Static and Dynamic Analysis of Circular Cable Roof Nets

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/305407889
Static and Dynamic Analysis of Circular Cable Roof Nets
Research · July 2016

DOI: 10.13140/RG.2.1.4665.2400
CITATIONS READS
0 1,166
1 author:
Mohamed Mohie Eldin

Beni Suef University
17 PUBLICATIONS 50 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Using of textile (natural and fabric) materials as reinforcement of concrete elemnts. View project
All content following this page was uploaded by Mohamed Mohie Eldin on 19 July 2016.
The user has requested enhancement of the downloaded file.

MANSOURA UNIVERSITY
FACULTY OF ENGINEERING
DEPARTMENT OF STRUCTURAL ENGINEERING
STATIC AND DYNAMIC ANALYSIS OF

CIRCULAR CABLE ROOF NETS
A THESIS SUBMITTED IN
The Partial Fulfill of The Requirement
for the Degree of Master of Science in
Structural Engineering
By
Eng. Mohammad Mohie Eldin Mohammad Ossman
B.Sc. Structural division
Civil Engineering Department
Alexandria University
SUPERVISORS
Professor Dr.
Nabil Sayed Mahmoud
Head of Structural Engineering Department
Faculty of Engineering,
Mansoura University
Professor Dr. Associate Prof. Dr.

Ahmed Badr Mohamed Naguib Abou El-Saad
Prof., Struct. Engrg. Dept., Associate Professor, Struct. Engrg. Dept.,
Faculty of Engineering, Faculty of Engineering,
Mansoura University Mansoura University
Mansoura
2003
Supervisors
Name, Title and Affiliation Signature

1- Prof. Dr. Nabil Sayed Mahmoud ----------------------------------
Head of Structural Engineering Department,
Mansoura University
2- Prof. Dr. Ahmed Badr ----------------------------------

Head of Structural Engineering Department,
Mansoura University
3- Prof. Dr. Mohamed Naguib Abou El-Saad ----------------------------------

Associate Prof., Structural Engineering Department,
Mansoura University
Date: 5 / 5 / 2003
i
Statement
This thesis is submitted to Mansoura University for the Degree of Master of

Science in Structural Engineering.
The work included in this thesis was carried out by the author in the
department of structural engineering , Mansoura University, from October
1996 till April 2003.
No part of this thesis has been submitted for a degree or a qualification at

any other University or Institution.
Date : 5 / 5 / 2003
Signature:
Name: Mohammad Mohie Eldin M. Ossman
ii
Information about the researcher
Name : Mohammad Mohie Eldin Mohammad Ossman
Date of Birth : 21 - 9 – 1970
Place of Birth: Alexandria
Qualification : B. Sc. Degree in Civil Engineering - Faculty of Engineering -

Alexandria University – 1992
Job : Structural Engineer
iii
Mansoura University
Faculty of Engineering
Structural Engineering Department
Abstract of the M. Sc. Thesis

Submitted by Eng. Mohammad Mohie Eldin M. Ossman
Title of Thesis
Static and Dynamic Analysis of Circular Cable Roof Nets
Supervisors
Prof. Dr. Nabil Sayed Mahmoud
Prof. Dr. Ahmed badr
Assoc. Prof. Dr. Mohamed Naguib
Registration Date: 24/ 10 / 1996 Examination Date: 5 / 5 / 2003
Abstract
The main aim of the present work is to investigate both the static and dynamic analysis
of circular cable suspended roofs taking into account the non-linearity of these systems.
The classification, pretensioning, protection, mechanical properties and the static analysis of
freely hanging cable under uniformly distributed load are given. Classifications, main
elements, materials, advantages and disadvantages of cable suspension roofs are included.
Static and dynamic loads which affect cable roofs are presented with special care for wind
because of its importance for cable roofs. Methods of analysis of cable suspended roofs are
presented with a brief review for the previous works in this field. Method of minimizing the
total potential energy (TPE) using the conjugate gradient (CG) technique to predict the static
response of cable roofs is introduced. Time domain analysis by expanding the method of
minimizing total potential energy to include dynamic loads and responses is completely
presented. Also, the method of the frequency domain analysis is given. An approach to
evaluate the dynamic response due to wind is included.
Many circular cable roofs have been analyzed using a computer program based upon the
minimization of (TPE) by the (CG) technique. These results are used to investigate the
factors that affect the design and the response of cable suspended roofs. Also, these results
have been used in making non-dimensional tables and graphs and to infer new relationships.
These tables, graphs and relationships form a good technique for the preliminary design of
circular cable suspended roofs either nets or grids. This good preliminary technique was
examined using several examples and the results are in good agreement with these solved
using the computer program based on minimizing the (TPE) using the (CG) technique for
the analysis of cable nets. Also, computer programs for the dynamic analysis were used to
investigate the factors that affect the frequencies of circular cable suspended roofs and to
express their dynamic behavior.
iv
Acknowledgement
I would like to express my gratitude and thanks to Allah for his help and
grant to complete this work.
My appreciation, thanks and grateful acknowledgement are due to Dr.

Mohamed Naguib Abou El-Saad, Structural Engineering Department, Faculty
of Engineering, Mansoura University, for his valuable advices and his
continuous assistance that helped me a lot to complete this work.
Thanks are also extended to Prof. Dr. Nabil Sayed Mahmoud, Head of
Structural Engineering Department, Faculty of Engineering, Mansoura
University, for his continuous guidance.
Thanks are also to Prof. Dr. Youssef Agag and to Prof. Dr. Ahmed Badr,
Structural Engineering Department, Faculty of Engineering, Mansoura
University, for their encouragement.
Finally, special acknowledgement to my parents, my wife, my daughters ,

my brothers and sisters and to my parents-in-law.
v
List of Contents
Page
Chapter (1) Introduction
1.1 General 1
1.2 Classification of cables 2
1.3 Cable pretensioning 3
1.4 Protective coating 3
1.5 Analysis of freely hanging cables 4
1.5.1 Previous work 4
1.5.2 Static analysis 5
1.6 Object of the work 8
1.7 Layout of thesis 8
Chapter (2) Literature Review

2.1 Introduction 10
2.2 Classification of cable Roofs 11
2.2.1 Classification according to its using manner 11
2.2.2 Classification according to geometry of suspended roofs 11
2.2.2.1 Classification according to the type of cladding supports 11
2.2.2.2 Classification according to number of layers 13
2.2.2.3 Classification according to the roof dimensions 14
2.2.3 Classification due the structural behavior of suspended roofs 14
2.2.3.1 Classification according to the cables pretensioning 14
2.2.3.2 Classification according to the self balance of the structure 14
2.2.3.3 Classification due to the mathematically treatment method 14
2.3 Main Elements of cable roofs and their material 15
2.3.1 The supporting structure 15
2.3.2 The roof cladding 15
2.4 Loads 15
2.4.1 Static loads 15
2.4.2 Dynamic loads 16
2.4.3 Loads combination 16
2.5 Wind (nature and response) 16
2.5.1 Wind speed 17
2.5.2 Wind pressure 19
2.5.3 Wind loading 20
2.5.4 Distribution and evaluation of static wind loads on cable roofs 21
2.6 Methods of analysis 22
2.6.1 Continuous approach (analogous membrane solutions) 22
2.6.2 Discrete approach 22
2.6.2.1 Stiffness method 23
2.6.2.2 Finite element method 24
2.6.2.3 Minimizing of total potential energy (TPE) method 25
2.6.2.4 Dynamic relaxation method 26
2.6.2.5 Approximate linear methods 27
2.7 Advantages and disadvantages of cable systems 28
2.7.1 Advantages of cable roofs 28
2.7.2 Disadvantages of cable roofs 28
vi
Chapter (3) Static analysis of cable roofs by energy method
3.1. Introduction 37
3.2. Mathematical formulation 38
3.3. Minimization techniques 38
3.4. Expressions for the TPE 40
3.5. Numerical ill-conditioning, scaling and convergence 44
3.5.1 Condition number and sources of ill-conditioning 44
3.5.2 Scaling and ill conditioning of the TPE surface 44
3.5.3 Convergence criteria (Termination procedures) 45
3.6. Summary of the iterative procedures 46
Chapter (4) Dynamic analysis of cable roofs

4.1 Introduction 51
4.2 Time domain analysis (Step by step integration method) 52
4.3 Frequency domain analysis 63
4.3.1 Mass, damping and stiffness matrices 65
4.3.1.1 Mass properties 65
4.3.1.2 Damping matrix 66
4.3.1.3 Stiffness matrix 66
4.3.2 Reduction of NDOF 66
4.3.2.1 Static condensation method 66
4.3.2.2 Dynamic condensation method 67
4.3.2.3 Modified dynamic conensation method 68
4.3.2.4 Reduction of the structure 68
4.3.3 An indication of dynamic sensitivity 68
4.3.4 An approach for the dynamic response due to wind 68
Chapter (5)
Static and dynamic analysis of circular cable roof nets
5.1 Introduction 72
5.2. Design of circular cable nets 73
5.2.1 Explanatory examples 73
5.2.2 The preliminary design of circular cable nets 75
5.2.3 Investigation of factors affect the response of cable nets 77
5.3. Design of circular cable grids 78
5.3.1 Explanatory examples 78
5.3.2 Factors affect the response of cable grids 79
5.3.3 The preliminary design of circular cable grids 83
5.4 Factors affect natural frequencies of circular cable roofs 92
5.4.1 Natural frequencies of circular cable nets 92
5.4.2 Natural frequencies of circular cable grids 92
5.5 Application of dynamic analysis using time-domain 93
5.5.1 Deflection vs. time 93
5.5.2 Velocity vs. time 93
5.5.3 Acceleration vs. time 93
5.5.4 Lengths of members vs. time 93
5.5.5 Tensions in members vs. time 94
5.6 Application on the approach of the wind response 94
vii
Chapter (6) Summary and Conclusion 128
6.1 Summary 128
6.2 Conclusions 129
References 132
viii
LIST OF FIGURES
Figure & Title Page
Fig. (1.1): Diagram of the North Carolina State Fair Arena. 9
Fig. (1.2): Types of cables: a) Strand ; b) Rope. 9
Fig. (1.3): Simply supported cable with uniformly distributed load along the horizontal 9
projection of the span.
Fig. (2.1) Simply suspended roofs. 30

(a) Cables in parallel planes.
(b) Cables suspended in radial planes between an inner tension ring and an outer
compression ring.
(c) The central cables in parallel planes and the cables at each end suspended
between an inner split tension ring and an outer curved beam.
Fig. (2.2): Cable beam structure with corrugated metal roof decking. 31
(a) Convex cable beam structure with corrugated metal roof decking.
(b) Concave cable beam structure with corrugated metal roof decking.
(c) Convex-concave cable beam structure with corrugated metal roof decking.
Fig. (2.3) Radial cable beam structure with inner tension ring and outer compression
ring /rings. 32
(a) Convex.
(b) Concave.
(c) Convex-concave.
Fig. (2.4) Multi-span cable beam structures. 33

(a) Convex.
(b) Concave.
(c) Convex-concave.
Fig. (2.5): Saddle-shaped cable net with edge cables. 33
Fig. (2.6): 34
(a) Saddle-shaped net with edge beams.
(b) Saddle-shaped net supported between two vertically erected arches.
(c) Saddle-shaped cable net supported by a space ring.
Fig. (2.7): 35
(a) Circular convex cable grid supported by a horizontal ring beam.
(b) Circular concave cable grid supported by two ring beams.
(c) Circular convex-concave cable grid supported by two horizontal ring beams.
Fig. (2.8): Discrete and continuous representation of suspension structures. 35

(a) Discrete. (b) Continuous
Fig. (2.9): Roof cladding. 36
ix
Fig. (2.10): Distribution of wind loading on cable roof. 36
Fig. (2.11): Diagrammatic representation of the stiffness method. 36
Fig. (3.1): Two-dimensional pretensioned cable system with one free joint. 47
Fig. (3.2): Contour lines of total potential energy for the structure shown in Fig. (3.1). 47
Fig. (3.3): Contour map of the TPE surface for a two DOF structure indicating
diagrammatically the basic method used in the analysis of nonlinear structures. 48
Fig. (3.4): Construction of two conjugate directions: The directions v k and v k 1

are k -orthogonal if v k 1 is parallel to the tangent v to the ellipse ( v T kv  constant  c ).
The equation of the straight line through v k and parallel to v k 1 is in a parametric
form: v  vk  tv k 1 where t  scalar quantity. 48
Fig. (3.5): Geometric representation of the conjugate gradient method for a two DOF
structure. 49
Fig. (3.6): Convergence in one direction for both methods of steepest descent and
conjugate gradients. 49
Fig. (3.7): Coordinates and force for link ab in:

(a) Pretensioned but not loaded cable assembly,
(b) Pretensioned and loaded cable assembly at stage of the iterative process where
the extension of the link ( ab ) is eab and
(c) Pretensioned and loaded cable assembly when the calculated strain is eab
and the strain due to the next step Sv is eab . 50
Fig (4.1) Assumed change in acceleration during a time step t in the linear
acceleration method 70
Fig (4.2) Assumed change in acceleration during a time step t in Wilson- method 70
Fig (4.3) Assumed change in acceleration during a time step t in the constant
acceleration method 70
Fig (4.4) Diagram showing the motion of a lumped-mass spring system with viscous
damping ratio  > 1.0 71
Fig (4.5) Variation of aerodynamic admittance factor A(n) with the reduced frequency
nB/U(10). 71
Fig. (5.1): Calculation of uniformly distributed load per unit length of the cable. 100
Fig. (5.2): Circular cable net with ( 10  10 ) spacing showing the numbering of joints. 100
x
Fig. (5.3): Non-dimensional relation between the central deflection and other
parameters in (10  10) circular cable net with sag/span ratio of 3%. 101
Fig. (5.4): Non-dimensional relation between the maximum cable force and other
Fig. (5.5): Non-dimensional relation between the central deflection and other parameters
in (10  10) circular cable net with sag/span ratio of 4%. 102
Fig. (5.7): Non-dimensional relation between the central deflection and other parameters
in (10  10) circular cable net with sag/span ratio of 5%. 103
Fig. (5.9): The relation between the central deflection and other parameters for different
sag/span ratios. 104
Fig. (5.10): The relation between the maximum cable force and other parameters for
different sag/span ratios. 104
Fig. (5.11): The relation between the central deflection and cable steel area for different
values of applying loads. 105
Fig. (5.12): The relation between the central deflection and cable steel area for different
values of the spacing between nodes. 105
Fig. (5.13): The relation between the central deflection and the pretension in cables. 106
Fig. (5.14): The relation between the maximum cable force and the pretension in cables. 106
Fig. (5.15): (10  10) Circular cable roof. 107
Fig. (5.16): The relation between the nodal deflection and the steel area of either
suspension or pretensioned cables in both convex and concave grids. 108
Fig. (5.17): The relation between the maximum cable forces and the steel area of either
suspension or pretensioned cables in both convex and concave grids. 108
Fig. (5.18): The relation between forces in struts and the steel area of either suspension or
pretensioned cables in a convex grid. 109
Fig. (5.19): The relation between forces in ties and the steel area of either suspension or
pretensioned cables in a concave grid. 109
Fig. (5.20): The relation between the strut steal area and the nodal deflections. 109
xi
Fig. (5.21): The relation between the steel area of ties and the nodal deflections. 110
Fig. (5.22): The relation between the steel area of ties or struts and the forces in cables. 110
Fig. (5.23): The relation between the steel area of either ties or struts and their final
forces. 110
Fig. (5.24): The assumed distribution of forces in cable beams / grids for preliminary
design calculations. 111
Fig. (5.25): The relation between the sag / span ratio and the nodal deflections of both
suspension and pretensioned cables in either concave or convex grids. 111
Fig. (5.26): The relation between the sag / span ratio and the maximum forces in both
Fig. (5.27): The relation between the rise / span ratio and the nodal deflections of both
Fig. (5.28): The relation between the rise / span ratio and the maximum forces in both
Fig. (5.29): The relation between (sag / span) or (rise / span) ratios and the maximum
forces in struts or ties. 113
Fig. (5.30): The relation between the pretension of the pretensioned cable and the nodal
deflections of both suspension and pretensioned cable in either concave or
convex grids. 113
Fig. (5.31): The relation between the pretension of the pretensioned cable and the
maximum forces in both suspension and pretensioned cable in either concave or
convex grids. 113
Fig. (5.32): The relation between the pretension of the suspension cable and the
nodal deflections of both suspension and pretensioned cable in either concave or
convex grids. 114
Fig. (5.33): The relation between the pretension of the suspension cable and the
maximum forces in both suspension and pretensioned cable in either concave or
convex grids. 114
Fig. (5.34): The relation between the pretension of the suspension or the pretensioned cables
and the maximum forces in struts or ties. 114
Fig. (5.35): The relation between the pretension in ties and:

(a) the nodal deflections of both suspension and pretensioned
cables,
(b) the maximum forces in both suspension and pretensioned
cables, and
xii
(c) the maximum forces in ties.
115
Fig. (5.36): The relation between the concurrent increasing of pretension in both
suspension and pretensioned cables and: (a) their nodal deflections and (b) their
maximum forces. 115
Fig. (5.37): The relation between the number of segments in the diametrical member and:
(a) the nodal deflections of both suspension and pretensioned cables in either
concave or convex grids;
(b) the maximum forces of both suspension and pretensioned cables in either
concave or convex grids. 116
Fig. (5.38): The relation between the number of segments in the diametrical member
and the maximum forces in ties or struts. 116
Fig. (5.39): The relation between the gap that lies between both the suspension and
the pretensioned cables of a concave grid and:
(a) the nodal deflections of both the suspension and the pretensioned cables of the grid,
and (b) the maximum forces in both the suspension and the pretensioned cables of
the grid. 116
Fig. (5.40): A plan of the concave part of (20  20) circular cable convex grid roof. 117
Fig. (5.41): Effect of increasing the applying load upon frequencies. 117
Fig. (5.42): Effect of increasing the pretension on frequencies in a circular cable net. 118
Fig. (5.43): Effect of increasing cable steel area on frequencies in a circular cable net. 118
Fig. (5.44): Effect of increasing sag/span ratio on frequencies in a circular cable net. 118
Fig. (5.45): Effect of increasing the spacing between nodes on frequencies in a circular
cable net. 119
Fig. (5.46): The first modeshape of a circular cable net. 119
Fig. (5.47): The second modeshape of a circular cable net. 119
Fig. (5.48): Effect of the sag/span ratio on natural frequencies of circular cable grid. 120
Fig. (5.49): Effect of the rise/span ratio on natural frequencies of circular cable grid. 120
Fig. (5.50): Effect of the spacing between hangars and natural frequencies of circular
cable grid. 120
Fig. (5.51): Effect of the steel area of the suspension cable on natural frequencies of
circular cable grid. 121
xiii
Fig. (5.52): Effect of the steel area of the pretensioned cable on natural frequencies of
Fig. (5.53): Effect of the steel area of the hangars on natural frequencies of circular
cable grid. 121
Fig. (5.54): Effect of the pretension of the suspension cable on natural frequencies of
Fig. (5.55): Effect of the pretension of the pretensioned cable on natural frequencies of
Fig. (5.56): Effect of the pretension of the hangars on natural frequencies of circular
cable grid. 122
Fig. (5.57): Nodal deflections during time of analysis in circular cable net. 123
Fig. (5.58): Comparison between central deflections during the time of analysis in
cable net, convex grid and concave grid. 123
Fig. (5.59): Nodal velocities during time of analysis in circular cable net. 124
Fig. (5.60): Velocity of the central joint in the suspension cable of a circular grid
during the first second of the time of analysis. 124
Fig. (5.61): Velocity of the central joint in the suspension cable of a circular grid
during the last second of the time of analysis. 124
Fig. (5.62): Nodal accelerations during time of analysis in circular cable net. 125
Fig. (5.63): Central deflections in circular cable net during the last second of the time
of analysis. 125
Fig. (5.64): Central velocities in circular cable net during the last second of the time
of analysis. 126
Fig. (5.65): Central accelerations in circular cable net during the last second of the time
of analysis. 126
Fig. (5.66): Lengths of members during the time of analysis in circular cable net. 126
Fig. (5.67): Final tensions in members during the time of analysis in circular cable net. 127
Fig. (5.68): Comparison between lengths and tensions in suspension and pretensioned
cables in circular convex grid. 127
xiv
LIST OF TABLES
Table & Title Page
Table (2.1) Experimental values for  and z g . 29
Table (2.2) Roughness length z 0 and surface drag coefficients k for various types
of terrain. 29
Table (2.3) The principal wind pressure, q, in (kg/m2). 29
Table (2.4) The exposing factor, k. 29
Table (5.1) Data assumptions for case study (1). 95
Table (5.2) Numbering of members of the cable net shown in Fig (5.2). 95
Table (5.3) Non-dimensional table for circular cable net with (10  10) spacing. 96
Table (5.4) (16  16) Circular net with  = 4% and [w  0.25w] . 97
Table (5.7) Example (6):

(10  10) Convex grid with sag = rise = 4% and [ w  0.80w ]. 98


(10  10) Concave grid with sag = rise = 4% and [ w  0.80w ]. 98



xv
xvi
Chapter (1)
Introduction
1.1 General
Cable structures are one of the two categories of tension structures, which include both
cable structures and membrane structures. In tension structures the main load-carrying
members transmit the applied loads to the foundations or other supporting structures by
1
direct tensile stress without flexure or compression ( ).
Cable suspension structure may be defined as the structure whose principal supporting
2
elements are cables (tension members) draped between anchorages ( ).
A historical review of suspended roofs suggests that the earliest version of a tension roof
was a tent, which even today is considered perfect example of a membrane roof with
predominantly tensile forces (3). The earliest recorded cable structures were bridges located
in the Far East as in China where some of the early suspension bridges were constructed
with rope and iron chains (4). The first modern suspension bridge was built by J.Finley
in1796 and was followed, 73 years later, in 1869 by the famous Brooklyn Bridge by John
and Washington Roebling.
The earliest cable roof was probably the (189x156m) Roman Colosseum built in 70 AD,
with 180m cables supported a sunshade that could be drown over them to cover the arena.
The Colosseum was not followed by any appreciable advance in the construction of cable
roofs until 1953 when the State Fair Arena at Raleigh designed by Nowicki, Fig. (1.1), was
built in North Carolina-USA (3). This building of State Fair Arena at Raleigh, USA was the
inspirer for architects and engineers to construct many cable roof structures everywhere in
the world and the real development in analysis, design and construction of cable roofs have
occurred only after it (3).
Suspended roofs present quite different problems from suspension bridges. Roofs are much
lighter due to the smaller loads acting on them. Because of their light dead weight, they are
much more sensitive than heavy bridges to the lifting forces caused by wind suction. Due to
this fact, cable roofs did not gain popularity until the understanding of cable behavior gained
through suspension bridges had been achieved. Simply, it seams that the impetus for the
design of modern cable roofs was attributed to the success of suspension bridges.
The development of high tensile steel cables has made it possible for designer to transmit
large axial forces in tension at a relatively low cost and so, cable suspended roof is the
economic solution to cover large open area without interior columns. However, cable
suspended roof becomes an economic competitor, for smaller span structures, alternative to
conventional forms of structures such as portal and space frames (5). They have been used to
cover different types of buildings such as stadiums and sport halls, swimming pools and
water reservoirs, concert halls, theatres, cooling towers, hangars, warehouses and factories.
Cable suspended roofs must be designed not only for static loads but also for dynamic loads.
The failure, of the 853-m Tacoma Narrows bridge in 1940, because of aerodynamic
oscillations, constant wind of 41m.p.h (4), showed that it was not adequate to design
structures for static stability only (5).
1
“Cables offer fresh architectural forms and rapid construction as well as the ability to span
long distances with minimum weight. The development of high-strength steel and the
availability of electronic computers for accurate analysis of complex cable configurations
have made these systems economically competitive with more conventional types of
construction” (6).
In Egypt, cable roofs are used to cover three sport halls in Cairo International Stadium with
diameters 41.6, 50.0, and 120.0 m. The biggest hall can contain 25,000 spectators. Also,
three cable bridges, which considered as cable-stayed roofs, are used. The first one is a part
of 6th October Bridge at Ghamra-Cairo. The second is the Quantara Bridge across Suez
Canal at height 70.0 m over the maximum water surface level to allow huge ships to pass
safely beneath it. The third is Aswan Bridge across the Nile River.
Finally, this chapter includes both mechanical and structural properties of single cables,
object of the work and layout of the thesis.
1.2 Classification of cables

“Cable” is a flexible steel tension member that may be a strand, rope, or a group of strands,
ropes, or parallel wires acting as a unit member and protected by a uniform coating, usually,
of pure zinc. Cables have very much small transverse dimensions comparing with its
longitudinal dimension (3). Hence, cables have negligible buckling strength and cannot
sustain any bending or torsion moments and they can transmit loads only by developing
direct tension. Cable material is entirely utilized because of:
(a) The tensile stress across the cable section is uniform; and
(b) The variation of tension along the length, for practical sag/span ratio, is small.
Cables may be classified into two main types, structural strands and ropes, Fig. (1.2).
a) Structural strand: It is an assembly of wires formed helically around a center wire
in one or more symmetrical layers. Wires number depends on the required diameter and
ultimate strength.
b) Structural rope: It is composed of several strands laid helically around a core that
may be either another strand or an independent wire rope. Number of wires per strand that
processed into wire rope depends upon the required diameter of the rope.
Rope vs. strand (7)
1- Strand is stronger than rope for the same nominal size;
2- Strand has a higher modulus of elasticity than rope;
3- Strand is less flexible than rope, therefore, is not used with small bend radii;
7
4- Strand has greater bending stiffness and, as a result, ( ) used where bending or
flexibility is not a major requirement on the contrary of rope;
5- Strand has larger diameter wire than a rope of the same nominal size;
6- Strand of a specific class of zinc coating is more resistant to corrosion because of the
large size wires;
7- Strand has a better strength/weight ratio (3); and
8- Rope is easier to grip than strand.
Economic selection of cables (7)
A criterion of economic cable selection theoretically would encompass:
1) Minimum number of strands;
2) Largest number of wires per strand; and
3) Largest diameter of individual wires.
2
1.3 Cable pretensioning
Cables, either ropes or strands, must be pretensioned to increase the initial modulus of
elasticity, (2) for the following reasons:
a) To insure predictable elastic behavior of the cables;
b) To enable the designer to, accurately, calculate the expected elastic deformations of the
structure when subjected to various loading conditions;
c) To get accurate values for cable lengths when subjected to a prescribed tension (2);
d) To permit a cable to be measured and marked, under specified loads, to facilitate field
installation (erection); and essentially
e) To eliminate the constructional stretch of the cable, which happens because of
manufacturing operations.
The pretension load does not exceed 55% of the rated minimum ultimate strength of strand
or 50% of the rated minimum ultimate strength of rope.
Pretensioning is achieved either by (3):

1) Preloading the cables, or by
2) Employing counter-stressed systems.
A simply suspended roof may also be stiffened by designing the cladding as a pretensioned
shell by putting pre-cast concrete slabs on the cables and loaded with predetermined weights
and then the spaces between slabs are grouted with concrete, and the applied loads are
removed when the concrete sets.
The factors govern the rate of pretensioning include (3):
1) The ratio of dead load to the live load,
2) The limitation on deflections, and
3) The structure type.
The pretensioning required can be determined by trial and error, on the basis of approximate
calculations, and can be considered satisfactory if, on application of applied loads:
a) The deflections are within the permissible limits; and
b) The elements do not become slack.
A large amount of pretension may become necessary in networks, with flat zones, to avoid
excessive deformations and slackening of members. It is also necessary to assign values of
pretension such that the system fundamental frequency is in a range that resonance with
applied dynamic loads is avoided. Finally, it is necessary to introduce the entire value of
pretension in a cable or a set of cables not in a single step but rather in a number of carefully
determined sequence steps to avoid the occurrence of large moments and deformations
which can occur during pretensioning, while the axial forces are built up gradually (3).
1.4 Protective coating

