
6
4.5 3.5 4
4.5
6
6
6
4.5
3.5
4
4.5
(d)
7.5
6
4.5
4
4.5
5
6
7.5 6
4.5 4 4.5
5
6
7.5
6
4.5
4
4.5
5
6
(e)
Figure 6.12: Tension tie force distribution along the edges and on the extra nodes
to obtain an adequate internal force distribution in the tension truss
of conﬁguration 1: (a) three, (b) four, (c) ﬁve, (d) six, and (e) seven
rings. Interior elements are drawn thinner for clarity.
134
6.3. OFFSET CONFIGURATIONS
Table 6.2: Tension truss element forces and their ratios (oﬀset conﬁguration 1). The
best ratios are in bold face.
n F X
A
ξ F
D
p
H
a
t
int
min
t
int
max
t
int
max
t
int
min
t
all
min
t
all
max
t
all
max
t
all
min
(m) (m) (m) (N) (N) (N) (N)
0 1.118 0.250 0.335 1.62 4.05 2.50 1.62 18.44 11.38
1.5 0.3 1.166 0.227 0.322 1.71 4.40 2.58 1.71 19.50 11.41
0.6 1.221 0.208 0.307 1.64 4.79 2.93 1.64 20.66 12.61
0 1.083 0.300 0.288 2.00 4.45 2.22 2.00 20.69 10.33
3 1.8 0.3 1.118 0.273 0.280 2.05 4.75 2.32 2.05 21.63 10.56
0.6 1.158 0.250 0.270 2.00 5.08 2.53 2.00 22.64 11.30
0 1.062 0.350 0.252 2.39 4.90 2.05 2.39 23.10 9.66
2.1 0.3 1.088 0.318 0.246 2.38 5.16 2.16 2.38 23.93 10.03
0.6 1.118 0.292 0.240 2.36 5.45 2.31 2.36 24.83 10.54
0 1.118 0.250 0.335 1.20 6.37 5.29 1.20 28.34 23.53
1.5 0.3 1.166 0.227 0.322 1.05 6.91 6.56 1.05 31.27 29.71
0.6 1.221 0.208 0.307 0.95 7.52 7.93 0.95 34.57 36.44
0 1.083 0.300 0.288 1.93 6.86 3.56 1.55 31.60 20.33
4 1.8 0.3 1.118 0.273 0.280 1.78 7.31 4.10 1.60 32.96 20.61
0.6 1.158 0.250 0.270 1.67 7.82 4.67 1.65 35.39 21.43
0 1.062 0.350 0.252 2.60 7.47 2.88 1.76 35.33 20.03
2.1 0.3 1.088 0.318 0.246 2.46 7.85 3.19 1.81 36.54 20.21
0.6 1.118 0.292 0.240 2.35 8.28 3.52 1.86 37.87 20.38
0 1.118 0.250 0.335 2.20 7.88 3.57 2.20 50.40 22.87
1.5 0.3 1.166 0.227 0.322 2.27 8.49 3.74 2.27 54.17 23.88
0.6 1.221 0.208 0.307 2.20 9.17 4.16 2.20 59.91 27.18
0 1.083 0.300 0.288 2.88 8.67 3.01 2.63 56.81 21.59
5 1.8 0.3 1.118 0.273 0.280 2.93 9.18 3.14 2.90 59.18 20.41
0.6 1.158 0.250 0.270 2.92 9.75 3.34 2.92 61.79 21.18
0 1.062 0.350 0.252 3.54 9.59 2.71 2.60 63.61 24.48
2.1 0.3 1.088 0.318 0.246 3.55 10.02 2.82 2.82 65.70 23.30
0.6 1.118 0.292 0.240 3.53 10.51 2.98 3.01 68.02 22.59
0 1.118 0.250 0.335 2.60 9.88 3.80 1.69 78.83 46.63
1.5 0.3 1.166 0.227 0.322 2.64 10.55 4.00 1.81 86.46 47.83
0.6 1.221 0.208 0.307 2.13 11.27 5.30 1.95 95.63 48.92
0 1.083 0.300 0.288 3.34 11.08 3.31 2.01 88.95 44.35
6 1.8 0.3 1.118 0.273 0.280 3.40 11.63 3.42 2.10 92.57 44.11
0.6 1.158 0.250 0.270 3.22 12.25 3.80 2.21 97.20 43.87
0 1.062 0.350 0.252 4.07 12.38 3.04 2.31 99.67 43.14
2.1 0.3 1.088 0.318 0.246 4.11 12.85 3.12 2.39 102.86 43.06
0.6 1.118 0.292 0.240 4.10 13.39 3.27 2.49 106.41 42.78
0 1.118 0.250 0.335 1.99 14.12 7.10 1.99 114.40 57.52
1.5 0.3 1.166 0.227 0.322 2.06 15.02 7.30 2.06 120.33 58.46
0.6 1.221 0.208 0.307 1.90 16.00 8.41 1.64 126.84 77.52
0 1.083 0.300 0.288 2.63 15.86 6.03 2.21 129.23 58.51
7 1.8 0.3 1.118 0.273 0.280 2.68 16.61 6.20 2.62 134.38 51.36
0.6 1.158 0.250 0.270 2.74 17.44 6.38 2.74 140.08 51.21
0 1.062 0.350 0.252 3.26 17.75 5.44 2.03 144.92 71.32
2.1 0.3 1.088 0.318 0.246 3.29 18.39 5.59 2.38 149.45 62.75
0.6 1.118 0.292 0.240 3.37 19.11 5.67 2.74 154.50 56.45
135
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
6.4 Eﬀects of Manufacturing Errors on the Re
ﬂector Accuracy
As described in section 5.4, the performance of a reﬂector antenna is mainly de
termined by the manufactured surface accuracy of the reﬂecting surface. For the
tension truss, the systematic deviation due to surface faceting, (5.29), can be said
to represent the minimum theoretical error. Unavoidably, the minimum achievable
surface error is higher due to errors in the manufacturing process. These errors are
randomly distributed in the structure and their inﬂuence on the surface accuracy
can, in most cases, not be determined analytically. In such cases the alternative
approach is the Monte Carlo technique, cf. [97].
Despite the signiﬁcant importance of manufacturing errors, only a few studies are
available in the literature. Hedgepeth [59] develops an equivalence between the sta
tistical errors and the natural frequencies of a reﬂector antenna structure assembled
by tetrahedral truss elements. Hedgepeth’s analysis results in formulae which can be
used for the preliminary design of such antennas. An interesting case of his analysis
is when the height of the truss structure approaches zero, thus becoming a geodesic
dome. The rms surface error of a geodesic dome, with an errorfree rim, is [59]:
δ
rms,z
= 2Fσ
ε
, (6.11)
where σ
ε
is the standard deviation of the member length imperfections. It should
be noted that (6.11) is based on a continuum approach and valid only for a large
number of facets. Greene [49] continues the analysis of the tetrahedral truss structure
using the Monte Carlo technique and a linear Finite Element (FE) model to obtain
results for a reﬂector with a low number of facets. Greene ﬁnds that the surface
deviation increases as the number of facet rings decreases. For two and four rings
it is about three and two times higher than that predicted by (6.11), respectively.
In addition to the surface accuracy, Greene also analyses the reﬂector defocus and
variation of the forces in the tetrahedral truss. It is found that the defocus, like the
surface accuracy, increases with a decreasing number of facet rings. More recently,
Hedgepeth et al. [62] uses the Monte Carlo technique for a nine ring geodesic dome
(F = 15.7 m) ﬁxed at its rim. The rms surface error is computed for two error
sources: member length imperfection and tension tie force variation. The average
rms surface error of 100 simulations is 0.394 mm for the former case. Inserting
F = 15.7 m and σ
ε
= 10
−5
in (6.11) gives δ
rms,z
= 0.314 mm, which agrees well with
their simulated value. The average rms surface deviation due to a tension tie force
variation of 0.1 N is 0.160 mm. These three studies provide important information
on the eﬀects of manufacturing errors on diﬀerent reﬂector parameters.
The element length imperfection and tension tie load variation are not the only
manufacturing errors. Random errors related to the present antenna are:
• Member length imperfections,
• Ring structure distortion,
136
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
• Tension tie load variation, and
• Random thermal strains.
First, it is important to ﬁnd how each of these errors alone aﬀects the surface
accuracy and defocus in order to establish the manufacturing tolerance level. Then,
the combined eﬀect of all the random errors has to be determined to conclude on the
achievable surface accuracy of the reﬂector. Other sources that aﬀect the accuracy
of the reﬂector are gravity during ground testing, attitude accelerations, etc.
6.4.1 Monte Carlo Technique for the Tension Truss
The technique used to calculate the eﬀects of manufacturing errors makes use both
of the force method and of the FEM. First, the conﬁguration of the errorfree tension
truss was generated by the routine in Appendix C. Then, the element forces under
the prescribed tension tie loading were computed by the force method. These forces
were used to compute the unstrained length
0
of element i as:
0,i
=
i
1 + t
i
/AE
i
, (6.12)
where t is the force in each element and AE the axial stiﬀness. To account for length
errors in the assembly of the tension truss, the unstrained length of element i was
modiﬁed as
˜
0,i
=
0,i
(1 + ε), (6.13)
where
ε ∼ N(0, σ
ε
), (6.14)
i.e. normally distributed with a zero mean value and standard deviation σ
ε
. The
normally distributed random strain ε was generated by the Matlab [89] routine
randn, which will theoretically produce over 2
1492
≈ 10
449
values before repeating
itself. Then, the tension truss with the imperfect element lengths and unchanged
tension tie loads was analysed by a ﬁnite element program using nocompression
catenary cable elements, cf. [171]. These catenary elements require a small self
weight to avoid numerical problems. However, its inﬂuence was neglected as it was
chosen to be much smaller than the tension tie loads. In the FE analysis it was
assumed that all elements have the same properties, although the edge elements
most likely have to be stiﬀer to sustain the higher forces.
6.4.2 BestFit Paraboloid Analysis
In section 5.4, the bestﬁt paraboloid, which minimises the rms surface deviation,
was introduced. The computation of the bestﬁt paraboloid to a set of points on
the reﬂector surface can be separated into two parts [49, 140]:
137
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
1. a rigid body ﬁt, where a ﬁtted paraboloid of ﬁxed focal length undergoes
translation along the Zaxis and rotates about the X and Y axis, followed by
2. a readjustment ∆F of the equivalent focal length.
Some results of bestﬁt paraboloid computations are given in [62] and [140]. They
show that the rotations about the X and Y axes are very small; one is about 510
−4
and the other about one order lower. Therefore, the rotations about the X and
Y axes were assumed to be negligible in the present analysis. As a result, the Zaxis
of the bestﬁt paraboloid was always parallel to that of the ideal paraboloid. These
simpliﬁcations may also be warranted by the use of (5.24), instead of (5.21), when
calculating the rms surface error. Lai [78] uses a similar approach to compute the
surface error for the CRTS reﬂector.
For a reﬂector surface represented by a discrete set of points, the equation of the
bestﬁt paraboloid is
Z =
X
2
+ Y
2
4F
+ ∆Z
P
, (6.15)
where F denotes the bestﬁt focal length and ∆Z
P
the translation of the apex of the
ideal paraboloid in the XY Z system. Let (X
1
, Y
1
, Z
1
), (X
2
, Y
2
, Z
2
), ..., (X
n
, Y
n
, Z
n
)
be the coordinates of n points on the reﬂector surface. If the n points are not equally
distributed over the aperture, the contribution of each point in the computation of
the bestﬁt paraboloid will vary. This is taken care of by weighting each point by
its associated area. Thus, dense points are given lower weights than sparse points.
A more reﬁned weighting can be done by also taking into account the radiation
pattern over the aperture [178]. However, this will not be included here considering
the simpliﬁcations already made. Substituting each of the n points into (6.15) and
multiplying with the horizontal projection of the surface area associated to each
point lead to the following overdetermined system of linear equations:
A
1
(X
2
1
+ Y
2
1
) A
1
A
2
(X
2
2
+ Y
2
2
) A
2
.
.
.
.
.
.
A
n
(X
2
n
+ Y
2
n
) A
n
¸
¸
¸
¸
¸
1/4F
∆Z
P
=
¸
¸
¸
¸
A
1
Z
1
A
2
Z
2
.
.
.
A
n
Z
n
. (6.16)
Equation (6.16) is solved by a standard least squares solution technique, cf. [156].
In the present bestﬁt analysis the sides of each triangular facet were divided into
N parts of equal length ∆ = /N, thereby creating N
2
subtriangles. The centre of
gravity of each subtriangle was taken as a point in the bestﬁt paraboloid compu
tation and the horizontal projection of the subtriangle area was taken as its weight
factor. Subsequently, the axial rms surface deviation is computed as:
δ
rms,z
=
¸
1
¸
n
i=1
A
i
n
¸
i=1
A
i
˜
Z
i
−Z
i
2
¸
1/2
. (6.17)
Note that for oﬀset reﬂectors the points should be measured in the coordinate system
of the parent paraboloid, XY Z. Another important parameter for the performance
of the reﬂector is the defocus, which is simply the total translation of the focal
point, ∆F + ∆Z
P
[49].
138
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
6.4.3 Extracting the Random Surface Deviations
The routine for computing the bestﬁt paraboloid, rms surface deviation and de
focus does not distinguish systematic errors from random errors, but computes the
total rms surface error. However, the systematic error due to faceting will always
exist so the contribution from the random rms surface error must be extracted from
the total rms surface error. A technique to extract a single or several random errors
from the total error will now be introduced. First, assume that the total error is
caused by one systematic and one random error source. In the following, z
sy
denotes
the zcoordinate of the reﬂector surface with only the systematic error present, z
bf
the zcoordinate of the bestﬁt paraboloid (to both errors), and
z
the additional
contribution of the random error in the zdirection. Equation (5.24) can be expanded
as follows:
δ
2
rms,z
=
1
A
a
A
a
(∆z)
2
dA
=
1
A
a
A
a
((z
sy
+
z
) −z
bf
)
2
dA
=
1
A
a
A
a
(z
sy
−z
bf
)
2
dA
. .. .
I
+
1
A
a
A
a
2
z
+ 2
z
(z
sy
−z
bf
)dA
. .. .
II
.
(6.18)
Integral I is approximately equal to the squared systematic rms surface deviation
due to the facet approximation. It is not exactly equal since the calculation of
the bestﬁt paraboloid also includes
z
. Integral II is the square of the additional
surface deviation due to manufacturing tolerances. For
z
→ 0, the rms surface
deviation approaches that of the ideal faceted reﬂector antenna. With some further
simpliﬁcations it is also possible to approximately ﬁnd the separate eﬀects of diﬀerent
random deviations. For the case of two random deviations,
z
=
z,1
+
z,2
, (5.24) is
written as:
δ
2
rms,z
=
1
A
a
A
a
(z
sy
−z
bf
)
2
+
1
A
a
A
a
2
z,1
(z
sy
−z
bf
)dA
+
1
A
a
A
a
2
z,2
(z
sy
−z
bf
)dA +
1
A
a
A
a
(
z,1
+
z,2
)
2
dA.
(6.19)
Terms of type
z,i
z,j
are very small and can be neglected, which means that the
fourth integral vanishes. Thus, the second and third integrals are the inﬂuences of
z,1
and
z,2
, respectively. If the individual random error
z,i
is small this technique
can be extended to an arbitrary number of random errors. Thomson [166] uses
a similar technique to compute the total rms surface deviation, or the rss surface
deviation as he calls it, from several error sources. Thomson’s results are reproduced
in Table 5.2.
139
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
6.4.4 Systematic Facet Surface Deviation
Before the inﬂuence of the random errors can be computed the systematic rms
surface deviation and defocus have to be determined. Although it was possible
to get a good estimate of the rms surface error by (5.29) it was necessary for the
following analyses to compute the error by the bestﬁt computation described above.
In this, and the following analyses of the surface deviation, the parameters of the
antenna were: D = 3 m, F/D = 0.4, 0.6 and 0.8, and δ
an
/δ
sag
= 0.6. The analysis of
the systematic rms error were done for three to nine rings. The nodal positions of the
tension truss were the ones generated by the mesh generation routine in Appendix C.
As the tension truss is sixfold symmetric, only one sixth needed to be analysed.
The systematic rms surface error was computed for 5
2
, 10
2
, 20
2
, 25
2
and 40
2
sub
triangles in each triangular facet. The results from all except the 25
2
case were used
in a Richardson extrapolation scheme, cf. [29], to determine the asymptotic values of
the rms surface error. They were then compared to those by Agrawal et al., (5.29).
The 25
2
subdivision was chosen to be used in the subsequent computations of the
random error as it produced values close to the asymptotic values and required less
computation time when the entire tension truss has to be analysed. The results of
the systematic error computations are shown in Table 6.3, where superscripts
Ag
and
25
indicate the values by (5.29) and a 25
2
subdivision, respectively. Values
without superscript are the Richardsonextrapolated ones. Throughout the range of
rings, both of the computed rms surface errors agree well with the one by (5.29). The
defocus decreased with increasing number of rings as it should, thus the reliability
of the bestﬁt routine was conﬁrmed.
6.4.5 Inﬂuence of Tension Tie Loading
While the computation of the systematic surface errors was done for three to nine
rings, the analyses concerned with the random errors were only done for three rings.
Since this is a feasibility study, although a quite extensive one, an estimation of
the required manufacturing tolerances can be done by analysing a representative
antenna conﬁguration. For example, Hedgepeth et al. [62] perform Monte Carlo
simulations on a nine ring geodesic dome with ﬂexible members in order to get
accuracy estimates for the real one with 27 rings. It was further assumed that the
dependency of the rms surface error and defocus on the number of rings was similar
to that of the tetrahedral truss structure studied in [49, 59]. Thus, a smaller number
of rings would produce a larger rms surface error and more severe defocus. Note,
however, that the validity of this assumption was not put to test in this thesis.
The computations of the systematic errors in the previous section were based on the
generated nodal positions. When the equilibrium conﬁguration of the tension truss
was computed using the FEM, the nodal positions diﬀered due to load deformations.
For signiﬁcant diﬀerences, the previously computed systematic error cannot be used
when extracting the random errors from the total error. To determine if the elastic
deformation was large enough to aﬀect the systematic error, the threering tension
truss was analysed for two diﬀerent interior tension tie loads: 1 and 10 N. The results
140
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
Table 6.3: Surface deviation (rms) and defocus of a paraboloid due to triangular
faceting.
n F/D δ
Ag
rms,z
δ
rms,z
δ
25
rms,z
∆F + ∆Z
P
∆F
25
+ ∆Z
25
P
(mm) (mm) (mm) (mm) (mm)
0.4 3.3620 3.3496 3.3417 22.4630 22.4728
3 0.6 2.2413 2.2331 2.2278 22.6142 22.6233
0.8 1.6810 1.6748 1.6709 24.9814 24.9909
0.4 1.8911 1.9305 1.9261 13.6208 13.6272
4 0.6 1.2607 1.2870 1.2841 14.0371 14.0435
0.8 0.9456 0.9652 0.9630 15.7322 15.7392
0.4 1.2103 1.2629 1.2601 8.4811 8.4852
5 0.6 0.8069 0.8419 0.8401 8.6046 8.6086
0.8 0.6052 0.6315 0.6301 9.5516 9.5559
0.4 0.8405 0.8897 0.8877 5.4803 5.4830
6 0.6 0.5603 0.5931 0.5918 5.3715 5.3740
0.8 0.4202 0.4448 0.4439 5.8324 5.8349
0.4 0.6175 0.6569 0.6555 3.6903 3.6921
7 0.6 0.4117 0.4379 0.4370 3.4574 3.4589
0.8 0.3088 0.3285 0.3277 3.6399 3.6416
0.4 0.4728 0.5011 0.5000 2.5954 2.5967
8 0.6 0.3152 0.3341 0.3333 2.3142 2.3153
0.8 0.2364 0.2506 0.2500 2.3489 2.3499
0.4 0.3736 0.3919 0.3910 1.9075 1.9085
9 0.6 0.2490 0.2613 0.2607 1.6223 1.6231
0.8 0.1868 0.1959 0.1955 1.5849 1.5856
are shown in Table 6.4. The higher tension tie load gave slightly lower values for the
defocus. This was due to the increased vertical deformation of the net. However,
the rms surface deviation was unaﬀected by the higher tie load. In the studies which
follow, an interior tie load of 5 N was used to further reduce the eﬀects of the small
selfweight of the catenary elements. In a zerogravity environment, a smaller tie
load can be used.
Table 6.4: Inﬂuence of deformation due to tension tie loading (threering tension
truss).
F/D 0.4 0.6 0.8
t
int
tie
(N) 1 10 1 10 1 10
δ
rms,z
(mm) 3.3417 3.3417 2.2279 2.2279 1.6710 1.6710
∆F + ∆Z
P
(mm) 22.4801 22.4695 22.6253 22.6115 24.9765 24.9575
141
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
6.4.6 Statistical Considerations
The statistical treatment of the results from the Monte Carlo trials was more or less
straightforward. The mean value of a parameter p is denoted ¯ p and its standard
deviation σ
p
. Here, the parameters are the rms surface error and defocus. From
the trials, the maximum and minimum values of p were also extracted. However,
since the number of trials was quite small the maximum computed value p
max
could
not be taken as the highest value likely to occur. To ﬁnd a representative maximum
value of the parameter, a probability limit had to be set. A maximum value often
used is ´ p = ¯ p+3σ
p
, which corresponds to a probability of 0.00135 that any simulated
value will be larger. This value was chosen to represent the worst possible value of
the studied parameter. The quality of the simulated values was estimated by the
standard deviation of all the values generated with randn, which should be close to
one. The standard deviation of S simulations was approximatively computed as
σ
randn
=
1
S
S
¸
i=1
σ
2
randn,i
1/2
, (6.20)
where σ
randn,i
is the standard deviation for simulation number i.
Another important statistical aspect is the required number of Monte Carlo simu
lations. Hedgepeth et al. [62] use 100 simulations for their two cases. Greene [49]
also uses 100 simulations for the rms surface error but ﬁnds it necessary to increase
the number of trials to 200 to determine the defocus with reasonable accuracy. To
determine how the surface error and defocus vary with the number of Monte Carlo
simulations, a tension truss with member length imperfections was chosen as a test
example. The rms surface error and defocus, and their corresponding standard devi
ations, were computed for 100 to 500 simulations, Table 6.5. Overall, the diﬀerences
were very small; only the standard deviation of the surface error had diﬀerences of
more than 1%. Hence, it was concluded that 100 simulations was suﬃcient for the
Monte Carlo analysis of the tension truss.
6.4.7 Inﬂuence of Member Length Imperfections
The ﬁrst of the random error sources is the individual lengths of the elements. The
fabrication tolerance that can be achieved depends highly on the manufacturing
costs that can be accepted. Hedgepeth [58] ﬁnds that “a value of σ
ε
of 10
−3
is
representative of ordinary careful practice, 10
−4
is characteristic of a highquality
machine shop, 10
−5
is achievable with welldesigned and operated hard tooling, and
10
−6
is very diﬃcult and costly.” However, length tolerances σ
ε
as low as 3 10
−7
has recently been achieved for the AstroMesh [166]. It was decided to analyse the
present threering tension truss for three tolerances: σ
ε
= 10
−4
, 10
−5
and 10
−6
.
The results for the rms surface error and defocus are given in Table 6.6. Of impor
tance to the performance of the antenna is also to maintain a suﬃcient stress level
in the tension truss. The lowest net element force was therefore sought for each
142
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
simulation. For F/D = 0.4 and σ
ε
= 10
−4
, 10
−5
and 10
−6
the interval and mean
value (in parenthesis) of the lowest element force were 0.0463–5.97 N (1.18 N), 5.57–
6.97 N (6.43 N), and 6.95–7.09 N (7.03 N), respectively. For F/D = 0.6 they were
0.0635–8.63 N (3.38 N), 8.82–10.4 N (9.78 N), and 10.3–10.5 N (10.4 N), respec
tively. For F/D = 0.8 they were 0.0642–11.8 N (6.57 N), 12.4–13.8 N (13.2 N), and
13.8–13.9 N (13.8 N). Apparently, a tolerance of σ
ε
= 10
−4
cannot be accepted.
6.4.8 Inﬂuence of Tension Tie Load Variation
The tension ties are socalled constantforce springs with relatively low stiﬀness so
that an elongation results only in a small change of force. Compared to the length
imperfections, the achievable force accuracy is much lower. Hedgepeth et al. [62] use
a standard deviation of 0.1 N for a magnitude of 1 N. In the following analysis, two
tie force variations were used: 0.1 and 0.05. The case of the lower accuracy, σ
τ
= 0.1,
yielded unsatisfactory results and therefore the higher accuracy was required.
The interval and mean value (in parenthesis) of the lowest element force for F/D =
0.4, 0.6, and 0.8 with σ
τ
= 0.1 were 0.177–5.85 N (3.52 N), 0.218–8.16 N (5.56 N),
and 1.04–11.7 N (8.18 N). There was a risk, although slim, that some cables become
slack. For F/D = 0.4 the lowest force was lower than 0.25 N in only two cases of the
100 simulations, otherwise it was higher than 1.25 N. Using the smaller tolerance,
σ
τ
= 0.05, gave intervals and mean values 3.51–6.54 N (5.56 N), 5.46–9.87 N (8.62 N),
and 9.26–13.4 N (11.7 N), for the same values of F/D. Clearly, σ
τ
= 0.05 provided
a better chance of keeping a good prestress in all elements, which is necessary to
avoid severe mesh saddling.
Table 6.5: Surface accuracy and defocus for a threering tension truss as a function
of the number of Monte Carlo simulations (D = 3 m, F/D = 0.4, ρ = 0.1
and σ
ε
= 10
−6
).
Number of
100 200 300 400 500
simulations
¯
δ
rms,z
(mm) 3.341749 3.341747 3.341744 3.341743 3.341743
Diﬀerence (%) — −0.000070 −0.000150 −0.000185 −0.000191
σ
δ
rms,z
(mm) 0.000055 0.000059 0.000060 0.000060 0.000062
Diﬀerence (%) — 7.43 9.56 9.23 11.56
∆F + ∆Z
P
(mm) 22.470104 22.469671 22.469767 22.469752 22.469726
Diﬀerence (%) — −0.000019 −0.000015 −0.000016 −0.000017
σ
∆F+∆Z
P
(mm) 0.004709 0.005031 0.005107 0.005186 0.005233
Diﬀerence (%) — 0.068 0.084 0.10 0.11
143
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
T
a
b
l
e
6
.
6
:
I
n
ﬂ
u
e
n
c
e
o
f
m
e
m
b
e
r
l
e
n
g
t
h
i
m
p
e
r
f
e
c
t
i
o
n
o
n
t
h
e
r
m
s
s
u
r
f
a
c
e
d
e
v
i
a
t
i
o
n
a
n
d
d
e

f
o
c
u
s
(
t
h
r
e
e

r
i
n
g
t
e
n
s
i
o
n
t
r
u
s
s
,
1
0
0
M
o
n
t
e
C
a
r
l
o
s
i
m
u
l
a
t
i
o
n
s
)
.
F
/
D
0
.
4
0
.
6
0
.
8
σ
ε
1
0
−
4
1
0
−
5
1
0
−
6
1
0
−
4
1
0
−
5
1
0
−
6
1
0
−
4
1
0
−
5
1
0
−
6
´ δ
ε r
m
s
,
z
(
m
m
)
0
.
4
1
1
2
0
.
1
1
2
7
0
.
0
3
9
2
1
0
.
4
3
8
8
0
.
1
0
1
3
0
.
0
3
6
5
6
0
.
4
9
8
9
0
.
1
0
3
5
0
.
0
3
4
6
9
¯ δ
ε r
m
s
,
z
(
m
m
)
0
.
2
1
9
5
0
.
0
2
5
9
0
.
0
1
8
2
8
0
.
2
6
5
8
0
.
0
2
9
9
0
.
0
2
0
0
3
0
.
3
3
2
2
0
.
0
4
0
9
0
.
0
2
2
3
9
σ
ε ∆
F
+
∆
Z
P
(
m
m
)
0
.
4
8
9
6
0
.
0
4
7
0
0
.
0
0
4
7
1
1
.
4
5
2
1
0
.
1
4
2
4
0
.
0
1
3
7
5
3
.
4
2
7
3
0
.
3
0
4
8
0
.
0
3
0
6
5
∆
F
+
∆
Z
P
(
m
m
)
2
2
.
4
3
9
7
2
2
.
4
7
2
1
2
2
.
4
7
0
1
0
2
2
.
3
8
0
6
2
2
.
5
9
4
6
2
2
.
6
1
2
5
9
2
3
.
9
6
3
9
2
4
.
9
7
1
6
2
4
.
9
6
3
7
1
σ
ε r
a
n
d
n
1
.
0
1
1
0
1
.
0
0
6
9
1
.
0
0
2
6
3
0
.
9
9
5
7
0
.
9
9
8
2
1
.
0
0
2
6
3
1
.
0
0
1
9
1
.
0
0
2
6
1
.
0
0
4
3
4
¯ δ
r
m
s
,
z
(
m
m
)
3
.
3
4
8
9
3
.
3
4
1
8
3
.
3
4
1
7
5
2
.
2
4
3
6
2
.
2
2
8
0
2
.
2
2
7
8
9
1
.
7
0
3
6
1
.
6
7
1
4
1
.
6
7
1
0
0
δ
m
a
x
r
m
s
,
z
(
m
m
)
3
.
3
6
9
3
3
.
3
4
3
1
3
.
3
4
1
9
1
2
.
2
6
9
1
2
.
2
3
0
3
2
.
2
2
8
1
0
1
.
7
3
1
0
1
.
6
7
3
9
1
.
6
7
1
2
1
δ
m
i
n
r
m
s
,
z
(
m
m
)
3
.
3
3
7
1
3
.
3
4
0
0
3
.
3
4
1
6
2
2
.
2
2
8
6
2
.
2
2
5
8
2
.
2
2
7
7
2
1
.
6
7
5
4
1
.
6
6
9
4
1
.
6
7
0
8
5
σ
δ
r
m
s
,
z
(
m
m
)
0
.
0
0
6
0
0
.
0
0
0
6
0
.
0
0
0
0
6
0
.
0
0
9
0
0
.
0
0
0
7
0
.
0
0
0
0
7
0
.
0
1
3
4
0
.
0
0
0
9
0
.
0
0
0
0
7
∆
F
(
m
m
)
9
.
1
3
9
7
9
.
1
6
9
3
9
.
1
6
8
1
0
1
3
.
5
2
8
4
1
3
.
7
2
9
8
1
3
.
7
4
6
3
3
1
7
.
3
4
7
6
1
8
.
3
2
4
5
1
8
.
3
1
6
1
9
∆
F
m
a
x
(
m
m
)
1
0
.
2
5
4
8
9
.
2
5
6
9
9
.
1
7
6
5
4
1
7
.
7
1
7
7
1
4
.
1
2
5
1
1
3
.
7
7
1
9
4
2
5
.
5
7
9
6
1
8
.
9
1
4
1
1
8
.
3
8
3
2
5
∆
F
m
i
n
(
m
m
)
8
.
2
0
7
9
9
.
0
6
2
1
9
.
1
5
6
0
1
1
0
.
4
5
9
4
1
3
.
3
3
5
9
1
3
.
7
0
8
5
2
8
.
7
7
7
4
1
7
.
4
6
5
2
1
8
.
2
4
9
8
2
σ
∆
F
(
m
m
)
0
.
4
2
1
9
0
.
0
4
0
1
0
.
0
0
4
0
3
1
.
3
5
2
8
0
.
1
3
2
3
0
.
0
1
2
7
7
3
.
2
7
8
3
0
.
2
9
2
2
0
.
0
2
9
1
7
∆
Z
P
(
m
m
)
1
3
.
3
0
0
0
1
3
.
3
0
2
8
1
3
.
3
0
2
0
0
8
.
8
5
2
1
8
.
8
6
4
8
8
.
8
6
6
2
6
6
.
6
1
6
4
6
.
6
4
7
1
6
.
6
4
7
5
2
∆
Z
m
a
x
P
(
m
m
)
1
3
.
5
2
7
3
1
3
.
3
2
1
1
1
3
.
3
0
3
8
0
9
.
2
1
3
9
8
.
9
0
0
2
8
.
8
6
8
7
5
7
.
0
5
9
7
6
.
6
7
9
1
6
.
6
5
1
3
9
∆
Z
m
i
n
P
(
m
m
)
1
3
.
1
4
9
4
1
3
.
2
8
5
6
1
3
.
2
9
9
8
5
8
.
6
0
6
1
8
.
8
3
0
7
8
.
8
6
3
1
9
6
.
2
2
6
2
6
.
6
0
7
3
6
.
6
4
3
7
0
σ
∆
Z
P
(
m
m
)
0
.
0
7
7
0
0
.
0
0
7
9
0
.
0
0
0
7
9
0
.
1
1
4
2
0
.
0
1
1
5
0
.
0
0
1
1
2
0
.
1
5
9
0
0
.
0
1
4
6
0
.
0
0
1
6
1
144
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
Table 6.7: Inﬂuence of tension tie force variation on the rms surface deviation and
defocus (threering tension truss, 100 Monte Carlo simulations).
F/D 0.4 0.6 0.8
σ
τ
0.1 0.05 0.1 0.05 0.1 0.05
´
δ
τ
rms,z
(mm) 0.1656 0.11030 0.1854 0.12845 0.2274 0.14592
¯
δ
τ
rms,z
(mm) 0.0366 0.02711 0.0790 0.05704 0.1281 0.06173
σ
τ
∆F+∆Z
P
(mm) 0.0765 0.03707 0.2968 0.14942 0.7357 0.43429
∆F + ∆Z
P
(mm) 22.4641 22.47280 22.5509 22.61540 24.8709 24.91352
σ
τ
randn
1.0027 1.00001 0.9957 1.00810 0.9986 0.99714
¯
δ
rms,z
(mm) 3.3419 3.34181 2.2292 2.22853 1.6758 1.67199
δ
max
rms,z
(mm) 3.3451 3.34319 2.2340 2.23053 1.6875 1.67601
δ
min
rms,z
(mm) 3.3379 3.34065 2.2240 2.22572 1.6684 1.66735
σ
δ
rms,z
(mm) 0.0013 0.00057 0.0021 0.00099 0.0035 0.00174
∆F (mm) 9.1629 9.17070 13.6891 13.74996 18.2262 18.26742
∆F
max
(mm) 9.3482 9.24285 14.2461 14.06349 19.9889 19.23906
∆F
min
(mm) 8.9964 9.08753 13.0219 13.41423 16.2504 17.19885
σ
∆F
(mm) 0.0680 0.03307 0.2798 0.14064 0.7116 0.41983
∆Z
P
(mm) 13.3012 13.30211 8.8618 8.86544 6.6447 6.64610
∆Z
max
P
(mm) 13.3253 13.31591 8.9073 8.88727 6.7302 6.68603
∆Z
min
P
(mm) 13.2816 13.29318 8.8138 8.84171 6.5870 6.59766
σ
∆Z
P
(mm) 0.0096 0.00470 0.0190 0.00994 0.0282 0.01703
6.4.9 Inﬂuence of Ring Structure Distortion
In the following analysis of the inﬂuence of the ring structure distortion, it was
assumed that the random distortions in x, y, and zdirections were normally dis
tributed with a zero mean value and the same standard deviation, σ
ρ
. In the real
antenna structure this type of error arises due to length errors in the members of the
ring structure, which are of two diﬀerent types: cables and struts. It is likely that
one of these can be manufactured to a higher precision than the other. Hence, the
least accurate of the types will give rise to the largest error at the ring nodes; this was
the basis of the assumption of using a single standard deviation. The member length
imperfection inﬂuence analysis indicated that the accuracy level 10
−4
could produce
cable slackening. Therefore, only the two higher accuracy levels, 10
−5
and 10
−6
, were
used in the ring structure distortion analysis. For each node, the nodal distortion
in each of the x, y and zdirections was computed as D
ρ
, where
ρ
∼ N(0, σ
ρ
).
The results from the ring distortion analysis are shown in Table 6.8. The interval
and mean value (in parenthesis) of the lowest element force for F/D = 0.4, 0.6,
and 0.8 with σ
ρ
= 10
−5
were 0.0387–5.26 N (0.716 N), 0.0342–8.08 N (0.725 N), and
145
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Table 6.8: Inﬂuence of ring structure distortion on the rms surface deviation and
defocus (threering tension truss, 100 Monte Carlo simulations).
F/D 0.4 0.6 0.8
σ
ρ
10
−5
10
−6
10
−5
10
−6
10
−5
10
−6
´
δ
ρ
rms,z
(mm) 0.4442 0.08298 0.8320 0.09032 0.6798 0.09899
¯
δ
ρ
rms,z
(mm) 0.2454 0.04088 0.4439 0.01156 0.3608 0.06146
σ
ρ
∆F+∆Z
P
(mm) 0.7685 0.00945 1.2530 0.04909 1.1724 0.04861
∆F + ∆Z
P
(mm) 21.5504 22.46750 23.6198 22.65153 24.5268 24.91198
σ
ρ
randn
0.9861 0.98562 1.0027 0.98153 0.9723 1.00060
¯
δ
rms,z
(mm) 3.3507 3.34195 2.2716 2.22783 1.7094 1.67198
δ
max
rms,z
(mm) 3.3623 3.34252 2.3308 2.22889 1.7707 1.67288
δ
min
rms,z
(mm) 3.3392 3.34130 2.2262 2.22673 1.6710 1.67054
σ
δ
rms,z
(mm) 0.0068 0.00026 0.0355 0.00060 0.0315 0.00060
∆F (mm) 8.5018 9.16349 14.6794 13.77344 17.6941 18.27448
∆F
max
(mm) 9.2674 9.18197 17.1475 13.84509 19.9586 18.37599
∆F
min
(mm) 7.3351 9.14827 12.7152 13.71798 13.7381 18.19524
σ
∆F
(mm) 0.5959 0.00695 1.1382 0.03537 1.2003 0.03994
∆Z
P
(mm) 13.0486 13.30401 8.9405 8.87810 6.8327 6.63750
∆Z
max
P
(mm) 13.3672 13.31591 9.1928 8.90543 7.3125 6.66412
∆Z
min
P
(mm) 12.7981 13.29415 8.7002 8.85469 6.4979 6.62382
σ
∆Z
P
(mm) 0.1832 0.00510 0.1364 0.01408 0.2301 0.00955
0.0368–10.4 N (1.85 N), respectively. With σ
ρ
= 10
−6
, the corresponding intervals
and mean values were 5.53–7.02 N (6.49 N), 5.16–10.4 N (7.77 N), and 10.9–13.7 N
(11.8 N), respectively. Evidently, the tension truss is more sensitive to distortions
of the supporting ring than to member length imperfections. Nevertheless, it is
possible to adjust the ring structure during the assembly process, which is not the
case with the individual member lengths.
6.4.10 Combined Inﬂuence of Manufacturing Imperfections
In the previous sections, the individual inﬂuences of member length imperfection,
tension tie force variation, and distortion of the supporting ring structure on the rms
surface deviation and defocus were examined. The manufacturing accuracy of the
real reﬂector surface is determined by the combined eﬀect of all these imperfections.
An estimate of the surface accuracy can be obtained by computing the total rms
surface deviation from all error sources as
δ
rms,z
=
¸
i
δ
2
rms,z,i
1/2
. (6.21)
146
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
It was shown above that a low tolerance may result in cable slackening even though
the decrease in surface accuracy is small; the algorithm used for computing the rms
surface deviation does not take cable slackening into account. The combined eﬀect
of the random errors will increase the risk of cable slackening, so high accuracy levels
are needed. Two tolerance levels were tested:
• I: σ
ε
= 10
−5
, σ
ρ
= 10
−5
, and σ
τ
= 0.1.
• II: σ
ε
= 10
−6
, σ
ρ
= 10
−6
, and σ
τ
= 0.05.
The results are given in Table 6.9. For level I, the interval and mean values of the
lowest element force were 0.0181–2.72 N (0.125 N), 0.0283–5.58 N (1.07 N), and
0.0728–10.5 N (5.27 N), for F/D = 0.4, 0.6 and 0.8, respectively. Thus, level I does
not achieve a suﬃcient level of accuracy as the combined eﬀect of the manufacturing
tolerances resulted in cable slackening. For level II, the intervals and mean values
were 0.938–6.00 N (3.99 N), 5.61–9.98 N (8.10 N), and 8.28–12.9 N (11.2 N), using
the same order as above. Hence, level II is the tolerance goal for an accurate reﬂector
antenna.
6.4.11 Inﬂuence of Thermal Strains
Another error source, that is somewhat related to the manufacturing accuracy, is the
straining of the elements due to thermal loading. During a mission, the spacecraft
will be both in and out of the Earth’s shadow and, thus, subjected to extreme
temperatures. To analyse the inﬂuence of thermal strains on the surface accuracy,
the maximum and minimum equilibrium temperatures of the inorbit tension truss
have to be found. The equilibrium temperature of a body in space is estimated by
an energy balance. Conservation of energy yields that the absorbed energy is equal
to the emitted energy. Each element of the tension truss is treated as a thin ﬂat plate
with no side insulated and whose surface normal is parallel to the solar rays and
passes through the centre of the Earth. In the energy balance of a body in space,
the heat inputs are the direct solar ﬂux, Earthreﬂected solar ﬂux (albedo), and
Earthemitted infrared (IR) ﬂux. The heat output is the emitted radiation energy
from the body. Setting up this energy balance, assuming the same IR emissivity
and solar absorptivity on the top and bottom surfaces of the net bands, and solving
for the worstcase hot temperature yields [119]
T
max
=
¸
q
IR
ε
IR
sin
2
ρ + G
S
α
S
(1 + aK
a
sin
2
ρ)
2ε
IR
σ
1/4
, (6.22)
where q
IR
is the Earth IR emission (237 ± 21 W/m
2
), ε
IR
the IR emissivity on the
band surface, ρ the angular radius of the Earth (sin ρ = R
E
/(h+R
E
)), R
E
the radius
of the Earth (6,378,140 m), h the altitude of the body, α
S
the solar absorptivity on
the band surface, G
S
the solar ﬂux (1326–1418 W/m
2
depending on season), a the
albedo of direct solar ﬂux (0.30±0.05), K
a
a correcting factor for the reﬂection of the
147
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Table 6.9: Combined inﬂuence of manufacturing imperfections on the rms surface
deviation and defocus (three ring tension truss, 100 Monte Carlo simu
lations).
F/D 0.4 0.6 0.8
Tolerance level I II I II I II
´
δ
c
rms,z
(mm) 0.6339 0.15472 0.4490 0.13489 0.6092 0.13946
¯
δ
c
rms,z
(mm) 0.3902 0.08536 0.2846 0.05896 0.4143 0.03420
σ
c
∆F+∆Z
P
(mm) 0.5271 0.04887 0.6758 0.15815 0.9932 0.45069
∆F + ∆Z
P
(mm) 21.8707 22.41913 22.3707 22.60387 24.0263 25.22523
σ
ε
randn
1.0039 0.99641 1.0042 1.00835 0.9951 1.00447
σ
ρ
randn
0.9945 1.01713 1.0084 0.98381 1.0254 0.98031
σ
τ
randn
1.0028 1.00164 1.0010 0.99887 0.9908 1.00538
¯
δ
rms,z
(mm) 3.3644 3.34279 2.2459 2.22858 1.7215 1.67120
δ
max
rms,z
(mm) 3.3963 3.34456 2.2813 2.23157 1.7743 1.67657
δ
min
rms,z
(mm) 3.3424 3.34066 2.2278 2.22500 1.6788 1.66679
σ
δ
rms,z
(mm) 0.0123 0.00083 0.0089 0.00110 0.0190 0.00182
∆F (mm) 8.7067 9.14152 13.4875 13.73870 17.5499 18.53619
∆F
max
(mm) 9.4895 9.23777 16.0529 14.02969 20.2019 19.47404
∆F
min
(mm) 7.4097 9.04804 11.8542 13.30536 14.4719 17.29273
σ
∆F
(mm) 0.4414 0.03721 0.6164 0.14884 0.9334 0.42772
∆Z
P
(mm) 13.1640 13.27761 8.8833 8.86517 6.4763 6.68904
∆Z
max
P
(mm) 13.3538 13.31202 9.0786 8.89629 6.6566 6.76164
∆Z
min
P
(mm) 12.9545 13.25132 8.7021 8.83677 6.2391 6.62063
σ
∆Z
P
(mm) 0.1013 0.01571 0.0801 0.01155 0.0877 0.03455
solar energy oﬀ the spherical Earth (K
a
= 0.664 + 0.521ρ −0.203ρ
2
), and σ Stefan
Boltzmann’s constant (5.67051 10
−8
W/m
2
K
4
). The worstcase cold temperature
occurs when the body is in the shadow of the Earth and out of view of any portion
of the sunlit parts of the Earth. For this condition there is no direct solar ﬂux,
G
S
= 0, or albedo, a = 0. The worstcase cold temperature is [119]
T
min
=
¸
q
IR
sin
2
ρ
2σ
1/4
. (6.23)
Note that the T
min
is independent of the surface properties. The above equation
give estimates of the equilibrium temperature for the worstcase conditions. For the
net elements, the angle between the solar rays and the surface normal of the bands
varies over the reﬂector surface resulting in diﬀerent equilibrium temperatures, which
further degrades the surface. A correction of the equilibrium temperature for each
individual band due to the angle of the solar rays was not made here as this analysis
serves only as an estimate.
148
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
The net elements are made of CFRP and it was assumed that their surface properties
were that of carbon black paint, [119], α
S
= 0.96 and ε
IR
= 0.88. The altitude
of the spacecraft to which the antenna is attached was h = 550 km, which gave
sin ρ = 0.9206. The worstcase hot temperature of the net bands, using q
IR
=
258 W/m
2
, G
S
= 1418 W/m
2
, and a = 0.35, was computed to T
max
= 374 K
or ϑ
max
= +101
◦
C. The worstcase cold temperature, using q
IR
= 216 W/m
2
, was
T
min
= 200 K or ϑ
min
= −73
◦
C. If the antenna was mounted on a satellite in a GEO
(H = 35,786 km, sin ρ = 0.1513), like the 12.25 m diameter AstroMesh antenna
on Thuraya, the worst case temperatures would be ϑ
max
= +69
◦
C and ϑ
min
=
−192
◦
C, using the same surface properties as above. According to Panetti [119],
a typical operating temperature range of a parabolic reﬂector is −160 to +95
◦
C.
The distortion test of a 6 m diameter AstroMesh uses worstcase temperatures of
−160
◦
C and +120
◦
C [166]. In the present analysis, the following temperatures
were used:
• Assembling temperature ϑ
ref
= +20
◦
C,
• Worst case cold temperature ϑ
min
= −80
◦
C, and
• Worst case hot temperature ϑ
max
= +110
◦
C.
If the equilibrium temperature variation due to the incident solar rays is neglected,
the inﬂuence of the thermal strains on the antenna accuracy is mainly characterised
by a change in average temperature [58]. The unstrained length of an element i
after a change in temperature is
0,ϑ,i
=
0,i
(1 + ε
ϑ
) (6.24)
where
0
is the unstrained length at the reference temperature ϑ
ref
, and the ε
ϑ
the
thermal strain. The thermal strain at a temperature ϑ is
ε
ϑ
= α
T
(ϑ −ϑ
ref
) , (6.25)
where α
T
is the CTE. One of the advantages of CFRP is that it is possible to theoret
ically achieve zero CTE by an appropriate choice of material and layup parameters.
In reality, manufacturing imperfections and variations in material properties result
in nonzero CTE. CFRP with α
T
less than 1.0µ/
◦
C is readily available, but with
out costly testing it is diﬃcult to make the variation of the CTE, σ
α
T
, less than
0.4µ/
◦
C, [62]. This variation results in a random thermal strain, even under a uni
form change in temperature over the tension truss. In an early study, Hedgepeth [56]
uses α
T
= 0.5µ/
◦
C for the CFRP. More recently, thinwalled composite tubes with
near zero CTE are developed [157]; depending on the layup of the carbon ﬁbre
sheets, the CTE of the tubes varies from −0.20µ to 0.16µ/
◦
C. Following this brief
study of the achievable CTE of CFRP for space structures, the following values for
the mean value and standard deviation of the CTE were used in the present analysis:
¯ α
T
= 0.5µ/
◦
C and σ
α
T
= 0.4µ/
◦
C.
Before computing the random errors of the thermal strains, it is necessary to examine
how a uniform temperature change aﬀects the rms surface error and defocus. In
149
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Table 6.10: Inﬂuence on the rms surface deviation and defocus of a uniform deter
ministic thermal strain in the threering tension truss.
F/D 0.4 0.6 0.8
Temperature ϑ
min
ϑ
max
ϑ
min
ϑ
max
ϑ
min
ϑ
max
δ
ϑ
rms,z
(mm) — 0.2220 — 0.2085 — 0.2029
∆F + ∆Z
P
(mm) 22.8387 22.1382 23.5076 21.8072 26.8721 23.2415
δ
rms,z
(mm) 3.3338 3.3491 2.2176 2.2375 1.6585 1.6831
∆F (mm) 9.3448 9.0088 14.3755 13.1798 19.8796 16.9042
∆Z
P
(mm) 13.4939 13.1294 9.1321 8.6274 6.9925 6.3373
t
min
all
(N) 7.14 7.07 10.58 10.41 14.10 13.75
t
max
all
(N) 46.46 46.35 67.92 67.57 89.83 89.05
this study, the CTE was assumed to be deterministic and equal to ¯ α
T
. The results
of this analysis are shown in Table 6.10. The net forces were more or less unaﬀected
by a uniform temperature change. Note that ϑ
min
produced a smaller systematic
rms surface deviation due to an increased focal length.
The results from the thermal analysis with a random CTE are given in Table 6.11.
For the worstcase cold temperature the interval and mean values (in parenthesis)
of the lowest element force for F/D = 0.4, 0.6, and 0.8 were 0.470–6.76 N (4.44 N),
4.30–9.95 N (7.89 N), and 8.28–13.3 N (11.1 N). For the worstcase hot temperature,
the intervals and mean values were 2.04–6.68 N (4.44 N), 4.91–10.0 N (7.86 N), and
8.32–13.2 N (11.1 N) in the same order as above. There is a slight possibility that
some cables go slack, but the average minimal forces are acceptable.
6.4.12 Achievable Reﬂector Accuracy
A comparison of the rms surface deviation results for member length errors with
(6.11) showed that
¯
δ
ε
rms,z
was about the same as the corresponding value of (6.11) for
10
−4
and 10
−5
, but diﬀered by approximately one order in magnitude for 10
−6
. This
was a bit puzzling since σ
δ
rms,z
always diﬀered with about one order of magnitude.
However, this anomaly was not considered to be a major problem as the surface
errors for σ
ε
= 10
−6
were lower than the resolution of common measuring systems
1
.
For σ
ε
= 10
−5
and 10
−6
,
´
δ
ε
rms,z
was almost constant at 0.11 and 0.04 mm, respectively.
As anticipated, the rms surface deviation was relatively insensitive to a variation of
the tension tie forces. This would not be the case with a kinematically indeterminate
network. For the lower tolerance, σ
τ
= 0.1,
´
δ
τ
rms,z
varied between 0.17 and 0.23 mm,
while for the higher one, σ
τ
= 0.05, it varied between 0.11 and 0.15 mm. These
errors were about the same order as the member length errors. The eﬀects of ring
distortions were more severe than the previous two manufacturing errors. For σ
ρ
=
10
−5
,
´
δ
ρ
rms,z
varied between 0.44 and 0.83 mm, where the highest magnitude was
1
The video photogrammetry system used for the AstroMesh had a resolution of 0.07 mm rms,
[166].
150
6.4. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY
Table 6.11: Inﬂuence on the rms surface deviation and defocus of a random thermal
strain (threering tension truss, 100 Monte Carlo simulations).
F/D 0.4 0.6 0.8
Temperature ϑ
min
ϑ
max
ϑ
min
ϑ
max
ϑ
min
ϑ
max
´
δ
ϑ
rms,z
(mm) 0.2282 0.2182 0.1898 0.2221 0.2402 0.2326
¯
δ
ϑ
rms,z
a
(mm) 0.0895 0.0732 0.1173 0.1079 0.1400 0.1285
σ
ϑ
∆F+∆Z
P
(mm) 0.1851 0.1808 0.5575 0.4946 1.2298 1.0949
∆F + ∆Z
P
(mm) 22.8507 22.1311 23.4806 21.8259 27.0176 23.2851
σ
ϑ
randn
1.0009 1.0009 1.0027 1.0004 0.9994 0.9990
¯
δ
rms,z
(mm) 3.3350 3.3499 2.2207 2.2401 1.6644 1.6880
δ
max
rms,z
(mm) 3.3401 3.3550 2.2294 2.2480 1.6755 1.6981
δ
min
rms,z
(mm) 3.3290 3.3427 2.2149 2.2340 1.6556 1.6807
σ
δ
rms,z
(mm) 0.0022 0.0021 0.0027 0.0028 0.0038 0.0037
∆F (mm) 9.3529 9.0032 14.3502 13.1987 20.0154 16.9481
∆F
max
(mm) 9.8201 9.4572 15.7129 14.1209 23.4995 19.0815
∆F
min
(mm) 8.9989 8.5603 13.2259 11.8385 16.4306 13.8639
σ
∆F
(mm) 0.1582 0.1536 0.5148 0.4595 1.1769 1.0493
∆Z
P
(mm) 13.4978 13.1280 9.1303 8.6272 7.0023 6.3371
∆Z
max
P
(mm) 13.5803 13.2154 9.2769 8.7171 7.1456 6.4529
∆Z
min
P
(mm) 13.4224 13.0506 9.0143 8.5167 6.8232 6.1939
σ
∆Z
P
(mm) 0.0310 0.0300 0.0474 0.0405 0.0585 0.0526
a
Note that the eﬀect of CTE variation is computed by (6.18) as before, but now with the
systematic error from Table 6.10.
obtained for F/D = 0.6. For σ
ρ
= 10
−6
,
´
δ
ρ
rms,z
varied between 0.08 and 0.10 mm,
where the higher value corresponded to F/D = 0.8.
As the rms surface error was very sensitive to the ring distortion, the ring structure
must be very accurately constructed; it may be necessary to provide some kind of
adjustability, to ﬁne tune its shape during ground testing. The combined eﬀects
of the three manufacturing errors was at most 0.64 mm at level I and 0.16 mm
at level II. Considering the slack cables at level I, it is obvious that level II must
be the manufacturing tolerance goal. However, it may be possible to use the lower
tolerance of the tension tie forces, σ
τ
= 0.1 and still have a high accuracy. The eﬀects
of the thermal loading,
´
δ
ϑ
rms,z
, were about the same for the three focal lengths: 0.19–
0.24 mm. These values were computed with respect to the systematic error of a
network with deterministic thermal strains. Taking the rss value of the systematic
and random thermal rms errors, the resulting maximum error was approximately
0.31 mm.
Taking all the random error sources above into account, the upper bound of the rms
surface deviation was about 0.35 mm, or 1.2 10
−4
D. This value can be compared
151
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
with Table 5.2, where the rss surface deviation of a 6 m diameter AstroMesh, due
to random error sources, is about 0.46 mm, or 7.7 10
−5
D.
At tolerance level II, the defocus, ∆F + ∆Z
P
, varied between 22.4 and 25.2 mm
depending on the focal length. Its variation, σ
∆F+∆Z
P
, at the same level, varied
between 0.05 and 0.45 mm. Since a defocus of up to one wavelength (31 mm
at 9.65 GHz) can be accepted without seriously aﬀecting the performance of the
antenna, [131], it cannot be considered a problem for the present antenna. In fact,
the defocus decreased signiﬁcantly as the number of net rings increased.
152
Chapter 7
Tensegrity Reﬂector Antennas
7.1 Introduction
A preliminary study of the new antenna concept is presented in references [168,169].
The present chapter will provide a more complete study of the concept starting from
the simple initial studies with the stiﬀened tensegrity module to the vibration char
acteristics of a full scale oﬀset reﬂector antenna. It was apparent from Figure 5.5
that the AstroMesh concept, Figure 2.15, is a major source of inspiration in the
development of the new concept. To better understand the similarities and diﬀer
ences between the two concepts, this chapter will start with a closer study of the
AstroMesh.
7.2 The AstroMesh Concept
Most details of the AstroMesh are found in the US patent [167] and in a techni
cal report from Astro Aerospace [62]. The main parts of the AstroMesh are the
triangular nets and the deployable ring structure.
7.2.1 Net Generation
The triangular nets of the AstroMesh have a conﬁguration similar to that of the
tension truss with three sets of bands oriented approximately 60
◦
apart. Ideally,
the bands should follow the geodesic lines of the surface, as a lateral loading on
the surface would not tend to shift the bands in that position. However, with
equally spaced nodes along the circumference, the intersection of a set of three
geodesic bands, connected to the boundary, would not necessarily coincide. To
minimise the number of intersections, a quasigeodesic net, with coinciding three
band intersections, is used instead of a true geodesic net. This quasigeodesic net
is called geotensoid. The generation of the geotensoid starts with a hexagonal array
of equilateral triangles, in this case with six rings, Figure 7.1(a). For a circular
153
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
(a) (b) (c)
Figure 7.1: Geotensoid generation: (a) sixring tension truss, (b) initial sixring
geotensoid, and (c) converged sixring geotensoid (redrawn from [62]).
aperture, the nodes of the sixring array are then mapped to lie on equally spaced
concentric circles, Figure 7.1(b). Finally, the free nodes are iteratively adjusted so
that the total length of the net bands is minimised, Figure 7.1(c).
7.2.2 Deployable Ring Structure
The deployable ring structure of the AstroMesh, shown in Figure 7.2(a), is composed
of a series of upper and lower bars which, when connected, form upper and lower
rings. The upper and lower rings, each composed of B segments, are separated by
vertical and diagonal bars. A bar and joint count for the ring truss yields b
ring
= 4B
and j
ring
= 2B. As it is obvious that this ring truss cannot be prestressed, s = 0,
the number of internal mechanisms is, (1.2),
m
ring
= 2B −6. (7.1)
Only for B = 3, when the ring becomes a triangular prism, is the structure kinemat
ically determinate. Usually the ring is divided into several bays, hence, the number
of internal mechanisms is quite large. The main purpose of the two triangular net
works is to eliminate the internal mechanisms of the ring truss. Ultimately, the
complete assembly should be statically determinate for easy prestressing.
7.2.3 Static and Kinematic Properties
Initially, the nodes on the circumference connect directly to the ring truss, Fig
ure 7.1(c); the number of bays is therefore equal to the number of net rings. The
total numbers of bars and joints of such an assembly, with 6n bays, are
b
6n
= 6n(3n −1)
. .. .
b
nets
+ 24n
....
b
ring
(7.2)
154
7.2. THE ASTROMESH CONCEPT
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 17 18
19
20
21
22
23
24
25
26
27
28
29
30
(a) (b)
Figure 7.2: Conﬁguration of the AstroMesh onboard the Thuraya satellite: (a) line
drawing created from Figure 2.15(b) and (b) highlighted rings of the top
net.
and
j
6n
= 6n(n + 1) + 2, (7.3)
respectively. Substituting (7.2) and (7.3) into the extended Maxwell’s rule, (1.2),
yields m− s = 0. As neither the ring truss nor the synclastic triangular nets can
sustain a state of selfstress, the resulting assembly is statically determinate. A
2.5 m diameter AstroMesh reﬂector (n = 4) is built according to this approach and
set up for various tests [165]. However, this approach will work only for small values
of n as the ring truss composes the major part of the reﬂector mass; for large values
of n, the antenna is too heavy. Schemes for connecting nets with many rings to
trusses with less than 6n bays must be developed. These schemes are presumably
available within TRW Astro Aerospace, but not in the open literature.
Figure 7.2(a) shows the deployable ring truss and complete top net of the 12.25
m diameter AstroMesh reﬂector. This ring truss has 30 bays. To better see the
conﬁguration of the top net and compare it with the tension truss, the ring truss
is removed and the net rings highlighted in Figure 7.2(b). Each net has nine rings
and is connected to the ring structure at 30 nodes. Referring to Figure 7.2(b), the
conﬁguration of the net at connections 1, 6, 11, 16, 21, and 26 is similar to that
of the tension truss. The remaining connections are similar but diﬀerent from the
previous ones. The net conﬁguration at these remaining connections is chosen to
render the complete reﬂector statically determinate. This conﬁguration may change
if the number of rings in the nets or the number of bays in the deployable ring
truss changes. Generally, the number of ring segments increases with the number
of net rings. Figure 7.3 shows eight conﬁgurations of ﬁve to ninering nets which
satisﬁes Maxwell’s rule, (1.1). It is observed that for an odd number of rings the
nets have the same type of ring truss connections. For an even number of net rings,
several alternative connections are possible. This small exercise clearly illustrates
the great ﬂexibility of the AstroMesh concept and the simple theory behind it. It
also emphasises the importance of static determinacy, which by virtual work implies
155
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
(a) n = 9 (b) n = 8 (c) n = 7
b = 1872
j = 626
b = 1440
j = 482
b = 1404
j = 470
30 bays 30 bays 24 bays 24 bays
b = 1188
j = 398
24 bays 18 bays 12 bays
b = 864
j = 290
b = 828
j = 278
b = 792
j = 266
(d) n = 6
18 bays
b = 648
j = 218
(e) n = 5
Figure 7.3: Statically determinate conﬁgurations of the AstroMesh.
kinematic determinacy, in the design of precision structures.
1
It will be seen that
the new concept, to which the rest of this chapter is devoted, was designed along
the lines of the AstroMesh design.
7.3 New Concept
The idea behind the new concept originates from a study by Knight et al. [71], where
an antenna design based on tensegrity is introduced. However, no speciﬁc details
about the design are given. When reference [71] was published the present author
was working at DSL with tensegrity formﬁnding methods and tensegrity masts.
1
The question “is static determinacy the key to the design of precision structures?” was discussed
at the IUTAMIASS Symposium on Deployable Structures (pp. 488–489 in [129]) Conclusions from
that discussion are that static determinacy is desirable as these structures are easier to model than
statically indeterminate ones, but it is not necessarily a solid factor to base the design on.
156
7.3. NEW CONCEPT
About the same time DERA contacted DSL for help with developing deployable
structures for the STRV. As mentioned earlier, a parabolic antenna was one of the
structures in which they were interested. The initial idea by Pellegrino [126] was that
an antenna based on tensegrity could be developed by using a hexagonal tensegrity
module and add nodes and bars inside the two hexagons to remove the internal
mechanisms. The resulting structure would have a potential to meet the stringent
requirements by DERA—the work with the tensegrity antenna was initiated.
7.3.1 Stiﬀened Hexagonal Module
Consider the regular hexagonal tensegrity module in Figure 7.4. This structure has
j = 12 joints and b = 24 bars. With c = 6, the extended Maxwell’s rule, (1.2), gives
m−s = 6. (7.4)
Like all tensegrity modules, s = 1 with only the longer members in compression.
Firstorder stiﬀness can be achieved by prestressing the structure, but this provides
only a relatively small stiﬀness; highprecision applications require dimensionally
accurate structures, i.e. kinematically determinate structures.
An improved version of the module is shown in Figure 7.5. This structure was
obtained by connecting the nodes of the top and bottom hexagons to two inter
connected, central joints. Note that these internal joints are not coplanar with the
hexagons, thus forming two triangulated surfaces that coarsely approximate to a
curved surface. The modiﬁed assembly has j = 14 joints and b = 37 bars. The
extended Maxwell’s rule now yields
m−s = −1. (7.5)
Since the same state of selfstress of the structure in Figure 7.4 also exists for this
structure and there is no other independent state of selfstress, s = 1 as before.
(a) (b)
Figure 7.4: Hexagonal tensegrity module: (a) threedimensional view and (b) top
view.
157
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
(a)
θ
(b)
Figure 7.5: Stiﬀened hexagonal tensegrity module: (a) threedimensional view and
(b) top view.
Hence, this structure has no internal mechanisms and is, therefore, potentially useful
for the present application. Closer analysis showed that in the initial conﬁguration,
where the joints of the top hexagon lie directly above the ones of the bottom hexagon,
all of the added bars were unstressed, meaning that they must be rigid. However, if
the upper hexagon was rotated clockwise with respect to the lower one by an angle
θ, i.e. the compression members became longer, the state of selfstress changed such
that all members became stressed. It turned out that all of the additional members
ended up in tension and can be substituted by cables. Thus, the resulting structure
has only six compression members and, if cables are used for the remaining members,
the structure can easily be folded by collapsing the struts.
Figure 7.6 shows how the force density in the members changes with the angle
of rotation θ. Four conﬁgurations, which diﬀered in the positions of the internal
interconnected joints, were analysed. The tension force in the elements forming
the triangulated surfaces, called net cables, increases from its initial zero value.
The compression force in the struts increases monotonically, i.e. they become more
compressed. The tension force in the base cables, however, decreases and they
eventually end up in compression; the angle when this happens depends on the
length of the interconnected element, ∆H. The triangulated surface in this example
is far too coarse to support a reﬂective mesh that approximates a paraboloid. To
reﬁne the surface, the simple triangulated surfaces are replaced by tension trusses.
7.3.2 Hexagonal Tensegrity Module and Tension Trusses
Consider a pinjointed bar structure consisting of the original ring structure, i.e. the
hexagonal module, plus two tension trusses. The total numbers of joints and bars
for that structure are, cf. section 6.1,
j = 2
1 + 6
n(1 + n)
2
+ 6
, (7.6)
158
7.3. NEW CONCEPT
0 5 10 15 20 25 30
2
1.5
1
0.5
0
0.5
1
1.5
2
0.01
0
.
0
1
0
.
1
0
0.10
0.25
0
.
2
5
0.50
∆
H
/
H
=
0
.
5
0
Struts
N
et cab
les
Base cables
q
i
q
lateral
θ (
◦
)
Figure 7.6: Force density values for the onestage tensegrity reﬂector for various
∆H/H. ∆H is the length of the interconnected element and H is the
total height of the module (q
lateral
= 1).
and
b = 6 4 + 2
¸
6
n(1 + 3n)
2
+ 6(n + 1)
, (7.7)
respectively. Substituting (7.6) and (7.7) into the extended Maxwell’s rule gives
m−s = 0. (7.8)
Again, at θ = 0
◦
no additional states of selfstress has been created, s = 1. This
means that there exists one internal mechanism. The immediate diﬀerence between
this structure and the previous one is that the interconnected element is missing.
Adding an additional bar, which connects the middle nodes of the nets, may remove
this single mechanism. However, a way of prestressing the complete structure must
be found. An obvious approach, based on the AstroMesh, is to connect the corre
sponding nodes of the two nets with tension ties. It must be emphasised that the
tension ties should not be counted as bars. The tension ties are springs providing
constant forces and can, therefore, be treated as external loads in an analysis. It
is a common misconception that the inclusion of the tension ties would result in a
structure that is highly statically indeterminate and, therefore, suﬀers from all sorts
of prestressing diﬃculties; the AstroMesh is a direct proof that this is not the case.
159
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
To obtain detailed information about the static and kinematic properties of the
antenna structure, its equilibrium matrix has to be analysed for diﬀerent values of
θ. But before an adequate conﬁguration of the antenna can be speciﬁed, one more
parameter must be determined—the separation between the nets.
7.3.3 Minimum Separation between Front and Rear Nets
The minimum distance between the closest nodes of the nets in the AstroMesh is de
termined by the shortest acceptable length of the tension tie connecting these nodes.
For the present concept, the minimum separation depends on both the curvature of
the nets, F/D, and the position of the struts, θ. Consider two paraboloids with the
same F/D ratio separated by a distance ∆H at the apexes, Figure 7.7. A system
xyz has its origin at the apex of the upper paraboloid. A strut connects points 1
and 2 on the rims of the paraboloids. The coordinates of these points are
p
1
=
D
2
cos θ
D
2
sin θ
D
2
16F
T
, (7.9)
p
2
=
D
2
cos 240
◦
D
2
sin 240
◦
z
2
T
. (7.10)
If the paraboloids are too closely spaced the strut will intersect them. Hence, the
minimum separation is when the strut lies in the tangential plane to the upper
paraboloid at p
1
(or to the lower one at p
2
). The normal to the upper paraboloid
at p
i
is
n(p
i
) =
−
x
i
2F
−
y
i
2F
1
T
, (7.11)
which yields the tangent plane at p
1
as
z −z
1
=
x
1
2F
(x −x
1
) +
y
1
2F
(y −y
1
) . (7.12)
Substituting (7.9) and (7.10) into (7.12) yields the vertical position z
2
of the rim of
the bottom paraboloid. Subtracting the height of the bottom paraboloid yields the
minimum separation
∆H =
D
2
8F
cos (60
◦
−θ) . (7.13)
Thus, the minimum total height of the antenna structure is
H =
D
2
8F
[1 + cos (60
◦
−θ)] . (7.14)
Note that even if the bottom paraboloid is made shallower than the top one, the
total height of the antenna will remain constant; the coordinate z
2
is determined by
the deepest paraboloid. The separation ∆H is zero at θ = −30
◦
, ∆H = D
2
/16F
at θ = 0
◦
, and ∆H = D
2
/8F at θ = 60
◦
; the last ﬁgure is easily veriﬁed. This
means that the deployed antenna will be rather deep, e.g. D = 3 m, F/D = 0.4,
and θ = 10
◦
yield H = 1.54 m.
160
7.3. NEW CONCEPT
1
x
z
y
2
n
1
x
y
n
1
2
1
Tangent plane at 1
(a) (b)
Figure 7.7: At the conﬁguration with the theoretical minimum separation between
the front and rear nets, the strut lies in the tangent planes of the
paraboloids: (a) threedimensional view and (b) top view.
7.3.4 ThreeRing AxiSymmetric Reﬂector
It is now possible to describe the conﬁguration of an axisymmetric antenna with
ﬁxed D and F/D by only one parameter, θ. An antenna with threering nets
was analysed. The net conﬁguration and the tension tie forces were according to
Chapter 6 for n = 3. The diameter D = 3 m and F/D = 0.4. Corresponding nodes
of the two nets were connected by only tension ties, hence, m−s = 0. For θ = 0
◦
, the
structure had one internal mechanism and large compressive forces were induced in
the ring structure; the state of selfstress for θ = 0
◦
was a prestressed ring structure
but unstressed nets. Like the stiﬀened tensegrity module, the whole structure can be
prestressed with θ > 0
◦
. However, contrary to the stiﬀened module, the static and
kinematic properties of the antenna assembly changed when going from the initial
to a rotated conﬁguration. At θ = 0
◦
, m = 1 and s = 1, but for θ > 0
◦
, m = 0
and s = 0, i.e. the structure is statically determinate. By itself, the rotated ring
structure could no longer be prestressed as s = 0. Nevertheless, when the complete
reﬂector structure was considered, including the prestressing forces applied by the
tension ties, it was found that only six of the 252 elements were in compression.
The variation of the internal forces in the threering antenna with the angle θ was
analysed. Throughout the studied range, 0–30
◦
, all of the net forces were in tension
161
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
0 5 10 15 20 25 30
0
1
2
3
4
5
6
8
9
10
12
16
17
18
19
F
o
r
c
e
(
N
)
θ (
◦
)
(a)
0 5 10 15 20 25 30
4
5
6
7
8
9
10
13
14
15
1
2
3 4 5
6 7
8 9 10 11 12
13
14
15
16
17 18
19
F
o
r
c
e
(
N
)
θ (
◦
)
(b)
Figure 7.8: Variation of the forces in the net cables for the threering conﬁguration.