Cables have to be protected with anticorrosive material. Suitable lubricants or rope
dressings provide the main corrosion protection, which may not be sufficient for permanent
structural applications. Therefore, an additional protection is provided either by metallic
coating, galvanizing or aluminizing the wires, or through an extruded plastic coating. With
the exception of permanent sheathing, say, high-density polyethylene, it is necessary to
regularly inspect surfaces to ensure that the lubricants or dressings have not been removed
by abrasion or chemical reaction and to insure that additional coats are applied when
necessary (5). Time interval between inspections will depend on the particular environment
in which cable is employed (5).
3
a) Galvanizing (zinc coating)
Galvanizing is the most common metallic protective coating which employ pure zinc as a
uniform coating with thickness up to 300 g/m2. Zinc coating life is directly proportional to
its weight (or thickness) (3). Zinc coating is classified into three categories, A, B, and C in
terms of weight per square length of wire surface.
Class A is the standard coating weight, B  2 A and C  3 A . The choice of the coating class
depends on the application and amount of protection required. However, it is observable
that, for the same cable cross section, the heavier zinc coating displaces a larger portion of
the steel area, and therefore the ultimate strength of the cable is reduced.
b) Aluminizing
The second metallic protective coating is the aluminizing using a suspension of aluminum
particles in a hydraulic resin, diluted by a solvent for application (3).
c) Plastic coating
It is made using an extruded plastic coat. Coating has to be removed at anchorages, and
after stressing the strand is re-protected by application of an air-drying plastic (3).
1.5 Analysis of freely hanging cables

1.5.1 Previous work
Krishna (3) illustrated the analysis of single cables either as a catenary, parabola, or as
straight element under uniformly distributed load or concentrated loads, and put a
generalized method for cable analysis under general case of loading.
Buchholdt (5) presented the governing Equations for freely suspended parabolic cable
under several types of distributed load along the cable span.
Greenberg (8) presented a mathematical representation of cable stress-strain curves.
Morales (9) used the shear-volume method to calculate the final cable length. This
method is valid for not very large sag/span ratio and for no loads parallel to the cable chord.
O’Brien (10) investigated the behavior of freely suspended cable under a single
concentrated load, applied in turn in each of the three principal directions.
Buchanan (11) derived Equations for the analysis of two-dimensional cable using the
perturbation technique.
Wilson and Wheen (12) and (13) developed general direct design expressions for
general cable under general loads distributed part way across the span.
Desai (14) developed a method for evaluating the central dip corresponding to an
applied load for a simple, and not initially stretched, flexible cable.
Judd and Wheen (15) described a trial-and-error procedure to calculate the horizontal
components of cable tension and then deduce the corresponding applied loads in the desired
configuration. It is an expansion of the work (12 and 13).
Huddleston (16) presented an iterative method. It consists of numerically integrating

the system of exact nonlinear differential Equations that describe the cable in final state.
4
Jayaraman and Knudson (17) presented a new small strain elastic catenary element for
the analysis of cable structures. The Newton-Raphson procedure has been used to solve the
Equations of Equilibrium and the central difference method has been used for the dynamic
computations.
Levinson and Kane (18) presented an algorithm to solve an inextensible uniform cable
supported at its two ends and loaded solely by its own weight.
1.5.2 Static analysis (3)

Complete analysis of a cable in any of its applications is divided into two steps:
(I) The establishment of the cable geometry under its own weight; and
(II) The cable analysis under applied loads.
The own weight is assumed to include any permanent uniform load superimposed on the
cable. For strict analysis, cable segment is considered as catenary and for approximation, it
may be considered either parabolic or straight segment.
The approximation of cable segment as straight element is valid only if:

(1) The cable is, or assumed to be, weightless along its length; and
(2) The ends of the segments are subjected to forces that are larger than the effects of the
loads along the segment.
The difference between catenary and parabolic segments is that the uniformly distributed
loads, i.e. own weight, are considered superimposed along the arc length of catenary and
along the span of the parabola respectively.
The approximation of a catenary as a parabola is justified in the following two cases:

(1) Sag/span ratio is small; i.e. the difference between the arc length and the span is very
small, and
(2) If the cable has permanent superimposed load uniformly distributed along its span
and as much heavier than the cable itself.
Practically, using the parabolic segment is sufficient for the following reasons:
(1) The parabolic segment analysis is much easier than catenary segment analysis.
(2) Many suspended roofs, such as bridges, have the load more uniformly distributed
along its span length than the length of the arc.
(3) The difference in shape between parabola and a catenary with the same span and
central sag is so small that it needs an exaggerated scale to be perceived.
(4) The practical sag/span ratios are sufficiently enough to assume that the load along
the span is the same as that along the cable itself.
Freely hanging cable under uniformly distributed loading (5)

The governing Equation of the parabolic segment, Fig. (1.3), is:
2z
H 2 w 0 (1.1)
 x
Where: H = The horizontal component of the tension in the cable;
w = The uniformly distributed load along the cable span; and respectively
x and z = The horizontal and vertical coordinates of a point along the cable .
5
By integrating the above Equation twice and for boundary conditions (z =0 at x =0 and z =Z
1 Z
at x =L):  Hz = w (Lx-x2)+H x (1.2)
2 L
w Z
 z [ Lx  x 2 ]  x (1.3)
2H L
M
Or z  x  x tan  d x  x tan (1.4)
H
Z
Where d x  z  x tan and tan  (1.5)
L
Eq. (1.4) indicates that for all transversely loaded cable segments, the deflected shape
assumed by the cable is the same as the shape of the moment diagram for an Equivalent
simply supported beam loaded in the same manner as the cable with values Equal to the
moment dividing by H, drawing beneath the chord of the cable which at distance x tan
from span. The maximum sag position is found from the condition:
  Z  w w
( d x )  ( z  x)  ( [ Lx  x 2 ])  ( L  2 x)  0 (1. 6)
x x L x 2 H 2H
 The position of maximum sag, d, occurs in the mid span, ( x  L / 2 )
L Z
Substituting x  at ( z   d ) in Eqn. (1.2) yields:
2 2
wL2
H (1.7)
8d
Finally, substituting the expression for H in Eq. (1.2) yields:
4d Z
z  2 ( Lx  x 2 )  x (1.8)
L L
The vertical component of the cable tension, V, is expressed as:
dz Z wL x Z 4f
V H H  (1  2 )  H [  ( L  2 x)] (1.9)
dx L 2 L L L
d
Where f   sag/span ratio
L
Eq. (1.8) shows that the cable shape is parabolic when the load is uniformly distributed
across the span rather than along the cable itself.
The length of the cable can be expressed as:
L
L dz 16d 2 8dZ Z
l   1  ( ) 2 dx   1  4 ( L2  4 Lx  4 x 2 )  3 ( L  2 x)  ( ) 2 dx (1.10)
0 dx 0 L L L
Expanding the square root in Eq. (1.10) by the binomial theorem, including the first three
terms only and by integrating between limits, yields the following expression:
8 d 32 d 1 Z 1 Z 4 d Z
l  L{1  ( ) 2  ( ) 4  ( ) 2  ( ) 4  ( ) 2 ( ) 2 } (1.11)
3 L 5 L 2 L 8 L 3 L L
The tension in the cable at distance x from the left-hand support can be found by applying
the geometric constraint condition:
dl dz
dl  (dz ) 2  (dx) 2 Or  1  ( )2 (1.12)
dx dx
dl dz
And Tx  H  H 1  ( )2  H 2  V 2 (1.13a)
dx dx
6
Using Eq. (1.9) Tx is expressed as:
4f Z
Tx  H 1  [ ( L  2 x)  ] 2 (1.13b)
L L
Z
Thus TA  H 1  (  4 f ) 2 (1.13c)
L
Z
And TB  H 1  (  4 f )2 (1.13d)
L
Values of H and d will change to ( H  H ) and (d  d ) if there are changes in:

a) Applied load; b) Temperature; c) Cable length or d) Movements of supports.
The change in (H) due to variation of (d) may, referring to Eq. (1.6), be written as:
( w  w)( L  L) 2
( H  H )  (1.14)
8(d  d ) 2
Where: w  The change in the distributed load per unit length of span; and
L  The change in the span.
For horizontal movements of supports

BY differentiating Eq. (1.11) with respect to L, this yields to:
120 L4  320d 2 L2  2304d 4  480d 2 Z 2  60Z 2 L2  45Z 4
d   L (1.15)
640dL3  3702d 3 L  320dZ 2 L
For change in cable length, l , caused by either (a), (b), or (c)

By differentiating Eq. (1.11) with respect to ( l ), this yields to:
15L3
d  l (1.16)
8(10 L2 d  5Z 2 d  48d 3 )
Where l is expressed as the following:
For (a), change in applied load:

l L
Tx H dz H 16
l   dl   1  ( ) 2 dx  ( L2  d 2  Z 2 ) (1.17a)
o
EA EA o dx EAL 3
For (b), change in temperature: l   t l (1.17b)
For (c), change in cable length: l  (1.17c)

Where   The coefficient of linear temperature expansion;
t  The change of temperature; and  The change in cable length.
Finally, the modified horizontal support reaction is calculated using Eq. (1.7) and the
corresponding modified cable tension using Eq. (1.13b).
The same approach may also be used for cables under non-uniformly distributed loads and,
however, will lead to very lengthy and cumbersome expressions. Therefore, if an elastic
analysis is required, it is better to consider the distributed load as a series of concentrated
loads and then find the forces and displacements by one of the methods given in chapter (2).
7
1.6 Object of the work
The object of this thesis is to investigate static and dynamic analysis of circular cable
suspended roofs. The following studies have been carried out for this purpose:
1. A general review of cable suspended roofs includes classifications of cable roofs,
presentation of the loads that affect these roofs including the wind in details, and
also the considered methods of analysis.
2. The development of a theory, based on minimization of the total potential work for
the nonlinear analysis of cable suspension roofs subjected to both static and dynamic
loads.
3. The development of a theory based on frequency domain method for the dynamic
analysis of cable suspension roofs.
4. Designing of some circular roofs using computer programs to verify the factors that
affect the response of such structures and using the results to infer a good method for
the preliminary design of circular cable suspended roofs based on non-dimensional
tables or graphs and a small number of formulas.
1.7 Layout of thesis

In chapter (1), classification, pretensioning and protection of the cables, which are the
basic elements of cable suspended roofs, are presented. The static analysis of freely hanging
cable under uniformly distributed load is also included.
In chapter (2), classifications of cable roofs are presented in detail. The main elements
of cable roofs and the types of applying loads are presented. Wind is presented in detail
including its nature, statistical properties and response. Also, a review of the methods used
for the design of cable suspended roofs is presented.
Finally, both advantages and disadvantages of cable suspension roofs are included.
In chapter (3), a method, for predicting static responses of cable suspended roofs, is
presented and is based on the minimization of the total potential work using the conjugate
gradients technique and valid for both linear and nonlinear analysis.
In chapter (4), the static method mentioned in chapter 3 is expanded to include

dynamic analysis. Also, a method of the frequency domain analysis for cable suspended
roofs is explained. Finally, an approach for predicting the response of cable roofs due to
wind in frequency domain is presented.
In chapter (5), some circular cable suspended roofs have been analyzed using a
computer program based on the minimization of total potential energy by conjugate gradient
technique. The results of such analysis are used to investigate the factors that affect the
design and the response of cable suspended roofs. Also, these results have been used in
making non-dimensional tables and graphs and to infer new relations. These tables, graphs
and relationships are used in the preliminary design of circular cable suspended roofs either
nets or grids. A computer program based on the frequency domain analysis is used for
calculating the natural frequencies of many circular cable roofs to examine the factors that
affect the natural frequencies of such structures. Another computer program based on
minimization of the total dynamic work is used for calculating the dynamic response of
circular cable suspended roofs.
In chapter (6), a general summary and conclusion of the thesis is presented.
8
Fig. (1.1): Diagram of the North Carolina State Fair Arena.
(a) (b)
Fig. (1.2): Types of cables: a) Strand ; b) Rope.
Fig. (1.3)
Simply supported cable with uniformly distributed load along the horizontal projection of the span.
9
Chapter (2)
LITERATURE REVIEW
2.1 Introduction:
Cable structures are highly redundant and the deformation of the individual systems
plays an important role in determining the stresses in the segments. The response of cable
systems to loads is generally nonlinear and, as a result, the principle of superposition is not
valid, although it can be admitted as a measure of approximation, and so, the use of
influence lines becomes inapplicable (3).
This nonlinearity may be either “geometric“ or “material“ nonlinearity or both. “Geometric

nonlinearity.“ results from the large deflections in the structure due to the cable flexibility
4
and “material nonlinearity“ ( ) results when the structure material is stressed beyond its
elastic limit.
It has been mentioned that cable cannot carry bending or twisting moments and it will
become slack under compressive forces. The only considered type of forces in cable
structures is the axial tension and therefore, cable systems are generally pretensioned to
insure that the entire system remains in tension under the worst conditions of loading.
A fundamental assumption made in the analysis of pretensioned cable roofs is that the
cables are the main load-carrying elements and both axial and shear stiffness effects of the
cladding are ignored.
“The structural analysis of a suspension system under static loading may be considered as
comprised essentially of two problems:
(1) Determining the initial position (shape) of the system under the initial loading.
(2) Computing the displacements of the system that result from changes of loading,
temperature, or support position” (2).
“The initial shape of a suspension system is its equilibrium position under the initial loading
which usually consists of the suspension system dead load plus cable pretension forces, if
any. The parameters that may be specified from the initial shape include:
a) Cable length,
b) Horizontal component of cable tension,
c) Magnitude, direction and sense of applied forces, and
d) The co-ordinates of significant points on the cable system” (2).
The exact analysis of cable systems is still recognized as a complex and lengthy process.
Cable structures may be treated mathematically as discrete or continuous systems. Both
methods of analysis generate nonlinear equations due to either geometric and / or material
nonlinearity.
The exact analysis of cable suspended roofs requires also an exact representation for the
applying loads especially dynamic loads; i.e. wind and earthquake loading.
This chapter represents a general review of the previous work about cable roof systems:
classifications, design and types of loading. Adding to these, a review of both the wind
(nature and response) and the different methods of analysis are represented. Finally, both
advantages and disadvantages of cable systems are also represented.
10
2.2 Classification of cable Roofs
3)
2.2.1 Classification according to its using manner (
a) Cable supported roofs (cable stayed roofs)
In cable supported (stayed) roofs, as in stayed bridges, the cables function is to provide
additional support for the roof elements, which are otherwise sufficiently strong to carry a
major proportion of the loading. In this type the cables are straight lines and the decking is
almost rigid and cable movements (change in geometry) is somewhat smaller than those for
corresponding unstiffened roofs (19).
b) Cable-cum-air supported roofs (pneumatic roofs)

In this type, the roofs can be supported either by air pressure only or by combination of
air pressure and cables. Air pressure stretches the roof skin to form a tensioned membrane,
which is attached to the cables, and they are tensioned too. The tension in cables and
membrane is transmitted by anchorages to the earth.
The basic elements of pneumatic roofs include structural membrane, cables, inflation
equipment, anchorages, entrances, and exits. Pneumatic roofs can be classified as:
I) Air inflated roofs
The roof is of tubular or cellular construction, which is capable of transmitting the applied
loads to the points of support. It is expensive since it needs a larger membrane area and air
pressure. In this type, the inner air-pressure equals outer air-pressure (19).
II) Air–supported structures
This type provides a single-wall enclosure, The membrane is attached to the support
along the periphery .The membrane is stretched and elevated by a slight increase in the
internal air pressure so that it can support the applied load. This system is often supported
on cables or ropes and the inner air-pressure must increase the external air pressure (19).
Pneumatic roofs are excellent as temporary constructions but with some limitations like
maintaining the pressure differential across the membrane, thermal and acoustic insulation
of thin membrane and the regular checking of the building so that the safety is not impaired.
c) Cable suspended roofs

In this type, the system of cable carries the roof load directly and as such has a primary
structural function and also serves as the falsework of the cladding erection. More details
will be mentioned in section (2.2.2).
2.2.2 Classification according to geometry of suspended roofs

2.2.2.1 Classification according to the type of cladding supports
a) Simply suspended cable roof systems Fig. (2.1)
In this type, roof cladding is supported by a series of simply suspended cables (20). Such
systems have no stiffness and to reduce movements the roof cladding must either be very
heavy or act as a shell. Reinforced concrete slab is the most suitable roofing material.
Unless the spaces between the slabs are filled with concrete, thus forming a concave shell,
the roofs are easily deformed and not very satisfactory (20). Its advisable to considerably
pretension the cables to stiffen the structure while construction and to prevent crack forming
in the finished roof. In the case of circular structures, jacking the central tension ring
downwards and releasing it after the concrete has set could do this. Also, prevention of
cracking, when prefabricated slabs are used, can be done by placing an over load on the roof
before inserting grout between the slabs, and then removing the load when the grout has set.
11
(5)
b) Pretensioned cable beam structures Fig. (2.2) & Fig. (2.3)
Lighter and stiffer systems than “simply suspended cable structures“ can be achieved by
connecting a second set of cables with reverse curvature to the suspension cables. In other
words, other methods of stiffening freely suspended cable systems are to introduce internal
ties or suspend heavy beam from cables (19). Such a system will be quite stiff if tensioned to a
level that the both sets of cables remain in tension under any combination of applied loading.
Pretensioned cable beams, which are considered as structural mechanisms, may be classified
as convex, concave and concave-convex beams. Also, multi-span cable beam structures, Fig.
(2.4), can be used to interconnect a number of large spans or where column-free interiors are
not required with the following advantages:
1) It reduces the size and hence the cost of the required anchors.
2) It decreases the height, dimensions and foundations of the flexural columns.
3) Reduction of height saves the cladding cost and reduces the internal volume, which
needs to be air-conditioned.
These previous advantages lie in the fact that the anchor forces are functions of the
maximum span size and not of the spans number.
c) Pretensioned cable net structures (2)

The net is a system where single-layer cables are arranged in plan as a mesh and may be
constructed in many patterns. The stiffness is achieved by tensioning the cables before
placing the roof cladding (19). The cables of a net may be orthogonal or oblique. Net
structures may be classified according to the roof curvature as:
i) Flat Roofs (5)
They are not practical for large span roofs because they are not very stiff and usually result
in areas of the roof, which are not properly tensioned. They may be used for small roofs
provided that the boundary makes it possible to provide a high level of pretensioning and
arrangement must be made for adequate drainage. Adequate drainage can be ensured by
having a sloping roof surface or by central drainage in the case of flat horizontal roofs.
ii) Synclastic roof surfaces (2)
It has the same kind of curvature in all directions; that is, concave or convex in any
direction (positive Gaussian curvature roof).
iii) Anticlastic roof surfaces (2)
It has opposite curvature in the principal orthogonal directions; that is, concave in one
direction and convex in the other (negative Gaussian curvature surface). Some times it is
referred to as saddle-shaped, Fig. (2.5). The concave cables are referred to as suspension
cables and the convex ones as pretensioning cable (19).
The two cable sets may be pretensioned against each other to produce the required structural
rigidity without the temporary application of gravity loads (19).
For saddle-shaped net supported by a space ring, Fig. (2.6), the ring must be relatively
flexible to permit redistribution of the forces in the cables. This redistribution is caused by
movements of the supports. These movements will oppose the loss of the tensions in the
pretensioning cables caused by the deflections of the net. The curvature of the net should be
relatively small in order to avoid excessive horizontal displacements of the ring beam (5).
Notes according to the classification of cable nets: (5)

1) To cover large spans, the net must be anticlastic or saddle shaped.
2) To prevent relative movement (slipping) of the cables, they are clamped at their points
of intersection (20), and clamps must exert sufficient pressure on them to prevent
slipping because any clamps movement will reduce the cable tension and
consequently the stiffness.
12
3) It is therefore advisable to arrange the geometry of the suspension and pretension
cables so that the resulting forces along the cables due to the action of clamps are
minimized and this will also reduce size and cost of the clamps.
4) For highly curved nets, the clamping must take place before tensioning the cables, as
it is otherwise extremely difficult to position the cables accurately (20).
5) If the roof net is only of a semi-permanent nature and stiffness is a minor
consideration it is possible to replace the edge members by cables, Fig. (2.5) (20).
6) For large nets, if, architecturally, “basket-formed “ or “flat areas“ can not be avoided,
additional stiffening means are used, internal or external ties, because “ baskets “ will
result in areas where the cables cannot be properly tensioned and “flat area“ will be
subjected to movement which may damage the cladding.
Generally, cable nets have to be designed as two-dimensional system with the spacing of the
cables ranging from 0.5m to 5.0m. The pitch is dictated by the design of the cladding. If the
cladding is supported by a secondary framework consisting of purlins or rectangular frames,
then the spacing of the cables can be greater than if it is resting directly on the cables. The
cost of an intermediate supporting system is, however, usually more costly than adopting a
smaller cable pitch, and should be therefore avoided. For nets with rigid boundaries or these
supported by elastic arches or space rings, sag and rise varying from 3% to 6% of the span
are usually sufficient. For nets with edge cables, the curvature may need to be greater (5).
(5)
d) Pretensioned cable grid structures
Cable grids may be considered as double-layer nets or as multi-directional systems of
intersecting cable beams, Fig. (2.7). Because of the geometry of cable beams, the use of grids
is limited to roofs with either circular or elliptical plan. Cable grids are best suited for self-
balancing structures as will be shown in section (2.2.3.2).
A disadvantage of cable grids:
It has slightly greater difficulty in detailing the ring beams at the points of cable attachments.
The advantages of cable grids are:
1) The beams in grid systems are self-aligning and then the use of the clamps may probably
be omitted or the used clamps exert small forces.
2) The using of longer and more distribution cable units, due to tension ring absence,
reduce the number of cable terminals required.
“It is more beneficial and also less costly to use larger cables rather than to increase the
pretension force. As a general guide, satisfactory stiffness will be achieved if the maximum
sag of the suspension (sagging) cable is of the order from 4-6% of the span, and the cross-
sectional areas of the cables are such that the maximum load in the cables is less than or
equal to half their ultimate strength. Since the stiffness effect of increasing pretension is only
marginal pretension should only be of sufficient magnitude to prevent cables from going
slack under any maximum specified load condition” (5).
2.2.2.2 Classification according to number of layers (2)

a) Single layer systems
They are systems consisting of a simple layer of cables. These systems include simply
suspended cable roofs and cable net roofs.
b) Double layer cable systems
They have cables of one layer directly below corresponding cables of the other layer and
can be utilized for forming anticlastic or synclastic surfaces. Double-layer systems include
both cable beams and cable grids.
13
2.2.2.3 Classification according to the roof dimensions (19)
a) One dimensional system
The system is large in one direction comparing with the other two directions, as in freely
(simply) suspended cables and cable beams either single or multi-span.
These systems are easy to design and erect and can cover spans up to 100m.
b) Two dimensional system
The system is large in two directions comparing with the third direction, as in cable nets.
These systems are difficult in both design and erection comparing with the previous type.
c) Three dimensional system
The system is equally large in the three directions such as cable grid structures. These
systems are very complex in both design and erection.
2.2.3 Classification due the structural behavior of suspended roofs (8)

2.2.3.1 Classification according to the cables pretensioning
a) Not pretensioned cables
Such system is used to cover small area. It is preferable to use rigid cladding like
reinforced concrete to stiffen the structure.
b) Pretensioned cables
Pretensioning of the cables, to stiffen the structure, is the only way to cover large areas.
Pretensioning is made either by preloading concept or by arranging the sets of cables to
pretension each other (counter-stressed systems). Two schools for degree of pretensioning
3
are as follows: ( )
1) A few cable elements could be permitted to become slack under adverse loads, which
expected to occur rarely with condition that the system over all stability would not be
affected.
2) It is not allowable for any cable to become slack under any loading condition.
The first method is economic, especially with knowing that a local buckling in a cable
structure will not spread because the cable structure is a tension system.
2.2.3.2 Classification according to the self balance of the structure (5)

a) Self-balancing structure
It is the structure supporting the cables, which its geometry permits the forces in cables to
be balanced internally as shown in Figures (2.1b, 2.1c, 2.3, 2.6 and 2.7).
b) Non self-balancing structure
It is the structure supporting the cables, which its geometry is unable to resist the cable
forces without the aid of ground anchors as shown in Figures (2.1a, 2.2, 2.4 and 2.5).
2.2.3.3 Classification due to the mathematically treatment method (2)

a) Discrete-approach system
In the discrete approach, the structure is represented by an assemblage of straight
elements interconnected at a finite number of nodal points at which the loading is assumed
concentrated as shown in Fig. (2.8a).
b) Continuous–Approach system
In this approach, the real structure is represented by curved segments, Fig. (2.8b).
Physically, this situation can exist only if loads are distributed along the cable.
So cladding effect is analytically considered.
14
2.3 Main Elements of cable roofs and their material (3)
Cable roof structures may be divided into two parts as follows:
2.3.1 The supporting structure :
It includes anchors, columns, rings or frames (21). It may be plain, reinforced, or prestressed
concrete, steel or a combination of the two.
Aesthetic and structural considerations and economics govern the material choice.
2.3.2 The roof cladding
Roof cladding may be divided into six parts, Fig. (2.9):
a) Cables; b) Fittings; c) Auxiliary framework;
d) Decking (cladding); e) Insulating layer and f) Roofing.
a) Cables
Cables are the main load-carrying elements in cable structures as shown in chapter 1.
b) Fittings
Fittings are the attachments used to grip the cable at the ends or along its length to facilitate
the connections of the cables to anchorages and to the structure (5).
c) Auxiliary framework
Auxiliary framework supports the decking and is placed over the cables. It is required only if
the decking cannot span the cable spacing. Using of these frames considerably adds to the
cost of the roof and can be avoided if a suitable cable pitch is adopted.
d) Decking (cladding)
The function of decking is to transmit loads to the supporting cable system. The cladding
may consist of a variety of materials such as corrugated sheets of steel or aluminum,
aluminum sandwich panels (21), concrete, timber, gypsum, or rigid insulation boards.
e) Insulating layer
Its function is to prevent the heat transfer through the roof, as well as the walls, to maintain
comfortable temperature inside the building and to economize air-conditioning and heating
costs. The basic material in most cases is a mineral wool, such as rock glass or slag wool;
vegetable or wood fiber; animal hair, etc. The varieties available are numerous.
f) Roofing
It is the external water proofing skin. It may be made of sheet metal, glass, plastics, or
similar materials. The sheet metal used for roofing includes steel, aluminum, stainless steel,
copper, etc. The choice of the roofing depends upon economics and functional requirements.
2.4 Loads
The structural system must be designed to resist both static and dynamic loads.
2.4.1 Static loads (3) Static loads include:
I) Dead load: It includes cladding weight, insulating material, cables and fittings...etc.
It ranges (25-75) kg/m2 for light cladding and (75-150) kg/m2 for heavy cladding and of
necessity; these values have to be higher for simply suspended systems.
II) Live load: Cable roof is, generally, inaccessible to people except for maintenance
purposes since it has a large span and has a curved surface. Therefore, it may be quite
justified to design the cladding units for the normal live load, but a lighter design live load
may be considered in designing the cable systems and the supporting structures.
III) Snow load: It should be taken into account in the areas where the snowfall occurs. Also
ponding of melted snow or rainwater should be treated as an additional load.
15
IV) Pretensioning load: It should be considered as dead load (because the dead load cable
forces, cannot, theoretically, be computed without accounting for pretension) and also
included in the intermediate erection load condition (2).
V) Erection load: (22) The forces transmitted from cables to other parts of the structure are
quite high even during intermediate construction stages, so they may be the critical loading
on a portion of the structure and should be considered by the designer.
VI) Creep and temperature effects: (3) The loss of pretension due to creep will be less
seriously felt using higher-tension steels. Increasing the temperature will result in relaxing
the cable tension and thus reducing the stiffness of the system.
2.4.2 Dynamic loads

Dynamic loads on cable systems include wind and / or earthquake loads and
sometimes blast loading. The design of cable roof system is almost governed by wind forces.
Generally, dynamic analysis is recommended to check for dynamic safety.
2.4.3 Loads combination (3)

The cable is designed so that cross-sectional areas may be adequate to sustain the
worst combination of loads. The stipulations of the AISI, Manual for structural applications
of steel cables for buildings, 1973, regarding the choice of load combinations as follows:
The effective design ultimate strength of the cable must be equal to or greater than the
largest value produced by the following cable tension conditions:
(a) 2.2 T1 (b) 1.6 T1+2.7 T2 (c) 2.2 T3
(d) 2.0 T4 (e) 2.0 T5 (f) 2.0 T6
Where
T1 = net tension in cable due to dead load and pretension,
T2 = change in cable tension due to application of live load,
T3 = net tension in cable due to dead load, pretension and live load,
T4 = net tension in cable due to dead load, pretension, live load, and wind or earthquake,
T5 =net tension in cable due to erection of structure,
T6 = net tension in cable due to dead load, pretension, and wind.
Temperature should be taken into account with T1. In calculating the above T values, dead
and live loads are added to or subtracted from the pretension in suspension and pretensioned
cables respectively. The same is done for wind load according to its type (pressure / suction).
2.5 Wind (nature and response)

Wind is a phenomenon of movement of air particles and results from atmospheric
pressure differences over the earth surface, caused mainly by the differential radiation of the
sun and in second-rate due to rotation and curvature of the earth. Away from the ground, the
pressure system is a relatively stationary. Near the ground, the balance of the pressure system
is disturbed by drag forces caused by the ground’s surface roughness, which occurs naturally
(mountains, hills…etc.) or by man-made (dams, buildings…etc.).
Ground surface roughness causes mechanical stirring of the air movements that: (23)
 The wind speed near the surface is retarded.
 The wind direction changes and is no longer parallel to the isobars that are the lines of
equal barometric pressure.
 The flow conditions become unsteady and the wind exhibits instantaneous random
variations in magnitude and direction.
16
The rougher the surface, the more prominent these effects are. The effects decrease with
increasing in height above the ground. The height at which the effects have virtually
vanished is referred to as “the gradient height” and ranges from 300m to 600m depending
on the degree of the surface roughness.
Although wind is unsteady and exhibits random fluctuations in both time and space domains,
it posses stationary characteristics. So, it is possible to describe its functions in statistical
terms. Research has revealed hat the long-term statistical properties of wind are general and
independent of type of terrain, wind strength and site location.
For linear analysis of structures, the statistical characteristics of the response are determined
in terms of the statistical properties of wind which is called “frequency domain analysis”.
For non-linear analysis of structures such as in cable roofs whose structural properties vary
with the amplitude of response and hence with time the structural properties may be updated
at the end of each time step or may be unchanged during the time step. This analysis form is
called “time domain analysis” and required generating of wind histories. Wind histories and
wind fields, simulating real time, are generated from spectral density functions of the
fluctuating component of wind speed either by superposition of harmonic waves or by
filtering sequences of white noise.
2.5.1 Wind speed

It is convenient to express the wind velocity as the sum of the mean velocity U ( z, x) in
the long-wind direction at height z and the fluctuating time-dependent velocity components
u( z, x, t ), u( z, y, t ) and u( z, z, t ) where x represents the along-wind, y the horizontal across-
wind and z the vertical across-wind directions at height Z . Hence
V ( z, x, t )  U ( z, x) u ( z, x, t ) 
V ( z, y, t )   0   u ( z, y, t ) (2.1)
     
V ( z, z, t )   0   u ( z, z, t ) 
In cases where the horizontal and vertical crosswind fluctuations are of secondary
importance, the instantaneous wind velocity can be treated as a scalar quantity, in which
case, omitting the direction indicator, the instantaneous velocity at height is given by:
V ( z, , t )  U ( z )  u ( z, t ) (2.2)
Where: U (z )  The mean wind velocity component, and
u( z, t )  The fluctuating wind speed component.
As a result, the response of structures to wind can be divided into two parts:
1) The quasi-static response caused by the mean wind velocity component, and
2) The response due to fluctuation wind velocity that is the source of dynamic excitation.
1) Mean wind speed

Wind speed increases with increasing altitude above the ground until it reaches the
velocity at the gradient height. Two main laws have been proposed to describe the way in
which the mean wind varies with height.
The first law is the so-called “power law” and is given by:
z
U ( z )  V g ( ) (2.3)
zg
Where: Vg  The gradient wind velocity, which is given as soon as the velocity at 10m
height for a place is prescribed (for example, by codes or experiments);
17
z g  The gradient height, which is a constant for a given place and a function
of the ground roughness; and
 Constant and is a function of the ground roughness.
Experimental values for  and z g are shown in Table (2.1).
(5)
The most generally law describing the way in which the mean velocity varies with height is
the “logarithmic law” and is given by:
U ( z )  2.5 u* ln( z / z 0 ) (2.4)
U (10)
Where u*  (2.5)
2.5 ln(10 / z 0 )
Or u*  U (10) k (2.6)
Where: u*  The shear velocity or friction wind velocity,
z 0  Roughness length, k  Surface drag coefficients and
U (10)  The reference-mean wind at 10m height.
Table (2.2) gives values of both z 0 and k (23).
If the surface drag coefficient k is known, then u* is known using Eqn (2.6) and the
corresponding value of z 0 can be expressed using Eqn (2.4) as:
z 0  z e U ( z ) / 2.5 u* (2.7)
The logarithmic law is applicable for heights  10m. Below this height the velocity is
assumed to be constant and equal to U (10) .
In some of the more recent codes the logarithmic law has been modified to be:
 
U ( z)  2.5u* ln( z / z 0 )  5.75( z / H g ) (2.8)
Where H g , the gradient height, can be determined from
H g  u* / 2 sin  (2.9)
Where:  is constant = 6,
  The angular rotation of the earth = 7.2722  10 5 rad / s , and
  The local angle of latitude.
Using Eqs (2.8) and (2.9) and the above values for  and  , it yields that:
U (10)  0.1254454 sin 
u* (2.10)
2.5 ln(10 / z 0 )
As  1  sin   1 , Eqn (2.10) can for most applications be simplified to the same form of
U (10)
Eqn (2.5), i.e. u * 
2.5 ln(10 / z 0 )
2) Fluctuating wind speed

Wind speed can be considered as a stationary random process. Thus, for any space point:
T
1
T 0
U (t ) dt  U (2.11)
T
1
T 0
u (t ) dt  0 (2.12)
18
Where, T varies from 10 minutes to 1 hour. Because of this, the characteristics of the
fluctuating component of wind can be expressed by statistical functions. One of these
statistical functions, which will be used later, is the spectral density function.
Spectral density function (power spectra) of longitudinal velocity fluctuations S u (n)

The expression for the power spectrum of the longitudinal fluctuations is mainly based
upon experimental data. The most important formulations are (23) as follows:
4u*2 f 2 1200n
Davenport: S u ( n)  , f  (2.13)
n(1  f ) 2 43
U (10)
Where n is the frequency of the harmonic (sinusoidal) velocity components of the fluctuating
part of the wind velocity.
4u*2 f 2 1800n
Harris: S u ( n)  , f  (2.14)
n( 2  f )2 56
U (10)
The above two expressions are independent of the height z. A convenient formulation that
varies with height is that suggested by Kamial (23) and given by:
200u*2 f ( z, n) zn
S u ( z , n)  , f ( z , n)  (2.15)
n[1  50 f ( z, n)] 53
U ( z)
This function approximates, in the higher frequency range, very closely to spectra of
recorded wind histories while it is suspect in the lower frequency range.
Another spectral density function that also varies with height is based on the current ESDU
(European Statistical Data Unit) model and yields much lower values for the power spectral
density function in the lower frequency range than do the other three spectra and is given by:
U (10) 4.4 (1  S top hs ) 2.66
S u ( z , n)  (2.16)
2500 z 10 25n1 3U ( z ) 2
Where S top is a topographic factor and  hs is the hill slope.
Generally, S u ( z, n) obtained by Eqs (2.15) and (2.16) decrease with increasing height.
Cross-spectral density and coherence functions S u j vk (n)

It is a measure of the degree to which the two histories u (t ) and v(t ) , recorded at stations
j and k respectively, are correlated in the frequency domain. It is given by:
S u j vk (n)  S ( j, k , n)  e  S u j (n)  S vk (n) (2.17)
Where
2n C x2 ( x j  xk ) 2  C y2 ( y j  y k ) 2  C z2 ( z j  z k ) 2
 (2.18)
U (z j )  U (zk )
Where originally from wind tunnel measurements (24) C x  C y  16 and C z  10 .
The value e  is known as the coherence function cohu j vk (n) and decreases with:
1) Increasing distance between the two points,
2) Increasing frequencies, and
3) Decreasing wind speeds.
2.5.2 Wind pressure

As mentioned before in Eqn (2.2), the velocity is expressed as:
V ( x, y, z, t )  U ( z)  u( x, y, z, t ) (2.19)
19
The total wind pressure is given by:
1
p( x, y, z, t )   C p ( x, y, z, t )(V ( x, y, z, t )  x (t )) 2 (2.20a)
2
1
Or p(t )   C p (V (t )  x ) 2 (2.20b)
2
Where: p( x, y, z, t )  Pressure per unit surface area,
  Mass density of the air  1.226 kg / m 3 ,
C p ( x, y, z, t )  Pressure coefficient for a given point, and
x (t )  The velocity of the structure at the same point and in the direction of the wind.
Coefficient C p ( x, y, z, t ) is probably time-dependent, but it is customary to neglect its

dependence in time and be approximated by static wind test.
Davenport (25) expressed a more exact expression for wind pressure than Eqn (2.20),
introducing the effect of the mass of surrounding air vibrating with the structure, as:
1 A dV (t )
p(t )   C pV 2 (t )  C m  0 (2.21)
2 D dt
Where: C m  The coefficient of the additional mass. It is dependent of the shape and
the vibrating mode of the structure and varies from 2.5 to 7.5 (23) and may be taken as 8/3 (26);
A0  The reference area for a virtual mass; and
D  The diameter of the object.
The surrounding air influences the frequencies of the vibrating roofs and can be considered
as an additional vibrating mass of the system, which could be calculated approximately as:
m  C m  a 3 (2.22)
Where: a  The dimension (radius) of the structure.
2.5.3 Wind loading

The integration of the pressure p(t ) over of the surface (area) of a structure or
structural element will yield the resultant exerted by the wind. The force components
Pp , Pd , Pl due to the pressure, drag and lift forces respectively are expressed as:
1
Pp   Ap C p (V (t )  x ) 2 (2.23)
d
2 d d
l l l
Where: Pd  The force component parallel to the along-wind direction,

Pl  The force component perpendicular to the along-wind direction (up  ),
Pp  The force component perpendicular to the along-wind direction (down  ),
Ad  The area projected on to a plane perpendicular to the wind direction by a unit
length of structure or structural element,
Al & Ap  The area projected on to a plane in the along-wind direction by a unit
length of structure or structural element, and
C p , Cd , and Cl  Pressure, drag and lift coefficients respectively.
A general expression for the wind force component in any direction i , [ i  1,2,3 ( x, y, z ) ] is:
20
1 ~ ~ ~
 (C pi  C di  Cli sin  t )(V  x ) 2
Pi  (2.24a)
2
1 ~ ~ ~
Or Pi   (C pi  C di  Cli sin  t )(V 2  (2V  x ) x ) (2.24b)
2
~ ~ ~
The aerodynamic coefficients C pi , Cdi , and Cli for an inclined member with wind along the
x-axis, i  1 , can be shown to be:
~
C pi  C p Ap cos  pi (2.25a)
~
Cdi  Cd Ad (2.25b)
~
Cli  Cl Al cos  li (2.25c)
Where: cos  pi is the direction cosine of the normal to the area relative to the i axis,
cos  li is the direction cosine of the lift force due to vortex shedding relative to the i axis.
  Angular frequency of vector-shedding (affecting certain structures such as chimneys,
towers, electrical transmission lines and suspended pipelines) and is given by:
2SU
 (2.26)
D
Where: S = Strouhal number and D = Diameter of the circular cylinder.
Note that: Pli  0 (at i  1 ) & Pdi  0 (at i  2,3 ) (2.27)
Using Eqn (2.24b), the aerodynamic damping is proportional to (2V  x ) x and the
equivalent viscous aerodynamic damping coefficient at any time is given by:
1 ~ ~ ~
cdi   (C pi  C di  Cli sin  t )(2V  x ) (2.28)
2
The static wind force is given, ignoring the structure movements, by replacing the total wind
velocity with the mean wind velocity in Eqs (2.23) or (2.24) as follows:
1 ~ ~ ~
Pswi   (C pi  C di  Cli sin  t )U 2 ( z ) (2.29)
2
This static wind force will cause static response.
The dynamic wind force is given by replacing the total wind velocity with the fluctuating
part in either Eqn (2.23) or Eqn (2.24), taking into account that:
V 2 ( x, y, z, t )  U 2 ( z)  2U ( z)u( x, y, z, t )  u 2 ( x, y, z, t ) (2.30)
1 ~ ~ ~
It yields to: Pdwi   (C pi  C di  Cli sin  t )[u 2 ( x, y, z, t )  2U ( z ).u ( x, y, z, t )] (2.31)
2
Notice: For roof structures pressure coefficients are used while both drag and lift
coefficients are only used for isolated braces, back-stays, etc. (5)
2.5.4 Distribution and evaluation of static wind loads on cable roofs

The distribution of static wind load on cable roofs is assumed to be as Fig. (2.10).
The Egyptian code (27) uses formula for evaluating the exterior wind pressure (downward  )
or suction (upward  ) on the roof as follows:
pe  ce k q (2.32)
Where: pe  The static wind pressure or suction per unit area of the roof,
q  The principal wind pressure and varies with the geographical location, Table (2.3),
k  Exposing factor with values differ with the height, Table (2.4),
ce  Distribution factor and may be taken as 0.5.
21
2.6 Methods of analysis
Two main approaches represent cable roofs, continuous and discrete approaches.
2.6.1 Continuous approach (analogous membrane solutions)
It simplifies the analysis by replacing the actual cable system by an equivalent thin
membrane. In the continuous approach simultaneous ordinary nonlinear differential or partial
differential equations are used to represent the real structure (2). Many approximations are
usually introduced in the derivation of these equations. Analytical and numerical methods are
used for the solution of these equations (5). It is assumed that the cable segments are curved,
Fig. (2.8b). Physically, this situation can exist only if loads are distributed along the cable.
Concentrated loads result in discontinuities of the slopes of the curves (2).
Since, the exact solution of membrane may differ significantly from the exact solution of
cable network; the membrane approach can fulfill the need for an approximate method in the
preliminary design of cable roofs (28). It is mentionable that membrane analogy should be
used to get the solution of free vibrations if all joints become node points and nodes are
unsymmetrical. In this case, frequencies are determined from the frequencies of a cable (29).
Although, in membrane analogy, the discrete structure is first represented by a continuum in
order to derive the governing equations, once the equations are obtained it then becomes
necessary to replace the continuous equations by a discrete form to obtain a solution (2).
Previous work
Leonard (1) presented static and dynamic behavior of membranes. Cable net is
approximated by an equivalent continuous flat membrane with modified material properties
and the finite element method is used to analyze this equivalent structure.
Sigh and Dhoopar (29) presented a general membrane analogy for both isotropic and
anisotorpic flat cable networks either orthogonal or non-orthogonal.
2.6.2 Discrete approach

“ In the discrete approach the real structure is represented by an assemblage of elements
interconnected at a finite numbers of nodal points at which the loading is assumed
concentrated, Fig. (2.8a). At each node equilibrium of forces and compatibility of
displacements must be satisfied. Mathematically, this approach may be represented by a set
of difference equations. Such equations are derived by application of the basic laws of
structural mechanics, generally, the virtual work Hamilton’s variational principles. When
expanded these yield a set of simultaneous algebraic equations with finite number of
unknowns and may be compared to the finite element method. Alternatively, the set of
algebraic equations may be formed directly by considering equilibrium at each node in terms
of the locations of adjacent nodes. For complex structures the set of algebraic equations may
be quite large” (2).
This set of nonlinear equations is to determine the unknown joint displacements. Discrete
approach is a general approach and preferred for the following advantages:
I) It is more accurate than continuous approach;
II) It enable actual representation of the real structure;
III) It accounts fully and accurately the nonlinear effects; and
IV) It is adaptable for computer analysis.
Generally, discrete analysis is divided into two main methods, displacement (stiffness)
methods and energy methods.
These methods are, generally, used for cases of cable slackening, cable rupture, cable
elements with nonlinear modulus of elasticity, temperature change, struts buckling, laterally
loaded cable elements, and for the flexibility of the supporting structure.
22
Review for techniques of discrete approach
Several techniques for static and dynamic analysis of cable structures are used.
Although energy and stiffness methods are the two main techniques, A.S.K. Kwan (30)
represented a brief outline of these techniques that, due to his vision, include:
1. Stiffness matrix method;
2. Finite element method;
3. Minimizing of total potential energy method;
4. Dynamic relaxation method; and
5. Approximate linear approach.
2.6.2.1 Stiffness method

“One of the two methods redundant structural analysis, flexibility (force or
compatibility) and stiffness (displacement or equilibrium) methods, the stiffness method has
proved to be more useful in the analysis of highly redundant cable systems. In stiffness
method, the fundamental unknowns of the equations formulated are the displacements of the
nodal points connecting segments” (1). Because of its advantages, the finite element method
of stiffness analysis supplanted the “lumped parameter method” (in which the effects are
replaced by forces in equivalent straight weightless springs, which may be nonlinear, and a
system of simultaneous algebraic equations is generated from the nodal equilibrium
conditions which are focused on the static equilibrium of nodes). Stiffness method is a linear
method and generally expressed as:
K d   P (2.33)
Where: [K] = Square stiffness matrix of the structure;
[d] = Vector of unknowns (displacements at joints); and
[p] = Vector of applied forces.
Previous work
Foster (4) found that initial pretension of cables had a minimal effect on final deflected
position as far as all cables not being pretensioned to exactly the same value.
Urelius and Fowler (6) used a modified form of the direct stiffness method to
investigate the geometrically nonlinear behavior of pretensioned cable truss structures.
Greenberg (8) presented an inelastic method for the analysis of suspension roofs.
Gaussian elimination, which is direct solution technique, is used to improve the convergence.
El-Ibiary, Shadia (22) developed a technique for solving nonlinear cable structures with
general edge conditions, using the truss direct element matrix.
Saafan (31 and 32) presented the finite deflection theory for analysis of spatial systems
of nets on the basis of the stiffness (equilibrium or displacement) method.
Jonatowski and Birnstiel (33) developed a procedure to determine the elasto-plastic

behavior of three-dimensional stiffened and unstiffened suspension structures.
Mollmann (34) presented an analysis of elastic cable beams composed of simple

tension members, simple compression members and shallow cable members.
Baron and others (35) presented an analysis method of geometrically nonlinear

suspension cables and three dimensional truss systems using tangent stiffness matrix.
23
C.G.Hood (36) presented a general stiffness method for the analysis of nonlinear cable
networks with distributed loading using Newton-Raphson technique.
Valente (37) presented a method of analysis of hanging roofs with stiffening beams in
the second order theory based on the deflection method. The deflection-load law is nonlinear
and the deflection vector is found by the Newton-Raphson method.
Godbole et al. (38) presented a procedure to study the support flexibility against the
performance of suspended cable roofs employing the stiffness method.
Abu-Hamad and Aly (39) developed a technique using modified form of the direct
stiffness method to generalize the solution of pretensioned cable roofs.
Kassimali et al. (40) studied the nonlinear behavior of plane counterstressed cable
trusses and a hyperbolic network using the tangent stiffness method.
Desai et al. (41) presented a new stiffness procedure for a three-node, parabolic cable
element, which is compatible with traditional structural finite elements, for the static
nonlinear analysis of three-dimensional cable-supported structures.
Swaddiwudhipong et al. (42) developed computer software package using the tension
coefficient method. Nonlinearity is attended by Newton-Raphson method.
Saafan and El–Rakabawy (43) investigated the behavior of the Olympic stadium roof
recently built in Cairo by the finite deflection theory of suspended roofs.
Shan et al. (44) presented an approach for the analysis of frame cable structures which
are pretensioned cable nets incorporating space frames.
El-Sadawy (45) developed a new technique using a modified form of the direct
stiffness method to generalize the solution of space cable roofs.
Eisenloffel and Adeli (46) present a micro computer-aided analysis of tensile networks.
The nonlinear equations system is solved using Newton-Raphson method.
Krishna (3) gives a complete stiffness method approach solved iteratively by the
Newton–Raphson method. For a pretensioned cable net in equilibrium, Eqn (2.33) is
corrected for geometric nonlinearity effects as follows:
K  d   P  R(d ) (2.34)
Where [K], [d] and [P] are as shown before, and [R (d)] is the vector of residual forces,
which is nonlinear function of (d), as shown in Fig. (2.11). Since the right-hand side contains
the term R, which is a function of the unknowns, an iterative numerical solution has therefore
to be adopted. The equilibrium is reached when R is a minimum.
2.6.2.2 Finite element method (FEM)

While the geometric stiffness matrix method or minimizing the total potential energy
method is strictly speaking a form of finite element method (FEM) in which each cable
segment is represented by a single element, some researches have applied a more general
FEM approach (30). A finite element solution can be roughly divided into four processes, each
having its own source of potential error in the modeling of the true continuum behavior (1):
24
1) Idealization of the continuum
The body is subdivided into an adequate number of finite elements. The number, shape, and
size of the elements should be determined in such a way that original body geometry is
simulated as closely as possible.
2) Discretization of displacements
Simple functions are selected to approximate the actual behavior at an arbitrary point within
an element.
3) Formulation of element stiffness
The stiffness matrix relates the nodal displacements to the applied forces transformed to the
nodes. This relationship is developed from a discrete form of the principle of virtual work,
which governs in the continuum. The sum of approximate energies in each element is
assumed equal to the total energy of the real continuum. If this discrete sum of energies is
minimized with respect to the nodal displacements, then overall equilibrium of the nodal is
guaranteed and the nodal displacements calculated are the best answers available for the
given approximating functions.
4) Solution of matrix equations
After assembly of stiffness and force matrices of each element into matrices for the entire
continuum, the final step is to apply discrete forms of the boundary conditions and solve the
matrix equilibrium equations (or equations of motion) for the nodal displacement as a
function of load (or time).
Previous work
Peyrot and Goulois (47) presented a general procedure for the static analysis of
structures include cable elements. The analysis procedure uses a single cable element
subprogram, which computes the element end forces, its geometry and local tangent
stiffness. A nonlinear FE procedure is used to assemble the cable elements with the rest of
the structures and solve for the node displacements.
Tabarrok and Qin (48) described a nonlinear FEM for form finding and load analysis of
tension structures. The complete procedure for both form finding and load analysis functions
is implemented in a program.
Mitsugi (49) formulated a stiffness matrix for the “hyper-cable” element (which is a
cable connected to intermediate pulleys along its length and also known as “active cable”) to
be incorporated into FEM solver.
2.6.2.3 Minimizing of total potential energy (TPE) method

The approach of minimizing the total potential energy (TPE) has been described by
many researchers. The theoretical approach is essentially the same in each case, but the
choice of the minimizing algorithm differs. This method is a unifying approach to the
analysis of structures of either linear or nonlinear behavior by considering the determination
of equilibrium as an iterative process of minimizing the total potential energy, the position of
equilibrium being reached when the TPE is minimum.
The total potential energy (TPE) can be expressed as:
Wt  WS  WP  (WS f  WS P )  WP (2.35)
M
WP  P T x And W S   wS (2.36)
25
Where:
WP = The potential energy of the applied loads;
P T = Transpose of the vector of the applied loads;
x = Displacement vector corresponding to all degrees of freedom in the structure;
WS f = Strain (elastic) energy stored in the flexural members (beams and columns);
WS P = Strain (elastic) energy stored in the pin-jointed members and cables;
M = Total number of cable elements; and
wS = Strain (elastic) energy stored in a particular element.
Mathematically, equilibrium condition in direction i at joint j may be expressed as:
W
0 [ i =1,2 and 3] (2.37)
x ji
Where x ji  the displacement of joint j in the direction i .
The method of minimizing TPE is fully illustrated in chapter 3.
Previous work
Buchholdt (5 and 50) put a complete theory of static and dynamic analysis for both
linear and nonlinear structures using three methods of minimizing the TPE, steepest descent
method, conjugate gradient method and Newton-Raphson method.
Murray and Willems (51) presented the analysis using two minimization techniques:
(1) the variable metric method and (2) the conjugate gradient method.
Buchholdt and Bhatlacharya (52) used the conjugate gradient method of minimizing
TPE for the analysis of cable structures with flexible boundaries and laterally loaded cables.
Monforton and El-Hakim (53) use the conjugate gradient method to minimizing the
TPE for analysis of general pin-ended truss and cable structures.
Fried (54) presented a high order nonlinear, extensible, cable finite element, which is
created by Gauss integration of the element TPE.
Stefanou et al. (55) used the conjugate gradient method of minimizing TPE to calculate
response of cable structures to static loads.
Stefanou and Nejad (56) used the conjugate gradient method of minimizing TPE to
analyze cable assemblies with fixed and flexible elastic boundaries.
Allam (21) presented the analysis of cladding-net work interaction in pretensioned

cable roof structures under static loads using minimizing of TPE method.
2.6.2.4 Dynamic relaxation method (30)

It is an attractive approach to the design of pretensioned cable networks because both
form finding and analysis can be carried out within a single analytical framework. The
technique traces the motion of the structure from the point of loading to the final steady
equilibrium state through the use of the D‘Alembert principle.
26
P(t )  Mx  Cx  Kx (2.38)
Where: P(t ) = time dependent vector of externally applied load;
M = Fictitious mass matrix;
C = Damping coefficients matrix;
K = Stiffness matrix; and
x = The vector of displacements.
The first two terms on the right-hand side of Eqn (2.38), ( Mx  Cx ), represent the residual,
or out of balance force, which is iteratively minimized to an acceptably low value, at which
point the structure reaches steady equilibrium. Iteration is carried out in small time steps
using a finite difference approach to find values for changes in nodal displacements.
The speed and stability of the whole process is critically governed by:
(1) The fictitious mass assigned to the nodes;
(2) The choice of damping coefficients; and
(3) The size of the time step.
Briefly, dynamic relaxation method may be considered as a technique of minimizing the total
dynamic potential energy TDPE of the structure.
Previous work
Eman (28) represents an energy approach using the dynamic relaxation technique to
obtain analytical solution for highly redundant nonlinear structures.
Lewis (57) investigated the relative efficiencies of the stiffness matrix and dynamic
relaxation methods in solving specific structural examples. The dynamic relaxation method
shows higher efficiency in term of the total CPU time required for convergence. The kinetic
damping routine gives real time saving over the viscous damping technique.
Lewis and Shan (58) presented a numerical approach for the analysis of cladding-
network interaction in pretensioned cable roof structures under static loads. The approach is
based on the sub-region mixed energy principle. The governing equilibrium and
compatibility are solved using the dynamic relaxation technique with kinetic damping.
2.6.2.5 Approximate linear methods