162
7.3. NEW CONCEPT
0 5 10 15 20 25 30
200
150
100
50
0
50
100
150
200
Lateral cable
Base cable
Strut
1.4
1.45
1.5
1.55
1.6
1.65
1.7
1.75
1.8
Ring height
F
o
r
c
e
(
N
)
H
e
i
g
h
t
(
m
)
θ (
◦
)
Figure 7.9: Variation of the forces in the ring structure and its height for the three
ring conﬁguration.
with a magnitude between about 1 and 10 N for tension tie loads of 1 N on the
internal joints and 2 N on the edge and additional joints, Figure 7.8. The forces in
the ring structure, however, were extremely large for low values of θ, but decreased
exponentially to acceptable levels at about 5
◦
, Figure 7.9. For this particular con
ﬁguration, the upper limit for θ, at which the base cables were no longer in tension,
was about 24
◦
. When looking for an adequate conﬁguration one should not only
take the magnitudes of the internal forces into account. Also shown in Figure 7.9 is
the height of the ring structure from (7.14). It went from 1.41 m at 0
◦
to 1.70 m at
24
◦
and the goal must be to keep it as low as possible. Another issue, which might
aﬀect the choice of θ, is the possible interference between the struts and the tension
ties. This problem was not considered in the present analysis, but when a physical
structure is built it is of primary importance that the struts are not interfering with
the tension ties, or vice versa. As the angle θ increased, the struts moved closer to
the centre of the of reﬂector. This means not only that the struts become longer
but that they are also more likely to interfere with the tension ties, at least during
the deployment procedure if not at the deployed state. Therefore, θ should be kept
small. Considering all of these issues, θ = 10
◦
seems like a good choice for the
present example.
163
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
0 cm 5 10 15 20
Figure 7.10: Paraboloidal plastic mold with the ﬁrst of the two threering nets at
tached. The radial cables are used to connect the net to the ring struc
ture.
7.4 Demonstration Model
To verify the feasibility of the proposed concept, a smallscale physical model was
constructed. Based on the analysis in the previous section, the model had threering
nets and a rotation angle of 10
◦
.
The triangular nets
2
were constructed on a paraboloidal mold of Vivak
3
with a
diameter of 0.45 m and focal length of 0.134 m, Figure 7.10, on which the position
of the net joints had been marked with a threeaxis computercontrolled machine.
The nets were made of 0.8 mm diameter Kevlar cords which were straightened and
taped to the molds; the cords were joined with Nylon loops at all crossover points
and bonded with epoxy resin. This manufacturing technique was not very accurate.
Systematic length errors were introduced when the cords followed the arc lengths
between nodes instead of the straight lines and at the crossover points where only
one of the three cords lay on the surface of the mold. The latter error could have
been minimised by using thin steel or CFRP tapes instead of the cords but since this
was the ﬁrst model, at a rather small scale, it was decided not to choose material not
readily available. To summarise, the total length error was estimated at about 1 mm
per net element, which is undesirably high, but with the available material it could
not be made smaller. Corresponding nodes of the two nets were later connected by
lengthadjusted ﬁshing line and steel springs.
2
The conﬁguration of each net was not exactly as given earlier; the additional 6 joints and 24
elements outside the edges were left out as the net conﬁguration of the demonstration model was
based on an earlier study, cf. [169].
3
Vivak is a registered trademark of Sheﬃeld Plastics Inc. for glycol modiﬁed polyethylene
terephthalate (PETG), a thermoplastic copolyester.
164
7.4. DEMONSTRATION MODEL
(a)
(b)
Figure 7.11: Demonstration model: (a) top view and (b) bottom view.
165
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
Based on the experience from the construction of the tensegrity masts, it was de
cided to manufacture special joints with precisiondrilled holes for the ring structure.
Twelve identical aluminium alloy, 30 mm long joint ﬁttings of cylindrical shape with
a diameter of 15 mm were manufactured. In each joint ﬁtting, holes with 2.0 mm
diameter in the direction of each ring cable and a radial net cable were drilled. A
1.0 mm Kevlar cord was used for the ring cables. The cords were connected to the
joint ﬁttings by epoxy resin. The struts were made of 6.4 mm diameter aluminium
rods, which ﬁtted in 20 mm long holes at the bottom of the joint ﬁttings. These
holes were coaxial with the joints. Grub screws held the joints on the struts. The
joints were well made and functioned satisfactorily, despite them being too large
compared to the rest of the model. A smaller, spherical joint would have been ideal,
but also more costly to produce.
The model worked quite well, considering it was the ﬁrst time that a structure of
this kind had been constructed in DSL. However, some of the cables in the two
nets remained slack after deployment and there was some interference between the
nets and struts, Figure 7.12. This was mainly due to the length errors in the nets,
which made them deeper. Since, in the vicinity of the ring connections, the nets
are close to the joints, even in an ideal structures, a large length error is bound to
result in struttonet contact. This interference could hardly have been avoided by
further separating the nets; the length errors were simply too large. In addition,
the relatively large diameter and cylindrical shape of the joint ﬁttings prevented
the nets from being attached close to ring structure, Figure 7.13. Correcting these
problems should be possible in a secondgeneration model, e.g., by using thin bands
for the nets, a more accurate mold and smaller joints in the ring structure.
7.5 Deployment Schemes
The success of any deployable structure lies in the actual deployment; it does not
matter how accurate or stiﬀ the structure is in its deployed state if it fails to deploy.
The deployment of the present structure relies entirely on the way the struts are
unfolded; the unfolding rate must be easy to control. Facing the identical problem,
Knight [72] lists four possible solutions for strut deployment:
• Hinged struts,
• Sliding coupling struts,
• Telescopic struts, and
• Inﬂatable struts.
Hinges of various types have been used on deployable systems for several decades. A
simple and reliable hinge is the TSR hinge described in section 4.5.1. Its automatic
locking capability makes it especially interesting for this application.
166
7.5. DEPLOYMENT SCHEMES
Figure 7.12: Side view of demonstration model.
Figure 7.13: Cylindrical joint ﬁtting with grub screw.
A sliding coupling, with a locking mechanism, is an alternative to the hinge. With
sliding couplings it takes minimal force to deploy the strut but signiﬁcant force to
stow it again. However, sliding couplings are fairly new and also introduce stiﬀness
nonlinearities [72].
167
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
Telescoping structures have not been used very often in space applications due to
excessive weight and drive force required [72]. However, Becchi and Dell’Amico, [7],
have developed an interesting 2.4 m telescopic mast consisting of seven tubes. For
increased stiﬀness, the tubes slide on tightﬁtting Vespel
4
pads, and a minimum
overlap of one tube diameter is maintained.
The last option is to use inﬂatable struts. This approach can minimise the stowed
volume, but the size and weight is comparable to the three previous schemes [72]. An
inﬂatable strut would use manufacturing and inﬂation techniques similar to those
of inﬂatable antennas. After inﬂation, the struts need to be rigidised to ensure their
structural integrity throughout the mission lifetime. A rigidised strut would have a
uniform crosssection and a minimum of stiﬀness nonlinearities [72].
One design issue, which is critical for deployable structures with cables, is snag
prevention. There is a potentially large risk that the long slack ring cables get
caught or looped around a strut during deployment. To avoid snagging they must
be stowed in a clever way. Knight [72] and Duﬀy et al. [35] study the possibility of
using highly elastic cables, which eﬃciently prevent snagging. However, in such an
approach the structure is subjected to very high stowage forces and stiﬀness creep.
In addition, no deployable structure can be allowed to have a too quick deployment
sequence as high shock and vibration may be introduced into the spacecraft [72]. A
slow, controlled deployment is desired.
Following this brief review of deployment options, it was decided to investigate the
folding of the small reﬂector model using, ﬁrst, hinged struts and, then, telescopic
struts.
7.5.1 Hinged Struts
The aluminium rods were replaced by wooden rods, each having two hinges along its
length. The hinges were made by small pieces of metal plate which easily could be
bent to a speciﬁc angle, Figure 7.14. More reﬁned struts were not made as it seemed
that the chances for this folding approach to work were very low given the almost
certain entanglement of the struts in the tension springs; still it had to be tried.
Figure 7.15 shows the reﬂector under two early stages of folding. The entanglement
started immediately and became more severe as the folding continued. On top of
that, the struts started to interfere with each other, and thereby prevented a compact
package. These problems can be seen in Figure 7.15(b). This simple test showed
that, for this structure, hinged struts were not an option. Even for the threering
conﬁguration, which had relatively few tension ties, the entanglement problems were
too severe.
4
Vespel is a registered trademark of E. I. du Pont de Nemours and Company.
168
7.5. DEPLOYMENT SCHEMES
Figure 7.14: Wooden struts with two hinges at diﬀerent stages of folding.
(a) (b)
Figure 7.15: Folding of the antenna using hinged struts ended in failure.
7.5.2 Telescopic Struts
Six telescopic struts, each 0.46 m long, were made by cutting oﬀ the sticks of six
identical foldable umbrellas. Special connections were made at the ends of the sticks
to make them ﬁt in the cylindrical joint ﬁttings. Each strut consisted of three tubes
of diﬀerent lengths as the umbrella sticks had to be shortened, Figure 7.16. The
stowed length of the strut, including the joint ﬁttings, was about 0.28 m. Of course,
169
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
Figure 7.16: Threepiece telescopic strut at diﬀerent stages of folding.
a more eﬃcient packaging could be achieved using custom made struts. The folding
improvement was immediate; no entanglement between struts and springs or inter
ference between struts occurred. To crudely simulate an actual folding, the model
was suspended by a ﬁshing line and then folded by hand, followed by a length adjust
ment of the ﬁshing line, Figure 7.17. This simple simulation looked very promising,
but problems could arise if the struts deploy in an unsynchronised manner. A conclu
sion from this simple test was that synchronously deployable telescopic struts, with
some kind of motor synchronisation, would be the best alternative for a successful
deployment.
7.6 Preliminary Design of 3 m Reﬂectors
In this section, the full scale antenna for a future STRV mission will be designed.
First, the design procedure will be described, including various simpliﬁcations and
engineering judgements. This procedure will then be applied to axisymmetric and
oﬀset reﬂector conﬁgurations. Finally, the dynamic characteristics of the best an
tenna conﬁguration will be computed.
7.6.1 Design Scheme
Network Density
As shown in section 5.4 the required network density, i.e. the number of rings in the
tension truss, depends on the wavelength λ and the focal length F. Due to mesh
saddling the actual facet surface error is higher than the ideal one. Hedgepeth [56]
sets the force t in the net elements equal to ten times the mesh tension p multiplied
by the facet side length . Equation (5.27) yields the surface error
δ
rms,z
= 0.01667
2
F
, (7.15)
170
7.6. PRELIMINARY DESIGN OF 3 M REFLECTORS
(a) (b)
(c) (d)
Figure 7.17: Simple folding simulation with the antenna model suspended in a ﬁsh
ing line.
which is only 3.3% larger than the ideal one, (5.29). To preserve this surface error,
the smallest internal force in the tension truss should be t = 10p. The corresponding
required triangle side length is
= 7.745
Fδ
rms,z
. (7.16)
As decided in section 5.4.6, two accuracy goals apply, with facet error contributions
of λ/100 and λ/50, respectively. At 9.65 GHz the wavelength is 31 mm, so the
allowable facet errors were 0.3 mm and 0.6 mm. Note, however, that the number of
rings was maximised to seven according to Chapter 6, so the lower error might not
be achievable with a deep reﬂector.
The nets were assumed to be constructed from CFRP band with a crosssection of
5 0.2 mm
2
. The density of the bands was 1740 kg/m
3
. This value was doubled
171
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
in the mass estimation to take the weight of the joints into account. It was further
assumed that the same CFRP bands were to be used throughout the net, although
the elements along the edges might be too highly stressed and need strengthening.
To take this strengthening into consideration, the total length of the net bands was
increased by 10% in the mass calculation.
Mesh
The RF reﬂective mesh were tensioned to either 2 or 10 N/m, as discussed in section
5.5.2. Previous results on the present reﬂector, cf. [169], are based only on the lower
mesh tension. A higher mesh tension will more eﬃciently smooth out the creases
formed during folding and give a more robust antenna structure as the prestress
level increases. A higher mesh tension will, however, result in a heavier structure.
To facilitate comparison, both values were used in the design study.
When estimating the antenna mass, the surface density of the mesh, 0.025 kg/m
2
,
was doubled to account for seams and surface treatment. The mesh area could be
approximated with that of a spherical cap of equal height and equivalent radius
as the reﬂector. However, for a more accurate mass comparison between diﬀerent
antenna conﬁgurations, it was decided to compute the actual area of the stretched
mesh.
Tension Tie Forces
At each node, each of the three bands running continuously through the node turns
an angle /2F, Figure 7.18. The required force in the tension tie to equilibrate these
net forces is
t
tie
= 1.5
t
F
. (7.17)
Using this value might result in net forces less than 10p, but as long as the majority
of the net elements have this force the surface degradation should be negligible. For
practical purposes t
tie
was given in full Newtons only, with a minimum value of 1 N.
Hence, the tension tie force by (7.17) was rounded to the nearest integer towards
inﬁnity.
Ring Structure
Once the required net tension and tension tie forces had been established, the forces
in the ring structure could be computed.
First, a safety factor with respect to material failure had to be set. This factor
depends on the acceptable risk of failure for the mission and may vary for diﬀerent
structural details, cf. [143]. For CFRP compression members, Hedgepeth et al. [62]
use a maximum stress of 200 MPa, which corresponds to a safety factor against
material failure between 5 and 10, depending on the strength of the CFRP. This
172
7.6. PRELIMINARY DESIGN OF 3 M REFLECTORS
R
t
tie
R ≈ 2F
Figure 7.18: Equilibrium of a node in the tension truss.
allowable stress value was also used here for the net elements, ring cables and struts,
for the case of material failure.
The struts were basically designed to resist Euler buckling, subject to constraints on
the slenderness, l
e
/r
g
, where l
e
is the eﬀective length of the strut and r
g
the radius
of gyration. The minimum slenderness ratio was found by equating the critical
buckling stress and the proportional limit stress σ
pl
:
l
e
r
g
min
=
π
2
E
σ
pl
. (7.18)
This slenderness was used only for conﬁrmation of the elastic buckling assumption.
For the present application, E = 227.5 GPa and σ
pl
≈ 1500 MPa, (l
e
/r
g
)
min
=
39. The maximum slenderness depends on the required axial stiﬀness and length
precision of the strut. If the strut is not straight, the axial stiﬀness is severely
degraded. For a strut with both ends pinjointed, an initial sinusoidal imperfection,
w = sin (πx/l), reduces the axial stiﬀness to [57]
AE
eﬀ
=
AE
1 +
1
2
r
g
2
. (7.19)
For a thinwalled tube of radius a and wall thickness τ
r
g
=
I
A
≈
πa
3
τ
2πaτ
=
a
√
2
. (7.20)
173
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
A 1% reduction of the axial stiﬀness corresponds to /r
g
= 0.14. A thinwalled tube
should have /2a < 1/10, [57]. However, it is believed that much less crookedness,
say /2a < 1/50, is needed for this application. During ground testing, the strut
sags due to gravity. The midpoint deﬂection of a simply supported horizontal strut
of length l
s
is
=
5
384
ρgl
4
s
Er
2
g
, (7.21)
where ρ is the density and g the gravity. The change in length, ∆l, of an in
compressible beam due to bending is
1
2
l
0
(dw/dx)
2
dx where, in this case, w(x) =
ρgx(l
3
−2lx
2
+ x
3
) /24Er
2
g
. The corresponding strain due to this change in length
is
∆l
l
s
=
17
40320
ρgl
s
E
2
l
s
r
g
4
. (7.22)
The eﬀective length of the struts is equal to their actual length, l
e
= l
s
, which
changes with the rotation angle θ. The length of a deployed strut in an axisymmetric
reﬂector can be written as:
l
s
=
D
2
4
1
2
+ cos θ
2
+
√
3
2
+ sin θ
2
¸
¸
+
D
2
8F
2
[1 + cos (60
◦
−θ)]
2
1/2
.
(7.23)
With D = 3 m and F/D = 0.4, the strut length varies between 2.95 and 3.38 m, as
θ varies between 0 and 30
◦
. Setting /r
g
= 2
√
2/50 and substituting l
s
= 3.38 m,
E = 227.5 GPa, ρ = 1740 kg/m
3
and g = 9.80665 m/s
2
into (7.21) yielded l
e
/r
g
=
257. Inserting these values into (7.22) produced ∆l
/l
s
= 1.2 10
−7
, which is about
one order better than the required ring distortion σ
ρ
= 10
−6
. Hedgepeth et al. [62],
use a maximum slenderness of 200 for 4.25 m long CFRP struts with identical
material properties as above. In the following analysis, the maximum slenderness
was, therefore, set to 200. Thus, an initial value for the minimal strut radius a was
computed as:
a
ini
=
l
e
√
2
200
. (7.24)
With the wall thickness ﬁxed to 0.5 mm, the strut radius was adjusted until l
e
/r
g
≤
200. A subsequent step checked that P
cr
≥ 10P, where 10 is the chosen safety factor
against buckling and P the design compression force.
To ﬁt in the launch envelope, 0.1 0.2 0.8 m
3
, the 3.38 m long struts had to be
collapsed to less than a quarter of their length. In the telescopic mast by Becchi and
Dell’Amico, [7], a minimum overlap of one tube diameter is maintained to achieve
adequate stiﬀness. Thus, struts up to about 3.1 m in length can be constructed from
four tubes. However, at this preliminary stage it is not possible to give a detailed
mass calculation of the struts. Therefore, the mass of the tubular struts of constant
crosssection was ampliﬁed by 50% to allow for tube overlap and the variation in the
crosssection of the telescopic struts. The mass of the deployment motor, latches,
cables and pulleys was estimated as 0.2 kg per strut [168].
The tendons of the ring structure are preferably CFRP bands, which easily can be
folded. Bands are also less likely to get tangled up during folding because basically
174
7.6. PRELIMINARY DESIGN OF 3 M REFLECTORS
only two folding directions are permitted. The maximum allowable stress in the
cables was, as above, 200 MPa. The mass of the connections between the cables and
the struts were estimated to 0.05 kg per connection.
7.6.2 AxiSymmetric Reﬂectors
In Table 6.3, the required number of rings for a certain facet surface accuracy was
given. The lower surface accuracy, 0.6 mm, was nearly obtained with seven rings
for F/D = 0.4, six rings for F/D = 0.6, and ﬁve rings for F/D = 0.8. The higher
surface accuracy, 0.3 mm, was nearly achievable with seven rings for F/D = 0.8.
These four combinations, denoted I–IV, were analysed in terms of internal forces for
diﬀerent values of the rotation angle θ and mesh tension p. Figure 7.19 shows the
results from the analysis of the internal forces when p = 10 N/m. In Figure 7.19(a),
the forces in the ring structure are shown. Like the threering reﬂector antenna,
they decreased rather rapidly in the beginning and are ended where a net force
ceases to be in tension. The minimum net forces in Figure 7.19(b) have almost
constant downward slopes; the steepness of the curves seems to depend on the
number of net rings. It was signiﬁcant that the minimum net force for the seven
ring conﬁgurations, I and IV, was much lower than for the other conﬁgurations. To
preserve the surface accuracy it ought to be about 10p = 21 N for seven rings, which
is about ten times the value at θ = 10
◦
. A closer study revealed that the elements
with the minimum force were located outside the highly stressed edge elements and
therefore not attached to the mesh. Slightly lower forces, say 2p, which increase
the additional rms surface error from 3.3% to 16.5%, may in some cases be also
accepted in elements connected to the mesh, if the aﬀected elements are in the
minority. Here, the forces were just too low and the two sevenring conﬁgurations, I
and IV, were discarded from further analysis. Thus, this eliminated the possibility
of achieving the higher surface accuracy, λ/50. Good designs for the remaining two
conﬁgurations, II and III, were obtained at θ = 10
◦
. At that angle the ring forces
had come down from their very large values at 5
◦
and the minimum net forces were
acceptable at 6.2 and 18.7 N, respectively. Mass estimates, following the scheme in
the previous section, for conﬁgurations II and III at θ = 10
◦
, are given in Table 7.1.
7.6.3 Oﬀset Reﬂectors
For oﬀset reﬂectors, the separation of the nets for the oﬀset reﬂector is computed by
(7.13), but with a corrected focal length F
eq
= ξF, where ξ is the ellipticity, (5.10).
This takes into account that for identical focal lengths and aperture diameters the
oﬀset reﬂector is shallower than the axisymmetric one. For example, D = 3 m,
F/D = 1.2, X
A
= 0 m, and θ = 10
◦
give a total ring height of 1.54 m in the axi
symmetric case but only 1.31 m in the oﬀset case (H
a
= 0.40 m and ∆H = 0.51 m).
Similarly, when computing the tension tie force from (7.17), F should be substituted
with F
eq
.
175
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
0 5 10 15 20 25 30
4000
3000
2000
1000
0
1000
2000
3000
0 5 10 15 20 25 30
5
0
5
10
15
20
25
III
II
I
IV
III
II
I IV
(a)
(b)
F
o
r
c
e
(
N
)
F
o
r
c
e
(
N
)
θ (
◦
)
θ (
◦
)
Figure 7.19: Forces in (a) strut and most stressed ring cable, and (b) least stressed
net member for four diﬀerent axisymmetric conﬁgurations with p = 10
N/m: (I) F/D = 0.4, n = 7; (II) F/D = 0.6, n = 6; (III) F/D = 0.8,
n = 5; (IV) F/D = 0.8, n = 7.
176
7.6. PRELIMINARY DESIGN OF 3 M REFLECTORS
The number of net rings for the required surface accuracy was computed using
(7.16). With focal lengths F = 1.5, 1.8, and 2.1 m the higher surface accuracy of
0.3 mm could not be achieved with less than eight rings. But, as stated before,
the number of rings was limited to seven. To achieve the lower surface accuracy of
0.6 mm seven rings were required for F = 1.5 m, while six rings were suﬃcient for
F = 1.8 and 2.1 m. In Table 6.2, the net forces for diﬀerent oﬀset conﬁgurations are
shown. However, a preliminary analysis with a threering oﬀset reﬂector showed that
the tension tie forces determined in the previous chapter do not produce the same
force distribution in the nets when the whole antenna structure was considered. In
addition, the mirror symmetry of the internal forces in the net was lost. This was
most likely a result of the loss of symmetry for the whole structure. The bottom
net was a mirror image of the top one in the aperture plane so that corresponding
nodes were connected by tension ties. Hence, neither mirror nor rotational symmetry
existed. However, the internal forces in the two nets were identical as the structure
still had the quasiﬂip symmetry, i.e. the eﬀect of turning the antenna upside down
so that the top net becomes the bottom net will only change the direction of the
struts and θ. Slight changes in the tension tie forces along the edges were, therefore,
necessary to reobtain a satisfactory internal force distribution.
For the present study, the reﬂectors had an oﬀset value of 0 or 0.3 m, as their force
relations were about the same for n = 6 and slightly worse for n = 7, as seen in
Table 6.2. Considering the increased prestressing problems discussed above, the
oﬀset antenna analysis were limited to six rings. Thus, only reﬂectors with focal
lengths of 1.8 and 2.1 m were studied. Figure 7.20 shows the modiﬁed tension tie
distribution along the edges for the sixring nets; the only change was an increase
from 4.5 to 5.0 for six of the ties. The results for the three oﬀset conﬁgurations are
shown in Figure 7.21. The variations of the forces in the ring structure were similar
for all three conﬁgurations. However, the variations of the minimum net forces
were drastically diﬀerent from those of the axisymmetric case. For conﬁguration
V, F = 1.8 m and X
A
= 0, the minimum net force was maximised at about 10
◦
,
while for VI, F = 2.1 m and X
A
= 0 m, and VII, F = 2.1 m and X
A
= 0.3 m, it
was maximised at 8
◦
and 12
◦
, respectively. For F = 1.8 m and X
A
= 0.3 m, it was
not possible to prestress the nets with the present tension tie distribution. Hence,
a feasible conﬁguration VIII did not exist. Note also that the minimum net force
was lower than the 10p required for good surface accuracy. However, the majority
of the net cables were stressed to the required value, so the overall surface accuracy
should not be seriously aﬀected by the lower tensions. Considering the magnitude of
the strut force and the strut–tie interference issues discussed earlier, it, once again,
seems that θ = 10
◦
is a good choice for the oﬀset conﬁgurations. Mass estimates
for conﬁgurations V–VII at θ = 10
◦
are given in Table 7.1. As the forces in the
ring structure were similar for all of the ﬁve remaining conﬁgurations, the resulting
masses were almost identical. Conﬁguration VI seems slightly better than the others
because of a larger minimum net force and a marginally lower weight.
177
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
T
a
b
l
e
7
.
1
:
M
a
s
s
e
s
t
i
m
a
t
e
s
o
f
r
e
ﬂ
e
c
t
o
r
c
o
n
ﬁ
g
u
r
a
t
i
o
n
s
I
I
,
I
I
I
,
a
n
d
V
–
V
I
I
a
t
θ
=
1
0
◦
.
D
e
s
c
r
i
p
t
i
o
n
C
o
n
ﬁ
g
u
r
a
t
i
o
n
I
I
I
I
I
V
V
I
V
I
I
F
o
c
a
l
l
e
n
g
t
h
(
m
)
1
.
8
2
.
4
1
.
8
2
.
1
2
.
1
O
ﬀ
s
e
t
d
i
s
t
a
n
c
e
(
X
A
)
(
m
)
—
—
0
0
0
.
3
N
u
m
b
e
r
o
f
n
e
t
r
i
n
g
s
6
5
6
6
6
M
e
s
h
t
e
n
s
i
o
n
(
N
/
m
)
1
0
2
1
0
2
1
0
2
1
0
2
1
0
2
I
n
t
e
r
n
a
l
t
i
e
f
o
r
c
e
a
(
N
)
6
2
6
2
5
1
5
1
5
1
M
i
n
i
m
u
m
n
e
t
f
o
r
c
e
(
N
)
6
.
2
2
.
1
1
8
.
7
6
.
2
5
.
4
1
.
1
8
.
2
1
.
6
6
.
0
1
.
2
M
a
x
i
m
u
m
n
e
t
f
o
r
c
e
(
N
)
3
6
7
1
2
3
2
9
8
9
9
4
4
6
9
0
4
7
1
9
5
5
0
5
1
0
1
R
i
n
g
c
a
b
l
e
f
o
r
c
e
(
N
)
1
6
8
1
5
6
1
1
3
2
3
4
4
1
1
6
9
2
3
3
9
1
8
4
6
3
7
0
1
9
0
1
3
8
1
S
t
r
u
t
f
o
r
c
e
(
N
)
−
3
1
0
0
−
1
0
3
4
−
2
5
2
0
−
8
4
0
−
3
2
2
0
−
6
4
4
−
3
5
4
5
−
7
0
9
−
3
6
7
7
−
7
3
6
E
ﬀ
e
c
t
i
v
e
s
t
r
u
t
l
e
n
g
t
h
(
m
)
2
.
9
1
2
.
8
3
3
.
0
4
2
.
9
6
3
.
0
0
S
t
r
u
t
d
i
a
m
e
t
e
r
b
(
m
m
)
8
4
5
8
7
7
5
3
8
8
5
1
8
9
5
2
9
1
5
3
E
u
l
e
r
b
u
c
k
l
i
n
g
f
o
r
c
e
(
k
N
)
3
0
.
3
1
9
.
9
0
2
4
.
6
5
7
.
9
7
3
1
.
9
6
6
.
1
4
3
4
.
8
8
6
.
8
7
3
6
.
3
1
7
.
0
9
M
e
s
h
a
r
e
a
(
m
2
)
5
.
0
6
5
.
0
7
5
.
4
8
5
.
3
7
5
.
5
0
N
e
t
l
e
n
g
t
h
(
m
)
2
0
1
.
3
6
1
7
0
.
3
3
2
0
9
.
2
4
2
0
6
.
5
9
2
0
9
.
1
3
R
i
n
g
c
a
b
l
e
s
l
e
n
g
t
h
(
m
)
3
0
.
0
2
2
9
.
3
1
3
1
.
9
4
3
1
.
0
7
3
1
.
7
0
S
t
r
u
t
s
l
e
n
g
t
h
(
m
)
1
7
.
4
6
1
6
.
9
8
1
7
.
9
2
1
7
.
5
4
1
7
.
7
1
R
i
n
g
h
e
i
g
h
t
(
m
)
1
.
0
3
0
.
7
7
0
.
9
4
0
.
8
3
0
.
8
1
M
a
s
s
c
o
m
p
o
s
i
t
i
o
n
M
e
s
h
(
k
g
)
0
.
2
5
0
.
2
5
0
.
2
7
0
.
2
7
0
.
2
8
N
e
t
(
k
g
)
0
.
7
7
0
.
6
5
0
.
8
0
0
.
7
9
0
.
8
0
C
o
n
n
e
c
t
i
o
n
s
(
k
g
)
0
.
6
0
0
.
6
0
0
.
6
0
0
.
6
0
0
.
6
0
M
o
t
o
r
s
a
n
d
l
a
t
c
h
e
s
(
k
g
)
1
.
2
0
1
.
2
0
1
.
2
0
1
.
2
0
1
.
2
0
R
i
n
g
c
a
b
l
e
s
c
(
k
g
)
0
.
4
7
0
.
1
6
0
.
3
6
0
.
1
5
0
.
5
0
0
.
1
1
0
.
5
4
0
.
1
1
0
.
5
5
0
.
1
1
S
t
r
u
t
s
(
k
g
)
5
.
9
8
4
.
1
2
5
.
3
3
3
.
6
5
6
.
4
3
3
.
7
1
6
.
3
6
3
.
7
0
6
.
5
7
3
.
8
1
T
o
t
a
l
(
k
g
)
9
.
2
7
7
.
1
0
8
.
3
9
6
.
5
1
9
.
8
0
6
.
7
0
9
.
7
6
6
.
6
7
1
0
.
0
0
6
.
8
0
a
T
e
n
s
i
o
n
t
i
e
f
o
r
c
e
s
a
r
e
r
o
u
n
d
e
d
t
o
n
e
a
r
e
s
t
i
n
t
e
g
e
r
(
N
)
t
o
w
a
r
d
s
i
n
ﬁ
n
i
t
y
.
b
B
u
c
k
l
i
n
g
f
a
i
l
u
r
e
h
a
s
a
h
i
g
h
e
r
s
a
f
e
t
y
f
a
c
t
o
r
t
h
a
n
m
a
t
e
r
i
a
l
f
a
i
l
u
r
e
t
o
a
c
c
o
u
n
t
f
o
r
i
m
p
e
r
f
e
c
t
i
o
n
s
a
n
d
c
r
o
s
s

s
e
c
t
i
o
n
a
l
v
a
r
i
a
t
i
o
n
s
.
T
h
e
b
u
c
k
l
i
n
g
f
o
r
c
e
m
a
y
i
n
s
o
m
e
c
a
s
e
s
p
r
o
d
u
c
e
a
s
t
r
e
s
s
a
l
i
t
t
l
e
h
i
g
h
e
r
t
h
a
n
2
0
0
M
P
a
,
b
u
t
t
h
i
s
i
s
n
e
g
l
e
c
t
e
d
t
o
f
a
c
i
l
i
t
a
t
e
a
f
a
i
r
m
a
s
s
c
o
m
p
a
r
i
s
o
n
.
c
C
r
o
s
s

s
e
c
t
i
o
n
a
l
a
r
e
a
s
a
r
e
r
o
u
n
d
e
d
t
o
n
e
a
r
e
s
t
i
n
t
e
g
e
r
(
m
m
2
)
t
o
w
a
r
d
s
i
n
ﬁ
n
i
t
y
.
178
7.6. PRELIMINARY DESIGN OF 3 M REFLECTORS
6
4.5
3.5
4
5
6
6
4.5
3.5
4
5
6
6
6
4.5
3.5
4
5
Figure 7.20: Tension tie forces along the edges for sixring oﬀset reﬂector antennas
(mirror symmetry). Interior elements are drawn thinner for clarity.
7.6.4 Stowage Considerations
Conﬁguration V has the longest struts, 3.04 m. If this strut is divided into four
pieces and the minimum overlap is one diameter, 88 mm, the stowed height of the
struts is 0.848 m. This is in excess of the limit 0.8 m. Only conﬁguration III with
the shortest struts, 2.83 m, came under the limit at 0.785 m. Using the smaller tube
diameter decreased the stowed height a little but only conﬁgurations II, III, and VI
stayed below the limit. The height limit can easily be met by dividing the strut into
ﬁve pieces or by accepting longer stowed struts. However, according to Figure 5.4
there is a hard limit of 891 mm on the stowed height, so the recommended limit
of 800 mm should not be exceeded and more tube segments only complicate the
strut design. Besides the height constraint, the stowed package must conform to the
width and depth requirements of 200 mm and 100 mm, respectively. Rectangular
stowage dimensions for a mesh reﬂector antenna are rather odd. Most mesh antennas
have a supporting structure composed of only stiﬀ elements so the package with the
stowed antenna tends to be cylindrical. The present structure, however, is not
bound to any particular shape of the stowed conﬁguration since the struts are not
connected to each other. Two natural ways of arranging the struts are shown in
Figure 7.22. The rectangular conﬁguration, Figure 7.22(a), is preferable in the
present case. With a maximum depth of 100 mm, the maximum strut diameter was
2 100/(2+
√
3) ≈ 53.5 mm. Note that this is only a theoretical value, not achievable
in practice, as space in between the struts must be provided for the folded mesh and
nets. At present, this space requirement cannot be quantiﬁed.
A circular conﬁguration, Figure 7.22(b), provides suﬃcient space in the middle for
the nets and mesh. The maximum diameter of the struts was 100/3 ≈ 33 mm, which
yields a slenderness of 257 for l
e
= 3 m. Hence, the circular conﬁguration was not
feasible. Returning to the rectangular conﬁguration it was immediately recognised
that a mesh tension of 10 N/m yielded a package that was too large to ﬁt into the
launch envelope. The lower mesh tension, 2 N/m, gave diameters 51–58 mm, which
better ﬁt the requirements on stowed volume. Considering both the stowage volume
and net forces, VI again seems to be the best conﬁguration.
179
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
0 5 10 15 20 25 30
4000
3000
2000
1000
0
1000
2000
3000
0 5 10 15 20 25 30
1
0
1
2
3
4
5
6
7
8
9
10
(a)
(b)
VI
VI
VII
VII
V
V
F
o
r
c
e
(
N
)
F
o
r
c
e
(
N
)
θ (
◦
)
θ (
◦
)
Figure 7.21: Forces in (a) strut and most stressed ring cable, and (b) least stressed
net member for three diﬀerent oﬀset conﬁgurations with n = 6 and
p = 10 N/m: (V) F = 1.8 m, X
A
= 0 m; (VI) F = 2.1 m, X
A
= 0 m;
(VII) F = 2.1, X
A
= 0.3 m.
180
7.7. VIBRATION ANALYSIS
R
R
R
7R
(
2
+
√
3
)
R
3R
(a)
(b)
Figure 7.22: Stowing of the struts: (a) rectangular and (b) circular.
7.7 Vibration Analysis
An important characteristic is the natural frequency of the deployed antenna. Here,
only conﬁguration VI, which was found to be slightly better than the others, was
analysed. For the vibration analysis, all data except the stiﬀness of the tension
ties are available. In the Monte Carlo simulations of Chapter 6, a maximum force
variation of σ
τ
= 0.1 was used. In general, the softer the springs of the tension
ties, the easier it is to control the force. However, too soft springs might cause
dimensional stability problems during ground testing, e.g. sagging of the bottom
net due to selfweight. During ground testing it is common to test the antenna
in two positions, cupup and cupdown, to estimate the eﬀects of gravity. Here,
the weight of the lower net was about 0.40 kg at a cupup position and 0.66 kg in
the cupdown position. In the cupdown position, the mass per internal node of the
lower net was 0.66/127 ≈ 0.005 kg or 0.05 N, in force units. By itself, this additional
loading will not cause severe degradation of the surface as its magnitude is of the
same order as the force variation σ
τ
= 0.1 N (t
tie
= 1 N). However, for a tie stiﬀness
of, say 10 N/m, the node was displaced 5 mm vertically by the 0.05 N load, which
is unacceptable. Assuming that a vertical displacement, i.e. defocus, of 0.2 mm
could be accepted in the cupdown position during ground testing, the required tie
stiﬀness was 250 N/m. Steel springs with this stiﬀness are readily available, but
some applications may require springs of another material with a smaller CTE.
In the following vibration analysis, the tension ties were assumed to be weightless.
The mass of the mesh was added to that of the top net, giving a length density
of 0.0064 kg/m, while the mass of the bottom net was 0.0038 kg/m. The mass of
the strut–cable connections were uniformly distributed over the total length of the
ring cables and the struts, and the mass of the motors and latches were uniformly
distributed over the total length of the struts. This gave a ring cable mass and
a strut mass of 0.0158 and 0.2919 kg/m, respectively. The total antenna weight
remained unchanged, 6.67 kg. As stated earlier, the attachment of the antenna
181
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
x
y
z
Figure 7.23: Supports for the reﬂector antenna in the vibration analysis. Symbols:
restrained in x, y and zdirection, ´ in x and y, • in x and ◦
unrestrained.
to the spacecraft was not investigated. Nevertheless, it is highly likely that the
antenna will be connected at point A, cf. Figure 5.7, and therefore the support
conﬁguration in Figure 7.23 was used in the vibration analysis. All element were
modelled using twonode bar elements with a consistent mass matrix, cf. [27]. The
ten lowest natural frequencies were 9.19, 13.32, 19.88, 26.58, 35.98, 47.27, 57.74,
65.77, 75.37, and 76.12 Hz. The modes for the two lowest frequencies are shown in
Figure 7.24. If the support with constraints in the x and ydirection, i.e. ´, was
constrained in the zdirection, i.e., became , the ﬁrst two frequencies increased to
13.30 and 17.78 Hz, respectively.
The lowest vibration frequency of a pinended strut is
f
s,1
=
π
2
EI
ml
4
. (7.25)
Assuming a constant tubular crosssection of 52 mm in diameter and 0.5 mm thick
ness, the 2.96 m long strut has f
s,1
= 37.33 Hz, which is well above the fundamental
frequency of the complete structure. The actual telescopic strut has lower bending
stiﬀness, which somewhat decreases the natural frequency. Figure 7.25 shows the
ratio of the buckling loads of a telescopic and a constant crosssection beam. The
beam was tubular with a maximum diameter of 52 mm and a tube thickness of
0.5 mm. The telescopic beam was divided into four segments and at each intersec
tion the diameter decreases ∆d; the minimum value of ∆d = 2t. It was assumed that
the tubes are free of any imperfections and have zero play at the intersections. The
buckling load was computed by the ﬁnite diﬀerence method, cf. [173], with a sub
division of l/100. For suﬃcient stiﬀness, the gap between the tube segments should
not be more than 3 mm (∆d = 4 mm). Hence, the safety factor of 10 for buckling
182
7.8. DISCUSSION
(a) f
1
= 9.19 Hz
(b) f
2
= 13.32 Hz
Figure 7.24: The two lowest vibration modes of the oﬀset antenna conﬁguration VI
with p = 2 N/m (the tension ties are removed for clarity).
has now decreased and will diminish further for the real strut due to manufacturing
errors, e.g. play between tube segments.
7.8 Discussion
It can be concluded that the requirements on the stowed dimensions are approx
imately met; the package has to be slightly deeper, maybe 140 mm, to make
room for the folded nets and mesh. This yields the following stowed dimensions:
d/D =
4 0.2 0.14/π/3 ≈ 0.06 and h/D = 0.8/3 ≈ 0.27. Only the 15 m diameter
HCA has so low stowed dimension values.
It is observed that the mass limit of 20 kg, set up by DERA, is easily achieved
by the present antenna at 6.67 kg or 1.24 kg/m
2
(A = 5.37 m
2
). Compared to
the 2.4 1.6 m
2
SAR reﬂect array with a mass of around 15.5 kg or 4.04 kg/m
2
,
183
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
0 2 4 6 8 10 12 14 16 18
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
l/4 l/4 l/4 l/4
d
d∆d
d2∆d
d3∆d
d = 52 mm, t = 0.5 mm
∆d (mm)
P
cr,t
P
cr,c
Figure 7.25: Buckling load of a fourpiece perfectly straight telescopic strut, P
cr,t
,
divided by the buckling load of a constant crosssection strut, P
cr,c
.
cf. [130], the reﬂector antenna is superior. Comparing the areal densities of the mesh
antennas in Table 2.1, which vary between 0.36 and 4.58 kg/m
2
, the present value
is amongst the best. The mass composition of the present antenna is 84.1% for the
ring structure (including connections, motors and latches), 11.8% for the nets, and
the remaining 4.1% for the mesh. In comparison, the 23.56 m diameter AstroMesh
(ξ = 1.118, n = 27) of reference [62] has a total weight of 144.9 kg, of which 67.0%
is the mass of the ring structure, 15.7% the mass of the nets and the remaining
17.3% belonging to the mesh. Although these two antennas diﬀer signiﬁcantly in
aperture size, it is apparent that the ring structure is the dominating factor when
considering the total mass. The relationship between the mass of the nets and that
of the mesh depends entirely on the number of net rings.
Earth observing systems typically require a fundamental natural frequency above
0.1 Hz [55]. With a lowest frequency of 9 Hz, the present antenna easily fulﬁls this
requirement. A high stiﬀness is also an advantage in ground testing as the gravity
compensation system can be simpliﬁed. In comparison, the SAR reﬂect array has a
frequency of 0.9 Hz [130].
Although it seems that the present antenna is suitable for a future STRV mission,
some issues remain that are either solved unsatisfactorily or not at all. Beside the
attachment to the spacecraft, which was mentioned earlier, the position of the struts
184
7.8. DISCUSSION
will always be an obstacle in the concept. The current position, where the struts go
right through the forest of tension ties, will seriously limit the number of net rings
and thereby the achievable surface accuracy. If the struts in some way could be
moved closer to the perimeter of the aperture, a great deal would be won, e.g. the
ring structure height could diminish. A possible reﬁnement of the ring structure is
to use a twostage tensegrity module instead of the onestage. The main advantages
are that the struts are closer to the boundary and that two parameters, the overlap
η and the rotation angle θ, can be used to change the prestress in the structure.
Consider a twostage tensegrity module with six struts per stage which has m = 7
and s = 1 by (4.3). Then add two interconnected, central joints and 13 bars in the
same fashion as in Figure 7.5. The resulting structure has m − s = −1. In the
conﬁguration given by (4.5), the stress in the added bars is zero. By changing η
and θ, the added bars can be stressed. Figure 7.26 shows the variation of the force
density in the bars. It is observed that the area of feasible conﬁgurations is very
small. Still, there is a possibility that such a conﬁguration will also work when the
extra bars are replaced with triangular networks.
Another aspect that needs further investigation is the layout of the nets for static
0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
.0
1
0
.0
1
0
.0
1
0
0
0
.5
0.5
0.5
1
1
1
1
base cables
net cables

1

1
1
2
struts
0
.2
5
0.25
0
.5
0.5
vertical cables
1
1
1
1
1
.
5
1
.5
s
a
d
d
l
e
c
a
b
l
e
s
configurations with s = 3 and m = 2
U
pper bound
L
o
w
er b
o
u
n
d
n
.c.
b.c.
b.c.
struts
s
.
c
.
v.c.
0
.1
0.1
0.2
0.2
n.c.
n.c.
θ (
◦
)
η
Figure 7.26: Force density values for the stiﬀened twostage tensegrity module (R =
H = 1, ∆H = H/100). The two thickest lines represent the upper
(q
base
= 0) and lower (q
net
= 0) bound for the overlap (q
diagonal
= 1).
185
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
determinacy of the complete structure and easy prestressing. In was shown in Chap
ter 6 that it became more diﬃcult to prestress the nets as the number of rings
increases. Finally, there are several manufacturing issues that will need to be inves
tigated before the antenna can be constructed. Those concerned with the achievable
manufacturing accuracies are especially important.
186
Chapter 8
Conclusions
8.1 Analysis Methods
One of the aims of the present work was to scrutinise the various methods for form
ﬁnding of tensegrity structures. It was found that the methods could be classiﬁed
into two groups—kinematic and static methods. Kinematic methods determine the
conﬁguration of either maximal length of the struts or minimal length of the cables,
while the length of the other type of element is kept constant. Static methods search
for equilibrium conﬁgurations that permit the existence of a state of prestress in the
structure with certain required characteristics.
The force density method and the energy method are found to be equivalent. In
the search for new conﬁgurations, the force density method is well suited since the
lengths of the elements of the structure are not speciﬁed at the start. However,
it is diﬃcult to control the variation in the lengths of the elements as the set of
force densities is varied. In general, the static methods seem to possess more usable
features than the kinematic ones. The reduced coordinate method or the force
method are suitable for problems where some parts of the geometry are known.
It is concluded that methods for analysing tensegrity structures are available, but
no single method is suitable for general problems. This is unsatisfactory and is
contrary to the formﬁnding of cable nets and membrane structures, where the most
popular methods, i.e. the force density and the dynamic relaxation methods, can
handle general structures. The present review of formﬁnding methods for tensegrity
structures oﬀers a base for anyone intending to further continue the subject.
The force method is very eﬃcient and highly suitable for the structural analysis
of large frameworks. It may be argued that the FEM is computationally more
eﬃcient, but these methods normally do not provide any information on the static
and kinematic properties of the framework; this information is essential in the early
stages of the structural design.
187
CHAPTER 8. CONCLUSIONS
8.2 Deployable Masts
Using recent advances in formﬁnding techniques, the equilibrium conﬁgurations of
tensegrity masts with adequate internal force distributions are easily found. This
design step is thus no longer a major obstacle.
The proposed manufacturing scheme worked well when all the early mistakes had
been eliminated. The use of joints with holes manufactured to the correct three
dimensional angles should produce a net with the required accuracy. A high length
accuracy is generally required, in the construction of cable structures, to obtain the
desired prestress and straightness of the mast.
The suggested strut deployment approach worked well. However, an important
deﬁciency of the present approach is that the mast do not achieve full stiﬀness until
the last stage has been deployed. Hence, the mast is very ﬂexible throughout the
entire deployment process and this seriously limits the applicability of the masts.
In the physical model, stiﬀness during deployment was provided by a central rod
which also controlled the rate of unfolding. For longer masts, some other means of
stiﬀening is needed.
The structural behaviour of the masts was in agreement with the ﬁndings of previous
studies. The masts were relatively stiﬀ axially but very ﬂexible in bending. Adding
cables to remove the internal mechanism did not improve the stiﬀness substantially.
While the additional cables improved the initial axial stiﬀness by about 50%, the
bending stiﬀness remained almost entirely unchanged, the ﬁrst cables becoming slack
under a small imposed load. This structural ineﬃciency of tensegrity structures with
noncontacting struts has also been observed in studies of doublelayer tensegrity
grids, cf. [53,179]. One way to stiﬀen tensegrities, which is suggested in these studies,
is to accept contacting struts. This will, however, reduce the excellent deployment
capabilities associated with discontinuous struts. The design will thus be a tradeoﬀ
between ease of deployment and compact packaging on one side and stiﬀness and
strength on the other.
8.3 Deployable Antennas
Miura and Pellegrino [105], inventors of several deployable structures concepts, claim
that “it is impossible to approach the ﬁeld of deployable structures with a single,
general concept or theory.” This statement is certainly true in the case of the antenna
proposed here. Several unique methods and concepts, each with its own particular
features, were combined to produce a robust structure which conforms well to the
requirements of a future space mission involving a small satellite.
The most critical constraint was the size of the stowed structure, and this restriction
more or less dictated the design of the ring structure. Due to its special characteris
tics and maturity, the tension truss was chosen to approximate the reﬂecting surface.
However, some modiﬁcations had to be done to the tension truss to render the assem
188
8.4. FURTHER RESEARCH
bly kinematically determinate, which is needed for the dimensional stability. One
aspect that needs further investigation is the layout of the nets, in order to ensure
static determinacy of the complete structure and easy prestressing. It was apparent
that the prestressing became more diﬃcult with the present layout, as the number
of rings increased. An accurate tension truss requires very tight length tolerances
for the elements. Typically, a maximum error of one part per million is required.
The variations of the coeﬃcient of thermal expansion and tension tie forces must
also be very low.
The proposed antenna concept is considered feasible for a future smallsatellite mis
sion as it approximately meets the requirements on the stowed dimensions and easily
meets the mass goal. The stiﬀness of the antenna, as indicated by the lowest natural
frequencies, is high compared to other antenna concepts. High stiﬀness is beneﬁcial
for ground testing as the gravity compensation system can be simpliﬁed.
Nevertheless, there are some issues that remain unsolved or solved unsatisfactorily.
One aspect is the attachment to the satellite bus and this has to be designed along
with the primary spacecraft structure. Another is the position of the struts. In their
current position the struts go right through the tension tie forest. This seriously
limits the number of net rings and thereby the achievable surface accuracy. There
are also several manufacturing issues that will need to be investigated before the full
scale antenna can be constructed. Critical factors are those aspects concerned with
the achievable manufacturing accuracies.
8.4 Further Research
To develop new tensegrity structures, suitable areas of application must be identiﬁed
and detailed requirements need to be formulated. Concerning the morphology of
tensegrities, the number of conceivable conﬁgurations exceeds the likely range of
applications. The main focus for future research must be on the implementation of
tensegrities. In this respect, some formﬁnding studies are certainly needed to ﬁnd
conﬁgurations that meet the requirements, but the primary objective must be to
solve the technological problems that still remain. This work should be aimed at
simplicity and reliability.
One speciﬁc issue that needs more research is the how a tensegrity structure can
be stiﬀened during deployment. The load bearing capacity during deployment is
generally nonexistent until the last cable has been prestressed. This mobility or
looseness is especially undesirable in the case of masts, which often support other
structures, such as solar panels and therefore must deploy in a predicted manner.
Stiﬀness during deployment is also required for the numerical simulation of the
process.
As indicated by the mast analysis, the bending strength of tensegrity masts must be
improved. Skelton et al. [150] analyse planar tensegrity structures that are eﬃcient in
bending. Threedimensional tensegrity structures with greater eﬃciency in bending
189
CHAPTER 8. CONCLUSIONS
need to be developed to make tensegrity structures useful for applications requiring
long slender masts.
The kinematic indeterminacy of tensegrities is not always a disadvantage. Since
the shape changes with the equilibrium of the structure only a small quantity of
control energy is needed to change its conﬁguration. The use of tensegrity structures
as sensors and actuators is therefore another area of application. This has been
explored by Skelton and Sultan [150, 159].
190
Bibliography
[1] AECAble Engineering Company, Inc. http://www.aecable.com (15
November 2001).
[2] Agrawal, P. K., Anderson, M. S., and Card, M. F. Preliminary
design of large reﬂectors with ﬂat facets. IEEE Transactions on Antennas and
Propagation AP29, 4 (1981), 688–694.
[3] Arianespace. “Ariane 4 ASAP” and “Ariane 5 payloads: ASAP”.
http://www.arianespace.com (21 June 2001).
[4] Baier, H., Datashvili, L., Gogava, Z., Medzmariashvili, E., and
Montuori, V. Building blocks of large deployable precision membrane re
ﬂectors. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dy
namics, and Materials Conference and Exhibit (Seattle, WA, USA, 16–19 April
2001). AIAA 20011478.
[5] Barnes, M. R. Applications of dynamic relaxation to the design of cable,
membrane and pneumatic structures. In Proceedings of Second International
Conference on Space Structures (Guildford, 1975), W. J. Supple, Ed.
[6] Barnes, M. R. Formﬁnding and analysis of prestressed nets and membranes.
Computers & Structures 30, 3 (1988), 685–695.
[7] Becchi, P., and Dell’Amico, S. Design and testing of a deployable,
retrievable boom for space applications. In Proceedings of the 23rd Aerospace
Mechanisms Symposium (Marshall Space Flight Center, Huntsville, AL, USA,
3–5 May 1989), pp. 101–112. NASA CP3032.
[8] Belkacem, S. Recherche de forme par relaxation dynamique des structures
r´eticul´ees spatiales autoconstraintes. PhD thesis, Universit´e Paul Sabatier de
Toulouse, Toulouse, France, 1987.
[9] Belvin, W. K. Quasistatic shape adjustment of a 15meterdiameter space
antenna. Journal of Spacecraft and Rockets 26, 3 (1989), 129–136.
[10] Boeing Satellite Systems. http://www.hughespace.com (28 November
2001).
191
BIBLIOGRAPHY
[11] Bouderbala, M., and Motro, R. Folding tensegrity systems. In IUTAM
IASS Symposium on Deployable Structures: Theory and Applications (Cam
bridge, UK, 6–9 September 1998), S. Pellegrino and S. D. Guest, Eds., Kluwer
Academic Publishers, Dordrecht, The Netherlands, 2000, pp. 27–36.
[12] Brown, K. W., and Prata, Jr., A. A design procedure for classical
oﬀset dual reﬂector antennas with circular apertures. IEEE Transactions on
Antennas and Propagation 42, 8 (1994), 1145–1153.
[13] Calladine, C. R. Buckminster Fuller’s “tensegrity” structures and Clerk
Maxwell’s rules for the construction of stiﬀ frames. International Journal of
Solids and Structures 14, 2 (1978), 161–172.
[14] Calladine, C. R., and Pellegrino, S. Authors’ closure. International
Journal of Solids and Structures 27, 4 (1991), 521–522.
[15] Calladine, C. R., and Pellegrino, S. Firstorder inﬁnitesimal mecha
nisms. International Journal of Solids and Structures 27, 4 (1991), 505–515.
[16] Calladine, C. R., and Pellegrino, S. Further remarks on ﬁrstorder
inﬁnitesimal mechanisms. International Journal of Solids and Structures 29,
17 (1992), 2119–2122.
[17] Campbell, G., Bailey, M. C., and Belvin, W. K. The development of
the 15meter hoop column deployable antenna system with ﬁnal structural and
electromagnetic performance results. Acta Astronautica 17, 1 (1988), 69–77.
[18] Chesley, B., Lutz, R., and Brodsky, R. F. Space payload design and
sizing. In Space Mission Analysis and Design, J. R. Wertz and W. J. Lar
son, Eds., third ed., Space Technology Series. Microcosm Press and Kluwer
Academic Publishers, El Segundo, CA, USA and Dordrecht, The Netherlands,
1999, ch. 9.3, pp. 255–265.
[19] Chironis, N. P. Mechanisms & Mechanical Devices Sourcebook. McGraw
Hill, New York, 1991, ch. 4, pp. 164–171.
[20] Chmielewski, A. B. et al. ARISE—mission and spacecraft description. Jet
Propulsion Laboratory Document 16330, 1998. 83 pages.
[21] Connelly, R. Rigidity and energy. Inventiones Mathematicae 66, 1 (1982),
11–33.
[22] Connelly, R. Rigidity. In Handbook of convex geometry, Vol. A, P. M.
Gruber and J. M. Wills, Eds. NorthHolland, Amsterdam, 1993, pp. 223–271.
[23] Connelly, R. Tensegrity structures: why are they stable? In Rigidity Theory
and Applications (New York, 1999), M. F. Thorpe and P. M. Duxbury, Eds.,
Plenum Press, pp. 47–54.
[24] Connelly, R., and Back, A. Mathematics and tensegrity. American
Scientist 86, 2 (1998), 142–151.
192
BIBLIOGRAPHY
[25] Connelly, R., and Servatius, H. Higherorder rigidity—what is the
proper deﬁnition? Discrete & Computational Geometry 11 (1994), 193–200.
[26] Connelly, R., and Terrell, M. Globally rigid symmetric tensegrities.
Structural Topology 21 (1995), 59–78.
[27] Cook, R. D., Malkus, D. S., and Plesha, M. E. Concepts and appli
cations of ﬁnite element analysis, third ed. John Wiley & Sons, New York,
1989.
[28] Crone, G. Large deployable reﬂector antenna for advanced mobile commu
nications. ESA/Industry Brieﬁng Meeting. 9 May, 2000.
[29] Dahlquist, G., and Bj
¨
ork, ˚A. Numerical methods. PrenticeHall, Engle
wood Cliﬀs, New Jersey, 1974.
[30] DatonLovett, A. J. An extendible member. Patent Cooperation Treaty
Application PCT/GB97/00839, Publication Number WO 97/35706. Filed 25
March 1997, Published 2 October 1997.
[31] Day, A., and Bunce, J. The analysis of hanging roofs. Arup Journal
(September 1969), 30–31.
[32] Defence Evaluation and Research Agency. Space Technology Re
search Vehicle (STRV). http://www.dera.gov.uk/html/space/strv/home.htm
(21 June 2001).
[33] Deutschen Zentrum f
¨
ur Luft und Raumfahrt. http://www.dlr.de (4
December 2001).
[34] Dietrich, R., and Pellegrino, S. Feasibility study of a deployable mesh
reﬂector: phase II. Tech. Rep. CUED/DSTRUCT/TR199, Department of
Engineering, University of Cambridge, Cambridge, UK, 20 June 2001.
[35] Duffy, J., Rooney, J., Knight, B., and Crane III, C. D. A review
of a family of selfdeploying tensegrity structures with elastic ties. The Shock
and Vibration Digest 32, 2 (2000), 100–106.
[36] El Naschie, M. S. Stress, stability and chaos in structural engineering: an
energy approach. McGrawHill, London, 1990.
[37] Emmerich, D. G. Selftensioning spherical structures: single and double
layer spheroids. International Journal of Space Structures 5, 3&4 (1990), 353–
374.
[38] Fichter, W. B. Reduction of rootmeansquare error in faceted space an
tennas. AIAA Journal 22, 11 (1984), 1679–1684.
[39] Fleeter, R. Design of lowcost spacecraft. In Space Mission Analysis and
Design, J. R. Wertz and W. J. Larson, Eds., third ed., Space Technology
Series. Microcosm Press and Kluwer Academic Publishers, El Segundo, CA,
USA and Dordrecht, The Netherlands, 1999, ch. 22, pp. 853–882.
193
BIBLIOGRAPHY
[40] Freeland, R. E. Survey of deployable antenna concepts. In Large Space
Antenna Systems Technology—1982 (Houston, TX, USA, 30 November–3 De
cember 1982), vol. 1, NASA Langley Research Center, pp. 381–421. NASA
CP2269.
[41] Freeland, R. E., and Bilyeu, G. D. INSTEP inﬂatable antenna exper
iment. In 43rd International Astronautical Federation Congress (Washington,
USA, 28 August–5 September 1992). IAF920301.
[42] Freeland, R. E., Bilyeu, G. D., and Mikulas, M. M. Inﬂatable de
ployable space structures—technological summary. In 49th International As
tronautical Congress (Melbourne, Australia, 28 September–2 October 1998).
IAF98I.5.01.
[43] Freeland, R. E., Bilyeu, G. D., and Veal, G. R. Development of
ﬂight hardware for a large, inﬂatabledeployable antenna experiment. In 46th
International Astronautical Congress (Oslo, Norway, 2–6 October 1995). IAF
95I501.
[44] Fuller, R. B. Tensileintegrity structures. United States Patent 3,063,521,
1962. Filed 31 August 1959, Granted 13 November 1962.
[45] Furuya, H. Concept of deployable tensegrity structures in space application.
International Journal of Space Structures 7, 2 (1992), 143–151.
[46] Geiger, D., Stefaniuk, A., and Chen, D. The design and construction
of two cable domes for the Korean Olympics. In Shells, Membranes and Space
Frames, Proceedings IASS Symposium, Vol. 2 (Osaka, Japan, 1986), pp. 265–
272.
[47] G
´
eradin, M., and Rixen, D. Mechanical vibrations, second ed. John Wiley
& Sons, New York, 1997.
[48] Grahn, S., and Rathsman, A. Astrid—An attempt to make the mi
crosatellite a useful tool for space science. http://www.ssc.se/ssd/papers/ (12
September 2001).
[49] Greene, W. H. Eﬀects of random member length errors on the accuracy
and internal loads of truss antennas. Journal of Spacecraft and Rockets 22, 5
(1985), 554–559.
[50] Greschik, G., Palisoc, A., Cassapakis, C., Veal, G., and Miku
las, M. M. Sensitivity study of precision pressurized membrane reﬂector
deformations. AIAA Journal 39, 2 (2001), 308–314.
[51] Guest, S. D., and Pellegrino, S. A new concept for solid surface de
ployable antennas. Acta Astronautica 38, 2 (1996), 103–113.
[52] Hachkowski, M. R., and Peterson, L. D. A comparative study of
the precision of deployable spacecraft structures. Tech. Rep. CUCAS9522,
194
BIBLIOGRAPHY
Center for Aerospace Structures, University of Colorado, Boulder, CO, USA,
December 1995.
[53] Hanaor, A. Engineering properties of doublelayer tensegrity grids. In Spa
tial Structures at the Turn of the Millennium. Proceedings of the International
IASS Symposium (Copenhagen, Denmark, 2–6 September 1991), T. Wester,
S. J. Medwadowski, and I. Mogensen, Eds., vol. 2, pp. 195–200.
[54] Hanaor, A. Geometrically rigid doublelayer tensegrity grids. International
Journal of Space Structures 9, 4 (1994), 227–238.
[55] Hearth, D. P., and Boyer, W. J. Flexibility of space structures makes
design shaky. Aerospace America 23, 5 (1985), 100–103.
[56] Hedgepeth, J. M. Accuracy potentials for large space antenna structures.
In 39th Annual Conference of the Society of Allied Weight Engineers, Inc. (St.
Louis, MO, USA, 12–14 May 1980). SAWE Paper No. 1375, ii+33 pages.
[57] Hedgepeth, J. M. Critical requirements for the design of large space struc
tures. NASA CR3484, November 1981.
[58] Hedgepeth, J. M. Accuracy potentials for large space antenna reﬂectors
with passive structure. Journal of Spacecraft and Rockets 19, 3 (1982), 211–
217.
[59] Hedgepeth, J. M. Inﬂuence of fabrication tolerances on the surface accuracy
of large antenna structures. AIAA Journal 20, 5 (1982), 680–686.
[60] Hedgepeth, J. M. Structures for remotely deployable precision antennas.
Astro Aerospace Corporation Technical Document, AACTN1154, 16 January
1989.
[61] Hedgepeth, J. M., and Adams, L. R. Design concepts for large reﬂector
antenna structures. NASA CR3663, 1983.
[62] Hedgepeth, J. M., Thomson, M. W., and Chae, D. Design of large
lightweight precise mesh reﬂector structures. Astro Aerospace Corporation
Technical Document, AACTN1164, 8 November 1991.
[63] Hirosawa, H. et al. Space VLBI satellite HALCA and its engineering ac
complishments. Acta Astronautica (2001). Paper IAF98Q1.01, In press.
[64] Ingber, D. E. The architecture of life. Scientiﬁc American 278, 1 (January
1998), 48–57.
[65] Iqbal, K., Pellegrino, S., and DatonLovett, A. Bistable composite
slit tubes. In IUTAMIASS Symposium on Deployable Structures: Theory and
Applications (Cambridge, UK, 6–9 September 1998), S. Pellegrino and S. D.
Guest, Eds., Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000,
pp. 153–162.
195
BIBLIOGRAPHY
[66] Itanami, T., Minomo, M., and Ohtomo, I. Fan rib type deployable mesh
antenna for satellite use. Acta Astronautica 17, 5 (1988), 561–566.
[67] Jenkins, C. H. M. Gossamer spacecraft: membrane and inﬂatable structures
technology for space applications, vol. 191 of Progress in Astronautics and
Aeronautics. AIAA, Reston, VA, USA, 2001.
[68] Jensen, F., and Pellegrino, S. Arm development—review of existing
technologies. Tech. Rep. CUED/DSTRUCT/TR198, Department of Engi
neering, University of Cambridge, Cambridge, UK, 25 June 2001.
[69] Kangwai, R. D., Guest, S. D., and Pellegrino, S. An introduction to
the analysis of symmetric structures. Computers & Structures 71, 6 (1999),
671–688.
[70] Kenner, H. Geodesic math and how to use it. University of California Press,
Berkeley, CA, USA, 1976.
[71] Knight, B., Duffy, J., Crane, C., and Rooney, J. Innova
tive deployable antenna developments using tensegrity design. In 41st
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Mate
rials Conference and Exhibit (Atlanta, GA, USA, 3–6 April 2000), pp. 984–994.
AIAA 20001481.
[72] Knight, B. F. Deployable antenna kinematics using tensegrity structure
design. PhD thesis, University of Florida, Gainesville, FL, USA, 2000.
[73] K
¨
otter, E.
¨
Uber die M¨oglichkeit, n Punkte in der Ebene oder im Raume
durch weniger als 2n −3 oder 3n −6 St¨ abe von ganz unver¨ anderlicher L¨ ange
unverschieblich miteinander zu verbinden. In Festschrift Heinrich M¨ uller
Breslau, A. Kr¨oner, Ed. Leipzig, 1912, pp. 61–80.
[74] Kuznetsov, E. N. On immobile kinematic chains and a fallacious matrix
analysis. Journal of Applied Mechanics, Transactions of the ASME 56, 1
(1989), 222–224.
[75] Kuznetsov, E. N. Discussion of “ﬁrstorder inﬁnitesimal mechanisms”. In
ternational Journal of Solids and Structures 27, 4 (1991), 517–519.
[76] Kuznetsov, E. N. Underconstrained structural systems. Springer, New York,
1991.
[77] Kwan, A. S. K. A pantographic deployable mast. PhD thesis, University of
Cambridge, Cambridge, UK, 1991.
[78] Lai, C.Y. Analysis and design of a deployable membrane reﬂector. PhD
thesis, University of Cambridge, 2001.
[79] Lai, C.Y., and Pellegrino, S. Feasibility study of a deployable mesh
reﬂector. Tech. Rep. CUED/DSTRUCT/TR186, Department of Engineering,
University of Cambridge, Cambridge, UK, 17 September 2000.
196
BIBLIOGRAPHY
[80] Lai, C.Y., and Pellegrino, S. Umbrellatype furlable reﬂector based on
tension truss concept. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference and Exhibit (Seattle, WA,
USA, 1619 April 2001). AIAA 20011481.
[81] Lai, C.Y., You, Z., and Pellegrino, S. Shape of deployable reﬂectors.
Journal of Aerospace Engineering 11, 3 (July 1998), 73–80.
[82] Lake, M. S. Launching a 25meter space telescope: are astronauts a key
to the next technically logical step after NGST? In 2001 IEEE Aerospace
Conference (Big Sky, MT, USA, 10–17 March 2001). IEEE Paper No. 2001
460.
[83] Lalvani, H. (Editor). Origins of tensegrity: views of Emmerich, Fuller and
Snelson. International Journal of Space Structures 11, 1&2 (1996), 27–55.
[84] Lee, S., and RahmatSamii, Y. Simple formulas for designing an oﬀset
multibeam parabolic reﬂector. IEEE Transactions on Antennas and Propaga
tion AP29, 3 (1981), 472–478.
[85] Linkwitz, K. Formﬁnding by the “direct approach” and pertinent strategies
for the conceptual design of prestressed and hanging structures. International
Journal of Space Structures 14, 2 (1999), 73–87.
[86] Linkwitz, K., and Schek, H.J. Einige Bemerkungen zur Berechnung von
vorgespannten Seilnetzkonstruktionen. IngenieurArchiv 40 (1971), 145–158.
In German.
[87] Love, A. W. Some highlights in reﬂector antenna development. Radio Sci
ence 11, 89 (1976), 671–684.
[88] Marks, R., and Fuller, R. B. The dymaxion world of Buckminster Fuller.
Reinhold Publications, New York, 1960.
[89] Mathworks Inc., The. MATLAB function reference—version 5.
http://www.mathworks.com, 1999.
[90] Maxwell, J. C. On the calculation of the equilibrium and stiﬀness of frames.
London, Edinburgh and Dublin Philosophical Magazine 27 (1864), 294–299.
[91] Medzmariashvili, E., Kinteraya, G., Datashvili, L., Bedukadze,
G., Siradze, N., Efremov, I., Chernyavski, A., and Kravchenko,
Y. Space experiment “reﬂector” on testing the largescale deployable
high precision oﬀset antenna reﬂector of a new generation at the or
bital station MIR. In AP2000 Millennium Conference on Antennas
& Propagation (Davos, Switzerland, 9–14 April 2000). Abstract 1508,
http://www.estec.esa.nl/ap2000/abstracts/a01508.html (7 November 2001).
[92] Meguro, A., Tsujihata, A., Hamamoto, N., and Homma, M. Technol
ogy status of the 13 m aperture deployment antenna reﬂectors for Engineering
Test Satellite VIII. Acta Astronautica 47, 2–9 (2000), 147–152.
197
BIBLIOGRAPHY
[93] Micheletti, A. On the kinematics of tensegrity towers. Oral presentation
at “Colloquium Lagrangianum—Strutture Tensegrity: Analisi e Progetti”, 6–8
May 2001, Rome, Italy.
[94] Micheletti, A. On the kinematics of tensegrity structures. Draft, 2 October
2001.
[95] Mikulas, M. M., and Thomson, M. State of the art and technology needs
for large space structures, vol. 1: New and Projected Aeronautical and Space
Systems, Design Concepts, and Loads of FlightVehicle Materials, Structures,
and Dynamics—Assessment and Future Directions. ASME, New York, 1994,
ch. 3, pp. 173–238.
[96] Mikulas Jr., M. M., Collins, T. J., and Hedgepeth, J. M. Prelimi
nary design considerations for 10–40 meterdiameter precision truss reﬂectors.
Journal of Spacecraft and Rockets 28, 4 (1991), 439–447.
[97] Miller, I., Freund, J. E., and Johnson, R. A. Miller & Freund’s
probability and statistics for engineers, sixth ed. Prentice Hall, Upper Saddle
River, New Jersey, 1987.
[98] Mitsugi, J., Ando, K., Senbokuya, Y., and Meguro, A. Deployment
analysis of large space antenna using ﬂexible multibody dynamics simulation.
Acta Astronautica 47, 1 (2000), 19–26.
[99] Mitsugi, J., and Yasaka, T. A modular approach to build a large space
antenna. In 42nd International Astronautical Congress (Montreal, Canada,
5–11 October 1991). IAF91315, 10 pages.
[100] Mitsugi, J., Yasaka, T., and Miura, K. Shape control of the tension
truss antenna. AIAA Journal 28, 2 (1990), 316–322.
[101] Miura, K. Concepts of deployable space structures. International Journal
of Space Structures 8, 1&2 (1993), 3–16.
[102] Miura, K., April 2001. Private communication.
[103] Miura, K., Furuya, H., and Suzuki, K. Variable geometry truss and its
application to deployable truss and space crane arm. Acta Astronautica 12,
7/8 (1985), 599–607.
[104] Miura, K., and Miyazaki, Y. Concept of the tension truss antenna. AIAA
Journal 28, 6 (1990), 1098–1104.
[105] Miura, K., and Pellegrino, S. Structural concepts. 1999. Draft.
[106] Miura, K., and Tanizawa, K. Tension truss antenna—concept, reality
and future. In IUTAMIASS Symposium on Deployable Structures: Theory
and Applications (Cambridge, UK, 6–9 September 1998), S. Pellegrino and
S. D. Guest, Eds., Kluwer Academic Publishers, Dordrecht, The Netherlands,
2000, pp. 291–300.
198
BIBLIOGRAPHY
[107] Miyasaki, A., Homma, M., Tsujigata, A., Nakamura, K., Yamada,
K., and Meguro, A. Design and ground veriﬁcation of large deployable
reﬂector. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dy
namics, and Materials Conference and Exhibit (Seattle, WA, USA, 1619 April
2001). AIAA 20011480.
[108] M
¨
obius, A. F. Lehrbuch der Statik, vol. 2. Leipzig, 1837.
[109] MoholyNagy, L. Von Material zu Architektur. Albert Langen Verlag,
M¨ unchen, 1929.
[110] Motro, R. Forms and forces in tensegrity systems. In Third International
Conference on Space Structures (Guildford, 1984), H. Nooshin, Ed., Elsevier,
pp. 180–185.
[111] Motro, R. Tensegrity systems and geodesic domes. International Journal
of Space Structures 5, 3&4 (1990), 341–351.
[112] Motro, R. Tensegrity systems: the state of the art. International Journal
of Space Structures 7, 2 (1992), 75–83.
[113] Motro, R. Structural morphology of tensegrity systems. International Jour
nal of Space Structures 11, 1&2 (1996), 233–240.
[114] Motro, R., Belkacem, S., and Vassart, N. Form ﬁnding numerical
methods for tensegrity systems. In Spatial, Lattice and Tension structures,
Proceedings of the IASSASCE International Symposium 1994 (Atlanta, USA,
1994), J. F. Abel, J. W. Leonard, and C. U. Penalba, Eds., ASCE, pp. 704–713.
[115] Murakami, H. Static and dynamic analyses of tensegrity structures. Part I.
Nonlinear equations of motion. International Journal of Solids and Structures
38 (2001), 3599–3613.
[116] Murakami, H., and Nishimura, Y. Initial shape ﬁnding and modal analy
ses of cyclic rightcylindrical tensegrity modules. Computers & Structures 79,
9 (2001), 891–917.