Pellegrino (30) proposed an approximate linear approach to geometric nonlinearity by
observing that a general load on a cable net can be decomposed into two parts, i.e. one part
which causes extensional displacements, and the second which excites inextensional
displacements (infinitesimal mechanisms).
Geometric nonlinearity arises from the latter which Pellegrino verified, by experiments, to be
small for pretensioned cable nets and hence a linear approximation that he proposed to
determine the magnitude of the second type of displacements was sufficiently good practical
purpose.
27
2.7 Advantages and disadvantages of cable systems (3)
2.7.1 Advantages of cable roofs
1) Cable roofs utilize the material of the structural member more efficiency than any other
systems, as all structural elements in suspension roofs are subjected to direct tension only.
2) Cable roofs represent the economic solution for large span covering comparing with
any other systems using rigid elements as structural elements.
3) Cable roofs are highly intermediate structures and being made from steel and as such
ductile members so they offer considerable overload capacity beyond the working load.
The ultimate load is as much as 60% to 170% higher than the nominal load.
4) Cable roofs eliminate problems of instability and the resulting reduction in allowable
stress.
5) Cable roofs are very useful and effective for temporary works. That is because they are
lightweight, collapsible, relocatable, and easy to transport and erect with low installation
cost (28).
6) Cable roofs may give any required form or slope for the roof.
7) In constructing cable roofs, little or no framework is necessary.
8) Construction and erection time is considerably less than many other forms and that
saves money.
9) If there is a local buckling, it will not be spread because of its being a tension system.
10) There is no problem, comparing with shells or latticed domes, of supporting
concentrated loads in cable roofs.
11) The fire proofing qualities need not always be a dominant factor since such roofs will
usually be quite high.
12) The long-span roofs have functional flexibility, which allows varying uses of the space
enclosed.
13) Cable roofs can with suitable design be put to repetitive use most efficiency.
2.7.2 Disadvantages of cable roofs

1) The analysis, design and details of cable roofs need a tough work.
2) The initial shape of the surface and coordinates of its main joints must be specified at
first.
3) Pretensioning of cable need a great care of construction especially when cables
pretension each other.
4) Cable roofs are sensitive to temperature changes since higher temperature causes the
cable material to creep and an overall decrease in the stiffness of the system hence large
deformations may occur.
5) Drainage systems must be studied carefully, especially for concave surfaces.
6) The safety of the roof must be ensured through conservative design.
7) Cable roofs need a maintenance checking programs through its lifetime.
8) The functional flexibility of the large-span roofs cannot be assumed for future
expansions.
28
Table (2.1): Experimental values for  and z g
Types of terrain Open terrain Suburban terrain Centers of large cities
 0.16 0.28 0.40
z g (meter) 275 400 520
Table (2.2):
Roughness length z 0 and the surface drag coefficients k for various types of terrain
Types of terrain z 0 (m) k  10 3
Sand 0.0001-0.01 1.2-1.9
Sea surface 0.005 0.7-2.6
Low grass 0.01-0.04 3.4-5.2
High grass 0.04-0.10 5.2-7.6
Pine forest 0.90-1.00 28.0-30.0
Suburban areas 0.20-0.40 10.5-15.4
Centers of towns 0.35-0.45 14.2-16.6
Centers of large cities 0.60-0.80 20.2-25.1
Table (2.3): The principal wind pressure, q, in (kg/m2):

The location q (kg/m2)
Marsa Matrouh 90
Alexandria - Hourghada 80
Cairo 70
Fayoum-Luxur-Aswan 60
Tanta-Mansoura-Damenhour 50
Table (2.4): The exposing factor, k:

Height in (m) k
0-10 1.0
10-20 1.1
20-30 1.3
30-50 1.5
50-80 1.7
80-120 1.9
120-160 2.1
More than 160m 2.3
29
(a) Cables in parallel planes.
(b) Cables suspended in radial planes between an inner tension ring and an outer compression ring.
(c) The central cables in parallel planes and the cables at each end suspended between an inner split tension
ring and an outer curved beam.
Fig. (2.1) Simply suspended roofs.
30
(a) Convex cable beam structure with corrugated metal roof decking.
(b) Concave cable beam structure with corrugated metal roof decking.
(c) Convex-concave cable beam structure with corrugated metal roof decking.
Fig. (2.2): Cable beam structure with corrugated metal roof decking.
31
(a) Convex.
(b) Concave.
(c) Convex-concave.
Fig. (2.3) Radial cable beam structure with inner tension ring and outer compression ring/rings.
32
Fig. (2.4) Multi-span cable beam structures.
(a) Convex, (b) Concave and (c) Convex-Concave.
Fig. (2.5): Saddle-shaped cable net with edge cables.
33
Fig. (2.6): (a) Saddle-shaped net with edge beams. (b) Saddle-shaped net supported between two vertically
erected arches. (c) Saddle-shaped cable net supported by a space ring.
34
Fig. (2.7): (a) Circular convex cable grid supported by a horizontal ring beam.
(b) Circular concave cable grid supported by two ring beams.
(c) Circular convex-concave cable grid supported by two horizontal ring beams.
(a) Discrete. (b) Continuous.
Fig. (2.8): Discrete and continuous representation of suspension structures.
35
Fig. (2.9): Roof cladding.
Fig. (2.10): Distribution of wind loading on cable roof, [+ Pressure] and [– Suction].
Fig. (2.11): Diagrammatic representation of the stiffness method.
36
Chapter (3)
Static analysis of cable roofs by energy method
3.1 Introduction
Energy method is a unifying approach to both linear and nonlinear analysis of
structures. It is an indirect method of analysis and valid for both small and large structures.
The energy method is applied to the analysis of general pin-ended truss and cable structures.
Both geometric and material nonlinearties are directly incorporated within the formulation,
thereby accounting for large displacement and strains as well as configuration changes due
to the structural response (53).
The solution of a problem is by determining the structure equilibrium as an iterative process
of minimizing the total potential energy (TPE); the position of equilibrium is when the TPE
is a minimum. The total potential energy is obtained by summing the energy contributions
from each member and formulated in terms of the displacements of the nodes or joints of
the structure. Knowing that the stable equilibrium configuration is the global minimum of
the TPE, function minimization techniques are employed to numerically locate the
minimum (51). It is now obvious that, in the energy method, the equilibrium of the system is
expressed indirectly through stationary condition of a scalar function, the energy of the
system in a variational form.
For linear analysis, “indirect methods” are advantageous since they do not require complex
techniques to reduce the computer storage, but they have the drawbacks that:
1) The rate of convergence may be slow;
2) The large amount of calculations may lead to accumulation of round–off errors; and
3) They require a criterion to terminate the procedure.
These are probably the reasons why direct methods are generally used in automatic analysis
of structures.
In case of nonlinear analysis, however, the intrinsic properties of indirect methods can be
better exploited. First, the stiffness matrix of the structure can be updated more frequently
than with commonly used instantaneous stiffness matrix methods and incremental loading
techniques. Thus, divergence problems that some times arise with the latter may be more
easily tackled. Second, the formulation of the terminating criterion is, in the case of
nonlinear problems, required by any solution method (21).
Energy method, by minimizing the TPE, which is as indirect method, has the following
advantages: (51)
(1) The powerful methods of mathematical programming are used;
(2) The divergence problems of finding solutions to nonlinear simultaneous equations are
avoided;
(3) The effort required to formulate the problem is reduced as TPE is simply the scalar sum
of the energy contribution of the individual elements and the change in potential of the
external forces; and
(4) The need for the assembled stiffness matrix of the structure is eliminated and a
significant saving in the computer storage requirement is realized.
In this chapter, the energy method by minimizing the TPE using the technique of conjugate
gradient method is completely represented.
Main sources of knowledge in this chapter are references (5), (20), (50), (55) and (56).
37
3.2 Mathematical formulation
The method of analysis is easily understood by considering a structure with two degrees
of freedom, Fig. (3.1). The total potential energy, W=TPE, of this structure is represented by
contour lines in a topographical map, Fig. (3.2), in which all the points on any one contour
line represent displacements of the free joint for which W is constant. For a three-
dimensional structure having three degrees of freedom, contour lines become surfaces.
The point of minimum W defines the equilibrium position of the loaded structure.
Mathematically, the equilibrium condition in the i direction at joint j may be expressed by:
W
 [ g ji ]  0 , i  1, 2 and 3 (3.1)
x ji
Where: x ji  The displacement of joint j corresponding to a particular degree of
freedom, direction, i , and
g ji  The corresponding gradient of the energy surface.
Mathematicians call the gradient g of the function W “residue”. In structural analysis, the
gradient g represents the force vector required to maintain an assembly in equilibrium in
displacement space x.
The location of minimum W is achieved by moving down the energy surface along a descent
vector v a distance Sv until W is a minimum in that direction, that is, to a point where:
W
0 (3.2)
S
From this point a new descent vector is calculated and the above process is repeated.
The step length S is, strictly speaking, the length when the descent vector v is a unit vector,
along v direction. However, for convenience S will be referred to as the step length even
when v is not a unit vector.
The method is shown diagrammatically in Fig. (3.3) and mathematically by expressing the
displacement vector at the (k+1) th iteration as:
xk 1  xk  S k vk (3.3)
Where: v k = The descent vector at the k th iteration from x k in displacement space, and
S k =The step length determining the distance along v k to the point of Wmin . .
3.3 Minimization techniques

Minimization techniques are the gradient methods (descent methods) for the
determination of descent direction. For any of these techniques, the gradient vector of the
TPE, [ g i ], with respect to all possible displacements, must be calculated at the beginning of
each iteration using Eqn (3.1).
The corresponding Euclidean norm R (the out-of-balance) of [g] is given by:

R  g  g 
T

1/ 2
(3.4)
The equilibrium will be reached when:
 W 
 
g ji   0 Or R  0.0 (3.5)
 x ji 
Generally, a descent method is an iterative process beginning from a starting point x0 and
 
will be ended at the point of equilibrium using a set of descent vectors v ji k .
38
Examples of descent methods used for the analysis of cable suspension roofs are: (56)
1. The method of steepest descent;
2. The methods of relaxed steepest descent;
3. The method of Runge-Kutta;
4. The method of Newton-Raphson;
5. The method of Fletcher-Powell-Davidon (variable metric method); and
6. The method of conjugate gradients.
Descent methods may be classified, according to the way of finding the descent direction v,
into three types as direct, first order and second order methods (21).
(a) Direct methods require no partial derivatives of W .
(b) First order methods require W and its first partial derivative, such as conjugate
gradient method.
(c) Second order methods require W and its first and second partial derivatives, such as
Newton-Raphson method.
The conjugate gradient method (CGM)

This method was originally developed for solution of a system of linear equations
having a symmetric, positive definite matrix of coefficients. In CGM, the descent vector
vk 1 at the beginning of the (k+1) th iteration is expressed in terms of the descent vector in
the previous iteration, v k . A set of direction vectors is said to be conjugate or more
precisely K-conjugate, K being constant if, Fig. (3.4):
vTk1 K vk  0 (3.6)
Or vTk1 g k 1  g k   0 (3.7)
The descent vector vk 1 is therefore determined by a set of linear combinations of

g k 1 and vk . Fletcher and Reeves proved that vk and vk 1 are conjugate if: (56)
vk 1  g k 1   k vk (3.8)
So  is written as: k 
g Tk1 g k 1  g k  (3.9)
vTk g k 1  g k 
The expression of  may be simplified, if the following hypotheses are valid:
(a) The line search along ( v ) is exact; and
(b) The TPE is quadratic.
If hypotheses (a) and (b) are both valid, then  may be written as:
g Tk 1 g k 1 R2k 1
k   (3.10)
g Tk g k R2k
This expression is the most commonly used and is strictly only applicable to linear analysis
of structures (quadratic TPE function) for which the TPE may be written as:
W  12 x K  x  F  x
T T
(3.11)
Also, this equation works well for pretensioned cable assemblies with the function of W is
non-quadratic.
If only hypothesis (a) is valid, then  may be written as:
g Tk 1 g K 1  g K  R2K 1  g TK 1 g K
K   (3.12)
g TK g K R2K
39
This equation is more applicable for assemblies with very low or no pretension such as
simply suspended roofs and guyed masts. Another expression for  to be used for
nonlinear functions is given by Polak and Ribiere as follows:
Rk21 g k 1  g k 
k  (3.13)
Rk2
Both Equations (3.3) and (3.8) provide the basic algorithm for minimizing the objective
function of the TPE by CGM. Previous work shows that convergence is affected by both the
expression for  and the elastic properties of the system and external loads.
Experience has shown that Polak’s equation gives, on the whole, better results (21). However,
it was shown that for analysis assemblies of flexural and cable elements, the expression
proposed by Fletcher and Reeves, Eqn (3.9), is more advantageous. A two-dimensional
example of conjugate descent directions is shown in Fig. (3.5). For linear analysis of
structures the solution is located in at most n steps, while for nonlinear analysis the set of
directions are only approximately conjugate, and the number of steps will be greater than n.
For a complete solution of the problem, (1) the TPE of the structure, (2) the step length,
and (3) the descent vectors have to be calculated at every iteration till the equilibrium.
Comment: “Comparing the methods it would appear that the conjugate gradients meets the
requirements for the analysis of large nonlinear structures. It demands modest computer
storage and converges relatively quickly. Compared with the Newton-Raphson method, the
CGM converges more slowly in terms of iterations, however, its compactness makes it more
suitable for large structures” . (56)
3.4 Expressions for the TPE

The total potential energy of a structure may be written as:
W= U+V (3.14)
Where: U= The elastic or strain energy stored in the structure, and
V= The potential energy of the loading.
The TPE may also be expressed as:
W  U f U p V (3.15)
Where: U f = The strain energy stored in the flexural elements (columns-beams), and
Up = The strain energy stored in pin-jointed members and cables.
In the following it is assumed that:
(a) The cables cannot resist bending.
(b) The tensile forces in the cables are of sufficient magnitude to ensure linear elastic
behavior of the cables between any two nodes.
(c) The axial strains are small compared with the lengths of the members.
(d) The movements of flexural members are sufficiently small for them to behave
linearly.
(e) The cables are of uniform cross sections (21).
Expression for V
V is expressed as:
N
V  F  x    Fn x n (3.16)
n 1
Where: Fn  Element in applied load vector;

xn  Element in displacement vector due to applied load only; and
40
N  Total number of degrees of freedom of all joints.
Expression for Uf
The stiffness matrix in the global system is:
K   T  K L T T (3.17)
Where T = the transformation matrix;
K = the 12  12 matrix of the element in global co-ordinates; and
KL= the local stiffness matrix of the element.
The strain energy stored in the element is given by:
 
f 12 12  
U F  x K  x      x s k sr xr 
1 T 1
(3.18)
2 n 1 s 1 r 1  2 n
Where: f  Number of flexural members;
x s or x r  Element of displacement vector of a flexural member including effect of
the pretension in the cables; and
k sr = Element of stiffness matrix in global co-ordinates of a flexural member.
Expression for UP
 p en

U p   U 0   T0  T de 
 (3.19)

n 1 

o n
Where: U0 = Initial strain energy in a pin-jointed member or cable link due to pretension;
T0 = Initial force in a pin-jointed member or cable link due to pretension;
T = Increment in force in a pin-jointed member or cable link due to applied
loads only; and
e = Elongation of pin-jointed member or cable links due to applied load only;
p = Number of pin-jointed members and cable links.
EA
But T  en (3.20)
L0
Where: E = Modulus of elasticity; A = area of the pin-jointed member or cable link; and
L0= The unstrained initial length of the pin-jointed member or cable link.
p
 EA 2 
 U p   U 0  To e  e  (3.21)
n 1  2 L0  n
Finally
Using Eqs (3.16), (3.18) and (3.21), the TPE of a structure, is expressed as:
f 12 12
1 
p
 EA 2  N
W     x s k sr x r    U o  To e  e    Fn xn (3.22)
n 1 s 1 r 1  2  n n 0  2 Lo  n n1
Further development of Eqn (3.21) requires an expression for the elongation e.
Expression for displacement e
Using Figures (3.7a and b), it can be seen that for the cable segment ab :
3
L20    X ai  X bi 
2
(3.23)
i 1
3
L0  e2    X ai  xai  X bi  xbi 2 (3.24)
i 1
Subtracting Eqn (3.24) from Eqn (3.23) yields:
 
3
2 L0 e  e 2   2 X ai  X bi  xai  xbi    xai  xbi 
2
(3.25)
i 1
41
Since eab  L0 ab , the part e2 may be neglected.
1 3 
 e   X ai  X bi xai  xbi   1 xai  xbi 2  (3.26)
L0 i 1  2 
The above equation gives the elongation of the link due to applied load only.
The elongation due to the original tension (pretension) is expressed as:
L
e0  T0 0 (3.27)
EA
WhereT0 is the pretension in the cable link.
The total elongation of the cable link is, then, expressed by:
1 3  2 L20T0 
eab  e0  e    X  X  x  x  
1
 x  x  
 EA  (3.28)
L0  i 1 
ai bi ai bi ai bi
2  
Expression for the gradient vector [g]
It was shown in Eqn (3.1) that the gradient vector at joint b is expressed by:
 
g i b   W  i =1,2 and 3 (3.29)
 xi  b
Differentiating Eqn (3.22) with respect to xb yields the following expression for the
elements of the gradient vector [g] of the TPE corresponding to point b.
   
g i b   kbr xr n    T0  EA e   en   Fi b
f b 12 pb
(3.30)
n 1 r 1 n 1  L0  n  xi  b
Where f b and p b are the number of flexural members and cable links respectively, which
meet at joint b.
Differentiating Eqn (3.26) to xbi , the displacement in the i th direction at joint b is:
e 1
  X ai  X bi  xai  xbi  (3.31)
xbi L0
e
Substituting for into Eqn (3.30) yields the following b th elements in g:
xbi
fb 12 pb
g bi   k br xr n   t ab  X ai  xai  X bi  xbi   Fbi (3.32)
n 1 r 1 n 1
Where, t ab , the tension coefficient of the force in member ab is expressed as:

 EA  
t ab   To  e  / L0  (3.33)
 L0   ab
The TPE at xk+1 in the displacement space

The expression for total potential energy at xk 1  xk  S k vk is found by substituting
the above expression for x k 1 in Eqn (3.22).
Omitting the iteration suffix for convenience, this yields:
f 12 12
1 
p
 EA 2  N
W     x  Sv s k sr  x  Sv r    U 0  To e  e    Fn x  Sv n
n 1 s 1 r 1  2  n n1  2 Lo  n n1
(3.34)
42
Where, from Eqn (3.28) and since eab <<< L0 ab :
 3  
   X ai  X bi  xai  Svai  xbi  Svbi   xai  Svai  xbi  Svbi 2   o o
1 1 L2 T
eab  
   EA 
L0  i 1 2 
(3.35)
Or 
eab  a1  a2 S  a3 S 2 / L0  (3.36)
3
  T
a1    X ai  X bi    xai  xbi  xai  xbi   L20 0
1
Where (3.37a)
i 1  2  EA
3
a 2    X ai  X bi    xai  xbi vai  vbi  (3.37b)
i 1
3
a3  
1
vai  vbi 2 (3.37c)
i 1 2
Substituting the expression for eab into Eqn (3.34) and carrying out the implied
multiplication yields a polynomial in S of the form:
W  C4 S 4  C3 S 3  C2 S 2  C1 S  C0 (3.38)
p
Where C 4   ( EAa 32 / 2 L30 ) n (3.39a)
n 1
p
C3   ( EAa 2 a3 / L30 ) n (3.39b)
n 1
       12 v k 
p f 12 12
C 2   t 0 a3  EA a 22  2a1a3 / 2 L30 s v 
sr r (3.39c)
 n
n
n 1 n 1 s 1 r 1
    x k
p f 12 12 N
C1   t 0 a 2  EAa1a 2 / L 3
0 n s v
sr s n   Fn vn (3.39d)
n 1 n 1 s 1 r 1 n 1
     12 x k 
p f 12 12 p
C0   U 0  t 0 a1  EAa12 / 2 L30 s sr x r    Fn x n (3.39e)
  n n1
n
n 1 n 1 s 1 r 1
e0 EA 2
And U 0   T0 de0  e0 (3.40)
0 2 L0
Determination of step length S
The value of (S) can now be found from the condition that (dW / dS )  0 at the
stationary point along the path of descent vector. Differentiating Eqn (3.38) with respect to S
and applying the above condition yields the following cubic equation in S from which the
step length can be calculated.
4C4S3 +3C3S2 +2C2S +C1 = 0 (3.41)
The complete solution of the above equation is not necessary as only the root corresponding
to the nearest stationary point along the descent vector is required.
This root can most conveniently be calculated using Newton’s approximation formula,
which in this case can be written as:
dW dS
S k 1  S k  2 (3.42)
d W dS 2
4C 4 S 3  3C3 S 2  2C 2 S  C1
Or S k 1  S k  (3.43)
12C 4 S 2  6C3 S  2C 2
Where k is an iteration suffix and S k 0  0 . From the above equations it may be noted that it
is not necessary to compute the coefficient C0.
43
Determination of member forces and displacements at xk+1 in displacement
space
To calculate the gradient vector and hence the descent vector and step length
coefficients at stage (k+1), we need to write down the expressions for the tension
coefficients and displacements at this stage. Using Eqs (3.33) and (3.36) it yields:
t ab k 1  t ab k  EA (a1  a 2 s  a3 s 2 ) ab
  3
(3.44)
L0 ab
Where a1 , a 2 and a 3 are given by Eqn (3.37) and the new displacement vector is given by
the relation xk 1  xk  S k vk where S k is obtained using Eqn (3.43) and v k is given by
one of the descent methods mentioned before in section (3.3).
3.5 Numerical ill-conditioning, scaling and convergence

3.5.1 Condition number and sources of ill-conditioning (56)
For a symmetric positive matrix [K], the spectral condition number C is defined as the
ratio of the largest to the smallest eigenvalue of the matrix.
 max
C (3.45)
 min
The convergence rate depends on the condition number C. So, the larger the condition
number, the narrow are the ellipses forming the TPE contours. When C  1 , ellipses of the
TPE contours approach circularity, a well-conditioned system. Ill conditioning in cable
structures with elastic boundaries arises from large variations in the various stiffness of the
elements of the system. When individual stiffness matrices are associated with the rotations
of joints along with the translations, the problem of ill conditioning becomes more apparent.
In the case of cable structures, this results in pronounced “ long narrow TPE valleys “. The
variations in the relative magnitudes of the stiffness of the elements may be due to the
inclusion of different types of members, member sizes, elastic modulus and the dimensional
units used.
3.5.2 Scaling and ill-conditioning of the TPE surface (56)

If the contour lines are transformed so that the TPE valleys are widened, ill-
conditioning can be remedied. This can be done by replacing the actual displacement vector
[x] with a transformed vector [u] where:
[x] =[H][u] (3.46)
And [H] is a diagonal “scaling“ matrix.
From Eqn (3.3), [u] at the (k+1) th intersection may be expressed as:
uk 1  uk  S vk (3.47)
It has been shown that when [x] is substituted in the TPE expression, [K] is transformed
to K  , where:
  
K   H T K H  (3.48)
In this case, Eqn (3.22) may be rewritten as:
 EA 2 
W  x  [ K ][ x ]  [ F ][ x]  u 0  T0 e 
1 T
e 
2  2 L0 
(3.49)
1 T  EA 2 
[u ] [ H ]T [ K ][ H ][u ]  [ F ][ H ][u ]  u 0  T0 e  e 
2  2 L0 
44
If [H] is chosen such that all the elements on the leading diagonal of K  are unity, and
K  is symmetric, with its off-diagonal terms tending to zero, then the eigen-values of
K  will be approximately equal. When this is so, the valleys will widen and the rate of
convergence of the conjugate gradient method will improve.
It may be noted that when K  is a unit matrix; both the methods of steepest descent and
conjugate gradients should converge in one iteration, Fig. (3.6).
Optimal scaling is a numerically difficult problem. In practice, however, considerable
benefit is obtained computationally by choosing [H ] as a diagonal matrix where:
H ii  1 K ii 
12
(3.50)
Where K ii is the i th element along the leading diagonal of the stiffness matrix.
Or H  diag{k111 / 2 , k 22
1 / 2 1 / 2
,............k nn } (3.51)
For minimizing the TPE using CGM, Eqn (3.8) will be written as:
[v]k 1  [ H ][ g ]k 1   k [v]k (3.52)
Where [v]0  [ g ]0 (3.53)
Eqs (3.10) and (3.12) for calculating  will be, respectively, rewritten as:
^
[ g ]Tk 1 [ H ]T [ H ][ g ]k 1 [ g ]Tk 1 [ K ][ g ]k 1
k   (3.54)
[ g ]Tk [ H ]T [ H ][ g ]k ^
[ g ]Tk [ K ][ g ]k
[ g ]Tk 1 [ H ]T H {[ g ]k 1  [ g ]k } [ g ]Tk 1 [ K ]{[ g ]k 1  [ g ]k }

^
And k   (3.55)
[ g ]Tk [ H ]T [ H ][ g ]k ^
[ g ]T [ K ][ g ]
k k
K   H  H T  H T H 
^
Where (3.56)
Experience has shown that updating of the scaling matrix during the iterative process
is only necessary for structures which are highly nonlinear, such as freely suspended
systems and systems with very low levels of pretension. Updating of the scaling matrix
requires that the conjugate gradient method be restarted. This limits the number of times it is
practical to rescale. Numerical investigations seem to indicate that if it is considered
necessary to rescale this should be done not more frequently than after every N iterations,
where N is the number of degrees of freedom.
3.5.3 Convergence criteria (Termination procedures)

The convergence criterions are needed to terminate the iterative process. The problem
is considered converged to a sufficient degree of accuracy when:
1) The TPE remains constant; or
2) The changes in forces and displacements are negligible; or
3) The elements in the product S k vk are approximately zero; or
4) The norm of the gradient vector, Rk  [ g ]Tk [ g ]k  1/ 2
, is reduced to a predetermined
value.
The last criterion is the simplest one and is considered of sufficient accuracy when
[ R]k  0.01  0.001[ R]1 (3.57)
45
3.6 Summary of the iterative procedures
The main steps in the iterative processes to achieve structural equilibrium by
minimizing of total potential energy are summarized as follows:
First, before start of the iteration scheme

a) Calculate tension coefficients for pretension forces in cables, Eqn (3.33).
b) Assume the elements in the initial displacement vector to be zero, x0  0 .
c) Calculate the lengths of all the elements in the pretensioned structure using Eqn
(3.23).
d) If either the method of steepest descent or the method of conjugate gradients is
used, calculate elements in the scaling matrix using Eqn (3.51).
The steps in the iterative process are as follows:

Step (1) Calculate the elements in the gradient vector of the TPE, Eqn (3.32).
Step (2) Calculate the Euclidean norm of the gradient vector, Rk  [ g kT g k ]1 / 2 , and check if
the problem has converged. If Rk  Rmin stop the calculations and print the results.
If not, proceed to step (3).
Step (3) Calculate the elements in the descent vector, v using Eqn (3.52) and either Eqn
(3.54) or Eqn (3.55).
Step (4) Calculate the coefficients in the step-length polynomial from Eqn (3.39) bearing in
mind that C0 need not be calculated.
Step (5) Calculate the step-length S using Eqn (3.43).
Step (6) Update the tension coefficients using Eqn (3.44).
Step (7) Update the displacement vector using Eqn (3.46).
Step (8) Repeat the above iteration by returning to step (1).
Note Attention should be paid to the accuracy to which S is calculated.

This is practically so when analyzing structures containing both cables and flexural
members.
46
Fig. (3.1): Two-dimensional pretensioned cable system with one free joint.
Fig. (3.2): Contour lines of total potential energy for the structure shown in Fig. (3.1).
47
Fig. (3.3): Contour map of the TPE surface for a two DOF structure indicating diagrammatically the basic
method used in the nonlinear analysis of structures.
Fig. (3.4): Construction of two conjugate directions: The directions v k and v k 1 are k -orthogonal if v k 1 is
parallel to the tangent v to the ellipse ( v kv  constant  c ). The equation of the straight line through v k
T
and parallel to v k 1 is in a parametric form: v  vk  tv k 1 where t  scalar quantity.
48
Fig. (3.5): Geometric representation of the conjugate gradient method for a two DOF structure.
Fig. (3.6): Convergence in one direction for both methods of steepest descent and conjugate gradients.
49
Fig. (3.7): Coordinates and force for link ab in: (a) Pretensioned but not loaded cable assembly,
(b) Pretensioned and loaded cable assembly at stage of the iterative process where the extension of the
link ( ab ) is eab and (c) Pretensioned and loaded cable assembly when the calculated strain is eab
and the strain due to the next step Sv is eab .
50
Chapter (4)
Dynamic analysis of cable roofs
4.1 Introduction
The purpose of dynamic analysis is to predict the dynamic response of the structure due
to dynamic loads. Dynamic loads may be harmonic, periodic or random loads. Random
loads include wind, earthquakes, waves and impulse loads.
Behavior of structures may be linear or nonlinear. Nonlinear behavior may be stiffening or

softening. In softening structures, such as cable roofs, the rate of change of displacements
will increase deformation while the reverse will be the case in stiffening structures such as
suspension bridges. It is mentioned that even structures of linear behavior may act non-
linearly when subjected to strong excitation.
Pre-tensioned cable roofs are lighter and more flexible than other forms of structures and, as
a result, they are more resistance to earthquakes and more sensitive to wind than
conventional structures.
The dynamic sensitivity of pretensioned cable system is generally increasing with increasing
the gravitational load. Experience has shown that cable roofs with adequate curvatures and
cable tensions are dynamically stable and vice versa.
A complete dynamic analysis includes: (5)

1) A frequency analysis;
2) Establishment and formulation of the dynamic loading;
3) Estimates and formulation of the structural damping; and
4) A dynamic analysis.
The dynamic analysis, due to random loading, may be carried out in either time
domain analysis or in frequency domain analysis. For structures of nonlinear behavior, cable
systems, the stiffness of the structure varies with the amplitude of response. As a result both
natural frequencies and damping will also vary. For this reason, the closed-form solutions to
obtain frequencies (frequency domain analysis) are no longer valid. During analysis,
structural damping is assumed to be constant because of lack of information, and because
the values given in the codes of practice tend to be conservative (23).
In this chapter, we will expose to the following:

(1) Time domain analysis of cable suspended roofs by minimizing the TPDE (total
potential dynamic energy) as an extension to the work done in chapter (3); and
(2) Frequency domain analysis of cable suspended roofs with which the dynamic analysis
in this thesis will be done.
The main sources of knowledge in this chapter are references (5), (23), (24), (59) and (60).
51
4.2 Time domain analysis (step by step integration method)
Time domain analysis (step by step integration method) is the general approach for
predicting the dynamic behavior , either linear or nonlinear, of structures to all types of
dynamic loading. This method is based on the incremental equation of motion, and arises
from various assumptions with respect to the change in acceleration during a time step t .
Other time domain methods, which are particularly suitable for structures of highly
nonlinear behavior such as guyed masts, membrane and cable roofs, are those in which
equilibrium of the dynamic forces at the end of each time step is reached by minimizing the
TPDE using one of the descent methods such as conjugate gradient or Newton-Raphson
methods. Any of the previous time domain methods requires the ability to generate dynamic
load histories, i.e., earthquake histories and wind histories. Then the dynamic response is
obtained by a forward integration of the load histories in the time domain.
Equation of motion is:
Mx  Cx  Kx  P(t ) (4.1)
Incremental equation of motion is:
Mx  Cx  Kx  P (4.2)
This equation is solved only if there exists a relationship between x, x and x .
The most commonly such relationships are: (23)
1. Linear change of acceleration method;
2. Wilson-  method;
3. Constant acceleration method; and
4. Newmark  -method.
(1) Linear change of acceleration method

The acceleration during time step is assumed to vary linearly, see Fig. (4.1). Using the
incremental displacement x as the basic variable for convenience, this leads to:
x(t  t )  x  x (4.3a)
3 t
x (t  t )  x  2 x  x (4.3b)
t 2
6 6
x(t  t )  2 x  x  2x (4.3c)
t t
This method becomes unstable if t  T / 2 , ( T  1 / f is the natural vibration period).
t has to be less than T/10 to ensure sufficient accuracy in the predicted response.
(2) Wilson  -method

In this method, the acceleration is assumed to vary linearly during a prolonged time step
t , where   1 as shown in Fig. (4.2). This assumption leads to the following relations:
1 1 t 2 3
x(t  t )  2 x  x  t (1  2 ) x 
 (4  ) x (4.4a)
  6 
3 3 3t 1
x (t  t )  3 x  (1  2 ) x  (1  ) x (4.4b)
 t  2 
6 6 3
x(t  t )  3 2 x  2 x  (1  ) x (4.4c)
 t  t 
For linear-behavior structures, the method is stable when   1.37 . In general, a value of
  1.4 is satisfactory. If (   1.4 ), the predicted amplitude of response is much
overestimating.
52
(3) Constant acceleration method
The acceleration is assumed to remain constant during the time step t and equal to
1
( x  x) as shown in Fig. (4.3). This assumption leads to the following relations:
2
x(t  t )  x  x (4.5a)
2
x (t  t )  x  x (4.5b)
t
4 4
x(t  t )  2 x  x  x (4.5c)
t t
(4) Newmark  -method

Newmark proposed expressions for velocity and displacement at time (t  t ) as:
x(t  t )  x  t (1   ) x  t  ( x  x) (4.6a)
1
x(t  t )  x  t x  t 2 (   ) x  t 2  ( x  x) (4.6b)
2
Where  and  are variable constants. Using these two above equations, all quantities at
time (t  t ) can be expressed as: (59)
x(t  t )  x  x (4.7a)
  
x (t  t )  x  (  1) x  t (  1) x (4.7b)
 t  2
1 1 1
x(t  t )  x  x  (  1) x (4.7c)
 t 2
 t 2
It is usually assumed that   1 / 2 . With   1 / 2 and   1 / 6 or   1/4, the Newmark
  method is identical to the linear acceleration or constant acceleration methods
respectively.
Choosing of the method

Experience indicates that, with the same time steps, the constant acceleration method
and Wilson   method with   1 / 4 yield similar results. The constant acceleration method
may thus be preferable since less computational effort is required. It must be noted that for
any of the above methods, the time step must be small enough to enable all the significant
harmonic components in the load history to be taken into account, and should not be greater
than 0.05 times the period of the system analyzed.
Incremental equation of motion

Assuming constant acceleration during time step  , Eqn (4.5), it yields that:
x  x(t  t )  x (4.8a)
2
x  x (t  t )  x  x  2 x (4.8b)
t
4 4
x  x(t  t )  x  2 x  x  2x (4.8c)
t t
Substituting with these relationships in Eqn (4.2) it yields:
2 4 4
[ K  C  2 M ]x  P  2Cx  M [ x  2x] (4.9)
t t t
53
Where: K, C and M = Static stiffness, damping and mass matrices for a multi-DOF structure
at time t ;
x = the incremental displacement vector;
x, x and x = Displacement, velocity and acceleration vectors at time t respectively; and
P  The incremental load vector.
The elements in vectors x, x and x at time (t  t ) are given by Eqs (4.5a,b and c).
Using Eqn (4.9), dynamic stiffness matrix and incremental dynamic load vectors are:
2 4
K d  [K  C  2 M ] (4.10)
t t
4
Pd  P  2Cx  M [ x  2x] (4.11)
t
Hence x  K d1Pd (4.12)
Evaluating of the damping

The work done by damping forces (forces resisting the motion) causes an energy loss
in vibration. This energy loss is caused by hysteretic in the members, friction in joints and
resistance by the surrounding mass of air. On a hand, damping forces have a marginal effect
on the first amplitude of vibration of a structure to a sudden random load such as wind gust
or shock loading. On the other hand, they have a very much effect on the rate of decay of the
subsequent oscillatory motion and the amplitudes of response due to regularly pulsating
forces, such as those due to vortex shedding. As it is not practical to include damping forces
separately, it is usual to express damping coefficients from knowledge of the modal
damping ratios of similar structures. Damping ratio is expressed as:
 
  (4.13)
4 2   2 2
Where  , the logarithmic decrement of damping, is given by:
1 x
 ln r (4.14)
s  r xs
x r and x s are the amplitudes at the end of the r th and s th oscillation respectively.
For r  0 and s  n , where n= number of oscillations, see Fig. (4.4), it yields:
1 x
  ln 0 (4.15)
n xn
Since xn  x0 e  n nT , where natural frequency  n  K M and T is the period, it yields
xr x r xr
that:   (4.16)
x s x s xs
x x x
So, we can substitute the value r in Eqn (4.14) by either r or r .
xs x s xs
C C
Knowing that:   (4.17)
Ccr 2 n M
Where: Ccr  the critical damping.
 C  2 n M (4.18)
Experimentally, it appears that:
For cable structures with cladding:   0.007 to 0.02.
For cable structures without cladding:   0.0016 to 0.005.
54
Incremental stiffness and load due to turbulent wind
Using Eqn (2.24), at time (t  t ) :
P  P  Fw [(V  V )  ( x  x )]2 (4.19)
1 ~ ~ ~
Where: Fw   (C Pi  C Di  C LI sin t ) (4.20)
2
Subtracting Eqn (2.24) from Eqn (4.19), ignoring the second order terms of smallness, yields
the following expression for the incremental load due to turbulent wind:
P  2Fw (V  x )(V  x ) (4.21)
Where x is calculated using Eqn (4.8b).
Substitution of the above expression for P into Eqn (4.9) yields:
2 4 4
[ K  C  2 M  Fw (V  x )]x 
t t t
(4.22)
4
2 Fw (V  x )(V  2 x )  2Cx  M ( x  2 x)
t
Which reveals that the wind, as well as exciting the structure, also increases the dynamic
stiffness (23), which alternatively may be written as:
2 4
K d  [ K  (C  2 Fw (V  x ))  2 M ] (4.23)
t t
This equation, comparing with Eqn (4.10), shows that wind not only excites a structure, but
also increases the damping coefficient by the term 2Fw (V  x ) .
Generalized method of minimizing TPDW

This method consists of a step-by-step response calculation (forward integration) in
the time domain using small time increments. During each increment, t , the equilibrium of
the dynamic forces at the end of each step is based upon minimizing of the total potential
dynamic work. This method is a generalization to the theory illustrated in chapter (3) for
both static and dynamic loads affected structures of either linear or nonlinear behavior.
Assumptions
(a) The deformations of the structural elements remain within the elastic limits and
their behavior obeys Hook’s law.
(b) The cables can be treated as assemblies of linear pin-jointed link elements.
(c) The structural damping may be expressed as equivalent viscous damping.
(d) The values of pressure, drag and lift coefficients are invariable.
(e) The dynamic loads may be applied as equivalent concentrated point loads.
(f) The histograms for ground acceleration at the position of the supports are
identical and without phase shifts.
(g) The direction of the wind is in the direction of the positive global x-axis.
(h) The rate of change of acceleration during a time step remains constant, Eqn (4.5).
The total potential (energy) work at time (t  t )

TPDW at any time is a function of:
a) The elastic energy stored in the structural elements (flexural or cable elements);
b) The kinetic energy of the structural system and any additional gravitational loads
such as ice and snow;
c) The energy dissipated by structural and aerodynamic damping; and
d) The work done by the gravitational and dynamic forces.
55
When only the response due to the gradually applied gravitational forces is required, the
work done by the inertia and damping forces may be ignored. In such cases the integration
process is reduced to one step in which the condition of equilibrium is determined by
minimization of the TPE only, as shown in chapter (3).
TPE or TPW= W  W{x}  W{x}  W{x}  W{V }  W{u}  W{P} (4.24)
Where: W {x]  The instantaneous TPE of the structure and gravitational loads;
W {x}  The total energy dissipated by the structural damping;
W {x}  The kinetic energy of the structure and the gravitational loads;
W {V }  The total work done by the wind;
W {u}  The total work done by the inertia forces due to support acceleration; and
W {P}  The total work done by other forms of time-dependent forces.
If the motion of a structure given by Eqn (4.2) is assumed to be relative to the supports, then
W may be written explicitly in terms of dynamic loading functions as:
f 12 12 p N
1
W   ( k sr ( ~ x  ~
x ) r (~ x ) s ) n   (U  Te  EAe 2 / 2 L) n  VF   Fs x s
x  ~
n 1 s 1 r 1 2 n 1 s 1
N N xs N N xs
 Ds    c sr ( x  x ) r d (x s )  I    msr ( x  x) r d (x s )  QV
s 1 r 1 0 s 1 r 1 0
N xs N x
s
1 ~ ~ ~
   (C P  C D  C L sin ~t ) s (V  x L  x L ) 2s d (x s )  Qu    msr us d (x s )
s 1 0 2 s 1 0
N xs
 Q p    P( t ) s d (x s )
~
s 1 0
(4.25)
Where:
f  The number of flexural members; p  number of pin-jointed members;
~
x  Total static and dynamic displacement at t ; N  number of degrees of freedom;
U n , Tn  Strain energy and tension in link n at time t ; F  gravitational forces;
Ln , en  The length at time t and the change in length during time step t in link n
respectively; VF  Potential energy of gravitational forces at time t ;
Ds , I , QV , Qu , Q p  Work done by: internal damping forces, inertia forces, wind, inertia
forces due to support acceleration and forcing function P(t ) from time t  0 to t ;
V , u  The average wind speed and support acceleration during the time step t ;
~ 1 ~ ~ ~
t  t  t; C P , C D , C L  Aerodynamic coefficients given by Eqn (2.25); and
2
x L  x L  The along-wind velocity of the joint associated with degree of freedom s.
The condition of dynamic equilibrium is given by:

W
gs  0 , s  1,2,3,....., N (4.26)
(x s )
Where: x s  The displacement in the direction s during a time increment t , and
N= Number of degrees of freedom.
The condition of dynamic equilibrium is logically determined by using a gradient
minimization technique. Newton-Raphson method is used if N is relatively small.
On the other hand, a conjugate gradient algorithm is more advantageous if N is large.
56
The iterative process required to locate the minimum may be expressed as:
xk 1  xk  S k vk (4.27)
Where: v  The descent vector; k  Iteration counter; and
S  The step length to the nearest minimum along v.
According to Eqs (4.24) and (4.26), it follows that:
g  g{x}  g{x}  g{x}  g{V }  g{u}  g{P} (4.28)
Descent vector v
Dynamic equilibrium at the end of each time increment is established by minimizing W
along v until g  0 , using one of the following algorithms:
1. Newton-Raphson method:
~
vk   (x)  k 1 ( ) g k (4.29)
~
Where: k ( )  k d  instantaneous dynamic stiffness matrix, Eqn (4.10).
2. Conjugate gradients method:
vk 1  g k 1   k vk (4.30)
Where  k is expressed by any of Eqs (3.10), (3.12) and (3.13) as shown:
g Tk 1 g k 1 R2k 1
k   (4.31)
g Tk g k R2k
g Tk 1 g K 1  g K  R2K 1  g TK 1 g K
K   (4.32)
g TK g K R2K
R2 g   g k 
 k  k 1 k 21 (4.33)
Rk
Step length S
Substituting the expression for xk 1 , Eqn (4.27), into the function of W, it yields the
following fourth-order polynomial in S:
W  G4 S 4  G3 S 3  G2 S 2  G1 S  G0 (4.34)
Where: G4  G4 {x}  G4 {x}  G4 {x}  G4 {V }  G4 {u}  G4 {P} (4.35a)
G3  G3 {x}  G3 {x}  G3 {x}  G3 {V }  G3 {u}  G3 {P} (4.35b)
G2  G2 {x}  G2 {x}  G2 {x}  G2 {V }  G2 {u}  G2 {P} (4.35c)
G1  G1{x}  G1{x}  G1{x}  G1{V }  G1{u}  G1{P} (4.35d)
G0  G0 {x}  G0 {x}  G0 {x}  G0 {V }  G0 {u}  G0 {P} (4.35e)
The value of S may be determined from the condition that:
dW
0 (4.36)
dS
As the position of the nearest minimum along v is required, it is usual to use Newton
approximation formula to determine values of S by assuming zero initial values of S.
In order to develop expressions for g , Eqn. (4.28), and the coefficients G0  G4 , Eqn.
(4.35), Eqn (4.25) must be written in terms of x rather than in terms of e, x and x .
To avoid lengthy expressions, the six terms in g in Eqn (4.28) and their corresponding six
terms in each of the expressions G0 to G4 in Eqn (4.35) are developed separately.
For convenience, the time and iteration counter will be omitted.
57
A) Total potential energy, TPE=W{x}
TPE = (Elastic energy stored in flexural members) + (Elastic energy stored in pin–jointed
members) + (The work done by the static loads).
f 12 12 p
1
W {x}  { [ k sr ( x  x ) r ( x  x ) s ]n }  { (U  Te  EAe 2 / 2 L) n }
~ ~ ~ ~
n 1 s 1 r 1 2 n 1
N
 {VF   Fs x s } (4.37)
s 1
According to Eqn (4.34) W {x} can be expressed as:
W {x}  G4 {x}S 4  G3 {x}S 3  G2 {x}S 2  G1{x}S  G0 (4.38)
And Gi {x}  Gi { f }  Gi { p}  Gi {F}, i  0,1,2,3,4 (4.39)
Where Gi { f } are corresponding to the flexural members; Gi ( p) are corresponding to pin-
jointed members and Gi (F ) are corresponding to the static loads.
A-1) The energy stored in the flexural members, W f (x)
f 12 12
1
W f {x}  [ k sr ( ~ x  ~x ) r (~
x  ~
x ) s ]n (4.40)
n 1 s 1 r 1 2
The s th element in the gradient vector is:
W f {x} f j 12 ~
g f {x}s    [( x  x) r k sr ]n (4.41)
(x s ) n 1 r 1
Where: f j  the total number of flexural members meeting at joint j.
Replacing x in Eqn (4.41) by (x  Sv) , the expressions of Gi { f } in Eqn (4.39) are:
G4 { f }  0
G3 { f }  0
f 12 12
1
G2 { f }   ( k sr v r v s ) n (4.42)
n 1 s 1 r 1 2
f 12 12
G1{ f }   [k sr ( ~
x  x) r v s ] n
n 1 s 1 r 1
f
12 12
1
G0 { f }   [ k sr ( ~ x  x) r ( ~
x  x) s ] n
n 1 s 1 r 1 2
A-2) The energy stored in pin-jointed and cable members, W p {x}

p
e 2
W p {x}   (U  T e  EA )n (4.43)
n 1 2L
W p {x} p j e e
g p {x}s    (T  EA ) n (4.44)
(x s ) n1 L x s
Where: p j  number of pin-jointed members meeting at joint j.
3
For a link ab : Lab   ( X bi  X ai ) 2 , i  1,2,3 (4.45)
i 1
3
( Lab  eab ) 2   ( X bi  xbi  X ai  xai ) 2 (4.46)
i  1
58
Subtracting Eqn (4.45) from Eqn (4.46), it yields:
 
3
2 Lab eab  eab   2 X bi  X ai xbi  xai   xbi  xai 
2 2
(4.47)
i 1
Since eab  Lab , the part eab may be approximated as:
1 3 
eab    X bi  X ai xbi  xai   1 xbi  xai 2  (4.48)
Lab i 1  2 
Differentiating Eqn (4.47) with x s where x s  xai , this yields:
eab
2( Lab  eab )  2( X bi  xbi  X ai  xai ) (4.49)
x s
eab  1
As Lab  eab :  ( X b  X a  xb  xa ) i (4.50)
x s Lab
eab
Substituting for into Eqn (4.44), this yields:
x s
pj
g p {x}  [(t ab  t ab )( X b  X a  xb  xa ) i ]n (4.51)

n 1
Where (t ab  t ab ) is the tension coefficient of the force in member ab and is equal to:
EA
(t ab  t ab )  (T  T ) ab Lab  (T  e) ab Lab (4.52)
L
If x in Eqn (4.48) is replaced by (x  Sv) , this yields:
1
eab  (a1  a 2 S  a3 S 2 ) (4.53)
Lab
3
1
Where a1 `[( X bi  X ai )(xbi  xai )  (xbi  xai ) 2 ] (4.54a)
i 1 2
3
a 2    X bi  X ai    xbi  xai vbi  vai  (4.54b)
i 1
3
a3  
1
vbi  vai 2 (4.54c)
i 1 2
Substituting the expression for eab given by Eqn (4.53) into Eqn (4.43), yields:
p
G4 { p}   ( EAa 32 / 2 L3 ) n (4.55a)
n 1
p
G3 { p}   ( EAa 2 a3 / L3 ) n (4.55b)
n 1
   
p
G2 { p}   t a3  EA a 22  2a1a3 / 2 L3 n
(4.55c)
n 1
 
p
G1{ p}   t a 2  EAa1a 2 / L3 n
(4.55d)
n 1
 
p
G0 { p}   U  t a1  EAa12 / 2 L3 n
(4.55e)
n 1
eab EA
And U  Tab deab  eab
2
(4.56)
0 2 Lab
59
A-3) The work done by static loads (gravitational loads), WF {x}
N
WF {x}   Fs x s  VF (4.57)
s 1
WF {x}
g F {x}s    Fs (4.58)
(x s )
By replacing x s in Eqn (4.57) by (x  Sv) , yields:
G 4 {F }  0
G 3 {F }  0
G 2 {F }  0 (4.59)
N
G1{F }   Fs v s
s 1
N
G0 {F }   Fs x s  VF
s 1
Substituting with Eqs (4.42), (4.55) and (4.59) into Eqn (4.39) gives the values of
coefficients G0  G4 for the total potential energy, TPE  W {x} , Eqn (4.37).
B) Total potential energy dissipation by structural damping, W {x}

If the internal forces opposing the motion are assumed to be proportional to the velocity of
vibration then the total potential energy dissipation due to friction and hysteretic losses and
according to Eqn (4.5b) may be written as:
N N xs
W {x}    c sr ( x  x ) r d (x s )  Ds
s 1 r 1 0
N N xs 2
   c sr ( x  x ) r d (x s )  Ds (4.60)
s 1 r 1
0 t
N N
1
  c sr ( x  x ) r x s  Ds
s 1 r 1 t
Differentiating Eqn (4.60) with respect to x s , yields the s th element in the gradient vector
as follows:
N
2
g{x}s   c sr ( x  x ) (4.61)
r 1 t
By replacing x s with (x  Sv) in Eqn (4.60), the coefficients are as follows:
G4 {x}  0
G3 {x}  0
N N
1
G2 {x}   c sr v r v s (4.62)
s 1 r 1 t
N N
2
G1{x}   c sr ( x  x ) r v s
s 1 r 1 t
N N
1
G0 {x}   c sr ( x  x ) r x s  Ds
s 1 r 1 t
60
C) Total potential work of inertia forces, W {x}
TPW of the inertia forces and according to Eqn (4.5c) is given by:
N N xs
W {x}    msr ( x  x) r d (x s )  I
0
s 1 r 1
N N xs
4 4
   x  x  x) r d (x s )  I
msr ( (4.63)
s 1 r 1
0 t 2
t
N N
2 4
  msr ( 2 x  x  x) r x s  I
s 1 r 1 t t
Differentiating Eqn (4.63) to x s yields the s th element in the gradient vector as:
N
4 4
g{x}s   m sr (
x  x  x) r (4.64)
r 1 t 2
t
By replacing x s with (x  Sv) in Eqn (4.63), the coefficients are as follows:
G4 {x}  0
G3 {x}  0
N N
2
G2 {x}   msr v r v s (4.65)
s 1 r 1 t
2
N N
4 4
G1{x}   m sr ( 2 x  x  x) r v s
s 1 r 1 t t
N N
2 4
G0 {x}   m sr ( 2 x  x  x) r x s  I
s 1 r 1 t t
D) Total potential work of wind, W {V }
TPW of wind may be written using aerodynamic coefficients given by Eqn (2.25) as:
N xs
1 ~ ~ ~
W {V }     (C P  C D  C L sin  t ) s (V  x L  x L ) 2s d (x s )  QV
~ (4.66)
s 1 0 2
For simplicity, it may be assumed that V  x L , Eqn (4.66) may be reduces to:
N xs
1 ~ ~ ~
W {V }     (C P  C D  C L sin ~t ) s (V  x L ) 2s d (x s )  QV
s 1 0 2
N
1 ~ ~ ~
   (C P  C D  C L sin  t ) s (V  x L ) 2s x s  QV
~ (4.67)
s 1 2
Differentiating Eqn (4.67) with respect to x s , yields the s th element in the gradient vector:
1 ~ ~ ~ ~
g{V }s    (C P  C D  C L sin  t ) s (V  x L ) 2s (4.68)
2
By writing (x  Sv) instead of x s in Eqn (4.67), the coefficients are as follows:
G4 {V }  0 G3 {V }  0 G2 {V }  0
N
1 ~ ~ ~
G1{V }    (C P  C D  C L sin  t ) s (V  x L ) 2s v s
~ (4.69)
s 1 2
N
1 ~ ~ ~
G0 {V }    (C P  C D  C L sin  t ) s (V  x L ) 2s x s  QV
~
s 1 2
61
E) TPW of inertia forces due to support movements, W {u}
The TPW of inertia forces caused by an earthquake, which gives rise to acceleration u of the
supports, can be written as:
N N xs N N
W {u}    msr ur d (x s )  Qu   msr ur x s  Qu (4.70)
s 1 r 1 0 s 1 r 1
Differentiating Eqn (4.70) with respect to x s yields:

N
g{u}   msr ur (4.71)
r 1
By writing (x  Sv) instead of x s in Eqn (4.70), the coefficients will be written as:
G4 {u}  0 G3 {u}  0 G2 {u}  0
N N
G 1 {u}   msr ur v s (4.72)
s 1 r 1
N N
G0 {u}   msr ur x s  Qu
s 1 r 1
F) TPW of independent dynamic forces, W {P}