[117] Nishimura, Y. Static and dynamic analyses of tensegrity structures. PhD
thesis, University of California at San Diego, La Jolla, CA, USA, 2000.
[118] Pak, H. Y. E. Deployable tensegrity structures. Fourthyear undergradu
ate project report, 24 May 2000. Department of Engineering, University of
Cambridge, Cambridge, UK.
[119] Panetti, A. Thermal. In Space Mission Analysis and Design, J. R. Wertz
and W. J. Larson, Eds., third ed., Space Technology Series. Microcosm Press
and Kluwer Academic Publishers, El Segundo, CA, USA and Dordrecht, The
Netherlands, 1999, ch. 11.5, pp. 428–458.
[120] Paz, M. Structural dynamics: theory and computation, fourth ed. Chapman
& Hall, New York, 1997.
199
BIBLIOGRAPHY
[121] Pellegrino, S. Mechanics of kinematically indeterminate structures. PhD
thesis, University of Cambridge, Cambridge, UK, 1986.
[122] Pellegrino, S. Analysis of prestressed mechanisms. International Journal
of Solids and Structures 26, 12 (1990), 1329–1350.
[123] Pellegrino, S. A class of tensegrity domes. International Journal of Space
Structures 7, 2 (1992), 127–142.
[124] Pellegrino, S. Structural computations with the singular value decomposi
tion of the equilibrium matrix. International Journal of Solids and Structures
30, 21 (1993), 3025–3035.
[125] Pellegrino, S. Large retractable appendages in spacecraft. Journal of
Spacecraft and Rockets 32, 6 (1995), 1006–1014.
[126] Pellegrino, S., 2001. Private communication.
[127] Pellegrino, S., and Calladine, C. R. Matrix analysis of statically and
kinematically indeterminate frameworks. International Journal of Solids and
Structures 22, 4 (1986), 409–428.
[128] Pellegrino, S., Green, C., Guest, S. D., and Watt, A. SAR advanced
deployable structure. Tech. Rep. CUED/DSTRUCT/TR191, Department of
Engineering, University of Cambridge, Cambridge, UK, 18 November 2000.
[129] Pellegrino, S., and Guest, S. D., Eds. IUTAMIASS Symposium on De
ployable Structures: Theory and Applications, vol. 80 of Solid Mechanics and
Its Applications. Kluwer Academic Publishers, Dordrecht, The Netherlands,
2000.
[130] Pellegrino, S., Kukathasan, S., Tibert, G., and Watt, A. Small
satellite deployment mechanisms. Tech. Rep. CUED/DSTRUCT/TR190, De
partment of Engineering, University of Cambridge, Cambridge, UK, 9 Novem
ber 2000.
[131] Pontoppidan, K. Electrical consequences of mechanical antenna character
istics. In Workshop on Mechanical Technology for Antennas (Noordwijk, The
Netherlands, 26–28 June 1984), ESA/ESTEC, pp. 41–47. ESA SP225.
[132] Pugh, A. An introduction to tensegrity. University of California Press, Berke
ley, CA, USA, 1976.
[133] Rebouillat, S., Peng, J. C. M., Donnet, J.B., and Ryu, S.K. Car
bon ﬁber applications. In Carbon Fibers, J. Donnet, T. K. Wang, J. C. M.
Peng, and S. Rebouillat, Eds., third ed. Marcel Dekker, New York, 1998, ch. 7,
pp. 463–542.
[134] Reibaldi, G. G. Antenna mechanical technologies within ESA. In Sec
ond ESA Workshop on Mechanical Technology for Antennas (Noordwijk, The
Netherlands, 20–22 May 1986), ESA/ESTEC, pp. 5–13. ESA SP261.
200
BIBLIOGRAPHY
[135] Reibaldi, G. G., and Bernasconi, M. C. QUASAT program: the ESA
reﬂector. Acta Astronautica 15, 3 (1987), 181–187.
[136] Reynolds, T. Small satellite deployment mechanisms requirement report.
DERA Report, DERA/KIS/SPACE/CR000495/1.0, September 2000.
[137] Rimrott, F. P. J., and Fritsche, G. Fundamentals of STEM mechan
ics. In IUTAMIASS Symposium on Deployable Structures: Theory and Ap
plications (Cambridge, UK, 6–9 September 1998), S. Pellegrino and S. D.
Guest, Eds., Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000,
pp. 321–333.
[138] Roederer, A. G., and RahmatSamii, Y. Unfurable satellite antennas:
a review. Annales des T´el´ecommunications 44, 9–10 (1989), 475–488.
[139] Rogers, C. A., Stutzman, W. L., Campbell, T. G., and Hedgepeth,
J. M. Technology assessment and development of large deployable antennas.
Journal of Aerospace Engineering, ASCE 6, 1 (1993), 34–54.
[140] Rusch, W. V. T. Analysis of reﬂector antennas. Academic Press, New York,
1970.
[141] Russell, A. G. Development of a large deployable space reﬂector structure.
Proceedings of the Institution of Mechanical Engineers, Part G—Journal of
Aerospace Engineering 206, G2 (1992), 111–123.
[142] Ruze, J. Antenna tolerance theory—a review. Proceedings of the IEEE 54,
4 (April 1966), 633–640.
[143] Sarafin, T. P., Doukas, P. G., McCandless, J. R., and Britton,
W. R. Structures and mechanisms. In Space Mission Analysis and Design,
J. R. Wertz and W. J. Larson, Eds., third ed., Space Technology Series. Mi
crocosm Press and Kluwer Academic Publishers, El Segundo, CA, USA and
Dordrecht, The Netherlands, 1999, ch. 11.6, pp. 459–497.
[144] Sarafin, T. P., and Larson, W. J., Eds. Spacecraft Structures and
Mechanisms—From Concept to Launch. Space Technology Series. Microcosm
Press and Kluwer Academic Publishers, El Segundo, CA, USA and Dordrecht,
The Netherlands, 1995.
[145] Satellite News Digest. Satellite outages and failures. http://www.sat
index.com/failures/ (24 August 2001).
[146] Sch
¨
afer, W. Stand der Technik auf dem Gebiet gr¨ oßerer entfaltbarer
ParabolantennenStrukturen f¨ ur Raumﬂugger¨ ate. Zeitschrift f¨ ur Flugwis
senschaften und Weltraumforschung 4, 5 (1980), 255–267.
[147] Schek, H.J. The force density method for form ﬁnding and computation
of general networks. Computational Methods in Applied Mechanics and Engi
neering 3 (1974), 115–134.
201
BIBLIOGRAPHY
[148] Sj
¨
oberg, F. The design and manufacturing of a thermally stable sub
millimetre wave spaceborne radio telescope. http://www.ssc.se/ssd/papers/
(24 August 2001).
[149] Skelton, R. E., and He, M. Smart tensegrity structure for NESTOR.
In Smart Structures and Materials 1997—Smart Structures and Integrated
Systems (San Diego, 36 March 1997), M. E. Regelbrugge, Ed., vol. 3041,
SPIE—The International Society for Optical Engineering, pp. 780–787.
[150] Skelton, R. E., Helton, J. W., Adhikari, R., Pinaud, J.P., and
Chan, W. An introduction to the mechanics of tensegrity structures. Draft,
19 April 2001. http://www.mae.ucsd.edu/research/reskelton/ (11 November
2001).
[151] Snelson, K. D. Continuous tension, discontinuous compression structures.
United States Patent 3,169,611, 1965. Filed 14 March 1960, Granted 16 Febru
ary 1965.
[152] Southwell, R. V. An introduction to the theory of elasticity. Dover Publi
cations, New York, 1969.
[153] Specht, N. Deployment simulation of a parabolic space antenna with rigid
panels. International Journal of Space Structures 8, 1&2 (1993), 63–70.
[154] Stern, I., November 2001. Private communication.
[155] Strang, G. Introduction to applied mathematics. WellesleyCambridge Press,
Wellesley, MA, USA, 1988.
[156] Strang, G. Linear algebra and its applications, third ed. Harcourt Brace
Jovanovich College Publishers, San Diego, USA, 1988.
[157] Strock, J. D. Development of zero coeﬃcient of thermal expansion com
posite tubes for stable structures. In Design of Optical Instruments (Orlando,
FL, USA, 22–24 April 1992), vol. 1690, SPIE—The International Society for
Optical Engineering, pp. 223–230.
[158] Stutzman, W. L., and Thiele, G. A. Antenna theory and design, sec
ond ed. John Wiley & Sons, New York, 1998, pp. 342–347.
[159] Sultan, C. Modeling, design, and control of tensegrity structures with appli
cations. PhD thesis, Purdue University, West Lafayette, USA, 1999.
[160] Sultan, C., Corless, M., and Skelton, R. E. Reduced prestressability
conditions for tensegrity structures. In 40st AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference (St. Louis, MO,
USA, 12–15 April 1999).
[161] Sultan, C., and Skelton, R. E. Force and torque smart tensegrity sensor.
In Smart Structures and Materials 1998—Mathematics and Control in Smart
Structures (San Diego, California, 2–5 March 1998), V. V. Varadan, Ed.,
202
BIBLIOGRAPHY
vol. 3323, SPIE—The International Society for Optical Engineering, pp. 357–
368.
[162] Sultan, C., and Skelton, R. E. Tendon control deployment of tenseg
rity structures. In Smart Structures and Materials 1998—Mathematics and
Control in Smart Structures (San Diego, California, 2–5 March 1998), V. V.
Varadan, Ed., vol. 3323, SPIE—The International Society for Optical Engi
neering, pp. 455–466.
[163] Swedish Space Corporation. Microsatellites. http://www.ssc.se/ssd/
msat/ (12 September 2001).
[164] Swedish Space Corporation. Odin—a Swedish small satellite project
for astronomical and atmospheric research. http://www.ssc.se/ssd/ssat/
odin.html (12 September 2001).
[165] Thomson, M. W. The AstroMesh deployable reﬂector. In Proceeding of the
Fifth International Mobile Satellite Conference (IMSC ’97) (Pasadena, CA,
USA, 16–18 June 1997), pp. 393–398.
[166] Thomson, M. W. The AstroMesh deployable reﬂector. In IUTAMIASS
Symposium on Deployable Structures: Theory and Applications (Cambridge,
UK, 6–9 September 1998), S. Pellegrino and S. D. Guest, Eds., Kluwer Aca
demic Publishers, Dordrecht, The Netherlands, 2000, pp. 435–446.
[167] Thomson, M. W., Marks, G. W., and Hedgepeth, J. M. Lightweight
reﬂector for concentrating radiation. United States Patent 5,680,145. Filed 12
March 1996, Granted 21 October 1997.
[168] Tibert, A. G., and Pellegrino, S. Deployable tensegrity reﬂectors for
small satellites, 2001. Submitted to Journal of Spacecraft and Rockets.
[169] Tibert, A. G., and Pellegrino, S. Furlable reﬂector concept for small
satellites. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dy
namics, and Materials Conference and Exhibit (Seattle, WA, USA, 16–19 April
2001). AIAA 20011261.
[170] Tibert, A. G., and Pellegrino, S. Review of formﬁnding methods
for tensegrity structures, 2001. Accepted by International Journal of Space
Structures.
[171] Tibert, G. Numerical analyses of cable roof structures. Licentiate thesis,
Royal Institute of Technology, Stockholm, Sweden, 1999.
[172] TRW, Inc. TRWbuilt AstroMesh reﬂector deployed aboard Thuraya space
craft. http://www.trw.com (5 December 2000).
[173] Ugural, A. C., and Fenster, S. K. Advanced strength and applied elas
ticity, third ed. PrenticeHall, New Jersey, 1995.
203
BIBLIOGRAPHY
[174] van ’t Klooster, K., Rits, W., Pagana, E., Mantica, P. G., and
Bernasconi, M. C. An inﬂatable parabolic reﬂector antenna: its realisation
and electrical predictions. ESA Journal 14, 2 (1990), 211–216.
[175] Vassart, N. Recherche de forme et stabilit´e des syst`emes r´eticul´es spatiaux
autocontraints: Applications aux syst`emes de tens´egrit´e. PhD thesis, Univer
sit´e des Sciences et Techniques du Languedoc, Montpellier, France, 1997.
[176] Vassart, N., Laporte, R., and Motro, R. Determination of mecha
nism’s order for kinematically and statically indetermined systems. Interna
tional Journal of Solids and Structures 37, 28 (2000), 3807–3839.
[177] Vassart, N., and Motro, R. Multiparametered formﬁnding method: Ap
plication to tensegrity systems. International Journal of Space Structures 14,
2 (1999), 147–154.
[178] von Hoerner, S., and Wong, W.Y. Gravitational deformation and astig
matism of tiltable radio telescopes. IEEE Transactions on Antennas and Prop
agation AP23 (1975), 689–695.
[179] Wang, B.B. A new type of selfstressed equilibrium cablestrut system made
of reciprocal prisms. International Journal of Space Structures 11, 4 (1996),
357–362.
[180] Watt, A. M. Lightweight deployable SAR structures. MPhil thesis, Univer
sity of Cambridge, Cambridge, UK. 2000.
[181] Watt, A. M., and Pellegrino, S. Tapespring rolling hinges. In Proceed
ings of the 36th Aerospace Mechanisms Symposium (Glenn Research Center,
Cleveland, OH, USA, 15–17 May 2002), NASA.
[182] Webb, J. E. Deployable lattice column. United States Patent 3,486,279.
Filed 30 November 1967, Granted 30 December 1969.
[183] Wertz, J. R., and Larson, W. J., Eds. Space Mission Analysis and
Design, third ed. Space Technology Series. Microcosm Press and Kluwer Aca
demic Publishers, El Segundo, CA, USA and Dordrecht, The Netherlands,
1999.
[184] Wested, J. H. Eﬀect of deviation from the ideal paraboloid shape of large an
tenna reﬂectors. In Design & Construction of Large Steerable Aerials for Satel
lite Communication, Radio Astronomy, and Radar (6–8 June 1966), pp. 115–
119. IEE Conference Publication Number 21.
[185] Wu, K. C., and Lake, M. S. Multicriterion preliminary design of tetrahe
dral truss platform. Journal of Spacecraft and Rockets 33, 3 (1996), 410–415.
[186] You, Z. Deployable structures for masts and reﬂector antennas. PhD thesis,
University of Cambridge, Cambridge, UK, 1994.
204
BIBLIOGRAPHY
[187] You, Z., and Pellegrino, S. Deployable mesh reﬂector. In Spatial, Lattice
and Tension Structures, Proceedings of the IASSASCE International Sympo
sium (Atlanta, USA, 1994), J. F. Abel, J. W. Leonard, and C. U. Penalba,
Eds., ASCE, pp. 103–112.
[188] You, Z., and Pellegrino, S. Cablestiﬀened pantographic deployable
structures Part 2: mesh reﬂector. AIAA Journal 35, 8 (1997), 1348–1355.
[189] Zarghamee, M. S. On antenna tolerance theory. IEEE Transactions on
Antennas and Propagation AP15 (1967), 777–781.
[190] Zwanenburg, R. Olympus solar array structure and mechanisms. In Pro
ceedings of the Fourth European Symposium: Photovoltaic Generators in Space
(Cannes, France, 18–20 September 1984), ESASP210, ESA, pp. 411–419.
205
Appendix A
Overlap Values for Tensegrity
Masts
To construct a tensegrity mast, the overlap between stages, which gives a feasible
selfstress, must be found. This overlap η is both dependent on the total number of
stages n of the tower and on the relative rotation θ. Deﬁnitions of η and θ are found
in Chapter 4. The overlap for the mast with three struts per stage is given for up
to 50 stages for θ from −10
◦
to 29
◦
at 1
◦
intervals. These overlap values might be
useful for any one who wants to construct a tensegrity mast.
207
APPENDIX A. OVERLAP VALUES FOR TENSEGRITY MASTS
Table A.1: Nondimensional overlap versus the total number of stages n for a tenseg
rity tower with 3n struts for diﬀerent values of the rotation angle θ.
n
θ (
◦
)
−10 −9 −8 −7 −6 −5 −4 −3 −2 −1
2 .60878 .59824 .58763 .57696 .56622 .55541 .54451 .53353 .52246 .51128
3 .51817 .50810 .49802 .48793 .47784 .46772 .45758 .44741 .43721 .42696
4 .48298 .47326 .46355 .45385 .44415 .43444 .42473 .41499 .40524 .39546
5 .46483 .45532 .44584 .43636 .42689 .41742 .40794 .39845 .38895 .37942
6 .45399 .44463 .43529 .42595 .41663 .40730 .39798 .38864 .37929 .36992
7 .44693 .43766 .42841 .41918 .40995 .40072 .39149 .38226 .37301 .36374
8 .44203 .43283 .42365 .41449 .40533 .39617 .38701 .37785 .36867 .35947
9 .43849 .42934 .42021 .41109 .40198 .39288 .38377 .37465 .36553 .35638
10 .43583 .42672 .41763 .40855 .39948 .39041 .38134 .37226 .36317 .35407
11 .43378 .42470 .41564 .40659 .39755 .38851 .37947 .37042 .36136 .35229
12 .43217 .42311 .41407 .40505 .39603 .38701 .37800 .36897 .35994 .35088
13 .43087 .42184 .41282 .40381 .39481 .38582 .37682 .36781 .35879 .34976
14 .42982 .42080 .41180 .40280 .39382 .38484 .37586 .36687 .35786 .34884
15 .42895 .41994 .41095 .40197 .39300 .38403 .37506 .36608 .35709 .34809
16 .42822 .41923 .41025 .40128 .39232 .38336 .37440 .36543 .35645 .34745
17 .42761 .41862 .40965 .40069 .39174 .38279 .37384 .36488 .35591 .34692
18 .42708 .41810 .40914 .40019 .39125 .38231 .37336 .36441 .35545 .34646
19 .42663 .41766 .40871 .39976 .39082 .38189 .37295 .36401 .35505 .34607
20 .42624 .41728 .40833 .39939 .39046 .38153 .37260 .36366 .35470 .34573
21 .42591 .41694 .40800 .39907 .39014 .38122 .37229 .36335 .35440 .34544
22 .42561 .41665 .40771 .39878 .38986 .38094 .37202 .36309 .35414 .34518
23 .42535 .41639 .40746 .39853 .38961 .38070 .37178 .36285 .35391 .34495
24 .42511 .41616 .40723 .39831 .38939 .38048 .37157 .36264 .35370 .34475
25 .42491 .41596 .40703 .39811 .38920 .38029 .37138 .36246 .35352 .34457
26 .42472 .41578 .40685 .39794 .38903 .38012 .37121 .36229 .35336 .34441
27 .42455 .41561 .40669 .39778 .38887 .37997 .37106 .36214 .35321 .34426
28 .42440 .41547 .40655 .39763 .38873 .37983 .37092 .36201 .35308 .34413
29 .42427 .41533 .40641 .39751 .38860 .37970 .37080 .36188 .35296 .34401
30 .42415 .41521 .40629 .39739 .38849 .37959 .37068 .36177 .35285 .34391
31 .42403 .41510 .40619 .39728 .38838 .37948 .37058 .36167 .35275 .34381
32 .42393 .41500 .40609 .39718 .38829 .37939 .37049 .36158 .35266 .34372
33 .42384 .41491 .40600 .39709 .38820 .37930 .37040 .36150 .35258 .34364
34 .42375 .41483 .40591 .39701 .38812 .37922 .37033 .36142 .35250 .34356
35 .42367 .41475 .40584 .39694 .38804 .37915 .37025 .36135 .35243 .34350
36 .42360 .41468 .40577 .39687 .38797 .37908 .37019 .36128 .35237 .34343
37 .42353 .41461 .40570 .39680 .38791 .37902 .37013 .36122 .35231 .34337
38 .42347 .41455 .40564 .39674 .38785 .37896 .37007 .36117 .35225 .34332
39 .42341 .41449 .40559 .39669 .38780 .37891 .37002 .36112 .35220 .34327
40 .42336 .41444 .40553 .39664 .38775 .37886 .36997 .36107 .35215 .34322
41 .42331 .41439 .40548 .39659 .38770 .37881 .36992 .36102 .35211 .34318
42 .42326 .41434 .40544 .39655 .38766 .37877 .36988 .36098 .35207 .34314
43 .42322 .41430 .40540 .39650 .38762 .37873 .36984 .36094 .35203 .34310
44 .42318 .41426 .40536 .39647 .38758 .37869 .36980 .36091 .35199 .34306
45 .42314 .41422 .40532 .39643 .38754 .37866 .36977 .36087 .35196 .34303
46 .42311 .41419 .40529 .39640 .38751 .37862 .36974 .36084 .35193 .34300
47 .42307 .41416 .40525 .39636 .38748 .37859 .36971 .36081 .35190 .34297
48 .42304 .41412 .40522 .39633 .38745 .37856 .36968 .36078 .35187 .34294
49 .42301 .41410 .40519 .39630 .38742 .37854 .36965 .36075 .35185 .34292
50 .42298 .41407 .40517 .39628 .38739 .37851 .36962 .36073 .35182 .34289
208
n
θ (
◦
)
0 1 2 3 4 5 6 7 8 9
2 .50000 .48861 .47709 .46544 .45365 .44170 .42959 .41730 .40482 .39213
3 .41667 .40632 .39590 .38542 .37485 .36420 .35344 .34258 .33159 .32047
4 .38564 .37579 .36588 .35592 .34590 .33580 .32563 .31536 .30499 .29451
5 .36987 .36028 .35065 .34097 .33123 .32143 .31156 .30160 .29155 .28140
6 .36053 .35110 .34163 .33212 .32255 .31293 .30323 .29346 .28359 .27364
7 .35445 .34513 .33577 .32637 .31691 .30740 .29782 .28816 .27842 .26858
8 .35025 .34100 .33172 .32239 .31301 .30357 .29407 .28449 .27483 .26508
9 .34721 .33802 .32878 .31951 .31018 .30080 .29135 .28183 .27223 .26254
10 .34494 .33578 .32658 .31735 .30806 .29872 .28931 .27983 .27028 .26063
11 .34319 .33406 .32489 .31568 .30643 .29712 .28774 .27830 .26877 .25915
12 .34181 .33270 .32356 .31437 .30514 .29585 .28650 .27708 .26758 .25799
13 .34070 .33161 .32249 .31332 .30411 .29484 .28551 .27611 .26662 .25705
14 .33980 .33072 .32162 .31247 .30327 .29401 .28470 .27531 .26584 .25629
15 .33905 .32999 .32090 .31176 .30257 .29333 .28403 .27465 .26520 .25565
16 .33843 .32938 .32029 .31117 .30199 .29276 .28346 .27410 .26465 .25512
17 .33791 .32886 .31978 .31067 .30150 .29227 .28299 .27363 .26420 .25467
18 .33746 .32842 .31935 .31024 .30108 .29186 .28258 .27323 .26381 .25429
19 .33707 .32804 .31898 .30987 .30072 .29151 .28224 .27289 .26347 .25396
20 .33674 .32771 .31866 .30955 .30040 .29120 .28193 .27260 .26318 .25367
21 .33645 .32743 .31837 .30928 .30013 .29093 .28167 .27234 .26292 .25343
22 .33619 .32718 .31813 .30903 .29989 .29070 .28144 .27211 .26270 .25321
23 .33597 .32696 .31791 .30882 .29968 .29049 .28123 .27191 .26250 .25301
24 .33577 .32676 .31772 .30863 .29950 .29031 .28105 .27173 .26233 .25284
25 .33559 .32659 .31754 .30846 .29933 .29014 .28089 .27157 .26217 .25269
26 .33543 .32643 .31739 .30831 .29918 .28999 .28075 .27143 .26203 .25255
27 .33529 .32629 .31725 .30817 .29905 .28986 .28062 .27130 .26191 .25243
28 .33516 .32616 .31713 .30805 .29892 .28974 .28050 .27119 .26180 .25232
29 .33505 .32605 .31701 .30794 .29882 .28964 .28040 .27108 .26169 .25221
30 .33494 .32594 .31691 .30784 .29872 .28954 .28030 .27099 .26160 .25212
31 .33484 .32585 .31682 .30775 .29863 .28945 .28021 .27090 .26151 .25204
32 .33476 .32576 .31673 .30766 .29854 .28937 .28013 .27082 .26144 .25196
33 .33468 .32568 .31666 .30759 .29847 .28929 .28006 .27075 .26137 .25189
34 .33460 .32561 .31658 .30752 .29840 .28923 .27999 .27069 .26130 .25183
35 .33454 .32554 .31652 .30745 .29833 .28916 .27993 .27062 .26124 .25177
36 .33447 .32548 .31646 .30739 .29828 .28911 .27987 .27057 .26119 .25172
37 .33441 .32543 .31640 .30734 .29822 .28905 .27982 .27052 .26113 .25167
38 .33436 .32537 .31635 .30729 .29817 .28900 .27977 .27047 .26109 .25162
39 .33431 .32532 .31630 .30724 .29812 .28896 .27973 .27042 .26104 .25158
40 .33427 .32528 .31626 .30719 .29808 .28891 .27968 .27038 .26100 .25154
41 .33422 .32524 .31622 .30715 .29804 .28887 .27964 .27034 .26096 .25150
42 .33418 .32520 .31618 .30711 .29800 .28884 .27961 .27031 .26093 .25146
43 .33415 .32516 .31614 .30708 .29797 .28880 .27957 .27027 .26090 .25143
44 .33411 .32513 .31611 .30705 .29793 .28877 .27954 .27024 .26086 .25140
45 .33408 .32509 .31608 .30701 .29790 .28874 .27951 .27021 .26084 .25137
46 .33405 .32506 .31605 .30698 .29788 .28871 .27948 .27018 .26081 .25134
47 .33402 .32504 .31602 .30696 .29785 .28868 .27946 .27016 .26078 .25132
48 .33399 .32501 .31599 .30693 .29782 .28866 .27943 .27013 .26076 .25129
49 .33397 .32498 .31597 .30691 .29780 .28863 .27941 .27011 .26074 .25127
50 .33394 .32496 .31594 .30688 .29778 .28861 .27939 .27009 .26071 .25125
209
APPENDIX A. OVERLAP VALUES FOR TENSEGRITY MASTS
n
θ (
◦
)
10 11 12 13 14 15 16 17 18 19
2 .37921 .36605 .35261 .33889 .32485 .31045 .29567 .28046 .26477 .24856
3 .30920 .29777 .28615 .27434 .26231 .25003 .23748 .22464 .21146 .19790
4 .28390 .27316 .26226 .25119 .23993 .22846 .21676 .20481 .19256 .17999
5 .27113 .26074 .25020 .23951 .22864 .21758 .20631 .19480 .18302 .17095
6 .26357 .25338 .24305 .23258 .22194 .21112 .20010 .18885 .17735 .16556
7 .25864 .24858 .23839 .22806 .21757 .20690 .19603 .18495 .17362 .16203
8 .25522 .24525 .23516 .22492 .21452 .20396 .19320 .18223 .17102 .15955
9 .25274 .24283 .23280 .22263 .21231 .20181 .19113 .18024 .16911 .15773
10 .25088 .24101 .23103 .22091 .21063 .20019 .18956 .17873 .16767 .15635
11 .24943 .23961 .22966 .21957 .20934 .19893 .18835 .17756 .16654 .15528
12 .24830 .23850 .22857 .21852 .20831 .19794 .18738 .17663 .16565 .15442
13 .24738 .23760 .22770 .21766 .20748 .19713 .18661 .17588 .16492 .15373
14 .24663 .23687 .22698 .21697 .20680 .19647 .18597 .17526 .16433 .15315
15 .24601 .23626 .22639 .21639 .20624 .19593 .18543 .17474 .16383 .15268
16 .24549 .23575 .22590 .21590 .20577 .19547 .18499 .17431 .16341 .15227
17 .24505 .23532 .22547 .21549 .20536 .19507 .18461 .17394 .16306 .15193
18 .24468 .23496 .22511 .21514 .20502 .19474 .18428 .17363 .16275 .15164
19 .24435 .23464 .22480 .21484 .20472 .19445 .18400 .17335 .16249 .15138
20 .24407 .23436 .22453 .21457 .20447 .19420 .18375 .17311 .16226 .15116
21 .24383 .23412 .22430 .21434 .20424 .19398 .18354 .17291 .16205 .15096
22 .24361 .23391 .22409 .21414 .20404 .19379 .18335 .17272 .16188 .15079
23 .24342 .23373 .22391 .21396 .20387 .19361 .18318 .17256 .16172 .15064
24 .24325 .23356 .22375 .21380 .20371 .19346 .18304 .17241 .16158 .15050
25 .24310 .23341 .22360 .21366 .20357 .19333 .18290 .17229 .16145 .15038
26 .24297 .23328 .22347 .21353 .20345 .19320 .18278 .17217 .16134 .15027
27 .24285 .23316 .22336 .21342 .20334 .19309 .18268 .17206 .16124 .15017
28 .24274 .23305 .22325 .21331 .20323 .19300 .18258 .17197 .16114 .15008
29 .24264 .23296 .22315 .21322 .20314 .19291 .18249 .17188 .16106 .15000
30 .24255 .23287 .22307 .21314 .20306 .19282 .18241 .17181 .16098 .14993
31 .24247 .23279 .22299 .21306 .20298 .19275 .18234 .17173 .16092 .14986
32 .24239 .23271 .22292 .21299 .20291 .19268 .18227 .17167 .16085 .14980
33 .24232 .23265 .22285 .21292 .20285 .19262 .18221 .17161 .16079 .14974
34 .24226 .23258 .22279 .21286 .20279 .19256 .18215 .17155 .16074 .14969
35 .24220 .23253 .22273 .21281 .20274 .19251 .18210 .17150 .16069 .14964
36 .24215 .23247 .22268 .21276 .20269 .19246 .18205 .17146 .16064 .14959
37 .24210 .23243 .22263 .21271 .20264 .19241 .18201 .17141 .16060 .14955
38 .24205 .23238 .22259 .21267 .20260 .19237 .18197 .17137 .16056 .14951
39 .24201 .23234 .22255 .21262 .20256 .19233 .18193 .17133 .16053 .14948
40 .24197 .23230 .22251 .21259 .20252 .19230 .18190 .17130 .16049 .14944
41 .24193 .23226 .22247 .21255 .20249 .19226 .18186 .17127 .16046 .14941
42 .24190 .23223 .22244 .21252 .20245 .19223 .18183 .17124 .16043 .14938
43 .24187 .23220 .22241 .21249 .20242 .19220 .18180 .17121 .16040 .14936
44 .24184 .23217 .22238 .21246 .20240 .19217 .18177 .17118 .16038 .14933
45 .24181 .23214 .22235 .21243 .20237 .19215 .18175 .17116 .16035 .14931
46 .24178 .23211 .22233 .21241 .20234 .19212 .18173 .17113 .16033 .14928
47 .24176 .23209 .22230 .21238 .20232 .19210 .18170 .17111 .16031 .14926
48 .24173 .23207 .22228 .21236 .20230 .19208 .18168 .17109 .16029 .14924
49 .24171 .23204 .22226 .21234 .20228 .19206 .18166 .17107 .16027 .14922
50 .24169 .23202 .22224 .21232 .20226 .19204 .18164 .17105 .16025 .14921
210
n
θ (
◦
)
20 21 22 23 24 25 26 27 28 29
2 .23175 .21428 .19604 .17694 .15682 .13552 .11282 .08842 .06191 .03273
3 .18393 .16948 .15449 .13889 .12258 .10543 .08731 .06802 .04730 .02480
4 .16706 .15372 .13992 .12559 .11064 .09499 .07850 .06102 .04232 .02212
5 .15855 .14577 .13256 .11887 .10462 .08972 .07405 .05748 .03981 .02078
6 .15347 .14101 .12816 .11484 .10100 .08655 .07138 .05536 .03830 .01997
7 .15013 .13788 .12525 .11218 .09861 .08445 .06961 .05394 .03730 .01943
8 .14778 .13568 .12321 .11031 .09692 .08296 .06834 .05294 .03658 .01904
9 .14606 .13406 .12170 .10892 .09566 .08186 .06740 .05219 .03604 .01876
10 .14475 .13283 .12055 .10786 .09470 .08101 .06668 .05160 .03563 .01853
11 .14373 .13187 .11965 .10703 .09395 .08033 .06610 .05114 .03529 .01835
12 .14291 .13110 .11893 .10636 .09334 .07979 .06564 .05077 .03502 .01820
13 .14225 .13047 .11834 .10581 .09284 .07935 .06526 .05046 .03480 .01808
14 .14170 .12995 .11785 .10536 .09242 .07898 .06494 .05020 .03461 .01798
15 .14125 .12952 .11744 .10498 .09207 .07866 .06467 .04997 .03445 .01789
16 .14086 .12915 .11709 .10465 .09178 .07840 .06443 .04978 .03431 .01782
17 .14053 .12883 .11680 .10438 .09152 .07817 .06423 .04962 .03419 .01775
18 .14025 .12856 .11654 .10414 .09130 .07797 .06406 .04948 .03408 .01769
19 .14001 .12833 .11632 .10393 .09111 .07779 .06391 .04935 .03399 .01764
20 .13979 .12813 .11612 .10375 .09094 .07764 .06377 .04924 .03391 .01759
21 .13960 .12795 .11595 .10358 .09079 .07750 .06365 .04914 .03383 .01755
22 .13944 .12779 .11580 .10344 .09065 .07738 .06354 .04905 .03376 .01751
23 .13929 .12764 .11567 .10331 .09054 .07727 .06345 .04897 .03370 .01748
24 .13916 .12752 .11555 .10320 .09043 .07717 .06336 .04889 .03365 .01745
25 .13904 .12741 .11544 .10310 .09033 .07709 .06328 .04883 .03360 .01742
26 .13893 .12730 .11534 .10300 .09025 .07701 .06321 .04877 .03355 .01739
27 .13884 .12721 .11525 .10292 .09017 .07693 .06315 .04871 .03351 .01737
28 .13875 .12713 .11517 .10285 .09010 .07687 .06309 .04866 .03347 .01735
29 .13867 .12705 .11510 .10278 .09003 .07681 .06303 .04862 .03344 .01733
30 .13860 .12698 .11503 .10271 .08997 .07675 .06298 .04858 .03341 .01731
31 .13854 .12692 .11497 .10265 .08992 .07670 .06294 .04854 .03338 .01729
32 .13848 .12686 .11492 .10260 .08987 .07666 .06290 .04850 .03335 .01727
33 .13842 .12681 .11486 .10255 .08982 .07661 .06286 .04847 .03332 .01726
34 .13837 .12676 .11482 .10251 .08978 .07657 .06282 .04844 .03330 .01724
35 .13832 .12671 .11477 .10246 .08974 .07654 .06279 .04841 .03327 .01723
36 .13828 .12667 .11473 .10243 .08970 .07650 .06276 .04838 .03325 .01722
37 .13824 .12663 .11470 .10239 .08967 .07647 .06273 .04835 .03323 .01720
38 .13820 .12659 .11466 .10236 .08964 .07644 .06270 .04833 .03321 .01719
39 .13817 .12656 .11463 .10232 .08961 .07641 .06268 .04831 .03320 .01718
40 .13813 .12653 .11460 .10230 .08958 .07639 .06265 .04829 .03318 .01717
41 .13810 .12650 .11457 .10227 .08955 .07636 .06263 .04827 .03316 .01716
42 .13808 .12647 .11454 .10224 .08953 .07634 .06261 .04825 .03315 .01715
43 .13805 .12645 .11452 .10222 .08951 .07632 .06259 .04823 .03314 .01714
44 .13802 .12642 .11449 .10220 .08948 .07630 .06257 .04822 .03312 .01714
45 .13800 .12640 .11447 .10218 .08946 .07628 .06255 .04820 .03311 .01713
46 .13798 .12638 .11445 .10216 .08944 .07626 .06254 .04819 .03310 .01712
47 .13796 .12636 .11443 .10214 .08943 .07624 .06252 .04817 .03309 .01711
48 .13794 .12634 .11441 .10212 .08941 .07623 .06251 .04816 .03308 .01711
49 .13792 .12632 .11440 .10210 .08939 .07621 .06249 .04815 .03307 .01710
50 .13790 .12630 .11438 .10209 .08938 .07620 .06248 .04813 .03306 .01710
211
Appendix B
Flat Cable Net for Constructing
Tensegrity Masts
D
S
V
(a) Twostage tensegrity mast
D
S
V
(b) Threestage tensegrity mast
Figure B.1: Threedimensional conﬁgurations of tensegrity masts. Saddle cable is
denoted S, diagonal cable D and vertical cable V.
213
APPENDIX B. FLAT CABLE NET FOR CONSTRUCTING TENSEGRITY MASTS
D
S
V
(a) Net 1
D
S
V
(b) Net 2
D
S
V
(c) Net 3
Figure B.2: Diﬀerent twodimensional conﬁgurations of twostage tensegrity masts.
214
D
S
V
(a) Net 1
D
S
V
(b) Net 2
D
S
V
(c) Net 3
Figure B.3: Diﬀerent twodimensional conﬁgurations of threestage tensegrity masts.
215
Appendix C
Mesh Generation Procedure for
the Tension Truss
In this appendix, the procedure used for generating the triangular mesh of the
paraboloidal cable nets is described in detail. The procedure is applicable to nets
with bases that form a regular polygon and is illustrated in Figure C.1. First, the v
sided polygon sides are divided into v triangular bays, Figure C.1(a). Each triangular
section is then subdivided into n triangles along each side, Figure C.1(b). Finally,
the horizontal triangular mesh with a predeﬁned sagtospan ratio ρ is projected
onto the paraboloidal surface giving the shape of the cable net, Figure C.1(c). In
the ﬁnal net, the number of triangles t, elements b and joints j are
t = vn
2
, (C.1)
b = v
n(1 + 3n)
2
, (C.2)
j = 1 + v
n(1 + n)
2
, (C.3)
respectively.
In the following, a description of the n subdivision of the triangular bays is given.
Given parameters are: the number of polygon sides v, the subdivision n, the radius
R, and the twodimensional sagtospan ratio ρ. Here, ρ is deﬁned as, Figure C.2:
ρ =
δ
sag
2R
0
tan (θ/2)
(C.4)
where δ
sag
is the sag of the edge cable, θ = 2π/v, and R
0
the eﬀective radius of the
net. Note that the span used in the deﬁnition, 2R
0
tan (θ/2), is diﬀerent from the
distance between the outer vertices which is 2Rtan (θ/2), usually used to deﬁne the
sagtospan ratio. This is, however, of minor importance since the threedimensional
sag of the edge cable is dependent on the curvature, i.e. the relation between focal
length and diameter of the reﬂector surface.
217
APPENDIX C. MESH GENERATION PROCEDURE FOR THE TENSION TRUSS
1
2
v
v −1
(a) v sided polygon.
n
n
(b) Triangular n subdivision.
(c) Vertical mapping to paraboloidal sur
face.
Figure C.1: Mesh generation of net.
With the sagtospan ratio known, R
0
is calculated by subtracting from R the fol
lowing lengths, Figure C.2:
∆
1
= R
1 −cos (θ/2)
cos (θ/2)
(C.5)
∆
2
=
δ
sag
cos (θ/2)
(C.6)
With Equations (C.4)–(C.6), the relation between R and R
0
is written as
R
R
0
=
1 + 2ρ tan (θ/2)
cos (θ/2)
(C.7)
The radius R is divided into n equal parts giving identical triangles in the ﬁrst n−1
rings. In the outer ring, the triangles are distorted by the sag of the edge cables.
218
R/n
R/n
R/n
δ
sag
R
0
∆
1
∆
2
θ
r
γ/n
γ/n
γ/n
Figure C.2: Triangular subdivision of a bay (n = 3).
219
APPENDIX C. MESH GENERATION PROCEDURE FOR THE TENSION TRUSS
The edge joints are equidistantly positioned on an arc with radius r and opening
angle γ, Figure C.2, which are given by
r =
δ
2
sag
+ R
2
sin
2
(θ/2)
2δ
sag
(C.8)
γ = 2 arccos
r −δ
sag
r
(C.9)
The horizontal projection of the length of the edge elements is 2r sin (γ/2n). It
should also be noted that for an odd n the actual twodimensional sag of the edge
elements will be slightly less than δ
sag
as shown in Figure C.2 where n = 3.
Note that in the present mesh generation routine the sagging edge cable only aﬀects
the outermost ring of triangles. Therefore, there is a limit on the maximum number
of rings n for a certain value of the sagtospan ratio ρ. From Figure C.2, the
maximum value for δ
sag
is
δ
max
sag
=
Rcos(θ/2)
n
(C.10)
Equations (C.4), (C.7) and (C.10) yield the maximum sagtospan ratio as
ρ
max
=
1
2(n −1) tan(θ/2)
(C.11)
220
Royal Institute of Technology
Department of Mechanics
SE100 44 Stockholm, Sweden
http://www.mech.kth.se
Deployable Tensegrity Structures for Space Applications
Doctoral Thesis by Gunnar Tibert
ERRATA
11 July 2002
The indices of the cable and strut lengths in Equations (3.2) and (3.3) on pages 35 and 36 should be changed from (i) t to c and (ii) c to s so that they read:
2 lc = 2R2 (1 − cos θ) + H 2 , 2 ls = 2R2 1 − cos θ +
(3.2a) (3.2b)
2πi v
+ H 2,
and
2 ls = 4R2 sin θ +
πi v
sin
πi 2 + lc . v
(3.3)
Equation (4.13) on page 83 should read:1 Pcr = 4 CM l (4.13)
As a result of the error in (4.13), the beginning of the second paragraph on page 84 should read: ...which yields Pcr = 1315 N using (4.13). Substituting the buckling load, ∆ = 3.65 mm and M = 13 Nm into (4.15) yielded P = 960 N. Furthermore, the end of the third sentence on page 85 should read: ...PEu = 5.28 kN or about 5.5P . Finally, the end of line 13 on page 87 should read: The buckling load of the struts, 960 N,...
1
Thanks are due to Dr. Zhong You for pointing out this error to the author.
Deployable Tensegrity Structures for Space Applications
by
Gunnar Tibert
April 2002 Technical Reports from Royal Institute of Technology Department of Mechanics SE100 44 Stockholm, Sweden
Akademisk avhandling som med tillst˚ av Kungliga Tekniska H¨gskolan i Stockand o holm framl¨gges till oﬀentlig granskning f¨r avl¨ggande av teknologie doktorsexaa o a men m˚ andagen den 15:e april kl 13.00 i Kollegiesalen, Administrationsbyggnaden, Kungliga Tekniska H¨gskolan, Valhallav¨gen 79, Stockholm. o a c Gunnar Tibert 2002
Calladine of the University of Cambridge for constructive comments on a draft paper and to Prof. in particular. JeanMarc Battini. Karin Forsell for meticulous proofreading of the manuscript. Thanks are due to Prof. o a e Anders Wiberg and those already mentioned above. UK for providing the reﬂector antenna design task and Ian Stern of Harris Corporation in Melbourne. I thank my supervisor Prof. Anders Salw´n. for their technical assistance. at the laboratory of the Department of Structural a Engineering in Stockholm. Roger Denston. Williams of Carnegie Mellon University in Pittsburgh. Anders o Olsson. First. Many grateful thanks go to Eric Pak for excellent teamwork during the design and construction of the mast models. Alistair Ross and the technicians at the workshop of the Engineering Department in Cambridge and to Olle L¨th and Daniel Hissing. Dr. Peter Andr´n. Jonatan PaulssonTralla. Tom Th¨yr¨. Anders Eriksson for his constant encouragement and guidance. Abraham Getachew. Rickard Johnson. I thank Tim Reynolds of the Defence Evaluation and Research Agency in Farnborough. Roger Ullmann. William O. Sergio Pellegrino for giving me the opportunity to work at his excellent Deployable Structures Laboratory (DSL). USA and Andrea Micheletti of Universit` a di Roma “Tor Vergata” in Rome for valuable discussions on tensegrity. to Gerard James for interesting discussions and painstaking proofreading of the manuscript. Dr. A debt of gratitude is owed to Peter Knott. and to Dr. enthusiasm. A thank goes to colleagues and former colleagues of the Department of Structural Engineering and Department of Mechanics. I also thank the members of DSL and the Structures Research Group for many helpful discussions and a iii . the Department of Structural Engineering and the Department of Mechanics at the Royal Institute of Technology between December 1999 and April 2002. encouragement and vast knowledge of structures have been invaluable throughout the course of this research.Preface The work presented in this thesis was carried out at the Department of Engineering at the University of Cambridge. I thank Prof. Anders Ansell and Dr. USA for the data on their deployable reﬂector antennas and for the interesting discussions on deployable tensegrities. His guidance. Christopher R. Anders e Ahlstr¨m.
Sangarapillai Kukathasan. Finally. Stockholm. Dr. Elizbar ‘Buba’ Kebadze. Alan Watt and Wesley Wong. Annette Fischer. Financial support from the Royal Institute of Technology and The Royal Swedish Academy of Sciences is gratefully acknowledged. Jason Lai. Guido Morgenthal. April 2002 Gunnar Tibert iv . Special thanks go to Dr. Marcus Aberle. Dr. I would like to thank my parents for their encouragement.great time in Cambridge. continuous love and support. Lin Tze Tan. Hannes Schmidt. Lars Ekstr¨m. o Dr. Andrew Lennon.
is adopted. A comparison is made between the various reﬂector antennas in terms of deployed and stowed sizes.Abstract This thesis deals with the development of deployable structures. The scheme using telescopic struts are identiﬁed as the most suitable and a preliminary design an antenna. A stateoftheart review of deployable masts and reﬂector antennas for space applications is presented. The analysis and design of deployable tensegrity masts. for applications in space. The kinematically determinate triangulated cable network is thoroughly analysed. kinematic and static methods. analysis. both to systematic errors arising from the triangular approximation of the surface and random manufacturing errors. A way to deploy the struts using selfdeployable hinges is introduced and demonstrated by four. The mast is relatively stiﬀ in the axial direction but very weak in bending. The force method.and eightstage mast models. which uses a triangulated cable network to approximate the reﬂecting surface. with a diameter of three metres. but no single method is suitable for general tensegrities. Diﬀerent schemes for deployment are investigated. Two of the statical methods seems to be identical. the tensegrity mast is compared with an existing deployable mast with respect to stiﬀness. design. and the advantages and disadvantages of each method are investigated. Numerical computations show that the antenna is stiﬀ and extremely light. v . with three struts per stage. is presented. The requirements for a deployable reﬂector antenna used on small satellites are formulated. cable net. based on the tensegrity concept. mass and accuracy. reﬂector antenna. It is concluded that several formﬁnding methods are available. is described. tensegrity. Finally. is evaluated. The achievable surface accuracy of the net. Keywords: deployable structures. for a future space mission is performed. spacecraft. mast. Several methods proposed for this step are scrutinised and classiﬁed into two groups. The underlying principles and the statical and kinematical properties of the new concept are presented. formﬁnding. A physical model is built to analyse the feasibility of the concept and to test various deployment schemes. The key step in the design of tensegrity structures is the formﬁnding analysis. A routine for the manufacturing of physical models is proposed and evaluated. A concept. for the analysis of the kinematic and static properties of large bar frameworks.
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Deployable Reﬂector Antennas . 29 Antenna Comparison . 13 Articulated Trusses . . . . . . . . . . .3.3. Tensegrity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2. . . . . . .4 Mesh Antennas . . . . . . . . . . . . . . .3 1. . . . . . . . .1 1. . . . . . . . . . . . . . . . . Scope and Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . .4 ThinWalled Tubular Booms . . Mechanics of Bar Frameworks . . . . .2 2. . . . . . . . .1 2. . . . . . . . . . . . . . .2 1. . . . . . . . . . . . . . Deployable Masts . . . .3 2. . . . . . . . . . . .1 2. . . . . . . . . . . . . . . . . . . . . .Contents Preface Abstract List of Symbols and Abbreviations 1 Introduction 1. . . . . . . . . . . . . . . . .5 Deployable Structures . . . Outline of Thesis . . . . . . . .1 2. . . 17 Solid Surface Antennas . . . . . .2. . . . . . . . . . . . . iii v xx 1 1 2 3 5 6 9 9 2 Deployable Space Structures 2. . 16 2. . . . . . . . . . . . . . . .3. . . . . . . .2 2. . . . . . . . . . . . . . . . . . . . . . . . . 26 Inﬂatable Antennas . . . . . . . . .2 Introduction . . . . . . . . . . . 30 vii . . . . . . . .3 2. 13 2.4 1. . . . . . . . . .2. . . . . . . . . .2.2. . . . . . . . 11 Telescopic Masts . . . . . . . . . . . 11 Coilable Masts . . . . . . . . . . . . . .
. 33 Kinematic FormFinding Methods . . . . . . 39 3. . . . . . . . . . . . .6 Discussion .1 4. . . . .3 Background . . . . . . . . . .2 3. . . . 67 MultiStage Tensegrity Masts . . . .3. . . . . 49 Spherical Tensegrities . . . . . . .2 33 Introduction . . . . 60 3. 65 Static and Kinematic Properties . . . . . . . . . . . . . . . .3 Analytical Solutions . . . .6 Equilibrium and Compatibility Matrices . . . . . . . . . .3 3. . . . . . . . . . . . . . . . . . .2 3.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . . . .3 Analysis Methods for Tensegrity Structures 3. . . . . . . . . . . 36 Dynamic Relaxation . . .2. . . . . . . . .3. 35 3. . . . . . . . . 57 Structural Computations . . . . . . . . . . . . . . . . . . 47 3. . . . . . . . . . . . . . . . . . . . 54 RigidBody Mechanisms . . . .4. . . . . . . . . . . . . . . . . . . . . . .2 3. . . . . . . . . . . . . . . . . . . . . . . 41 Reduced Coordinates . . . . . 65 FormFinding . . .3 Static FormFinding Methods . . . . . . . . . . . .2. . . 39 Force Density Method . . . . . . 68 viii . . .3 3. . . . . . . . . . . . . . . . . . . . . . 51 3. . . . . . . . . . . . 56 Internal Mechanisms . . . 53 3. 35 NonLinear Programming . . . .1 3. . . . . .1 3. . . .5. . . . . . . . . . . . . . .2 4. . . . . . . . . 40 Energy Method . . . . . . . . . . . . . . . .1 4. . . . . . . .5 3. .4. . . . . .2. . .3 A TwoDimensional Example . . . . . . . . . . . .4 Implementation of the Force Density Method . . . . . . . .3.1 3.4 3. . . . . . . . . . . . .4.2 3. . . . . . . . . . . . . . . . . . . . . 44 3. 54 Static and Kinematic Properties . . .3. .5 The Force Method for Analysis of Bar Frameworks . . . . . 67 4. . . . . . . .3. . . . . . .1 3. . . . . . . . . . . . . . . . . 47 Tensegrity Prisms . . . . . . . . . . . . . . . . . . 61 65 4 Deployable Tensegrity Masts 4. . . . .4 Analytical Solutions . . . . . . . . . . . . . 59 Example: Hanging Triangular Net . . . . . . .5. . .3. . . . . . . . . . . . . . . .1 3. . . . . . . . 37 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 TwoStage Tensegrity Mast . .5. . . . . . . . . . .5. . . .
. 117 ix . . . . .3 4. . . . . . . . . . . . . . . .2 5. . . . . . .6. . . .2 5. . .4. . . . . . . . . . . . . . . . .4 4. . . . .1 4. . . . . . 104 5. . . . . . . . . . . 102 Tensegrity Reﬂector Concept .3 Geometry of Parabolic Reﬂector Antennas . . . . . . . . . .2. . . 95 5. . . . . . . . .1 4. . . . . . . . . . . . . . . 100 5. . . . . . . . . . . . . . . . 86 Static Analysis . . . . .3. . . . . . . . . . . . 73 4. . . . . . . . . . . . . 98 Future STRV Missions . . . . . . .3 5. 101 5. 91 95 5 Design Prerequisites for a Deployable Reﬂector Antenna 5.3 5. . . . . .2 Strut Deployment . . . . . . . . . . .6. . . . . . . . . . . . . . . . . 115 Ground Resolution . . . . .2 Existing Concepts with Passive Structure . . . . . . . . . .4. . . . . . . . . .5. . . .2 4. . . . . . . 74 Demonstrator Masts . 78 4. . . . . . . . . . . . . . . . .1. . . . . . . .5 Manufacturing Technique . . . . . . . . . 75 4. . . .4 Reﬂector Antenna Theory . . . . . 70 Deployment . . . .1. . . . . . .4. . . . . . . .6. .1 5. .1. . . . . . .2 5. . . . . . .7 Discussion . 102 5. . . 83 Vibration Analysis . . .4. 104 Oﬀset Reﬂector . . . . . . . . . . . .4 Initial Equilibrium Element Forces . . . . . 110 Eﬀects of Random Surface Errors . . . . . . 79 Preliminary Design of Struts and Cables . . . 96 Small Satellite Odin . . . .3. . 110 5. . . . . . . . . . . . . . . .3.2 New Reﬂector Concept for Small Satellites . . . . 109 5. . . . . . .6 Structural Analysis . . . . . . . . . . . . . . . . . . . .4. . . . . . . 112 Systematic Surface Error of Faceted Paraboloids . . . . . . . . 105 Values for F/D and XA . . . . . . . . . . . . . . 113 Allowable Surface Error of Reﬂector Antennas . . . . . . . . . . . . . . . . .4 MicroSatellites Astrid1 and 2 . . . . . . . . . . . . . . . . . . . . .1 5.1 5. . . . . . . .2. . .3 AxiSymmetric Reﬂector . . .4. . . . 89 4. . . . . .1 5.1 Small Satellites and Deployable Structures . . . . . . . . . . . .1. . . . . . . . . . . . . .4 5.5. . . . . . . . 97 Space Technology Research Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . . . . . . .5 Antenna Gain . . . . . . . . . . . . . . . . . . . . . . .
.10 Combined Inﬂuence of Manufacturing Imperfections . . . . . . . . . . . . 140 Inﬂuence of Tension Tie Loading . . . .4. . 123 6. . . . . .8 6. . 150 7 Tensegrity Reﬂector Antennas 7. . . . . . . . . . . . . . . . . . . . . .9 Monte Carlo Technique for the Tension Truss . . . . . . .3 Oﬀset Conﬁgurations . .4.4. . . . . . . . . . . . . . . . 136 6. . . .1 5. . . . . .1 6. . . . . 137 Extracting the Random Surface Deviations . . . . . . .3. . . . . . . . .4. . . . .2. . . . . . . . .3. . 133 6. . .2 6. . . 123 Position of Additional Nodes .2 6. 134 6. . . . . . . . . . . . . . . . . . . . . .1 6. . . .4. . . . . . . . . . . . . . 140 Statistical Considerations . . . . . . . . .6 5. . . . . 139 Systematic Facet Surface Deviation . . . . . . . . . . . . . .4 6. . . . . . . . .1 6. . . . . . . . .1 6. . .6 6. .3 6. . . . . .2 Focal Length and Oﬀset Distance . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . .4. . .5. . . . . . . . . . . . . . . . . 126 6. . . .11 Inﬂuence of Thermal Strains . . . . . . . . . 147 6. . . .2. . . . . . .1 153 Introduction . .7 6. 126 Tension Tie Force Distribution . . . .2. . . . . . . . . 145 6. . . 137 BestFit Paraboloid Analysis . . . . .12 Achievable Reﬂector Accuracy . . . . . . . . . . 118 Selection of Materials . . .5. . . 118 5. .5 Accuracy Goals for the Present Antenna . . . . . . . . . . . .2 Static and Kinematic Properties . . .4. .5. . . . . . . . 142 Inﬂuence of Member Length Imperfections . . 118 Materials for the RF Reﬂective Surface . . . . .4. . . . . . .4. . . . . . 133 Two Symmetric Oﬀset Conﬁgurations . . . . . . . . . . . .5 6. . . . . 122 AxiSymmetric Conﬁgurations . . .4.2 Materials for the Antenna Structure . . . . . 119 121 6 Analysis of Tension Trusses 6. . . . . .3 SagtoSpan Ratio . . 153 x . . . . . . . .4 Eﬀects of Manufacturing Errors on the Reﬂector Accuracy . . . 142 Inﬂuence of Tension Tie Load Variation . 146 6. . . . . . . . . . . .4. . . . . . . . . . 143 Inﬂuence of Ring Structure Distortion . . .4. . . . . . . . . . . . . . . . . . . . . .
. . .2. . . . . . . . 189 191 207 Bibliography A Overlap Values for Tensegrity Masts xi . . . . . 164 Deployment Schemes . . . . . 153 Deployable Ring Structure . . . . .2 7. . . . . . . . . . .2. 154 7. . . . 154 Static and Kinematic Properties . . .6. . . . . . . .7. . . . . . . . . . . .5. . . . . . . . . .4 7. . . . . . . . . .4 Analysis Methods . . . . . . . . 168 Telescopic Struts . . . . . . 175 Oﬀset Reﬂectors . . . . .3 New Concept . . . .7 7. . . . . . . . . . . . . . . . . . . . . . . . . . . .2 The AstroMesh Concept . . . . . . .2 7. . . . . . . . . . . .5 Demonstration Model . . . . . . . . . . . . . . . . . . . . . . 169 7. . . . . . . . . . . . . . . . . . . . . . . 175 Stowage Considerations .1 8. . .1 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . 157 Hexagonal Tensegrity Module and Tension Trusses . . . . . . .2 8. . . . . . . .4 Design Scheme . . . . . . . . 153 7. . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . 188 Deployable Antennas . . . . . . . . . . . . . . 161 7. . . 187 Deployable Masts . . . . . . . . . . . .3 7. . . . . 166 7. . . . .6 Preliminary Design of 3 m Reﬂectors . . . . . 179 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 7. . . . . . . . . . . . 188 Further Research . . . . . . . . . .3. . . . . . . .3 Net Generation . . . . . . . . . . . . . . . . . . . 181 Discussion . . . . . . .3 8. . . . . . . . .8 Vibration Analysis . . . . 170 AxiSymmetric Reﬂectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Hinged Struts . . 160 ThreeRing AxiSymmetric Reﬂector . . 170 7. . . . 158 Minimum Separation between Front and Rear Nets . . . . .6. . . . . .3. . . . .2. . .3 7. . . . . . . . . . . . . . .4 Stiﬀened Hexagonal Module . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . 183 187 8 Conclusions 8. . . . .1 7. . . . . . . . .1 7.1 7. . . . . . . . . . . . . . . 156 7.6. .
B Flat Cable Net for Constructing Tensegrity Masts C Mesh Generation Procedure for the Tension Truss 213 217 xii .
p. p. p. p. 111 Projected blockage area. 123 Number of kinematic constraints. p. p. p. p. 40 Incidence matrix of free coordinates. 4 Number of additional bars for kinematic determinacy. p. p. 106 Diameter of parent paraboloid. p. 54 Incidence matrix of constrained coordinates. 105 Diameter of stowed antenna. p. 46 Aij = ∂lj /∂gi . p. p. 30 Diameter of oﬀset reﬂector. p. 111 Side length of equilateral triangle. 37 Vector of nodal accelerations (3j − c). 57 Force density matrix Vector of nodal displacements (3j − c). 4 Compatibility matrix (b × 3j − c). p. 37 Vector of nodal velocities (3j − c).List of Symbols and Abbreviations Roman Symbols Aa Ab a A b ban c C Cf Cs Cu D Da Dp d D Drb D d ˙ d ¨ d Aperture area. 37 xiii . 40 Aperture diameter. p. 56 Matrix of rigidbody mechanisms (mrb × 3j − c). 30 Matrix of internal mechanisms (3j − c × m). p. 45 Number of pinjointed elements (bars). p. 40 Incidence matrix. p.
60 Dimension of space. p. 54 An integer < v. p. p.de D e e e0 F f1 fop f G g g Gi g H H0 Ha h H i j jan ¯ k K l le Vector of nodal displacements compatible with e (3j − c). p. p. 3 Number of additional joints for kinematic determinacy. 32 Vector of nodal loads (3j − c). p. 111 Generalised coordinate. p. 38 Vector of bar elongations (b). p. 173 Side length of facet. 59 Focal length. 35 Height of axisymmetric reﬂector. p. p. 106 Stage overlap. p.80665 m/s2 ). p. p. 58 Vector of generalised coordinates. 37 Gain of reﬂector antenna. p. p. p. p. 5 Buckling length. 114 m Number of internal mechanisms. p. 4 xiv . 37 Length. p. p. p. p. 30 Operating frequency. 123 Fictitious stiﬀness. 104 Height of oﬀset reﬂector. 174 Matrix of geometric loads (3j − c × m). 44 Height. 54 Vector of initial bar elongations (b). 38 Stiﬀness matrix (3j − c × 3j − c). 44 Gravity (9. 35 Number of frictionless joints (nodes). 41 Element elongation. p. 104 Lowest natural frequency. p. p. p. p. 46 Equilibrium matrix (3j − c × b). p.
43 Dimension of nullspace. p. 4 Matrix of states of selfstress (b × s). 114 Nodal coordinates. 173 Rank of H. p.mrb M n n nmax N n N P p pi q Q R Ra r rg rH rR R r s S T t t tf U Number of of rigidbody mechanisms. p. p. p. p. 56 Temperature (K). p. p. p. (xi yi zi )T . p. p. p. 5 Vector of bar axial forces (b). 59 Left singular matrix (3j − c × 3j − c). p. 37 Number of nodes. 39 Diagonal matrix of force densities. 65 Maximum number of net rings for a given sagtospan ratio. 37 Translation vector (3). p. 112 Radius of gyration. p. 54 Vector of bar forces in equilibrium with f (b). 35 Force density. p. p. 5 Tension in RF reﬂective mesh. 40 Number of mast stages. 57 Rotation vector (3). p. 57 Mass matrix (3j − c × 3j − c). 35 Radius of oﬀset reﬂector. 147 Axial element force. p. 57 Number of selfstress states. p. 43 External force. 40 Radius. p. p. 55 xv . p. p. p. 105 Distance from centre of reﬂector. p. 55 Rank of R. 57 Matrix of kinematic constraints (c × 6). p. 126 Damping matrix (3j − c × 3j − c). p.
p. 138 Separation of front and rear nets. 65 Total antenna eﬃciency factor. 41 Greek Symbols α αT ∆F ∆H ∆n ∆℘ ∆ZP ∆z δ δ δrms δrms. z x xf xu Number of vertices of regular polygon (or struts per stage). p. p. 111 Aperture blockage eﬃciency factor. ZO x. p. p. 55 Coordinates in global system (parent paraboloid). p. 41 Vector of free xcoordinates. p. 158 Surface deviation in normal direction. y. p. p.z ε η ηa ηb ηr Azimuth angle. 105 Vector of xcoordinates. 45 Coeﬃcient of thermal expansion. 104 Coordinates in local system. 111 Random error eﬃciency factor. YO . p. 105 Oﬀset distance.v W X. p. p. 112 Axial rms surface deviation. 112 Translation of paraboloid apex. p. 112 Strain. 40 Vector of constrained xcoordinates. z x. p. p. p. p. p. Y. p. p. 149 Diﬀerence between ideal and bestﬁt focal lengths. p. 112 Phase error of reﬂected rays. 137 Overlap ratio of mast stages. 105 Coordinates in global system. 106 Coordinates of origin (oﬀset reﬂector). y. Z XA XO . p. p. 138 Surface deviation in axial direction. 112 xvi . 112 General displacement. p. 45 Radiometric rms surface deviation. 5 Colatitude. p. p. 35 Right singular matrix (b × b). p.
145 Standard deviation of tension tie loads. p. 106 Density Sagtospan ratio. p. 30 Ariane Structure for Auxiliary Payloads. p. p. 18 xvii . 20 Able Deployable Articulated Mast. 105 Matrix of bar ﬂexibilities (b × b). 55 Angle between x axis and xaxis (oﬀset reﬂector). p. p. p. p. p. 42 Vector of element stresses (equivalent to q).ηt θ θ0 θ∗ ϑ λ ν ξ ρ ρ Aperture taper eﬃciency factor. p. 32 σ σε σρ στ Σ φa Φ ω Ω ω Singular value. 125 Areal density. p. p. 149 Wavelength of operating frequency fop . p. p. 55 Standard deviation of length imperfections. 69 Temperature (◦ C). 35 Angle between zaxis and rim (axisymmetric reﬂector). p. 20 Advanced Radio Interferometry between Space and Earth. 111 Rotation angle. 96 Applications Technology Satellite. 16 Advanced Folding Rib Antenna. 42 Matrix of element stresses ω. p. 111 Relative error. p. 59 Element stress (equivalent to q). p. p. p. 136 Standard deviation of ring structure distortion. p. p. 38 Ellipticity of oﬀset reﬂector rim. 42 Abbreviations ACeS ADAM AFRA ARISE ASAP ATS Asia Cellular Satellite. 143 Matrix of singular values (3j − c × b). p. p. p. p. 104 Rotation angle of interior stages.
iii EnergiaGPISpace. p. 20 Hubble Space Telescope. 9 Engineering Test Satellite. p. p. p. p. p. 25 Coeﬃcient of Thermal Expansion. p. 136 Finite Element Method. 20 HingedRib Antenna. p. p. 14 Finite Element. p. 9 Inﬂatable Antenna Experiment. p. p. p. p. 11 Carbon Fibre Reinforced Plastic. p. 11 u Deployable Structures Laboratory. p. p. 99 Highly Advanced Laboratory for Communications and Astronomy. p. 23 xviii . 100 Geosynchronous Transfer Orbit. p. 23 Hoop/Column Antenna. p. 21 European Space Agency. 58 Geosynchronous Orbit. p. 23 Folding Articulated Square Truss. p. 13 Collapsible RibTensioned Surface. p. 101 CableStiﬀened Pantographic Deployable Antenna. 96 Deutschen Zentrum f¨r Luft. p. 118 Collapsible Tubular Mast.BRC CFRP CM CRTS CSPDA CTE CTM DAISY DERA DLR DSL EGS ESA ETS FAST FE FEM GEO GPS GTO HALCA HCA HRA HST IAE IR ISAS Bistable Reeled Composite. 11 Deployable Antenna Integral System.und Raumfahrt. p. p. p. 147 Institute of Space and Astronautical Science. 29 Infrared. 11 Coilable Mast. 27 Defence Evaluation and Research Agency. 96 Global Positioning System.
15 NASA Jet Propulsion Laboratory. 112 RigidRib Antenna. 18 NASA Langley Research Center. 117 Synthetic Aperture Radar. 96 Singular Value Decomposition. p. 102 Low Earth Orbit. 29 WrapRib Antenna. 16 Space Technology Research Vehicle. p. 54 Tracking and Data Relay Satellite. 95 Medium Density Fibreboard National Aeronautics and Space Administration. 29 International Space Station. p. p. 1 Radio Frequency. 17 rootmeansquare. 20 Large Deployable Reﬂector. 11 Shuttle Radar Topography Mission. p. p. 18 IEEE Frequency Bands HF 3–30 MHz xix . p. p. p. 23 Next Generation Space Telescope. p. p. p. 17 TapeSpring Rolling Ultraviolet. p. p. p. p. p. 17 rootsumofsquares. 9 National Space Development Agency of Japan. p. 100 SpringBack Antenna. 96 Solid Surface Deployable Antenna. p.ISRS ISS JPL LaRC LDR LEO MDF NASA NASDA NGST RF rms RRA rss SAR SBA SSC SSDA STEM STRM STRV SVD TDRS TSR UV WRA Inﬂatable Space Rigidised Structure. p. p. p. p. 21 Swedish Space Corporation. 28 Storable Tubular Extendible Member. p.
VHF UHF Lband Sband Cband Xband Ku band Kband Ka band Millimetre wave band 30–300 MHz 300–1000 MHz 1–2 GHz 2–4 GHz 4–8 GHz 8–12 GHz 12–18 GHz 18–27 GHz 27–40 GHz 40–300 GHz xx .
An important issue in the design of all deployable space structures is the tradeoﬀ between the size of the packaged structure and its precision in the deployed state. 1957. the 25 m diameter Next Generation Space Telescope (NGST) [82]. which are considerably lower. Considerable research has been conducted over the last decade on erectable structures. Deployable structures have many potential applications both on Earth and in space. but possess the disadvantage of requiring risky inspace construction. Both aspects are usually critical to the mission performance.g. large state. compact state to a deployed. deployable structures have been used since the former Soviet Union launched its ﬁrst satellite Sputnik on October 4. Another. e. not easily recognised. deployable structures are the only practical way to construct large. but are sometimes conﬂicting requirements. In space. A well known example is the umbrella. 1 . Obvious advantages of deployable structures are savings in mass and volume.1 Deployable Structures Deployable structures are structures capable of large conﬁguration changes in an autonomous way. constantly is growing. beneﬁt is that the structure can better withstand the launch loads in the stowed conﬁguration. When required. the structure is only subjected to the orbital loads. In the foreseeable future. Erectable structures are versatile and can be compactly stowed. the capacity of launch vehicles will remain unchanged while the need to launch very large spacecraft. temporary or emergency structures have been used for a long time. extendible. In the beginning. they are deployed into their service conﬁguration. developable and unfurlable structures.95]. [82. lightweight structures for remote locations in space. these structures are used for easy storage and transportation. As the spacecraft grew bigger so did the launch vehicles. In its deployed conﬁguration. Deployable structures are sometimes known under other names like expandable. which are assembled in space by astronauts or robots. cf. In civil engineering. but not at the same rate. all spacecraft were small by virtue of the limited capacity of the launch vehicles.Chapter 1 Introduction 1. Most common is that the conﬁguration changes from a packaged. A more recent application is retractable roofs of large sports stadia. Therefore. Usually.
Even though the concept of tensegrity is more than ﬁfty years old. Increasing the forces in the elements of a tensegrity structure will increase its strength and load bearing capacity. e. INTRODUCTION 1. Geiger’s cable domes [46. adds to the above deﬁnition the notion that “as well as imparting tension to all cables. This provides the possibility to fold these members and hence the structure can be compactly stowed.” Hanaor [54] describes tensegrity structures as “internally prestressed. A more extensive investigation into the origin of tensegrity is given in reference [83]. While Snelson was the inventor. Snelson. In a tensegrity structure the struts have the role of the air and the cables that of the skin. Pugh [132] deﬁnes a tensegrity system as: “A tensegrity system is established when a set of discontinuous compressive components interacts with a set of continuous tensile components to deﬁne a stable volume in space. Ingber [64] argues that tensegrity is the fundamental architecture of life. Fuller [44] describes a tensegrity structure as “an assemblage of tension and compression components arranged in a discontinuous compression system. 2 . also by Miura and Pellegrino.” To explain the mechanical principle of tensegrity structures. Today. the state of prestress serves the purpose of stabilising the structure. this structure was not a true tensegrity structure. If the enclosed air is at higher pressure than the surrounding air it pushes outwards against the inwardspulling balloon skin.. mainly for art exhibitions. Ioganson. represents the birth of the tensegrity concept. B. 123]. but it bore a close resemblance. constructed in 1948. tensegrity structures are very interesting since the compressive elements are disjointed. Figure 1. it is generally regarded that K. was coined by R. A broader interpretation by Miura and Pellegrino [105] is “that a tensegrity structure is any structure realised from cables and struts.g. in which the cables or tendons are tensioned against a system of bars or struts”. Fuller [44] was the ﬁrst to look upon tensegrity structures from an engineering point of view. thus providing ﬁrstorder stiﬀness to its inﬁnitesimal mechanisms.2 Tensegrity The word tensegrity. Tensegritylike structures in cells have been observed in cell biology experiments.1(b). If the air pressure inside the balloon is increased. [113].CHAPTER 1. who is a sculptor. MoholyNagy [109] explains “that if the string was pulled the composition would change to another position while maintaining its equilibrium. The meaning of the word is vague and different interpretations are possible.” Referring to the work by Fuller. Fuller in his patent from 1962 [44]. Pugh [132] uses a balloon analogy. to which a state of prestress is imposed that imparts tension to all cables. The origin of tensegrity structures can be pinpointed to 1921 and a structure called Study in Balance made by the Russian constructivist K. few applications exist. In recent years. From the deployable structures point of view. which is a contraction of tensile integrity. engineers and biologists. the stresses in the skin become greater and the balloon will be harder to deform. Figure 1. the concept has received new attention from mathematicians.1(a).” A narrower interpretation.. cf. has since built numerous tensegrity structures. D. freestanding pinjointed networks.” According to the deﬁnitions above. Snelson’s XPiece structure.
M¨bius [108] was the ﬁrst to show that a general plane framework consisting of j o frictionless joints has to have at least 2j − 3 bars in order to be rigid. 90]: “In those cases where stiﬀness can be produced with a smaller number of lines. tensegrity structures can be treated in a similar manner as frameworks with pinjointed bars. About three decades later. transforms the framework into a ﬁnite mechanism. Maxwell also ano ticipates exceptions to the rule by stating [13. while a space framework needs 3j − 6.3 Mechanics of Bar Frameworks When properly prestressed. as a small disturbing force may produce a displacement inﬁnite in comparison with itself. Central to an understanding of the structural mechanics of any framework is the concepts of static and kinematic determinacy.3. MECHANICS OF BAR FRAMEWORKS (a) (b) Figure 1. M¨bius also observes that the removal of a bar from a framework with the o minimum number of bars. according to the rule. rendering the case one of a maximum or minimum value of one or more of its lines. M¨bius is aware of exceptions to this rule and he observes o that the determinant of the equilibrium equations of the nodes vanishes in those cases.” Although it was introduced 3 .1: Early tensegritylike structures: (a) Karl Ioganson’s Study in Balance and (b) Kenneth Snelson’s XPiece. 1. But he further observes that no additional degree of mobility is introduced if the length of the removed bar is either maximum or minimum [121]. Maxwell [90] rediscovers M¨bius’ rule. The stiﬀness of the frame is of an inferior order. certain conditions must be fulﬁlled.1.
CHAPTER 1.2). (1. Calladine studies a physical model of a tensegrity structure with 12 joints and 24 bars.2 with two bars and three joints of which two are fully ﬁxed (c = 4). o 3j − b − c = 0. Consider the twodimensional framework in Figure 1.2(b). Figure 1.2) where m is the number of internal mechanisms and s the number of states of selfstress. Equation (1. K¨tter introduces an analytical way of evaluating whether or not a plane assembly o with b < 2j − 3 is rigid [121]. (1. 75 years after M¨bius’ ﬁndings. yet. s = 1. The method is rather cumbersome and does not give any general statements. Both bars can now be prestressed to the same magnitude. INTRODUCTION l/2 δ l/2 l/2 l/2 (a) (b) P Figure 1. Figure 1. b the number of bars and c the number of kinematic constraints (c ≥ 6 in three dimensions) is now widely known as Maxwell’s rule. Calladine [13] went back to the paper by Maxwell [90] for an explanation of the mechanics of bar frameworks. Resolving the vertical equilibrium at the node yields δ3 δ P ≈ 4t0 + 8AE 3 .1) where j is the number of joints.2: A twobar framework in two conﬁgurations. all nodes lie along the same line. according to Maxwell’s rule. but it introduces a clear explanation of the fundamental mechanics of bar frameworks. In 1978. which should be loose with 6 degrees of freedom. the rule for the construction of rigid threedimensional frameworks. by M¨bius. The key result from his study is the new version of Maxwell’s rule. In the ﬁrst conﬁguration. In the second conﬁguration. but on the complete speciﬁcation of the geometry of the framework [127]. The values of m and s depend not only on the numbers of bars and joints. it is stiﬀ. l l 4 (1. and the assembly has no stiﬀness against vertical loads. nor even on the topology of the connections. the nodes do not lie along the same line and it is easily seen that this framework is rigid. The extended Maxwell’s rule can be illustrated by a simple example.2(a). does not by itself solve m and s of a general bar framework. K¨tter [73] presents the ﬁrst study entirely o o concerned with the static and kinematic indeterminacy of pinjointed frameworks. which hereafter will be referred to as the extended Maxwell’s rule. which includes all possible special cases: 3j − b − c = m − s. In 1912. m = 1.3) .
been considered suitable for space applications.. Thus. but the advantages and limitations of each method have not been clearly described. In many of these. that is frequently overlooked. hence. in the absence of prestress. Thus.1. the framework has zero vertical stiﬀness in the initial conﬁguration. Deployables can be used for increased power. Several formﬁnding methods for tensegrity structures have been proposed. Many of the deployable structure concepts available are very complex. In the case of tensegrity structures. the prestress stabilises the structure by providing additional stiﬀness. The ﬁrst step in the design of a tensegrity structure is to ﬁnd a prestressed conﬁguration.. Tensegrity structures have. the volume of the stowed structure must be very small to ﬁt the spacecraft bus. For a small deﬂection δ the stiﬀness is proportional to δ 2 . the size constraints for a small satellite may be completely diﬀerent. In most missions. The most challenging phase in the development of new deployable structures is certainly the study of the deploy 5 . but those concepts that have ﬂown are actually very simple and use simple deployment mechanisms. The primary aim of this work is to develop deployable tensegrity structures for space applications. Commercial satellites are in many cases quite small. Typically. 1. It is. as soon as the node is displaced. The basis for the present study is the assumption that the stowed volume of deployable structures can be reduced by using the tensegrity concept. aperture or to position sensitive instruments away from the interference caused by the satellite. developed for a certain scale. To achieve this goal requires the mastering of analysis methods and construction techniques not encountered in normal design work. Most deployable space structures are aimed for large satellites and. the mechanism is inﬁnitesimal. if the framework is prestressed the stiﬀness is proportional to the level of prestress. This step is called formﬁnding. is that deployable reﬂectors are also beneﬁcial for these satellites.4. but with the same reliability as their predecessors. but no study has considered the entire process from initial idea to working prototype. complex spacecraft. therefore. This illustrates what Maxwell means by “stiﬀness of . for a long time. which often have several inﬁnitesimal mechanisms. However. an inferior order”. some kind of deployable structure was needed to make them ﬁt into the launch vehicle compartment. This community requires structural systems that are cheaper. SCOPE AND AIMS where t0 is the prestressing force and AE the axial stiﬀness. not just a matter of scaling down an existing structure. A new era has now begun where the spacecraft production for commercial applications exceeds those of the military. but an important aspect. However.4 Scope and Aims Conventional satellite technology has for several decades been focused on a small number of large. The aim is therefore to scrutinise and classify the existing formﬁnding methods for tensegrity structures. every attempt is made to circumvent the use of deployables because of their perceived high risk and high cost. the elongation of the bars will stiﬀen the structure.
1. weight. e. The geometries of axisymmetric 6 . The deployment approach is demonstrated by four. orbit type. communication links. The method is linear and usually applicable only to assemblies which undergo small displacements. geometry. In Chapter 4. dynamics. is introduced.. etc. This requires studies of suitable hardware and also experiments with physical models. kinematic and static.e. surface accuracy. The parts of this chapter concerned with formﬁnding are reproduced from reference [170] with some minor modiﬁcations. which uses a triangulated cable network to approximate the reﬂecting surface. using selfdeployable hinges. the analysis and design of deployable tensegrity masts are described.and eightstage mast models. the tensegrity mast is compared with an existing deployable mast with respect to stiﬀness. The ﬁrst part of the chapter is concerned with various systems for deployable masts. In Chapter 3. A routine for the manufacturing of physical models is proposed and evaluated. In the second part of the chapter. Finally. an eﬃcient method for structural analysis of large bar frameworks is presented. analysis methods for tensegrity structures are presented. i. Deployable reﬂector antennas from three groups—mesh. Unless they can be translated into structural terms which directly aﬀect the design of the secondary spacecraft structures considered here. construction. loading. a mast and a reﬂector antenna. No attention is put to the loadcarrying primary structure of the spacecraft or to other issues normally related to spacecraft design. Various methods used to ﬁnd the initial equilibrium conﬁguration of tensegrity structures are described. propulsion. A concept. etc. two speciﬁc deployable space structures. strength. These methods are classiﬁed into two groups. solid surface and inﬂatable antennas—are studied in the second part. The chosen mast conﬁguration is that of Snelson [151] with three struts per stage. the current stateoftheart of deployable structures for space applications is reviewed.e.g. are to be designed. electric power. related to the secondary structures of a spacecraft. Two of the static methods are linked to each other. natural frequencies. The present research is concerned with structural aspects. is proposed. stiﬀness. and the advantages and disadvantages of each method are investigated.CHAPTER 1. deployables. The antenna systems are compared with respect to the deployed and stowed size. etc. materials. Diﬀerent schemes for deployment are investigated and a way to deploy the struts. i. In Chapter 5. guidance. In order to thoroughly evaluate proposed ideas and concepts. Two diﬀerent conﬁgurations of the mast are analysed: one with equal strut lengths and one with uniform element forces. Another aim is therefore to develop simple and reliable deployment mechanisms for tensegrities. INTRODUCTION ment process.5 Outline of Thesis In Chapter 2. the requirements for a deployable reﬂector antenna aimed for small satellites are formulated.
are highlighted. In Chapter 7. Subsequently. First. Another crucial aspect for the performance is the choice of suitable material for the diﬀerent parts of the antenna. First. OUTLINE OF THESIS and oﬀset reﬂector antennas are given. In Chapter 6. A preliminary design of a three metre diameter antenna is performed. Finally. overlap values for two. the features of the current stateoftheart deployable mesh antenna. Then.1. The eﬀects of diﬀerent types of errors on the antenna performance are reviewed along with allowable surface accuracies of existing deployable reﬂector antennas.5. Low weight and high stiﬀness are important properties. the underlying principles and the static and kinematic properties of the new concept are presented. conﬁguration details for adequate prestress distributions in the nets are determined. the achievable surface accuracy of the net. Chapter 8 concludes the study and gives some suggestions for further research. a detailed description of the new antenna concept is given. The scheme using telescopic struts is identiﬁed as being the most suitable. In Appendix A. This gives a clear indication on how accurate the manufacturing process must be for a satisfactory performance of the antenna. the various conﬁgurations of the twodimensional cable net used in the construction of tensegrity masts are presented. In Appendix C. is investigated. 7 . A good performance of the antenna requires high accuracy of the surface. a method for making the networks kinematically determinate is described. In Appendix B. A physical model is built to analyse the feasibility of the concept and to test various deployment schemes. on which the new concept partly is based. the triangulated cable network is thoroughly analysed.to 50stage tensegrity masts with equallength struts are given. both to systematic errors arising from the triangular approximation of the surface and random manufacturing errors. the routine for generating the triangular network for the reﬂector surface is described.
.
Chapter 2 Deployable Space Structures
2.1 Introduction
Many space missions have been completed, or are currently being planned, that involve spacecraft which are signiﬁcantly larger than the volume capacity on available launch vehicles. The cargo compartment of the National Aeronautics and Space Administration’s (NASA) Space Shuttle has a diameter of 4.6 m and a length of 18.3 m and the Ariane5 launcher of Arianespace and the European Space Agency (ESA) has a diameter of 4.56 m and is about 11 m in length [77]. As these dimensions are not likely to change very soon, some components of the spacecraft have to be folded to ﬁt into the cargo bay of the launcher and then deployed once the spacecraft is in orbit. Another limiting factor is the maximum allowable spacecraft weight. This depends on the type of launch vehicle and destination of the satellite and are, contrary to the volume constraint, generally negotiable. Over the past three decades, a signiﬁcant amount of research has been carried out in the ﬁeld of deployable space structures. There are some diﬀerences between the existing concepts: some structures can be retracted again after they have been deployed, others rely on stored strain energy for deployment and some structures are stiﬀ during deployment. Retraction is not a necessity in space, but may be required in some cases, cf. [125]. Structures not depending on stored energy are deployed by external means, e.g. a motor. Most of the structures do not obtain full stiﬀness until fully deployed while others can immediately sustain loads. Despite the amount of research into deployable structures, several highproﬁle failures have occurred in the last two decades [187]. Failures in space are very expensive and extremely diﬃcult to correct. The most wellknown example is the Hubble Space Telescope (HST) which, beside its lens problems, experienced unexpected levels of mechanical vibration caused by the thermal loading of its solar arrays. One reason for these failures is an incomplete understanding of the behaviour of the structure. Another reason, probably more rare, is that the concept itself is poor. Miura [101], inventor of several deployable structures, emphasise that “creating a rational structural concept should be the ﬁrst step in the process of designing a structure and it
9
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
precedes the practical design step.” The robustness of the concept has to be ﬁrst proven by analytical and physical models. The actual design procedure that follows involves just a tradeoﬀ between packing eﬃciency, structural stiﬀness and precision of the deployed structure. The ﬁnal steps in the construction of a space structure involve extensive ground and, possibly, ﬂight testing, which are extremely costly but unavoidable to ensure mission success. In the aerospace industry there are three main types of deployable structures: • Masts, • Antennas, and • Solar panels. Masts are typically used for separating electronic instruments to reduce interference [95] or for supporting other structures such as solar arrays [190]. A vast number of mast concepts exist, cf. [95], whereof only a few will be presented in this chapter. All satellites need to communicate with Earth and therefore need some type of antenna. Amongst the many antenna types available, the parabolic reﬂector antenna is the most common one mainly due to its high gain, which enables high data rate transmission at low power [158]. Antennas are used not only for communication but also for Earth observation and astronomical studies. The current ﬂexible solar cell technology can produce 223 W/m2 [136], which means that the solar arrays have to be quite large to produce enough power for the ever increasing number of instruments aboard a satellite. One of the most sophisticated spacecraft is the HST, which requires 4.7 kW for its instruments. Four 2.39 × 6.06 m2 solar arrays provide this power [77]. Deployable concepts for solar panels are not within the scope of this thesis and will therefore not be reviewed. For information on solar arrays see [125]. The remaining of this chapter is concerned with the most important concepts available for deployable masts and reﬂector antennas.
2.2
Deployable Masts
Deployable masts can be divided into the following four groups [95]: • Thinwalled tubular booms, • Telescopic masts, • Coilable masts, and • Articulated trusses. A few reviews on deployable masts exist: Mikulas and Thompson [95] present structures developed in the U.S.A., Pellegrino [125] covers concepts that also are retractable, and Jensen and Pellegrino [68] present the most recent and extensive one.
10
2.2. DEPLOYABLE MASTS
2.2.1
ThinWalled Tubular Booms
Thinwalled tubular booms are probably the earliest types of deployable and retractable structures. They make use of the elastic deformability of thinwalled shells. Typical materials for thinwalled booms are stainless steel, CopperBeryllium and Carbon Fibre Reinforced Plastic (CFRP) [68, 125]. Storable Tubular Extendable Member The Storable Tubular Extendible Member (STEM) was invented in Canada in the early 1960s [137]. It is an extension of the principle used for coilable selfstraightening steel tape measure. While the tape measure only subtends a small angle of the complete cylinder formed by its radius of curvature, the STEM covers more than 360◦ , Figure 2.1(a). The choice of overlap depends on several factors but is at least 50◦ [125]. A STEM is quite stiﬀ axially and in bending, but because of the open tubular section it has low torsional stiﬀness. Increased torsional stiﬀness can be provided if there is suﬃcient friction in the overlap region. The STEM is rolled up and ﬂattened onto a drum within a cassette for stowage. It is deployed by rotating the drum whereby the stored elastic energy automatically brings it back into the tubular conﬁguration. The ploy region, where it goes from ﬂat to tubular crosssection, is contained with the stowage cassette. An extension of the STEM is the biSTEM with two identical strips placed one inside the other, Figure 2.1(b). A biSTEM has higher bending stiﬀness and better mechanical damping behaviour. Other advantages of a biSTEM over a STEM of similar stiﬀness and length are shorter drums and ploy length, which give a more compact stowage cassette [125]. One version of the biSTEM has interlocking STEMs, Figure 2.1(c), which increase the torsional stiﬀness. A recently developed tubular boom is the Bistable Reeled Composite (BRC) [65], which looks exactly like a STEM but is stable in both the stowed and deployed conﬁguration. This means that the stowage cassette can be made smaller and lighter. In addition, the retraction mechanism can be simpliﬁed. Collapsible Tubular Mast The Collapsible Tubular Mast (CTM) is made from two STEMs bonded at the edges, which create a boom with higher torsional stiﬀness than the STEM. CTMs are rolled up in a similar way as STEMs. In the deployed conﬁguration, the tube is unstressed and has a lenticular crosssection [125]. A 14 m long CTM to be used for solar sails was recently developed by the German Aerospace Centre (DLR). It is made of CFRP and weighs only 0.1 kg/m [33].
2.2.2
Telescopic Masts
Telescopic masts normally consist of a series of concentric, thinwalled cylindrical tubes that are nested inside one another. Limiting factors in terms of length are
11
CHAPTER 2. DEPLOYABLE SPACE STRUCTURES
(a)
(b)
(c)
Figure 2.1: Tubular booms: (a) STEM, (b) biSTEM, and (c) interlocking biSTEM.
Figure 2.2: Collapsible tubular mast (Courtesy of DLR).
tube thickness and overlap length. Typical thickness of aluminium alloy and CFRP is 0.5 mm. Telescopic masts can be deployed either sequentially using a spindleandnut technique, Figure 2.3(a), or synchronously, Figure 2.3(b), by combining the previous technique with cables and pulleys. A 40 m long telescopic mast of CFRP has been developed by the company Dornier [77]. The mast consists of 18 segments of about 0.5 m in diameter and has a stowed length of 3 m. A 2.4 m long synchronously deployed telescopic mast consisting of seven tubes has been developed for the Tethered Satellite [125].
12
2. structural eﬃciency and precision than the previous concepts. The mast is stowed by coiling the longerons.2. which is about two mast diameters higher than the retracted mast length. They are capable of higher stiﬀness. 2.4. R. For longer masts.3 Coilable Masts In 1967. i. 2. the tip of the mast rotates during deployment. where a section of the mast goes from stowed to deployed conﬁguration. As a result. which pays out to control the deployment and reels in to retract the mast.4. The ﬁrst method relies on the stored elastic energy in the coiled longerons and the rate is controlled by a lanyard. The bracing consists of members perpendicular to the longitudinal ones and diagonal members.3: Deployment of a telescopic mast: (a) sequential and (b) synchronous (Reproduced from [125]). the motor driven method is used whereby the mast is stowed inside a special canister.4 Articulated Trusses Articulated trusses are widely used for space applications and are available in many diﬀerent conﬁgurations. Figure 2. Figure 2. the retracted length is about 2–3% of the deployed length.75 m have been constructed and the practical limit is considered to be about 1 m [95]. this method is only suitable for shorter masts. One thing that popular truss masts have 13 . an axial cable attached to the tip of the mast. It can be deployed in two diﬀerent ways: self or motor driven extension. the Coilable Mast (CM) was invented by H. typically less than 3 m [125]. contains a motor driven rotating nut.2.2. Another advantage with the canister is that it is the nut rather than the tip of the mast that rotates. A CM is very eﬃcient from a stowage viewpoint. The transition zone. DEPLOYABLE MASTS Figure 2. A coilable mast normally consists of three longitudinal elements braced at regular intervals. The canister. is contained within the canister so the part leaving the canister has full strength. Masts with diameters up to 0.e. As the stiﬀness of the mast is lower during deployment than in the deployed conﬁguration. Mauch of the Astro Research Corporation [182].
As for CMs. each deployed bay has full stiﬀness. but it is the author’s opinion that most of them have fallen into disfavour partly because of their changing diameters.CHAPTER 2.). Hence. e.g. During folding.4: Deployment of a coilable mast from a canister by a rotating nut (Courtesy of AECAble Engineering Company. Figure 2. Inc. Figure 2. DEPLOYABLE SPACE STRUCTURES Figure 2. The mast has revolute hinges along the longerons with axes parallel to the sides of the square bays and two pairs of diagonal bracing cables on each face of the bays [105]. the Variable Geometry Truss [103] and the CableStiﬀened Pantographic Mast [77. half of the bracing cables become slack as the bows bend.5(b) and (c). The strain energy stored in the bows actuates the deployment of one bay of the mast.5(a). the 14 . The cables are prestressed by four lateral bows. 186]. in common is a constant diameter during deployment. A large number of other ingenious truss concepts have been invented. Folding Articulated Square Truss Mast The Folding Articulated Square Truss (FAST) mast was developed by AECAble Engineering Company.
6% of the deployed length [1]. The canister length is 2. Figure 2.09 m in diameter and 34.5: Principle of the FAST mast (Courtesy of S.). Pellegrino). 15 .75 m in length. support the solar arrays on the International Space Station (ISS). Figure 2.3 m or about 6.2. retracted mast and transition zone are enclosed by a canister.2. DEPLOYABLE MASTS Deployed (a) Transition Force to retract (b) (c) Folded Figure 2. Inc.6.6: FAST mast for the ISS (Courtesy of AECAble Engineering Company. Eight FAST masts. each 1.
42 m and the length of the canister is 2. Pellegrino [125]. Like in CMs. and Hachkowski and Peterson [52]. The reviews by Freeland and Mikulas and Thomson cover only concepts developed in USA.7(a). Figure 2. ADAM uses no bows and only one pair of cross bracing cables on each face of a bay.3 Deployable Reﬂector Antennas Several reviews of deployable reﬂector antennas are available: Freeland [40]. Compared to the FAST mast. Mikulas and Thomson [95]. Roederer and RahmatSamii [138]. Figure 2. Pellegrino focuses on concepts. Able Deployable Articulated Mast AECAble Engineering Company recently developed the Able Deployable Articulated Mast (ADAM) for applications requiring very long and stiﬀ masts. while Roederer and RahmatSamii also include European ones. Figure 2.).92 m or 4. The STRM ADAM. Rogers et al.7: ADAM deployment sequence (Courtesy of AECAble Engineering Company. which maps the Earth.CHAPTER 2. has a diameter of 1. DEPLOYABLE SPACE STRUCTURES (a) (b) Figure 2.8% of full mast length [1]. Special latches on the diagonal cables stop the deployment and stiﬀen each bay. A 60 m long ADAM is used in the Shuttle Radar Topography Mission (STRM). Spherical hinges are ﬁtted at the ends of the longitudinal members and the rigid lateral square rotates almost 90◦ during deployment. Its stowed height is 1. Inc.7(b). 2. whose prime function is to separate two radar antennas. from 16 . a canister/nut technique is used for deployment. [139].7.12 m and consists of 87 bays.
the antenna concepts will be summarised and compared in terms of packaging eﬃciency. The ribs are hinged only at the hub and they simply fold towards the feed structure.3. the most important deployable mesh antenna concepts are considered. Deployable mesh antennas are available in many conﬁgurations which diﬀer in the way the mesh is supported. RigidRib Antenna Harris Corporation developed the RigidRib Antenna (RRA) for the NASA Tracking and Data Relay Satellite 1 (TDRS) and the NASA Galileo mission to Jupiter. In the following three sections. The latest one by Hachkowski and Peterson is concerned with the precision of the antennas and includes 50 structures. 1 17 . which can achieve higher surface accuracy. 2. Europe and Japan. areal mass. is therefore necessary for a more complete background. In recent years a small number of new. Umbrellatype designs are still prevalent but several new concepts. In each category. in the previous reviews. In the following.3. and • Inﬂatable antennas. very important concepts have been developed which are not included.8(a). are now being developed. including only the most important structures. the most important concepts from each category will be presented. or only mentioned brieﬂy. Basically. tubular CFRP ribs attached to a central hub and an RF reﬂective mesh between the ribs. Finally. • Solid surface antennas. In its The rigid rib antennas by Harris were used on TDRSA through G and then replaced by the springback antenna for TDRSH through J. surface accuracy and natural frequency. The antennas constructed for the TDRS and Galileo are almost identical with a diameter of 5 m. not only reﬂectors. which gives a stowed height about the same as the antenna radius. some of which have ﬂown while others never left the ground. The RRA is an umbrellatype antenna with 16 parabolic. several diﬀerent concepts have been proposed but only a limited number have proven to be viable and even fewer have actually ﬂown. it can reﬂect radio frequency (RF) waves up to about 40 GHz [95]. DEPLOYABLE REFLECTOR ANTENNAS USA.2. The most common antenna design is an inverted umbrella with curved ribs emanating from a hub and the mesh suspended between the ribs. there are three diﬀerent types of deployable antennas: • Mesh antennas. Although the mesh is discontinuous. A new review. which are retractable.1 Mesh Antennas The most common type of deployable antennas is the mesh antenna with a reﬂective surface composed of a knitted lightweight metallic mesh. Figure 2.
146]. In vacuum.1 m diameter WRA was launched in 1974 with the Applications Technology Satellite 6 (ATS6). failed to deploy as commanded on April 11. Figure 2. which were stowed in a 2. stowed conﬁguration. The failure is believed to be caused by very high friction between restraint pins and their receptacles.45 m. the ribs are rotated on vertical hinge pins and then tangentially wrapped around the hub. 138.CHAPTER 2.9(b). including the ribs. 1989. The entire antenna structure.7 m. be accomplished using an antenna with lower gain and various enhancements to the communication link. 1991. parabolic CFRP ribs of lenticular crosssection. In the stowed position. A 9. Note that it is also possible to retract the antenna in orbit by reversing the drive. 18 .0 m diameter hub with a height of 0. reﬂector surface.9(b).9 and 2. WrapRib Antenna NASA Jet Propulsion Laboratory (JPL) and Lockheed Missiles and Space Company developed the WrapRib Antenna (WRA) in the 1970s. The mission objectives could.2 mil diameter 10 openings/inch (a) (b) Figure 2. feed and deployment mechanisms that fold and unfold the structure. 2 From measurement in ﬁgure on page 415 of [40]. and an RF reﬂective mesh. The deployment mechanism is shown in Figure 2. The whole antenna weighed about 60 kg [87. The RRA antenna onboard the Galileo spacecraft. This is an umbrellatype concept with a central hub. this sequence takes about two seconds. Springloaded doors are opened and the stored energy in the ribs causes them to unwrap into their original position. weighs 24 kg. The ribs are deployed by cutting a restraining cable placed around the hub. respectively2 . The ATS6 antenna consisted of 48 ribs. however. the diameter and height is 0. DEPLOYABLE SPACE STRUCTURES Subreflector Equipment compartment 18 GRP ribs Hub Struts Goldplated molybdenum wire 1. which was launched on October 18.8: The 5 m diameter rigid rib antenna for the Galileo mission (Courtesy of Harris Corporation).
1 m diameter ATS6 wraprib antenna (Courtesy of Lockheed Missiles and Space Company). drive Wrap ribs (b) Wraprib deployment mechanism. Figure 2./refurl.9: The wraprib reﬂector antenna. RF reflective mesh Unfurl. DEPLOYABLE REFLECTOR ANTENNAS (a) The 9. 19 .2. drive Furled wrap ribs/ RF reflective mesh pack b1 Ri Tape spool Rib 1 Rib Tape 1 Unfurl.3./refurl.
the cable network supports and shapes the RF reﬂective mesh surface [40]. In addition to its structural contributions. an outward extending preload segment is deployed at the bottom of the column. A 15 m diameter HCA. but with lower mass and smaller stowed diameter and height [154]. the hoop deploys by means of motors mounted in eight of the 24 hoop joints.11(a) and (b). 2000. The total weight of the antenna and boom is 127 kg. has been built and tested on ground. In the stowed conﬁguration. with the same deployed diameter as the ACeS HRA. A central column and a largediameter hoop are the compression members. the antenna ﬁts inside a cylinder of 2.7 m in height and 0. The total weight of the 20 . In order to provide a smaller package diameter. The antenna deploys in three principal steps: the column deploys simultaneously from both top and bottom. HingedRib Antenna Harris Corporation developed the HingedRib Antenna (HRA) for the Asia Cellular Satellite (ACeS) system. Figure 2. the ribs are straight and the parabolic mesh surface is shaped by standoﬀ elements along the ribs. called Advanced Folding Rib Antenna (AFRA). Figure 2. DEPLOYABLE SPACE STRUCTURES Figure 2.5 m. launched February 13. which produce a smaller height of the stowed package.5 m. In this concept.10.CHAPTER 2.9 m in diameter. which pretensions all cables and the mesh. Hoop/Column Antenna NASA Langley Research Center (LaRC) and Harris Corporation developed the Hoop/Column Antenna (HCA) around 1980. with a height of 9. The stowed diameter and height of each antenna are 0. Figure 2. which maintain pretension in a cable network. to demonstrate the feasibility of a large space antenna. the ribs are hinged along their length. The concept is a simple tension and compression preloaded structure.10: The 12 m diameter hingedrib antenna for the ACeS system (Courtesy of Harris Corporation).86 and 4. Harris is currently working on a reﬁned version of the HRA. respectively. Two 12 m diameter HRAs are mounted on the ACeS Garuda1 satellite.11(c)–(e). Finally.
is 35 kg. was developed by the Hughes Space and Communication Company 3 . The stowed diameter and height of the antenna are 0. The EGS antenna consists of a circular pantograph ring and radial tensioned membrane ribs connecting to a central hub. The total weight. The SpringBack Antenna (SBA) consists of a thin graphite mesh dish with an integral lattice of ribs and a stiﬀening hoop along the rim. but only tested on the ground [4.4 m was tested in space on the Russian orbital station MIR. A 13.3. including both the mechanical and electrical systems. EGS Antenna The RussianGeorgian company EnergiaGPISpace (EGS) has developed a new concept aimed at deployable reﬂector antennas in the diameter range 5–25 m.0 m. The weight of the antenna. 21 . The antenna is elastically folded like a taco shell and 3 Now Boeing Satellite Systems.11: Deployment sequence of the hoop/column antenna. Reproduced from [146]. including the instrumentation and feed. SpringBack Antenna A totally diﬀerent antenna concept.12.6 m by 6. respectively. 91]. Inc.2. DEPLOYABLE REFLECTOR ANTENNAS (a) (b) (c) (d) (e) Figure 2. antenna structure alone is 291 kg. is 410 kg [9].5 m diameter antenna has also been built.6 and 1. without moveable connections. An elliptical reﬂector antenna with dimensions 5. Figure 2.
25 m elliptical shape and weighed 20 kg. Tension Truss Antenna The Tension Truss concept was developed by Miura in 1986 [104] to meet demands for high precision large deployable reﬂectors.). respectively. Figure 2. In the stowed conﬁguration. the reﬂector deploys by cutting the tie cables. 22 . Photo taken from the orbital station MIR on July 28. Outofplane forces. the two SBAs are rolled together into a 4. The stowed volume of this antenna is quite large compared to other antennas and the stowed height is about the same as the reﬂector diameter. which releases its stored elastic strain energy. the novelty of the tension truss concept was to use ﬂexible members for the sides of the triangles so that the whole assembly could be easily folded. NASA’s next generation TDRS. the geodesic dome. 105]. Each of the two SBAs had a 6.8 m by 5. the reﬂector is divided into triangular facets instead of gores as in an umbrellatype reﬂector. 1996. held in that conﬁgurations by tie cables along the rim. was the ﬁrst to use the SBA.5 and 3. The MSAT1 satellite. launched April 20.9 m high truncated cone on top of the spacecraft. In this concept. The top and bottom diameters of the cone are 1. DEPLOYABLE SPACE STRUCTURES Figure 2.0 m. The idea of approximating bowlshaped surfaces by polygons was not new.6 m diameter EGS antenna in space. 1999 (Courtesy of EnergiaGPISpace. and had been previously applied in space antenna applications. Ltd. In orbit. cf.6 m diameter SBAs [10.CHAPTER 2. e. the 5 m diameter RRAs are replaced by 4.12: A 5.g.13. In a more recent application. However.
the mass of the mesh and cables is about 11 kg and that of the six deployable masts is about 100 kg [63]. CMs could not be used because of the high compressive forces. is about 170 kg [92].2.g. The ETSVIII antenna is based on a modular approach with 4. including supporting booms and deployment mechanics.8 m diameter hexagonal tension truss modules.3. Miura. The stowed size is 1 m in diameter and 4 m in height.14. The total weight of a 14module antenna. Figure 2. by decreasing the size of the triangles. without increasing the number of supports. but the total dimension of the antenna is 19. The main diﬀerence between them is the supporting structure for the triangular network. The National Space Development Agency of Japan (NASDA) developed another antenna for their Engineering Test Satellite VIII (ETSVIII). A reﬂector with an eﬀective aperture diameter of 8 m was developed by the Institute of Space and Astronautical Science (ISAS) and ﬂown on the Highly Advanced Laboratory for Communications and Astronomy (HALCA) satellite. e. In Japan. springs.2 × 16. Maximum diameter of the antenna is 10 m and the total mass is 246 kg. would be used to prestress the triangular network.7 m2 . the tension truss is deployed by six truss masts4 .13: Folded springback antenna (Courtesy of Hughes Space and Communications Company). 4 23 . A tension truss The exact mast type is not known to the author. but according to K. From the main bus. The triangular prestressed network is the support for the reﬂective mesh. A main advantage of the tension truss antenna over the umbrellatypes is that the surface accuracy can be increased. DEPLOYABLE REFLECTOR ANTENNAS Figure 2. two diﬀerent types of antennas using the tension truss concept have been developed. [102]. A 14module antenna has an aperture diameter of 13 m.
CHAPTER 2. In the stowed conﬁguration. This assembly is prestressed by tension ties connected to mirroring nodes of the two networks.15(c). antenna developed in USA will be described next. Astro Aerospace Corporation5 started developing what is now the current stateoftheart deployable reﬂector antenna—the AstroMesh. AstroMesh Antenna In 1990.3 and 3. 24 . the diameter and height are 1. has a diameter of 12.15(a). which was launched on December 5.14: Tension truss antenna for HALCA (Courtesy of ISAS). The RF reﬂective mesh is attached to the backside of the front net.15(b) and (c). This antenna. the AstroMesh is based on the tension truss concept and its components are shown in Figure 2. respectively [172]. Two identical paraboloidal triangular networks are attached to a deployable ring truss. Deployment synchronisation is achieved through special joints at the truss connections where only three members meet [166]. which can be clearly seen in Figure 2. 2000.8 m. 5 Now TRW Astro Aerospace.25 m and weighs 55 kg. As mentioned in the previous section. The latest application of the AstroMesh is onboard the telecommunication satellite Thuraya. The antenna is deployed by shortening a cable which continuously runs through the telescopic diagonal members of the ring truss. DEPLOYABLE SPACE STRUCTURES Figure 2. which is shown in Figures 2.
W. Thomson and TRW Astro Aerospace). which supports the RF reﬂective mesh. A 3.25 m diameter reﬂector for Thuraya (Courtesy of M. is 25 .16. CableStiﬀened Pantographic Antenna DSL developed the CableStiﬀened Pantographic Deployable Antenna (CSPDA). The pairs of rods are connected at their end points to form a circular pantographic structure that can be folded.3. Crucial to a successful folding is accurate positioning and manufacturing of the scissor and end joints. A double layer cable network. Figure 2. DEPLOYABLE REFLECTOR ANTENNAS Front Net Mesh Tension Ties Truss Rear Net (a) (b) (c) Figure 2. The deployable ring structure consists of three diﬀerent pairs of rods connected by scissor joints.15: The AstroMesh deployable reﬂector antenna: (a) concept. (b) and (c) the 12.5 m diameter model has twelve sides and is composed of 48 pantograph elements.2.
2 m. respectively [188]. attached to the ring structure. Most of the solid surface deployable reﬂector concepts consist of a central hub with rigid curved panels. DEPLOYABLE SPACE STRUCTURES (a) (b) (c) Figure 2. Like the AstroMesh.16: Deployment sequence of the 3. You and S. deployable high precision reﬂectors are limited to approximately 10 m in diameter [95].CHAPTER 2. arranged as radial petals. In the stowed conﬁguration. The layout of the network is chosen such that the total structure is statically determinate. 26 .3. The concepts diﬀer by the manner in which the petals fold [51]. the diameter and height are 0. Because of mechanical complexity and launch vehicle size constraints.6 and 1. or indeterminate to only a small degree.2 Solid Surface Antennas Antenna applications operating at frequencies over about 40 GHz require high surface accuracy. 2. Solid material is usually chosen for their reﬂective surface [95]. an active cable is used to deploy the ring structure. often CFRP face sheets over an aluminium honeycomb core.5 m diameter model of the cablestiﬀened pantographic deployable antenna (Courtesy of Z. Pellegrino).
18. respectively [51. The position and orientation of the hinges are determined by extensive deployment simulations to achieve good packaging and avoid interference between the panels during deployment. however. with panels nested in front of the hub. 105]. which allows the panels to fold by nesting in front of the hub. 186]. which has a stowed diameter and height of 2.15 and 1. Here. DEPLOYABLE REFLECTOR ANTENNAS Sunﬂower The ﬁrst concept with rigid panels was the Sunﬂower by TRW.4 m in diameter and 6.6 m in height [60]. seen in Figure 2. 27 . 52]. This version is.4 m. A 4. respectively [51. with a diameter of 15 m.8 m. During folding. but does not achieve a great reduction in size. An 8 m diameter engineering model has been built. Figure 2. The folding conﬁguration is similar to the DAISY.1 m. plan view (Courtesy of TRW).7 m diameter MEA model has a stowed diameter and height of 1. respectively [51.17: Sunﬂower antenna: (a) folded.9 m diameter model was built.9 and 4. also by Dornier and ESA. (b) deployed. each panel is connected to the central hub by a revolute joint.18. DAISY and MEA The second concept is the Deployable Antenna Integral System (DAISY) by ESA and Dornier.17. The third concept. provides additional stiﬀness for better surface accuracy. with a stowed diameter and height of 2.2.3. An extended version of the Sunﬂower. is MEA. each panel folds towards the hub and twists at the same time. Figure 2. The paneltopanel connecting rods compensate for the increased kinematic freedom introduced by the twoaxis joints and synchronise the motion of the panels. A 4. Figure 2.7 and 2. Each panel is connected to the hub by a joint that allows rotation about two axes and to neighbouring panels by rods with spherical joints. stows into a cylinder of 4. more complex that the original Sunﬂower. The truss structure on the back of each panel. It folds using simple revolute joints between the panels.
Each wing is further subdivided into panels which are connected by revolute joints. D.36 and 0. In this concept.18: DAISY antenna: (a) folded. Figure 2. the surface is split into wings rather than radial petals.81 m. is quite diﬀerent from the others. (a) (b) (c) (d) Figure 2. DEPLOYABLE SPACE STRUCTURES (a) (b) Figure 2. Solid Surface Deployable Antenna The fourth concept. yield 0. Guest).19: Deployment sequence of the solid surface deployable antenna: (a)–(d) (Courtesy of S.19.5 m diameter SSDA model with six wings and ﬁve panels per wing had a stowed diameter and height of 0. Estimates show that increasing the number of panels from ﬁve to seven. (b) deployed (Courtesy of Dornier GmbH).CHAPTER 2. A 1.56 and 0. 28 . This concept is extendable to any number of wings and panels and will package eﬃciently as the curved panels nest inside one another.75 m as stowed dimensions [51]. the Solid Surface Deployable Antenna (SSDA) by DSL. respectively.
1 × 0. The canopies were constructed of 6. which support the feed.3.46 m3 . Mylar is a registered trademark of E.5. I.5 µm Mylar 7 ﬁlm. Inﬂatable structures have a high deployment reliability because of their simplicity and a low number of failure modes.0 × 1.1 Pa.3. the front canopy was left transparent while the back was aluminised for reﬂectivity. The main disadvantage of inﬂatable antennas is the diﬃculty of achieving high shape accuracy for the reﬂective surface [186]. impregnated with an epoxytype resin. Figure 2. no attention was paid to space rigidised material. du Pont de Nemours and Company. which is easily achieved when the structure is exposed to direct sunlight. The resin hardens in six hours at a temperature of 110◦ C. JPL initiated the Inﬂatable Antenna Experiment (IAE). Figure 2. Three engineering models of the antenna. A smooth reﬂective surface could be obtained at a very low pressure of 2.21. with diameters 3. du Pont de Nemours and Company. no ﬂight test was performed and the programme has now been cancelled [28]. which housed the stowed inﬂatable structure. the inﬂated antenna is positioned so that its faces the Sun. As the antenna would only be used for a short time. DEPLOYABLE REFLECTOR ANTENNAS 2. It is constructed from a thin ﬂexible material which is folded prior to launch and then deployed by inﬂation. have been built. a zero gravity environment can be simulated through the use of Helium [40]. the inﬂation gas is expelled [135]. was 2. The torus and struts were made of 0. It is stiﬀened by an inﬂatable torus along the edge. the 12 m diameter oﬀset reﬂector. When in orbit. Together with the company L’Garde they developed and manufactured a 14 m diameter inﬂatable antenna with a low pressure canopy structure. I. The size of the box container. The torus and struts were both pressurised to 6. The total weight of the inﬂatable structure was 60 kg. Inﬂatable Space Rigidised Reﬂector ESA and the company Contraves developed an antenna using Inﬂatable Space Rigidised Structure (ISRS) technology for a joint ESA/NASA programme requiring a large antenna [135]. Inﬂatable Antenna Experiment To demonstrate the potential of inﬂatable structure technology for large antennas. The reﬂector structure is like a circular paraboloidal cushion with a transparent front side and a reﬂective rear side.20 shows the latest model. The structure can be made more rigid by impregnating the membrane material with a resin. An early design study showed that a 20 m diameter reﬂector would have a mass of 134 kg [135]. For testing the structures on the ground. The antenna structure is made of Kevlar 6 membranes. which slowly cures at high temperature or by the Sun’s ultraviolet (UV) rays.9 kPa. Deployment starts with 6 7 Kevlar is a registered trademark of E.3 Inﬂatable Antennas An inﬂatable antenna would give the smallest package size and potentially the lowest mass. high pressure torus and three high pressure struts.2. However.3 mm thick Kevlar. 29 . 6 and 12 m. After hardening.
Low mass is necessary because of the extremely high launch costs in the order of 10.4 Antenna Comparison A comparison between the presented antenna concepts is given in Table 2. 43].20: The 12 m diameter model of the inﬂatable space rigidised reﬂector antenna (Courtesy of A. the stowed diameter d varies between 6 and 22% of the deployed diameter D and the stowed height h varies between 5 and 93%. which includes the development of a 25 m diameter inﬂatable antenna pressurised to 2.000 USD/kg [185]. DEPLOYABLE SPACE STRUCTURES Figure 2. the canopy. The large mesh antennas in Table 2. Unfortunately. 1996 and thereby provided the ﬁrst demonstration of the potential of inﬂatable precision structures [41. it should be noted that the AstroMesh has an areal density signiﬁcantly lower than other mesh antennas and comparable to those of inﬂatable antennas. Then.8 Pa [20]. The IAE was launched on May 29. The solid surface antennas have.1 have a fundamental frequency f1 lower than 1 Hz. no value 30 . Currently. G. 2. JPL and L’Garde are working on a project called Advanced Radio Interferometry between Space and Earth (ARISE). as expected. which is desirable to separate the structural and attitude control system frequencies [55].1. Roederer. In that respect. ESA). the struts are inﬂated followed by the torus and ﬁnally. For mesh antennas. as expected.CHAPTER 2. The IAE has. Low mass combined with a high structural stiﬀness give a high lowest natural frequency of the structure. an ejection of the package from its container. the best packaging ratio with ratios of 6 and 14% in diameter and height.3. Another crucial parameter is the mass and surface density. worse packaging ratios than their mesh counterparts.
25 m diameter AstroMesh reﬂector is given in references [165.2. The last parameters. which can be compared to 17 Hz of the EGS antenna of about the same size. DEPLOYABLE REFLECTOR ANTENNAS Figure 2. A 6 m diameter AstroMesh is reported to have a fundamental frequency of 2.). 31 . the surface accuracy and operating frequency. for the deployed frequency of the 12. Inc. will be discussed in Chapter 5 in connection to the development of the tensegrity reﬂector antenna.21: The 14 m diameter IAE reﬂector antenna in space (Courtesy of L’Garde.0 Hz [166]. 166].3.
8 3.93 0. 172] [154] [154] [87.42 0.06 [20] [135] [43] I ARISE ISRS IAE FS EM O 25 20 14 a b M: mesh.52 2.98 1.5 0.4 m Dmin = 5.56 — — 11.6 4 2 — — 8.44 0.64 5.34 291 170 55 127 120 60 230 35 20 24 11 — 1.22 0.92 4.38 0.31 0.7 1.65 0.31 0.39 0.2 4 0. 134] [51.18 0.07 0. Dmax = 6.36 0.1: Comparison of diﬀerent deployable reﬂector antenna concepts (sorted by type and aperture size).25 5 4. EM: Engineering Model.00 FS EM EM EM EM 15 8 4.8 0.73 0.068 0.18 0. FS: Feasibility Study c Surface area: πD2 /4 or πDmax Dmin /4 if elliptic.06 — — 0.3 3.51 0.12 1.14 — 0.14 192 134 60 0.6 m.43 0. 146] [63] [91] [10] [40.5 0.05 — 0.08 0.35 0.5 0.9 4. 52] [52.22 — 0. S: solid and I: inﬂatable.10 0.25 m. 153] [51] Target surface accuracy d/D = 4 · 1.51 0.008 3000 0. Comment 2) (m) (kg) (kg/m (Hz) (mm) (GHz) Quadaperture conﬁguration 14 modules Dmin = 12.25 m.06 0.13 0.06 0.1 8 5.46/π/14 ≈ 0.36 0.4 — — — 0. 15 m2 HCA TT ETSVIII AstroMesh HRA ACeS AFRA WRA TT HALCA EGS SBA RRA TDRS TT ETSVIII CSPDA EM EM O O — O O O O O EM EM 15 13 12. O: Orbit.06 0.42 — — — — — — — — 0.58 1.37 0. . Sunﬂower DAISY Sunﬂower MEA SSDA — — 0.81 0.29 0.37 0.39 0.2 30 — — 86 22 — [60] [51.6 5.051 60 0.25 12 12 9.35 — — 17 — — — — 1.8 m One module.11 — 0. 17] [92] [166. c D d/D h/D Mass f1 δrms fop Statusb Ref.25 22 — 2 15 — — [9.6 — — 0.36 1.1 · 0. 52] [98] [188] S Ext. Dmax = 6.Typea Structure Table 2.71 1.07 0.18 0. Dmax = 16 m Status not known M Dmin = 5.54 0.73 — 0.17 0.24 0.54 — — 31 94 — — 6 1.
Hence. Figure 3. cf. circuit and zigzag.1 Introduction A major obstacle in the analysis and design of tensegrity structures is the determination of their equilibrium conﬁguration. Snelson [151] and Emmerich [37] into the form of tensegrity structures. convex polyhedra as the basis for ﬁnding new conﬁgurations. 147]. by Fuller [44. proper formﬁnding methods are needed to ﬁnd the equilibrium conﬁguration of even the simplest tensegrity structure [112].1. are found in reference [132]. the selfstressed shape of a tensegrity is not identical to that of the polyhedron and. but the various methods have not previously been classiﬁed or linked. Early studies. This happens. However. 88]. A large number of diﬀerent tensegrities. For general tensegrity structures. eﬃcient formﬁnding methods have been available for a long time.2. use mainly regular. however. for example. [5. [160]. This key step in the design procedure is usually known as formﬁnding. and the expandable octahedron (icosahedron). This purely geometric research has resulted in a large number of conﬁgurations which are classiﬁed by Pugh [132] into three pattern types: diamond. with detailed schemes and advice on how to build them.177] 33 . Formﬁnding methods for tensegrity structures have been investigated by many authors. both for the truncated tetrahedron. and recently by Connelly and Terrell [26]. therefore. and Sultan et al. Vassart and Motro [177]. Figure 3. For other tension structures.Chapter 3 Analysis Methods for Tensegrity Structures 3. physical models of these structures show that the shape of the tensegrity.160. such as membrane and cable nets. In the ﬁrst two sections of this chapter the existing methods were classiﬁed into two broad families—kinematic and static methods—and advantages and limitations of each method were identiﬁed. is diﬀerent from that of the polyhedron. Diﬀerent approaches are proposed by these authors. Closer scrutiny of the seemingly diﬀerent approaches in references [26. corresponding to a particular polyhedron. the formﬁnding process has proven to be more complicated.
(a) Side view (b) Top view Figure 3. Note the distortion of the hexagonal faces. revealed some links: indeed Vassart and Motro’s force density method was linked directly to the more abstract energy approach by Connelly. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES (a) Side view (b) Top view Figure 3.CHAPTER 3.2: Comparison of expandable octahedron and icosahedron (drawn with dashed lines). the static and kinematic properties.1: Comparison of truncated tensegrity tetrahedron and the polyhedron from which it originates (drawn with dashed lines). Once the formﬁnding step is completed. the states of selfstress and internal mechanisms. must be found. A suitable method 34 . i.e.
A compact description of the geometry of this problem. 12. This approach mimics the way in which tensegrity structures are built in practice. (3. 3. where i is an integer smaller than v.2. In the starting conﬁguration the lateral cable. without explicitly requiring that the cables be put in a state of pretension.1 Analytical Solutions Consider a simple structure consisting of cables arranged along the edges of a regular prism. taking advantage of its symmetry. 13. 2πi v T T p2 = R cos θ R sin θ H p3 = p4 = p5 = R cos θ + R cos R cos 2π v 2π v 2πi v (3. 12. as follows. by considering the square of the lengths of the lateral cable. 3.2. (3. T (3. the strut lengths may be kept constant while the cable lengths are decreased until they reach a minimum.1e) R sin θ + 2π v 2π v 0 −R sin R sin 0 T The kinematic formﬁnding proceeds as follows.3. As this method will be used frequently in the remainder of the thesis. Figure 3. KINEMATIC FORMFINDING METHODS for this task is the Force Method developed by Pellegrino and Calladine [127].3 shows the elements connected to one of the nodes of the bottom polygon. Alternatively. regular top and bottom polygons for which a tensegrity structure is obtained.2b) 35 . . The coordinates of nodes 1–5 are: p1 = R 0 0 T .2 Kinematic FormFinding Methods The characteristic of these method is that the lengths of the cables are kept constant while the strut lengths are increased until a maximum is reached.1d) (3. was introduced by Connelly and Terrell [26].2a) + H 2. and strut. . it is described in the last part of this chapter. Figure 3.1b) H . 2πi 2 lc = 2R2 1 − cos θ + v (3. is vertical and the angle between the ends of the strut is 2πi/v. 2 lt = 2R2 (1 − cos θ) + H 2 .1c) (3.1a) . Depending on the value of v and the oﬀset between vertices connected by a strut.3. plus a number of struts connecting the v vertices of the bottom polygon to vertices of the upper polygon. there is a special rotation angle θ between the plane.
2. one or more struts are elongated. proposed by Pellegrino [121]. z) subject to gi (x.2b) can be rewritten as: πi πi 2 2 sin + lt . nonsymmetric. until a conﬁguration is reached in which their lengths are maximised.3. (3. (3.5) where the objective function f (x.3) v v For a given cable length lc the length of the strut ls is maximised for θ=π 1 i − 2 v . y. z) is. z) = 0 for i = 1.. Equation (3.4) The simplicity of the kinematic method for structures with vfold symmetry is mirrored by the static method. The general constrained minimisation problem has the form: Minimise f (x. see section 3. Starting from a system for which the element connectivity and nodal coordinates are known. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES 3 z 2 H 4 1 x θ R 5 2πi v y Figure 3. 3. but for other. Pellegrino [121] applies this method to two tensegrities: the triangular prism and the truncated tetrahedron. z) are the ﬁxed lengths of the cables. the negative length of one of the struts and constraint equations gi (x. n. turns the formﬁnding of any tensegrity structure into a constrained minimisation problem.. 36 . respectively. .CHAPTER 3. y. maintaining ﬁxed length ratios. radius R and height H.2 NonLinear Programming This general method. where the subscripts c and s denote cable and strut. y. cases the present formulation becomes infeasible due to the large number of variables required to describe a general conﬁguration. for example.3: Elements meeting at node 1 of structure with vfold symmetry. y. lc = 4R2 sin θ + (3..
although diﬀerent geometric conﬁgurations of structures with the same topology can be found by specifying diﬀerent relationships between the lengths of the struts. hence three of its six nodes are ﬁxed in space. This problem can be solved. . has nine cables of length lc = 1 and three struts of equal length. equal to the negative length of one of the struts.3 Dynamic Relaxation The method of dynamic relaxation.4679 obtained from (3. there is no direct way of controlling the corresponding variation in the state of prestress. The ﬁnal length of the struts is 2. However.2. d and d. N a damping matrix. The ﬁnal length of the struts is 1. Note that the strut length obtained from a purely geometric analysis of the ideal truncated tetrahedron √ is 5 ≈ 2. that had already been successfully used for membrane and cable net structures [6. compared to the √ theoretical value of 1 + 2/ 3 ≈ 1. For a structure in a given initial conﬁguration and subject to given general forces the equilibrium conﬁguration can be computed by integrating the following ﬁctitious dynamic equations ¨ ˙ Md + Nd + Kd = f .7) where K is a stiﬀness matrix. v = 3 and i = 1. lc6 − 1 = 0 2 2 l − l 2 = 0 s2 s1 2 l − l 2 = 0 s3 s1 subject to (3. the number of constraint equations increases with the number of elements so this approach is not feasible for larger systems. velocity and 37 . . was put forward by Motro [110] and Belkacem [8] as a general formﬁnding method for tensegrity structures. s2 and s3 the struts. One of the base triangles is ﬁxed. the constrained minimisation problem has the form: Minimise − l2 s1 2 lc1 − 1 = 0 2 l − 1 = 0 c2 ..4).3) and (3. and d. 31]. Figure 3. for example. c2.2361. hence the slight warping of the hexagonal faces leads to an increase of the strut length by 0. . KINEMATIC FORMFINDING METHODS A triangular prism. f the vector ¨ ˙ of external loads. M a mass matrix. An advantage of the nonlinear programming approach is that it makes use of general purpose. (3. for the case of the truncated tetrahedron there are six struts and 18 cables. In Cartesian coordinates. Similarly. 15 on the cable lengths plus 5 on the struts.2507. Also.3. 3. The objective function to be minimised. c6 denote the six remaining cables and s1. has to satisfy 20 constraint equations. are the vectors of acceleration.1..7%.2. standard software.6) where c1.468. using the constrained optimisation function fmincon in Matlab [89]..
Alternatively. the method becomes rather cumbersome if several diﬀerent ratios between strut lengths and cable lengths are desired. There are several ways of carrying out a formﬁnding analysis. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES displacement from an initial conﬁguration. The relative error ν in the nodal coordinates is.3) and those for the expandable octahedron to the results of a static method. whereby the undamped motion of the structure is traced. The coeﬃcients of the damping matrix are usually all assigned the same value. (3. chosen such as to maximise the speed of convergence to the equilibrium conﬁguration. ¯ of the element. This error is computed for each node and the largest error is taken to represent the structure. respectively. small axial stiﬀness. (3. Motro [110] applies the dynamic relaxation method to the formﬁnding of the triangular tensegrity prism. as above. nodal equations of equilibrium are used to compute outofbalance forces from which the current acceleration can be obtained through (3. all velocity components are set to zero. a technique called kinetic damping can be used. [114] later conclude that the dynamic relaxation method has good convergence properties for structures with only a few nodes but is not eﬀective when the number of nodes increases. which is close to the value determined by Pellegrino [121]. for example by prescribing for each element of the structure a constitutive relationship of the type ¯ t = t0 + ke. The resulting system of uncoupled equilibrium equations can then be integrated using a centred ﬁnite diﬀerence scheme.9) where ˜ denotes the approximate values. until a state of prestress was set up in the structure. The relative errors in the nodal coordinates for the three structures are 0. The latter technique usually gives a faster convergence. measured from the initial conﬁguration. The lengths of the cables were held constant while the struts were gradually elongated.7).468. which 38 . The process is then repeated. Belkacem [8] analyses the triangular and square tensegrity prisms. are obtained. Motro et al. and also the expandable octahedron.CHAPTER 3. In any current conﬁguration of the structure. Both M and N are assumed to be diagonal. 4 and 2%. until the peak kinetic energy becomes suﬃciently small [6]. for node i.8) where t is the axial force and e the extension. An analysis of the truncated tetrahedron [111] yields a ratio ls /lc slightly greater than 2.24. Also. The results for the tensegrity prisms are compared with the theoretical values obtained from (3. For the tensegrity prisms. respectively. computed as νi = (xi − xi )2 + (yi − yi )2 + (zi − zi )2 ˜ ˜ ˜ 2 x2 + yi + zi2 i . and the velocities and displacements are initially set to zero. When a local peak in the total kinetic energy of the structure is detected. respectively. starting from the current conﬁguration.2. errors in the rotation angle θ of 1 and 8%. This analysis converged to ls /lc = 1. t0 is the desired prestress and k a ﬁctitious. for simplicity.
Note that (3. To set up a system of linear equilibrium equations.4. spherical tensegrities with polyhedral geometries.12) for which the cables are in tension is [26] θ=π 1 i − 2 v . as expected.5 and Group Theory to obtain closedform expressions for the equilibrium geometry of highly symmetric spherical tensegrities.13) is identical to (3. were also analysed using the same approach. 3.3. The distance between the struts in each pair is exactly half the strut length. Connelly and Terrell [26] use an equilibrium approach to ﬁnd the prestress stable form of vfold symmetric tensegrity prisms.e. 3.12) 2πi v = 0.3. Nishimura [117] uses the force method.10) and (3.11) give q12 sin θ − sin θ + 2πi v = 0. Other. The values of θ for tensegrity prisms with v going from 3 to 6 are given in Table 3. the cuboctahedron and the icosidodecahedron. [152]. Figure 3. Equations (3. However.2.3. described in section 3.1 Analytical Solutions Kenner [70] uses node equilibrium and symmetry arguments to ﬁnd the conﬁguration of the expandable octahedron. as variable for each element.and ydirection can be written as q12 H + q13 H = 0 and q12 R sin θ + q13 R sin θ + respectively.4).13) At the rotation given by (3.10) (3. cf. (3. Denoting by qij the force density in element ij—note that q14 = q15 due to symmetry—the equilibrium of node 1 in the z. i. whose six identical struts are divided into three pairs of struts which are mutually perpendicular.3 Static FormFinding Methods The general characteristic of these methods is that a relationship is set up between equilibrium conﬁgurations of a structure with given topology and the forces in its members.11) The only solution of (3. i. (3.13) the resultant force from cable 12 and strut 13 is radial. force divided by length. they use force density1 . (3. the same restriction applies to kinematic methods in general.1. 1 Also called tension coeﬃcient. This relationship is then analysed by various methods. This is discussed further in section 3. more complex.3.e. STATIC FORMFINDING METHODS restricts its applicability to less regular structural forms. 39 .
of size b × n.16) can be written also in terms of the y. For a general structure with b elements and n nodes the equilibrium equations in the xdirection can be written as CT QCs x = fx . e. uses a simple mathematical trick to transform the nonlinear equilibrium equations of the nodes into a set of linear equations. Q a diagonal matrix containing the force densities.16) (3.18) . 4. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES Table 3. v 3 4 5 6 1 30 45 54 60 2 −30 0 18 30 i 3 — −45 −18 0 4 — — −54 −30 5 — — — −60 3. Cs can be partitioned as Cs = [Cu Cf ] . (3.and zcoordinates. x a column vector of xcoordinates. whose value needs to be known at the start of the formﬁnding process. the equilibrium equation in the xdirection for node i is tij (xi − xj ) = fix .2 Force Density Method The force density method for cable structures. describes the connectivity of the structure. lij (3.16) can now be written as CT QCu xu = fx − CT QCf xf . Equation (3. These equations can be linearised by introducing for each element the force density qij = tij /lij . 6. see below. these nodes are attached to a foundation.1: Values of θ (◦ ) for v = 3. and fx a column vector of external nodal forces in xdirection.17) where the restrained nodes have been put at the end of the numbering sequence.3. 147]. it is actually nonlinear because the lengths lij in the denominator are also functions of the coordinates.g. If the coordinates of some of the nodes are given.14) j where node i is connected to j nodes by cables or struts and tij is the tension in element ij.15) where Cs is the incidence matrix. u u 40 (3. 5. s (3. ﬁrst proposed by Linkwitz and Schek in 1971 [85. Equations identical in form to (3. if an element connects nodes i and j. 86. Although this may appear to be a linear equation in the nodal coordinates. The incidence matrix Cs . then the corresponding row of Cs has +1 in column i and −1 in column j.CHAPTER 3. For example.
of at least 1 since the row and column sums are zero. with a nullity. The same approach can be extended to the formﬁnding of membrane structures by converting the stresses in the membrane into forces in a virtual cable net [5. i. several complications arise during formﬁnding. i. . where D = CT QCs . the D matrix for a tensegrity is semideﬁnite and. (3. without going through Cs and Q. the external loads are zero during formﬁnding. Hence. Analogous equations hold in the y. cables cannot increase in length. thus.20) Dij = q if i = j. due to the presence of compression elements. STATIC FORMFINDING METHODS where xu and xf are the column vectors of unknown and given xcoordinates. j (3. Further details will be given in section 3. pn . (3. there are usually no foundation nodes or external loads. but as these structures are selfstressed.3 Energy Method In the following.4.16) becomes Dx = 0. (3. and hence CT QCu is positive deﬁnite and. Unlike the matrix CT QCu for a cable net attached to foundation nodes. the dimension of the nullspace.e. struts cannot decrease in length and bars cannot change length. Usually. a strut or a bar. A similar formulation can be applied to the formﬁnding of tensegrity structures. Equation (3. respectively. In a structure consisting of cables only. cable net. Note that the n × n matrix D is always singular. invertible. can be solved to ﬁnd the nodal coordinates. A stress state ω for G(P) is a selfstress if the following condition holds at each node i: ωij (pj − pi ) = 0. there is always a unique solution to the formﬁnding problem. e. with qij < 0.and zdirections.22) 41 .20). s The force density matrix D can be written directly [177].18).3. k=i ik 0 if i and j are not connected. was presented by Vassart [175].3. 81].g. by (3. . see page 120 of reference [147].e.19) 3. A practical procedure for ﬁnding a set of force densities that yield a matrix D with the required rank. qij > 0. (3. u Therefore. all tension coeﬃcients are positive. which is u positive deﬁnite.and zdirections. some key main ﬁndings by Connelly [22] will be summarised using as far as possible the original terminology.21) A tensegrity framework G(P) is the graph on P where each edge is designated as either a cable. following the scheme −qij if i = j. A conﬁguration of n ordered points in Ddimensional space is denoted by P = p1 p2 .3. together with analogous equations for the y.
reference [21] deﬁnes the following energy form associated with the stress ω: 1 ωij pj − pi 2 . .25) E(P) = pT 2 Ω where the elements of Ω are given by −ωij if i = j. Ωij = ω if i = j. (3. The function in (3. followed by the ycoordinates. have a rest length of zero while the struts. (3. which only take compression. Mathematically. but not suﬃcient. Then.22) with the equilibrium equations for the same node written in terms of force densities. Ω (3. 1)T . have an inﬁnite rest length. ωij ≤ 0 for struts. the cables. . because—as already noted above for D—the nullspace of Ω contains at least the nontrivial vector (1 1 . A necessary condition for the tensegrity framework to be prestress stable in conﬁguration P is that the quadratic form E(P) has a local minimum at P. Let x p = y z (3. A basic principle in the analysis of the stability of structures is that the total potential energy functional should be at a local minimum for a given conﬁguration to be stable. k=i ik 0 if there is no connection between i and j.23) can be written as the quadratic form: Ω 1 p. which only take tension. In analogy with the total potential. but expecting positive deﬁniteness is unrealistic. it is obvious that the stresses ωij are identical to the force densities qij . The positive deﬁniteness of E(P) is directly related to that of Ω. Comparing (3.26) Note that Ω is identical to D. hence the above formulation provides a deeper insight into the characteristics of the force density method and how it can be used to ﬁnd stable equilibrium conﬁgurations of tensegrity structures. etc. energy builds up as a function of the square of the extension. containing the xcoordinates of P.CHAPTER 3.24) be a column vector. condition for the tensegrity framework to be in a stable equilibrium conﬁguration. All members are assumed to behave as linear elastic springs.23) E(P) = 2 ij The idea is that when the end points of an element are displaced.23) is set up to have an absolute minimum corresponding to the rest length of the element [22]. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES where ωij ≥ 0 for cables. Satisfying the above equilibrium condition is a necessary. 42 . of length Dn. (3. and no condition is stipulated for the bars. The link between the force density method and the energy minimisation was ﬁrst pointed out by K¨tter [73] o and later by Schek [147].
4 is super stable [23]. Hence the tensegrity structure in Figure 3. i. using group theory. 43 . A is skew symmetric.4. A complete catalogue of all the tensegrity structures that are possible for each symmetry group is produced. A further assumption is that there are two types of cables but only one type of strut.e. Ω= (3. When describing a rigid body motion. to design a super stable tensegrity framework one has to ﬁnd a set of force densities such that the nullity N of Ω is D + 1. is n − D − 1. by symmetry. A = −AT . consider the twodimensional (D = 2) tensegrity structure in Figure 3. the catalogue contains many solutions that were previously unknown. Such a mechanism can be described by the linear map di = Api + n. i. The maximal rank of Ω for a structure in Ddimensional space that does not in fact lie in a subspace of smaller dimension.4: Snelson’s Xframe The strongest type of prestress stability. A further condition for super stability is that there are no aﬃne inﬁnitesimal ﬂexes of the tensegrity framework G(P). Although some of the structures in the catalogue have struts that go through each other. matrix is [21]: 1 −1 1 −1 −1 1 −1 1 .4 where the outside edges are cables and the diagonals are struts. Connelly and Back [24] analyse tensegrity structures with diﬀerent types of symmetry using this method. The stress. requires prestress stability with the additional condition that Ω is positive semideﬁnite with maximal rank. and therefore are of limited practical interest. see examples in section 3. Aﬃne inﬁnitesimal ﬂex is another name for a linear inﬁnitesimal mechanism. Their initial assumption is that there is a symmetric state of selfstress with a force density of 1 in each cable and −ωs in each strut. The force density in the strut is chosen such that the structure is super stable [23. STATIC FORMFINDING METHODS 4 3 1 2 Figure 3. where A is a D × D matrix and n a translation vector [22].3. arranged such that satisfying equilibrium at only one node of the structure implies.27) 1 −1 1 −1 −1 1 −1 1 which is positive semideﬁnite with nullity 3.e. force density. 24]. A stress equal to 1 in the cables and −1 in the struts is a selfstress for this structure. that it is satisﬁed also at all other nodes.3.1. named super stability by Connelly [23]. Hence. For example.
. ∂gi (3. The change of length of cable j is N δlj = i=1 ∂lj δgi .CHAPTER 3. A set of equilibrium equations relating the forces in the cables. generalised coordinates g = (g1 g2 . ANALYSIS METHODS FOR TENSEGRITY STRUCTURES (a) (b) Figure 3. [160].5 for some structures from their catalogue. 44 . hence N = 3 × O.4 Reduced Coordinates This method was introduced by Sultan et al. a set of N independent. the cable forces t = (t1 t2 .28) Considering all cables. can be obtained from virtual work. which deﬁne the position and orientation of these struts. if D = 3 then N = 5 × O. Hence. 3. Consider a tensegrity structure whose b elements consist of M cables and O struts. (3.5: Symmetric tensegrities from Connelly and Back’s catalogue.2 Consider a state of selfstress for the structure and let tj be the axial force in a generic cable element j. Consider a virtual displacement δg of the structure that involves no extension of the struts. tM )T are in equilibrium with appropriate forces in the struts and zero external loads. 2 (3. gN )T is deﬁned. The struts are considered as a set of bilateral constraints acting on the cable structure.3. .29) If D = 2 three generalised coordinates are required for each struts.28) gives δl = AT δg. See Figure 3. but without showing explicitly the forces in the struts. . .
31) For the structure to be in equilibrium. i.3. e.34) (3. Maple or Mathematica. This gives the following reduced equilibrium equations At = 0. . i.32) General analytical conditions that govern the form of a tensegrity structure of given topology can be obtained by analysing (3. tj > 0 for j = 1. Having shown that a structure in which the rigid plates have been replaced by cables. is tT δl = (At)T δg.g. (3. Sultan [159] applies this method to a tensegrity mast of which Figure 3. The same structure had been previously considered by Snelson [151]. By using symbolic manipulation software. the angle between the strut and the zaxis.30) Because the extensions of the struts are zero. the ﬁnal shape of the structure is still unknown and the existence of a prestressable conﬁguration is dependent on ﬁnding a suitable set of strut lengths. but involves a smaller number of unknown cable forces. has the same equilibrium conﬁguration as the original structure. and diagonal—between two rigid triangular plates at the top and bottom..3. the angle between the vertical plane containing the strut and the x–z plane. M . and the colatitude δj . in Figure 3. i. The mast consists of three struts per stage. 2. Sultan analyses this simpler problem. from the cables only. this must be zero for any virtual displacement δg. Sultan [159] reduces the number of 45 . the azimuth angle αj . At this stage.33) (3. and • three translation and three rotation parameters deﬁning the position and orientation of the rigid plate at the top with respect to the bottom plate. where only solutions that are entirely positive are of interest.33) and (3. The 18 coordinates chosen by Sultan [159] for the twostage mast are • for each strut.6 shows a simple twostage example.e.6 note the deﬁnition of the overlap h.34). ∂gi (3. The ﬁrst step in the formﬁnding process is to identify a set of generalised coordinates which describes the conﬁguration of this structure. STATIC FORMFINDING METHODS where the elements of the N × M matrix A are Aij = ∂lj .. in symbolic form.. the virtual work in the struts is also zero and so the total internal work. vertical. the length of each cable can be expressed in terms of the 18 coordinates and then diﬀerentiated to obtain the 18 × 18 matrix A. held in place by three sets of cables—saddle. (3. For this equation to have a nontrivial solution it is required that rank A < M.e.e.
to 3 × 3 with the forces in the diagonal. Finally. and (c) top view. and by considering a ﬁxed position of the top plate. Then. (3. independent generalised coordinates by considering only symmetric conﬁgurations. applying to this reduced matrix the condition for the existence of nontrivial solutions. A particular symmetric conﬁguration. with the same azimuth. and l the length of each strut.35) π . and colatitude.6: Sultan and Skelton’s two stage tensegrity tower: (a) threedimensional view. a is the side length of the equilateral triangles at the top and bottom of the mast. δ.CHAPTER 3. is deﬁned by the following relationship between δ and α: π 2a . in which all the nodes lie on the surface of a cylinder. yields a quadratic equation that could be solved for the overlap h: π a2 1 − 3l2 sin2 δ cos2 α + π 3 6 2 tan δ cos α + 6 h= a π − √ + l sin δ cos α + if α = 6 3 l cos δ if α = 2 (3. the problem could be reduced even further.37) δ = arcsin √ sin α + 3 l 3 46 .36) Here. 3 π . by assuming a special symmetry in t. (b) side view. saddle and vertical cables remaining as the only unknowns. rank A = 2 equivalent to det A = 0. α. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES Base ru St t h al on ag Di Vertical (b) Sad dle y z y x α x (a) (c) Figure 3. 3 (3.
the 47 . For example. Case N = 1 Most sets of force densities yield a D matrix with nullity one. The following examples show how this is done in practice. Vassart and Motro [177] list three techniques for ﬁnding a set of force densities that achieve the required nullity: (i) intuitive. The third technique is the most eﬀective. This section will deal with procedures to actually ﬁnd superstable tensegrities. For it to be super stable. A was derived using symbolic software. and will be illustrated in section 3. 3.4.4.4. the nullity of D has to be three. but it is interesting to consider also the cases N = 1. the second technique is based on a trialanderror. Sultan [159] successfully applied this formﬁnding method to masts with up to nine stages.7. D is analysed in symbolic or semisymbolic form. or more reﬁned search for a set of force densities that yield the required nullity.3. (ii) iterative and (iii) analytical. with positive semideﬁnite matrix D with nullity N = D + 1. i. IMPLEMENTATION OF THE FORCE DENSITY METHOD 3 2 2 3 9 8 1 7 4 1 4 6 5 5 6 Figure 3.2. 3.1. For masts with more than two stages. in the case of systems with a large number of elements [177].1 A TwoDimensional Example Consider the hexagonal tensegrity shown in Figure 3.4 Implementation of the Force Density Method The force density method has been outlined in section 3. for structures of increasing complexity.e.7: Twodimensional hexagonal tensegrity.3. but (3. the ﬁrst is suitable for systems with only a few members.3 it was stated that one must look for sets of force densities that make rank D = n − D − 1.35) was then solved numerically. 2 to better understand why in section 3.3. Of these three techniques.