The TPW of independent dynamic forces, that is, work done by time-variable forces which
intensities at any time are independent of the motion of the structure, may be written as:
N xs N
W {P}    P( t ) s d (x s )  QP   P( t ) s x s  QP
~ ~ (4.73)
s 1 0 s 1
Differentiating Eqn (4.73) with respect to x s yields:

g{P}s   P(~t )s (4.74)
By writing (x  Sv) instead of x s in Eqn (4.73), the coefficients will be written as:
G4 {P}  0 G3 {P}  0 G2 {P}  0
N
G 1 {P}   P(~
t ) s vs (4.75)
s 1
N
G 0 {P}   P( t ) s x s  QP
~
s 1
Convergence and scaling

As it was shown in chapter (3), static analysis for problems containing both flexible
and cable members based upon minimization of TPE using the conjugate gradient method
converge badly and need to be scaled. Convergence is usually less of a problem in dynamic
analysis than in the static one. However, the use of scaling tends to increase the overall
efficiency significantly. The scaling in dynamic analysis is the same as in static analysis
except that the i th element along the diagonal of the dynamic stiffness matrix will be
expressed by the incremental dynamic stiffness, Eqn (4.10), as follows:
~ 2 4
k ii  k ii  cii  2 mii (4.76)
t t
The minimization process may convergence to a “false” equilibrium position if, at any time
during the iterative process, one or more cable elements go into “compression”.
This difficulty may be overcome by determining the “false” equilibrium position and then
restarting the iterative process, having first reduced Young’s modulus of the elements in
compression to nominal values, such as 10 20 N / mm2 .
62
Stability and accuracy versus the magnitude of the time step
The size of the time step must be sufficiently small to ensure that the energy content of
the highest-frequency component that is likely to excite the structure is accurately described.
Experience has shown that this requires the period of the smallest significant harmonic load
component to be divided into between 20 and 30 time steps. In addition, the time step t
must be sufficiently small to take into account the highest mode in which the structure is
likely to respond.
1
Generally, experience has shown that the step-by-step method given is stable if t  Tmin ,
2
where Tmin is the smallest periodic time of the structure, and tends to be unstable if
1
t  Tmin . Also, experience has shown that the use of time steps much smaller than those
2
required to ensure stability only marginally increases the accuracy of the predicted response.
1
If Wilson  -method is used, the integration process becomes stable at t  Tmin . Size of
2
t , however, is always limited by the necessity of describing the dynamic load accurately.
4.3 Frequency domain analysis

This method is limited to the analysis of linear-behavior structures, although in
practice it is also applied to some nonlinear structures by taking only the nonlinear response
due to the mean wind speed component into account. When the frequency domain method is
applied to determine the dynamic response of nonlinear-behavior structures such as cable
roofs, whose stiffness and frequencies are functions of the degree of deformation, the natural
frequencies should be determined for the deformed state due to the mean wind component
and not for the case when there is no load on the structure. Apart from the assumptions with
respect to the statistical characteristics of wind, the main assumption made in order to make
the method possible is that the amplitudes of the fluctuating component of wind are
sufficiently small compared to the mean wind speed and can be ignored. Generally, this
assumption is justified except for sites in mountainous areas, where fluctuations of the same
order of magnitude as the mean wind speed have been observed. This method is based on
the spectral density function that enables the use of closed-form solutions, of the random
loading. For this reason, the accuracy of analysis will vary with the type of spectral density
function used and it may be advisable to construct spectral density functions, for important
structures, from recordings at site, concerned. For nonlinear analysis, frequency domain
method will underestimate the response in the case of softening structures and overestimate
it in the case of stiffening structures. An N DOF structure has N eigenvalues and
eigenvectors associated with the system of equations that defines its motion.
Eigenvalues are equal to the square of the natural angular frequencies of the structure,
which indicate whether or not the structure is likely to respond dynamically.
Eigenvectors corresponding to the natural frequencies represent the natural modes or

modeshapes in which the structure can vibrate and, hence, the best positions for placing
artificial dampers if required.
The determination of eigenvalues is of fundamental importance to the frequency domain

method of analysis, in which the distribution of energy of random forces such as wind are
given as functions of their frequency content in terms of power spectra.
63
Structural damping is usually not included when one is formulating the eigenvalue
problem, as it increases the numerical effort considerably and has only a second effect on
the calculated frequencies.
The eigenvalue equation may be written in general matrix notation as:

K   2 M  0 (4.77a)
Or K  M  0 (4.77b)
Where: K  The tangent stiffness matrix at the static equilibrium position;
M  The mass matrix, which is diagonal when the structure mass is lumped
together at nodes as the case for cables;
  The modeshape matrix (eigenvectors); and
 2    N  N diagonal matrix of the square of the natural angular frequencies
(eigenvalues) corresponding to the mode shape matrix  .
In the iterative method the eigenvalues  2 and eigenvectors  are determined by optimizing
an assumed modeshape vector through an iterative procedure on either:
 2  M 1 K (4.78)

Or  K 1 M (4.79)
 2
Iterations on Eqn (4.78) will cause the assumed eigenvector to converge towards the mode
corresponding to the highest eigenvector and hence the highest frequency; iterations on Eqn
(4.79) will cause the assumed eigenvector to converge towards the eigenvector
corresponding to the lowest frequency.
Eqn (4.78) involves the inversion of the mass matrix M, which when the matrix is diagonal
is achieved by simply inverting each of the elements on the leading diagonal.
Eqn (4.79) involves the inversion of the stiffness matrix K that is banded and needs more
time to be inverted. This problem can be avoided by calculating the lowest eigenvalue and
eigenvector as follows:
Let 
B i   I  M 1 K  i (4.80)
Where:   Constant larger than the highest eigenvalue;
I  The unit matrix; and
B  Square matrix of the same order as M and K.
From Eqn (4.78) it follows that:
M 1 Ki   i2 Ii (4.81)
Substituting of the expression for M 1 K i given in Eqn (4.81) into Eqn (4.80) yields:
  
Bi     i2 I i     i2 i (4.82)
Assuming an initial vector  i , iteration on Eqn (4.82) will yield the highest value of
 
  i2 and hence the lowest possible value of  i2 . Thus
 i2   l2
(4.83)
i  l
Iteration algorithm based on Eqs (4.78) and (4.82) will yield the highest and lowest natural
frequencies and corresponding modeshape for any structure.
64
The Rayleigh quotient
Pre-multiplication of each term of Eqn (4.77a) by  T yields:
 T K   2 T M  0 (4.84)
 K
T
Hence 2  (4.85)
 T M
The expression for  2 given by Eqn (4.85) is called Rayleigh quotient. It has the property
that for even approximately correct values of eigenvectors or modeshape vectors the values
for the frequencies are reasonably correct. This can be seen simply by pre-multiplying each
term in Eqn (4.84) by ½. This yields:
1 T 1
 K   2 T M (4.86)
2 2
1
Which states that the maximum strain energy (  T K ) is equal to the maximum kinetic
2
1
energy (  2 T M ) due to the mode shape vector  .
2
4.3.1 Mass, damping and stiffness matrices

4.3.1.1 Mass properties
Mass matrix of cable element may be expressed in one of the following methods:
1) Lumped-mass matrix
Assuming that the entire mass of a structure is concentrated at the points at which the
translational displacements are defined, the lumped mass matrix for a cable element is
given by:
1 0 0 0 0 0
0 1 0 0 0 0 
 
mL 0 0 0 0 0 0
M L     (4.87)
2 0 0 0 1 0 0 
0 0 0 0 1 0 
 
0 0 0 0 0 0
Where m is the mass per unit length of the cable element.
2) Consistent-mass matrix
2 0 0 1 0 0
0 2 0 0 1 0 
 
 0 0 2 0 0 1 
M C   mL   (4.88)
6 1 0 0 2 0 0 
0 1 0 0 2 0 
 
0 0 1 0 0 2
Lumped-mass system is less effort than consistent-mass system for two reasons:
i) Lumped-mass matrix is diagonal.
ii) Rotational degrees of freedom can be eliminated from a lumped-mass analysis,
wheares all rotational and translational degrees of freedom must be included in
the consistent-mass analysis.
65
4.3.1.2 Damping matrix
Damping matrix is expressed as:
C   2 M  (4.89)
Thus the damping matrix will only be diagonal if the mass matrix is also diagonal.
Eqn (4.89) implies that the damping forces at different points in a structure are proportional
to the distribution of mass and that the damping ratios decrease and are very small in the
higher modes of vibration.
4.3.1.3 Stiffness matrix (60)

The stiffness matrix of pin-jointed pretensionmed link is given by:
I I
K   EA  T0 GG T  GGT  T0
T T
(4.90)
L0  GG GG L0  I I
Where I is a unit matrix of dimension (3  3) , and
G  l m n where l , m and n are the direction cosines of the member.
4.3.2 Reduction of NDOF

When the mass of a structure is assumed to be concentrated at the nodes, it is usual to
consider only the inertia due to transnational movements and to ignore that due to rotation.
This assumes that the lumped masses are concentrated as point masses with radii of gyration
equal to zero. Thus in the case of flexible structures, such as cable roofs, where the joints
rotate, the elements on the leading diagonal of the mass matrix corresponding to the
rotational degrees of freedom will be zero. In such cases the mass matrix can not be
inverted. Therefore the elements related to the rotation need to be eliminated by condensing
the stiffness matrix. Condensation or reduction of stiffness matrix may also be desirable to
reduce the overall degree of freedom, of structures with a very large number of DOF in
order to reduce the numerical problem. Four condensation methods are mentioned below:
4.3.2.1 Static condensation method

Stiffness equation of any structure may be written using partition of matrices as:
K  Kx    0
 (4.91)
K x  K xx  x P
Where   and xare the displacement vectors corresponding to  and x degrees of
freedom, respectively, where  are the secondary coordinates to be condensed and x are
the primary coordinates (remaining coordinates).
Carrying out a Gauss-Jordan elimination Eqn (4.91) may be written as follows:
I   T    
0
(4.92)
0 K  x P
It should be noted that in Eqn (4.91) it is assumed that at the dependent degrees of freedom
 the external forces are zero.
Eqn (4.92) is equivalent to the following two relationships:
  Tx (4.93)
Kx  P (4.94)
Where T is the transformation matrix given by:
T   K 1 Kx  (4.95)
66
In Eqn (4.94), that shows the relationship between the displacement vector x and the force
vector P, K is the reduced stiffness matrix and may be expressed by the following
transformation of the system matrix:
K  T T KT (4.96)
T 
Where T   (4.97)
I 
Similarly, the reduced mass and damping matrices will be expressed as:
M  T T MT (4.98)
C  T CT
T
(4.99)
Eqn (4.93), which expresses the relationship between displacement vectors x and  , may
also be rewritten, to calculate the modal shape matrix of the system, as:
  T 
  x (4.100)
x I 
4.3.2.2 Dynamic condensation method (60)

Eqn (4.91) will be dynamically extended and rewritten as:
K    i2 M   Kx    i2 M x    0
 (4.101)
K x    i2 M x  K xx    i2 M xx  x 0
Where  i2 is the approximation of the i th eigenvalue which was calculated in the preceding
step of the process. To start the process an approximate or zero value is taken for the first
eigenvalue  12 . After carrying out a Gauss-Jordan elimination of the unknown rotations  ,
Eqn (4.101) will be rewritten as:
 I  T    0
     (4.102)
0 Di   x  0
Where: Di  The reduced dynamic equation and expressed as:
Di  K xx   i2 M xx (4.103)
The reduced mass and damping matrices are expressed respectively as:
M i  Ti M Ti
T
(4.104)
Ci  Ti CTi
T
(4.105)
Where transformation matrix Ti is calculating using Eqs (4.93).
Using Eqn (4.103), the reduced stiffness matrix will be expressed as:
K i  Di   i M i
2
(4.106)
Finally, according to Eqs (4.102) and (4.106), the reduced eigenproblem is:
[ K i   i2 M i ]x  0 (4.107)
This equation solved to obtain an improved eigenvalue  i2 , and also an approximation for
the next order eigenvalue  i21 . The i th modal shape  i is given, using the eigenvector x for
the reduced system, by Eqn (4.100).
67
4.3.2.3 Modified dynamic condensation method (60)
Firstly setting  i2  0 in Eqn (4.106), it yields an unchangeable reduced stiffness matrix
K . Secondly, an approximated value is taken for the first eigenvalue  12 to calculate the
corresponding reduced dynamic matrix Di using Eqn (4.103).
The corresponding reduced mass matrix for the i th mode is given by:
M i   12 K   Di  (4.108)
wi
The reduced eigenproblem will be:
[ K   i2 M i ]x  0 (4.109)
As mentioned before, this equation solved to obtain an improved eigenvalue  i2 , and also an
approximation for the next order eigenvalue  i21 . Also, the i th modal shape  i is given
using Eqn (4.100).
4.3.2.4 Reduction of the structure

For large nets and cable grids, reduction of the structure may be done by lumping the
cables together and distributing the total mass between a smaller number of joints. This form
of simplification will usually, depending upon the degree of simplification, lead to a
reduction in the values of the significant frequencies. The errors incurred may to a large
extent be reduced by reducing the mass of the dead load or by increasing the stiffness until
the potential energy of the equivalent system at equilibrium is equal to that of the original
one. The reduction of joints and the presence of damping may both cause changing in the
order in which the mode shapes appear in the original undamped system.
4.3.3 An indication of dynamic sensitivity

A structure may be examined whether it is likely to be dynamically sensitive or not
without carrying out a full eigenvalue analysis by only calculating the frequency of the first
symmetric mode.
Using Eqn (4.85) and assuming that the first symmetric mode is of the same shape as the
static displacement due to the applied uniformly distributed load, an approximated
expression for the natural frequency of the first symmetric mode for a pure cable structure
may be expressed as:
12
 p EAe 2 N

    ( ) n  (mx 2 ) n  (4.110)
 n 1 L0 n 1 
Where e is the elongation and x the displacement vector due to a uniformly distributed load
of mass m per joint.
Eqn (4.110) may also be used as a check when carrying out a frequency analysis of a
reduced structure.
4.3.4 An approach for the dynamic response due to wind (23)

Assuming that the roof responds only in one mode and that the statistical properties of
the external loading are known.
The magnitude of the amplitude of the maximum fluctuation that may occur within a given
time interval T is expressed as:
z (t )   z (4.111)
68
Where:  is the peak factor and expressed as:
  2 ln(TR )  0.577 / 2 ln(TR ) (4.112)
12
 

Where:    n 2 S z (n)dn /  S z (n)dn (4.113)
 0 0 
And TR is the duration at which, the value of U (10) is recorded and averaged and is taken
equal to 3600 seconds (one hour).
Assuming that the structure is lightly damped,  may be assumed to be:
  f n  n   n / 2 (4.114)
Where: n  f n = the fundamental natural frequency.
When this is the case:
z (t ) max  [ 2 ln( nTR / 2 )  0.577 / 2 ln( nTR / 2 ) ] z (4.115)
And for weakly damped structures, the variance  z2 may be approximated to:
z2  n
 z2  st2 A(n) S z (n) (4.116)
U 
Where:
z st  The static deflection under mean wind velocity,
U  Mean wind speed,
  Damping ratio varies from 0.007 to 0.02 and for cable structures without cladding from
0.0016 to 0.005,
S z (n)  Power spectral density function and
A(n)  Aerodynamic admittance factor, Fig. (4.5), varies with the reduced frequency
nB /U (10) where B is the reference dimension of the structure, i.e. its diameter.
So, the maximum deflection will be expressed as:
z max  z st   z (4.117)
The steps of calculation are as follows: (26)

Step (1) Calculate the natural frequencies of the structure and estimate practical value for
the damping of the structure.
Step (2) Assume mean wind velocity at 10m-height and roughness length z 0 and calculate
U * using Eqn (2.5). Then calculate the mean wind speed at the average height of the
structure using Eqn (2.4).
Step (3) Calculate the power spectral density function S u (n) using one of Equations from
(2.13) to (2.16).
Step (4) Solve the structure statically to get the static deflection, x st , under mean wind
velocity.
Step (5) Calculate the variance, Eqn (4.116), and the peak factor, Eqn (4.112).
Step (6) Calculate the response due fluctuating wind speed from Eqn (4.111).
Step (7) Calculate the maximum response using Eqn (4.117).
69
Fig. (4.1): Assumed change in acceleration during a time step t in the linear acceleration method.
Fig. (4.2): Assumed change in acceleration during a time step t in the Wilson-  method.
Fig. (4.3): Assumed change in acceleration during a time step t in the constant acceleration method.
70
Fig. (4.4): Diagram showing the motion of a lumped-mass spring system with viscous damping ratio
  1.0
Fig. (4.5): variation of aerodynamic admittance factor A(n) with the reduced frequency nB / U (10) .
71
Chapter (5)
Static and dynamic analysis of circular cable roof nets
5.1 Introduction
Circular roofs are the most common forms of suspended cable roofs since its roundness
gives the architect the flexibility in both the planning and the usage of such structures.
Circular as well as other forms of cable suspended roofs carries only, adding to dynamic
loads, its own weight and a small live load to allow for repair the cladding.
Circular roofs in plan may be radial or of intersecting cables/cable beams.

Radial roofs include both simply suspended cable roofs, Fig. (2-1b), and pretensioned cable
beams roofs, Fig. (2-3). Intersecting of cables leads to a pretensioned cable net roof of
counter-stressed cables, Fig. (2-6c), while intersecting of cable beams leads to a
pretensioned cable grid of counter-stressed cable beams, Fig. (2-7).
The cables or cable beams in circular roofs are supported by an outer RC compression ring
and, for radial roofs, a small steel tension ring at the center. The design of such a roof
basically consists of the design of the cables and the design of a RC ring beam.
In this chapter, many examples are solved to examine the factors that affect the design of
cable suspended roofs, i.e. applying loading, cable rigidity ( EA ), cable pretension ( T0 ),
sag/span and rise/span ratios, the effect of hangars and the rigidity of the supporting
structure. The relations between the above factors affecting the design and the response of
such roofs to static loads are given in non-dimensional graphs or tables. Analyzing and
comparing these non-dimensional graphs and tables reach many useful relationships. These
relationships adding to the non-dimensional graphs or tables are the fundamentals of a good
technique for the preliminary design of circular cable suspended roofs, either nets or grids,
reached by the author.
The above tables and figures are made for ( 10  10 ) circular cable roofs, which mean that the
diametrical cable is divided into 10 segments.
Although, H.A.Buchholdt (5) mentioned that simplification of the mathematical model by

lumping cables or beams together is allowable, the new preliminary design method uses
formulae, instead of lumping, to express the deflections and forces of any (n  n) roof by the
guidance of ( 10  10 ) cable roof.
Natural frequencies for different types of circular cable roofs are calculated, using frequency
domain method, to examine the factors that affect these natural frequencies.
Finally, circular cable net, convex grid and concave grid are solved dynamically to compare
both static and dynamic responses for these types of cable roofs.
The work done in this thesis is made using computer programs in FORTRAN (61) based upon
the theories mentioned in chapters 3 and 4 using minimizing the total potential energy, either
static or dynamic. Also, another program for the frequency domain analysis is used to
calculate both eigenvalues and eigenvectors of some studied roofs.
72
5.2 Design of circular cable nets
5.2.1 Explanatory examples
In this section, many examples are carried out using the computer program that mentioned
before to reach and explain both the good technique of the preliminary analysis of circular
cable nets and the factors affect the analysis. For simplicity, it is assumed that the roofs are
with fixed supports (infinite rigid supports.)
Calculation of uniformly distributed load along cable length
Assuming that the roof (net or grid) is orthogonal in plan, Fig. (5.1), then the loaded area
that surround joint j is equal to s 2 , where s is the spacing between two successive cables
or cable beams. So, the total load that assumed to be concentrated at joint j is equal to
( w s 2 ) where w is the intensity of loading per unit area of the roof.
If this concentrated load is assumed to be uniformly distributed along the two s-length cable
segments intersected at joint j , then the uniformly distributed load per unit length of the
cable or cable beam is:
S
w  w  (5.1)
2
The pretension as an equivalent load
The pretension will be taken as an equivalent uniformly distributed load related to the
applying load (w) and varies from zero to (w) values.
Assuming that the pretension is equivalent to a load per unit length w , then the pretension
in all cables are assumed to be equal to:
wL2
T0  (5.2)
8d
Where: L is the roof diameter and d is the sagging of the roof.
Many cases of parametric studies are presented in the following sections:
Case (1):
Three examples are solved with different values for the parameters that affect the design or
the response of cable nets, which include: The diameter of the roof ( L ), the spacing between
nodes ( S  L / n ), the applying loading per unit area ( w ) or per unit length ( w ), the steel
area ( A ), the modulus of elasticity ( E ), the sag/span ratio ( d / L ), the pretension per unit
length of the cables ( w ), the maximum required central deflection ( z ) and the maximum
final tension in the cables ( Tmax ).
Two examples are of ( 10  10 ) spacing circular cable nets, Fig. (5.2), while the third is
(16  16) cable net with the data shown in Table (5.1). Sequent steel areas in the first two
examples are chosen so that the corresponding values of ( EA / wL2 ) are equal.
In the following, it will be considered that:
wLS 1
  (d / L)% ,   (Tmax / Tmin . break. )% and f (z )  
d z EA d z
EA( ) 2 ( ) n( 2 )( ) 2 ( )
L L wL L L
Where: Tmin . break. = The minimum (breaking) ultimate strength of cables. It may be taken
equal to an approximate value ( 14  A ).
Numbering of the cable net members according to their ends is given in Table (5.2).
The results for examples (1) and (2) in case (1) are the same and given in Table (5.3), while
the results of example (3) is given in Table (5.4).
73
The comparison between the obtained results for examples (1) and (2), shows that:
For two circular cable nets of the same arrangement (n  n) :
(1) At the same sag/span ratio ( d / L ) and the same ratio of pretension to the applying load
( w w ), the values ( z L ), (  ) and f (z ) in the two roofs, will be invariable if the
value ( EA / wL2 ) is constant in the both nets.
(2) The central deflection is always equal zero, if the equivalent U.D.L of the pretension is
equal to the applied load, i.e. w  w .
According to the above two notes and to the results of other examples:
The following conclusions will be generalized:
(a) Circular cable nets with the same arrangement ( n  n ) will act identically, i.e. ( z L )
and (  ) are invariable, under the following conditions:
1. The value ( EA / wL2 ) is constant.
2. The sag/span ratio ( d / L ) is constant.
3. The ratio of pretension to the applying load ( w w ) is constant.
(b) According to condition (a), it yields that the steel area is proportional to both the
applying load and the square of the roof diameter and is inversely proportional to its
modulus of elasticity:
A w (5.3)
A L 2
(5.4)
1
A (5.5)
E
(c) Circular cable nets with the same arrangement ( n  n ), that act identically, have a
wLS 1
constant value of the quantity ( f (z )   ), where n is
d z EA d z
EA( ) 2 ( ) n( 2 )( ) 2 ( )
L L wL L L
the number of segments in the diametrical member; i.e. n  L / S and S is the spacing
between nodes.
According to the this deduction, it yields that the steel area is inversely proportional to
both the required ratio of deflection to span, the square of the sag/span ratio, and to the
number of the segments in the diametrical member:
1
A (5.6)
(z / L)
1
A (5.7)
( d / L) 2
1
A (5.8)
n
z 1
 (5.9)
L ( d / L) 2
(d) If the equivalent value of pretension exceeds the applying load (w), deflection of the
central node will reflect to be upward, hence it is recommended to take the equivalent
load of pretension less than that of the applying load.
8T d
[ w  02 ]  w (5.10)
L
74
(e) Comparing the values of Tables (5.3) and (5.4), it yields that:
For two circular nets ( n1  n1 ) and ( n2  n2 ) with the same properties (diameter, steel
area, loading, modulus of elasticity and sag/span ratio), the relation between deflections
and tensions in the two roofs may be approximately expressed as:
n S
z 2  1  z1  2  z1 (5.11)
n2 S1
n S
(Tmax ) 2  1  (Tmax )1  2  (Tmax )1 (5.12)
n2 S1
These deductions are useful in making non-dimensional tables or graphs that will be used in
the preliminary design of circular cable nets as will be shown later.
5.2.2 The preliminary design of circular cable nets

The previous deductions are used in utilizing non-dimensional tables or graphs for a
specified arrangement ( n  n ) to the preliminary design of circular cable nets as follows:
1) Figs. (5.3) to (5.8) are non-dimensional graphs corresponding to the preliminary
design of ( 10  10 ) circular cable nets.
2) Table (5.3) is non-dimensional table for the preliminary design of ( 10  10 ) circular
cable nets, and it includes most, not all, of the data given by the above Figs.
3) According to relations (5.11) and (5.12), the response of cable nets of different
arrangements ( n  n ) may be estimated by using the above non-dimensional tables or
graphs as will be shown in case (2).
Case (2):
Table (5.5) includes the data of two cable nets (examples 4 and 5) with spacing (16  16) and
(8  8) respectively to confirm the obtained non-dimensional relations in case (1).
Preliminary design of example (4)

w  0.25  0.30  0.075 t / m
0.075  80 2
Using Eqn (5.2): T0   18.75 ton
8  3.2
(z) allowable  8000 / 250  32 cm
Using Eqn (5.11), the allowable deflection corresponding to ( 10  10 ) circular net is:
16
z1010  32   51.2 cm
10
z1010 51.2
( )%   100  0.64%
L 8000
Using Table (5.3), or Figs. (5.5) and (5.6), corresponding to the preliminary design of
( 10  10 ) circular net with 4% sag/span ratio and equivalent pretension w  0.25w , it
yields that for (z / L)%  0.64% :
EA
 65.52 and (Tmax . / Tmin . break. )%  26.578
wL2
 Areq  65.52  0.12  80 2 / 1663  30.26 cm 2
75
Tmax . T 10
Using Eqn (5.12): ( )1616  ( max . )1010   16.61%
Tbreak. Tbreak. 16
 Tmax .  0.1661 (30.26  14)  70.37 ton
Exactitude of the analysis

To examine the above results, the designed steel area (A) adding to the other characteristics
of the example are used as data in the computer program as follows:
Data
L= 80 m, S = 5 m, d = 3.2 m, w= 0.12 t/m2, T0 = 18.75 t, A = 30.26 cm2 and E = 1663 t/cm2.
Results
The results are:
z max .  33.19 cm  (z / L)  0.00414  0.414%
Tmax .
Tmax .  73.48 ton   73.48 /(30.26  14)  17.34%
Tmin . break.
These results show how much the accuracy of the used method of the preliminary design is.
Preliminary design of example (5)

w  0.35  0.20  0.07 t / m
0.07  32 2
Using Eqn (5.2): T0   9.33 ton
8  0.96
(z) allowable  3200 / 250  12.80 cm
8
z1010  12.8   10.24cm  0.32% L
10
Using Table (5.3), or Figs. (5.3) and (5.4), for the preliminary design of ( 10  10 ) circular
net with 3% sag/span ratio and T0 corresponding to 0.25w and 0.50w respectively, it yields
that for (z / L)%  0.32% :
For w  0.25w  ( EA / wL2 )  248.83 and (Tmax . / Tmin . break. )%  9.92
For w  0.50w  ( EA / wL2 )  160.48 and (Tmax . / Tmin . break. )%  14.92
Using the proportion:

EA
For w  0.35w   213.49 and (Tmax . / Tmin . break. )%  11.92
wL2
 Areq  213.49  0.10  32 2 / 1663  13.15 cm 2
Tmax . T 10
Using Eqn (5.12): ( )1616  ( max . )1010   14.90%
Tbreak. Tbreak. 8
 Tmax .  0.1490  (13.15  14)  27.43 ton
76
As done in the previous example, the designed steel area (A) adding to other characteristics
Data
L=32 m, S =4 m, d = 0.96 m, w =0.10 t/m2, T0 = 9.33 ton, A= 13.15 cm2 and E= 1663 t/cm2.
Results
The results are:
z max .  12.35 cm  (z / L)  0.00386  0.386%
Tmax .
Tmax .  25.14 ton   25.14 /(13.15  14)  13.66%
Tmin . break.
These exact results confirm the accuracy of this technique for the preliminary design.
5.2.3 Investigation of factors affect the response of cable nets