Denoting 48 . columns four. For example. and hence that the nullspace of D2 is spanned by (−1 −1 0 1 1 0)T and (2 2 1 0 0 1)T . q3 = (2 2 2 2 2 2 −1 −1 −1)T yields 3 −2 0 1 0 −2 −2 3 −2 0 1 0 0 −2 3 −2 0 1 . 0 −3 0 −4 0 8 0 −4 1 −5 0 9 0 −5 2 (3. . α and β can take arbitrary values. and (2 2 1 0 0 1)T . and is again of little practical interest.39) 1 0 −2 3 −2 0 0 1 0 −2 3 −2 −2 0 3 0 −2 1 It can be readily veriﬁed that columns ﬁve and six are dependent.8(a). This solution corresponds to conﬁgurations of the structure where all the nodes coincide and so the whole structure is reduced to a single point. Here. hence the nullspace of D3 is spanned by (1 2 2 1 0 0)T . were prescribed. For example. x6 )T = (α α α α α α)T and y = (y1 .CHAPTER 3. D3 = (3. in a ratio of two to one. for q2 = (2 2 2 2 2 2 − 1 − 1 − 3)T 3 −2 0 1 0 −2 −2 3 −2 0 1 0 0 −2 1 −2 0 3 . give the conﬁguration x = (x1 . .38) The nullspace of D1 is spanned by (1 1 1 1 1 1)T —see reference [156] for further details on how to compute a basis for a nullspace—which. the conﬁguration of the system is described by x = (−α + 2β −α + 2β β α α β)T and y = (−γ +2δ −γ +2δ δ γ γ δ)T . D2 = (3.40) 1 0 −2 3 −2 0 0 1 0 −2 3 −2 −2 0 1 0 −2 3 Here. as shown in Figure 3. denoting by α. This conﬁguration corresponds to all nodes lying on a straight line. Case N = 2 Next. uniform force densities both in the cable elements and in the struts. (−2 − 3 − 2 0 1 0)T . uniform force densities in all cable elements and in two of the struts force densities of half those in the cables were prescribed.19) and the analogous equation in the ycoordinates. δ their ycoordinates. β the xcoordinates of nodes 5. ﬁve and six are dependent. the force density in the third cable was arbitrary. Case N = 3 Finally. . through (3. and by γ. Hence. 6 respectively. . y6 )T = (β β β β β β)T . ANALYSIS METHODS FOR TENSEGRITY STRUCTURES arbitrary set q1 = (1 2 3 4 5 6 0 −1 0 D1 = 7 0 −6 − 7 − 8 − 9)T produces −1 0 7 0 −6 −5 −2 0 8 0 −2 −4 −3 0 9 . which is of little practical interest.
and ζ = −1/2. γ = √ √ 1/2. A set of force densities with threefold symmetry is prescribed as follows.2 Tensegrity Prisms Consider a structure with the topology shown in Figure 3. 2 5 (a) N = 2 (b) N = 3 6 3 1 3 2 4 1 5 (c) N = 3 6 Figure 3. ζ their ycoordinates.4. δ = 0. (3.7 is reobtained for α = −1. The force densities in the cable forming the bottom and top triangles are qb and qt . 6 4 1. 6. IMPLEMENTATION OF THE FORCE DENSITY METHOD 2 4. 49 . the system conﬁguration is given by x = (α − 2β + 2γ 2α − 3β + ζ)T .3. respectively. 5 3. However. note that. and ζ = − 3/2. an alternative choice was q4 = (1 2 1 2 1 2 −2/3 −2/3 −2/3)T . equilibrium perpendicular to the planes of the triangles gives qs = −ql . cf. symmetric state of force densities is because the element lengths are not explicitly set in the force density formulation. by carrying out a symbolic analysis of the force density matrix other solutions were subsequently found. β = −1/2. 5. 2γ 2α − 2β + γ α β γ)T and y = (δ − 2 + 2ζ 2δ − 3 + 2ζ 2δ − 2 + ζ δ The original solution in Figure 3. γ the free xcoordinates of nodes 4.4. by α. = − 3/2. as in the original conﬁguration) is that shown in Figure 3.7 the force densities must have a particular distribution. β = −1/2. this solution also produces the conﬁguration shown in Figure 3. respectively. to satisfy nodal equilibrium. The reason why it is possible to ﬁnd less symmetric or even asymmetric conﬁgurations for a given.8(c). However.10). Assuming that the top and bottom triangles are parallel. etc. for α = −1.8: Conﬁgurations of twodimensional hexagonal tensegrities. √ δ = 0. despite the force densities q3 being symmetric. = − 3/2. β. .8(b). β. for which a particular conﬁguration (with α. γ = 1. and δ. which has only twofold symmetry. For example. it is noted that the particular q3 considered above was obtained after noticing that in the conﬁgurations shown in Figure 3.9. 3. In concluding. they are ql in the lateral cables and qs in the struts.
rank D = 2 and so q ∗ = 0. if q ∗ = 0. Any set of positive cable force densities that satisﬁes the condition ql2 − 3qb qt = 0 (3. the last four coordinates can take arbitrary values. In case (ii) the conﬁguration of the structure is described by 50 . In both cases. Hence. rank D is either four. the D matrix can be set 2qt −qt −qt D= −ql 0 ql up in terms of only three force densities −qt −qt −ql 0 ql 2qt −qt ql −ql 0 −qt 2qt 0 ql −ql . √ in case (i) the conﬁguration of the δ.e. ql 0 2qb −qb −qb −ql ql −qb 2qb −qb 0 −ql −qb −qb 2qb (3. If super stability is required. denoting those xcoordinates. or two. U= 0 0 0 0 q ∗ −q ∗ 0 0 0 0 0 0 0 0 0 0 0 0 (3.41) By Gaussian elimination D is reduced to the upper echelon form [156] ql 0 −ql −qb −qb 2qb 0 ql −ql qb −2qb qb 0 0 0 −q ∗ 0 q∗ . √ structure is described by x = (α+(β +γ −2δ)/ 3 α+(−β +2γ −δ)/ 3 α β γ δ)T . γ. Since ql = 0.42) where q ∗ = (ql2 − 3qb qt )/ql .43) is possible. but Vassart [175] presents two interesting cases: (i) qt = qb and (ii) qt = ql . similarly for y and z. if q ∗ = 0. β.CHAPTER 3. i. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES 2 2 3 3 1 1 12 6 10 5 11 5 8 7 4 4 6 9 Figure 3. then N = 4. by α. for example.9: Tensegrity prism.
Figure 3. and qs = −ql 3 4 Figure 3. was found 51 . The rotationally symmetric conﬁgurations.10. Figure 3. Further applications of the same method. Earlier analysis. many asymmetric conﬁgurations can be found. Truncated tetrahedron An equilibrium conﬁguration for the truncated tetrahedron. In addition to the symmetric conﬁguration of Figure 3.10: Top views of two diﬀerent rotationally symmetric tensegrity prisms x = (α + (β + γ − 2δ)/3 α + (−β + 2γ − δ)/3 α β γ δ)T . equilibrium in the strut direction yields qs = −3qc /2. qb = ql /3.1.4. ql = √ 3qb . etc. has shown that the distance between parallel struts is half the length of the strut. there is only one type of cable and one type of strut. and qs = −ql (b) qt = ql . 3. are shown in Figure 3. An analysis of D produces two possible solutions for N = 4: qs = −3qc /2 and qs = −2qc .2.3 Spherical Tensegrities The earlier part of this section has shown applications of the force density method to the formﬁnding of some relatively straightforward tensegrity structures. Expandable octahedron For the expandable octahedron. many geometrically nonsymmetric conﬁgurations may be found by appropriate choices of the free nodes. obtained by giving appropriate values to α. [70]. to slightly more complex systems will be presented next. but D is positive semideﬁnite only for the ﬁrst one. several symmetric conﬁgurations that had already been found by other methods were thus reobtained. cf.4. Again.2. changes the relative sizes of the top and bottom triangles.3. IMPLEMENTATION OF THE FORCE DENSITY METHOD 5 2 5 2 1 1 6 4 6 3 (a) qt = qb . Changing the relationship between qt and qb while keeping ql ﬁxed.
and also a second solution qs /qt = −2. the condition for N = 4 was found to be 2 (1 + γ) qs qt 2 + [3 + 2γ (3 + γ)] qs + γ (3 + 2γ) = 0. 1. For that conﬁguration. As shown in Figure 3. is ﬁnally solved analytically. the nodes of the truncated tetrahedron lie in four diﬀerent horizontal planes.3795 and −0. and −1. γ = 1 as in reference [24]. unlike the structures analysed so far.6672. e. ql for the other cables. diﬀerent conﬁgurations. For diﬀerent values of γ. and qs for the struts. but less than ideal for structures with some known.CHAPTER 3. This method resembles the approach used by Connelly and Back [24]. or desired element lengths. icosahedron. qt (3. the corresponding force densities are 1. However.5016 in the struts. a. In the case of the truncated tetrahedron. the condition for a nontrivial selfstress is [117] √ 4κ2 cos α + 4 cos α 1 + cos2 α + 4 3 sin α (cos α − 1) √ √ 3 sin α + 2 − 2 3 sin α − 12κ cos2 α + cos α √ + 9 cos α − 3 sin α = 0. To complete the formﬁnding of spherical tensegrities it should be mentioned that Nishimura [117] uses the force method together with group theory to ﬁnd the initial equilibrium conﬁgurations of such structures. octahedron.3795 and thus qs /qt = −0. it is concluded that the force density method is an excellent method for ﬁnding the conﬁguration of new tensegrities. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES in reference [121]. An independent formﬁnding study showed that. For a force of 1 in the cable elements forming the triangle faces. there is a force of 1. and hence the corresponding conﬁgurations are unstable.3795 in the remaining cables. for that solution the D matrix has negative eigenvalues. it was diﬃcult to specify the ratio γ to ﬁnd a conﬁguration with a particular ratio between the lengths of the cables. Since the strut and cable lengths are 2. (3. Nishimura [117] notes that as κ → 0. and α is the angle of rotation of the triangles with respect to an inertial system xyz.5. With Nishimura’s contribution. cube. the single state of selfstress was computed by the force method described in section 3. of the truncated tetrahedron were obtained. Therefore.2507 and 1. By setting ql = γqt . no relationships between the force densities could be obtained from equilibrium statements without considering the geometric symmetry conditions. the general state of selfstress is characterised by three diﬀerent force density values: qt for cables forming the triangles.5022. Hence. and after carrying out a Gaussian elimination on the matrix D. respectively.44) To reobtain the earlier results γ = 1. and dodecahedron. However.1. with unequal cable lengths.45) where κ is the ratio of the side length of the truncating tetrahedron and the side length of the original tetrahedron. 52 .g.6672. the formﬁnding of tensegrities corresponding to the truncated versions of the following regular polyhedra: tetrahedron.
46) (3.48) yields the strut length ls = 2. is qt = −qs 3 − 4κ (1 − cos α) . a framework of pinjointed bars. 6κ cos α √ cos α + 3 sin α [3 − 4κ (1 − cos α)] √ .3348. As described in Chapter 1. with three degrees of freedom at each joint.49) and (3. This value is very close to 2. The state selfstress of the truncated tetrahedron. and the struts are lt = κa. Three of the ﬁve parameters. The method.2507. a 9 + 24κ(κ − 1) − 4κ [(3 − 4κ) cos α + κ cos 2α].46)–(3. a reﬁned version of the method was presented in [15]. is ﬁrst described in [127]. were missing. (1. Following the criticism. The lengths of the cables forming the truncating triangles. the cables connecting the triangles. In reference [127]. the number of bars. the computational scheme to ﬁnd the 53 . The speciﬁc case where the lengths of the cables are equal is considered.49) (3.5 The Force Method for Analysis of Bar Frameworks Once the prestress stable form of the tensegrity structure has been found its kinematic and static properties are sought. κ.2). 3. κ = 0. the basic positive deﬁnitiveness conditions. ql = −qs cos α 6 − 4κ 2 + cos α + 3 sin α (3. 18 (3. as presented.3.379421 and qs /qt = −0.5.50) gives ql /qt = 1.1127. A method to investigate the properties of a general framework of pinjointed bars is developed by Pellegrino and Calladine in a series of articles [14–16. but the questions around that were ﬁnally resolved in [16]. 122. 127]. which is now known as the force method.48) respectively. and α subject to ll = lt = 1 yields a = 2. depend on the geometrical conﬁguration of the framework. and as κ → 1/2. Simultaneously solving (3.9873.256274. are prescribed but the other two. joints and kinematic constraints. was incomplete as it failed to correctly analyse relatively simple frameworks. ll = 3 2 √ a ls = √ 9 + 3 − 4κ − 2κ cos α − 3 sin α . again in accordance with the previous analysis. 124. found by Pellegrino [121]. expressed in terms of force densities. Kuznetsov [74] found that the method. for prestress stability. Substituting these value into (3. can be completely characterised by the extended Maxwell’s rule. but not suﬃcient. and α = 0. the number of selfstress states and internal mechanisms. Kuznetsov [75] came up with what looked as a counterexample to the improved method.50) Inserting the present values into (3. α → 0. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS α → π/6. however.47) (3.48) for a.667142. It included the matrix rank conditions that are necessary.
(3. the previous method seems more advantageous. t is the internal force vector of length b. because what really is needed in the static and kinematic analysis of very large frameworks.51) where H is the 3j − c × b equilibrium matrix containing the direction cosines. but has several advantages which will be explained later. 3.2 Static and Kinematic Properties The equilibrium and compatibility matrices contain essential information about the framework. in the x. The set of equilibrium equations for the framework is written as Ht = f .1 Equilibrium and Compatibility Matrices Consider a threedimensional pinjointed bar framework with j joints and b bars acted upon by external forces at the joints. The SVD scheme is computationally more expensive than the Gaussian elimination scheme. based on a series of articles by him over the past three decades.52) where C is the b × 3j − c compatibility matrix. The set of linear compatibility equations is Cd = e (3. y and z directions. Wellknown in the linearalgebraic treatment of matrices are the four 54 . 3. 2 2 Equating the external work and strain energy yields HT = C. (3. The framework is in its deployed state and restrained in the threedimensional space by c kinematic constraints.53) Equation (3. Pellegrino [124] presents a scheme based on the Singular Value Decomposition (SVD) of the equilibrium matrix. Another method of analysing the properties of bar frameworks is presented by Kuznetsov in [76]. 155]. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES number of selfstress states and mechanisms was based on Gaussian elimination. In the next ﬁve sections the method by Pellegrino and Calladine [14–16. the method does not lend itself to eﬃcient numerical analysis as well as the previous one. Therefore.5. and e is the bar elongation vector of length b. 127] is described. If the structural deformations under the external loads are small. a linear analysis is sufﬁcient.CHAPTER 3. is a robust and eﬃcient numerical method. 124. 122.5. In comparison. d is the joint displacement vector of length 3j − c.53) expresses the static–kinematic duality of bar frameworks [124. and f is the external force vector of length 3j − c. The work done by the external loads is 1 f T d and the strain energy stored in the framework is 1 tT e. involving several hundred elements. but such a scheme will have problems with illconditioned equilibrium matrices. of each element. The external forces give rise to displacements which must be compatible with the elongations of the bars.
⊥ Nullspace: states of selfstress.5. vector subspaces. which are the square roots of the eigenvalues of HHT and HT H.3. s = rH Column space: loads which can be equilibrated in the initial conﬁguration. in which the equilibrium matrix H is factorised as H = UΣWT . (Solutions of Ht = 0) Compatibility C Column space: compatibility bar elongations rH = Bar space Rb ⊥ Left nullspace: incompatible bar elongations.11. (Solutions of Cd = 0) m = Figure 3. suggested by Pellegrino [124]. for further information on subspaces see [156].11: The four fundamental subspaces associated with H and C [122]. the number of pivots after a Gaussian elimination of H. The rank rH of the equilibrium matrix H is the number of nonzero diagonal elements in Σ.. W is a b × b orthogonal matrix. cf. In practice. is the SVD. Joint space R3j−c ⊥ Left nullspace: loads which cannot be equilibrated in the initial conﬁguration ⊥ Nullspace: inextensional displacements. rH ) on the leading diagonal and all other elements zero. Another approach. (3. The coeﬃcients of Σ are the singular values of H.. the diagonal of 55 . THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS Dim. Equilibrium H Row space: bar tensions in equilibrium with the loads in the column space. = Row space: extensional displacements.. and Σ is a 3j − c × b matrix with rH positive elements σii (i = 1. . The physical signiﬁcance of the four vector subspaces associated with H and C are brieﬂy explained in Figure 3.54) where U is a 3j − c × 3j − c orthogonal matrix. The standard way to ﬁnd the bases for the four subspaces is through Gaussian elimination.
. Assembly type I II III IV Static and kinematic properties s=0 m=0 s=0 m>0 s>0 m=0 s>0 m>0 Statically determinate and kinematically determinate Statically determinate and kinematically indeterminate Statically indeterminate and kinematically determinate Statically indeterminate and kinematically indeterminate Σ often contains up to min(3j − c.55) s = b − rH and m = 3j − c − rH . It is therefore not possible to just pick out the rigidbody mechanisms from D. What matters in the classiﬁcation is only if s or m is zero or not. arising from inadequate kinematic restraints of the structure. none of these values are actually equal to zero. u3j−c . A basis S for the states of selfstress is given by the last s columns of W.2: Classiﬁcation of structural assemblies.CHAPTER 3. Similarly.3 RigidBody Mechanisms The mechanisms in D can either be internal mechanisms or rigidbody mechanisms. These two types of mechanisms are fundamentally diﬀerent but the SVD algorithm makes no distinction between them. 3. D = urH +1 . A scheme to separate the internal 56 . (3. wb . a basis D for the mechanisms is given by the last m columns of U. .57) (3. Values smaller than a set tolerance have to be treated as zero. In a case when the structure is not fully restrained. D generally contains combinations of the two mechanism types. A wide gap in an otherwise continuous range of singular values is another sign to look for.58) (3.2. . Table 3. S = wrH +1 . but notes that a lower tolerance may be required near critical points where the rank of the equilibrium matrix drops.5.56) Pellegrino [122] classify structural assemblies into four groups depending on their degree of static and kinematic indeterminacy. but some are much smaller than others. Once rH has been decided the number of selfstress states and mechanisms are (3. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES Table 3. . Pellegrino [124] sets the tolerance to 10−3 · σ11 for most structural assemblies. b) singular values of decreasing magnitude.
3. (3.3. (3..4 Internal Mechanisms Once the structure has been adequately ﬁxed. but only m − mrb of them are independent. . It can cope with up to six rigidbody mechanisms.. In a ﬁnite mechanism the joints can move with no change in lengths of 57 . However.61) r where R has size c × 6. The corresponding set of rigidbody mechanisms to those motions is found by (3. the number of rigidbody mechanisms is (3..k . m) each independent rigidbody mechanisms drb. THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS mechanisms from the rigidbody ones was proposed by Pellegrino [121.5. Now each of the rigidbody mechanisms of Drb has to be removed from each column of D. Physically. the system of c equations in six unknowns is n R = 0. This is done by the GramSchmidt orthogonalisation procedure [156]: for each mechanism dj (j = 1. Any rigidbody displacement in threedimensional space may be described by a translation n and a rotation r..62) mrb = 6 − rR . (3. A basis for the nullspace of R is an independent set of mrb rigidbody motions in terms of n and r.5. A matrix Drb of size mrb × 3j − c containing the rigidbody mechanisms is formed. (3.60) where pi is the position vector of point i in the original conﬁguration.k (k = 1.k k=1 D is now transformed into a set of internal mechanisms. Thus. there are two types of internal mechanisms: inﬁnitesimal and ﬁnite. The displacement di of a point i in such a rigidbody motion is given by di = n + r × pi .k drb. mrb ) is removed by use of the formula mrb dT dj rb. When dealing with complex frameworks it is important to be able to diﬀerentiate between rigidbody mechanisms and internal mechanisms.59b) .. the main aim here is to eﬃciently restrain any structure so that D does not contain any rigidbody motions. If the structure has a total of c constrained degrees of freedom.59a) (3. The rank rR of R counts how many of the c kinematic constraints that suppress the rigidbody degrees of freedom. where n = n x ny nz r = rx r y r z T T . 127]. . the remaining mechanisms are only the internal ones.k rb.63) dj := dj − dT drb.. A subsequent Gaussian elimination step can be used to remove those dependent on others [121].60).
Also shown in that example was that a state of selfstress gave the assembly a positive ﬁrstorder stiﬀness. as shown in the simple example in section 1. 122]. [176] present an analytical method in two steps. Such inﬁnitesimal mechanisms are associated with secondorder changes of bar lengths. This is normally done by the Finite Element Method (FEM) but can.and higherorder mechanisms. inﬁnitesimal mechanisms tighten up when activated. This framework is classiﬁed as a ﬁrstorder inﬁnitesimal mechanisms but still cannot be stabilised by a state of selfstress [16]. they have been given some attention. A negative deﬁnite GT D corresponds to a selfstress −t. 58 . ANALYSIS METHODS FOR TENSEGRITY STRUCTURES the bars. they later found a certain framework with two states of selfstress and two internal mechanisms. For simple problems this can be done symbolically. mechanisms which cannot be stabilised by a state of selfstress are of second or higher order or are ﬁnite. of second or higher order. the test for ﬁrstorder mechanisms is to check that the symmetric matrix GT D is positive or negative deﬁnite. Each state of selfstress has its own matrix G. one of ﬁrstorder and one of higherorder.3. In many cases this method gives more insight into the problem.e. Such a routine is presented by Calladine and Pellegrino [15].e. i. For assemblies with s > 1. What is left is to analyse the response of the framework to external loads. the tools to completely classify a pinjointed bar framework have been provided.64) where Bdj is similar to the equilibrium matrix H but with coeﬃcients of type (dj − kx dj )/lkl instead of (xk − xl )/lkl . then the mechanisms are ﬁrstorder inﬁnitesimal. Vassart et al. (3. For assemblies with s = 1 and m > 0. Despite the stability issues related to second. each corresponding to a selfstress state. Hence. For each state of selfstress ti and each mechanism dj a vector of geometric loads gij can be computed as gij = Bdj ti . for certain frameworks. Thus. geometry and energy characterisation. If a state of selfstress can impart positive ﬁrstorder stiﬀness to every mechanism of a structure. the test becomes slightly more complicated as a linear combination of all matrices GT D. At this point.CHAPTER 3. which is positive deﬁnite has to be found. Now if the only objective is to determine if the mechanisms are ﬁrstorder inﬁnitesimal or not there is a better approach proposed by Calladine and Pellegrino [15. In terms of displacements. The m geometric loads for selfstress state i form the columns of the geometric load matrix Gi of size 3j − c × m. this change in length is. while in an inﬁnitesimal mechanism there are some small length changes of the bars when the joints move. if an assembly has some mechanisms they must be ﬁrstorder inﬁnitesimal to ensure the stability of the structure. However. i. be done by the present method. an automatic numerical routine is required. However. displacements of mechanism j instead of nodal lx coordinates. but for larger problems. generally. which can determine the order of the mechanisms and identify ﬁnite mechanisms. Connelly and Servatius [25] discuss the various deﬁnitions of order of rigidity and mechanisms in literature and redeﬁne higherorder rigidity.
however. The assumptions of the method. are that geometrical and material nonlinearities can be neglected. i). THE FORCE METHOD FOR ANALYSIS OF BAR FRAMEWORKS 3. the bar elongations e. as it is presented below. undergo signiﬁcant increases of prestress during loading and deform less than predicted by the linear method. If the framework passes this test.51) is t = tf + Sα. Assemblies of type IV. Hence.67) gives the resulting internal forces t. Before starting the computations of the internal bar forces t.69) stating that the elongation e must vanish in the subspace of incompatible strains. Nevertheless. σii (3. Figure 3. where e0 is the vector of initial bar elongations. In analogy with (3. Table 3. (3. Substituting (3.67) the general solution to (3. which in the case of a deployable structure is equal to the deployed one.52) is d = de + Dβ.67) into (3. First.65) e = e0 + Φt. This method has been found to be very accurate for assemblies of type II. and the displacements d one has to check that the framework can carry the load f .124] presents a method for computing the forces and displacements of a pinjointed bar framework due to external loads. (3. the external load f must be zero in the subspace of loads which cannot be equilibrated.2.69) and solving for α yields α = −(ST ΦS)−1 ST (e0 + Φtf ) Finally. cf. (3. substituting α into (3.5. the method is linear and relates to the initial conﬁguration of the assembly. the linear method is presented below with respect to a type IV assembly.5 Structural Computations Pellegrino [122. the element forces t are related to the bar elongations e by the ﬂexibility matrix Φ as (3.71) (3.66) DT f = 0. The b × b diagonal ﬂexibility matrix Φ has φi = li /Ai Ei as its entry of position (i.5.65) and (3.11: (3.70) 59 .67) where tf contains the internal forces in equilibrium with the load f and α is a vector of s free parameters. the force and displacement systems have to be connected: assuming linearelastic material.3. tf is eﬃciently computed as rH tf = i=1 uT f i wi . the general solution to (3.68) The value of α is determined by the following compatibility condition ST e = 0.
72) Note that the term Dβ in (3. The set of equilibrium equations for the hanging net is written as: 60 .74) Substituting β back into (3. 3.73) and solving for β yields β = −(GT D)−1 GT de . The horizontal projections of the members all have equal length l and the middle triangle lies a distance h = l/4 beneath the plane of the supports. (3. Three nodes are fully ﬁxed in space. de is computed as rH de = i=1 T wi e ui .71) into (3. Substituting (3.71) gives the resulting joint displacements d. which is obtained from virtual work considerations. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES l 5 z l 2 1 2 y l x l 4 l 3 1 W l 6 3 W W Figure 3.71) represents a general inextensional displacement satisfying Cd = 0. (3. β is determined by the orthogonality between the matrix of geometric loads G and d.12: Hanging triangular net where de is any vector of displacements compatible with e and β is a vector of m free parameters. Similarly to tf .5. A weight W hangs in each of the unconstrained nodes.CHAPTER 3.73) GT d = 0.12 with six pinjointed bars.6 Example: Hanging Triangular Net Consider the hanging triangular net in Figure 3. σii (3.
76) Using AE/W = 103 . Theoretically.15 −22.31 4. [69].13 shows the variation of the two lowest singular values with the height h of the triangular platform. (3.12 4.75) where ti is the tension force in bar i and the right hand side the external loads. which search for equilibrium conﬁgurations that permit the existence of a state of prestress in the structure 61 .74) gives β ≈ 0 in the present numerical precision. t4 −W t5 0 t6 0 −W (3. The computation of the internal forces are straightforward as S = 0. (3.9 103 T .01 < h/l < 0. Figure 3. Thus. An SVD of the equilibrium matrix gives s = 0 and m = 3.9 −0.667 1. which determine the conﬁguration of either maximal length of the struts or minimal length of the cable elements.6. the displacements of the assembly are due to extensional deformation only: the displacements are d= l 1. The ﬁrst category contains kinematic methods.9 −0.6 Discussion Seven formﬁnding methods for tensegrity structures have been reviewed and classiﬁed into two categories. It appear that a lower tolerance must be used if the geometry has to be accurately determined. To ﬁnd the symmetric mechanisms a special approach is needed.31 2.667 −1. 3.68) gives t = W 2.31 2. The second category contains static methods. all bars lie in one plane.33 0 −22.3.15 −22. the numbers of selfstress states and mechanisms change to s = 1 and m = 4 using the recommend tolerance 10−3 · σ11 for the small singular values. (3.77) If the level of the triangular middle platform is set to zero. (3. the assembly has a state of selfstress only for h = 0 but with the present tolerance for the singular values the range of prestressability is about −0. The mechanisms in D are orthogonal but not necessarily symmetric with respect to the geometry.12 T √ √ 0 0 3/2 0 3/2 −4/ 17 −1/2 0 1/2 0 0 0 √ 0 0 0 −1/ 17 0 0 √ √ − 3/2 0 0 0 2/ 17 0 √ √ 1/2 1 0 0 −2 3/ 17 0 √ 0 0 0 0 −1/ 17 0 √ √ 0 0 − 3/2 0 0 2/ 17 √ √ 0 −1 −1/2 0 0 2 3/ 17 √ 0 0 0 0 0 −1/ 17 .e. while the length of the other type of element is not allowed to vary. DISCUSSION √ 0 0 t1 −W t2 0 t3 = 0 .01. hence assembly type II. but this will not be used here.12 4. i. cf. This need to be considered when analysing the tensegrities in the following chapters.
is an eﬃcient numerical method which makes use of the advantages of the SVD.4. as the set of force densities is varied. The main strength of the force density method is that it is well suited to situations where the lengths of the elements of the structure are not speciﬁed at the start. Thus. The linear force method. nodal coordinates.15 0. σ55 and σ66 . The three remaining static methods.05 0 0. e. suitable only for simple or very symmetric structures. with the height h.CHAPTER 3.13: The variation of the two lowest singular values.25 h/l Figure 3. However.25 0. Each category includes an analytical method. i.g. as it is presented here. sections 3. neither of the two can be applied to problems that are not completely deﬁned.1 0. are in fact only two. with certain required characteristics.05 0.2 0. when the lengths of all cables are not known in the formulation where the lengths of the struts are maximised. but it is diﬃcult to control the variation in the lengths of the elements.1 0.2 0. of structures that were already essentially well known.2–3. The reduced coordinates method oﬀers a greater control on the shape of the structure. but involves more extensive symbolic computations. is useful only for 62 .3.e.3. new conﬁgurations can be easily produced. The nonlinear optimisation approach and the method of pseudo dynamical relaxation have both successfully been used to determine conﬁguration details. since the force density method and the energy method are equivalent.15 0. ANALYSIS METHODS FOR TENSEGRITY STRUCTURES 10 10 10 10 0 σ55 /σ11 2 4 6 σ66 /σ11 σii /σ11 10 10 10 10 10 10 8 10 12 14 16 18 0. for analysis of the structure subsequent to formﬁnding. It should be emphasised that the method.
A FEM. DISCUSSION structures that behave linearly. This will be used in some of the following analyses. suitable for cable structures where cable slackening can occur. is presented in reference [171].3. In cases where the structures undergo large displacements. 63 . a geometrically nonlinear FEM must be used.6. this is suﬃcient for many problems. but not all. Now that the necessary tools to analyse a general tensegrity structure have been provided it is time to move on to the development of deployable tensegrity structures for space applications. As will be seen later.
.
η = h/H. Counting the number of joints and bars in the mast yields j = 2vn (4. [151]. The direction of rotation of the prisms vary so that every second prism is rotated clockwise and every other counterclockwise.1 Background In Snelson’s US patent “Continuous tension.1) 65 . This mast is created by assembling triangular prisms on top of each other.1. Fuller presents a tetrahedral mast in his patent from 1962. discontinuous compression structures” from 1965. constructed according to the scheme in Figure 4. The height of each module is H. An important characteristic of Snelson’s tensegrity structures is that they have a single state of selfstress.1 for a threestage mast.2 Static and Kinematic Properties The ﬁrst step in the analysis of any bar framework is the determination of its static and kinematic properties. 4. They merge into a mast by substituting their individual base cables by the saddle cables. less interesting from the point of deployability. as has been shown by Snelson. One of these structures is a mast with three struts per stage. This procedure is illustrated in Figure 4. the construction of highly complex tensegrity structures by simple modules is described. therefore. [117].Chapter 4 Deployable Tensegrity Masts 4. Hence. It can be shown that for a cylindrical mast this overlap can be expressed as a ratio of the module height. This mast has struts that are connected to each other and is. [44]. but the height of the mast is lower than 3H due to the overlap h of the saddle cables. Consider an nstage tensegrity mast with v struts per stage. the length of only one element has to be adjusted to prestress the structure. This is a key property to the practical implementation of tensegrity structures. cf.
2). (4.1: Assembling a threestage tensegrity tower with three struts per stage from basic tensegrity modules: (a) three modules are (b) assembled by replacing the cables of the bases with saddle cables and ﬁnally (c) adding diagonal cables to prestress the structure.3) Substituting (4. m − s = 2v − 6. under the assumption of only one state of selfstress. DEPLOYABLE TENSEGRITY MASTS Top cables z z = H(3 − 2η) cab les z = H(2 − η) Vertical cables St ru ts Sad cab dle les z = H(2 − 2η) Dia gon al Stage 3 Stage 2 z=H z = H(1 − η) Stage 1 z=0 y x Base cables (b) (c) (a) θ Diameter D (d) Figure 4.2) and c = 6 into the extended Maxwell’s rule. 66 . The middle module is rotated counterclockwise the same angle.t.r.CHAPTER 4. yields which is independent of the number of stages n. s = 1.4) For most applications the stiﬀness is important. the number of mechanisms is m = 2v − 5. (4. The top base of the upper and lower modules is rotated an angle θ w. the bottom base. Hence.2) (4.1). (4. and b = 2v(3n − 1). (1. hence the masts with three struts per stage are preferable as they have the lowest number of internal mechanisms.
However. [159]. Nishimura [117]. From a practical viewpoint it would be more diﬃcult to provide adequate restraints at the base. one might think that the masts with twostruts per stage would be better. The extended Maxwell’s rule for this mast yielded m − s = 0. 4.5) where R1 and R1 are the base and top radii of the ﬁrst stage.5) simpliﬁes to η 2 cos π 1 +θ − 3 2 − η cos π 1 +θ + 3 2 67 + cos π + θ = 0. Sultan [159] presents the ﬁrst mathematical treatment of Snelson’s multistage tensegrity mast with three struts per stage. hence the same number of mechanisms as the masts with three struts per stage. the focus will.6) .1). Some special cases of (4. respectively. closer examination of the masts with v = 2 yielded j = 4n. Therefore. η= π 1 + 2 cos v 2 cos and (ii) for v = 3 and R1 = R1 .3. Following the pioneering work by Sultan.3 FormFinding Snelson [151] has built highly complex tensegrity masts for several decades. 4.1 TwoStage Tensegrity Mast Sultan. FORMFINDING Another aspect is to keep the total number of struts low as they will comprise the major part of the mass of the mast. Shortly thereafter. The masts with three struts per stage have triangular bases which can easily be constrained. derives the following general equation for the overlap ratio η of a twostage tensegrity mast with vstruts per stage: π η 1 − (1 − η) cos R1 π v cos . Nishimura [117] presents closedform solutions for the equilibrium conﬁguration of multistage tensegrity masts with v struts per stage. The mathematical conditions for the existence of a prestressable conﬁguration have. Symmetry implies R1 = R2 and R1 = R2 . +θ = π R1 v 2 1 − cos (1 − η) + η 2 v (4. (4. therefore.3.4. for this case with two struts per stage. be on masts with three struts per stage. 3 (4.7) (4.5) are noted: (i) for θ = 0 and R1 = R1 . and θ the angle of relative rotation. shows that the initial equilibrium solution of a cylindrical twostage tensegrity mast with three struts per stage and equal stage height. reduces to a quadratic equation in η with θ as the only remaining variable. the overlap is π v . but b = 6(2n − 1) as the bottom and top bases now only consist of a single cable each. In the following. for the majority of these masts. been unknown. in accordance with (4.
2 5 2 10 0 0.4 0. 68 . The mast conﬁguration was then generated by the scheme in [159].25 0. which was implemented by the author and H.2 0. E. 0.35 0.3 15 10 0.5 .3 0.2 MultiStage Tensegrity Masts Sultan [159] uses a symbolic approach to ﬁnd the overlap of multistage masts. In order to analyse masts with a very large number of stages.2: Contour lines of the dimensionless overlap η for multistage tensegrity masts with three struts per stage. which solves a system of nonlinear equations by the method of least squares.1 25 20 0. The initial equilibrium conﬁguration of the tensegrity mast is the solution to (3.4 0.B.6) indicates that the overlap span of the twostage mast is quite small: η = 1/2 for v = 3.35 0. The determinant of the 6(3n−1)×6(3n− 1) equilibrium matrix H was then computed (N.1 0.3 0. starts with an initial value for η. and η → 2/3 as v → ∞. Hence. i.15 0. The Matlab [89] function fsolve. Y. even though symmetry conditions are used to reduce the size of the matrices.05 35 Number of stages 30 0.3. As the number of stages increases.05 0. say 100. det H(η) = 0. 25 5 0 5 10 15 20 25 30 θ(◦ ) Figure 4.45 35 04 0. a numerical approach based on the force method is suggested here. This approach.4 0. v = 3). masts with up to nine stages are analysed in [159]. the task is to ﬁnd the overlap η which renders the equilibrium matrix singular.CHAPTER 4. f = 0. 4.1 5 0. DEPLOYABLE TENSEGRITY MASTS Equation (4. the computations becomes too complex for the symbolic mathematical softwares.05 0.15 0. 50 0. Pak.1 0.25 0.2 45 40 0.51) in the absence of external forces.e.
η∗ = π 1 + cos v cos θ∗ = π 1 1 − 2 v . Care must therefore be taken in choosing the initial interval. based on successive interval bisectioning. 69 .4. To analyse masts with more stages. The overlap values of these masts are given in Appendix A and Figure 4. and the overlap ratio η (= η ∗ ). The fact that the overlap decreases with the number of stages may not be ideal from a manufacturing point of view as new stages cannot be added without changing the geometry of the complete mast. Instead of implementing a sophisticated linesearch routine. was implemented.9) while if η ∗ = 0. masts with up to 15 stages could be solved. Investigating multistage tensegrity masts. With the fsolve function.2. (4. the rotation angle of the interior stages θ∗ . but only one with the correct prestress. Note that the rotation angle of the second and second to last stages is the same as that of the interior stages. Recall that for a given relative rotation θ. Several overlap values may exist.4). Nishimura [117] ﬁnds that even though all symmetry groups are used to simplify the prestressability condition.10) which is identical to the relative rotation of a tensegrity prism with i = 1. but the termination tolerance TolFun had to be set as low as 10−40 to achieve convergence. which give a zero determinant.8) = 0. The geometry of a cylindrical multistage tensegrity mast with the same selfstress for interior stages and constant stage height (H = H ∗ ) is thus described by three parameters [117]: the rotation angle of the ﬁrst and last stages θ. An interior stage is deﬁned as any stage but the ﬁrst and the last stage in a mast [117]. (3. the relationship between the overlap ratio η ∗ and relative rotation θ∗ is [117] η ∗ 2 cos π π − cos + θ∗ v v + η ∗ 1 − cos π π 1 + cos + θ∗ v v π π + θ∗ 1 − cos − cos v v (4. the overlap ratio η decreased as the number of stages increased. If θ∗ = 0. Nishimura shows that it is possible to keep a constant overlap ratio for any number of stages. no analytical solution could be obtained for masts with more than four stages using current symbolic software. Nishimura then investigates a class of tensegrity masts with the same selfstress for the interior stages independent of the number of stages. For the interior stages. This approach may not be the most elegant way of ﬁnding the initial equilibrium conﬁgurations. The equilibrium conﬁgurations for masts with up to 50 stages were computed by this improved numerical method. By having diﬀerent rotation angles for the interior and the ﬁrst and last stages.3. another approach was needed. Figure 4. The convergence was slow but stable. It is observed that the overlap diminishes with the number of stages. which also may fail.2 shows a graphical representation of these values. FORMFINDING was used to solve det H(η) = 0 for η. [29]. but it nevertheless yields the correct solution. a more rudimentary routine. (4. π v . The state of selfstress also changes with n and θ. cf.
CHAPTER 4. However. This is in contrast to the conventional way of construction.. i. Note that it is possible to design multistage masts with θ = θ∗ and constant η. An obvious disadvantage of the conventional method is the diﬃculty in retaining the required precision when the whole structure has to be constructed in three dimensional space. R1 3 − cos v (4. where. The ﬁrst step in the manufacturing procedure is to map the threedimensional net of cables onto a twodimensional plane. The separation of the tension and compression members would result in more freedom in the manipulation of the ﬂexible cables.4 Manufacturing Technique The practical implementation of tensegrity masts requires an eﬃcient and accurate manufacturing technique. Micheletti [93. 94] shows that the rotation of the end stages is actually found using the prestressability condition for the twostage mast.11) The multistage tensegrity masts by Nishimura are better suited for applications as the forces are relatively uniform through the mast. is especially awkward. which means that the saddle cables had been inverted. The present manufacturing technique was inspired by Pugh’s [132] illustration of the diamond pattern system.5) with R1 = R1 . a mast is built from node to node and stage by stage.2(a) and (b) for a twostage mast and in Figures B.. because of the intrinsic diﬀerence between the cables and struts in terms of material properties. it was decided to separate the construction of the tension and compression members. To ﬁnd the geometry of a cylindrical tensegrity mast with uniform interior selfstress. which in the case with ﬂexible structures. At ﬁrst. Net 2 agrees 70 . Then θ is computed by (4. cf. the base and top radii have to change according to (4. (4. R1 /R1 = 1. symbolically or numerically. the number of net conﬁgurations is restricted. net 1 and 2 for the twostage mast seemed equal.8). For example. However. These conﬁgurations are shown in Figures B.3(a) and (b) for a threestage mast. Conceptually. the characteristic equation of a 7 × 7 matrix (Equation (41) in [116]). exist. 4.5).g. but the nodes of these masts do not all lie on the surface of a cylinder. which otherwise would require complicated joint designs. as done by Pugh. A direct analogy to the proposed manufacturing method is the separation of tensile and compression members in the construction of tension roof structures. It was found that only two cable net conﬁgurations. Figure 4.e. the structure is easy to make since the compression members do not touch each other. without changing any cable lengths.5). DEPLOYABLE TENSEGRITY MASTS After ﬁnding the interior overlap ratio η ∗ by (4. the relative rotation of the ﬁrst and last stages is subsequently determined by solving. However. close scrutiny of net 1 showed that the diagonal cables had changed place with the vertical cables.3. [171]. θ = θ∗ = 0 gives R1 2 = π. Since the cable net is composed mainly of triangles connected to each other in a special way. e. one speciﬁes θ∗ to ﬁnd η ∗ by (4.8). which preserve all cable lengths.
Figures B.3(c). The changes in length seemed to depend on the way the cables were looped around the nodes. which forces the nodes to rotate to adopt to the threedimensional conﬁguration. three struts per stage mast. the connections were secured by epoxy resin. By relaxing the length preservation condition slightly.2(c) and B. which is highly undesirable.3(b). the angles between the members in net 3 should agree better with those in three dimensions. it was soon realised that because there existed more than two nodes with an odd number of cables connected to them. The direct cause was the use of net 1. MANUFACTURING TECHNIQUE more with the layout of threedimensional version of the cable net. by simple Graph Theory. The horizontal distance between the nodes along the saddles is identical to the length of the base cables. The two key aspects of the manufacturing method required to obtain good precision are to ensure accurate element lengths and to make certain that the angles in the twodimensional net do not diﬀer too much from the corresponding angles in three dimensions.4. This setting yielded that the distance between the nodes at the bases was too long and. so an attempt was made to ﬁnd a way of completing the entire tracing of the net with only one or two separate cables. this layout has overlapping triangles. could not be constructed along with the rest of the net. However. it was impossible to use one cable only. the looping direction decided if the cable lengthened or shortened.3: Illustration of the Pugh’s diamond pattern systems for a threestage. In this respect net 2 is better. However. the saddle overlap in the twodimensional cable net was set equal to ηH. but still not ideal. [132]. Subsequent to the tracing of the net. A B C D D to D to B joins to A to C Figure 4. Net 1 may satisfy the ﬁrst requirement but certainly not the latter one. Holes were then drilled at all intersection points and small threaded rods were positioned in the holes. Considering accuracy issues. Unfortunately.1(a).4. According to [118]. despite accurate cable nets. The construction of three tensegrity masts by Pak and the author is described in reference [118]. Stiﬀ Kevlar cords were strung between the intersection points and fastened by looping the cords once around each threaded rod. but it was not the heart of the problem. 71 . which produces an overlap in two dimensions that is larger than that in three dimensions. thus. In hindsight. It is reported that these masts suﬀer from large cable length inaccuracies. For a threestage mast. It was identiﬁed as sensible to have as few openends of the cables as possible. the length errors were caused by uneven lengthening or shortening of individual cables. Figure B. net 3 is the preferred layout. Figure B. These nets were constructed by ﬁrst gluing the plotted cable net to a wooden board. the deﬁciencies of net 1 was not recognised until all three masts had been constructed.
diﬃcult to design and will easily fail.4. the experience from the previous constructions indicated that the connections must be manufactured to a higher precision. the cables of the twodimensional cable net should lie in the same plane and go through the centre of the nodes. It was immediately recognised that the eightstage mast was far more accurate than the previous models. however. The aluminium tubes had been predrilled with eight 1 mm coplanar holes. By adjusting the distance between the nodes in the twodimensional net according to the layout of the holes in the spherical joint it would still be possible to construct the net in two dimensions. The modiﬁcation of the struts to enable mast folding and deployment will be presented next. in the threedimensional net they do not necessarily intersect. Thus. which eﬃciently bonded to the thin steel cable. approximately normal to the board. To eliminate this error source it is necessary to manufacture joints. DEPLOYABLE TENSEGRITY MASTS To examine the feasibility of net 3. preferably spherical.CHAPTER 4. The cables were made of a 0. Thus. 1 72 . Figure 4. After the tracing of the cables. However. where the forces can be in equilibrium.45 mm diameter stainless steel cable composed of 49 individual wires for maximum ﬂexibility. Two nuts were placed on the threaded rods to enable length adjustment. from the experience from earlier models. Preferably. The net for the eightstage mast was plotted on an A0 paper (1189×841 mm2 ). an eightstage mast was constructed by the author. However. were drilled through the board at the cable net intersections. using almost the same method as the previous masts.e. the geometry of the mast will change to a conﬁguration where they intersect. 6–8 May 2001. surprisingly large length adjustments were needed to prestress the nets. The struts were made of the same aluminium tubes as the connections. No net was. To summarise. One reason for the inaccuracies is shown in Figure 4. Nevertheless. which was to separate the construction of the cable net and that of the bars. looping of the cables around the nodes must be avoided. since they are usually large. were positioned in the holes. in the direction of the cables through the centre of the tube.4(b). which was fastened on a 40 mm thick Medium Density Fibreboard (MDF). all the cables in the ﬁnal net would lie in the same plane and each cable would go through the centre of the node. Holes of 4 mm in diameter. However.4(a). constructed according to this principle. with holes drilled in the directions of the cables in the threedimensional net conﬁguration. In the twodimensional net. 44 mm in length. i. but the cables will no longer lie in the same plane1 . 14 mm long aluminium tubes of 6 mm outer and 4 mm inner diameter were glued. E. One of the initial aims was to make a mast with the minimum number of adjustment mechanisms. This solution was pointed out to the author by R. it was considered necessary to include some adjustment possibility. To the parts of the rods sticking up from the board. the cables connected to a node lie in the same plane and go through the same point. Threaded 4 mm rods (M4). the connections were secured by ﬁlling the aluminium cylinders with highstrength epoxy resin. worked well after some early mistakes. Figure 4. the basic idea of the construction scheme. Skelton at “Colloquium Lagrangianum— Strutture Tensegrity: Analisi e Progetti” (Tensegrity Structures: Analysis and Design) in Rome.
but such a scheme will probably lead to problems as it is no longer obvious how the struts should be stacked for a troublefree deployment. For modes (i) and (ii). Second. a large number of devices is needed to control the lengths of the cables. the focus would therefore be on the struts. In the few studies on foldable or deployable tensegrities. If the struts are stacked vertically the diameter need not change. although the latter one produces a more compact package. each strut 73 . the size of the triangular bases changes simultaneously during deployment. using a cableactivated deployment means that the diameter of the mast will increase if the struts are stacked horizontally on top of each other in the stowed conﬁguration. In addition. There are. only the lengths of the strut are changed and in the second one. In the ﬁrst mode. this has not always been the case. a few disadvantages with this approach. however. The cable mode is found to be less complex than the strut mode.5. only the cable lengths.4: Cables which coincide in two dimensions (a) may not coincide in three dimensions (b). First.2. the lengths of the cables are changed in a way so that the structure at every step is in an initial equilibrium conﬁguration. 162].4. Bouderbala and Motro [11] analyse diﬀerent approaches to tensegrity folding: (i) strut mode. In the deployment approach by Skelton and Sultan [159. Furuya [45] analyses three approaches for the deployment of a tensegrity mast with connected struts: (i) ﬁxed lateral cable length. (ii) cable mode and (iii) mixed mode.5 Deployment An nstage tensegrity mast with three struts per stage has 5−2/n times more cables than struts. In the third mode both lengths are changed. The folding of expandable octahedron assemblies are studied. 4. DEPLOYMENT (a) (b) Figure 4. which usually is more convenient for space applications. Telescopic struts are suggested for realising the mechanisms of this mode. Only in mode (iii) is it possible to deploy the mast sequentially. the structure is in stable equilibrium throughout the deployment. As observed in section 2. the popular masts for space applications have a constant diameter during deployment.4. (ii) ﬁxed strut length and (iii) ﬁxed base cable length. Hence. Concerning deployment and packaging eﬃciency. the struts are the only stiﬀ members and also the longest ones.
2 kg. a much lower moment is required to continue folding the tape. increase the stack height. This is a drawback. A key property of the tapespring is that it can be signiﬁcantly deformed several times without permanent damage. The resulting hinge will be quite stiﬀ in bending but weaker in torsion.5. some parts of them will lie on top of each other and. In order to fold these tapes. i. Its basic components are two tapesprings and two Rolamite hinges2 . In the search for a deployment procedure that is simpler than the cableactivated approach. USA. this deployment procedure is neither synchronous nor sequential but rather a combination of both.105 kg [181]. or the rollingband. a space qualiﬁed acetal resin. Having so many motors presents too large a risk. the The Rolamite. A tapespring hinge. [19] 3 Delrin is a registered trademark of E. one on each side of the tapesprings. has been developed at DSL. Subsequent to buckling.5. A telescopic alternative. thereby. the tapes provide a small but constant restoring moment to eventually lock into the straight position.e. and also the restoring moment. du Pont de Nemours and Company. which also is torsionally stiﬀ. As described in reference [149]. a procedure based on strut deployment will be proposed. Wilkes of the Atomic Energy Commission’s Sandia Laboratory. two tapesprings can be placed a certain distance apart with the concave sides facing each other. Including attachments to the struts the TSR hinge has a total mass of 0. The motors and the reels produce a very reliable cableactivated deployment. If the struts are longer. a large moment is required to initially buckle the tape. would require a motor on each of the 3n struts in the mast. the curved metal tape found in tape measures has been considered for a long time in the design of selfdeploying hinges [128]. with approximately half the density of aluminium. I. To increase the buckling moment. the struts can ﬁt horizontally beside each other in a canister with a diameter equal to that of the mast. or tapesprings. Other advantages of this hinge is simple assembling and low friction. The overall weight of the hinge depicted in Figure 4. a hinge with selfdeployable characteristics would be preferable. The failure of only one motor would end the deployment. but hopefully equally reliable. F. the device consists of a metal band looped around two rollers. TapeSpring Rolling (TSR) Hinge The use of carpenter tapes. The Rolamite hinges are made of steel cables and Delrin 3 .CHAPTER 4. DEPLOYABLE TENSEGRITY MASTS has one electric motor to control the length of the saddle cables and three reels on which the vertical and diagonal cables are stored to prevent entanglement [149]. Albuquerque. concept was invented by D.1 Strut Deployment If the strut length is equal to the diameter and folded by a midpoint hinge. NM. In its basic form. Figure 4. Thus. 2 74 . During deployment. but they add complexity to the mast. although not too serious.5 is 0. however. 4. Telescopic struts can be collapsed into shorter lengths.
one on the top base cable and the other on the bottom cable. Once folded into a bent conﬁguration. while it is strained in the rolledup conﬁguration. Another. cf. two with three stages and one with four stages. BiStable Tube Hinge A wellknown property of tapesprings is that they are stable only in the straight position. Watt). was invented by DatonLovett [30. A strut with a hinge made of a bistable tube is shown in Figure 4. In the straight conﬁguration the tube is unstrained. More information about the TSR hinge is found in [180. DEPLOYMENT (a) (b) Figure 4. which is stable in both the bent and straight positions. was built by H. latter eliminates the need for lubrication. A fourstage tensegrity mast.6. 65]. The mast was prestressed by two adjustable plastic ties. more distant. they need external restraints to keep them in that position. The bistability is obtained by a special layout of the ﬁbres in the composite material. E. [118].5: The tape spring rolling hinge: (a) deployed and (b) folded (Courtesy of A. with bistable hinges. The ﬁrst threestage mast was not foldable and was only made to gain some experience with the manufacturing procedure. Pak and the author. were built at DSL.2 Demonstrator Masts As mentioned earlier. The torsional stiﬀness can be increased by closing the open crosssection with velcro [30]. alternative to bistable hinges is to manufacture the whole strut as a bistable tube. The second threestage mast had struts made of aluminium rods and singleblade tapesprings as hinges. three tensegrity masts.5.g. To move the tube from the rolledup position to the straight position requires only a small quantity of energy. the BRC tube. M. When the mast was 75 . e. Recently. Y. 181]. the cassette containing the rolledup carpenter tape. 4.5.4.
This ﬂattening signiﬁcantly reduced the stiﬀness of the curved tape and yielded the mast useless. However. thanks to the bistability of the hinges. For the following masts. The mast could easily be folded by hand. was the ﬁrst one that could be folded in a satisfactory way. In order to achieve increased axial stiﬀness. which led to local buckling of the tubes. that functioned well.CHAPTER 4. several struts buckled in torsion. The folded mast is shown in Figure 4.8. FourStage Mast The fourstage mast. the rods were attached to the outside of the tubes.6. The struts were made of solid 6. high local stresses were induced at the connections between bistable composite tubes and the aluminium rods. had turnbuckles at the ends that enabled length adjustment.35 mm diameter aluminium rods and bistable tubes were used as hinges. the mast with the bistable struts was much heavier which led to cable slackening. prestressed. However. The bistable tubes are disproportionately large in comparison to the complete mast. The mast was reasonably straight considering the accuracy problems of the cable net discussed earlier. Figure 4. The torsional buckling problems with the previous model was eliminated by the bistable tubes. which had a length accuracy of ±1 mm. DEPLOYABLE TENSEGRITY MASTS Figure 4. no bistable tubes of smaller diameter. Compared to the mast with tapespring hinges. by shortening the top and bottom cables. This signiﬁcantly reduced the compressive strength of the struts and the mast could not be prestressed to the desired level. 76 . The struts. were available at the time. The deployed mast is shown in Figure 4. which was the third tensegrity mast built at DSL [118].7. Closer examination revealed that the blades had been ﬂattened by the poprivets which connected the blades with the rods.6: Strut with a bistable hinge in straight and bent conﬁgurations. a heavier boltandnut alternative was used instead of poprivets.
4. the eightstage mast was equipped with tapespring hinged struts.7: The deployed fourstage mast with bistable struts. (a) (b) Figure 4. DEPLOYMENT (a) (b) Figure 4.8: The stowed fourstage mast with bistable struts. EightStage Mast To fully evaluate the concept of selfdeployable struts. Two tapesprings (19 mm wide) were 77 .5.
The stiﬀness of the ADAM is presumably higher than that of the tensegrity mast. similar to that used for the CM. DEPLOYABLE TENSEGRITY MASTS connected to two aluminium tubes. The ADAM supports a 360 kg antenna at its tip and carries 200 kg of electric and ﬁbreoptic cables and a gas line along its length [1]. but none do a comparison with existing deployable masts. [45.117.e.CHAPTER 4. However. Another problem was how to control the deployment of the mast. One immediate problem was how to keep the struts in the folded conﬁguration. the longer base cable is shortened by a motorised turnbuckle and the mast is prestressed. deployable structures. as it uses stiﬀ square frames between each bay.4. Data for the STRM ADAM is given in Table 4.7. However. Finally.10 and 4. When the rod was pushed from underneath the base. talk about the potential of the tensegrity masts as lightweight. e. with the concave sides of the tapesprings facing each other. to prestress the whole structure. Twice the deployment had to be stopped to resolve some entanglement problems. the slender aluminium rod could not provide enough force and the tapespring hinges could not provide suﬃcient moment to deploy the bottom stage. the new hinges were much stiﬀer in both torsion and bending. i. Recall that the tensegrity masts with three struts per stage are kinematically indeterminate to one degree. Figure 2. The mast was not fully prestressed until the bottom stage was fully deployed.9. Hence. This stiﬀening approach will be tested here. The tube would activate the deployment and stabilise the unstressed mast during deployment. An approach suggested by Pellegrino [126] was to use an inﬂatable tube inside the mast. it deployed the mast sequentially.1.6 Structural Analysis Several studies. This must be regarded as very satisfactory considering no precautions had been taken to avoid entanglement.g. All the tapespring hinges can then be deployed since they do not have to prestress the mast.159]. 78 . For the present mast. Their stiﬀness is therefore dependent on the prestress level. The current stateoftheart mast is the primary deployable structure for the STRM— the 60 m long ADAM. 4. A ﬁnal prestressing of the mast can be obtained by allowing one of the base cables to be a little longer during deployment. the inﬂatable tube was replaced by an aluminium rod through the base of the canister and connected to the three nodes at the top of the mast. it is possible to remove the inﬁnitesimal mechanisms of the tensegrity mast by adding cables in such a way that completely triangulates the bottom stage. To satisfactorily prove the applicability of tensegrity masts as eﬃcient deployable structures. Figure 4.11. The mast could thus be given the necessary prestress to reduce the eﬀects of gravity. This was solved by enclosing the stowed package by a canister. they must be compared to a realised and successful mast. The complete deployment of the eightstage mast is shown in Figures 4. Compared with the tapespring hinges of the second threestage model. Figure 2. only two motors would be needed: one that shortens the base cable and one that deploys the structure by actuating and controlling the mast deployment.
6975 m. the maximum and minimum cable and strut forces versus the number of stages. For D = 1. rotation angle θ and overlap η. are shown 79 . [118]. Then. 4. for a mast with equallength struts.9: Eightstage mast stowed in a canister: (a) side view and (b) top view. a mast with uniform interior forces was analysed. the relationship between the bay and strut length must be found.12) The length of the struts at the end stages for a mast with uniform interior forces is found by setting θ∗ := θ. ls . one of the input values is the strut length. First. cf. is computed as: 1/2 D2 √ 2 + cos θ∗ + 3 sin θ∗ + ls = 4 2 Hbay 1 1−η 1− n 2 . bay length Hbay . STRUCTURAL ANALYSIS (a) (b) Figure 4. a tensegrity mast with struts of equal lengths was studied. To generate a mast with a given bay length. In the present mast generation routine.1 Initial Equilibrium Element Forces An appropriate initial equilibrium mast conﬁguration should have fairly uniform internal forces. for given values of the mast diameter D.12 m and Hbay = 0. For an nstage mast with three struts per stage. In order to facilitate an accurate comparison.4. the length of the struts.6. (4. length and number of bays as the ADAM. the tensegrity mast should have the same diameter.6.
DEPLOYABLE TENSEGRITY MASTS 80 Figure 4. .10: Rodcontrolled deployment of the eightstage mast: ﬁrst four stages.CHAPTER 4.
6. STRUCTURAL ANALYSIS Figure 4.81 4. .11: Rodcontrolled deployment of the eightstage mast: last four stages.
6975 87 290 0.0228 0.0059 0.12 0. This is a serious disadvantage and the only conclusion must be that long masts with equallength struts are unfeasible for applications.7 Stiﬀness Bending stiﬀness (MNm2 ) Torsional stiﬀness (MNm2 ) Shear stiﬀness (MN) First bending mode (Hz) First torsion mode (Hz) Strength Bending strength (Nm) Torsional strength (Nm) Shear strength (N) 13 0.CHAPTER 4.10 0. 20 10 0 ti 10 tbase 20 θ = −10◦ θ = 0◦ θ = 10◦ θ = 20◦ 30 40 0 5 10 15 20 25 30 Number of stages 35 40 45 50 Figure 4. the relationships between element forces are undesirably large. The overlap values were taken from Appendix A.12: Maximum and minimum forces in the cables and struts in a multistage mast with struts of equal length.49 0.1: Data for the 60 m long STRM ADAM. Already for seven stages. the maximum strut force was ten times the force in the base cables at θ∗ = −10◦ . Geometry Length (m) Diameter (m) Bay length (m) Number of bays Mass (kg) Mechanical stability Bending (◦ /N) Twist (◦ /Nm) Axial (µm/N) 60. DEPLOYABLE TENSEGRITY MASTS Table 4.68 1. The maximum cable and strut force increased linearly with the number of stages while the minimum cable force decreased only slightly. [1].17 8140 305 400 in Figure 4.12. in a multistage mast with equallength struts. 82 .15 0. Hence. The maximal and minimal forces were normalised by the force in the base cables.
In the following analysis of the Nishimura mast it was assumed that each stage had the same radius.14) yields P = 1 1 ∆ + Ml/2 Pcr . l (4.13) where CM is the spring stiﬀness and l the column length. The total number of diﬀerent force values is ten. their design is therefore of most importance. η was found by (4.14(a). height.5) and (4. diagonal. Figure 4. For a given value of the interior stage rotation θ∗ . was computed by (4. It was immediately noted that the forces were lower than those of the equallength strut masts.6.15) 83 . and saddle cables (N.5).B. Hence. θ. with the corresponding lateral midpoint displacement ∆ ≈ ψini l/2. vertical. Rearranging (4. η ∗ = η. As the struts most likely would be slender. the element forces in an nstage mast (n ≥ 5) was generated by two sets of element forces. The midpoint moment Ml/2 in the imperfect column due to the compressive load P is Ml/2 = P ∆ 1 1− P Pcr .8) are valid for any number of stages.8). Ri = Ri .6. Subsequently. As before. The buckling load of the column is [36] Pcr = CM . a suitable conﬁguration should have uniform element forces.4.14) where Pcr the buckling load according to (4. As (4. the ﬁrst set corresponds to stages 1 and n and the second set corresponds to the intermediate stages. Now consider that the column has an initial imperfection. diagonal and saddle cables at the interior stages. This remarkable characteristic signiﬁcantly simpliﬁed the analysis of multistage masts. the rotation of the end stages. (4. These requirements restricted the interval of possible solutions to θ∗ = 0–10◦ . 4. H ∗ = H and overlap ratio. their load bearing capacity is governed by buckling. is probably more suitable for practical uses. ψini . Set 1 is the struts and base. which has uniform forces for the interior stages. (4. Therefore. two values) at the ﬁrst (or last) stage and set 2 the struts and vertical. The variations of the normalised forces with θ∗ are shown in Figure 4.2 Preliminary Design of Struts and Cables The struts of the mast with uniform interior forces were subjected to the largest forces.13). Consider a perfectly straight column of two rigid bars and a rotational spring at the connecting hinge.13. STRUCTURAL ANALYSIS The mast conﬁguration by Nishimura [117]. the forces in an nstage mast were determined by analysing a ﬁvestage mast. the magnitudes of the forces in the mast were restricted: (i) no cable force was allowed to be lower than that in the base cables and (ii) the maximum compressive force in the struts could not be larger than ﬁve times the tension force in the base cables.
Considering the extremely high cost of launching a spacecraft. Formulae for predicting the stiﬀness indicate that the latter hinge has higher stiﬀness than the former one. the error arising from the alignment of the hinge attachments is probably more severe. CM . no safety factor against buckling was considered in the following analyses and sometimes the compression load was allowed to be higher than the buckling load in order to study the eﬀects of increased prestress. which yields Pcr = 328 N using (4. ∆ = 3. In reference [181].1 = 1. the buckling moment.12 m. and rotational stiﬀness. the aim for the strut crookedness was ∆ ≤ l/400. The test for the rotational stiﬀness was done with a hinge of slightly diﬀerent geometry than that which was tested for the buckling moment. the strut length at the ﬁrst stage is ls.13: Cable and strut forces in the ﬁrst (and last) and interior stages in a multistage mast with uniform interior forces.CHAPTER 4. As this value is very low. At θ∗ = 0◦ . In order to obtain a high load capacity. in the following calculation of the load bearing capacity of the hinged strut.13).12). of the TSR hinge are measured to 13 Nm and 480 kNmm/rad.6975 m and n = 87 in (4. the manufactured strut must be as close to straight as possible. Mcr . While the two bars of the strut can be made nearly straight. Nevertheless. D = 1. respectively. The tubes connected to the hinge were made of CFRP with E = 210 GPa and ρ = 1660 kg/m3 [96]. the measured values were used. The minimum wall thickness of the tubes 84 . Substituting the buckling load. DEPLOYABLE TENSEGRITY MASTS 4 3 ver dia gon al 1 sadd le 1 saddl diag tica 2 1 0 ti tbase l1 e inte r ior r strut length vertical interio r ona l int erio 1 2 3 4 5 6 30 strut s1 str in uts ter ior 20 10 0 θ ( ) ∗ ◦ 10 20 30 Figure 4. high manufacturing tolerances are justiﬁed.65 mm and M = 13 Nm into (4.46 m. Hbay = 0.15) yielded P = 300 N.
with l = 1460−88 = 1372 mm. with l = 88 mm.6. and the computed stiﬀness of the CFRP tubes. the struts cannot be too slender.28 kN or about 17P . A strut diameter of 25 mm yielded. [181]. STRUCTURAL ANALYSIS P P δ l 2 l 2 P ψ CM ψini l 2 l 2 (a) (b) (c) Figure 4. The cables were assumed to be made of thin CFRP tape 5 mm wide and 0. imperfect state. Its properties were the same as for the tubes above. which gave an allowable force of 500 N. The eﬀective axial stiﬀness of the strut was computed as AE l AE l AE l TSR AE + l TSR AE l = eﬀ CFRP . The axial stiﬀness of the cables was 0.16) CFRP where (AE/l)TSR is the stiﬀness of the TSR hinge and (AE/l)CFRP the stiﬀness of the CFRP tubes. 85 .4.5 mm thick. With regard to length precision requirements.14: Buckling of a twolink column: (a) initial state. The measured TSR hinge stiﬀness is 7223 N/mm. for a pinended strut of constant tubular crosssection and length ls. By (4. The allowable stress was set to 200 MPa with safety factor between 5 and 10. (4. (b) buckled state and (c) initial. PEu = 5. is 11540 N/mm.16) the eﬀective axial stiﬀness of the strut was 4442 N/mm or 6.525 MN.485 MN for a length of 1460 mm.1 . was set to 1 mm to ensure adequate toughness for handling and assembly [96].
3 Vibration Analysis A few earlier studies have been concerned with the vibrational characteristics of tensegrity masts.17) was rewritten as [47]: ˜ Kd = where ˜ d. 86 .20) K = LT K−1 L. e.CHAPTER 4. does not change signiﬁcantly with the prestress level. The second mode is a bending mode and its frequency is much higher than the ﬁrst one. An nstage mast of this type has s = m = n.22) Although not optimised for eﬃciency. the Matlab function eig was used.18) (4. therefore. this approach required signiﬁcantly less computer time when analysing larger problems. The natural frequencies are computed by the classical eigenvalue problem of the homogeneous linear systems (4. Since the lowest eigenvalues were sought for the present problem. M the mass matrix. For simplicity.21) ω2 The lower triangular matrix L was obtained by a Choleski factorisation of M: M = LLT . Not surprisingly. Furuya [45] analyses a ﬁvestage mast with contacting struts made of simple triangular prisms. i. was evenly distributed along the length of each strut. The eigs function computes the N largest eigenvalues and their corresponding eigenvectors. (4. While the frequency of the ﬁrst mode can be increased by increasing the prestress.2 kg. such as the 87stage mast. Murakami [115] performs a modal analysis of a sixstage mast with v = 3. a global axial mode.e. the mass of each hinge. d the displacement vector and ω the angular frequency. which has a nonzero elastic energy.025 kg. the frequency of the ﬂexural mode. each weighing 0. plays an important part in providing the mast with geometric stiﬀness.19) (4. The prestress. The total mass of these joints was evenly distributed along the total length of the cables and the struts.17) Kd = ω 2 Md. The joints between the struts and the cables were assumed to be 25 mm diameter aluminium spheres. the function eigs was used instead. the frequencies of the zero energy modes increase with the level of prestress. 0. DEPLOYABLE TENSEGRITY MASTS 4.6.g. tenstage mast. (4. where N is a userdeﬁned value. For larger problems. In the present study. For problems with a small number of degrees of freedom. = and 1 . (4. ˜ d = LT d. the mast and element dimensions were those determined in the previous section. where K is the tangent stiﬀness matrix. The fundamental mode is the one corresponding to the internal mechanism. (4. which is characterised by adjacent countertwisting of each stage.
that only the frequency of the mode corresponding to the “axial” internal mechanism.5714 397 194 −776 −623 500 2. but slightly shorter struts by (4.8478 5.975 13. For a cantilever beam.2359 12.037 Hz.15(b) and (c).8468 4. This shift in fundamental mode can be seen in Table 4. However.9618 15. Axial mode frequencies in bold face. the ﬂexural modes were not aﬀected by the increased prestress.23) and solving for EI yielded EI = 2.12). From the opposite point of view.8562 3.7 tbase f1 f2 f3 f4 f5 tmax tmin tmax tmin c c s s 50 1. which can be compared with 0. the lowest bending frequency is [120] f1. for very long masts the fundamental mode is a bending mode and its frequency is independent of the level of prestress. Thus. ml4 (4.0952 99 48 −194 −156 100 1. As before.1 MNm2 . The buckling load of the struts. a continuous increase of the prestress eventually resulted in that the frequency of the fundamental.2819 118. Figure 4. 100. as anticipated. Increasing tbase to 100 N gave a fundamental mode shift at 16 stages. was chosen for further studies on the eﬀects of the prestress level. The ﬁrst two modes were ﬂexural modes with frequencies of 0. axial mode becoming higher than those of the two ﬁrst ﬂexural modes. θ∗ = 10◦ . f1. The results for the 87stage mast are shown in Table 4. This mast had the same diameter and bay length as the former one. were unaﬀected by a change of θ∗ .6260 199 97 −388 −312 200 2. STRUCTURAL ANALYSIS To gain some understanding of the vibrational characteristics of the tensegrity masts. into (4.8214 2.15(a).10 Hz.516 2π EI . a tenstage mast was ﬁrst analysed. First.6138 994 484 −1939 −1559 87 .2412 6. The tenstage mast was analysed for four levels of prestress: 50. One conﬁguration. n θ∗ (◦ ) η Length (m) Mass (kg) mc (kg/m) ms (kg/m) lc (m) ls (m) 10 10 0.1. The frequencies of the ﬂexural modes.1371 12.4.13 Hz for the lowest prestress.6442 2.3006 2.0136 0. These observations agrees with the ones by Murakami [115]. varied with the prestress in the structure (which changes with θ∗ ). However.8581 13.80 kg/m and l = 60.2: Data and results from the modal analysis of the tenstage mast. The frequency of the axial mode was 0.8307 2.8245 2.6. the axial mode was the fundamental mode up to 24 stages for tbase = 50 N. the frequency given in Table 4.2 40. the mast was analysed for diﬀerent values of θ∗ between 0 and 10◦ and a constant prestress of 50 N in the base cables. m = 4. 300 N. 200 and 500 N. was disregarded in this analysis.10 Hz for the STRM ADAM.1 0.cb = 0.6006 9. This analysis yielded. Figures 4.68 m.cb ≈ 3.23) Substituting the data for the STRM ADAM.3. which is about 6 times lower than the value given in Table 4.8499 6.1787 2.8224 2. Note that this particular conﬁguration did not have the highest frequencies but had the lowest internal forces and lowest total mass.2 when going from 200 to 500 N.