These factors that are mainly affect the stiffness of a cable structure include (5):
a) The curvature of the cables (sag/span ratio);
b) The cross-sectional area of the cables (the steel area);
c) The level of pretension; and
d) The stiffness of the boundary and supporting structure.
Generally, the data of Table (5.3), especially the values corresponding to   4% and
w / w  50% , are used to investigate the effect of the above factors.
a) Sag/span ratio (   d / L )
According to Eqs (5.8) and (5.10) and to Figs. (5.9) and (5.10), it is observed that:
1) Increasing ( ) , for the same steel area and same pretension, decreases the net
deflections and consequently decreases the final cable tensions.
2) Increasing ( ) , for the same required deflection, decreases the required steel area.
3) The effect of ( ) upon either the steel area or the displacements is quadratic. This
means a large effect upon the response of cable nets.
Generally, sag/span ratios between 4 and 6% will give satisfactory structural behavior. (5)
b) The cable cross sectional area

Figures (5.3) to (5.8), (5.11) and (5.12) and Equations (5.4), (5.7) and (5.9) showed that:
1) Increasing the steel area decreases the nodal displacements.
2) Increasing the applying load, for the same required deflections, requires an increasing
in the steel area.
3) Increasing the number of segments in the diametrical member (decreasing the spacing
between nodes), for the same diameter and the same required deflection leads to a
decreasing in the required steel area.
4) Increasing the steel area decreases the final tensions in the cables while increasing the
pretension, Fig. (5.4), has a very negligible effect.
This means that increasing the steel area, over the required value, decreases the utilizing
of the cable cross-sectional area and also means that the pretension has a negligible effect
in utilizing the cross-sectional area comparing with the steel area. In other words, the
steel area, not the pretension, is the decisive in carrying the applying loads.
77
As a general guide, the cross-sectional area of the cables is considered satisfactory, if the
maximum tension force in the cables is less than or equal to the half of their breaking
(ultimate) strength. (5)
c) The level of pretension

The results presented in Figs. (5.3) to (5.8) and Figs. (5.13) and (5.14) showed that:
1) Increasing the pretension force, for the same value of steel area, decreases the nodal
displacements.
2) Increasing the pretension has a negligible effect comparing with increasing the steel
area upon the relative final forces in cables.
Since the stiffening effect of increasing pretension is only marginal, the pretension should
only be of sufficient magnitude to prevent the cables from going slack under any maximum
specified load condition. (5) As mentioned before in Eqn (5.10), it is recommended that the
equivalent pretension is less than or equal to the applying load.
d) The stiffness of the boundary and support structure

Non-dimensional Tables (9.1) and (9.2) [ref. 5], indicates the response of circular saddle-
shaped (11  11) cable nets with concrete ring beam.
According to these tables:
1) Increasing the cross-sectional area of the ring beam leads to a decreasing in the
vertical deflection of the central joint.
2) Increasing the cross-sectional area of the ring beam leads to an increasing in the final
forces in the cables.
5.3 Design of circular cable grids

5.3.1 Explanatory examples
In this section, as done before for circular cable nets, some examples are solved using
the computer programs that mentioned before to explain both the good technique of the
preliminary analysis of circular cable grids and the factors affect the analysis. For simplicity,
it is assumed that the following roofs are with fixed supports (infinity rigid supports).
Case (3):
Six examples will be solved for different steel area values of either suspension or
pretensioned cables ( As or A p ) corresponding to a constant steel area value AConstan t for
either ( A p or As ) respectively. All cables are of modulus of elasticity E  1663 t / cm 2 where
ties and struts are of modulus of elasticity E  2100 t / cm 2 . For all examples (sag/span) and
(rise/span) ratios are taken equal to 4%, the uniformly distributed load is taken equal to 0.10
t/m2 and the pretension is taken equivalent to 80% of the uniformly distributed load.
The data assumptions are shown in Table (5.6).
Results of examples (6) and (7) are the same and shown in Tables (5.7) and (5.8).
Results of examples (8) and (9) are the same and shown in Tables (5.9) and (5.10).
Results of example (10) are shown in Table (5.11).
Results of example (11) are shown in Table (5.12).
According to the results of the previous examples shown in Tables (5.7) to (5.12), the
following deductions, which are similar to these of the circular cable nets, will be
generalized as follows.
78
Conclusions
a) Circular cable grids with the same arrangement ( n  n ) will act identically, i.e.
( z L ) and (  ) for both sagging and hogging cables are invariable, under the following
conditions:
1. The values ( E s As / wL2 & E p Ap / wL2 ) are constant.
2. The sag/span and rise /span ratios ( d / L & R / L ) are constant.
3. The ratios of pretension to the applying load ( ws w & wp w ) are constant.
b) Circular cable grids with the same arrangement ( n  n ), that act identically, have a
constant value of the quantities f (z s ) and f (z p ) where
wLS 1
f (z s )   and
d 2 z s E s As d 2 z s
E s As ( ) ( ) n( )( ) ( )
L L wL2 L L
wLS 1
f (z p )   .
R 2 z p E p Ap R 2 z p
E p Ap ( ) ( ) n( )( ) ( )
L L wL2 L L
Due to this, equations from (5.4) to (5.9) are still valid for both sagging and hogging cables.
These conclusions are useful in making non-dimensional tables or graphs that will be used in
the preliminary design of circular cable grids as will be shown later.
c) For two circular grids ( n1  n1 ) and ( n2  n2 ), either concave or convex, with the same
properties (diameter, steel area, loading, modulus of elasticity and sag (rise)/span ratio),
approximate relations (5.11) and (5.12) are still valid, for both the suspension and the
pretensioned cables.
Also, the forces in the hangers (struts or ties) are approximately expressed as:
( F / EA) hang2 n S
 ( 1 )2  ( 2 )2 (5.13)
( F / EA) hang1 n2 S1
This equation is satisfied as long as the steel area of the suspension cable is more than or
equals that of the pretensioned cable ( AS  AP ).
5.3.2 Factors affect the response of cable grids

These factors that are mainly affect the stiffness of a cable grid include:
a) The cross-sectional area of the cables and the hangers (ties or struts);
b) The curvature of the cables (sag/span and rise/span ratios);
c) The level of pretension in both sagging and hogging cables and the ties;
d) The number of the hangers (ties / struts) or the spacing between nodes; and
e) The stiffness of the boundary and supporting structure
Data of Tables (5.7) to (5.12) are used, as needed, to investigate the effect of the above
factors.
a) Cross sections of the cables and the hangers

According to the results of both examples (6) and (8) shown in Tables from (5.7) to (5.10),
Figures from (5.16) to (5.23) are drawn.
Area of cables against their deflections (Fig. 5.16)
1) Increasing of steel area of either suspension or pretensioned cables leads to
decreasing in the nodal deflections.
79
2) Cable grid, with the area of the suspension cable is more than the area of the
pretensioned cable, deflects less than cable grid with the area of the suspension cable is
less than the area of the pretensioned cable.
3) Deformations of the suspension cable is slightly more than these of the pretensioned
cable in the concave grid, while the deformations of the suspension cable is slightly less
than these of the pretensioned cable in the convex grid, see also Figs. (5.20) and (5.21).
Area of cables against their final tensions (Fig. 5.17)

4) Increasing the steel area of the suspension cable leads to a quick decreasing in the
ratio between the final tension and the minimum ultimate strength (  ) of the suspension
cable and leads to a slight increasing of this ratio in the pretensioned cable.
5) Increasing the steel area of the pretensioned cable leads to decreasing in the ratio (  )
for both cables. Rate of decreasing is much more for the case of the pretensioned cable.
6) The ratios (  ) of the suspension cables are always much more than these of the
pretensioned cables; see also Figure (5.23).
Area of cables against final forces of ties or struts (Figs 5.18 and 5.19)
7) Increasing the steel area of the suspension cable increases the final forces in the
hangers (compression of the convex grid struts and tension of the concave grid ties).
8) Increasing the steel area of the pretensioned cable decreases the final forces in the
hangers.
Area of ties/struts against deflections and forces (Figs 5.20  5.23)

9) Increasing the cross sectional area of the hangers either ties or struts has
approximately no effect upon the deformations and the final forces of the two cables.
10) Increasing the cross sectional area of the ties or struts decreases their final forces.
Conclusions
a. The suspension cable is the main-load carrying cable.
b. The cross section of the suspension cable has not to be less than this of the
pretensioned cable.
c. Increasing the steel area of the suspension cable leads to more efficiency in utilizing
the cross sections of both the pretensioned cable and the ties / or struts.
d. Since increasing the areas of the ties / struts has a slight effect upon the final
deflections and forces, it is economical to use the minimum cross section that allow the
distribution of forces in cable grids which assumed by Buchholdt (5), see Fig. (5.24).
b) The curvature of the cables (sag/span and rise/span ratios)

Examples (6) and (8) are solved for different values of sag/span ratio at rise/span ratio equal
to 4% and also solved for different values of rise/span ratio at sag /span ratio equal to 4%
and in all cases ( AS  AP  6.55 cm 2 ). The results are as shown in Figs. (5.25) to (5.29).
The first observation about these Figures that both convex and concave grids act identically
when they respond to the increasing of sag or rise / span ratios.
Increasing sag/span ratio against deflections (Fig. 5.25)
1) Increasing sag/span ratio causes the decreasing of the nodal deformations.
Increasing rise/span ratio against deflections (Fig. 5.27)

2) The deflections decrease either by increasing the rise over the sagging or by
decreasing the rise under the sagging.
80
3) This means that the case of maximum deflections is when ( Rise  Sag ).
4) So, it is more economic and logic to use ( Rise  Sag ).
5) The changes in deflections that happened due to the change in rise/span ratio are
relatively small. So, it is economically recommended to take rise/span ratio as minimum
as possible.
Increasing sag/span ratio against final cable forces (Fig. 5.26)

6) Increasing the sag/span ratio decreases the ratio (  ) in the suspension cables and
increases it in the pretensioned cables.
7) This means that increasing the sagging improves the utilizing of the cross sectional
area of the pretensioned cable.
Increasing rise/span ratio against final cable forces (Fig. 5.28)

8) Increasing the rise/span ratio decreases the ratio (  ) in the pretensioned cables. In
other wards, it decreases the utilizing of the entire cross sectional area of the
pretensioned cable.
9) The ratio (  ) in the suspension cables decrease either by increasing the rise over the
sagging or by decreasing the rise under the sagging.
10) This means that the case of maximum forces in the suspension cables is when
( Rise  Sag ).
11) The changes in the forces of the suspension cables are extremely small. So, the
efficiency of utilizing their cross sections does not approximately change.
12) This confirms that it is better to take rise/span ratio as minimum as possible.
Increasing sag or rise/span ratio against forces of ties or struts (Fig. 5.29)
13) Increasing the sagging or the rise increases the forces in the ties or struts. So, the
efficiency of utilizing their cross sections is increasing.
14) The forces in ties or struts when ( sag  rise ) are more than their forces when
( sag  rise ).
Conclusions
e. The sagging of the suspension cable is the principal factor comparing with the rise
of the pretensioned cable.
f. The sagging of the suspension cable has not to be less than the rise of the
pretensioned cable.
g. The rise of the pretensioned cable has to be taken as minimum as the grid reaches
the required stiffness (rigidity).
h. In concave cable roofs, the optimum values of sag and rise to span ratios are 5%
and 4%, respectively. (62) In convex cable roofs, the sag and rise to span ratios ranges
between 4% and 6%. (63) In concave-convex cable roofs, the optimum values of sag and
rise to span ratios are 5% and 3%, respectively. (64)
c) The level of pretension in cables and the ties

Examples (6) and (8) are solved for different values of pretension in both suspension and
pretensioned cables and also in ties of the concave grids.,( AS  AP  6.55 cm 2 ),
( TP or TS0  15 ton ) and ( Sag  Rise  4%L ). The results are as shown in Figs. (5.30) to
0
(5.36). These Figures show that both convex and concave grids act identically when they
respond to the increasing of pretension in suspension or pretensioned cables.
81
According to this, the relations between the pretension in one of the cables and the forces
and deflections in both the cables seem to be linear.
Increasing the pretension against deflections (Figs. 5.30, 5.32 & 5.36a)
1) Increasing the pretension of the pretensioned cable ( TP0 ) leads to an increasing in the
deflections of both suspension and pretensioned cables.
2) Increasing the pretension of the suspension cable ( TS0 ) leads to decreasing in the
deflections of both suspension and pretensioned cables.
3) This means that the suspension cable is the main load carrying cable and the act of the
pretensioned cable is to stiffen the structure, not to decrease the deformations.
4) Increasing the pretension in both the pretensioned and suspension cables, in the same
time, leads to a little decreasing in the deflections of both suspension and pretensioned
cables.
Increasing the pretension against final cable forces (Figs. 5.31, 5.33 and 5.36b)
5) Increasing the pretension in either the pretensioned or the suspension cables ( TP0 and
TS0 ) or in both of them leads to an increasing in the ratios between final forces in both
suspension and pretensioned cables and their minimum ultimate strength (  s and  p ).
6) The values of the relative final forces in the suspension cables are much more than
these of the pretensioned cables (  s <<  p ). This confirms that the suspension cable is
the main load carrying cable as mentioned before.
Increasing the pretension against the response of ties or struts (Figs. 5.34)
7) Increasing the pretension in either the pretensioned or the suspension cables ( TP and0
TS0 ) leads to an increasing of the final forces in both ties of concave grids and struts of
convex grids.
8) Final forces in struts or ties if ( TP < TS0 ) are more than forces if ( TP > TS0 ).
0 0
Increasing the pretension of ties against the response of the cables (Figs. 5.35)
9) Increasing the pretension of the ties ( Ttie0 ) has a very negligible effect upon the final
forces in both suspension and pretensioned cables.
10) The deflections of the cables, as long as the final forces of the ties, change up and
down, during the increasing of ( Ttie0 ), in a range that can be neglected.
Conclusions
i. The values of pretension in either the pretensioned or the suspension cables have
not to allow compression in the pretensioned cable.
j. The pretension of the pretensioned cable has to be as minimum as possible with the
value that allow no compression in it.
k. Increasing the pretension of the suspension cable is the economic and the efficient
way for decreasing the nodal deflections comparing with increasing the pretension in
both the cables in the same time.
82
d) Number of ties or struts or the spacing between nodes
Examples (6) & (8) are solved for different values of spacing between nodes,
AS  AP  6.55 cm 2 and Sag  Rise  4%L . Results are shown in Figs. (5.37) and (5.38).
As mentioned before, both convex and concave grids act identically when they respond to
change in the spacing between nodes (number of segments).
1) Increasing the number of segments (decreasing the spacing between nodes) decreases
the nodal deflections of both suspension and pretensioned cables.
2) Increasing the number of segments decreases (  s ), the relative final tension of the
suspension cable, and increases the relative final tension of the pretensioned cable (  p ).
3) Increasing the number of segments decreases the final forces of ties or struts.
e) The stiffness of the supporting structure (ring beam & columns)

Non-dimensional Tables (8.6) and (8.7) [ref. 5], indicates the response of (10  10) circular
convex grid, of equal sag and rise to span ratios, with concrete ring beam. Fixing the level of
pretension, sag and rise to span ratios, and the applying loading, the effect of increasing the
dimensions of the ring beam is as follows:
1) Deflections change, up and down, in a very small range that does not suit the
increasing of the ring beam dimensions and generally may be considered neglected.
2) Final forces in cables increase very slightly with increasing the ring beam dimensions.
This means that the ring beam of these types of structures only marginally affect their
stiffness. (5) In other words, cables not the supporting structure, which have the principal
effect upon the stiffness of the cable roof. Also, it means that assuming finite rigid supports
(without ring beam) is valid when making the preliminary design of the cable roofs.
f) The gap between sagging and hogging cables of concave grid

Example (8) is solved for different values of gap between the upper and lower cables,
( AS  AP  6.55 cm 2 ) and ( Sag  Rise  4%L ). The results are as shown in Fig. (5.39).
1) At gap equals zero; deflections of both upper and lower cables are minimum.
2) Increasing the gap to a certain value leads to a slight increasing of deflections. Over
this value, deflections vibrate up and down with negligible values.
3) Generally, changing the gap has approximately no effect on deflections and final
forces. However, it is recommended for the value of gap to be as minimum as possible to
increase the stiffness of the grid.
5.3.3 The preliminary design of circular cable grids

In the preliminary design of cable grids or beams, it is initially sufficient to estimate both the
maximum central deflections and the forces in the cables when the structure is subjected to:
(a) Pretension + self weight + uniformly distributed snow load (if exists); and
(b) Pretension only.
According to this, it is usually assumed that (5):
a) Uniformly distributed load can be replaced by an equivalent system of concentrated
loads acting at the joints.
b) The forces in all struts and /or ties are equal in magnitude when subjected to
pretension only.
c) The tensile forces in the pretensioned cables are reduced to zero when the beams are
subjected to the maximum loading conditions, since the level of pretension only
marginally affects the stiffness.
83
Fig. (5.24) represents the assumed force distribution in cable beams / grids under a system of
equal concentrated loads ( W  wa ).
When the beams are fully loaded it follows from the above assumptions that the forces in the
vertical members are equal to zero in concave beams and equal to (W) in convex beams.
Also, the magnitudes of the pretension forces are such that the numerical value of forces in
all the struts and ties are equal to ( W / 2 ) when the applied loads are removed.
According to assumption (a), it follows that the maximum horizontal components of the
cable tensions are given by:
H max  wL2 / 8(d S  d ) (5.14)
Where: d S  The maximum sag of the suspension cable in the unloaded (but
pretensioned) conditions; and
d  The deflection of the suspension cable from the pretensioned
equilibrium condition due to the load ( wL / 2 ).
Thus the maximum cable tensions are given by:
1 L2
Tmax  wL(1  )1 2 (5.15)
2 16(d s  d ) 2
From assumptions (b) and (c) it follows that the horizontal components of the pretension
forces in the suspension and pretensioning cables are given respectively by:
wL2 wL2
H S0   (5.16)
8(d S  d ) 16(d S  d )
d
H P0  H S0 ( S ) (5.17)
dP
Where: d P  The rise of the pretensioned cable.
The value ( d ) may be determined using Eqs (1.16) and (1.17a), while, in this case, the
value ( H ) in Eqn (1.17a) is calculated due to a uniform load equal to ( w / 2 ).
Alternatively, the maximum displacement may be assumed, and the maximum cable tension
and hence the size of cables required calculated using Eqn (5.15).
The corresponding pretension can then be determined using a non-dimensional graph.
Due to Eqn (5.16), the pretension in the suspension cable is approximately equivalent to a
uniformly distributed load equal to ( 0.5w ) while the value ( d ) can be neglected.
According to Eqs (5.16) & (5.17) and to the examples solved by the author, the relation
between the pretension of the suspension cables and the minimum final tension of the
pretensioned cables is as follows:
At wS0  0.5w  TPmin  0.0 (5.18a)
At wS0  0.5w  TPmin  0.0 (5.18b)
Since compression is not allowable in, it yields that the pretension of suspension cables has
to be in the following range:
wS0  0.5w (5.19)
84
Useful notes
The following notes, observed and examined by the author, are corresponding to an
equivalent pretension load in the suspension cables equal to ( 0.5w ) and its corresponding
pretension in the pretensioned cable due to Eqn (5.17).
1) For cable grid of ( sag  rise ): the deflections of both sagging and hogging cables are
approximately equal and equal these of a circular cable net with the same sag, equivalent
pretension and other characteristics. In such a case, final forces in suspension cables are
approximately equal to these of the corresponding cable net, while the final forces in the
pretensioned cables are nearly zero.
2) For cable grid with ( sag  rise ):
1
(z S  z P )  (z Sag  z Rise ) (5.20)
2
1
And TS  (TSag  TRise ) (5.21)
2
Where:
z S  The deflection of the suspension cable of the grid.
z P  The deflection of the pretensioned cable of the grid.
z Sag  The deflection of a cable net with sag equal to this of the suspension cable of
the grid.
z Rise  The deflection of a cable net with sag equal to the rise of the pretensioned
cable of the grid.
TS  The final tension in the suspension cable of the grid.
TSag  The final tension in a cable net with sag equal to this of the suspension cable
of the grid.
TRise  The final tension in a cable net with sag equal to the rise of the pretensioned
cable of the grid.
According to Eqn (5.9), it yields that:

z Sag Rise 2
 (5.22)
z Rise Sag 2
So, Eqn (5.20) may be rewritten as follows:
z Sag Sag 2
(z S  z P )  [ (1  )] (5.23)
2 Rise 2
Rise 2
Or z sag  2z all ( ) (5.24)
Sag 2  Rise 2
Where: z all  The allowable deflection in both suspension and pretensioned cables.
According to Eqs (5.22) and (5.24), it yields that:
Sag 2
z Rise  2z all ( ) (5.25)
Sag 2  Rise 2
According to the above notes, the non-dimensional tables or graphs for the preliminary
design of circular cable nets can be used in the preliminary design of circular cable grids as
will be shown in the following examples.
85
For convex grid
wS 2 wS
Astrut   2( ) (5.26)
fC fC
Where f C is the allowable compression stress of the struts.

For concave grid
wS 2 wS
Atie   2( ) (5.27)
ft ft
Where f t is the allowable tension stress of the ties.
Case (4):
Three examples for circular cable grids of spacing (16  16) , (10  10) and (20  20)
respectively with data assumptions given in Table (5.13) to confirm the obtained relations
for the preliminary design of circular cable grids. The plan of the (20  20) roof is as shown
in Fig. (5.40). Modulus of elasticity is assumed to be 165 KN / mm2 (1663 t/cm2).
Preliminary design of example (12):

w  D.L  L.L  0.12 t / m 2  w  0.12  (5 / 2)  0.30 t / m
w  0.50  0.30  0.15 t / m and (Sag = Rise = 3.20 m)
0.15  80 2
(z) allowable  8000 / 400  20 cm   T0   35.29 ton
8(3.2  0.2)
16 z 32
z1010  20   32 cm   ( 1010 )%   100  0.4%
10 L 8000
Using Table (5.3), or Figs. (5.5) and (5.6), corresponding to the preliminary design of
( 10  10 ) circular net with 4% sag/span ratio and T0 corresponding to w  0.5w , it yields
that for (z / L)%  0.4% :
 Areq  72.90  0.12  80 2 / 1663  33.67 cm 2
Tmax . T 10
Using Eqn (5.12): ( )1616  ( max . )1010   15.39%
Tbreak. Tbreak. 16
 Tmax .  0.1539  (33.67  14)  72.55 ton
If convex grid and assuming struts of steel pipes with working compression stress
f C  150 kg / cm 2 , it yields that the required cross-sectional area of the pipe is:
0.12(5) 2
Astrut   20 cm 2
0.15
86
If concave grid and assuming ties of steel cables with allowable tension stress
f t  1663 kg / cm 2 , it yields that the required cross-sectional area of the tie is:
0.12(5) 2
Atie   1.8 cm 2
1.663
To examine the above results, the designed steel area (A) adding to the other characteristics
Data
L = 80 m, S = 5 m, d S  d P  3.2 m, w = 0.12 t/m2, TS0  TP0  35.0ton
AS  AP  33.67 cm2 and E = 1663 t/cm2.
Results
The results for either concave or convex grid are:
Z S  Z P  20.40 cm  (z / L)  0.00255  0.255% =1/392
Tmax .
TS max .  75 ton   75 /(33.67  14)  15.90%
Tmin . break.
TP min  0.0
These results show how much the accuracy of the used method of the preliminary design is.

w  D.L  L.L  0.10 t / m 2  w  0.10  (3.2 / 2)  0.16 t / m
Sag = 1.60 m, Rise = 0.96 m, and w  0.08 t / m
(z) allowable  3200 / 250  12.80 cm
0.08  32 2
 TS0   5.926 ton
8  (1.6  0.128)
1.60
So, according to Eqn (5.17)  TP0  5.926   9.877 ton
0.96
32
Using Eqn (5.24): z Sag  2  12.80  ( )  6.78 cm
5 2  32
z 6.78
( )%   100  0.21%
L 3200
Using Table (5.3) or Figs. (5.7) and (5.8), corresponding to the preliminary design of
( 10  10 ) circular net with 5% sag/span ratio and w  0.50w , it yields that for
EA
(z / L)%  0.21% :  92.7 and (Tmax . / Tmin . break. )%  16.30%
wL2
 Areq  92.7  0.10  32 2 / 1663  5.7 cm 2
 TSag  0.163  (5.7  14)  13.0 ton
87
( 1010 ) circular net with 3% sag/span ratio and T0 corresponding to w  0.50w , it yields
EA
that for (  92.7 ): (z / L)%  0.493 & (Tmax / Tmin . break. )  24.3%
wL2
 z Rise  0.00493  3200  15.78 cm & TRise  0.243  (5.7  14)  19.39 ton
According to Eqs (5.20) and (5.21):

(z P  z S )  [0.5(6.78  15.78)  11.28 cm]
TS  [0.5(13  19.39)  16.20 ton]
If convex grid and assuming struts of steel pipes with working compression stress
f C  150 kg / cm 2 , it yields that the required cross-sectional area of the pipe is:
0.10(3.2) 2
Astrut   6.83 cm 2
0.15
If concave grid and assuming ties of steel cables with allowable tension stress
f t  1663 kg / cm 2 , it yields that the required cross-sectional area of the tie is:
0.10(3.2) 2
Atie   0.616 cm 2
1.663
Data
L = 32 m, S = 3.2 m, Sag =1.60 m, Rise = 0.96 m, w = 0.10 t/m2, TS0  5.926 ton ,
TP0  9.877 ton , AS = Ap = 5.7 cm2 and E = 1663 t/cm2.
Results
The results are approximately the same for both concave and convex grids as follow:
z S  z P  10.03 cm  (z / L)  0.00313  0.313%
TS max .
TS max .  15.6 ton   15.6 /(5.7  14)  19.55%
Tmin . break.
TP min .
TP min .  4.75 ton   4.75 /(5.7  14)  5.95%
Tmin . break.
The exact results confirm the efficiency of this preliminary method.
Notice
This example may be resolved with unequal steel area of both suspension and pretensioned
cables as follows:
52
Using Eqn (5.25): z Rise  2  12.80  ( 2 )  18.82 cm
5  32
z 18.82
 ( )%   100  0.588%
L 3200
88
( 10  10 ) circular net with 3% sag/span ratio and w equal 0.50w , it yields that for
EA
(z / L)%  0.588% :  72.85 and (Tmax . / Tmin . break. )%  30.12%
wL2
 AP req  72.83  0.10  32 2 / 1663  4.48 cm 2
Resolving the problem using the computer program for the previous data with
AS  5.7 cm 2 & AP  4.48 cm 2 , it yields that:
z S  z P  10.6 cm  (z / L)  0.00331  0.331%
Tmax .
TS max .  16.0 ton   16.0 /(5.7  14)  20.05%
Tmin . break.
TP min .
TP min .  5.67 ton   5.67 /( 4.49  14)  9.02%
Tmin . break.