24) f1.15: First three eigenmodes of a tenstage tensegrity mast (tbase = 50 N Table 4. DEPLOYABLE TENSEGRITY MASTS (a) (b) (c) and θ∗ = 10◦ ): (a) f1 = 1.0374 0.1324 0. n θ∗ (◦ ) η Length (m) Mass (kg) mc (kg/m) ms (kg/m) lc (m) ls (m) 87 10 0. Inserting the values of the 87stage (or the tenstage) tensegrity mast in (4.2 0. 88 .0134 0.847 Hz.683 114.0374 0.0375 0. By Raleigh’s method the ﬁrst frequency of a cantilever beam with a point mass M at its tip can be written as [120]: 3EI 1 .2340 100 49 −196 −158 100 0. (b) f2 = 2.821 Hz and (c) f3 = 2.0374 0.6 MNm2 .2342 0.23) produced EI = 0.2348 0.2412 60.1 may include the 360 kg antenna attached to the tip. Axial mode frequencies in bold face.9 357 tbase f1 f2 f3 f4 f5 tmax tmin tmax tmin c c s s 50 0. Figure 4.2339 0.0374 0.2646 399 196 −785 −631 500 0. (4.2342 0.1872 0.179 Hz.4174 997 490 −1963 −1577 Table 4.2340 0.2347 0.24) yielded EI = 12.0375 0.3: Data and results from the modal analysis of the 87stage mast.2805 1047.2341 199 98 −393 −315 200 0.0374 0.0374 0.11 MNm2 .pm ≈ 33 2π 3 M + l ml 140 Substituting M = 360 kg into (4. which is much closer to the stated value. the STRM ADAM is more than 100 times stiﬀer than the tensegrity mast in bending.CHAPTER 4. Hence.cb.
but the axial stiﬀness was also computed. (c) bending in direction B1 and (d) bending in direction B2. Therefore.16. its static properties must be investigated. The torsional stiﬀness is identical in the clockwise and counterclockwise directions. bending and torsional stiﬀness of a fourstage mast. as the external forces increase. the focus of this study was on the bending properties of the ten. The bending stiﬀness of the four stage mast is constant until a cable goes slack. the bending stiﬀness increases.e. decreasing colatitude. Skelton and He [149] compute the axial stiﬀness and the buckling load of a 186 m long.6. [150]. As anticipated. the struts become more horizontal.6. and the prestress increases the axial stiﬀness in the case of small external forces. Their ﬁnal observation is that prestress does not aﬀect the bending stiﬀness of the mast provided the cables are taut.e. The special catenary element is advantageous in problems where cable F/3 F/3 F/3 F/3 F/3 F/3 F/3 F/3 F/3 F/3 F/3 F/3 (a) (b) (c) (d) Figure 4.and 87stage masts. the stiﬀness increases with the prestress level. i. The masts were analysed by a geometrically nonlinear FEM. the bending stiﬀness drops and becomes a nonlinear monotonically decreasing function of the applied loading.4 Static Analysis For a complete analysis of the tensegrity mast. 89 . but does delay the onset of cable slackening. The studies above clearly show that the strength and stiﬀness in bending are the critical properties of tensegrity masts. Pak [118] analyses the axial.16: Load cases for the tensegrity mast: (a) tension. Figure 4. As this analysis was not intended to be a parametric study. with the struts modelled by twonode bar elements and the cables by nocompression catenary elements. i. four load cases were considered. the eﬀects of prestress can be neglected.4. Sultan and Skelton [161] analyse the axial and torsional stiﬀness of a twostage mast. sixstage tensegrity mast for a deep sea project. cf. Static analyses of tensegrity masts with three struts per stage have previously been performed by several researchers. The axial stiﬀness increases as the struts become more vertical. Again it is found that the axial stiﬀness is almost equal in tension and compression and increases with the external loading. After a cable has gone slack. [171]. The bending stiﬀness is constant until one or more cables go slack. As the colatitude increases. STRUCTURAL ANALYSIS 4. A more thorough study of the twostage mast stiﬀness is presented by Skelton et al. only conﬁgurations with θ∗ = 10◦ were considered. (b) compression. Following the approach in reference [150].
18 and 4. The resulting mast now had three independent states of selfstress. the stiﬀened mast was stiﬀer in tension although the initial stiﬀness was about the same.38 0. Since the internal mechanism of the masts makes them weak in the axial direction.03 1.25) The α’s were found by equating the element forces in the base cables to unity.12 1 0. the diagonal cable on the compressed side at the second stage became 90 . Among these states of selfstress a rotationally symmetric combination ssym was sought. Three additional cables were added to the ﬁrst stage so that it became fully triangulated. However. The behaviour of the unstiﬀened mast under axial loading was about the same in tension and compression.17.42 0. In the original conﬁguration. slackening may occur.55 0.98 1. One way to prestress the cables was to further rotate the ﬁrst stage so that the struts became longer.98 0. The initial stiﬀness of the unstiﬀened mast was 275 kN/m. DEPLOYABLE TENSEGRITY MASTS 1. In bending. the mast got stiﬀer as the load increased.19 (a) (b) Figure 4.85 2.72 3. a way of stiﬀening the mast was investigated.91 0.97 3.72 1. The response of the stiﬀened mast diﬀered slightly in tension and compression. s = 3.88 1. For the tenstage mast the element forces in the ﬁrst two stages are shown in Figure 4. and no internal mechanism.99 1.19.CHAPTER 4.93 1.99 3.37 1. ssym = α1 s1 + α2 s2 + α3 s3 . 420 kN/m.99 0. Figure 4.82 1 1. The element forces in the mast now changed. which cannot be accepted. (4. Generally.85 1.3 N three diagonal cables on stage 2 went slack (marked as a discontinuity in the curve) and the stiﬀness decreased. at a loading of 16. these cable were unstressed.17: Normalised forces in the two lowest stages of tenstage tensegrity masts: (a) unstiﬀened and (b) stiﬀened.17.14 0. both for a normal and a stiﬀened mast with an additional 15◦ rotation of the ﬁrst stage.43 2. In compliance with the ﬁndings by others. Under compression no cable became slack. The results for the tenstage masts are shown in Figures 4.
2 N for load direction B1. it was not necessary to perform a complete analysis of the 87stage mast. For B2 it was a diagonal cable at stage 1 at the load of 8. the same cable went slack at 11. a lateral load F = 1 N would give a tip displacement of 677 mm.2 N. A further increase of the load resulted in more and more cables becoming slack and the bending stiﬀness dropped suddenly at every occurrence as noted in Figure 4.18: Load–displacement curves for an unstiﬀened and a stiﬀened tenstage tensegrity mast subjected to axial loading. hence the mast is much too ﬂexible.8 N.7 Discussion Using the remarkably simple relationship by Nishimura [117]. After a cable went slack. For B1 the next cable to go slack was a diagonal cable at stage 3 and this occurred at 9.9 N. one of the additional cables on the compressed side of the mast already went slack at 3. The case of the stiﬀened mast was a bit diﬀerent. DISCUSSION 50 ed st if fe n 40 30 20 Axial force F (N) 10 0 10 20 30 40 50 80 60 40 20 0 20 Axial displacement (mm) 40 42 / 0N mm m N/m 275 60 80 Figure 4.0 N. For example. slack at F = 11.4. The bending stiﬀness was about the same in each direction before and after cable slackening. multistage tensegrity masts with an adequate prestress distribution is no 91 . the bending stiﬀness immediately dropped to 22 kNm2 . The bending stiﬀness of the mast with no slack cables was 110 kNm2 .7. for direction B2 it happened at 4. As regards the low bending stiﬀness of the tenstage mast. For direction B1.6 N. as determined by the vibration analysis. 4. the formﬁnding of straight.19. For direction B2.
should eliminate the last obvious source of inaccuracy. Since most of the stages are equal. 92 . The proposed manufacturing scheme eventually turned out to be satisfactory when all the early mistakes had been eliminated. Tapes have a natural folding direction and it is presumably attainable to arrange the cables such that they do not get caught by struts during deployment. One approach to reduce the risk of the cables getting tangled up in the struts is to use tapelike instead of cordlike cables. longer an obstacle in design. it must be possible to give the saddle cables the correct prestress and the saddle nodes the proper restraints in order to prestress the deployed part of the mast. The suggested strut deployment approach worked well. One inherent problem with tensegrity masts are that they do not reach full stiﬀness until the last stage is deployed. This needs to be studied further.19: Load–displacement curves for an unstiﬀened and a stiﬀened tenstage tensegrity mast subjected to lateral loading. Each discontinuity indicates that yet another cable has become slack. with nodes lying on a regular hexagon. However. The use of joints.CHAPTER 4. with holes drilled to the correct threedimensional angles. it might not be that diﬃcult to provide the deployed portion of the mast full stiﬀness. These restraints must take on the role of the struts and the cables of the stage that is next in line to be deployed. DEPLOYABLE TENSEGRITY MASTS 20 18 Lateral force F at top (N) 11 0 kN 16 14 12 10 8 6 4 2 0 0 5 m 22 kN 2 m2 B1s B1 B2 B2s 10 15 20 Lateral displacement at top (mm) 25 30 Figure 4. This seriously limits the applicability of these masts.
the masts are relatively stiﬀ axially but very weak in bending. DISCUSSION The structural behaviour of the multistage mast concur with the predictions of studies by other researchers.7. but the bending stiﬀness cannot be said to have increased at all as the ﬁrst cable went slack at a very low load level. namely.4. The removal of the internal mechanism did not improve the behaviour signiﬁcantly. 93 . the initial axial stiﬀness increased by about 50%.
.
the new antenna structure is introduced. Recently these satellites have been complimented by systems which use several smaller satellites in Low Earth Orbit 1 (LEO) [39]. but they were not A LEO often has an altitude below 1. Microsatellites.1 Small Satellites and Deployable Structures Conventional satellite technology has for several decades been focused on a small number of large. This chapter can be seen as an introduction to reﬂector antennas with a requirement speciﬁcation for the particular antenna. and surface accuracies of existing deployable antennas will be presented. One particular aspect. have been used for technology tests and amateur radio for almost two decades. accuracy characteristics of diﬀerent concepts. are studied. Stateoftheart antenna concepts. A parabolic reﬂector antenna can be given two diﬀerent conﬁgurations: axisymmetric or oﬀset.Chapter 5 Design Prerequisites for a Deployable Reﬂector Antenna This and the following two chapters will provide an indepth description of the development of a deployable reﬂector antenna aimed for small satellites. complex spacecraft. 5. which will be developed in Chapters 6 and 7. the required smoothness of the reﬂector surface. with a mass less than 50 kg.. where the radiation from the Van Allen belt is low (p. The eﬀects of the surface accuracy on the antenna performance. suitable materials for the antenna structure will be selected. the geometries of these conﬁgurations will be presented. will signiﬁcantly aﬀect the choice of structural concept. the background and motivation behind the new concept are described along with examples of other small satellites using deployable structures.000 km.e. 1 95 . Finally. which may be suitable for the present application. In the present chapter. i. This background information is necessary to fully appreciate decisions and judgments made later on in the thesis. 181 in [183]). Then.
1 × 1. e. matching the Earth’s rotational motion. 39]. Figure 5. they are volume rather than mass constrained. the highgain antenna on Galileo. Two of the carrier systems are the Get Away Special and Hitchhiker.g. and several failures [136]. 1995 from the Plesetsk Cosmodrome in Russia as a piggyback passenger on a Kosmos3M launch vehicle to an altitude of about 1.1. Deployable structures have been used on small satellites since the beginning of the space programme. the four solar panels. However. The ASAP on the Ariane5 rocket. along with a large spacecraft payload. The period of a GEO is 23 hours 56 minutes and 4 seconds.CHAPTER 5.1 MicroSatellites Astrid1 and 2 Sweden’s ﬁrst microsatellite. In total. it should be kept in mind that these secondary payloads do not have a guaranteed launch date [39]. Astrid1.29 m2 . Modern small satellites are launched at low cost as secondary payloads. but.1 × 0. thereby providing global mail service. 2 96 . which even a large satellite in a Geosynchronous Orbit 2 (GEO) cannot provide [39]. With solar panels stowed. microsatellites in LEO were recognised due to their advantage in digital communication. In Europe. it is important to recognise the risk associated with such structures from the onset of the mission programme [136]. was launched on January 24. in general.45 × 0. NASA’s Space Shuttle has several ways of accommodating payloads in its payload bay. produced 42 W. small satellites are developed for shorttime missions and quickly put together by small teams. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA viewed as useful for space science research because of their small size [48].000 km. Mounted on the Ariane4 rocket the ASAP can carry six small satellites. Small satellites are therefore an aﬀordable way to space for the military. or socalled piggyback passengers. each weighing less than 120 kg [3. The mass and dimensions were chosen because they represented about half the maximum permitted values of a microsatellite on the Ariane4 ASAP [48]. Today. The use of fewer and more contemporary components also helps to reduce the cost of the spacecraft [39]. each with a mass of up to 50 kg. Users send a message to the satellite during its passage and the message could then be delivered throughout the world. In the 1980s.29 m3 . and with solar panels deployed 1.45 × 0. the dimensions was approximately 0. complexity. A few examples of small and microsatellites with deployable structures are given below. using cylindrical containers of diﬀerent sizes mounted on the bay wall [39]. Astrid1 was designed and developed by Swedish Space Corporation (SSC). Launch resources for secondary payloads often have a single price for payloads under a certain mass limit. can accommodate eight satellites. Figure 5.39 × 0.1. The total mass of Astrid1 was 27 kg. In order to facilitate the design of reliable mechanisms. Arianespace developed the Ariane Structure for Auxiliary Payloads (ASAP) to accommodate secondary payloads on the Ariane rockets. universities and industry.46 m3 (including antennas). each 0. In recent years they have fallen into disfavour due to their higher risk.2(a). 5.
It appears from Figure 5. Astrid2 also had a twohinged deployable boom.1 metre nondeployable. 5. and the total antenna mass. 1999 contact with the satellite was lost and never reestablished despite several attempts [163]. SMALL SATELLITES AND DEPLOYABLE STRUCTURES STRV ASAP Figure 5. The design lifetime for Astrid2 was one year.30 m3 . and to a similar altitude as Astrid1. each 0.9 kg. The total mass of Astrid2 was 29 kg.3 m3 . 97 . and with solar panels deployed 1. On Astrid2. The second Swedish microsatellite. including the subreﬂector and supporting structure.95×0.45×0. a height of 2. Figure 5. The design is simple and reliable.29 m2 .1: DERA’s small satellites STRV1 C and D on Ariane5 ASAP for launcher ﬁt check (Courtesy of DERA).7 × 1. The primary payload on Odin is a 1. oriented along its spin axis. also designed by SSC.2 Small Satellite Odin Another satellite developed by the SSC is the small satellite Odin.1. which carried a device for attitude determination plus one of the scientiﬁc payloads. one pair of hinges with a common axis of rotation on each solar panel.39 × 0. and a base area of 1. was 80 W. but on July 24. 1998 from the same site.1. is 9. a similar deployment technique was used for four of its six solar panels. oﬀset reﬂector antenna. was launched on December 8. The total power of the six solar panels. With solar panels stowed.0 m. solid surface.1 × 1. on a similar launch vehicle. while the remaining two panels were nondeployable. the solar panels are the only deployable structures.5.1 m2 with stowed solar panels. Astrid2. Odin has a total mass of 250 kg. On Astrid1.2(b). the size was 0.1 × 0. The mass of the main reﬂector is 5.2(b) that the boom unfolds in a plane and that a cable is used to stop the unfolding and to correctly position the boom.5 kg.
The four solar panels.8 m2 with deployed solar panels. given the high operating frequency of the antenna (580 GHz). a deployable solid surface antenna could have been used instead. low cost satellites could oﬀer aﬀordable spacebased research to civil government.CHAPTER 5. 2001 from Svobodny.3(a). A piggyback launch would have been a more economically viable alternative.1. DERA Farnborough was renamed QinetiQ on July 2. Russia. Odin was launched on February 20. industry. Figure 5. 2001.2: The Swedish microsatellites (a) Astrid1 and (b) Astrid2 with solar panels deployed (Courtesy of Swedish Space Corporation). The Odin satellite is signiﬁcantly larger than the Astrid satellites in order to accommodate the reﬂector antenna. The project objectives were to show that small. but will in this thesis still be referred to as DERA. and also provide support to military defence research objectives [32]. the Defence Evaluation and Research Agency (DERA) at Farnborough3 in United Kingdom started the Space Technology Research Vehicle (STRV) programme as a single microsatellite project.8 × 3.3(b). 3 98 . Figure 5.75 m2 .4 × 0. provided 340 W of power. solar reﬂectors and other scientiﬁc instruments. 5. only a solid reﬂector surface can give the required surface accuracy (10 µm). as the only payload on a Start1 launch vehicle to an altitude of 600 km.3 Space Technology Research Vehicles In 1989. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA (a) (b) Figure 5. However. Further. but that would have compromised the scientiﬁc objectives and led to a more complex satellite design [164]. and 3. each approximately 1. and academic organisations. but at a higher mission risk. The design lifetime for Odin is two years.
STRV1 C and D were launched on November 15. which is more than twice the power available on STRV1 A and B. each approximately 0. allowing students to operate the satellites [32]. 99 . STRV1 A and B had no deployable structures. Figure 5. 1994 from the Guiana Space Centre. STRV1 A and B. However. providing a total power of 31 W [32]. Each satellite weighed about 100 kg. including recovery commands [145]. STRV1 C and D.5. e. respectively. only two weeks after launch. the mission was brought to a premature ending when the receivers of both satellites were cut from their power source because of a design mistake. The dimensions of both STRV1 A and B were 0. while the mass was 50 and 53 kg. The ﬁrst two microsatellites of the programme.45 × 0. 2000 from the same site as their precursors. due to diﬀerent scientiﬁc payloads.40 × 0.3: The Swedish small satellite Odin with solar panels (a) stowed. solar panels. Following the success of STRV1 A and B.45 × 0. DERA designed two new satellites.47 m3 .7 × 0. They were piggyback passengers on an Ariane4 launch vehicle to a Geosynchronous Transfer Orbit (GTO) [32]. The satellites had solar panels on four of its sides.4. which could accommodate more scientiﬁc payloads and thereby widen the scope of mission objectives [32]. but they continued conducting orbital operations until September 1998 when they were donated to the University of Colorado at Boulder.7 m3 .g. Unlike the Astrid satellites. Kourou in French Guiana. again as auxiliary payloads. were launched on June 17.6 × 0. but this time on an Ariane5 launch vehicle to a GTO. capable of generating 80 W. The dimensions of STRV1 C and D were 0. The design lifetime of STRV1 A and B was one year. SMALL SATELLITES AND DEPLOYABLE STRUCTURES (a) (b) Figure 5. USA. The new payload required a larger spacecraft structure. and (b) deployed (Courtesy of Swedish Space Corporation). This means that the satellites cannot receive anything sent from the ground.45 m2 .1.
etc. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA 5. performed by Pellegrino and the author. • Planar array. and 3. This study will now be signiﬁcantly 100 . By using the motion of the spacecraft and advanced signal processing techniques. These requirements can be divided into three categories: 1. three novel deployable structures. needed to run the SAR antenna. In DERA’s requirements. but placing a large antenna in space is extremely expensive. feed. This is. it was soon found that new technology was needed to meet the requirements by DERA.65 GHz) as most recent systems use this frequency. The preliminary study of the SAR reﬂector antenna.CHAPTER 5. related to categories 2 and 3 above. Figure 5. in very simple terms. DSL was given the task of designing deployable structures to meet the requirements by DERA. a higher resolution is obtained using a larger antenna. and miniature remote sensing [32]. These tasks would necessitate the use of deployable structures and outline requirements are formulated for a range of such structures by DERA in reference [136]. In remote sensing. DERA started the plans for future STRV missions. no speciﬁc mission had been identiﬁed.1. Including electronics. Various SAR antenna options are available [136]: • Parabolic reﬂector. the principle behind the SAR technology. one of the biggest limitations in performing a SAR mission using a small satellite. is the problem of stowing a large antenna in a small volume. so the stowage constraints for the deployable structures were assumed to be deﬁned by the envelope around the STRV1 C and D satellites on the Ariane5 ASAP [136]. 2. were developed. These are presented in reference [130]. support. a larger antenna can be simulated. the author was involved in the development of a 3 m diameter parabolic reﬂector SAR.5–2 kW. a high image resolution is the primary goal. Booms for space based GPS applications.4 Future STRV Missions In the spring of 2000. In response to these requirements. Generally. As most of the existing technologies of deployable structures are aimed for large spacecraft applications. is presented in references [130] and [169]. Synthetic Aperture Radar (SAR) antenna for remote sensing. Hence. the total payload mass goal is 150 kg [136]. or • Reﬂect array. The performance of a SAR antenna is aﬀected by a number of parameters but one especially important is its size. Solar arrays of 1.4. In that study. The main areas of interest include a space based Global Positioning System (GPS) for attitude and orbit determination. The mass goal for the main reﬂector alone is 20 kg. The operating frequency for this antenna is assumed to be in the Xband (9.
the required stowed size immediately rule out all umbrellatype mesh antennas as the hubs of these are too large.4: Envelope of the STRV on ASAP ring (Courtesy of DERA). A concept which would easily conform to these dimensions is the inﬂatable reﬂector.2 New Reﬂector Concept for Small Satellites The main diﬃculty in meeting the requirements by DERA is the very stringent constraints on the dimensions of the packaged envelope. brieﬂy described in section 2. the HRA. The stowed size of these antennas also tend to be more cubic than oblong that is needed here. Many umbrellatype antennas also rely on active control of the surface shape which complicates the overall design. An inﬂatable antenna is therefore not considered as an alternative for the present application. without active control. [42.2. e. Much work is currently being done on inﬂatable reﬂectors. A better design would be one that passively.g. e. 5.e. mainly in the USA. 800 × 200 × 100 mm3 . cf. In addition. NEW REFLECTOR CONCEPT FOR SMALL SATELLITES 85 WAVER (100) 700 (100) 91 800 ASAP 349 ∅ 140 ∅ 200 900 ∅ 140 ∅ 200 (a) Front view (b) Side view Figure 5. i. the Collapsible RibTensioned Surface (CRTS) reﬂector [78]. This naturally leads us to the concept of the tension truss.5. 67].1. expanded.g. but still they cannot be regarded as a mature technology. The aim of the following and the next two chapters is to provide a thorough feasibility study of the new reﬂector concept.3. provides the required surface accuracy. 101 91 ASAP 800 85 WAVER .
i. If the height of the rear net is halved. possible to shorten the vertical struts by making the rear net shallower. Figure 5. it is considered that an adaptation of the AstroMesh concept oﬀers the greatest potential for meeting the requirements with a lowcost system. it is composed of three main parts: • a deployable ring structure.e. In general. however. 102 . • two identical cable nets (front and rear nets).5(a).2 Tensegrity Reﬂector Concept The new reﬂector antenna concept. say half the depth of the front net. Figure 5. i.5(b). again. the complete antenna structure is quite stiﬀ. the stowed height is 1. a large hub and is therefore not suitable for the present application. attached to the front net. connected by tension ties.22 m.5. [188]. has. with. The tension truss antenna for the HALCA satellite. like the umbrellatype antennas. a 3 m diameter reﬂector with F/D = 0.e. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA 5. The complete antenna structure is composed of a large number of tension elements. is based on the tensegrity and tension truss concepts. with identical front and rear nets and no separation between the nets.14. Although none of the existing designs are directly applicable. Figure 5. Figure 5.2.2. For example. Figure 5. but as will be shown in the next chapters. without active control. An alternative ring structure based on a pantograph concept. 5. Figure 2.e. For a given aperture and number of bays it is not possible to change the length of the horizontal struts. the struts. However. the net cables and constanttension springs.4 and a ring truss divided into 18 bays would. The deployable ring structure is a onestage tensegrity module with six struts which has one state of selfstress and seven internal mechanisms of inextensional deformation.5(c). The current stateoftheart of high precision antennas with a passive structure is the AstroMesh antenna. have a stowed height of 1. cf. and only six compression elements. It is. The stowed height of the AstroMesh is given by the length of one vertical plus one horizontal strut. this requires a large number of joints which add a signiﬁcant mass to the structure. The main parts of the new concept is presented below.1 Existing Concepts with Passive Structure The research on antennas which passively. The tensegrity module alone is very ﬂexible. i. Like the AstroMesh antenna. if larger forces in the ring are accepted. [96].5(b).CHAPTER 5. achieve the required surface accuracy started in the early 1980s with NASA’s Large Deployable Reﬂector (LDR) programme. 18 bays would have a packaged height of about 0. and • the reﬂecting mesh.46 m.9 m. cf. an AstroMesh reﬂector with small F/D requires a high ring truss with too high stowed height for the present application.
5: The new tensegrity reﬂector concept for small satellites: (a) the complete structure with RF reﬂective mesh attached to the front net. and (d) the tension ties connecting the nets.2. (b) the tensegrity ring structure.(a) (b) 5. . NEW REFLECTOR CONCEPT FOR SMALL SATELLITES 103 (c) (d) Figure 5. (c) the front and rear nets.
especially useful as an antenna [158]: • Incident rays parallel to the reﬂector axis converge to a spot known as the focal point. [71]. the focal point lies in the plane passing through the reﬂector rim. and design maturity. where the reﬂecting surface is a paraboloid of revolution. Conversely. xyz. which give the size and curvature. this surface is described by z= x2 + y 2 . 5. a reﬂector antenna consists of two components: a reﬂecting surface that is large relative to the operating wavelength and a much smaller subreﬂector. the distance from the apex of the paraboloid to the focal point. 4F (5. 4F (5. it consists of several known substructures put together to meet the requirements of stowage size. 16F (5. low complexity and cost.6.3 Geometry of Parabolic Reﬂector Antennas Reﬂector systems are the basic types for satellites because of their low mass. placed close to the focal point of the main reﬂector. respectively [158]. i. cf. The main diﬀerence compared to previous concepts must be the deployable ring structure.25. In a Cartesian coordinate system.3) 104 . The angle θ0 from the Zaxis to the reﬂector rim is θ0 = 2 arctan D . The role of the subreﬂector is to redirect the rays reﬂected in the main reﬂector to the feed. which is computed as: H0 = D2 . all rays leaving the focal point are parallel to the reﬂector axis after reﬂection from the parabolic surface. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA Although the concept above is new.1 AxiSymmetric Reﬂector The simplest antenna has an axisymmetric parabolic reﬂecting surface.2) As F/D → ∞. • All path lengths from the focal point to the reﬂector and onto the aperture plane are the same and equal to 2F . Basically.CHAPTER 5. Figure 5. It is sometimes useful to know the height of the reﬂector H0 . Two important properties make the parabolic reﬂector. where F is the focal length.1) It is most common to specify the reﬂector in terms of the diameter D and focallengthtodiameter ratio F/D. the reﬂector becomes planar and when F/D = 0. 5.e. The present ring structure was inspired by new ideas of using tensegrity for deployable space antennas.3.
the surface of the oﬀset reﬂector is visualised by the intersections of several concentric cylinders and the parent paraboloid.3 0.1).3. Then a cylinder of radius Ra .4 0.7. Simplifying gives the following quadratic 105 .5) and (5. In Figure 5.4) φa = arctan 2F x z and x z = cos φa sin φa − sin φa cos φa x .4. Also shown are the coordinate systems xyz and x y z . To eliminate or reduce aperture blocking reﬂector antennas with oﬀset feeds are used [158]. 0. y ) the z coordinate is found by substituting (5.5 0 0. (5.6) = X Z − XO 2 XO /4F . the following relations are derived [78]: XO .6: Parabola shapes for F/D = 0.3. GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS 0. The intersection of this cylinder and the parent paraboloid is a plane ellipse which deﬁnes the rim of the oﬀset reﬂector. 0. Referring to Figure 5. For a given point (x . both with origins at O in the centre of the oﬀset reﬂector.5D A disadvantage of axisymmetric reﬂectors is that the feed antenna and its support structure in many cases are bulky and block the incident rays. (5.6) into (5.5) Note that Y = y = y . Figure 5.3.7.8.5. is deﬁned in the ﬁrst and fourth quadrants of the XY plane. An oﬀset parabolic reﬂector may also simplify the satellite design as the feed can be contained within the satellite structure. 5. resulting in reﬂector performance degradation.2 Oﬀset Reﬂector The oﬀset parabolic antenna is constructed by considering a parent paraboloid with diameter Dp .5D 0 Figure 5. 0. z (5.5D θ0 F/D = 0.5. without the need of a support structure [183]. whose axis is parallel to the Zaxis.
calculated as: ξ= 1+ Da + 2XA 4F 2 with the solutions √ .. (5.5) and (5. The depth of the oﬀset reﬂector. y′ O X A P Figure 5. (5.7) B 2 − 4AC . i.9) Ha = 16ξF z1. corresponds to the intersection of the z axis and the part of the parent paraboloid in the second and third quadrants.. the ratio of the major and minor axis of the reﬂector aperture.3). Ha . (5. i. is reobtained for XA = −Da /2.6) as: 2 Da . X < 0.CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA Z Parent paraboloid z z′ Offset reflector B x′ x F Y y.2 = −B ± where ξ is the ellipticity.10) Note that the height of the axisymmetric reﬂector. (5. 106 .e. (5. can be found by (5. equation [78]: sin2 φa (z )2 − [4F cos φa + 2 sin φa (XO + x cos φa )] z A 2 B + (y ) + x 2XO cos φa + x cos2 φa − 4F sin φa C = 0.e. much larger value.8) 2A Only the “−” solution is of interest as the other.7: Threedimensional view of the oﬀset reﬂector antenna on parent paraboloid.
9(b) the reﬂector centre O does not coincide with the centre of the elliptic rim. XO .12) 107 . For the CRTS reﬂector. Figure 5.9(b): xA + xB = 0. Figure 5. 4F (5.11c) The coordinates of point C in the x y z system are obtained by (5. Figure 5. Lai [78] observed that the standard conﬁguration is not ideal and therefore proposed a new conﬁguration. The ﬁrst view. This reﬂector conﬁguration is therefore called the standard conﬁguration [78].9. Note that in Figure 5. is determined by the following relation. is along the zaxis. Two views of the standard conﬁguration are shown in Figure 5.9(b).9. yC = Ra sin ϕ. Ra zC = (Ra + 2XO cos ϕ) . (5.11b) (5.3. In the xyz system. Standard Conﬁguration The normal choice for O is XO = (XB + XA )/2 = XA + Ra . the central hub conﬁguration. Central Hub Conﬁguration For this conﬁguration point O is in the centre of the elliptic reﬂector rim. The position of the central hub. the coordinates of a point C on the reﬂector rim. the second view.11a) (5. with two axis of symmetry. Figure 5.6). GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS Z z F z′ Parent paraboloid B x′ ZO φa Offset reflector A O x X P XA XO XB Figure 5. is along the z axis.5.8: Oﬀset reﬂector geometry deﬁnitions. are xC = Ra cos ϕ.9(a).
108 . DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA y y′ C Ra C ϕ A B x A ϕ B x′ (a) (b) Figure 5.7). system the x and y coordinates of a point on the elliptic reﬂector rim are x = xA cos ϕ . the function fzero in Matlab [89].15) (5. Equation (5.14) . Note that for the central hub conﬁguration. As anticipated by Lai [78]. the local system must be rotated an angle ∆φa .12) is solved numerically for XO using e. the central hub conﬁguration produces a better membrane prestress distribution in the CRTS reﬂector than the standard conﬁguration. To make it normal. In the x y z coordinate. (5.13) (5. y = Ra sin ϕ .9: Standard oﬀset conﬁguration projected onto (a) xy plane and (b) x y plane.16) while the z coordinate is found by (5.g. the z axis is not normal to the elliptic aperture plane which is the case for the standard conﬁguration.CHAPTER 5. see [78] for further details. which is the diﬀerence between φa for the standard conﬁguration and φa for the central hub conﬁguration. where XO xA = (XA − XO ) cos arctan 2F and xB = (2Ra + XA − XO ) cos arctan XO 2F 2 XO (2Ra + XA )2 − XO sin arctan + 4F 2F 2 2 XA − XO XO + sin arctan 4F 2F (5.
focal length and oﬀset distance are determined by taking a number of aspects. This conﬁguration. GEOMETRY OF PARABOLIC REFLECTOR ANTENNAS y y′ C Ra C ϕ A B x A ϕ B x′ (a) (b) Figure 5.3. it should be large enough to eliminate the feed blockage. For the circular conﬁguration. 109 . a larger oﬀset means a larger ellipticity factor which may have some implications for the present concept concerning the prestressability of the tension truss. When it comes to the F/D ratio and oﬀset distance XA . than the two previous conﬁgurations.5. called the circular conﬁguration. both electrical and structural.3. The circular conﬁguration will therefore not be further considered in this study. a large F/D ratio means a bulky supporting structure for the subreﬂector which unnecessary adds weight to the whole assembly. Regarding the oﬀset distance.3 Values for F/D and XA The values for the antenna diameter. While this aspect was not a problem for the CRTS umbrellatype reﬂector. XA .10: Central hub oﬀset conﬁguration projected onto (a) xy plane and (b) x y plane. it is probably undesirable for a deployable reﬂector with a supporting ring structure. the reﬂector rim is not coplanar as in the other conﬁgurations. However. 5. The simplest way to get an idea of the appropriate values for these antenna parameters is to look at values used for other deployable space antennas. It is always desirable to have as large a diameter as possible for increased gain. Circular Conﬁguration Lai [78] also considered a third conﬁguration where the aperture projection on the x y plane is a circle instead of an ellipse. was chosen for the CRTS reﬂector as it produces a better membrane prestress distribution and more uniform loading of the ribs. From a structural viewpoint. into account. 84]. it seems that these are determined by special optimisation routines not immediately accessible to structural engineers [12.
The antennas for which all parameters were available are listed in Table 5. For oﬀset reﬂectors the diameter of the parent reﬂector. S: solid. [62].4. This was.53 1. as it is well beyond the scope of this thesis.62 11. Dp = 2(D + XA ).CHAPTER 5. it means that a ﬁner ground resolution can be obtained.40 5.1 Antenna Gain In communication systems.423 0. and I: inﬂatable. The factor describing the performance of the antenna is its gain. d The value of 1. within that theory there are some aspects directly related to the structural design of the antenna.56 3. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA Table 5. Nevertheless.00d 0.55 2.137 [20] — [174] M: mesh.50 0.10 1.120 [141] 1.307 [141] 1.04 0. XA /Da varies from 0.166. 5. In sensing systems. Typea Structure/satellite Aerospatiale AstroMeshc antenna Fan rib antenna BAe/Surrey antenna ODIN radiometer SSDA D (m) 5.167 and 0.056–1.118 [62] 1.273 0. an increase in aperture size means an increase in the amount of information that can be sent or received. however.231 0.265 0. 110 . 5.1. c Values for the AstroMesh onboard the Thuraya satellite is not known.167 0. with most values around 1. not such a simple task as in many cases only parts of the sought information were provided. The F/D ratio varies between 0. The ellipticity interval is 1.50 3. M 1. this section will not aim at describing the complete theory of reﬂector antennas.385 ξ Ref.70 2.70 3.00 23. however.363 0. to 0.059 [66] 1.00 Space rigidised antenna 5.05 — 0.48 F (m) 1.307. High gain antennas allow for high transmission data rates at low power and improve the signal to noise ratio [158].5 m given in Table 3 in [141] is corrected to 3.1.00 — F/Db 0.61 0.115 — [148] [51] S I a b ARISE antenna 25. is used instead.90 15. which means that they are quite deep.00 1. which is the absolute minimum.92 0. so the values for the antenna from the ﬁrst study. is used. It is likely that such information is classiﬁed and cannot be made public.286 0.13 XA (m) 0. These aspects will be presented here.363 for the parent paraboloid.1: Geometrical data for some axisymmetric and oﬀset deployable reﬂector antennas.0 m stated elsewhere.4 Reﬂector Antenna Theory Despite the title.
111 .89 [158].e.18) (5. The eﬃciency factor due to aperture blockage. aperture blockage and phase errors.65.g. According to Stutzman and Thiele [158]. This takes into account that the central part of the aperture usually is more aﬀected by the blockage. This maximum gain is reduced by a number of factors. many applications have ηa ≈ 0. Random errors are caused by the fabrication tolerances in the manufacturing process and are unpredictable within given statistical limits. The bestﬁt paraboloid. Considering their impact on the antenna radiation pattern. Pontoppidan [131] divides the surface deviations into the following three groups: 1. it is shown that ηt ≈ 0. The inﬂuence of random errors on the antenna performance is treated in the following section.17) where λ is the wavelength of the operating frequency. and their eﬀect on the performance are completely described by a defocus. 2. Deviations due to slowly varying distortions. Systematic surface deviations are inherent in the construction of the antenna and can be predicted. Deviations which change the desired paraboloid into another bestﬁt paraboloid. (5.5. Other systematic errors. may deviate from the ideal one up to about one wavelength without seriously aﬀecting the performance of the antenna. Values of ηa between 0. [174]. systematic deviations cannot be considered similar to random errors.19) where Ab is the blockage area projected onto the aperture area Aa . Phase errors arise due to surface deviations from the ideal paraboloid.43 and 0. and ηt the aperture taper eﬃciency factor. 0 < ηi ≤ 1.4. 3. to G0 = ηa Gth where ηa is the product of all antenna eﬃciency factors ηi . i.72. In general. are reported by van ’t Klooster et al. ηb . Other random errors. REFLECTOR ANTENNA THEORY For a reﬂector antenna. the theoretical maximum gain is Gth = πD λ 2 . a paraboloid which in a leastsquare sense best approximates to the ideal one. depending on the frequency. For optimum antenna performance. but have to be treated separately. like thermal distortions. e. can be estimated as [158]: ηb = 1 Ab 1− ηt Aa 2 (5.
The radiometric rms surface deviation is deﬁned as [50]: δrms = 1 Aa ∆℘ 2 2 1/2 dA .CHAPTER 5. can be written as: 2 4πδrms . i. the surface deviations are measured in a direction axial or normal to the reﬂector surface for reasons of simplicity.g. He assumed that the surface deviation at any point is a random sample from a single Gaussian distribution with zero mean and a standard deviation equal to the rootmeansquare (rms) of the surface deviation of the reﬂector.20) in general underestimate the gain and therefore can be considered to be a worst case.20) relates only to random surface deviations.21) Aa where ∆℘ is the phase error. (5.20). 142. it is with good approximation found to be valid for any type of deviation measured relative to the best ﬁt paraboloid [131].e. (5. For shallow reﬂector surfaces.4.21) simpliﬁes to δrms. 142]. Under these assumptions.z in (5. (5.23) where r is the distance from the centre of the axisymmetric reﬂector to the point of measure.24) Equation (5. [2. (5.4 and r = D/2 gives 1 + (r/2F )2 = 1.22) ∆℘ = 1 + (r/2F )2 while for the deviation in the normal direction. ηr . the random error eﬃciency factor.24) is likely to give a good approximation 112 . DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA 5.z ≥ δrms .. (5.39. Assuming that the denominator can be set to unity. it is ∆℘ = 2∆n 1 + (r/2F )2 . (5. ∆z. and ∆℘ is [50.24) is simpler than (5.20) ηr = exp − λ where δrms is the radiometric rms surface deviation [50. and Aa the aperture area. Although (5. Thus. F/D = 0.z = 1 Aa 1/2 (∆z)2 dA Aa .2 Eﬀects of Random Surface Errors An approximate method to compute the eﬀects of random surface deviations on the antenna gain is presented by Ruze [142].50]. Usually. will overestimate the gain loss as δrms.21) and frequently used to predict the performance of a reﬂector antenna. the factor 1 + (r/2F )2 is close to unity and even for deeper ones it is small. Using δrms. 189] show that the assumptions related to (5. ∆n. cf. later studies [184. However. (5. The relation between the deviation in the axial direction. pathlength diﬀerence of a reﬂected ray due to the surface imperfections. e. 184] 2∆z . A further assumption is that the surface deviations are correlated in small regions.
58] provides the ﬁrst and most thorough study of rms errors of faceted mesh antennas.27) . To ﬁnd an expression for the relation between the surface error and the facet size some assumptions have to be made [2. (5. a six bay triangular division. divide the reﬂector surface into ﬂat facets rather than gores as is the case for umbrellatypes. the δrms. the maximum facet size required to meet a speciﬁc surface accuracy is sought.11(a). Substituting λmax into (5. Maximum gain is obtained at the wavelength of λmax = 4πδrms . it can be closely approximated by a sphere with radius 2F . increases as the square of the frequency until the surface deviation eﬀect takes over and then a gain deterioration occurs. At the intersection between adjacent facets on the paraboloid. 5.20) is that if a given reﬂector operates at increasing frequency. AstroMesh [165] and ETSVIII [107]. ( /D)2 δrms. this study also includes eﬀects of mesh saddling. Figure 5. results in nearly equilateral triangles. with focal length F . With no lateral loading an isotropic uniformly tensioned membrane must have zero Gaussian curvature.33 p t . saddle shape. Hedgepeth [56.g.z of the bestﬁt facet and a sphere of radius 2F is. As shown in the previous section.11 shows a shallow spherical cap approximated by triangular and hexagonal ﬂat facets. REFLECTOR ANTENNA THEORY of the radiometric rms surface deviation and antenna gain loss and can therefore safely be used. The resulting lateral loading tends to curve the supporting element inwards.12. e. i. An important result from (5. Figure 5. If the reﬂector is shallow. at ﬁrst. For a shallow reﬂector. the gain.4.20) gives the maximum gain Gmax ηa ≈ 43 D δrms 2 (5.e. (5. [56].25) . In general. 58].3 Systematic Surface Error of Faceted Paraboloids Many reﬂector antennas. Figure 5.01614 D F/D 113 1 + 0.z calculation for an equilateral triangle on a spherical surface is assumed to be a good approximation for the actual geometry.z = 0.26) which is proportional to the square of the manufacturing accuracy D/δrms [142]. Faceting introduces a systematic deviation of the actual surface from the desired surface which degrade the performance of the reﬂector. it is not possible to map regular polygons onto a curved surface. Thus. the parameter governing the performance degradation is the rms surface deviation.4. Thus. [2] and found to be valid. the mesh tension changes direction.5. This assumption was checked by Agrawal et al. The corresponding normalised rms surface deviation δrms. Beside the surface deviation due to the facet size.
27) as p /t → 0.. p (a) (b) Figure 5. The rms deviation of facets including mesh saddling was recently reanalysed by Tanizawa et al.28) for p /t < 1. For equilateral triangles it is 1 ( /D)2 δrms.660 p + 0. square. D 16 15 F/D (5. A Taylor series expansion of the square root part of (5. the rms deviation of the equilateral triangle was found to be ( /D)2 δrms. which is similar to the corresponding part of (5.e.27) is negligible. p the mesh tension. for the deformation of the supporting tension element.. (5. By applying a Fourier series solution.28) around p /t = 0 gives 1 + 0. as reported in [106]. the tendon force must be large compared to the mesh tension. instead of a polynomial as Hedgepeth [56]. i. To minimise the eﬀects of mesh saddling. Figure 5.11: Approximating a shallow spherical dish with ﬂat facets of (a) triangular and (b) hexagonal shape..z the required side length is F δrms.01614 D F/D 1 + 0. and hexagonal shape and a sphere. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA (a) (b) Figure 5.133 t p t 2 (5.12: Mesh saddling of triangular facets on a synclastic surface.z = 0.CHAPTER 5.z = 7.e.29) i.872 . and t the force in the supporting element.28) and (5. Agrawal et al. allow the largest facet size.33p /t + O[(p /t)2 ]. where is the side length of the triangle.z = √ .30) D D D 114 . the diﬀerence between (5. identical to (5.27). For a given value of δrms. [2] compute the rms surface deviation between bestﬁt ﬂat facets of triangular.12. For practical values of p /t.
The optimum solution for a cable supported facet is. the triangles have the largest side length for a given rms deviation tolerance. Although hexagonal facets yield a better radiation pattern.31) where is the gore width at the reﬂector rim.046. Referring to the research by Agrawal et al.z . It is found that hexagonal facets provide a slightly better radiation pattern than triangular facets. triangular facets are structurally desirable and require the fewest number of structural members to meet a given surface tolerance. respectively. Including the mesh saddling between the parabolic ribs.817 F δrms.z . A reﬂector surface made out of square facets with identical δrms. respectively. therefore.0082 2 /F .5. the ﬂat facet. an acceptable value of rms surface distortion will be identiﬁed. which is about 49% below the ﬂat facet rms error. which yields p /t = −2. their conclusion is that triangles appear to be the best option for approximating a paraboloidal reﬂector surface. Only the surface accuracy can be considered at this point.4.0069 2 /F for a boundary radius of 1. the boundary members must be in compression which requires elastic rods rather than cables along the facet boundaries. of 12 mm (λ/25). the rms surface deviation is ( /D)2 δrms. Hence. δrms. For isotropic mesh tension.32) D A very large number of ribs is needed to obtain a high surface accuracy.4. the value 7. to realise this optimum facet as well as the congruent one. [2].01076 D F/D 1+ p1 p2 . Hence. and p1 and p2 the membrane tensions in the radial and circumferential directions. Considering all factors. the required gore width is = 6.z = 0. The requirement report by DERA [136] does not directly specify the required surface 115 . it is found that δrms. The minimum rms deviation was found to be δrms.474.z yield a worse radiation pattern. Hedgepeth [56] also studies the rms surface deviation of umbrellatype reﬂectors having gores rather than facets. [2] also analyse the electrical performance of a proposed large reﬂector antenna (D = 660 m and F = 573 m) with an operation frequency of 1 GHz and an allowable surface error. which is explained by the variation of the surface deviation over the reﬂector.872 is substituted by 6.4F .z = 0. D D (5. Fichter [38] proposes a method of changing the boundary shape of the facets so that the total rms error could be improved. p1 = p2 . Agrawal et al. For congruent facets with a boundary radius of 2F . REFLECTOR ANTENNA THEORY For squares and regular hexagons.4 Allowable Surface Error of Reﬂector Antennas The main design points for deployable reﬂector antennas are deployment feasibility and high surface accuracy. (5.160 and 4.z = 0. This optimum solution can also be reobtained by minimising (5. In this section. 5.28) with respect to p /t.
[62] also note that an umbrellatype antenna would require 170 ribs to yield the same accuracy. A closer look at the practical surface accuracy of structures based on this concept is therefore of great value in determining an allowable rms surface accuracy for the present reﬂector.20) it is evident the there can be a tradeoﬀ in the choice of reﬂector diameter and rms surface accuracy. however.3 mm. the design of a 23. the choice of reﬂector diameter is also constrained by aspects related to the radiation pattern [138]. Diﬀerent surface error distributions with the same total rms surface accuracy can give diﬀerent radiation patterns. (5. The ﬁrst contribution to the total surface error comes from the faceting of the reﬂector surface. the rms surface deviation for the entire reﬂector is estimated to less than 1. (5. In the extensive and more recent review of the precision of deployable antennas by Hachkowski and Peterson [52].286) is described. engineering models. From (5. The allowable surface accuracy is somewhat dependent on the antenna type. Hedgepeth et al. In [62]. and engineering feasibility studies. Roederer and RahmatSamii [138] observe that rms surface accuracy goals lie typically in the interval λ/100–λ/25. ﬂight models. less accurate reﬂector can provide the same peak gain as a smaller more accurate one. well beyond the scope of this thesis. • Centrifugal forces due to satellite spinning. Miyasaki et al. The acceptable rms deviation error due to faceting is set to λ/200 or 0.CHAPTER 5. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA accuracy for the present reﬂector but states that an rms accuracy of at least λ/20– λ/10 is necessary.29). and Values computed from 24 of the 50 listed antenna structures for which both λ and δrms.4 mm and 0. The allowable rms surface deviation is taken as λ/50. 4 116 .5 m. For oﬀset reﬂectors. again showing the advantage of the tension truss concept. Analysis and test of a 3module assembly result in an rms surface accuracy of 0. is 0.25 mm for the operating frequency. where the concept of the AstroMesh reﬂector is introduced. the rms surface deviation ranges from4 λ/353 to λ/6. Both of these studies include values from antennas at diﬀerent stages of development: in orbit. • Loading of the tension ties. The required side length of the triangles. for symmetrical reﬂectors the peak gain is the most important parameter [138].32).56 m diameter oﬀset reﬂector antenna (F/Dp = 0.4 mm at Sband) as the allowable surface deviation for the 13 m ETSVIII antenna. requiring a tension truss with 27 rings.18) and (5. or 1 mm at Cband. Other systematic contributions to the total surface error are distortions due to • Gravity during ground testing.z are given. A larger. It is generally not suﬃcient to specify a total rms surface accuracy. The tension truss concept was earlier identiﬁed as the most viable solution for passively achieving a reﬂector surface with high accuracy. [107] uses λ/50 (2. Using these results. respectively.5 mm. A study of these radiation patterns is.
33) Nadir is the direction from the spacecraft to the centre of the Earth.2 rpm.7 0. or from pretensioning. The rms surface error due to random member length errors in fabrication and thermal expansion coeﬃcient variation are evaluated by the Monte Carlo technique and found to be less than 0. The rms surface distortion due to gravity during ground testing is smaller than the resolution of the photogrammetry measuring system. 117 .0 < 0.08 1.5 N. 0.33 32.i / 2 i δrms.4. The rms surface error due to mesh saddling.5 mm.5 Ground Resolution The gain is not the only factor deciding the antenna diameter.4 mm if the rim is supported during ground testing. More recent results on the surface accuracy are available for the 6 m AstroMesh reﬂector [165.2 mm for a strain of 254.6 mm or 1. REFLECTOR ANTENNA THEORY Table 5. 0.i δrms. For Earth observing systems the ground resolution is of signiﬁcant importance [18].44 5 hλ D (5.41 50. which is 0.6 mm with a tension tie force of 4.2: Surface distortions for the 6 m AstroMesh reﬂector [166].4. A total rms surface error estimate is not given for the reﬂector but this extensive study clearly shows the inﬂuences of diﬀerent error sources. h. section 5.9 0. For a satellite at altitude. The total rootsumofsquares (rss) error from all sources is less than 0. Error source Surface faceting Manufacturing Mesh saddling Deployment repeatability Thermal extremes Total rootsumsquare a 2 Computed as δrms. is believed to be achievable with existing materials and manufacturing technology.20 12.0 · 10−4 D. The computed rms error estimates for these sources are: 0.12 mm at a spinning rate of 3.2. and around 0.5. at nadir 5 is χ = 2.2 mm.6 100.0 < 0. thus having very little inﬂuence on the total error.4 µm/m in each net element. is less than λ/500.4.10 3.4. that would give an rms surface accuracy equal to λ/20. 166]. A total surface distortion of 2. from a large temperature change. the ground resolution.5 · 10−5 D. 5.4 < 0. χ.07 mm. Table 5. For a 12 m reﬂector operating at 40 GHz.0 • Large strains arising due to systematic error in net fabrication. Random errors due to uncertainty in the tension tie loading yield an rms error of 0.z (mm) (%)a 0. The ground resolution is expressed as the size of an object that can be distinguished from the background.
thermal expansion. the ground resolution would be χ = 13. f = 9. It should be noted. 5.5 m [136]. Material with a modulus of 115–124 GPa is readily available [62. the SAR antenna will be able to simulate an aperture with D = 27. that a SAR can provide ground resolutions similar to visual systems. 5. 5.g.4. 183]. By controlling the direction of the graphite ﬁbres.5 Selection of Materials The space environment is extremely harsh with very low vacuum levels. density. normally are anisotropic. 183]. micrometeoroids and debris.CHAPTER 5.1 Materials for the Antenna Structure Aluminium alloy is the most commonly used metal for spacecraft structures. composites. Therefore. For precision structures. surface accuracy goal for the present application would be a total rms error of λ/50 with a budget of λ/100 for the faceting rms surface deviation. independent of the range and wavelength by synthesising the required aperture. 141]. availability. However. stiﬀness. The main disadvantage of aluminium is its high CTE. but realistic. According to Hedgepeth et 118 . The selection of materials is based on several factors [183]: strength.e. Advantages of aluminium are low density. Most metals are isotropic while nonmetals. since the present reﬂector will use SAR technology. D = 3 m. a lower surface accuracy can be accepted [4. the relevant material properties are high modulus of elasticity and low Coeﬃcient of Thermal Expansion (CTE). etc. Spacecraft structures typically contain both metallic and nonmetallic materials. very high stiﬀnesstoweight ratios are obtained. with a budget of λ/50 for the facet error. For the present antenna. The selection of materials is of cardinal importance to the survivability of a spacecraft in this environment. will also be considered. and machinability. have lower CTE but provide other disadvantages. and temperatures as low as 4 K [144.65 GHz. cost. titanium or magnesium. high strengthtoweight ratio. Other metals. high stiﬀness. however. CFRP oﬀer better material properties than metals and can be tailored for high strength.9 km and with the ground resolution goal of χ = 1. λ = 0. e. and extremely low CTE. such as reﬂector antennas. i. 22µ/◦ C [183]. a lower accuracy goal of λ/25. which uses signal processing to achieve a high resolution.5.8 km. magnetic and gravitational ﬁelds.5 µm. and h = 550 km. low cost. high doses of thermal and particle radiation.6 Accuracy Goals for the Present Antenna This review of surface accuracies of deployable mesh antennas shows that a high. DESIGN PREREQUISITES FOR A DEPLOYABLE REFLECTOR ANTENNA where h can be replaced by the slant range to determine the ground resolution away from nadir.
5.5. SELECTION OF MATERIALS
al. [62], CFRP with a modulus of 227.5 GPa can be produced at extra cost which for highprecision antenna applications may be justiﬁed. For example, using the less stiﬀ CFRP for the 23.5 m reﬂector in reference [62] would have resulted in doubled rms surface errors due to ground testing and pretensioning, and lower vibration frequencies. Typical tensile strength for highmodulus CFRP is about 1500 MPa, but higher strengths are available [133]. As CFRP is a very brittle material, it can be assumed that the proportional limit is above 90% of the ultimate strength. CFRP is also very light; densities between 1660 and 1740 kg/m3 are available [62, 96]. These densities are a little higher than raw CFRP, to allow for an impermeable coating of the struts [96]. CFRP with a CTE less than 1.0µ/◦ C can easily be procured. Although the main antenna structure will be composed of composite material, details such as end ﬁttings are made of metal. These details are preferably attached by bonding as bolting would give rise to undesirable stress concentrations. Other more complex moving details, e.g. hinges, can be made of several diﬀerent materials where each material aﬀect the overall performance of the detail. For example, the TSR hinge, section 4.5.1, contains tapes of spring steel, thin steel cables, aluminium alloy spacers, steel bolts, and a main body of Delrin.
5.5.2
Materials for the RF Reﬂective Surface
First of all, the material used for the RF reﬂective surface must be easy to compactly fold and require a low density. The most common surface material for space reﬂectors of moderate precision is a mesh knitted from metallic or synthetic ﬁbres that have been plated with RF reﬂective material. A mesh provides the compliance necessary to conform to the doubly curved surface without wrinkling. A high compliance also means that the mesh can be connected directly to the net structure without any special interface [61]. Because of its openings, the mesh is limited to frequencies up to 40 GHz. At higher frequencies, the losses through the mesh are too great. Solid membrane surfaces have been developed for higher frequency applications, but without success. The very low inplane compliance of membrane means that it will not easily conform to a doubly curved surface without developing wrinkles, which degrade the antenna performance [95]. For the present antenna, it is assumed that the mesh is knitted from goldplated molybdenum wires, which are readily available. This mesh has a surface density of 25 g/m2 [62]. To assure good electrical conductivity between the mesh wires, the tension in the mesh must be fairly isotropic and uniform. Mesh tensions of 2.0–2.5 N/m were reported in earlier studies [61,62]. Recently, values of 10–11 N/m have been used [34, 165]. A higher mesh tension will more eﬀectively smooth out the creases formed in the mesh during folding and, thereby, give a better antenna performance. In addition, it has been shown, [61, 62], that the strength of a structure designed for a mesh tension value of 2.5 N/m is suﬃcient to withstand lateral accelerations many times greater than those experienced in orbit without severe distortion of the reﬂector surface.
119
Chapter 6 Analysis of Tension Trusses
In the previous chapter it was found that a new reﬂector antenna concept is needed in order to meet the stowage requirements of a future STRV mission. Ideally, the required accuracy of the reﬂector surface should be achieved with a passive structure, i.e. without active control. Miura’s tension truss [104] was identiﬁed as the best candidate for this task as its surface accuracy can be altered without major changes to the supporting structure. As the performance of the antenna is a function of the condition of the reﬂecting surface, it was considered necessary to analyse the behaviour of tension trusses separately from that of the antenna structure. A tension truss is basically a geodesic dome with elements that are ﬂexible rather than stiﬀ. In that way, it can easily be folded. External forces are applied at the nodes to provide a state of tensile forces in the assembly, Figure 6.1. In practice, the external forces are provided by springs, called tension ties. The key feature of the tension truss is that its shape is more or less predetermined by the lengths and arrangement of its elements. The elastic deformation eﬀects on the shape are of secondary importance. For this to hold, the assembly must be kinematically determinate. Despite the clear advantages of the tension truss for lightweight, highprecision deployable structures, only a few studies consider its fundamental characteristics, cf. [100,104]. Other studies are mainly concerned with the implementation of the tension truss in various deployable structures, cf. [99,106,165]. Recently, basic research on the characteristics and applicability of the tension truss with diﬀerent types of support conditions has been undertaken at the DSL, cf. [34,79,80,169]. These studies provide new insight into the prestressability of the tension truss. In this chapter, the preliminary study of references [168, 169] is signiﬁcantly expanded. First, a way of making an nring tension truss kinematically determinate is presented. Then, studies on the prestressability of axisymmetric and oﬀset net conﬁgurations are undertaken. Finally, the eﬀects of systematic and random manufacturing errors on the antenna performance parameters are investigated.
121
CHAPTER 6. ANALYSIS OF TENSION TRUSSES
Figure 6.1: Tension truss prestressed by external loads acting approximately normal to the net surface.
6.1
Static and Kinematic Properties
Consider an axisymmetric tension truss generated according to the scheme in Appendix C. The number of elements b and joints j in the assembly are n(1 + 3n) , 2 n(1 + n) j =1+v , 2 b=v (6.1) (6.2)
respectively. Substituting (6.1) and (6.2) into the extended Maxwell’s rule, (1.2), yields the following number of kinematic constraints required for static determinacy: c = 3 + vn. (6.3)
Along the boundary there are 3vn degrees of freedom, which is more than suﬃcient. However, in both the HALCA antenna, Figure 2.14, and the present one, Figure 5.5, supports are provided only at the v outermost vertices, Figure 6.1. Inserting c = 3v into (1.2) yields m − s = v(n − 3) + 3. (6.4) As a state of selfstress generally cannot exist in a net of synclastic shape, i.e. bowl shape, s = 0. Thus, the number of internal mechanisms in the tension truss is m = v(n − 3) + 3. (6.5)
122
Setting jan = 1 gives ban = n+1. the minimum number of extra bars is ban = n − 2. AXISYMMETRIC CONFIGURATIONS Note that for n < 3. If it can be shown that all of the mechanisms can be removed. Here.6. the additional nodes and bars should be located along the boundary and connected in a manner that preserves the vfold symmetry of the original tension truss. In tensile roof applications. 123 . and possibly more nodes. a quick study showed that the resulting conﬁgurations are not ideal. to the assembly. A further increase of jan produces more complicated conﬁgurations involving crossing elements. Solving (6. Most important of these is the number of rings n. 6. However. However. v (6. Denoting the number of additional nodes and bars per bay by jan and ban . With a limited number of supports. For a given diameter D and curvature F/D. It is therefore correct to say that no feasible state of selfstress can exist in a synclastic net. it is not possible to construct a statically determinate assembly. it approximates an axisymmetric reﬂector. The above analysis provide an indication of the feasibility of the new conﬁguration. (1. This alternative is far better considering that there are n+1 nodes along the boundary of each bay to which the added bars can be connected directly. none of the selfstress combinations produced tension in all elements.2) yields for kinematic determinacy. respectively. thus disproving the statement s = 0 above. Other parameters are the sag of the boundaries and the position of the additional nodes for the best internal force distribution. the tension truss is a type III assembly.1 SagtoSpan Ratio The sagtospan ratio of the edge cables is the ﬁrst property to be determined. while a large ratio cuts away too much of the reﬂector surface and distorts the boundary triangles.2. the only way to remove the internal mechanisms is to add more bars.2.6) For v > 3. which can be accurately analysed by the linear force method described in section 3. Preferably. assuming s = 0. cf. several states of selfstress can exist in the synclastic net.2. the ﬁrst task for a designer is to specify the remaining parameters in the mesh generation routine.2 AxiSymmetric Conﬁgurations The main advantage of the tension truss is its ability to conform to a multitude of shapes. s = v(n − 3) + 3 + v(3jan − ban ). A small ratio will give rise to undesirably large forces in the edge elements. which is given by the required surface accuracy.6) for ban gives ban = 3+s + 3(jan − 1) + n. 6.5.7) (6. a sagtospan ratio of 8–12% is common. For the case v = 6 and s = 3. Table 3.
(3.8 1. By increasing the edge forces.7 1.2 3.5 1. it can be shown that each of the three mechanisms is orthogonal to the particular sets of tension tie forces used in the analysis.1 1.6 18.3 0.8 1.6 2 1.3 3.7 9. The analysis of the edge sag will be done with a threering tension truss (D = 3 m. Figure 6.8 1.7 1.1 1. which is most practical as identical constanttension springs would be used in all of the tension ties.6 (d) 17. some elements are in compression.9 1.CHAPTER 6.2 5. The threering tension truss was analysed for three sagtospan ratios: 5. An almost uniform force distribution was obtained for edge forces of 4 N.4 5.6 1.9 1.8 1.4 0.6 (b) 13 14.5 9. For the case where the tension tie forces are all equal to 1 N. For each ratio.8 1.9 8.2: Forces in a 3 m diameter threering net with 5% sagtospan ratio (only one sixth of the complete net is shown).7 1.8 1. Loads on the inner nodes are 1 N and loads on the edge nodes are: (a) 1 N.7 1. the largest force along the edge was over 15 N.5 1.5 2. worse. 10 and 15%. However.2. As the force 124 .4) were the additional nodes and elements have been taken out.2 0.5 1. (c) 3 N and (d) 4 N. some members are in compression.6 1.2(b)–(d). The resulting net is kinematically indeterminate with m = 3. Figure 6.4 1.1 1.6 (a) 1. cf.4 1.6 (c) Figure 6. When the sagtospan ratio was increased to 10% there was still compression for tension tie forces of 1 N. If the force pattern in the net is irregular or.5 1. This is done to remove the inﬂuence of the additional nodes on the net forces.7 1. the compression forces gradually become smaller and then tensile. Figures 6. F/D = 0.5 2.2(a).7 1.1 14.1 2 1.6 0.6 1.3(a).8 2.6 1. [171]. ANALYSIS OF TENSION TRUSSES 5.7 0. the tension tie forces have to be adjusted.8 0. the initial setting of the tension tie forces was 1 N throughout. (b) 2 N.8 1. However.7 1. cf.5 18.7 4.66). The results for a 5% sagtospan ratio are shown in Figure 6.1 2 1.
3 8.6 1.6 (b) 1.6 1.4: Forces in a 3 m diameter three ring net with 15% sagtospan ratio.6 1.8 8.7 1.6 2. Figure 6.1 1. Loads on inner nodes: 1 N. the forces in the edge ties gave an almost uniform force pattern. the further loss of reﬂecting area was not justiﬁed.8 Figure 6.5 1.5 N. AXISYMMETRIC CONFIGURATIONS 5 4.3 2 1.6.6 1. the maximum sagtospan ratio is ρmax = This can be rewritten as n=1+ 1 .1 1. and the forces in the edge elements were slightly smaller than for 10%.5 1.2.2 0.2 9.7 2. Figure 6.5 4.5 5 0.6 9. Note also that with the present mesh generation routine there is a limit to the maximum number of rings n. When increased to 2 N.7 1.8 0.9 Figure 6.1 2.4(a). 4.2 2.9) 125 .2 4 0.6 (a) 1.6–2. for a certain value of the sagtospan ratio ρ.3(b).6 1.8 1. loads on edge nodes: (a) 1 N and (b) 2 N.5 0. in the edge ties was increased to 2 N an acceptable distribution of net forces was obtained and the edge forces were smaller than for the 5% ratio.2 0.6 (a) (b) 1.5 1.7 1. loads on edge nodes: (a) 1 N and (b) 2 N. 2ρ tan(π/v) (6.9 1. Although a sagtospan ratio of 15% produced a better force pattern than the 10% ratio.6 2. According to Appendix C.8 3. in the range 1. A further increase of the sagtospan ratio to 15% yielded no compressed elements even for the case of uniform 1 N tension tie loads.7 8 1.7 1. 2(n − 1) tan(π/v) 1 .9 3.1 1.8 1.9 1.5 1.9 2 1.3: Forces in a 3 m diameter threering net with 10% sagtospan ratio.7 1.9 1.2 3.8) (6.9 1.8 7.8 1.7 1. Loads on inner nodes: 1 N.8 1.
10 yields nmax = 9. ANALYSIS OF TENSION TRUSSES δan δsag Figure 6. they were positioned in the middle of the supports.5.5. As the number of rings increases. the position of the additional nodes should not be too close to the edge cable in order to avoid problems that 126 . Tangentially. any value between 0 and δsag from the edge cable was possible. Radially the position varied. 6. the best conﬁgurations were obtained for δan /δsag between 0. it is likely that more than two diﬀerent tie forces are needed to obtain an adequate prestress. 6. The tie forces were equal to 1 N on interior nodes and 2 N on the edge and additional nodes.3 Tension Tie Force Distribution It was shown for the threering tension truss that the tension ties on the edges had to provide twice the force of the interior ties to give a desirable internal force pattern. The maximum number of rings nmax for a given sagtospan ratio is nmax = 1 + 1 .4 and 0. theoretically.2. 2ρ tan(π/v) (6. It appears that for all the three F/D ratios. so with ρ = 0. threering trusses with diﬀerent focal lengths were analysed. Figure 6.15 yields nmax = 6 and ρ = 0. ρ = 0.7).10) where the ﬂoor function x yields the largest integer ≤ x. For practical purposes. Also.32 m using (C. The ratios between the maximum and minimum forces in the interior and all elements. is that it leaves more room for the additional elements along the boundary. a sagtospan ratio of 10% was selected for the tension truss in the analyses that follow. The outer diameter D of the tension truss was ﬁxed to 3 m.2. Vertically.CHAPTER 6. Figure 6. respectively. For v = 6. were used to identify the best conﬁguration.10. the eﬀective aperture diameter was 2. Hence. To determine the best position of the additional nodes. the additional nodes lay on the paraboloid. it is necessary to restrict the number of diﬀerent ties and their diﬀerences in force.2 Position of Additional Nodes Another beneﬁt of a larger sagtospan ratio.6.5 N. however. If possible. only four diﬀerent types of ties should be used within a structure and the minimal diﬀerence in force was set to 0.5: The position of the extra node in the radial direction. The aim was an interior force ratio close to unity and a fairly low ratio for all forces.
127 . for each number of rings.6.9 Figure 6. 8 7 All elements 6 5 F/D = 0. found by a trialanderror procedure. In all cases.4 0. The corresponding element forces and their ratios for three diﬀerent positions of the additional nodes are shown in Table 6.2.6 δan /δsag 0. Throughout this analysis. AXISYMMETRIC CONFIGURATIONS may arise when connecting many members to a single node. Another issue that will limit the number of rings is that the struts of the ring structure in the antenna concept.8 0.1.7 0. The maximum number of rings in the analysed tension trusses was seven.5 0. The prestressing problems became signiﬁcantly severe for more than seven rings.6 F/D = 0.6: Ratios of the maximum and minimum forces in the tension truss. the best ratios were obtained for δan /δsag = 0.10. a minimum value of 0. Hence.3 0.8 tmax 4 tmin 3 2 1 0 0. The tie force combinations on the edge and additional nodes were. so this was taken as a limit.4 F/D = 0.2 Interior elements 0. Figure 5. D = 3 m and the sagtospan ratio was ﬁxed to 10%. pass through the tension tie forest. The results from the trial analysis are shown in Figures 6. the problem of prestressing an nring tension truss has been reduced to ﬁnding (i) a tension tie force combination on the edge and additional nodes and (ii) a position for the additional nodes.4.7–6.5.4 of the total sag seemed appropriate. After a few tests on a tension truss with four rings it became clear which edge ties aﬀected certain elements. Two cases: interior elements and all elements.
4 0.97 6.10 3.08 17.4 0.6 0.40 57.88 8.313 2.25 28.53 9.62 9.6 0.24 4.67 51.51 1.62 29.72 7.86 7.469 6 1.74 1.04 13.34 62.45 2.24 62.8 0.45 1.70 2.89 2.16 3.78 39.15 2.92 1.58 3 1.4 0.4 0.84 1.79 1.70 tint max (N) 2.6 0.86 29.09 57.84 59.75 425.43 1.23 4.234 1.48 0.08 13.313 2. The best ratios are in bold face.39 2.30 39.83 2.91 1.5 0.70 13.16 3.30 0.6 tint min (N) 1.59 1.93 3.4 0.80 10.84 2.13 39.93 2.16 14.02 3.6 0.8 0.30 1.28 2.6 0.5 0.31 2.83 1.42 2.88 82.4 0.43 5.8 0.5 0.8 0.38 0.8 0.80 33.14 3.8 0.30 27.34 1.81 15.97 3.62 3.4 0.4 0.26 5.234 1.55 107.5 0.80 4.69 1.93 22.98 3.94 tall max (N) 9.63 9.4 0.5 0.36 1.77 5.92 1.55 2.59 1.74 1.88 5.00 2.4 0.55 1.64 10.77 3.07 82.07 8.40 7.234 1.83 5.22 3.8 0.82 16.52 9.6 0.2 0.10 3.21 8.78 3.25 55.54 4.59 1.32 5.39 2.07 6.91 3.28 43.84 2.469 7 1.5 0.16 3.2 0.75 81.66 4.77 4.78 2.4 0.42 2.70 16.75 0.94 82.13 3.03 2.17 2.6 0.17 2.20 30.38 1.32 2.00 1.6 0.234 128 .76 7.24 30.71 3.50 4.6 0.89 4.56 1.19 81.34 2.04 57.16 2.5 0.4 0.98 14.52 18.79 5.26 14.69 2.96 23.97 29.28 5.10 1.33 4.91 12.54 1.05 6.21 14.25 3.71 107.17 17.5 0.5 0.67 51.33 34.6 0.5 0.46 81.53 18.06 1.18 1.45 1.93 11.4 0.35 1.20 5.59 3.6 0.2 0.63 3.6 0.24 41.6 0.00 57.06 0.61 17.57 2.34 1.97 13.28 2.24 39.00 8.74 2.4 0.04 1.86 5.00 2.6 0.72 43.4 0.6 0.30 39.6 0.06 5.80 2.05 2.95 3.82 56.09 2.83 2.CHAPTER 6.6 0.8 0.07 6.43 5.22 107.30 0.60 3.469 4 1.36 10.75 3.21 7.18 429.21 27.4 0.49 17.40 1.12 3.4 0.8 0.69 417.37 37.49 2.2 F Dp 0.51 3.92 3.44 2.313 2.33 2.2 0.18 17.68 3.07 37. n F (m) 1.39 3.469 5 1.21 43.89 6.23 2.5 0.5 0.01 30.8 0.13 2.57 0.4 0.35 tint max tint min 1.4 0.6 0.4 0.46 10.39 2.19 1.92 13.51 5.86 22.83 7.77 3.4 0.43 7.46 3.85 15.92 0.4 Ha (m) 0.36 62.38 9.58 2. ANALYSIS OF TENSION TRUSSES Table 6.18 12.60 33.469 δan δsag 0.98 2.41 12.6 0.97 10.5 0.4 0.91 tall max tall min 6.09 3.47 2.4 0.34 7.34 3.83 15.5 0.234 1.10 2.20 27.94 6.11 1.16 3.90 29.49 1.79 1.23 39.63 58.10 2.65 2.74 1.03 2.313 2.58 2.52 2.64 2.45 1.32 2.1: Tension truss element forces and their ratios (axisymmetric conﬁguration).69 28.06 tall min (N) 1.11 57.03 2.313 2.34 51.4 0.86 2.5 0.92 1.99 6.24 2.
4 1. Multiplication factor for element forces was 2.6 1.4 1.2.4.2 3.4 2 0. ρ = 0.2 7.2 1.4 0.2 0.1 1 1 1 1 1 1 1 1 1 1 1 (a) 2.7: Normalised prestress distribution (a) under tension tie forces (b).4.1 1.2 5.1 1.5 (b) Figure 6. 129 . AXISYMMETRIC CONFIGURATIONS 3. F/D = 0.6.10 and δan /δsag = 0.1 1.6 1.1 1.2 1.2 7.6 5.1.5 2 2.5 2. D = 3 m.
1 1. 130 .8 1 3.3 0.9 1.2 1.8 1. F/D = 0. Multiplication factor for element forces was 2.10 and δan /δsag = 0.2 1 1. D = 3 m.7 1 1.5 (b) Figure 6.9 1 5.5 7.8: Normalised prestress distribution (a) under tension tie forces (b).5 3.2 1.2 1.1 1 1 1 1 1 1 1 1 1 1 (a) 3.7 0.2 1.1 1 1 1.7 1 0.1 1.1 0.9 0.5 2.9 1.1 0.8 1.5 7.5 3.2 2.3 3.7 10.3 0. ρ = 0.4.7. ANALYSIS OF TENSION TRUSSES 10.4.5 2.CHAPTER 6.3 0.
4 1.1 0.1 1 1.2 1 0. F/D = 0.2.6 1.4 1.6 1.3 9.6 11.1 1 1 1 1 1 1 1 (a) 4.1 1 1 0.5 3. AXISYMMETRIC CONFIGURATIONS 3.5 1.5 4 4.1 1.6 1.1 1.4 1.1 1.4 13.5 3.1 1 1.5 (b) Figure 6.1 1.5 9.6 11.2.4.4.8 1 1.1 1 1 1 1 1 1 1 1 1 1 1.6.1 1 1 1 1 1 1.7 13.3 0.8 0. D = 3 m. Multiplication factor for element forces was 3.8 1.6 1.10 and δan /δsag = 0.5 1.4 1. ρ = 0.1 1 1. 131 .5 4 4.9: Normalised prestress distribution (a) under tension tie forces (b).5 1.
9 1 1.3 2.5 (b) Figure 6.8 1.9 1 1.9 3.3 1 1.1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0.CHAPTER 6. F/D = 0.2 10.10 and δan /δsag = 0.4.1 0.8 1 1.5 4 5 4 5 4.7 1.4 13.1 0.7 11.9 0.1 0. 132 .6 15.1 (a) 5 4.1 0.6 15.2 1. ρ = 0.10: Normalised prestress distribution (a) under tension tie forces (b). Multiplication factor for element forces was 3.1 1 1 1 1 1 1 1 1 1 1 1 1 0.3 0.7.2 0.1 1 0. ANALYSIS OF TENSION TRUSSES 3.1 1.4.5 1.8 1.8 1.1 1 1.9 1.1 1.8 1.3 0.7 1.9 0.9 1. D = 3 m.4 13.5 1.9 0.1 1.8 1 1.1 1 0.7 1.7 1.2 0.2 11.7 1.1 1.4 0.
3. A preliminary analysis of a threering oﬀset tension truss showed that a focal length less than 1. If.10. From a blockage viewpoint. The vertical joint positions are recalculated using (5. e. the tension tie forces were changed.8 and 2. however. To determine the sensitivity of the force distribution in the tension truss to the ellipticity.363 and XA /Da varies between 0 and 0.2. Thus. 6. in the present case where the minor axis of the ellipse is equal to D. Three diﬀerent focal length were also used: 1.167 and 0. only the xcoordinate needs changing. 1.5. Initially. (5. a larger value of XA is better. where ξ is the ellipticity. F/Dp varies between 0.5 m was more diﬃcult to prestress. (a) (b) Figure 6. OFFSET CONFIGURATIONS 6. Other issues.g.10).11: Two symmetric oﬀset conﬁgurations for the tension truss: (a) conﬁguration 1 and (b) conﬁguration 2.1 Focal Length and Oﬀset Distance For the oﬀset antennas in Table 5.3. 133 . due to the change in geometry.8).17. the size of the spacecraft bus and the antenna support structure.6.1 and 0.1 m. 0. the tension tie forces applied to the edge and additional nodes of the oﬀset tension trusses were identical to those of the axisymmetric cases.7–6. three diﬀerent realistic values of XA /Da were used: 0.1. may prescribe both the focal length and oﬀset distance. Figures 6. some of the elements ended up in compression. ξ. x := ξx.3 Oﬀset Conﬁgurations The oﬀset version of the tension truss is created by simply mapping the positions of the joints of the circular aperture onto an elliptic one.
5 2 2. The tension tie forces on the edge and additional nodes had to be changed slightly to give a satisfactory force pattern in the truss. The second alternative. The results for the oﬀset tension trusses are given in Table 6. based on the assumed simpler attachment.5 4 4.2 Two Symmetric Oﬀset Conﬁgurations There are only two ways of arranging the tension truss in the oﬀset conﬁguration which preserve a certain symmetry. is to let two opposite support nodes lie along the major axis of the elliptic aperture.5 2 2. The new values for three to seven rings are given in Figure 6.5 3 2 2.11(b). in terms of prestressability. conﬁguration 1 may be easier to attach to the spacecraft with nodes along the major ellipse axis.5 4 4.12.5 4 4. A preliminary analysis with a threering net showed that.11(a). only conﬁguration 1 will be used in the following analyses.5 3. Figure 6. ANALYSIS OF TENSION TRUSSES 6.5 3. (b) four. The value XA = 0 gives the lowest ratio for the internal forces. no conﬁguration could be said to be better than the other.5 2.5 3 2. Figure 6. (d) six.2.5 4.CHAPTER 6.5 6 5 7.5 4.5 6 4.5 3 4.5 3. The ﬁrst alternative.5 3.5 4.11.5 6 4 4. (c) ﬁve.5 3 2.5 6 6 4. Figure 6. but not necessarily the lowest for all forces.5 6 4.5 5 7.5 (c) 6 5 7.12: Tension tie force distribution along the edges and on the extra nodes to obtain an adequate internal force distribution in the tension truss of conﬁguration 1: (a) three. Interior elements are drawn thinner for clarity.5 A (a) (b) 3.5 4.5 6 4. but this is only an assumption.5 4. 134 .5 3 2 3 2. and (e) seven rings.5 2 2 2.5 2.5 6 4 4.5 3.5 6 4 3.5 6 3.5 6 3. diﬀers from the ﬁrst by a 90◦ rotation so that the opposite nodes now lie along the minor axis. Since there was hardly any diﬀerence in the force distribution between the two conﬁgurations.5 (d) (e) Figure 6.3.5 2. 2 2.5 2 2 B 2. From a more practical reasoning.
63 12.158 1.2: Tension truss element forces and their ratios (oﬀset conﬁguration 1).33 20.118 1.307 0.56 11.64 2.52 6.62 2.37 6.292 Ha (m) 0.34 3.79 4.92 2.38 22.1 1.21 2.64 2.05 4.240 0.221 1.53 2.86 106.227 0.92 149.20 2.5 7 1.252 0.083 1.67 2.227 0.5 6 1.221 1.40 54.37 tint max (N) 4.273 0.34 2.55 3.95 1.52 58.300 0.58 2.6 0 0.45 154.19 3.59 10.252 0.318 0.55 1.32 62.6 0 0.252 0.3 0.88 8.80 3.35 44.05 2.35 2.6 0 0.74 tall max (N) 18.76 1.86 2.280 0.82 2.00 2.6 0 0.54 23.270 0.3 0.18 61.288 0.350 0.158 1.270 0.227 0.60 2.322 0.32 2.40 120.6 0 0.6 0 0.63 2.55 11.44 5.118 1.307 0.39 2.38 11.350 0.240 0.90 5.78 57.49 9.25 12.60 32.088 1.158 1.39 35.51 9.118 F Dp 0.22 2.250 0.31 5.75 18.27 7.322 0.221 1.166 1.88 2.67 2.03 6.02 10.06 1.63 47.288 0.30 3.06 1.92 44.158 1.02 16.99 2.3.20 99.3 0.250 0.250 0.280 0.20 2.083 1.208 0.52 3.322 0.26 3.62 1.6 0 0.83 28.288 0.45 6.85 13.6.88 10.46 77.29 3.208 0.062 1.118 1.270 0.42 3.85 8.38 5.083 1.61 17.31 7.00 5.240 tint min (N) 1.74 3.03 2.00 2.288 0.57 97.280 0.40 4.38 2.36 1.307 0.59 5.118 1.062 1.40 3.91 7.05 0.95 1.8 2.00 2.166 1.300 0.062 1.221 1.12 3.088 1.78 1.240 0.81 59.1 1.80 4.288 0.08 4.240 0.1 1.69 21.12 15.8 2.41 21.45 4.86 16.14 3.335 0.280 0.93 2.75 56.158 1.8 2.51 51.45 3 1.44 20.50 2.250 0.27 2.13 3.10 2.3 0.246 0. The best ratios are in bold face.083 1.208 0.322 0.38 140.166 1.33 36.30 22.062 1.3 0.65 1.74 4.04 3.5 5 1.33 10.61 21.6 ξ 1.81 1.246 0.50 20.93 3.10 1.01 3.44 17.300 0.20 1.75 9.54 3.250 0.20 2.71 1.43 20.30 9.250 0.14 43.60 1.66 20.41 114.64 2.16 5.02 78.60 2.46 2.5 XA (m) 0 0.30 8.34 31.166 1.01 1.56 4.6 0 0.86 7.21 71.61 65.118 1.41 6.05 2.3 0. OFFSET CONFIGURATIONS Table 6.82 7.18 24.06 42.292 0.246 0.250 0.98 3.56 7.41 12.3 0.31 2.05 2.318 0.10 23.96 35.70 68.49 1.6 0 0.81 1.350 0.38 2.53 2.3 0.307 0.99 2.83 86.3 0.10 7.71 1.75 5.6 0 0.335 0.38 12.118 1.67 102.322 0.05 0.91 56.63 2.11 4.118 1.20 2.292 0.3 0.22 4.088 1.95 2.93 1.33 126.03 20.8 2.21 2.69 1.8 2.307 0.53 29.57 3.36 1.118 1.38 2.11 43.63 22.227 0.92 3.27 2.88 3.66 10.87 23.273 0.3 0.63 88.6 0 0.08 144.67 tall min (N) 1.118 1.250 0.335 0.335 0.47 7.82 3.270 0.17 59.84 129.60 2.208 0.3 0.6 0 0.088 1.11 tint max tint min 2.252 0.246 0.6 0 0.250 0.318 0.335 0.74 2.3 0.00 2.48 23.083 1.27 34.90 2.273 0.17 8.23 134.118 1.01 2.87 43.20 1.10 4.64 2.166 1.16 3.62 1.27 11.227 0.64 23.300 0.67 9.95 92.350 0.088 1.1 1.90 2.87 50.68 2.71 36.318 0.318 0.118 1.00 15.59 46.79 63.83 48.31 3.18 9.6 0 0.21 20.50 tall max tall min 11.18 21.54 37.46 95.88 27.246 0.221 1.118 1.36 51.250 0.3 0.28 7.39 2.39 2.270 0.292 0.20 6.273 0.118 1.39 19.93 2.280 0.07 4.52 58.93 24.252 0.350 0.062 1.300 0.44 19.1 135 .03 10.29 6.57 31.292 0.61 10.5 4 1.16 2.59 20.39 14.3 0.118 1.08 11.208 0.71 2. n F (m) 1.273 0.
394 mm for the former case. [97]. The element length imperfection and tension tie load variation are not the only manufacturing errors.7 m and σε = 10−5 in (6. Greene [49] continues the analysis of the tetrahedral truss structure using the Monte Carlo technique and a linear Finite Element (FE) model to obtain results for a reﬂector with a low number of facets. can be said to represent the minimum theoretical error. The rms surface error is computed for two error sources: member length imperfection and tension tie force variation. the systematic deviation due to surface faceting. (6.z = 0. Random errors related to the present antenna are: • Member length imperfections. which agrees well with their simulated value.11). only a few studies are available in the literature. The average rms surface error of 100 simulations is 0. The rms surface error of a geodesic dome. • Ring structure distortion.CHAPTER 6. In addition to the surface accuracy. not be determined analytically. Hedgepeth [59] develops an equivalence between the statistical errors and the natural frequencies of a reﬂector antenna structure assembled by tetrahedral truss elements. For two and four rings it is about three and two times higher than that predicted by (6. [62] uses the Monte Carlo technique for a nine ring geodesic dome (F = 15. is [59]: δrms.11) is based on a continuum approach and valid only for a large number of facets. It is found that the defocus.11) gives δrms. Unavoidably. In such cases the alternative approach is the Monte Carlo technique.29).4 Eﬀects of Manufacturing Errors on the Reﬂector Accuracy As described in section 5. Greene also analyses the reﬂector defocus and variation of the forces in the tetrahedral truss. More recently. in most cases. It should be noted that (6. thus becoming a geodesic dome. These three studies provide important information on the eﬀects of manufacturing errors on diﬀerent reﬂector parameters. Greene ﬁnds that the surface deviation increases as the number of facet rings decreases. For the tension truss. the performance of a reﬂector antenna is mainly determined by the manufactured surface accuracy of the reﬂecting surface. with an errorfree rim.11) where σε is the standard deviation of the member length imperfections. increases with a decreasing number of facet rings.7 m) ﬁxed at its rim. cf.1 N is 0. These errors are randomly distributed in the structure and their inﬂuence on the surface accuracy can. like the surface accuracy. (5.z = 2F σε .314 mm.4.160 mm. Hedgepeth’s analysis results in formulae which can be used for the preliminary design of such antennas. the minimum achievable surface error is higher due to errors in the manufacturing process. The average rms surface deviation due to a tension tie force variation of 0. 136 . Inserting F = 15. An interesting case of his analysis is when the height of the truss structure approaches zero. Despite the signiﬁcant importance of manufacturing errors. respectively. ANALYSIS OF TENSION TRUSSES 6. Hedgepeth et al.
The normally distributed random strain ε was generated by the Matlab [89] routine randn.6.4. although the edge elements most likely have to be stiﬀer to sustain the higher forces.14) i. etc. the tension truss with the imperfect element lengths and unchanged tension tie loads was analysed by a ﬁnite element program using nocompression catenary cable elements. These forces were used to compute the unstrained length 0 of element i as: 0. 6. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY • Tension tie load variation.i (1 + ε). Then. the conﬁguration of the errorfree tension truss was generated by the routine in Appendix C.13) where ε ∼ N(0.e. In the FE analysis it was assumed that all elements have the same properties. To account for length errors in the assembly of the tension truss. These catenary elements require a small selfweight to avoid numerical problems. which minimises the rms surface deviation. the combined eﬀect of all the random errors has to be determined to conclude on the achievable surface accuracy of the reﬂector. σε ).i = i 1 + ti /AEi . The computation of the bestﬁt paraboloid to a set of points on the reﬂector surface can be separated into two parts [49. normally distributed with a zero mean value and standard deviation σε .12) where t is the force in each element and AE the axial stiﬀness. Other sources that aﬀect the accuracy of the reﬂector are gravity during ground testing. [171]. (6.i = 0. Then. was introduced. its inﬂuence was neglected as it was chosen to be much smaller than the tension tie loads. (6. (6. the element forces under the prescribed tension tie loading were computed by the force method. which will theoretically produce over 21492 ≈ 10449 values before repeating itself. First. However. the bestﬁt paraboloid.2 BestFit Paraboloid Analysis In section 5. and • Random thermal strains. it is important to ﬁnd how each of these errors alone aﬀects the surface accuracy and defocus in order to establish the manufacturing tolerance level.1 Monte Carlo Technique for the Tension Truss The technique used to calculate the eﬀects of manufacturing errors makes use both of the force method and of the FEM.4. First. the unstrained length of element i was modiﬁed as ˜0.4. 6.4. cf. Then. 140]: 137 . attitude accelerations.
For a reﬂector surface represented by a discrete set of points. 2 2 An (Xn + Yn ) An A n Zn Equation (6. ∆Z .z = 1 n i=1 n Ai ˜ Ai Zi − Zi i=1 2 1/2 . a readjustment ∆F of the equivalent focal length. . one is about 5·10−4 and the other about one order lower. P . followed by 2. Y2 . 138 . this will not be included here considering the simpliﬁcations already made.15) + ∆ZP . Y1 . Z= 4F where F denotes the bestﬁt focal length and ∆ZP the translation of the apex of the ideal paraboloid in the XY Z system. a rigid body ﬁt. Zn ) be the coordinates of n points on the reﬂector surface. (Xn . thereby creating N 2 subtriangles. . Z2 ).17) Note that for oﬀset reﬂectors the points should be measured in the coordinate system of the parent paraboloid. . A more reﬁned weighting can be done by also taking into account the radiation pattern over the aperture [178].and Y axes are very small.16) is solved by a standard least squares solution technique. when calculating the rms surface error. which is simply the total translation of the focal point.15) and multiplying with the horizontal projection of the surface area associated to each point lead to the following overdetermined system of linear equations: 2 A 1 Z1 A1 (X1 + Y12 ) A1 A2 Z2 A2 (X 2 + Y 2 ) A2 2 2 1/4F = . Let (X1 . where a ﬁtted paraboloid of ﬁxed focal length undergoes translation along the Zaxis and rotates about the X. XY Z. (6.and Y axes were assumed to be negligible in the present analysis. Z1 ). the axial rms surface deviation is computed as: δrms. However.16) . As a result. Subsequently. the Zaxis of the bestﬁt paraboloid was always parallel to that of the ideal paraboloid. the contribution of each point in the computation of the bestﬁt paraboloid will vary. ∆F + ∆ZP [49]. Yn . Another important parameter for the performance of the reﬂector is the defocus. (6.. Thus.CHAPTER 6. Substituting each of the n points into (6. The centre of gravity of each subtriangle was taken as a point in the bestﬁt paraboloid computation and the horizontal projection of the subtriangle area was taken as its weight factor. Lai [78] uses a similar approach to compute the surface error for the CRTS reﬂector. . (X2 . .. This is taken care of by weighting each point by its associated area.. They show that the rotations about the X.24). If the n points are not equally distributed over the aperture. the rotations about the X.and Y axis. cf. ANALYSIS OF TENSION TRUSSES 1. These simpliﬁcations may also be warranted by the use of (5. instead of (5. Some results of bestﬁt paraboloid computations are given in [62] and [140].21). dense points are given lower weights than sparse points. In the present bestﬁt analysis the sides of each triangular facet were divided into N parts of equal length ∆ = /N . Therefore. [156]. . the equation of the bestﬁt paraboloid is X2 + Y 2 (6. .
2.1 and z. zsy denotes the zcoordinate of the reﬂector surface with only the systematic error present. respectively.24) is written as: 2 δrms. However. It is not exactly equal since the calculation of the bestﬁt paraboloid also includes z .2 ) 2 1 + Aa Aa 1 2 z.z = 1 Aa 1 Aa 1 Aa (∆z)2 dA Aa = = ((zsy + Aa z) − zbf )2 dA (6.1 dA.4.6.j are very small and can be neglected. Integral II is the square of the additional surface deviation due to manufacturing tolerances. Thomson [166] uses a similar technique to compute the total rms surface deviation. In the following. or the rss surface deviation as he calls it.i z. With some further simpliﬁcations it is also possible to approximately ﬁnd the separate eﬀects of diﬀerent random deviations. from several error sources. and z the additional contribution of the random error in the zdirection. First. zbf the zcoordinate of the bestﬁt paraboloid (to both errors). II I Integral I is approximately equal to the squared systematic rms surface deviation due to the facet approximation. the systematic error due to faceting will always exist so the contribution from the random rms surface error must be extracted from the total rms surface error.19) + z.i is small this technique can be extended to an arbitrary number of random errors.4. assume that the total error is caused by one systematic and one random error source. (5.2 .24) can be expanded as follows: 2 δrms.1 (zsy − zbf )dA (6.3 Extracting the Random Surface Deviations The routine for computing the bestﬁt paraboloid. Thus. the second and third integrals are the inﬂuences of z. z = z. If the individual random error z.2 . 139 . the rms surface deviation approaches that of the ideal faceted reﬂector antenna. rms surface deviation and defocus does not distinguish systematic errors from random errors. Equation (5. which means that the fourth integral vanishes. Thomson’s results are reproduced in Table 5.18) 1 Aa 2 z Aa (zsy − zbf )2 dA + Aa + 2 z (zsy − zbf )dA .z = 1 Aa (zsy − zbf )2 + Aa 1 Aa 2 Aa z. A technique to extract a single or several random errors from the total error will now be introduced. For z → 0. Terms of type z.1 + z. For the case of two random deviations. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY 6.2 (zsy − zbf )dA + Aa ( Aa z. but computes the total rms surface error.
The results 140 . When the equilibrium conﬁguration of the tension truss was computed using the FEM. The results of the systematic error computations are shown in Table 6.4. The analysis of the systematic rms error were done for three to nine rings. the analyses concerned with the random errors were only done for three rings. The nodal positions of the tension truss were the ones generated by the mesh generation routine in Appendix C. to determine the asymptotic values of the rms surface error. however. 102 . Hedgepeth et al. that the validity of this assumption was not put to test in this thesis. The computations of the systematic errors in the previous section were based on the generated nodal positions.6. Note.. 202 . Thus. where superscripts Ag and 25 indicate the values by (5.29) and a 252 subdivision. For signiﬁcant diﬀerences.4 Systematic Facet Surface Deviation Before the inﬂuence of the random errors can be computed the systematic rms surface deviation and defocus have to be determined. and δan /δsag = 0. The defocus decreased with increasing number of rings as it should. thus the reliability of the bestﬁt routine was conﬁrmed. the threering tension truss was analysed for two diﬀerent interior tension tie loads: 1 and 10 N.59]. To determine if the elastic deformation was large enough to aﬀect the systematic error. 6.4. In this. the previously computed systematic error cannot be used when extracting the random errors from the total error.6 and 0. the nodal positions diﬀered due to load deformations.5 Inﬂuence of Tension Tie Loading While the computation of the systematic surface errors was done for three to nine rings. The 252 subdivision was chosen to be used in the subsequent computations of the random error as it produced values close to the asymptotic values and required less computation time when the entire tension truss has to be analysed. only one sixth needed to be analysed. ANALYSIS OF TENSION TRUSSES 6. respectively. cf. an estimation of the required manufacturing tolerances can be done by analysing a representative antenna conﬁguration. The systematic rms surface error was computed for 52 .29). although a quite extensive one. As the tension truss is sixfold symmetric. For example. (5. The results from all except the 252 case were used in a Richardson extrapolation scheme.29) it was necessary for the following analyses to compute the error by the bestﬁt computation described above. [62] perform Monte Carlo simulations on a nine ring geodesic dome with ﬂexible members in order to get accuracy estimates for the real one with 27 rings.8.4.3. a smaller number of rings would produce a larger rms surface error and more severe defocus. 252 and 402 subtriangles in each triangular facet. both of the computed rms surface errors agree well with the one by (5. Throughout the range of rings. 0. Since this is a feasibility study. It was further assumed that the dependency of the rms surface error and defocus on the number of rings was similar to that of the tetrahedral truss structure studied in [49.29). and the following analyses of the surface deviation. Values without superscript are the Richardsonextrapolated ones. Although it was possible to get a good estimate of the rms surface error by (5.CHAPTER 6. F/D = 0. the parameters of the antenna were: D = 3 m. [29]. They were then compared to those by Agrawal et al.
6.3736 0.4728 0.8877 0.4 0.5000 0.1959 25 δrms.4589 3.3341 0.6 0.4.0435 15.3417 22.2413 1.9305 1.4801 10 3.3: Surface deviation (rms) and defocus of a paraboloid due to triangular faceting.7322 8. the rms surface deviation was unaﬀected by the higher tie load.6046 9.2613 0.4 0.5559 5.2500 0.5849 25 ∆F 25 + ∆ZP (mm) 22. In the studies which follow.3277 0.z (mm) 3.6253 0.8419 0.6903 3.4370 0.2364 0.4630 22.2490 0.4448 0.6272 14.2506 0.4439 0.2607 0.3088 0. F/D tint tie δrms.8405 0.6710 24.2279 22.2841 0.3417 2. an interior tie load of 5 N was used to further reduce the eﬀects of the small selfweight of the catenary elements. a smaller tie load can be used.6810 1.4379 0.6315 0.9075 1.6 0.4 0.6 0.5954 2.3740 5.3499 1.2629 0.6 0.8324 3.2870 0.6710 24.6555 0.6569 0.5011 0.9085 1.6223 1.6 0.8069 0.z ∆F + ∆ZP 0.7392 8. The higher tension tie load gave slightly lower values for the defocus.5967 2.5931 0.9456 1.z (mm) 3.8 0.3496 2.8 0.6416 2.1868 δrms.4: Inﬂuence of deformation due to tension tie loading (threering tension truss).3715 5.8911 1. n F/D 0.5856 are shown in Table 6.4117 0.2279 22.8897 0.5603 0.4728 22.3417 22.9261 1.8 0.6231 1.9909 13.3620 2.3153 2.4 0.4852 8.2607 0.4695 1 2.6709 1.9575 141 .2278 1.6921 3.4.4 0.2331 1.9630 1.4202 0.4 0.6115 1 1.4574 3.4 (N) (mm) (mm) 1 3.8349 3.8 0.5516 5.6399 2.9765 0.6052 0.3142 2.4 0. This was due to the increased vertical deformation of the net.4803 5.6208 14.6748 1.8401 0.3285 0.0371 15.4830 5. However.5918 0.3910 0. In a zerogravity environment.3333 0. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY Table 6.6301 0.1955 ∆F + ∆ZP (mm) 22.6 0.6 0.3919 0. Table 6.3152 0.3489 1.8 3 4 5 6 7 8 9 Ag δrms.6175 0.6 10 2.2601 0.8 0.9814 13.z (mm) 3.6142 24.8 0.2103 0.9652 1.8 10 1.4811 8.6233 24.6086 9.
To ﬁnd a representative maximum value of the parameter. Another important statistical aspect is the required number of Monte Carlo simulations.4. were computed for 100 to 500 simulations.i i=1 1/2 . Greene [49] also uses 100 simulations for the rms surface error but ﬁnds it necessary to increase the number of trials to 200 to determine the defocus with reasonable accuracy. Hence. The quality of the simulated values was estimated by the standard deviation of all the values generated with randn. Here. since the number of trials was quite small the maximum computed value pmax could not be taken as the highest value likely to occur. Hedgepeth [58] ﬁnds that “a value of σε of 10−3 is representative of ordinary careful practice.7 Inﬂuence of Member Length Imperfections The ﬁrst of the random error sources is the individual lengths of the elements. The standard deviation of S simulations was approximatively computed as σrandn = 1 S S 2 σrandn. only the standard deviation of the surface error had diﬀerences of more than 1%. A maximum value often used is p = p +3σp . This value was chosen to represent the worst possible value of the studied parameter. The rms surface error and defocus. It was decided to analyse the present threering tension truss for three tolerances: σε = 10−4 . [62] use 100 simulations for their two cases. it was concluded that 100 simulations was suﬃcient for the Monte Carlo analysis of the tension truss. (6. a tension truss with member length imperfections was chosen as a test example. length tolerances σε as low as 3 · 10−7 has recently been achieved for the AstroMesh [166]. the diﬀerences were very small.i is the standard deviation for simulation number i. Of importance to the performance of the antenna is also to maintain a suﬃcient stress level in the tension truss. a probability limit had to be set.20) where σrandn. The fabrication tolerance that can be achieved depends highly on the manufacturing costs that can be accepted.6 Statistical Considerations The statistical treatment of the results from the Monte Carlo trials was more or less straightforward. the parameters are the rms surface error and defocus. Hedgepeth et al.5.CHAPTER 6.4.” However.00135 that any simulated ¯ value will be larger. Table 6. However. and their corresponding standard deviations. ANALYSIS OF TENSION TRUSSES 6. which should be close to one. and 10−6 is very diﬃcult and costly. The results for the rms surface error and defocus are given in Table 6. which corresponds to a probability of 0. the maximum and minimum values of p were also extracted. 10−5 is achievable with welldesigned and operated hard tooling. 10−5 and 10−6 . 10−4 is characteristic of a highquality machine shop. Overall.6. The lowest net element force was therefore sought for each 142 . From the trials. To determine how the surface error and defocus vary with the number of Monte Carlo simulations. The mean value of a parameter p is denoted p and its standard ¯ deviation σp . 6.
The interval and mean value (in parenthesis) of the lowest element force for F/D = 0.43 N). 12.23 22.341749 — 0.4.26–13.1.469767 −0.6.4 the lowest force was lower than 0. and 10. The case of the lower accuracy.8 N). yielded unsatisfactory results and therefore the higher accuracy was required.4.1 were 0.43 22.470104 — 0.005031 0.56 22.469671 −0.4 N). for the same values of F/D.6.25 N.000017 0.005107 0.0463–5.18 N).000191 0.6 they were 0. Hedgepeth et al. 5.05.03 N).54 N (5.85 N (3.11 143 .97 N (6. For F/D = 0. 6.z Diﬀerence σδrms.3–10.000015 0.0642–11. which is necessary to avoid severe mesh saddling.8 with στ = 0. ρ = 0. Apparently. 5. the achievable force accuracy is much lower.78 N).000059 7.4 and σε = 10−4 .7 N (8. There was a risk.177–5.000019 0.09 N (7. Table 6. Compared to the length imperfections.218–8.4.000062 11.8–13.z Diﬀerence ∆F + ∆ZP Diﬀerence σ∆F +∆ZP Diﬀerence 100 (mm) (%) (mm) (%) (mm) (%) (mm) (%) 3.068 300 3.000060 9. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY simulation.341747 −0. that some cables become slack.1 N for a magnitude of 1 N.8 N (13.25 N in only two cases of the 100 simulations.000185 0.8 Inﬂuence of Tension Tie Load Variation The tension ties are socalled constantforce springs with relatively low stiﬀness so that an elongation results only in a small change of force.000070 0. In the following analysis.51–6. a tolerance of σε = 10−4 cannot be accepted.5: Surface accuracy and defocus for a threering tension truss as a function of the number of Monte Carlo simulations (D = 3 m. 10−5 and 10−6 the interval and mean value (in parenthesis) of the lowest element force were 0. 0.004709 — 200 3.7 N).4–13.5 N (10. F/D = 0.82–10.57– 6.469726 −0.4 N (11. στ = 0. For F/D = 0. and 13.1 and 0.04–11.341743 −0.8 N (6.005186 0. respectively. otherwise it was higher than 1.05. respectively. For F/D = 0. 0.084 400 3. and 0.000150 0. although slim.87 N (8.46–9.0635–8.2 N).63 N (3.000060 9.56 N). στ = 0.4.005233 0.1 and σε = 10−6 ).18 N).8 they were 0.05 provided a better chance of keeping a good prestress in all elements.16 N (5.56 22.4 N (9.341743 −0. gave intervals and mean values 3.469752 −0. Number of simulations ¯ δrms.000016 0.9 N (13. Using the smaller tolerance. and 9.56 N). στ = 0. and 6.38 N). two tie force variations were used: 0.97 N (1.57 N). For F/D = 0. [62] use a standard deviation of 0.10 500 3.341744 −0.000055 — 22. and 1.52 N).95–7.62 N). 8. Clearly.
2436 2.0060 ∆F (mm) 9. 100 Monte Carlo simulations).61259 23.0110 ¯ δrms.00161 F/D σε 10−4 (mm) 0.z ¯ δε (mm) 0.22789 1.00434 1.5946 22.02917 6.0134 17.0259 0.3476 25.67121 1.8307 0.3528 8.03921 0.01277 8. ANALYSIS OF TENSION TRUSSES 144 max δrms.2280 2.03065 24.6073 0.38325 18.65139 6.64752 6.z (mm) 3.7177 9.31619 18.1251 13.0115 0.4896 ∆F + ∆ZP (mm) 22.7310 1.0770 .6739 1.3431 3.30200 0.5273 max ∆ZP min ∆ZP (mm) 13.2548 ∆F min (mm) 8.6471 6.34162 2. 0.77194 13.70852 0.3359 0.3000 (mm) 13.0621 0.0299 0.96371 1.6061 0.3048 24.3400 0.6: Inﬂuence of member length imperfection on the rms surface deviation and defocus (threering tension truss.6714 1.1397 ∆F max (mm) 10.0069 3.3028 13.74633 13.2922 6.2783 6.6754 0.34191 2.4 10−5 0.2569 9.2139 8.9002 8.3371 σδrms.86319 0.00263 0.5284 13.4521 0.6694 0.4721 1.7298 14.0597 6.64370 0.2079 σ∆F (mm) 0.4594 9.00006 0.2303 2.4388 0.16810 13.9639 1.47010 22.00263 22.8521 9.3322 0.02003 0.67085 0.z (mm) 3.9957 0.30380 13.00079 13.0401 13.3418 3.67100 1.8 10−6 0.1590 0.6164 7.86626 8.0006 9.2258 3.2286 2.7774 3.2691 2.15601 10.6791 6.Table 6.3489 CHAPTER 6.00007 13.34175 2.2856 0.4112 ε δrms.01828 0.7036 1.29985 13.4219 ∆ZP (mm) 13.0007 3.03469 0.3245 18.1323 8.z (mm) 3.01375 3.9141 17.9982 1.4652 0.86875 8.1142 9.9716 1.00112 3.1693 9.2658 0.00403 1.8648 8.24982 0.0019 1.02239 0.1494 σ∆ZP (mm) 0.6 0.0009 18.2195 rms.4397 ε σrandn 1.4989 0.z (mm) 0.z ε σ∆F +∆ZP (mm) 0.2262 0.00007 18.1424 0.3211 13.3806 22.0079 0.0090 0.5796 8.1035 0.4273 0.0470 22.0026 1.1013 0.22810 2.3693 min δrms.0409 0.00471 1.22772 0.17654 17.03656 0.0146 10−6 10−4 10−5 10−6 10−4 10−5 0.1127 0.
0366 0. and zdirections were normally distributed with a zero mean value and the same standard deviation.2816 13.34319 3. y. 10−5 and 10−6 .05704 0.64610 6.0035 0.9964 0. The results from the ring distortion analysis are shown in Table 6.z τ σ∆F +∆ZP ∆F + ∆ZP (mm) τ σrandn ¯ δrms.00057 9.4641 22.0013 9.0680 σδrms.2240 0.22572 0.9986 1.06349 13.30211 13.00099 24.99714 1. Hence. the nodal distortion in each of the x.14592 0. In the real antenna structure this type of error arises due to length errors in the members of the ring structure.23906 16.0387–5.05 0.z min δrms.2798 8.00470 σ∆ZP 6.26742 19.31591 13.2262 18.7: Inﬂuence of tension tie force variation on the rms surface deviation and defocus (threering tension truss. were used in the ring structure distortion analysis.43429 0. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY Table 6.00994 18.00001 3. The member length imperfection inﬂuence analysis indicated that the accuracy level 10−4 could produce cable slackening.88727 8.12845 0.29318 0.z max δrms.716 N).4.7357 rms. Therefore.3482 8.8.1 0. which are of two diﬀerent types: cables and struts. y.23053 2.2461 14.67199 1.66735 0. and 0.0027 1.00174 (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) 3.2504 17.22853 2.6 0.1656 0.59766 0.6447 6.6875 1.86544 8.4 0.0790 0.2968 0. and 145 .14942 0.2274 0.9073 8.11030 0.6.05 0.0219 13.and zdirections was computed as D ρ .z ¯ δτ 0.8618 8.2340 2. 100 Monte Carlo simulations).3253 13.0765 0.74996 14.1281 0.7302 6.05 0. For each node.47280 1.34181 3.67601 1. σρ ). 0.8709 24. only the two higher accuracy levels.03707 0.17070 9.91352 0.61540 0.01703 13.1854 0.6891 13.5509 22.9957 2. It is likely that one of these can be manufactured to a higher precision than the other.1 (mm) (mm) (mm) 0.0282 0.9 Inﬂuence of Ring Structure Distortion In the following analysis of the inﬂuence of the ring structure distortion.z 22.7116 6.41423 0.68603 6.5870 0.4.1629 9. this was the basis of the assumption of using a single standard deviation.84171 0.34065 0.6.8 with σρ = 10−5 were 0.41983 6.0190 0.6684 0. the least accurate of the types will give rise to the largest error at the ring nodes.0021 1. 0.9889 19.8 0.z ∆F ∆F max ∆F min σ∆F ∆ZP max ∆ZP min ∆ZP 13.2292 2.8138 0.4.24285 9.03307 22.00810 2. The interval and mean value (in parenthesis) of the lowest element force for F/D = 0.1 0.3012 13.3419 3.0342–8.02711 0. where ρ ∼ N(0.3451 3.06173 0.14064 8.3379 0.6758 1. F/D στ τ δrms.0096 0. it was assumed that the random distortions in x.26 N (0.19885 0. σρ .08753 0.08 N (0.725 N).
respectively.00955 0.z max δrms.2530 23.2262 0.19524 0.6 10−6 0.08298 0.67054 0.3608 1.4 N (7.29415 0. Nevertheless.0486 13.16–10.7094 1.00026 9.8320 0.63750 6.03994 6.4442 0.3125 6. 5.22673 0.00945 22.85 N).3672 12.65153 0.4 N (1.1382 8.34195 3.71798 0.0368–10. With σρ = 10−6 .4979 0. (6.6794 17.6941 19.18197 9.3392 0.00510 10−5 0.3351 0.77344 13.66412 6.22889 2.5018 9.01408 10−5 0. 100 Monte Carlo simulations).0027 2. the individual inﬂuences of member length imperfection.z ¯ρ δrms.98562 3.27448 18.31591 13.3623 3.7981 0.8: Inﬂuence of ring structure distortion on the rms surface deviation and defocus (threering tension truss.2301 ρ σ∆F +∆ZP ρ σrandn ¯ δrms.8 10−6 0.00060 18.6198 1.1475 12. and distortion of the supporting ring structure on the rms surface deviation and defocus were examined. and 10.67198 1.9–13.z min δrms.CHAPTER 6.04088 0.84509 13. F/D σρ ρ δrms.0355 14.00060 13.09032 0.7685 21.7002 0.2674 7.4439 1. ANALYSIS OF TENSION TRUSSES Table 6.34252 3.z ∆F + ∆ZP (mm) σδrms.46750 0.02 N (6.98153 2.85469 0.9861 (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) 3.06146 0. 6.i .5959 13.2003 6.2716 2. the corresponding intervals and mean values were 5.0315 17. An estimate of the surface accuracy can be obtained by computing the total rms surface deviation from all error sources as 1/2 δrms. tension tie force variation.34130 0.01156 0.7381 1.22783 2.z ∆F ∆F max ∆F min σ∆F ∆ZP max ∆ZP min ∆ZP σ∆ZP 0. the tension truss is more sensitive to distortions of the supporting ring than to member length imperfections.4.3308 2.90543 8.3507 3.6710 0.7152 1.1724 24.8 N). Evidently.1928 8.z 0.10 Combined Inﬂuence of Manufacturing Imperfections In the previous sections.04861 24.77 N).37599 18.00060 1.67288 1.30401 13.8327 7.21) 146 .z = i 2 δrms.04909 22.14827 0.9586 13.53–7.91198 1.87810 8.2454 0.9405 9.49 N).4 10−5 (mm) (mm) (mm) 0.1832 10−6 0.03537 8.z.16349 9.6798 0.00695 13. respectively. which is not the case with the individual member lengths.5504 0.7707 1.0068 8.1364 0.62382 0.09899 0.9723 1. it is possible to adjust the ring structure during the assembly process. The manufacturing accuracy of the real reﬂector surface is determined by the combined eﬀect of all these imperfections.5268 0.7 N (11.
The combined eﬀect of the random errors will increase the risk of cable slackening. and solving for the worstcase hot temperature yields [119] Tmax qIR εIR sin2 ρ + GS αS (1 + aKa sin2 ρ) = 2εIR σ 1/4 .10 N). so high accuracy levels are needed. The equilibrium temperature of a body in space is estimated by an energy balance.4. assuming the same IR emissivity and solar absorptivity on the top and bottom surfaces of the net bands.58 N (1.140 m).938–6.27 N).4. 5. For level II. To analyse the inﬂuence of thermal strains on the surface accuracy. using the same order as above. During a mission.05). Setting up this energy balance. for F/D = 0. In the energy balance of a body in space. εIR the IR emissivity on the band surface. subjected to extreme temperatures. Each element of the tension truss is treated as a thin ﬂat plate with no side insulated and whose surface normal is parallel to the solar rays and passes through the centre of the Earth.8.4. 0. Ka a correcting factor for the reﬂection of the 147 . ρ the angular radius of the Earth (sin ρ = RE /(h+RE )).0283–5. and 0.28–12. thus.30±0. Hence. the algorithm used for computing the rms surface deviation does not take cable slackening into account.0181–2. GS the solar ﬂux (1326–1418 W/m2 depending on season). αS the solar absorptivity on the band surface. 6. the interval and mean values of the lowest element force were 0. that is somewhat related to the manufacturing accuracy.6 and 0.99 N).6. is the straining of the elements due to thermal loading. σρ = 10−5 .07 N). the heat inputs are the direct solar ﬂux.05. RE the radius of the Earth (6.378.1.9 N (11.5 N (5. and Earthemitted infrared (IR) ﬂux. Earthreﬂected solar ﬂux (albedo). 0. level II is the tolerance goal for an accurate reﬂector antenna. Conservation of energy yields that the absorbed energy is equal to the emitted energy.11 Inﬂuence of Thermal Strains Another error source. Thus. (6. the maximum and minimum equilibrium temperatures of the inorbit tension truss have to be found. the intervals and mean values were 0. The heat output is the emitted radiation energy from the body. a the albedo of direct solar ﬂux (0.22) where qIR is the Earth IR emission (237 ± 21 W/m2 ). For level I.61–9.00 N (3. h the altitude of the body. and στ = 0. and 8. the spacecraft will be both in and out of the Earth’s shadow and.98 N (8.72 N (0. respectively. • II: σε = 10−6 . EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY It was shown above that a low tolerance may result in cable slackening even though the decrease in surface accuracy is small. Two tolerance levels were tested: • I: σε = 10−5 .2 N). and στ = 0.125 N).0728–10. level I does not achieve a suﬃcient level of accuracy as the combined eﬀect of the manufacturing tolerances resulted in cable slackening. The results are given in Table 6. σρ = 10−6 .9.
67657 1.89629 8.8833 9.00182 18.23157 2.03721 13.0254 0.98381 0.62063 0.3902 0.05896 0.67120 1.8707 1.z c σ∆F +∆ZP ∆F + ∆ZP (mm) ε σrandn ρ σrandn τ σrandn ¯ δrms.2459 2.60387 1.664 + 0.01155 I 0. For the net elements.9945 1.01713 1.7067 9.ring tension truss.30536 0.34279 3.5499 20. a = 0. The above equation give estimates of the equilibrium temperature for the worstcase conditions.9932 24.4719 0.6788 0.3644 3.66679 0.15472 0.1013 II 0.0089 13.13946 0.03455 0.23777 9.0084 1.83677 0.2278 0.31202 13.0801 0.4763 6.04804 0.00538 1.67051 · 10−8 W/m2 K4 ).6566 6.23) Note that the Tmin is independent of the surface properties.2391 0.01571 I (mm) (mm) (mm) 0.0877 rms.3963 3. F/D Tolerance level c δrms.13489 0.3424 0.41913 0.9908 1.2846 0.6 II 0.25132 0.z σδrms.6339 0.68904 6.4875 16.4143 0.1640 13. For this condition there is no direct solar ﬂux.04887 22. the angle between the solar rays and the surface normal of the bands varies over the reﬂector surface resulting in diﬀerent equilibrium temperatures.86517 8.4097 0.29273 0. The worstcase cold temperature occurs when the body is in the shadow of the Earth and out of view of any portion of the sunlit parts of the Earth. or albedo.2019 14.8 II 0.9545 0.22500 0.521ρ − 0.14152 9.27761 13. and σ StefanBoltzmann’s constant (5. GS = 0.2813 2.99641 1.34066 0.6164 8.0042 1.42772 6. 148 .8542 0.22523 1.9334 6.15815 22.08536 0.47404 17.0010 2.45069 25.7215 1. 100 Monte Carlo simulations).00083 9.0123 8. ANALYSIS OF TENSION TRUSSES Table 6.22858 2.03420 0.00164 3.53619 19.00835 0.3538 12.34456 3.0263 0.14884 8.4414 13. which further degrades the surface.7021 0.00447 0.z ¯ δc 0.5271 21.02969 13.3707 1.6092 0.7743 1.z min δrms.4895 7.203ρ2 ). The worstcase cold temperature is [119] Tmin qIR sin2 ρ = 2σ 1/4 .99887 2. A correction of the equilibrium temperature for each individual band due to the angle of the solar rays was not made here as this analysis serves only as an estimate.9: Combined inﬂuence of manufacturing imperfections on the rms surface deviation and defocus (three.9951 1.CHAPTER 6.0529 11.00110 13.6758 22. (6.0039 0.98031 1.4490 0.0786 8.z max δrms.0028 (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) 3.76164 6.z ∆F ∆F max ∆F min σ∆F ∆ZP max ∆ZP min ∆ZP σ∆ZP solar energy oﬀ the spherical Earth (Ka = 0.73870 14.0190 17.4 I 0.
i = 0. If the equilibrium temperature variation due to the incident solar rays is neglected. In an early study. using qIR = 258 W/m2 . σαT . and • Worst case hot temperature ϑmax = +110 ◦ C. Hedgepeth [56] uses αT = 0. Following this brief study of the achievable CTE of CFRP for space structures. which gave sin ρ = 0. The distortion test of a 6 m diameter AstroMesh uses worstcase temperatures of −160 ◦ C and +120 ◦ C [166].20µ to 0. In 149 . In the present analysis.ϑ. • Worst case cold temperature ϑmin = −80 ◦ C. ¯ Before computing the random errors of the thermal strains. using the same surface properties as above.25 m diameter AstroMesh antenna on Thuraya.6. According to Panetti [119]. CFRP with αT less than 1. The thermal strain at a temperature ϑ is εϑ = αT (ϑ − ϑref ) .25) where αT is the CTE.i (1 + εϑ ) (6. [62]. [119].24) where 0 is the unstrained length at the reference temperature ϑref .35. like the 12. depending on the layup of the carbon ﬁbre sheets.786 km. the worst case temperatures would be ϑmax = +69 ◦ C and ϑmin = −192 ◦ C.9206. the inﬂuence of the thermal strains on the antenna accuracy is mainly characterised by a change in average temperature [58]. less than 0.16µ/◦ C. the CTE of the tubes varies from −0. thinwalled composite tubes with near zero CTE are developed [157]. even under a uniform change in temperature over the tension truss. was Tmin = 200 K or ϑmin = −73 ◦ C. In reality. it is necessary to examine how a uniform temperature change aﬀects the rms surface error and defocus. the following values for the mean value and standard deviation of the CTE were used in the present analysis: αT = 0. and the εϑ the thermal strain.4.5µ/◦ C and σαT = 0. a typical operating temperature range of a parabolic reﬂector is −160 to +95 ◦ C. GS = 1418 W/m2 . The unstrained length of an element i after a change in temperature is 0. manufacturing imperfections and variations in material properties result in nonzero CTE. but without costly testing it is diﬃcult to make the variation of the CTE. using qIR = 216 W/m2 .88. and a = 0.4µ/◦ C. (6. The worstcase cold temperature. The altitude of the spacecraft to which the antenna is attached was h = 550 km.1513). The worstcase hot temperature of the net bands. sin ρ = 0. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY The net elements are made of CFRP and it was assumed that their surface properties were that of carbon black paint.96 and εIR = 0.0µ/◦ C is readily available.4µ/◦ C. One of the advantages of CFRP is that it is possible to theoretically achieve zero CTE by an appropriate choice of material and layup parameters. αS = 0. This variation results in a random thermal strain. the following temperatures were used: • Assembling temperature ϑref = +20 ◦ C.5µ/◦ C for the CFRP. If the antenna was mounted on a satellite in a GEO (H = 35. was computed to Tmax = 374 K or ϑmax = +101 ◦ C. More recently.
11. but the average minimal forces are acceptable.89 N). δrms. but diﬀered by approximately one order in magnitude for 10−6 .83 mm.04–6.6831 16. There is a slight possibility that some cables go slack. and 8.z ∆F ∆ZP tmin all tmax all 23.1321 10.46 ϑmax 0.11) showed that δrms. 1 150 .12 Achievable Reﬂector Accuracy A comparison of the rms surface deviation results for member length errors with ¯ε (6. the intervals and mean values were 2.8721 1.3 N (11.6585 19. δrms.14 46.z 0.35 ϑmin — 0.58 67. The net forces were more or less unaﬀected by a uniform temperature change. ANALYSIS OF TENSION TRUSSES Table 6.44 and 0.0 N (7. For the worstcase hot temperature. στ = 0.3338 9.8387 3. This would not be the case with a kinematically indeterminate τ network. this anomaly was not considered to be a major problem as the surface errors for σε = 10−6 were lower than the resolution of common measuring systems1 .1294 7.10.11) for −4 −5 10 and 10 .8 were 0.2176 14. 4.6. respectively.8072 2.1798 8.2375 13.2 N (11.41 67.3755 9.86 N).10 89. and 0.04 mm. the CTE was assumed to be deterministic and equal to αT .8796 6.1 N) in the same order as above.91–10. where the highest magnitude was The video photogrammetry system used for the AstroMesh had a resolution of 0.z always diﬀered with about one order of magnitude. [166].z varied between 0. δrms. 6.z varied between 0.z was about the same as the corresponding value of (6.05.10: Inﬂuence on the rms surface deviation and defocus of a uniform deterministic thermal strain in the threering tension truss.CHAPTER 6.30–9.6274 10. 4.68 N (4.3448 13.470–6. However.07 46. Note that ϑmin produced a smaller systematic rms surface deviation due to an increased focal length.76 N (4.1 N). and 8.5076 2.0088 13.92 26.83 this study. The results from the thermal analysis with a random CTE are given in Table 6. the rms surface deviation was relatively insensitive to a variation of the tension tie forces.28–13.1.2085 21. These errors were about the same order as the member length errors.32–13.2220 22.15 mm.2029 23.23 mm.44 N).3373 13.11 and 0.2415 1. This was a bit puzzling since σδrms. The eﬀects of ring distortions were more severe than the previous two manufacturing errors.4 ϑmin (mm) (mm) (mm) (mm) (mm) (N) (N) — 22.9042 6.4939 7.6 ϑmax 0. 0.11 and 0.75 89. For σρ = ρ 10−5 .4. it varied between 0. F/D Temperature ϑ δrms.4.44 N). For the worstcase cold temperature the interval and mean values (in parenthesis) of the lowest element force for F/D = 0.1382 3. ε For σε = 10−5 and 10−6 .57 ϑmin — 0.95 N (7.z was almost constant at 0.8 ϑmax 0. while for the higher one.17 and 0.05 ∆F + ∆ZP δrms.07 mm rms. As anticipated.9925 14. στ = 0. The results ¯ of this analysis are shown in Table 6. For the lower tolerance.3491 9.
0022 9.10 mm.z varied between 0.4572 8.6807 0.8.8201 8. were about the same for the three focal lengths: 0.1209 11. Considering the slack cables at level I.2 · 10−4 D.16 mm at level II.0493 6.0732 0.6.4306 1.4529 6. However. 100 Monte Carlo simulations).3529 9.2221 0.6880 1.2402 0.8232 0.7171 8.0028 13. Taking all the random error sources above into account.0154 23.0009 (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) 3. The eﬀects ϑ of the thermal loading.1456 6.z ∆F ∆F max ∆F min σ∆F ∆ZP max ∆ZP min ∆ZP σ∆ZP a Note that the eﬀect of CTE variation is computed by (6.1769 7. These values were computed with respect to the systematic error of a network with deterministic thermal strains.6755 1.2769 9.z .4.1079 0.9994 1.0143 0.6981 1.9481 19.2851 0.2207 2.1311 1. it may be possible to use the lower tolerance of the tension tie forces. ρ obtained for F/D = 0.8639 1.1851 22.1987 14.5803 13. where the higher value corresponded to F/D = 0. it may be necessary to provide some kind of adjustability.1898 0.9989 0.0815 13.1582 13.6.24 mm.4806 1.0004 2.0526 0. or 1. the upper bound of the rms surface deviation was about 0. This value can be compared 151 .4224 0.1808 22.0009 3.18) as before. but now with the systematic error from Table 6.8385 0.9990 1.3502 15. F/D Temperature ϑ δrms.2480 2.0027 2.z max δrms.1173 0.8507 1.0027 14.5167 0.8259 1.0474 0. As the rms surface error was very sensitive to the ring distortion.5603 0.z ϑ σ∆F +∆ZP ∆F + ∆ZP ϑ σrandn ¯ δrms.0038 20.8 ϑmax 0.6 ϑmax 0. the ring structure must be very accurately constructed.2182 0.3401 3. στ = 0.4995 16.2340 0.3427 0. it is obvious that level II must be the manufacturing tolerance goal.4595 8.4 ϑmin (mm) (mm) (mm) (mm) 0.1 and still have a high accuracy.1280 13.19– 0.2298 27.0176 0.z ¯ δϑ a 0.0032 9. For σρ = 10−6 .2259 0.5575 23.0949 23.6644 1.2326 0.4946 21.0895 0.1400 1.3350 3.2149 0.0405 ϑmin 0.1285 1. The combined eﬀects of the three manufacturing errors was at most 0.2282 0.3290 0.64 mm at level I and 0.0300 ϑmin 0.3371 6.z min δrms.11: Inﬂuence on the rms surface deviation and defocus of a random thermal strain (threering tension truss. δrms.35 mm.1939 0.0037 16.31 mm.7129 13.1536 13.4978 13. EFFECTS OF MANUFACTURING ERRORS ON THE REFLECTOR ACCURACY Table 6. to ﬁne tune its shape during ground testing.2294 2.0585 rms. Taking the rss value of the systematic and random thermal rms errors.6272 8.1303 9.3550 3.3499 3. the resulting maximum error was approximately 0.5148 9.z σδrms.10.0023 7.2154 13.0310 ϑmax 0.0021 9.6556 0.2401 2. δrms.08 and 0.0506 0.
Since a defocus of up to one wavelength (31 mm at 9.4 and 25.05 and 0. it cannot be considered a problem for the present antenna.45 mm. at the same level.7 · 10−5 D. varied between 0. or 7. ANALYSIS OF TENSION TRUSSES with Table 5.46 mm. the defocus decreased signiﬁcantly as the number of net rings increased. where the rss surface deviation of a 6 m diameter AstroMesh.2 mm depending on the focal length. At tolerance level II. Its variation.65 GHz) can be accepted without seriously aﬀecting the performance of the antenna. is about 0. 152 .2. the defocus. varied between 22. due to random error sources.CHAPTER 6. σ∆F +∆ZP . In fact. ∆F + ∆ZP . [131].
Chapter 7 Tensegrity Reﬂector Antennas 7. the bands should follow the geodesic lines of the surface. with equally spaced nodes along the circumference.15. would not necessarily coincide. is used instead of a true geodesic net. this chapter will start with a closer study of the AstroMesh.1(a). To minimise the number of intersections.1 Net Generation The triangular nets of the AstroMesh have a conﬁguration similar to that of the tension truss with three sets of bands oriented approximately 60◦ apart.169]. The present chapter will provide a more complete study of the concept starting from the simple initial studies with the stiﬀened tensegrity module to the vibration characteristics of a full scale oﬀset reﬂector antenna. 7. It was apparent from Figure 5. Figure 2. connected to the boundary. To better understand the similarities and diﬀerences between the two concepts.2 The AstroMesh Concept Most details of the AstroMesh are found in the US patent [167] and in a technical report from Astro Aerospace [62]. with coinciding threeband intersections. the intersection of a set of three geodesic bands. However. as a lateral loading on the surface would not tend to shift the bands in that position. in this case with six rings. is a major source of inspiration in the development of the new concept. The generation of the geotensoid starts with a hexagonal array of equilateral triangles. 7. This quasigeodesic net is called geotensoid.5 that the AstroMesh concept.2. For a circular 153 . The main parts of the AstroMesh are the triangular nets and the deployable ring structure. Figure 7.1 Introduction A preliminary study of the new antenna concept is presented in references [168. a quasigeodesic net. Ideally.
2.2) 154 . is composed of a series of upper and lower bars which.3 Static and Kinematic Properties Initially. s = 0. A bar and joint count for the ring truss yields bring = 4B and jring = 2B. The total numbers of bars and joints of such an assembly. aperture.2). TENSEGRITY REFLECTOR ANTENNAS (a) (b) (c) Figure 7. with 6n bays. are separated by vertical and diagonal bars.2 Deployable Ring Structure The deployable ring structure of the AstroMesh. the nodes of the sixring array are then mapped to lie on equally spaced concentric circles.1) Only for B = 3. the nodes on the circumference connect directly to the ring truss. the free nodes are iteratively adjusted so that the total length of the net bands is minimised. 7. shown in Figure 7. Usually the ring is divided into several bays. mring = 2B − 6. hence. the number of bays is therefore equal to the number of net rings. the complete assembly should be statically determinate for easy prestressing. are b6n = 6n(3n − 1) + 24n bnets bring (7.1: Geotensoid generation: (a) sixring tension truss.1(c). when connected. 7. (7. As it is obvious that this ring truss cannot be prestressed. is the structure kinematically determinate. Finally.1(c). The main purpose of the two triangular networks is to eliminate the internal mechanisms of the ring truss. Figure 7. Figure 7. the number of internal mechanisms is quite large. The upper and lower rings. Figure 7. the number of internal mechanisms is. each composed of B segments. form upper and lower rings. (1.1(b). and (c) converged sixring geotensoid (redrawn from [62]). Ultimately.CHAPTER 7.2. when the ring becomes a triangular prism. (b) initial sixring geotensoid.2(a).
2). Substituting (7. 11. and 26 is similar to that of the tension truss. However.3) respectively. It is observed that for an odd number of rings the nets have the same type of ring truss connections.3 shows eight conﬁgurations of ﬁve. for large values of n.3) into the extended Maxwell’s rule. 16. Referring to Figure 7. (1. The remaining connections are similar but diﬀerent from the previous ones.7. It also emphasises the importance of static determinacy.25 m diameter AstroMesh reﬂector. but not in the open literature. A 2. Each net has nine rings and is connected to the ring structure at 30 nodes.2) and (7. Generally. the number of ring segments increases with the number of net rings. this approach will work only for small values of n as the ring truss composes the major part of the reﬂector mass. To better see the conﬁguration of the top net and compare it with the tension truss.2(a) shows the deployable ring truss and complete top net of the 12. The net conﬁguration at these remaining connections is chosen to render the complete reﬂector statically determinate. 21. Schemes for connecting nets with many rings to trusses with less than 6n bays must be developed. THE ASTROMESH CONCEPT 23 24 25 26 27 28 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 29 5 30 4 1 2 3 (a) (b) Figure 7.2: Conﬁguration of the AstroMesh onboard the Thuraya satellite: (a) line drawing created from Figure 2. This ring truss has 30 bays.1).to ninering nets which satisﬁes Maxwell’s rule. Figure 7. which by virtual work implies 155 .2. the antenna is too heavy. the resulting assembly is statically determinate. (1. the ring truss is removed and the net rings highlighted in Figure 7. This conﬁguration may change if the number of rings in the nets or the number of bays in the deployable ring truss changes. the conﬁguration of the net at connections 1. This small exercise clearly illustrates the great ﬂexibility of the AstroMesh concept and the simple theory behind it. 6.2(b). For an even number of net rings. These schemes are presumably available within TRW Astro Aerospace.5 m diameter AstroMesh reﬂector (n = 4) is built according to this approach and set up for various tests [165]. As neither the ring truss nor the synclastic triangular nets can sustain a state of selfstress. (7. Figure 7.2(b).15(b) and (b) highlighted rings of the top net. several alternative connections are possible. and j6n = 6n(n + 1) + 2. yields m − s = 0.
was designed along the lines of the AstroMesh design. in the design of precision structures. but it is not necessarily a solid factor to base the design on. 1 156 .CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS 30 bays 30 bays 24 bays 24 bays b = 1872 j = 626 (a) n = 9 b = 1440 j = 482 (b) n = 8 b = 1404 j = 470 b = 1188 j = 398 (c) n = 7 24 bays 18 bays 12 bays 18 bays b = 864 j = 290 b = 828 j = 278 (d) n = 6 b = 792 j = 266 b = 648 j = 218 (e) n = 5 Figure 7. When reference [71] was published the present author was working at DSL with tensegrity formﬁnding methods and tensegrity masts. [71]. 7.3 New Concept The idea behind the new concept originates from a study by Knight et al.1 It will be seen that the new concept. kinematic determinacy. no speciﬁc details about the design are given. where an antenna design based on tensegrity is introduced.3: Statically determinate conﬁgurations of the AstroMesh. to which the rest of this chapter is devoted. 488–489 in [129]) Conclusions from that discussion are that static determinacy is desirable as these structures are easier to model than statically indeterminate ones. The question “is static determinacy the key to the design of precision structures?” was discussed at the IUTAMIASS Symposium on Deployable Structures (pp. However.
gives m − s = 6. The modiﬁed assembly has j = 14 joints and b = 37 bars. the extended Maxwell’s rule. but this provides only a relatively small stiﬀness.4.e. central joints. i. kinematically determinate structures. The extended Maxwell’s rule now yields m − s = −1. highprecision applications require dimensionally accurate structures.4 also exists for this structure and there is no other independent state of selfstress.2). As mentioned earlier.3. An improved version of the module is shown in Figure 7.4: Hexagonal tensegrity module: (a) threedimensional view and (b) top view.5.3.4) Like all tensegrity modules.1 Stiﬀened Hexagonal Module Consider the regular hexagonal tensegrity module in Figure 7. Firstorder stiﬀness can be achieved by prestressing the structure. (a) (b) Figure 7. a parabolic antenna was one of the structures in which they were interested.5) Since the same state of selfstress of the structure in Figure 7.7. 157 . The initial idea by Pellegrino [126] was that an antenna based on tensegrity could be developed by using a hexagonal tensegrity module and add nodes and bars inside the two hexagons to remove the internal mechanisms. 7. s = 1 with only the longer members in compression. NEW CONCEPT About the same time DERA contacted DSL for help with developing deployable structures for the STRV. (7. Note that these internal joints are not coplanar with the hexagons. (1. This structure has j = 12 joints and b = 24 bars. s = 1 as before. This structure was obtained by connecting the nodes of the top and bottom hexagons to two interconnected. With c = 6. (7. thus forming two triangulated surfaces that coarsely approximate to a curved surface. The resulting structure would have a potential to meet the stringent requirements by DERA—the work with the tensegrity antenna was initiated.
where the joints of the top hexagon lie directly above the ones of the bottom hexagon. were analysed. called net cables. j =2 1+6 n(1 + n) +6 . the resulting structure has only six compression members and. i. The total numbers of joints and bars for that structure are. the angle when this happens depends on the length of the interconnected element. the simple triangulated surfaces are replaced by tension trusses. To reﬁne the surface. Four conﬁgurations.e.3.CHAPTER 7. decreases and they eventually end up in compression. Hence. It turned out that all of the additional members ended up in tension and can be substituted by cables. section 6. However. The compression force in the struts increases monotonically. The tension force in the elements forming the triangulated surfaces. The triangulated surface in this example is far too coarse to support a reﬂective mesh that approximates a paraboloid. Figure 7. therefore. if the upper hexagon was rotated clockwise with respect to the lower one by an angle θ. 2 (7. plus two tension trusses. 7. The tension force in the base cables. Thus.1. cf. this structure has no internal mechanisms and is. the state of selfstress changed such that all members became stressed. however. i.6) 158 .5: Stiﬀened hexagonal tensegrity module: (a) threedimensional view and (b) top view. Closer analysis showed that in the initial conﬁguration.2 Hexagonal Tensegrity Module and Tension Trusses Consider a pinjointed bar structure consisting of the original ring structure. if cables are used for the remaining members. meaning that they must be rigid. i. increases from its initial zero value.6 shows how the force density in the members changes with the angle of rotation θ. TENSEGRITY REFLECTOR ANTENNAS θ (a) (b) Figure 7. the compression members became longer. potentially useful for the present application. the hexagonal module.e. they become more compressed. which diﬀered in the positions of the internal interconnected joints.e. ∆H. the structure can easily be folded by collapsing the struts. all of the added bars were unstressed.
It must be emphasised that the tension ties should not be counted as bars.10 0.01 0.50 1 Struts 1.25 qi qlateral 0 Net cables 0.7) respectively. Substituting (7. the AstroMesh is a direct proof that this is not the case. The tension ties are springs providing constant forces and can. a way of prestressing the complete structure must be found.0 1 0.5 0.5 2 0 5 10 15 20 25 30 θ() Figure 7. therefore.2 1 0 0.5 0.8) Again. Adding an additional bar. NEW CONCEPT 2 1. 159 . The immediate diﬀerence between this structure and the previous one is that the interconnected element is missing. is to connect the corresponding nodes of the two nets with tension ties. based on the AstroMesh. An obvious approach.7. which connects the middle nodes of the nets. It is a common misconception that the inclusion of the tension ties would result in a structure that is highly statically indeterminate and.3.6: Force density values for the onestage tensegrity reﬂector for various ∆H/H. However. ◦ and b=6·4+2 6 n(1 + 3n) + 6(n + 1) . therefore. ∆H is the length of the interconnected element and H is the total height of the module (qlateral = 1).1 0. (7.6) and (7. may remove this single mechanism. 2 (7.50 =0 /H ∆H 5 0. This means that there exists one internal mechanism. at θ = 0◦ no additional states of selfstress has been created. suﬀers from all sorts of prestressing diﬃculties. s = 1.5 Base cables .7) into the extended Maxwell’s rule gives m − s = 0. be treated as external loads in an analysis.
160 . Figure 7. ∆H = 8F Thus. 2F 2F (7. A strut connects points 1 and 2 on the rims of the paraboloids.9) and (7. θ. e. and ∆H = D2 /8F at θ = 60◦ .11) n(pi ) = − 2F 2F which yields the tangent plane at p1 as z − z1 = x1 y1 (x − x1 ) + (y − y1 ) . TENSEGRITY REFLECTOR ANTENNAS To obtain detailed information about the static and kinematic properties of the antenna structure.10) D D cos 240◦ sin 240◦ z2 2 2 If the paraboloids are too closely spaced the strut will intersect them.7.4.9) . Consider two paraboloids with the same F/D ratio separated by a distance ∆H at the apexes. and the position of the struts. A system xyz has its origin at the apex of the upper paraboloid.3 Minimum Separation between Front and Rear Nets The minimum distance between the closest nodes of the nets in the AstroMesh is determined by the shortest acceptable length of the tension tie connecting these nodes. the total height of the antenna will remain constant. F/D = 0. one more parameter must be determined—the separation between the nets.54 m. But before an adequate conﬁguration of the antenna can be speciﬁed. (7. The coordinates of these points are p1 = p2 = D D D2 cos θ sin θ 2 2 16F T . 7. T (7. the minimum separation depends on both the curvature of the nets. D = 3 m.CHAPTER 7. the last ﬁgure is easily veriﬁed. This means that the deployed antenna will be rather deep.3. 8F (7. Subtracting the height of the bottom paraboloid yields the minimum separation D2 (7. The separation ∆H is zero at θ = −30◦ .g. For the present concept. Hence. The normal to the upper paraboloid at pi is T yi xi − 1 .13) cos (60◦ − θ) .10) into (7. its equilibrium matrix has to be analysed for diﬀerent values of θ.12) yields the vertical position z2 of the rim of the bottom paraboloid. the minimum total height of the antenna structure is H= D2 [1 + cos (60◦ − θ)] . the minimum separation is when the strut lies in the tangential plane to the upper paraboloid at p1 (or to the lower one at p2 ).12) Substituting (7. ∆H = D2 /16F at θ = 0◦ . and θ = 10◦ yield H = 1. (7. the coordinate z2 is determined by the deepest paraboloid. F/D.14) Note that even if the bottom paraboloid is made shallower than the top one.
m = 1 and s = 1. it was found that only six of the 252 elements were in compression. An antenna with threering nets was analysed. i. when the complete reﬂector structure was considered.3. the structure had one internal mechanism and large compressive forces were induced in the ring structure. m−s = 0.4. However. θ. The variation of the internal forces in the threering antenna with the angle θ was analysed. the static and kinematic properties of the antenna assembly changed when going from the initial to a rotated conﬁguration.7: At the conﬁguration with the theoretical minimum separation between the front and rear nets. contrary to the stiﬀened module. By itself.4 ThreeRing AxiSymmetric Reﬂector It is now possible to describe the conﬁguration of an axisymmetric antenna with ﬁxed D and F/D by only one parameter. hence. Like the stiﬀened tensegrity module. including the prestressing forces applied by the tension ties. but for θ > 0◦ . 7.3. The diameter D = 3 m and F/D = 0. the state of selfstress for θ = 0◦ was a prestressed ring structure but unstressed nets. Nevertheless. NEW CONCEPT y z n1 Tangent plane at 1 1 y x n1 1 x 2 2 (a) (b) Figure 7.7. Throughout the studied range. 0–30◦ . For θ = 0◦ . the whole structure can be prestressed with θ > 0◦ . m = 0 and s = 0. At θ = 0◦ .e. all of the net forces were in tension 161 . the strut lies in the tangent planes of the paraboloids: (a) threedimensional view and (b) top view. the structure is statically determinate. Corresponding nodes of the two nets were connected by only tension ties. The net conﬁguration and the tension tie forces were according to Chapter 6 for n = 3. the rotated ring structure could no longer be prestressed as s = 0.
CHAPTER 7.8: Variation of the forces in the net cables for the threering conﬁguration. 162 . TENSEGRITY REFLECTOR ANTENNAS 6 16 5 4 Force (N) 19 18 3 8 10 17 2 1 12 9 0 0 5 10 15 20 25 30 θ (a) 10 (◦ ) 9 16 19 17 9 6 14 10 18 11 7 5 2 15 12 13 15 8 13 8 Force (N) 7 3 4 6 1 5 14 4 0 5 10 θ (◦ ) (b) 15 20 25 30 Figure 7.
This problem was not considered in the present analysis. 163 Height (m) Base cable .70 m at 24◦ and the goal must be to keep it as low as possible. Considering all of these issues. at which the base cables were no longer in tension. with a magnitude between about 1 and 10 N for tension tie loads of 1 N on the internal joints and 2 N on the edge and additional joints.8.14). θ should be kept small.41 m at 0◦ to 1.7. is the possible interference between the struts and the tension ties. Therefore. the upper limit for θ.9: Variation of the forces in the ring structure and its height for the threering conﬁguration. at least during the deployment procedure if not at the deployed state. Another issue. but decreased exponentially to acceptable levels at about 5◦ . however.9. were extremely large for low values of θ. As the angle θ increased. When looking for an adequate conﬁguration one should not only take the magnitudes of the internal forces into account.8 1. NEW CONCEPT 200 150 100 50 Ring height Lateral cable 1. Also shown in Figure 7.9 is the height of the ring structure from (7. but when a physical structure is built it is of primary importance that the struts are not interfering with the tension ties.5 1.3. was about 24◦ . Figure 7. This means not only that the struts become longer but that they are also more likely to interfere with the tension ties. It went from 1. Figure 7.6 1. The forces in the ring structure. θ = 10◦ seems like a good choice for the present example. the struts moved closer to the centre of the of reﬂector. or vice versa.75 1. For this particular conﬁguration. which might aﬀect the choice of θ.4 Force (N) 0 50 100 150 200 0 5 10 15 θ (◦ ) 20 25 30 Figure 7.45 1.55 Strut 1.65 1.7 1.
The nets were made of 0. the total length error was estimated at about 1 mm per net element.45 m and focal length of 0.10. To summarise. TENSEGRITY REFLECTOR ANTENNAS 0 cm 5 10 15 20 Figure 7. 3 Vivak is a registered trademark of Sheﬃeld Plastics Inc. The radial cables are used to connect the net to the ring structure. The triangular nets2 were constructed on a paraboloidal mold of Vivak 3 with a diameter of 0. [169]. 2 164 . the model had threering nets and a rotation angle of 10◦ .4 Demonstration Model To verify the feasibility of the proposed concept. Figure 7. Systematic length errors were introduced when the cords followed the arc lengths between nodes instead of the straight lines and at the crossover points where only one of the three cords lay on the surface of the mold. The latter error could have been minimised by using thin steel or CFRP tapes instead of the cords but since this was the ﬁrst model. a smallscale physical model was constructed. the cords were joined with Nylon loops at all crossover points and bonded with epoxy resin. at a rather small scale. cf. it was decided not to choose material not readily available. Based on the analysis in the previous section. 7. which is undesirably high. the additional 6 joints and 24 elements outside the edges were left out as the net conﬁguration of the demonstration model was based on an earlier study. but with the available material it could not be made smaller. This manufacturing technique was not very accurate. Corresponding nodes of the two nets were later connected by lengthadjusted ﬁshing line and steel springs.CHAPTER 7.10: Paraboloidal plastic mold with the ﬁrst of the two threering nets attached.8 mm diameter Kevlar cords which were straightened and taped to the molds. for glycol modiﬁed polyethylene terephthalate (PETG). The conﬁguration of each net was not exactly as given earlier. a thermoplastic copolyester. on which the position of the net joints had been marked with a threeaxis computercontrolled machine.134 m.
4.11: Demonstration model: (a) top view and (b) bottom view.7. 165 . DEMONSTRATION MODEL (a) (b) Figure 7.
the relatively large diameter and cylindrical shape of the joint ﬁttings prevented the nets from being attached close to ring structure. some of the cables in the two nets remained slack after deployment and there was some interference between the nets and struts. Knight [72] lists four possible solutions for strut deployment: • Hinged struts. This was mainly due to the length errors in the nets. The struts were made of 6.CHAPTER 7. in the vicinity of the ring connections. despite them being too large compared to the rest of the model. a more accurate mold and smaller joints in the ring structure. Figure 7. TENSEGRITY REFLECTOR ANTENNAS Based on the experience from the construction of the tensegrity masts. which made them deeper.5 Deployment Schemes The success of any deployable structure lies in the actual deployment.5.1.. Figure 7. 166 . Twelve identical aluminium alloy. Correcting these problems should be possible in a secondgeneration model. which ﬁtted in 20 mm long holes at the bottom of the joint ﬁttings. A 1. by using thin bands for the nets. 7. the nets are close to the joints. A simple and reliable hinge is the TSR hinge described in section 4. the length errors were simply too large. Since. A smaller. Its automatic locking capability makes it especially interesting for this application. In addition. it was decided to manufacture special joints with precisiondrilled holes for the ring structure.4 mm diameter aluminium rods.13. However. This interference could hardly have been avoided by further separating the nets. the unfolding rate must be easy to control. and • Inﬂatable struts. but also more costly to produce.0 mm Kevlar cord was used for the ring cables. even in an ideal structures. 30 mm long joint ﬁttings of cylindrical shape with a diameter of 15 mm were manufactured. The joints were well made and functioned satisfactorily. In each joint ﬁtting. Grub screws held the joints on the struts. spherical joint would have been ideal.g. The deployment of the present structure relies entirely on the way the struts are unfolded. These holes were coaxial with the joints. Hinges of various types have been used on deployable systems for several decades. Facing the identical problem. a large length error is bound to result in struttonet contact. • Telescopic struts. The model worked quite well. considering it was the ﬁrst time that a structure of this kind had been constructed in DSL. holes with 2. The cords were connected to the joint ﬁttings by epoxy resin.12.0 mm diameter in the direction of each ring cable and a radial net cable were drilled. it does not matter how accurate or stiﬀ the structure is in its deployed state if it fails to deploy. • Sliding coupling struts. e.
7.5. is an alternative to the hinge. sliding couplings are fairly new and also introduce stiﬀness nonlinearities [72]. Figure 7. with a locking mechanism. 167 . DEPLOYMENT SCHEMES Figure 7.12: Side view of demonstration model.13: Cylindrical joint ﬁtting with grub screw. However. A sliding coupling. With sliding couplings it takes minimal force to deploy the strut but signiﬁcant force to stow it again.
have developed an interesting 2. The last option is to use inﬂatable struts. For increased stiﬀness. A slow. the tubes slide on tightﬁtting Vespel 4 pads. This approach can minimise the stowed volume. More reﬁned struts were not made as it seemed that the chances for this folding approach to work were very low given the almost certain entanglement of the struts in the tension springs. the struts started to interfere with each other. Figure 7. the entanglement problems were too severe. [35] study the possibility of using highly elastic cables. I. then. ﬁrst. TENSEGRITY REFLECTOR ANTENNAS Telescoping structures have not been used very often in space applications due to excessive weight and drive force required [72]. each having two hinges along its length. hinged struts and.5. and thereby prevented a compact package. However. du Pont de Nemours and Company. 168 . but the size and weight is comparable to the three previous schemes [72]. Knight [72] and Duﬀy et al. is snag prevention. One design issue. To avoid snagging they must be stowed in a clever way. which had relatively few tension ties. A rigidised strut would have a uniform crosssection and a minimum of stiﬀness nonlinearities [72]. still it had to be tried.14. Following this brief review of deployment options. An inﬂatable strut would use manufacturing and inﬂation techniques similar to those of inﬂatable antennas. 4 Vespel is a registered trademark of E.4 m telescopic mast consisting of seven tubes. On top of that. for this structure. [7]. which eﬃciently prevent snagging. telescopic struts. the struts need to be rigidised to ensure their structural integrity throughout the mission lifetime. The hinges were made by small pieces of metal plate which easily could be bent to a speciﬁc angle. Even for the threering conﬁguration.15 shows the reﬂector under two early stages of folding. no deployable structure can be allowed to have a too quick deployment sequence as high shock and vibration may be introduced into the spacecraft [72]. hinged struts were not an option.CHAPTER 7. The entanglement started immediately and became more severe as the folding continued. Figure 7. These problems can be seen in Figure 7.1 Hinged Struts The aluminium rods were replaced by wooden rods. it was decided to investigate the folding of the small reﬂector model using. After inﬂation. controlled deployment is desired. 7. However. There is a potentially large risk that the long slack ring cables get caught or looped around a strut during deployment.15(b). This simple test showed that. which is critical for deployable structures with cables. and a minimum overlap of one tube diameter is maintained. In addition. Becchi and Dell’Amico. in such an approach the structure is subjected to very high stowage forces and stiﬀness creep.
169 .15: Folding of the antenna using hinged struts ended in failure. DEPLOYMENT SCHEMES Figure 7. was about 0.5.14: Wooden struts with two hinges at diﬀerent stages of folding.16. The stowed length of the strut. Of course. including the joint ﬁttings.2 Telescopic Struts Six telescopic struts. Figure 7. Special connections were made at the ends of the sticks to make them ﬁt in the cylindrical joint ﬁttings.5. each 0. 7. Each strut consisted of three tubes of diﬀerent lengths as the umbrella sticks had to be shortened.46 m long. were made by cutting oﬀ the sticks of six identical foldable umbrellas.7. (a) (b) Figure 7.28 m.
6 Preliminary Design of 3 m Reﬂectors In this section. 7. no entanglement between struts and springs or interference between struts occurred.CHAPTER 7.e. followed by a length adjustment of the ﬁshing line. the design procedure will be described.16: Threepiece telescopic strut at diﬀerent stages of folding. but problems could arise if the struts deploy in an unsynchronised manner. depends on the wavelength λ and the focal length F . The folding improvement was immediate. A conclusion from this simple test was that synchronously deployable telescopic struts.z = 0.17.01667 F . the number of rings in the tension truss. i.4 the required network density. Figure 7.1 Design Scheme Network Density As shown in section 5. would be the best alternative for a successful deployment. including various simpliﬁcations and engineering judgements.6. 7. Hedgepeth [56] sets the force t in the net elements equal to ten times the mesh tension p multiplied by the facet side length . the full scale antenna for a future STRV mission will be designed.15) 170 . Equation (5. (7.27) yields the surface error 2 δrms. a more eﬃcient packaging could be achieved using custom made struts. This procedure will then be applied to axisymmetric and oﬀset reﬂector conﬁgurations. the model was suspended by a ﬁshing line and then folded by hand. TENSEGRITY REFLECTOR ANTENNAS Figure 7. the dynamic characteristics of the best antenna conﬁguration will be computed. This simple simulation looked very promising. with some kind of motor synchronisation. First. To crudely simulate an actual folding. Finally. Due to mesh saddling the actual facet surface error is higher than the ideal one.
the smallest internal force in the tension truss should be t = 10p . PRELIMINARY DESIGN OF 3 M REFLECTORS (a) (b) (c) (d) Figure 7. so the allowable facet errors were 0. This value was doubled 171 . The nets were assumed to be constructed from CFRP band with a crosssection of 5 × 0.2 mm2 . two accuracy goals apply. To preserve this surface error.16) As decided in section 5. Note.6 mm.6.z .29).745 F δrms.4. (7.17: Simple folding simulation with the antenna model suspended in a ﬁshing line. The density of the bands was 1740 kg/m3 . The corresponding required triangle side length is = 7. which is only 3.3% larger than the ideal one. that the number of rings was maximised to seven according to Chapter 6. At 9. (5. however.6.65 GHz the wavelength is 31 mm.3 mm and 0. with facet error contributions of λ/100 and λ/50. respectively.7. so the lower error might not be achievable with a deep reﬂector.
with a minimum value of 1 N.18. The mesh area could be approximated with that of a spherical cap of equal height and equivalent radius as the reﬂector. 0. Ring Structure Once the required net tension and tension tie forces had been established. To facilitate comparison. depending on the strength of the CFRP.025 kg/m2 . each of the three bands running continuously through the node turns an angle /2F . Mesh The RF reﬂective mesh were tensioned to either 2 or 10 N/m.CHAPTER 7. it was decided to compute the actual area of the stretched mesh. for a more accurate mass comparison between diﬀerent antenna conﬁgurations. are based only on the lower mesh tension. Figure 7. which corresponds to a safety factor against material failure between 5 and 10. result in a heavier structure. This factor depends on the acceptable risk of failure for the mission and may vary for diﬀerent structural details. cf. Previous results on the present reﬂector. cf. Hence. This 172 . [169].17) ttie = 1. A higher mesh tension will more eﬃciently smooth out the creases formed during folding and give a more robust antenna structure as the prestress level increases. TENSEGRITY REFLECTOR ANTENNAS in the mass estimation to take the weight of the joints into account. although the elements along the edges might be too highly stressed and need strengthening.5 . was doubled to account for seams and surface treatment. To take this strengthening into consideration. It was further assumed that the same CFRP bands were to be used throughout the net. a safety factor with respect to material failure had to be set. however. Tension Tie Forces At each node. However.2.5. [62] use a maximum stress of 200 MPa. the total length of the net bands was increased by 10% in the mass calculation. The required force in the tension tie to equilibrate these net forces is t (7. both values were used in the design study. as discussed in section 5. but as long as the majority of the net elements have this force the surface degradation should be negligible. the forces in the ring structure could be computed. the tension tie force by (7. the surface density of the mesh. Hedgepeth et al. A higher mesh tension will. [143]. First.17) was rounded to the nearest integer towards inﬁnity. F Using this value might result in net forces less than 10p . When estimating the antenna mass. For CFRP compression members. For practical purposes ttie was given in full Newtons only.
where le is the eﬀective length of the strut and rg the radius of gyration. PRELIMINARY DESIGN OF 3 M REFLECTORS R R ≈ 2F ttie Figure 7.6.18) This slenderness was used only for conﬁrmation of the elastic buckling assumption. (le /rg )min = 39.18: Equilibrium of a node in the tension truss. allowable stress value was also used here for the net elements.5 GPa and σpl ≈ 1500 MPa. For the present application. le /rg . If the strut is not straight.19) For a thinwalled tube of radius a and wall thickness τ rg = I ≈ A πa3 τ a =√ . an initial sinusoidal imperfection. 2πaτ 2 173 (7.20) . the axial stiﬀness is severely degraded. ring cables and struts. (7. E = 227. σpl (7. The struts were basically designed to resist Euler buckling. The maximum slenderness depends on the required axial stiﬀness and length precision of the strut.7. subject to constraints on the slenderness. w = sin (πx/l). reduces the axial stiﬀness to [57] AEeﬀ = AE 1 1+ 2 rg 2. The minimum slenderness ratio was found by equating the critical buckling stress and the proportional limit stress σpl : le rg = min π2E . For a strut with both ends pinjointed. for the case of material failure.
+ sin θ + ls = 4 2 2 8F (7. use a maximum slenderness of 200 for 4. To ﬁt in the launch envelope. a minimum overlap of one tube diameter is maintained to achieve adequate stiﬀness. The tendons of the ring structure are preferably CFRP bands.21) 2 384 Erg where ρ is the density and g the gravity.5 GPa.5 mm. set to 200. struts up to about 3.22) produced ∆l /ls = 1. which is about one order better than the required ring distortion σρ = 10−6 .21) yielded le /rg = 257. of an inl compressible beam due to bending is 1 0 (dw/dx)2 dx where.24) aini = 200 With the wall thickness ﬁxed to 0.14. Setting /rg = 2 2/50 and substituting ls = 3.1 × 0. [57]. ρ = 1740 kg/m3 and g = 9. where 10 is the chosen safety factor against buckling and P the design compression force. the strut radius was adjusted until le /rg ≤ 200.25 m long CFRP struts with identical material properties as above.22) ls 40320 E rg The eﬀective length of the struts is equal to their actual length. (7. w(x) = 2 2 ρgx (l3 − 2lx2 + x3 ) /24Erg . The change in length. which easily can be folded. The midpoint deﬂection of a simply supported horizontal strut of length ls is 4 5 ρgls = .1 m in length can be constructed from four tubes.38 m.38 m. the 3. cables and pulleys was estimated as 0. In the following analysis. le = ls . The corresponding strain due to this change in length is 2 4 ρgls ls 17 ∆l = .2 × 0. 0. Hedgepeth et al. During ground testing. at this preliminary stage it is not possible to give a detailed mass calculation of the struts. the maximum slenderness was. therefore.2 kg per strut [168].38 m long struts had to be collapsed to less than a quarter of their length. Therefore. which changes with the rotation angle θ.4. an initial value for the minimal strut radius a was computed as: √ le 2 . A subsequent step checked that Pcr ≥ 10P . Thus. ∆l.CHAPTER 7. (7. latches. A thinwalled tube should have /2a < 1/10. [7]. (7. as √ θ varies between 0 and 30◦ . E = 227. The length of a deployed strut in an axisymmetric reﬂector can be written as: 1/2 √ 2 2 D2 2 2 D 3 1 + cos θ + [1 + cos (60◦ − θ)]2 . Inserting these values into (7. the strut length varies between 2.95 and 3. [62]. in this case. say /2a < 1/50. it is believed that much less crookedness. the mass of the tubular struts of constant crosssection was ampliﬁed by 50% to allow for tube overlap and the variation in the crosssection of the telescopic struts.2 · 10−7 . Thus. However. is needed for this application. However. In the telescopic mast by Becchi and Dell’Amico.23) With D = 3 m and F/D = 0. the strut sags due to gravity.8 m3 .80665 m/s2 into (7. The mass of the deployment motor. TENSEGRITY REFLECTOR ANTENNAS A 1% reduction of the axial stiﬀness corresponds to /rg = 0. Bands are also less likely to get tangled up during folding because basically 174 .
In Figure 7. The higher surface accuracy.6. 0.5%. following the scheme in the previous section. the required number of rings for a certain facet surface accuracy was given.31 m in the oﬀset case (Ha = 0.17).51 m). were discarded from further analysis. I and IV. 0. Slightly lower forces.2 AxiSymmetric Reﬂectors In Table 6. Mass estimates. were analysed in terms of internal forces for diﬀerent values of the rotation angle θ and mesh tension p.19 shows the results from the analysis of the internal forces when p = 10 N/m. D = 3 m. 175 . which increase the additional rms surface error from 3.13). the separation of the nets for the oﬀset reﬂector is computed by (7. For example.6 mm. The minimum net forces in Figure 7. were obtained at θ = 10◦ . respectively. was nearly obtained with seven rings for F/D = 0.3 Oﬀset Reﬂectors For oﬀset reﬂectors. The maximum allowable stress in the cables was. F should be substituted with Feq . Thus. was nearly achievable with seven rings for F/D = 0. This takes into account that for identical focal lengths and aperture diameters the oﬀset reﬂector is shallower than the axisymmetric one. and θ = 10◦ give a total ring height of 1. may in some cases be also accepted in elements connected to the mesh. II and III. A closer study revealed that the elements with the minimum force were located outside the highly stressed edge elements and therefore not attached to the mesh. The mass of the connections between the cables and the struts were estimated to 0. are given in Table 7. PRELIMINARY DESIGN OF 3 M REFLECTORS only two folding directions are permitted.8. These four combinations. (5. Figure 7.3 mm.3% to 16.2. if the aﬀected elements are in the minority.4. The lower surface accuracy. six rings for F/D = 0. but with a corrected focal length Feq = ξF . the forces in the ring structure are shown. 200 MPa. this eliminated the possibility of achieving the higher surface accuracy. Similarly. It was signiﬁcant that the minimum net force for the sevenring conﬁgurations. and ﬁve rings for F/D = 0.2 and 18.3. the steepness of the curves seems to depend on the number of net rings.10). At that angle the ring forces had come down from their very large values at 5◦ and the minimum net forces were acceptable at 6. was much lower than for the other conﬁgurations. say 2p .19(a). where ξ is the ellipticity.7. I and IV. F/D = 1. XA = 0 m. denoted I–IV.6.1.7 N. Good designs for the remaining two conﬁgurations. Like the threering reﬂector antenna.8.54 m in the axisymmetric case but only 1. they decreased rather rapidly in the beginning and are ended where a net force ceases to be in tension.6. 7. when computing the tension tie force from (7. the forces were just too low and the two sevenring conﬁgurations. which is about ten times the value at θ = 10◦ .6.19(b) have almost constant downward slopes. as above. Here. for conﬁgurations II and III at θ = 10◦ . 7.05 kg per connection. λ/50.40 m and ∆H = 0. To preserve the surface accuracy it ought to be about 10p = 21 N for seven rings.
4. (IV) F/D = 0. TENSEGRITY REFLECTOR ANTENNAS 3000 2000 1000 Force (N) IV I (a) II III 0 1000 2000 3000 4000 0 5 10 15 θ (◦ ) 20 25 30 25 20 (b) 15 Force (N) III 10 II 5 IV I 0 5 0 5 10 θ 15 (◦ ) 20 25 30 Figure 7. n = 5. (II) F/D = 0. n = 7.8. (III) F/D = 0. n = 7. n = 6. 176 .CHAPTER 7. and (b) least stressed net member for four diﬀerent axisymmetric conﬁgurations with p = 10 N/m: (I) F/D = 0.8.6.19: Forces in (a) strut and most stressed ring cable.
5 m. a feasible conﬁguration VIII did not exist.8 and 2. With focal lengths F = 1. This was most likely a result of the loss of symmetry for the whole structure. Note also that the minimum net force was lower than the 10p required for good surface accuracy. In Table 6.3 m.1. However.7. seems that θ = 10◦ is a good choice for the oﬀset conﬁgurations. In addition. a preliminary analysis with a threering oﬀset reﬂector showed that the tension tie forces determined in the previous chapter do not produce the same force distribution in the nets when the whole antenna structure was considered. However.2.6.1 m and XA = 0. However. F = 1.8. while for VI. neither mirror nor rotational symmetry existed. it. For F = 1.1 m were studied. The variations of the forces in the ring structure were similar for all three conﬁgurations.e.1 m and XA = 0 m. therefore. However. The bottom net was a mirror image of the top one in the aperture plane so that corresponding nodes were connected by tension ties. only reﬂectors with focal lengths of 1.1 m the higher surface accuracy of 0. i.0 for six of the ties. 177 .20 shows the modiﬁed tension tie distribution along the edges for the sixring nets. as stated before.16). Mass estimates for conﬁgurations V–VII at θ = 10◦ are given in Table 7.21.2. To achieve the lower surface accuracy of 0.1 m. the variations of the minimum net forces were drastically diﬀerent from those of the axisymmetric case. the majority of the net cables were stressed to the required value. necessary to reobtain a satisfactory internal force distribution. Slight changes in the tension tie forces along the edges were. For the present study. as their force relations were about the same for n = 6 and slightly worse for n = 7. The results for the three oﬀset conﬁgurations are shown in Figure 7. the internal forces in the two nets were identical as the structure still had the quasiﬂip symmetry.5 to 5. Considering the magnitude of the strut force and the strut–tie interference issues discussed earlier. the net forces for diﬀerent oﬀset conﬁgurations are shown.6 mm seven rings were required for F = 1. the eﬀect of turning the antenna upside down so that the top net becomes the bottom net will only change the direction of the struts and θ. For conﬁguration V. once again. so the overall surface accuracy should not be seriously aﬀected by the lower tensions.8 m and XA = 0. 1.8 and 2. as seen in Table 6. respectively. But. Hence. the number of rings was limited to seven. while six rings were suﬃcient for F = 1. As the forces in the ring structure were similar for all of the ﬁve remaining conﬁgurations. the minimum net force was maximised at about 10◦ .3 m.3 m. the mirror symmetry of the internal forces in the net was lost. it was maximised at 8◦ and 12◦ .3 mm could not be achieved with less than eight rings. the reﬂectors had an oﬀset value of 0 or 0. Considering the increased prestressing problems discussed above. it was not possible to prestress the nets with the present tension tie distribution. the only change was an increase from 4. F = 2. PRELIMINARY DESIGN OF 3 M REFLECTORS The number of net rings for the required surface accuracy was computed using (7.5. Figure 7. and VII. Thus. and 2. F = 2. the oﬀset antenna analysis were limited to six rings.8 m and XA = 0. Hence. the resulting masses were almost identical. Conﬁguration VI seems slightly better than the others because of a larger minimum net force and a marginally lower weight.
06 201.02 17.94 VI 2.27 0.60 1.00 6. The buckling force may in some cases produce a stress a little higher than 200 MPa.27 0.87 5. Buckling failure has a higher safety factor than material failure to account for imperfections and crosssectional variations.3 6 10 2 5 1 6.65 0.11 6. Description (m) (m) 10 6 6.54 0.20 0.07 170.25 0.59 31.80 0.60 1.98 9.80 6.25 0.1 0.27 0.79 0.70 Focal length Oﬀset distance (XA ) Number of net rings Mesh tension Internal tie forcea Minimum net force Maximum net force Ring cable force Strut force Eﬀective strut length Strut diameterb Euler buckling force Mesh area Net length Ring cables length Struts length Ring height Mass composition Mesh Net Connections Motors and latches Ring cablesc Struts Total (N/m) (N) (N) (N) (N) (N) (m) (mm) (kN) (m2 ) (m) (m) (m) (m) 0.09 5.77 Conﬁguration V 1.16 4.8 — 6 CHAPTER 7.4 1.60 1.1 446 90 1692 339 −3220 −644 3.7 6.47 5.77 0. and V–VII at θ = 10◦ .55 0.07 17.24 31.4 — 5 10 2 6 2 18.70 17.83 VII 2.54 0.81 a b Tension tie forces are rounded to nearest integer (N) towards inﬁnity.57 3.80 0.1 0 6 10 2 5 1 8.2 298 99 1323 441 −2520 −840 2.37 206.81 10.31 16. .1: Mass estimates of reﬂector conﬁgurations II.48 209.13 31.36 3.0 1.04 88 51 31.Table 7.36 0.83 77 53 24.92 0. but this is neglected to facilitate a fair mass comparison.00 91 53 36.20 0.11 6.8 0 6 10 2 5 1 5.71 0.43 3. TENSEGRITY REFLECTOR ANTENNAS 178 (kg) (kg) (kg) (kg) (kg) (kg) (kg) 0.28 0.97 5.70 9.11 6.90 5.94 17. c Crosssectional areas are rounded to nearest integer (mm2 ) towards inﬁnity.65 8.96 89 52 34.96 6.03 0.36 30.60 1.67 2 2 2.14 5.33 29.39 6.46 1.71 9.50 0.60 1.31 7.65 7.6 471 95 1846 370 −3545 −709 2.80 III 2.76 6.50 209.2 1.33 3.20 0.10 0.20 0.51 0.98 0.91 84 58 30.12 7.1 123 561 −1034 2.31 9. III.2 367 1681 −3100 II 1.15 5.88 6.20 0.2 505 101 1901 381 −3677 −736 3.
7. the stowed package must conform to the width and depth requirements of 200 mm and 100 mm.04 m.5 4 5 6 4. gave diameters 51–58 mm. the circular conﬁguration was not feasible. With a maximum depth of 100 mm. came under the limit at 0. Besides the height constraint. the stowed height of the struts is 0.5 6 Figure 7. A circular conﬁguration.5 mm. The present structure. The rectangular conﬁguration. the maximum strut diameter was √ 2·100/(2+ 3) ≈ 53. is preferable in the present case. 7.5 6 3. and VI stayed below the limit. which better ﬁt the requirements on stowed volume. If this strut is divided into four pieces and the minimum overlap is one diameter. Most mesh antennas have a supporting structure composed of only stiﬀ elements so the package with the stowed antenna tends to be cylindrical. Hence. Figure 7. not achievable in practice.5 4.8 m. according to Figure 5.83 m. so the recommended limit of 800 mm should not be exceeded and more tube segments only complicate the strut design. Figure 7. Returning to the rectangular conﬁguration it was immediately recognised that a mesh tension of 10 N/m yielded a package that was too large to ﬁt into the launch envelope.4 Stowage Considerations Conﬁguration V has the longest struts. 88 mm.22.785 m.22(a). is not bound to any particular shape of the stowed conﬁguration since the struts are not connected to each other. Two natural ways of arranging the struts are shown in Figure 7. However. The height limit can easily be met by dividing the strut into ﬁve pieces or by accepting longer stowed struts.6. Interior elements are drawn thinner for clarity. 179 .5 4 5 4. The lower mesh tension. At present. The maximum diameter of the struts was 100/3 ≈ 33 mm. Rectangular stowage dimensions for a mesh reﬂector antenna are rather odd.22(b). Considering both the stowage volume and net forces. as space in between the struts must be provided for the folded mesh and nets. however. which yields a slenderness of 257 for le = 3 m.20: Tension tie forces along the edges for sixring oﬀset reﬂector antennas (mirror symmetry). this space requirement cannot be quantiﬁed. Note that this is only a theoretical value.4 there is a hard limit of 891 mm on the stowed height. This is in excess of the limit 0. 2. respectively. provides suﬃcient space in the middle for the nets and mesh. III. Using the smaller tube diameter decreased the stowed height a little but only conﬁgurations II. PRELIMINARY DESIGN OF 3 M REFLECTORS 6 6 5 4 6 3. 2 N/m.848 m.6. Only conﬁguration III with the shortest struts. 3. VI again seems to be the best conﬁguration.5 3.
and (b) least stressed net member for three diﬀerent oﬀset conﬁgurations with n = 6 and p = 10 N/m: (V) F = 1.1 m. XA = 0 m. (VII) F = 2. 180 . TENSEGRITY REFLECTOR ANTENNAS 3000 2000 1000 VI V (a) VII Force (N) 0 1000 2000 3000 4000 0 5 10 θ 15 (◦ ) 20 25 30 10 9 8 7 VI (b) 6 Force (N) 5 4 3 2 1 0 1 0 5 10 15 θ (◦ ) V VII 20 25 30 Figure 7. XA = 0.21: Forces in (a) strut and most stressed ring cable.1. (VI) F = 2.8 m.3 m.CHAPTER 7. XA = 0 m.
the node was displaced 5 mm vertically by the 0. for a tie stiﬀness of.66 kg in the cupdown position. Steel springs with this stiﬀness are readily available.1 N (ttie = 1 N). As stated earlier. while the mass of the bottom net was 0. and the mass of the motors and latches were uniformly distributed over the total length of the struts. In the cupdown position. of 0. respectively. the required tie stiﬀness was 250 N/m. but some applications may require springs of another material with a smaller CTE. was analysed. the softer the springs of the tension ties.7. in force units.2 mm could be accepted in the cupdown position during ground testing. the weight of the lower net was about 0. In general. The mass of the strut–cable connections were uniformly distributed over the total length of the ring cables and the struts.1 was used. In the following vibration analysis. e. giving a length density of 0. The mass of the mesh was added to that of the top net. cupup and cupdown. the easier it is to control the force. In the Monte Carlo simulations of Chapter 6.0064 kg/m.g. only conﬁguration VI. which was found to be slightly better than the others. By itself. which is unacceptable. too soft springs might cause dimensional stability problems during ground testing. However.05 N.e. 7. During ground testing it is common to test the antenna in two positions.0038 kg/m.40 kg at a cupup position and 0. This gave a ring cable mass and a strut mass of 0. Here. However. VIBRATION ANALYSIS R R 3)R √ R 3R (2 + 7R (a) (b) Figure 7. The total antenna weight remained unchanged. the attachment of the antenna 181 . Assuming that a vertical displacement. sagging of the bottom net due to selfweight. say 10 N/m. to estimate the eﬀects of gravity.2919 kg/m.7 Vibration Analysis An important characteristic is the natural frequency of the deployed antenna. this additional loading will not cause severe degradation of the surface as its magnitude is of the same order as the force variation στ = 0.22: Stowing of the struts: (a) rectangular and (b) circular. Here. 6.005 kg or 0. the tension ties were assumed to be weightless.05 N load. all data except the stiﬀness of the tension ties are available.0158 and 0. the mass per internal node of the lower net was 0. defocus. For the vibration analysis.66/127 ≈ 0. i.7.67 kg. a maximum force variation of στ = 0.
and therefore the support conﬁguration in Figure 7. . became . The ten lowest natural frequencies were 9.23: Supports for the reﬂector antenna in the vibration analysis.and ydirection. 13.e. For suﬃcient stiﬀness. [173]. Figure 5.5 mm thickness. cf. i.98. which is well above the fundamental frequency of the complete structure.23 was used in the vibration analysis.74.and zdirection. All element were modelled using twonode bar elements with a consistent mass matrix. The lowest vibration frequency of a pinended strut is fs. 35.5 mm. 26. i.19. and 76.30 and 17.88. the minimum value of ∆d = 2t.58. 65. The modes for the two lowest frequencies are shown in Figure 7. it is highly likely that the antenna will be connected at point A. TENSEGRITY REFLECTOR ANTENNAS y z x Figure 7. cf.78 Hz. 19.32.27.24.33 Hz. 75. y. The buckling load was computed by the ﬁnite diﬀerence method. the gap between the tube segments should not be more than 3 mm (∆d = 4 mm). It was assumed that the tubes are free of any imperfections and have zero play at the intersections. in x and y.25) Assuming a constant tubular crosssection of 52 mm in diameter and 0. the safety factor of 10 for buckling 182 .12 Hz. The actual telescopic strut has lower bending stiﬀness. Symbols: restrained in x. the ﬁrst two frequencies increased to 13. Nevertheless.7.1 = π 2 EI . which somewhat decreases the natural frequency. to the spacecraft was not investigated. respectively. [27].37. Figure 7. cf.e. If the support with constraints in the x.CHAPTER 7. was constrained in the zdirection.96 m long strut has fs. Hence. 47. with a subdivision of l/100. The telescopic beam was divided into four segments and at each intersection the diameter decreases ∆d.1 = 37.77. • in x and ◦ unrestrained.25 shows the ratio of the buckling loads of a telescopic and a constant crosssection beam. The beam was tubular with a maximum diameter of 52 mm and a tube thickness of 0. the 2.. 57. ml4 (7.
7.8. DISCUSSION
(a) f1 = 9.19 Hz
(b) f2 = 13.32 Hz
Figure 7.24: The two lowest vibration modes of the oﬀset antenna conﬁguration VI with p = 2 N/m (the tension ties are removed for clarity).
has now decreased and will diminish further for the real strut due to manufacturing errors, e.g. play between tube segments.
7.8
Discussion
It can be concluded that the requirements on the stowed dimensions are approximately met; the package has to be slightly deeper, maybe 140 mm, to make room for the folded nets and mesh. This yields the following stowed dimensions: d/D = 4 · 0.2 · 0.14/π/3 ≈ 0.06 and h/D = 0.8/3 ≈ 0.27. Only the 15 m diameter HCA has so low stowed dimension values. It is observed that the mass limit of 20 kg, set up by DERA, is easily achieved by the present antenna at 6.67 kg or 1.24 kg/m2 (A = 5.37 m2 ). Compared to the 2.4 × 1.6 m2 SAR reﬂect array with a mass of around 15.5 kg or 4.04 kg/m2 ,
183
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
1 0.9 0.8
l/4 l/4 l/4 l/4 d
d∆d
d2∆d
d3∆d
0.7 d = 52 mm, t = 0.5 mm 0.6
Pcr,t Pcr,c
0.5 0.4 0.3 0.2 0.1 0
0
2
4
6
8
10
12
14
16
18
∆d (mm)
Figure 7.25: Buckling load of a fourpiece perfectly straight telescopic strut, Pcr,t , divided by the buckling load of a constant crosssection strut, Pcr,c .
cf. [130], the reﬂector antenna is superior. Comparing the areal densities of the mesh antennas in Table 2.1, which vary between 0.36 and 4.58 kg/m2 , the present value is amongst the best. The mass composition of the present antenna is 84.1% for the ring structure (including connections, motors and latches), 11.8% for the nets, and the remaining 4.1% for the mesh. In comparison, the 23.56 m diameter AstroMesh (ξ = 1.118, n = 27) of reference [62] has a total weight of 144.9 kg, of which 67.0% is the mass of the ring structure, 15.7% the mass of the nets and the remaining 17.3% belonging to the mesh. Although these two antennas diﬀer signiﬁcantly in aperture size, it is apparent that the ring structure is the dominating factor when considering the total mass. The relationship between the mass of the nets and that of the mesh depends entirely on the number of net rings. Earth observing systems typically require a fundamental natural frequency above 0.1 Hz [55]. With a lowest frequency of 9 Hz, the present antenna easily fulﬁls this requirement. A high stiﬀness is also an advantage in ground testing as the gravity compensation system can be simpliﬁed. In comparison, the SAR reﬂect array has a frequency of 0.9 Hz [130]. Although it seems that the present antenna is suitable for a future STRV mission, some issues remain that are either solved unsatisfactorily or not at all. Beside the attachment to the spacecraft, which was mentioned earlier, the position of the struts
184
7.8. DISCUSSION
will always be an obstacle in the concept. The current position, where the struts go right through the forest of tension ties, will seriously limit the number of net rings and thereby the achievable surface accuracy. If the struts in some way could be moved closer to the perimeter of the aperture, a great deal would be won, e.g. the ring structure height could diminish. A possible reﬁnement of the ring structure is to use a twostage tensegrity module instead of the onestage. The main advantages are that the struts are closer to the boundary and that two parameters, the overlap η and the rotation angle θ, can be used to change the prestress in the structure. Consider a twostage tensegrity module with six struts per stage which has m = 7 and s = 1 by (4.3). Then add two interconnected, central joints and 13 bars in the same fashion as in Figure 7.5. The resulting structure has m − s = −1. In the conﬁguration given by (4.5), the stress in the added bars is zero. By changing η and θ, the added bars can be stressed. Figure 7.26 shows the variation of the force density in the bars. It is observed that the area of feasible conﬁgurations is very small. Still, there is a possibility that such a conﬁguration will also work when the extra bars are replaced with triangular networks. Another aspect that needs further investigation is the layout of the nets for static
1
0.9
0.8
0.7
configura
Upper
0.6
Lowe
bound
0.5
0.5
b.c.
r bou nd
tions wit hs=3a nd m = 2 1
1
0.0
1 b.c.
1
0.5
0.1 0.2 0.25
n.c. n.c.
η
base cables
0
0.5
vertical cables net ca bles
0.4
0.5 v.c. 0.01 n.c.
1.5
. s.c
cab
0.3
sad
les
1
strut s
0.5
0.250.2 1.5 0.1 0 1
1
dle
1
1 0.01
1
0.2
0.1
struts
1
0
0
5
1
2
30
10
15
20
◦
25
θ()
Figure 7.26: Force density values for the stiﬀened twostage tensegrity module (R = H = 1, ∆H = H/100). The two thickest lines represent the upper (qbase = 0) and lower (qnet = 0) bound for the overlap (qdiagonal = 1).
185
CHAPTER 7. TENSEGRITY REFLECTOR ANTENNAS
determinacy of the complete structure and easy prestressing. In was shown in Chapter 6 that it became more diﬃcult to prestress the nets as the number of rings increases. Finally, there are several manufacturing issues that will need to be investigated before the antenna can be constructed. Those concerned with the achievable manufacturing accuracies are especially important.
186
the force density and the dynamic relaxation methods. i. but these methods normally do not provide any information on the static and kinematic properties of the framework.e.Chapter 8 Conclusions 8. this information is essential in the early stages of the structural design. but no single method is suitable for general problems. In the search for new conﬁgurations. It was found that the methods could be classiﬁed into two groups—kinematic and static methods. It is concluded that methods for analysing tensegrity structures are available. The force density method and the energy method are found to be equivalent. The reduced coordinate method or the force method are suitable for problems where some parts of the geometry are known. It may be argued that the FEM is computationally more eﬃcient. where the most popular methods. the force density method is well suited since the lengths of the elements of the structure are not speciﬁed at the start. The force method is very eﬃcient and highly suitable for the structural analysis of large frameworks.1 Analysis Methods One of the aims of the present work was to scrutinise the various methods for formﬁnding of tensegrity structures. Static methods search for equilibrium conﬁgurations that permit the existence of a state of prestress in the structure with certain required characteristics. while the length of the other type of element is kept constant. it is diﬃcult to control the variation in the lengths of the elements as the set of force densities is varied. However. 187 . Kinematic methods determine the conﬁguration of either maximal length of the struts or minimal length of the cables. can handle general structures. This is unsatisfactory and is contrary to the formﬁnding of cable nets and membrane structures. The present review of formﬁnding methods for tensegrity structures oﬀers a base for anyone intending to further continue the subject. In general. the static methods seem to possess more usable features than the kinematic ones.
The masts were relatively stiﬀ axially but very ﬂexible in bending. were combined to produce a robust structure which conforms well to the requirements of a future space mission involving a small satellite. Several unique methods and concepts. For longer masts.” This statement is certainly true in the case of the antenna proposed here. [53. This design step is thus no longer a major obstacle. One way to stiﬀen tensegrities. Due to its special characteristics and maturity. however. Hence. The most critical constraint was the size of the stowed structure. an important deﬁciency of the present approach is that the mast do not achieve full stiﬀness until the last stage has been deployed. some modiﬁcations had to be done to the tension truss to render the assem 188 . reduce the excellent deployment capabilities associated with discontinuous struts. cf.2 Deployable Masts Using recent advances in formﬁnding techniques. some other means of stiﬀening is needed. claim that “it is impossible to approach the ﬁeld of deployable structures with a single.179]. the equilibrium conﬁgurations of tensegrity masts with adequate internal force distributions are easily found. the bending stiﬀness remained almost entirely unchanged. general concept or theory. 8. This will.3 Deployable Antennas Miura and Pellegrino [105]. is to accept contacting struts. A high length accuracy is generally required. inventors of several deployable structures concepts. the ﬁrst cables becoming slack under a small imposed load.CHAPTER 8. which is suggested in these studies. to obtain the desired prestress and straightness of the mast. The structural behaviour of the masts was in agreement with the ﬁndings of previous studies. The proposed manufacturing scheme worked well when all the early mistakes had been eliminated. and this restriction more or less dictated the design of the ring structure. CONCLUSIONS 8. This structural ineﬃciency of tensegrity structures with noncontacting struts has also been observed in studies of doublelayer tensegrity grids. While the additional cables improved the initial axial stiﬀness by about 50%. In the physical model. the mast is very ﬂexible throughout the entire deployment process and this seriously limits the applicability of the masts. The design will thus be a tradeoﬀ between ease of deployment and compact packaging on one side and stiﬀness and strength on the other. The suggested strut deployment approach worked well. the tension truss was chosen to approximate the reﬂecting surface. However. in the construction of cable structures. Adding cables to remove the internal mechanism did not improve the stiﬀness substantially. However. each with its own particular features. The use of joints with holes manufactured to the correct threedimensional angles should produce a net with the required accuracy. stiﬀness during deployment was provided by a central rod which also controlled the rate of unfolding.
suitable areas of application must be identiﬁed and detailed requirements need to be formulated. the bending strength of tensegrity masts must be improved.4 Further Research To develop new tensegrity structures. The variations of the coeﬃcient of thermal expansion and tension tie forces must also be very low. This mobility or looseness is especially undesirable in the case of masts. Typically. Nevertheless. The proposed antenna concept is considered feasible for a future smallsatellite mission as it approximately meets the requirements on the stowed dimensions and easily meets the mass goal. some formﬁnding studies are certainly needed to ﬁnd conﬁgurations that meet the requirements. 8. In this respect. such as solar panels and therefore must deploy in a predicted manner. It was apparent that the prestressing became more diﬃcult with the present layout. as indicated by the lowest natural frequencies. [150] analyse planar tensegrity structures that are eﬃcient in bending. there are some issues that remain unsolved or solved unsatisfactorily. Threedimensional tensegrity structures with greater eﬃciency in bending 189 . This seriously limits the number of net rings and thereby the achievable surface accuracy. Stiﬀness during deployment is also required for the numerical simulation of the process. which often support other structures. Skelton et al. which is needed for the dimensional stability. High stiﬀness is beneﬁcial for ground testing as the gravity compensation system can be simpliﬁed. One aspect that needs further investigation is the layout of the nets. As indicated by the mast analysis. a maximum error of one part per million is required. FURTHER RESEARCH bly kinematically determinate. One aspect is the attachment to the satellite bus and this has to be designed along with the primary spacecraft structure. An accurate tension truss requires very tight length tolerances for the elements.4. The stiﬀness of the antenna. as the number of rings increased. but the primary objective must be to solve the technological problems that still remain. The load bearing capacity during deployment is generally nonexistent until the last cable has been prestressed.8. In their current position the struts go right through the tension tie forest. One speciﬁc issue that needs more research is the how a tensegrity structure can be stiﬀened during deployment. Concerning the morphology of tensegrities. is high compared to other antenna concepts. The main focus for future research must be on the implementation of tensegrities. Another is the position of the struts. There are also several manufacturing issues that will need to be investigated before the full scale antenna can be constructed. the number of conceivable conﬁgurations exceeds the likely range of applications. in order to ensure static determinacy of the complete structure and easy prestressing. Critical factors are those aspects concerned with the achievable manufacturing accuracies. This work should be aimed at simplicity and reliability.
159].CHAPTER 8. CONCLUSIONS need to be developed to make tensegrity structures useful for applications requiring long slender masts. Since the shape changes with the equilibrium of the structure only a small quantity of control energy is needed to change its conﬁguration. The kinematic indeterminacy of tensegrities is not always a disadvantage. The use of tensegrity structures as sensors and actuators is therefore another area of application. 190 . This has been explored by Skelton and Sultan [150.
R. R.. M. 688–694. K. 1975).. membrane and pneumatic structures. ASAP”. P. WA.Bibliography [1] AECAble Engineering Company. Formﬁnding and analysis of prestressed nets and membranes.aecable. M. 101–112. S. Huntsville. [3] Arianespace. retrievable boom for space applications. F. E. L. [9] Belvin. AL. K.. In Proceedings of the 23rd Aerospace Mechanisms Symposium (Marshall Space Flight Center. In Proceedings of Second International Conference on Space Structures (Guildford. 3–5 May 1989). Journal of Spacecraft and Rockets 26. H. 3 (1988). [5] Barnes. [10] Boeing Satellite Systems. Z. PhD thesis. 16–19 April 2001). NASA CP3032. Datashvili. and Montuori. Computers & Structures 30.hughespace.. [4] Baier. IEEE Transactions on Antennas and Propagation AP29. Universit´ Paul Sabatier de e e e Toulouse. Design and testing of a deployable. Structural Dynamics. 129–136. Preliminary design of large reﬂectors with ﬂat facets. J. W. [2] Agrawal. P. 3 (1989). M. USA.com (28 November 2001). Applications of dynamic relaxation to the design of cable. [8] Belkacem. and Card. http://www.. W. Recherche de forme par relaxation dynamique des structures r´ticul´es spatiales autoconstraintes. “Ariane 4 ASAP” and “Ariane 5 payloads: http://www. http://www. Supple. France. M. Gogava. AIAA 20011478. and Materials Conference and Exhibit (Seattle.arianespace.. 1987. S. Anderson. [6] Barnes. Toulouse. Building blocks of large deployable precision membrane reﬂectors. 191 . USA.com (15 November 2001). Inc. V. pp. Medzmariashvili. 4 (1981). and Dell’Amico. S. Quasistatic shape adjustment of a 15meterdiameter space antenna. 685–695.com (21 June 2001).. [7] Becchi. Ed. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures.
. ARISE—mission and spacecraft description. 4 (1991). M. R. and Back. R. S. A. Plenum Press. M. 521–522. A. A.. R. R. P. and Brodsky. and Pellegrino. Rigidity and energy. R. B. R. and Prata. In Handbook of convex geometry. K. Dordrecht. Kluwer Academic Publishers. S. 1 (1982). R. M. et al. Pellegrino and S. Gruber and J. Bailey. Lutz. 83 pages. W. New York.. 2119–2122. K. F. [24] Connelly. The development of the 15meter hoop column deployable antenna system with ﬁnal structural and electromagnetic performance results. International Journal of Solids and Structures 27. Inventiones Mathematicae 66. ch. C. M.. [14] Calladine. International Journal of Solids and Structures 27.. pp. 161–172. R. Jet Propulsion Laboratory Document 16330. [21] Connelly.. 142–151. [23] Connelly. McGrawHill. Buckminster Fuller’s “tensegrity” structures and Clerk Maxwell’s rules for the construction of stiﬀ frames. [17] Campbell. International Journal of Solids and Structures 29. Larson. 4 (1991). and Pellegrino. 1993. A design procedure for classical oﬀset dual reﬂector antennas with circular apertures. 8 (1994). The Netherlands. 255–265. Vol. 1998. R. The Netherlands. Space Technology Series. Authors’ closure. [20] Chmielewski. [18] Chesley. 2000. CA. 4. [22] Connelly. In Space Mission Analysis and Design. UK. 47–54.BIBLIOGRAPHY [11] Bouderbala. Further remarks on ﬁrstorder inﬁnitesimal mechanisms. pp. pp.3. N. American Scientist 86. [16] Calladine. A. Thorpe and P. Microcosm Press and Kluwer Academic Publishers. P. Mechanisms & Mechanical Devices Sourcebook. R. [13] Calladine.. 1999. [19] Chironis... 505–515. 69–77. 1 (1988). and Belvin. Rigidity. W. C. S.. Duxbury. J. 2 (1978). and Motro. S. Guest. 1999). Eds. Wills. NorthHolland. Space payload design and sizing. 2 (1998). 164–171. C. Tensegrity structures: why are they stable? In Rigidity Theory and Applications (New York. In IUTAMIASS Symposium on Deployable Structures: Theory and Applications (Cambridge. 192 . Eds. Firstorder inﬁnitesimal mechanisms. pp. 27–36. pp. 11–33. Amsterdam. D. J. R.. and Pellegrino. F.. C. 17 (1992).. M. Folding tensegrity systems. [15] Calladine.. IEEE Transactions on Antennas and Propagation 42. G. Jr. 1991. Mathematics and tensegrity. International Journal of Solids and Structures 14. El Segundo. USA and Dordrecht. third ed. 1145–1153. R. 9.. ch. Eds. C. B. [12] Brown. Wertz and W. Eds. 6–9 September 1998). 223–271. Acta Astronautica 17. M.
M. Eds. [27] Cook. D. B. 20 June 2001.. 59–78. S. Feasibility study of a deployable mesh reﬂector: phase II. 22. Higherorder rigidity—what is the proper deﬁnition? Discrete & Computational Geometry 11 (1994). C. Tech. Design of lowcost spacecraft. Space Technology Series. J. 3&4 (1990).BIBLIOGRAPHY [25] Connelly. 1989. A. J. Space Technology Research Vehicle (STRV). stability and chaos in structural engineering: an energy approach. D. Cambridge. Structural Topology 21 (1995). Large deployable reﬂector antenna for advanced mobile communications. Microcosm Press and Kluwer Academic Publishers.. UK. J. The Netherlands. Filed 25 March 1997. USA and Dordrecht.. 193 . G.und Raumfahrt.gov.. H.. Published 2 October 1997.. 2000. J. E. Wertz and W. ch. R. J. and Bjork. The Shock and Vibration Digest 32. Rep. Reduction of rootmeansquare error in faceted space antennas. [36] El Naschie. An extendible member. G. CUED/DSTRUCT/TR199. London. and Crane III. third ed. [28] Crone. [31] Day. [38] Fichter. [39] Fleeter.. 853–882. McGrawHill. Knight. ESA/Industry Brieﬁng Meeting. PrenticeHall. and Terrell. http://www. 100–106. John Wiley & Sons.. 30–31. http://www. D. Arup Journal (September 1969). New York. 2 (2000). M. Concepts and applications of ﬁnite element analysis. R. R. 193–200. University of Cambridge. Numerical methods. [37] Emmerich. ˚. third ed. B. Selftensioning spherical structures: single and double layer spheroids. Stress.. 9 May. 11 (1984). [34] Dietrich. [30] DatonLovett. 1999. R. 1974. 1990. and Servatius. M. pp. D. and Pellegrino. A. R. S.dlr. The analysis of hanging roofs. S.htm (21 June 2001). In Space Mission Analysis and Design. CA. ¨ [33] Deutschen Zentrum fur Luft.uk/html/space/strv/home. EngleA wood Cliﬀs. [32] Defence Evaluation and Research Agency. 353– 374.. J. International Journal of Space Structures 5. Patent Cooperation Treaty Application PCT/GB97/00839. Globally rigid symmetric tensegrities.de (4 December 2001). AIAA Journal 22.dera. ¨ [29] Dahlquist. Rooney. New Jersey. and Plesha. R. 1679–1684. A review of a family of selfdeploying tensegrity structures with elastic ties. El Segundo. and Bunce. Department of Engineering. Malkus. [35] Duffy. [26] Connelly. W. Publication Number WO 97/35706. Larson.. G..
H. Eﬀects of random member length errors on the accuracy and internal loads of truss antennas. S. D. United States Patent 3. D. Inﬂatable deployable space structures—technological summary.063. ´ [47] Geradin. Acta Astronautica 38.. L. W. pp. NASA CP2269. Bilyeu. A. D. H. D. 143–151. M. G. R. [41] Freeland. Australia. G. A. G. Rep. M. [48] Grahn. E. 30 November–3 December 1982).01. [45] Furuya. Journal of Spacecraft and Rockets 22. Tech. R.. R. S. [52] Hachkowski. G... and Chen. B.ssc. R. 28 September–2 October 1998). 554–559. 194 . 2–6 October 1995). 2 (2001).. R. M. John Wiley & Sons. Tensileintegrity structures. Cassapakis.se/ssd/papers/ (12 September 2001). In 46th International Astronautical Congress (Oslo. The design and construction of two cable domes for the Korean Olympics.. TX. D. D. USA. G. Concept of deployable tensegrity structures in space application.521. and Mikulas. G. and Rixen. 381–421. 1.. C. 5 (1985). D. New York. NASA Langley Research Center..BIBLIOGRAPHY [40] Freeland. and Bilyeu. 1986). [51] Guest. Veal. Proceedings IASS Symposium. [43] Freeland. 28 August–5 September 1992). Filed 31 August 1959. In Large Space Antenna Systems Technology—1982 (Houston. 2 (1992). In 49th International Astronautical Congress (Melbourne. International Journal of Space Structures 7. Astrid—An attempt to make the microsatellite a useful tool for space science.. and Mikulas. [46] Geiger. Development of ﬂight hardware for a large. vol. pp.5. IAF95I501. USA. second ed. CUCAS9522. Stefaniuk.. 1997. Japan. [50] Greschik. 308–314. and Peterson. M. Palisoc. D. and Rathsman. In Shells. M. R. Bilyeu. inﬂatabledeployable antenna experiment. A comparative study of the precision of deployable spacecraft structures. 265– 272. Membranes and Space Frames. http://www. IAF98I. INSTEP inﬂatable antenna experiment.. E. R. Norway. E. 2 (1996).. and Veal. A new concept for solid surface deployable antennas. S. AIAA Journal 39. E... Vol. IAF920301. A. 103–113. 2 (Osaka. [49] Greene. In 43rd International Astronautical Federation Congress (Washington. [42] Freeland. 1962. M. [44] Fuller. and Pellegrino. Granted 13 November 1962. Mechanical vibrations. Survey of deployable antenna concepts.. Sensitivity study of precision pressurized membrane reﬂector deformations.
680–686. Journal of Spacecraft and Rockets 19. Astro Aerospace Corporation Technical Document. USA. UK. In press. P. J. D. pp. 1983. University of Colorado. [58] Hedgepeth. Geometrically rigid doublelayer tensegrity grids. [53] Hanaor. AACTN1164.. Astro Aerospace Corporation Technical Document. 48–57. 100–103. 5 (1982). 227–238. vol. and Chae. 4 (1994). J. W. [55] Hearth. [60] Hedgepeth. The Netherlands. November 1981. S. J. Engineering properties of doublelayer tensegrity grids. Dordrecht. H. International Journal of Space Structures 9. Eds. Wester. J. 16 January 1989. The architecture of life. Medwadowski. 3 (1982). D.BIBLIOGRAPHY Center for Aerospace Structures. M. 195 .01. 2000. J. J. NASA CR3484. Aerospace America 23. Structures for remotely deployable precision antennas. Bistable composite slit tubes. In Spatial Structures at the Turn of the Millennium. L. Pellegrino. SAWE Paper No. et al. In 39th Annual Conference of the Society of Allied Weight Engineers. Paper IAF98Q1. Design of large lightweight precise mesh reﬂector structures. A. In IUTAMIASS Symposium on Deployable Structures: Theory and Applications (Cambridge. [61] Hedgepeth. and I. Critical requirements for the design of large space structures. and Adams. D. [56] Hedgepeth. J. 1 (January 1998). [59] Hedgepeth. Louis. AACTN1154. [64] Ingber. Denmark. MO. [65] Iqbal.. 12–14 May 1980). T. R. Flexibility of space structures makes design shaky. E. Design concepts for large reﬂector antenna structures. Eds. and Boyer. 2–6 September 1991). USA.. 211– 217. [57] Hedgepeth. 2. S. Proceedings of the International IASS Symposium (Copenhagen. J. M. Acta Astronautica (2001). M.. 1375. 6–9 September 1998). S. [63] Hirosawa. (St. Space VLBI satellite HALCA and its engineering accomplishments. Scientiﬁc American 278... A. [54] Hanaor. J. M. M. Accuracy potentials for large space antenna reﬂectors with passive structure. D. M. Pellegrino and S. K. [62] Hedgepeth. W. and DatonLovett. pp. Boulder. Thomson. AIAA Journal 20. 153–162. A.. Accuracy potentials for large space antenna structures. 5 (1985). Mogensen. Guest. Kluwer Academic Publishers. 195–200. M. CO. Inﬂuence of fabrication tolerances on the surface accuracy of large antenna structures. Inc. ii+33 pages. December 1995. M. 8 November 1991. NASA CR3663..
25 June 2001. o [74] Kuznetsov. T. University of California Press. S. Discussion of “ﬁrstorder inﬁnitesimal mechanisms”. AIAA 20001481. B. University of Cambridge. A. Uber die M¨glichkeit.Y. An introduction to the analysis of symmetric structures. E. S. Transactions of the ASME 56. Crane. VA. University of Cambridge.. CA. 1912. 1991. 2001. [75] Kuznetsov. UK. Acta Astronautica 17. 222–224.. vol. Rep. F. Duffy. [72] Knight.. H. 561–566.. Underconstrained structural systems. E. Minomo. C. AIAA. J. In Festschrift Heinrich M¨lleru Breslau. Deployable antenna kinematics using tensegrity structure design. [67] Jenkins. ¨ ¨ [73] Kotter. FL.. M. C. Journal of Applied Mechanics. USA. 3–6 April 2000). Computers & Structures 71. CUED/DSTRUCT/TR198. UK. USA.BIBLIOGRAPHY [66] Itanami. Department of Engineering. University of Florida. Department of Engineering. Springer. Leipzig. and Pellegrino. Rep. B. J. 2001. Berkeley. [77] Kwan. Analysis and design of a deployable membrane reﬂector. 984–994. New York. UK. K. S. Fan rib type deployable mesh antenna for satellite use.Y. C. 1991. [76] Kuznetsov. and Ohtomo. H. R. Tech. Arm development—review of existing technologies. Gossamer spacecraft: membrane and inﬂatable structures technology for space applications. M. 6 (1999). PhD thesis.. Innovative deployable antenna developments using tensegrity design. 1976. 1 (1989). D.. PhD thesis. Feasibility study of a deployable mesh reﬂector. 61–80. and Rooney. On immobile kinematic chains and a fallacious matrix analysis. PhD thesis. Kr¨ner. 196 .. USA. 2000. 4 (1991). Reston.. C. In 41st AIAA/ASME/ASCE/AHS/ASC Structures. 17 September 2000. 671–688. University of Cambridge. E. [78] Lai. International Journal of Solids and Structures 27. I. 191 of Progress in Astronautics and Aeronautics. and Materials Conference and Exhibit (Atlanta. Geodesic math and how to use it. and Pellegrino. Gainesville. Cambridge. 5 (1988). n Punkte in der Ebene oder im Raume o durch weniger als 2n − 3 oder 3n − 6 St¨be von ganz unver¨nderlicher L¨nge a a a unverschieblich miteinander zu verbinden. N. University of Cambridge. E. USA. Tech. pp. GA. [79] Lai. N. Guest. [68] Jensen. S. and Pellegrino. Cambridge. 517–519. [71] Knight. F. pp. [69] Kangwai. D. A. [70] Kenner. Ed. CUED/DSTRUCT/TR186. Cambridge. N. Structural Dynamics. S. A pantographic deployable mast.
S. 294–299. Edinburgh and Dublin Philosophical Magazine 27 (1864). (Editor).. [85] Linkwitz. A. On the calculation of the equilibrium and stiﬀness of frames.. S.BIBLIOGRAPHY [80] Lai. Formﬁnding by the “direct approach” and pertinent strategies for the conceptual design of prestressed and hanging structures. 27–55. N. USA. H. Reinhold Publications. W. A. Switzerland. 2 (1999). 73–87. USA. B. MT.. [81] Lai. New York. Hamamoto. K. International Journal of Space Structures 14. [87] Love. [89] Mathworks Inc.. Abstract 1508. C. Umbrellatype furlable reﬂector based on tension truss concept..estec. [91] Medzmariashvili. Tsujihata. [88] Marks. 73–80. and Homma. Structural Dynamics. 671–684. In German. Einige Bemerkungen zur Berechnung von vorgespannten Seilnetzkonstruktionen. 3 (July 1998). You.Y. MATLAB function reference—version 5. [84] Lee. 145–158. 1960.mathworks. Efremov.esa. Journal of Aerospace Engineering 11. Bedukadze. 1&2 (1996). Some highlights in reﬂector antenna development.. 10–17 March 2001). J. Shape of deployable reﬂectors. and RahmatSamii. 197 . [86] Linkwitz. Datashvili..com. M. G. S.. and Pellegrino. IngenieurArchiv 40 (1971).. Simple formulas for designing an oﬀset multibeam parabolic reﬂector. L.. Technology status of the 13 m aperture deployment antenna reﬂectors for Engineering Test Satellite VIII.. 147–152..Y. and Schek. Acta Astronautica 47. London. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures. R. Siradze. 3 (1981). Launching a 25meter space telescope: are astronauts a key to the next technically logical step after NGST? In 2001 IEEE Aerospace Conference (Big Sky.. E. International Journal of Space Structures 11. WA. Radio Science 11. G. 2–9 (2000).nl/ap2000/abstracts/a01508. 2001460.. http://www. AIAA 20011481. [83] Lalvani. In AP2000 Millennium Conference on Antennas & Propagation (Davos. 89 (1976). [82] Lake.. C. C. The dymaxion world of Buckminster Fuller. Space experiment “reﬂector” on testing the largescale deployable high precision oﬀset antenna reﬂector of a new generation at the orbital station MIR. Fuller and Snelson. 472–478. Kinteraya. IEEE Paper No. The. Y. N.J. A.. [92] Meguro. http://www. and Kravchenko. IEEE Transactions on Antennas and Propagation AP29. H. R. 1619 April 2001). 9–14 April 2000). [90] Maxwell. Chernyavski. and Pellegrino. S. and Fuller. A. M.html (7 November 2001). and Materials Conference and Exhibit (Seattle. Origins of tensegrity: views of Emmerich. 1999. Z. I. Y.. K.
E. and Johnson. K. pp. Yasaka.. Canada.. 1994. T. [99] Mitsugi. [98] Mitsugi. 5–11 October 1991). and Yasaka. and Hedgepeth.. and Suzuki. T. Acta Astronautica 47. UK. Dordrecht. Prentice Hall. and Miura. A. Ando.. K.. [101] Miura. Variable geometry truss and its application to deployable truss and space crane arm. 173–238. A.. 198 . IAF91315.... 7/8 (1985).. and Miyazaki. Concept of the tension truss antenna. M. [103] Miura. J. K. Pellegrino and S. and Meguro.. AIAA Journal 28. J. International Journal of Space Structures 8. K. J. T. [106] Miura. M. Miller & Freund’s probability and statistics for engineers. and Loads of FlightVehicle Materials. Eds. K. 316–322. Design Concepts. J..BIBLIOGRAPHY [93] Micheletti. Upper Saddle River. 1987. K. [96] Mikulas Jr. Deployment analysis of large space antenna using ﬂexible multibody dynamics simulation. pp. In 42nd International Astronautical Congress (Montreal. The Netherlands. A modular approach to build a large space antenna. R. and Pellegrino. 3. April 2001. 19–26. Draft. Acta Astronautica 12. 2 (1990).. Furuya. Guest. Structures. Y. Italy. 1999. Rome. Private communication. ch. New Jersey. [100] Mitsugi. 6 (1990). Oral presentation at “Colloquium Lagrangianum—Strutture Tensegrity: Analisi e Progetti”.. vol. sixth ed. 2 October 2001. State of the art and technology needs for large space structures. Senbokuya. and Tanizawa. 6–9 September 1998).. Structural concepts. 439–447. M. Tension truss antenna—concept. [102] Miura. Shape control of the tension truss antenna. Collins. D. 1&2 (1993). Journal of Spacecraft and Rockets 28. 291–300. A. K. 4 (1991). K. H. S. New York. 2000. 3–16. K.. 6–8 May 2001. 599–607. 10 pages. Preliminary design considerations for 10–40 meterdiameter precision truss reﬂectors.. reality and future.. [94] Micheletti. ASME. J. [97] Miller. AIAA Journal 28. and Dynamics—Assessment and Future Directions. K. Freund. In IUTAMIASS Symposium on Deployable Structures: Theory and Applications (Cambridge. M. A. On the kinematics of tensegrity towers. M. and Thomson. Draft. [95] Mikulas. On the kinematics of tensegrity structures. [104] Miura. Concepts of deployable space structures. S. I. Y. M. [105] Miura. 1 (2000). 1: New and Projected Aeronautical and Space Systems. 1098–1104.. J. Kluwer Academic Publishers.
Homma. Chapman & Hall. Albert Langen Verlag. 1999. International Journal of Space Structures 7. International Journal of Space Structures 5. Y.. [114] Motro. ASCE. Ed. H. Eds.. CA. Cambridge. R. In 42nd AIAA/ASME/ASCE/AHS/ASC Structures. Nooshin. pp. F. USA. 180–185. Von Material zu Architektur. M¨nchen. Y. UK. S. ch. 428–458. U. 3&4 (1990). Belkacem. fourth ed.. 24 May 2000.. R. N. K. University of Cambridge. A. USA. Tsujigata. M. Penalba. Leonard. Yamada... [120] Paz. H. Proceedings of the IASSASCE International Symposium 1994 (Atlanta. Elsevier. Tensegrity systems: the state of the art. H. A. The Netherlands. Design and ground veriﬁcation of large deployable reﬂector. Part I. L. [109] MoholyNagy. PhD thesis. A. 2000. In Spatial. 3599–3613. [115] Murakami. Eds. AIAA 20011480. ¨ [108] Mobius... [116] Murakami. R. H. A. and Materials Conference and Exhibit (Seattle. u [110] Motro. 704–713. [119] Panetti. New York. Forms and forces in tensegrity systems.BIBLIOGRAPHY [107] Miyasaki. [112] Motro. J. [113] Motro. pp. CA. third ed. 2. Static and dynamic analyses of tensegrity structures. Nonlinear equations of motion. 1619 April 2001). 11.. J.. [118] Pak. Abel. 2 (1992). Fourthyear undergraduate project report. Lattice and Tension structures. Space Technology Series. 1&2 (1996). University of California at San Diego. Deployable tensegrity structures. 1994). USA. R. Y. Wertz and W. 1837. K. Larson. [117] Nishimura. 199 . 1929. 891–917. Lehrbuch der Statik. Tensegrity systems and geodesic domes. Leipzig. Structural Dynamics. Structural morphology of tensegrity systems.. J. Computers & Structures 79. vol. 233–240. La Jolla. Form ﬁnding numerical methods for tensegrity systems. El Segundo. In Space Mission Analysis and Design. M. Nakamura. Initial shape ﬁnding and modal analyses of cyclic rightcylindrical tensegrity modules. In Third International Conference on Space Structures (Guildford. A. R. International Journal of Space Structures 11. Static and dynamic analyses of tensegrity structures. 1984).. and Vassart. J. W. 1997. and C. 341–351. International Journal of Solids and Structures 38 (2001). R. pp. E. and Meguro. [111] Motro. Structural dynamics: theory and computation. and Nishimura. F. WA. Microcosm Press and Kluwer Academic Publishers. 9 (2001). USA and Dordrecht. 75–83. Thermal. Department of Engineering.5.
J. A. Structural computations with the singular value decomposition of the equilibrium matrix. Guest. [134] Reibaldi. The Netherlands. ESA/ESTEC. K. and S. Wang. 2001.. ch. IUTAMIASS Symposium on Deployable Structures: Theory and Applications. Eds. CUED/DSTRUCT/TR190. In Carbon Fibers. Mechanics of kinematically indeterminate structures. 1329–1350. Department of Engineering. [129] Pellegrino.K. International Journal of Solids and Structures 22.B. ESA SP225. Marcel Dekker. A. S. New York. S. Tech. G. S. [124] Pellegrino. [122] Pellegrino. S.BIBLIOGRAPHY [121] Pellegrino. 1976. [123] Pellegrino. Antenna mechanical technologies within ESA. pp. G. [125] Pellegrino. C.. C. International Journal of Space Structures 7.. [126] Pellegrino. Cambridge. Donnet. University of Cambridge. Journal of Spacecraft and Rockets 32. [132] Pugh. S.. USA. S. ESA/ESTEC.. pp. Rep. 20–22 May 1986). In Workshop on Mechanical Technology for Antennas (Noordwijk. CUED/DSTRUCT/TR191. [127] Pellegrino. Cambridge. Kukathasan. 18 November 2000. S. 463–542. [130] Pellegrino. pp.. International Journal of Solids and Structures 26. A class of tensegrity domes. and Guest. S. S. UK. Cambridge.. [131] Pontoppidan. C. 2 (1992). In Second ESA Workshop on Mechanical Technology for Antennas (Noordwijk. 80 of Solid Mechanics and Its Applications. and Ryu. 1998. and Watt. 12 (1990). Peng. SAR advanced deployable structure. [128] Pellegrino. R. University of Cambridge. 2000... An introduction to tensegrity. CA. Large retractable appendages in spacecraft. 4 (1986).. 127–142.. third ed. PhD thesis. and Calladine. Kluwer Academic Publishers. The Netherlands. Private communication. Donnet. 6 (1995). Tech. 1006–1014. T. 1986. J. Department of Engineering. vol. Carbon ﬁber applications. 21 (1993). S. Green. A. Matrix analysis of statically and kinematically indeterminate frameworks. 26–28 June 1984). The Netherlands. 7. University of Cambridge. Analysis of prestressed mechanisms.. 41–47. Rep. S. M. 9 November 2000.. Eds. UK. K. J. International Journal of Solids and Structures 30. 5–13. D. C. 3025–3035. Berkeley. Rebouillat. ESA SP261. and Watt. J. Dordrecht. G. D. M. S. S. UK.. 409–428. S. Peng. Small satellite deployment mechanisms. University of California Press. Electrical consequences of mechanical antenna characteristics. Tibert. S. [133] Rebouillat. 200 .
Academic Press. Stutzman. In Space Mission Analysis and Design. P. S. A. T. Space Technology Series.J. R. 1 (1993). C. Unfurable satellite antennas: a review. G. T. and RahmatSamii. T.. [140] Rusch. [142] Ruze. Annales des T´l´communications 44. CA. Larson. L. J.. Satellite outages and failures. [137] Rimrott. Development of a large deployable space reﬂector structure. [144] Sarafin. 475–488. Eds. G. Y. ch. 4 (April 1966). Analysis of reﬂector antennas.. J. 1999. P.. Part G—Journal of Aerospace Engineering 206. pp. In IUTAMIASS Symposium on Deployable Structures: Theory and Applications (Cambridge. ASCE 6. September 2000. and Fritsche.. DERA Report. 201 . G. Space Technology Series. W. 6–9 September 1998). 115–134.. Proceedings of the Institution of Mechanical Engineers. QUASAT program: the ESA reﬂector. J.. DERA/KIS/SPACE/CR000495/1. USA and Dordrecht. The Netherlands. 34–54. UK. 321–333. F. Zeitschrift f¨r Flugwisu a u senschaften und Weltraumforschung 4. M. 459–497. Wertz and W.. El Segundo. and Bernasconi. G. [145] Satellite News Digest. T. The force density method for form ﬁnding and computation of general networks. Microcosm Press and Kluwer Academic Publishers. R. 5 (1980). and Britton. W. Microcosm Press and Kluwer Academic Publishers. P.6.BIBLIOGRAPHY [135] Reibaldi. 11. Journal of Aerospace Engineering. [141] Russell. J. New York. Proceedings of the IEEE 54. Guest. 9–10 (1989). The Netherlands. Technology assessment and development of large deployable antennas. [136] Reynolds.. Kluwer Academic Publishers. J.0. R. ¨ [146] Schafer. 1995. Small satellite deployment mechanisms requirement report. Eds. 1970. Acta Astronautica 15. M. CA. A. W. third ed. G. 111–123. Antenna tolerance theory—a review.com/failures/ (24 August 2001). [138] Roederer. G.. and Hedgepeth...satindex. USA and Dordrecht. Pellegrino and S. 255–267.. V. Campbell. W. D. H. The Netherlands. 2000. McCandless. 633–640. Stand der Technik auf dem Gebiet gr¨ßerer entfaltbarer o ParabolantennenStrukturen f¨r Raumﬂugger¨te. Eds. Doukas. J. El Segundo. 181–187. Dordrecht. P. and Larson. G2 (1992). J. Computational Methods in Applied Mechanics and Engineering 3 (1974). ee [139] Rogers. T. Spacecraft Structures and Mechanisms—From Concept to Launch.. Fundamentals of STEM mechanics. A. G. Structures and mechanisms. W. pp. [143] Sarafin. C. 3 (1987). [147] Schek. http://www.
C. Deployment simulation of a parabolic space antenna with rigid panels. M. Louis. SPIE—The International Society for Optical Engineering. Draft. pp. N..611. Force and torque smart tensegrity sensor. 1988. vol.. 19 April 2001. Reduced prestressability conditions for tensegrity structures. [149] Skelton. A. Regelbrugge. E. New York.ucsd. Modeling. G. Wellesley. [150] Skelton. PhD thesis. 1690. MO. 1965. and Materials Conference (St. 1&2 (1993). E. WellesleyCambridge Press.. FL. G. USA. and He. Purdue University.. [156] Strang. November 2001. [157] Strock. [160] Sultan. Introduction to applied mathematics. 1998. United States Patent 3.. New York..P. C. In Smart Structures and Materials 1998—Mathematics and Control in Smart Structures (San Diego. 342–347. R. [151] Snelson. John Wiley & Sons. International Journal of Space Structures 8. vol. and Chan. 36 March 1997). In Smart Structures and Materials 1997—Smart Structures and Integrated Systems (San Diego. 1969. and Skelton. 1999. J. G.edu/research/reskelton/ (11 November 2001). E. USA. V.. 202 . D. Harcourt Brace Jovanovich College Publishers. and Thiele... M. SPIE—The International Society for Optical Engineering. 2–5 March 1998). Smart tensegrity structure for NESTOR. California. Granted 16 February 1965. Linear algebra and its applications..BIBLIOGRAPHY ¨ [148] Sjoberg. K. In 40st AIAA/ASME/ASCE/AHS/ASC Structures. Ed. The design and manufacturing of a thermally stable submillimetre wave spaceborne radio telescope. D. E. An introduction to the mechanics of tensegrity structures. [152] Southwell. [159] Sultan. http://www. Adhikari. Dover Publications. MA.se/ssd/papers/ (24 August 2001). design.169. pp. 63–70. Corless. and control of tensegrity structures with applications. C. [161] Sultan. [154] Stern. Antenna theory and de