w  D.L  L.L  0.10 t / m 2  w  0.10  (5 / 2)  0.25 t / m
Sag = 5.00 m, Rise =3.00 m, and w  0.125 t / m
(z ) allowable  40.00 cm
0.125  100 2
 TS0   28.94 ton
8  (5  0.40)
5
So, according to Eqn (5.17)  TP0  28.94   48.23 ton
3
Preliminary design of an equivalent ( 10  10 ) grid
20
z (1010) allowable  40   80.0cm
10
32
Using Eqn (5.24): z Sag  2  80  ( 2 )  42.35 cm
5  32
z 42.35
 ( )%   100  0.4235%
L 10000
Using Figs. (5.7) and (5.8), corresponding to the preliminary design of ( 10  10 ) circular net
with 5% sag/span ratio and T0 corresponding to 0.50w , it yields that for
(z / L)%  0.4235% :

EA
 48.50 and (Tmax . / Tmin . break. )%  32.65%
wL2
 Areq  48.50  0.10  100 2 / 1663  29.16 cm 2
& TSag  0.3265  (29.16  14)  133.31 ton
89
a) Assuming that (AS = AP)
Using Figs. (5.3) and (5.4), corresponding to the preliminary design of ( 1010 ) circular net
with 3% sag/span ratio and T0 corresponding to w  0.50w , it yields that for
EA
(  48.50 ): (z / L)%  0.792 & (Tmax / Tmin . break. )  46.0%
wL2
 z1010  0.00792  10000  79.20 cm & T1010  0.46  (29.16  14)  187.79 ton
10 10
 z Rise  79.20   39.60 cm & TRise  187.79   93.90 ton
20 20
According to Eqs (5.20) and (5.21):
(z P  z S )  [0.5(40.00  39.60)  39.80 cm]
TS  [0.5(133.31  93.90)  113.61 ton]

As done before, the designed steel area (A) adding to the other characteristics of the example
are used as data in the computer program as follows:
Data
L = 100m, S = 5.00 m, Sag =5.00 m, Rise = 3.00 m, w = 0.10 t/m2, TS0  28.94 ton ,
TP0  46.23 ton , A = 29.16 cm2 and E = 1663 t/cm2.
Results
The results are as follows:
z max S  28.84 cm  (z / L)  0.002884  0.2884% (Joint 1).
z max P  29.56 cm  (z / L)  0.002956  0.2956% (Joint 306).
Where joint (306) is the center of the convex part of the grid.
TS max .
TS max .  75.72 ton   75.72 /( 29.16  14)  18.55% (Member 573)
Tmin . break.
TS min .
TS min .  43.762 ton   43.762 /( 29.16  14)  10.72% (Member 582)
Tmin . break.
TP max .
TP max .  38.64 ton   38.64 /( 29.16  14)  9.47% (Member 1230)
Tmin . break.
TP min .
TP min .  20.89 ton   20.89 /( 29.16  14)  5.12% (Member 649)
Tmin . break.
Where members (649) and (1230) in the convex part are corresponding to members (1) and
(582) in the concave part respectively.
90
b) Assuming that ( AS  AP )
This example may be resolved with unequal steel area of both suspension and pretensioned
cables as follows:
20
z (1010) allowable  40   80.0cm
10
52
Using Eqn (5.25): z Rise  2  80  ( 2 )  117.65 cm
5  32
z 117.65
 ( )%   100  1.177%
L 10000
( 10  10 ) circular net with 3% sag/span ratio and T0 corresponding to 0.50w , it yields that
EA
for (z / L)%  1.177% :  23.82 and (Tmax . / Tmin . break. )%  84%
wL2
 AP req  23.82  0.10  100 2 / 1663  14.32 cm 2
Data
The previous data are the same except that AS  29.16 cm 2 & AP  14.32 cm 2
Results
The results are as follows:
z max S  33.29 cm  (z / L)  0.003329  0.3329% (Joint 1).
z max P  34.07 cm  (z / L)  0.003407  0.3407% (Joint 306).
TS max .
TS max .  81.135 ton   81.135 /( 29.16  14)  19.87% (Member 573)
Tmin . break.
TS min .
TS min .  44.182 ton   44.182 /( 29.16  14)  10.82% (Member 582)
Tmin . break.
TP max .
TP max .  42.44 ton   42.44 /(14.32  14)  21.17% (Member 1230)
Tmin . break.
TP min .
TP min .  32.41 ton   32.41 /(14.32  14)  16.17% (Member 649)
Tmin . break.
91
5.4 Factors affect natural frequencies of circular cable roofs
Using a computer program based on the frequency domain method mentioned in
chapter (4), [ref. 60], natural frequencies and mode-shapes of many circular cable roofs
(nets, concave grids and convex grids) are calculated. These natural frequencies are
recalculated for different values of sag/span ratio, rise/span ratio, spacing between the nodes,
the steel area of the sagging cable, the steel area of the hogging cable, the steel area of the
hangars, the pretension of the sagging cable, the pretension of the hogging cable and the
pretension of the hangars. To carry out the influence of any factor, it is considered that the
other factors are kept constant. The results are shown in non-dimensional graphs, Figs.
(5.41) to (5.56).
The analyzed roofs (nets, concave grids and convex grids) are of the following properties:
Diameter = 40.0m, Spacing = 4.0m, Sag = 1.60m (4%), Rise = 1.20m (3%), w = 0.1 t/m2,
TP 0  TS 0  TH 0  6.25 ton, steel area of all members = 13.92 cm2 and Height = 10.00m.
5.4.1 Natural frequencies of circular cable nets

The results presented in Figs. (5.41) to (5.45) showed that:
1. Increasing the applying uniformly distributed loads decreases the natural frequencies.
2. Increasing the pretension increases the natural frequencies.
3. Increasing the cable steel area does not change the first natural frequency, whereas
higher natural frequencies increase in a very slight rate that can be considered
unchanged also.
4. Increasing the sag/span ratio decreases the natural frequencies.
5. Increasing the spacing between nodes decreases the natural frequencies.
Figs. (5.46) and (5.47) represent the first two mode-shapes of the circular cable net with the
above properties.
5.4.2 Natural frequencies of circular cable grids

The results presented in Figs. (5.48) to (5.56) show that:
1. Increasing the sag/span ratio increases the natural frequencies of both concave and
convex grids in a slight rate.
2. Increasing the rise/span ratio has approximately no effect on convex grid, whereas it
increases the natural frequencies of the concave grid in a slight rate.
3. Increasing the spacing between hangars decreases the natural frequencies.
4. Increasing the steel area of one of the two cables increases the natural frequencies in
a very slight rate so that it has practically no effect. Also, hangars, either ties or
struts, have no effect on the natural frequencies.
5. Increasing the pretension of either the suspension or the pretensioned cable increases
the natural frequencies. The same thing will be happened when increasing the
pretension of the hangars, but in a less rate.
Conclusions:
1. Increasing the steel areas of cables or hangars has approximately no effect upon the
dynamic stability of the structure. This is due to their very small own weight
comparing with applying loading.
2. Increasing sag/span ratio of a cable net makes the structure more dynamically
excitable, whereas increasing sag or rise to span ratios in cable grids slightly
increases the dynamic stability of the structure.
3. The most efficiency and economic solution to make the circular cable roof more
dynamically stable, is to increase the pretension of the cables and the hangars.
92
5.5 Application of dynamic analysis using time-domain
Using a computer program based on minimizing the total dynamic potential energy as
mentioned before in chapter (4) and considering static deflections as deflections due to the
static part of the dynamic loading, three circular cable suspended roofs, i.e. net, convex grid
and concave grid, Fig. 5.15, are analyzed using the computer program. Numbering of cable
net members as shown in Table (5.2). The static data of the three roofs are as the same as in
section 5.4. Time of analysis is taken as “1” minute and time step as 0.01 sec., so number of
time steps is 6000. Damping ratio is assumed as 0.014. Natural frequencies are calculated
using the computer program of frequency domain analysis as 0.4911, 0.7384 and 0.7584
(c.p.s) for cable net, convex grid and concave grid respectively.
Resulting relations are shown in Figs 5.57 to 5.68
5.5.1 Deflection vs. time Figs. (5.57) and (5.58)

Dynamic deflections increase very quickly in a very small time to an approximately average
value, see Fig. (5.57). After this, deflections vibrate up and down in a small rate. It is
obvious that the amplitude of the vibration of a joint deflection is inversely proportional to
the location of the joint to the center. Fig. (5.58) represents the dynamic deflections of the
central joint in cable net, convex grid and concave grid. At time equal zero (static response)
the deflections of the cable net are more than these of the convex grid, which are more than
these of the concave grid, i.e. z net  z convex  z concave .
In the beginning of the time of analysis, for a few seconds, it was mentioned that
z convex  z net  z concave . With passing the time of analysis, this relationship was changed
to be z net  z convex  z concave and finally to the form z net  z concave  z convex .
Generally this indicates that convex grid is dynamically better than concave grid and both
are better than cable nets .
5.5.2 Velocity vs. time Figs. (5.59) to (5.61)

Velocities vibrate about an average value, Fig. (5.59). The amplitude of the vibration of joint
velocity is inversely proportional to the location of the joint to the center. Fig. (5.60) shows
the velocities of the central joints in both convex and concave grids corresponding to the
first second of the time of analysis whereas Fig. (5.61) is corresponding to the last second in
the time of analysis. These two figures indicate that the velocities are going to be regular by
the end of the time of analysis and that the unloaded cable vibrate much more than the
loaded cable which seems, by the end of the time of analysis, to have no vibration.
5.5.3 Acceleration vs. time Fig. (5.62)

Accelerations vibrate about an average value. The amplitude of the vibration of joint
acceleration is inversely proportional to the location of the joint to the center.
Figs. (5-63) to (5.65) present respectively deflection, velocity and acceleration of the central
joint in the studied cable net corresponding to the last second of the time of analysis.
5.5.4 Lengths of members vs. time Fig. (5.66)

At the beginning of the time of analysis, lengths of members increase. With going the time,
lengths of the members near the roof axis, i.e. members (1-9-16), decrease whereas lengths
of the members far the axis and near the roof edge, i.e. members (121-123-125), increase.
93
5.5.5 Tensions in members vs. time Fig. (5.67)
Values of the tensions vibrate up and down. The amplitude of vibration increases for the
members near the roof center. On the contrary of the static analysis, the values of tensions in
members that lie on the same cable increase when going far away the center and toward the
roof edge.
Fig. (5.68) shows lengths and tensions of suspension and pretensioned cables of circular
convex grid where members (157), (167) and (277) on the convex part of the roof are
corresponding to members (1), (11) and (121) of the concave part of the roof.
It is obvious that the values of two opposite members are vibrating in two opposite
directions.
5.6 Application on the approach of the wind response

Data and results
Mean velocity at 10m height is 25m/s and the structure is in suburban area.
The structure is as example (14) with height equal to 20m.
According to Table (2.2), z 0  0.20  0.40m (taken 0.30m)
25
Using Eqn (2.5): u*   2.85 m/s
2.5 ln(10 / 0.3)
The mean wind speed at the average height of the structure is given by Eqn (2.4) as:
U (20)  2.5  2.85 ln(20 / 0.3)  29.94 m / s
The fundamental natural frequency is given as:
 6.056284
n0  0   0.96 Hz
2 2
The spectral density function is calculated using Eqn (2.15) as follows:
20  0.96
f ( z , n)   0.64
29.94
200  2.85 2  0.64
S u ( z , n)   3.19 m 2 / s
0.96[1  50  0.64] 53
The reduced frequency nD / U (10)  0.96 100 / 25  3.84Hz

Hence, using Fig. (4.5), the aerodynamic admittance factor A(n)  0.1
The maximum static response at the center, assuming that ( AS  AP ), is xst  28.84cm
Assuming that the static damping ratio is 0.014 and using Eqn (4.116), the variance is
expressed as:
0.2884 2   0.96
 x2    0.1  3.19  0.00638m 2
29.94 2 0.014
 x  0.08m
Using Eqn (4.112), the peak factor is expressed as:
0.577
  2 ln(0.96  60)   3.05
2 ln(0.96  60)
xmax  xst   x  0.2884  3.05  0.08  0.5319m
In general, frequency domain approach leads to a maximum response larger than that
obtained in time domain analysis.
94
Table (5.1): Data assumptions for case study (1):
Ex. Types of Diameter (D.L+L.L)
2
w Group of areas Sag/Span
arrangement (m) kg/m (t/m) (cm2) % ratios
5.031, 5.722, 6.55, 7.64, 9.058,
1 10  10 40.00 100.00 0.20 11.03, 13.92, 18.48, 27.07, 50.10 3%,
16.98, 19.31, 22.11, 25.79, 30.57, 4% &
2 10  10 60.00 150.00 0.45 37.23, 46.98, 62.37, 91.36, 169.10 5%
5.031, 5.722, 6.55, 7.64, 9.058,
3 16  16 40.00 100.00 0.125 11.03, 13.92, 18.48, 27.07, 50.10 4%
Table (5.2): Numbering of members of the cable net shown in Fig. (5.2):
Mem 1st 2nd Mem 1st 2nd Mem 1st 2nd Mem 1st 2nd Mem 1st 2nd Mem 1st 2nd
end end end end end end end end end end end end
1 1 2 27 34 35 53 61 62 79 2 24 105 50 54 131 27 80
2 2 3 28 35 36 54 62 63 80 3 25 106 52 55 132 32 81
3 3 4 29 37 38 55 63 64 81 4 26 107 53 56 133 36 82
4 4 5 30 38 39 56 33 65 82 5 27 108 40 57 134 39 83
5 6 7 31 1 40 57 65 66 83 23 28 109 41 58 135 60 84
6 7 8 32 40 41 58 66 67 84 24 29 110 42 59 136 64 85
7 8 9 33 41 42 59 37 68 85 25 30 111 43 60 137 67 86
8 9 10 34 42 43 60 68 69 86 26 31 112 57 61 138 69 87
9 11 12 35 6 44 61 1 6 87 27 32 113 58 62 139 20 88
10 12 13 36 44 45 62 2 7 88 28 33 114 59 63 140 21 89
11 13 14 37 45 46 63 3 8 89 29 34 115 60 64 141 22 90
12 14 15 38 46 47 64 4 9 90 30 35 116 61 65 142 19 74
13 16 17 39 11 48 65 5 10 91 31 36 117 62 66 143 15 73
14 17 18 40 48 49 66 6 11 92 33 37 118 63 67 144 37 91
15 18 19 41 49 50 67 7 12 93 34 38 119 65 68 145 38 92
16 20 21 42 50 51 68 8 13 94 35 39 120 66 69 146 39 93
17 21 22 43 16 52 69 9 14 95 40 44 121 5 70 147 36 83
18 23 24 44 52 53 70 10 15 96 41 45 122 10 71 148 32 82
19 24 25 45 53 54 71 11 16 97 42 46 123 15 72 149 55 94
20 25 26 46 20 55 72 12 17 98 43 47 124 19 73 150 56 95
21 26 27 47 55 56 73 13 18 99 44 48 125 22 74 151 54 79
22 28 29 48 23 57 74 14 19 100 45 49 126 43 75 152 51 78
23 29 30 49 57 58 75 16 20 101 46 50 127 47 76 153 68 96
24 30 31 50 58 59 76 17 21 102 47 51 128 51 77 154 69 97
25 31 32 51 59 60 77 18 22 103 48 52 129 54 78 155 67 87
26 33 34 52 28 61 78 1 23 104 49 53 130 56 79 156 64 86
95
Table (5.3): Non-dimensional table for circular cable net with (10  10) spacing
Equivalent EA   3%   4%   5%
pretension
wL 2
z z z
w / w % % f ( z ) % % f ( z ) % % f (z )
L L L
52.291 l.42 36.00 149.7 1.00 31.66 119.5 0.704 27.59 108.7
59.473 1.30 32.94 143.9 0.90 28.34 116.8 0.624 24.53 107.8
68.079 1.18 29.14 138.4 0.80 25.19 114.8 0.548 21.64 107.2
79.408 1.054 25.70 132.8 0.70 22.00 112.4 0.471 18.74 107.0
0%
94.147 0.927 22.32 127.3 0.60 18.87 110.6 0.396 15.94 107.3
114.64 0.794 18.91 122.0 0.50 15.77 109.0 0.323 13.20 108.2
144.68 0.656 15.49 117.0 0.40 12.71 108.0 0.251 10.53 110.1
192.08 0.514 12.07 112.6 0.30 9.72 108.5 0.183 7.974 113.6
281.36 0.362 8.54 109.3 0.20 6.72 111.1 0.119 5.457 120.0
520.73 0.194 4.78 109.8 0.10 3.66 120.0 0.058 2.941 133.0
52.291 1.085 37.75 195.8 0.770 32.37 155.3 0.541 27.83 141.5
59.473 0.995 33.93 187.9 0.692 28.89 151.8 0.479 24.69 140.5
68.079 0.905 30.31 180.4 0.618 25.60 148.7 0.420 21.64 139.9
79.408 0.810 26.61 172.9 0.540 22.28 145.9 0.360 18.79 139.8
25 %
94.147 0.713 23.00 165.6 0.463 19.07 143.5 0.303 15.97 140.5
114.64 0.611 19.39 158.6 0.385 15.89 141.8 0.246 13.20 142.0
144.68 0.505 15.79 152.1 0.306 12.76 141.1 0.191 10.51 144.7
192.08 0.395 12.23 146.5 0.229 9.73 141.9 0.139 7.947 149.8
281.36 0.277 8.60 142.6 0.152 6.71 146.1 0.090 5.436 158.8
520.73 0.148 4.78 144.2 0.076 3.65 159.0 0.044 2.931 176.6
52.291 0.733 40.11 290.1 0.527 33.27 226.9 0.370 28.11 206.6
59.473 0.674 35.84 277.4 0.474 29.57 221.7 0.328 24.88 205.4
68.079 0.615 31.83 265.6 0.423 26.12 217.3 0.287 21.86 204.9
79.408 0.551 27.78 253.8 0.369 22.63 213.2 0.246 18.86 205.0
50 %
94.147 0.486 23.86 242.8 0.316 19.30 209.9 0.206 15.98 206.2
114.64 0.418 19.98 232.1 0.263 16.01 207.7 0.167 13.18 208.9
144.68 0.345 16.16 222.4 0.209 12.82 206.9 0.129 10.49 213.9
192.08 0.270 12.43 214.4 0.156 9.74 208.9 0.092 7.92 227.0
281.36 0.189 8.67 209.2 0.103 6.70 215.7 0.060 5.415 236.0
520.73 0.100 4.78 212.8 0.051 3.63 235.3 0.029 2.921 262.6
52.291 0.369 43.30 576.6 0.272 34.40 439.8 0.192 28.45 398.9
59.473 0.341 38.42 548.7 0.245 30.43 429.8 0.170 25.0 396.8
68.079 0.312 33.86 523.1 0.218 26.75 420.6 0.148 22.00 396.3
79.408 0.281 29.32 497.5 0.191 23.07 413.2 0.1237 18.93 397.4
75 %
94.147 0.249 24.98 474.4 0.163 19.58 406.7 0.106 16.00 400.8
114.64 0.214 20.74 452.4 0.135 16.18 403.1 0.086 13.18 406.9
144.68 0.178 16.62 432.7 0.107 12.88 402.8 0.064 10.46 432.0
192.08 0.139 12.67 416.9 0.080 9.752 408.0 0.048 7.893 436.1
281.36 0.097 8.75 407.1 0.053 6.684 423.1 0.031 5.391 466.1
520.73 0.051 4.78 416.3 0.026 3.618 466.1 0.015 2.91 524.3
96
Table 5.4: (16  16) Circular net with  = 4% and [ w  0.25w ]
( EA / wL2 ) 52.291 59.473 68.079 79.408 94.147 114.64 144.68 192.08 281.36 520.73
(z / L)% 0.511 0.455 0.400 0.345 0.291 0.237 0.185 0.135 0.086 0.041
 % 21.35 18.97 16.74 14.49 12.34 10.22 8.164 6.184 4.238 2.287
f (z ) 155.3 151.8 148.7 145.9 143.5 141.8 141.1 141.9 146.1 159.0
Table (5. 5): Data assumptions for case study (2).

Ex. Arrangement Diam. Spacing D.L L.L w w w z allow E
(m) (m ) kg/m2 kg/m2 t/m2 Sag t/cm2
t/m
w L
4 16  16 80.00 5.0 80 40 0.12 0.30 4% 25% 1/250
5 32.00 4.0 70 30 0.10 0.20 3% 35% 1/250 1663
8 8
Table (5.6): Data assumptions for case study (3).

Type and Diam w Group of areas ( As or A p ) (cm2) T0 Atie &
Ex. arrangement (m) (t/m’) (ton) Astrut
( 10  10 ) Convex 3, 5, 6.55, 8.0, 10.0 &12.0 Steel pipes
6 Fig. (5.15a) 40.00 0.20 AConstan t = 6.55 20 8.40 cm2
( 10  10 ) Convex 6.75, 11.25, 14.74, 18.0, 22.5 & 27.0 Steel pipes
7 Fig. (5.15a) 60.00 0.30 AConstan t =14.74 45 18.90 cm2
( 10  10 ) Concave Steel cable

8 Fig. (5.15b) 40.00 0.20 3, 5, 6.55, 8.0, 10.0 &12.0 20 0.968 cm2
f S  f p  0.40m AConstan t = 6.55 T0 =3.0 t
( 10  10 ) Concave Steel cable
9 Fig. (5.15b) 60.00 0.30 6.75, 11.25, 14.74, 18.0, 22.5 & 27.0 45 2.178 cm2
f S  f p  0.60m AConstan t =14.74 T0 =6.75 t
3, 5, 6.55, 8.0, 10.0 &12.0 Steel pipes
10 ( 16  16 ) Convex 40.00 0.125 AConstan t = 6.55 12.5 8.40 cm2
Steel cable
11 ( 16  16 ) Concave 40.00 0.125 3, 5, 6.55, 8.0, 10.0 &12.0 12.5 0.968 cm2
f S  f p  0.40m AConstan t = 6.55 T0 =3.0 t
97
Table 5.7: Example (6)
(10  10) Convex grid with sag = rise = 4% and [ w  0.80w ]
As cm2 3.00 5.00 6.55 8.00 10.00 12.00
( EAs / wL2 ) 31.18 51.97 68.08 83.15 103.94 124.73
(z s / L)% 0.6265 0.51 0.4475 0.402 0.3533 0.3153
(z p / L)% 0.6275 0.511 0.4485 0.403 0.3543 0.3163
s % 66.5 43.786 35.136 29.830 24.821 21.321
p % 6.368 8.877 10.262 11.287 12.421 13.293
(C / A) strut 0.2488 0.2607 0.2679 0.2738 0.2798 0.2845

A p cm2 3.00 5.00 6.55 8.00 10.00 12.00
( EA p / wL2 ) 31.18 51.97 68.08 83.15 103.94 124.73
(z s / L)% 0.5588 0.492 0.4475 0.4108 0.367 0.3298
(z p / L)% 0.5598 0.493 0.4485 0.4118 0.3678 0.3308
s % 38.146 36.227 35.136 34.155 32.988 32.007
p % 33.69 16.07 10.26 7.098 4.479 2.923
(C / A) strut 0.288 0.275 0.268 0.262 0.255 0.249

(10  10) Concave grid with sag = rise = 4% and [ w  0.80w ]
As cm2 3.00 5.00 6.55 8.00 10.00 12.00
( EAs / wL2 ) 31.18 51.97 68.08 83.15 103.94 124.73
(z s / L)% 0.628 0.510 0.450 0.406 0.358 0.320
(z p / L)% 0.624 0.507 0.447 0.403 0.355 0.317
s % 66.75 44.00 35.30 30.00 24.95 21.45
p % 6.55 9.08 10.44 11.44 12.57 13.46
(T / EA) tie 3.13 E-4 3.65 E-4 4.09 E-4 4.36 E-4 4.65 E-4 4.90 E-4
98
A p cm2 3.00 5.00 6.55 8.00 10.00 12.00
( EA p / wL2 ) 31.18 51.97 68.08 83.15 103.94 124.73
(z s / L)% 0.562 0.495 0.450 0.413 0.369 0.332
(z p / L)% 0.559 0.492 0.447 0.410 0.366 0.329
s % 38.23 36.47 35.30 34.34 33.18 32.20
p % 34.02 16.30 10.45 7.24 4.61 3.03
(T / EA) tie 5.168 E-4 4.46 E-4 4.09 E-4 3.78 E-4 3.42 E-4 3.06 E-4

As cm2 8.00 10.00 12.00
( EAs / wL2 ) 83.15 103.94 124.73
(z s / L)% 0.2539 0.222 0.197
(z p / L)% 0.2543 0.222 0.197
s % 19.052 15.856 13.617
p % 6.49 7.29 7.926
(C / A) strut 0.1063 0.1085 0.1102

As cm2 3.00 6.55 8.00
( EAs / wL2 ) 31.18 68.08 83.15
(z s / L)% 0.254 0.223 0.198
(z p / L)% 0.2508 0.2193 0.1942
s % 19.21 15.98 13.76
p % 6.655 7.462 8.10
(T / EA) tie 1.794 E-4 1.9058 E-4 2.05 E-4
Table (5.13): Data assumptions for case study (4)

Sag/span Equivalent
Example The grid Diam S D.L. L.L. & pretension Allowable
arrangement (m) (m) (t/m2) (t/m2) Rise/span w / w deflection
12 ( 16  16 ) 80.00 5.0 0.08 0.04 Sag =4% 50 % (1/400) of
Rise = 4% the span
13 (10  10) 32.00 3.20 0.07 0.03 Sag =5% 50 % (1/250) of
Rise = 3% the span
14 (20  20) 100.0 5.00 0.065 0.035 Sag =5% 50 % (1/250) of
Fig. (5.40) Rise =3% the span
99
Fig. (5.1): Calculation of uniformly distributed load per unit length of the cable.
Fig. (5.2): Circular cable net with ( 10  10 ) spacing showing the numbering of joints.
100
Fig. (5.3): Non-dimensional relation between the central deflection and other parameters in (10  10) circular
cable net with sag/span ratio of 3%.
Fig. (5.4): Non-dimensional relation between the maximum cable force and other parameters in
(10  10) circular cable net with sag/span ratio of 3%.
101
102
103
Fig. (5.9): The relation between the central deflection and other parameters for different sag/span ratios.
Fig. (5.10): The relation between the maximum cable force and other parameters for different sag/span ratios.
104
Fig. (5.11): The relation between the central deflection and cable steel area for different values
of applying loads.
Fig. (5.12): The relation between the central deflection and cable steel area for different values of the spacing
between nodes.
105
Fig. (5.13): The relation between the central deflection and the pretension in cables.
Fig. (5.14): The relation between the maximum cable force and the pretension in cables.
106
Fig. (5.15): (10  10) Circular cable roof.
107
Fig. (5.16): The relation between the nodal deflection and the steel area of either suspension or pretensioned
cables in both convex and concave grids.
Fig. (5.17): The relation between the maximum cable forces and the steel area of either suspension or
pretensioned cables in both convex and concave grids.
108
Fig. (5.18): The relation between forces in struts and the steel area of either suspension or pretensioned cables
in a convex grid.
Fig. (5.19): The relation between forces in ties and the steel area of either suspension or pretensioned cables in
a concave grid.
Fig. (5.20): The relation between the strut steal area and the nodal deflections.
109
Fig. (5.21): The relation between the steel area of ties and the nodal deflections.
Fig. (5.22): The relation between the steel area of ties or struts and the forces in cables.
Fig. (5.23): The relation between the steel area of either ties or struts and their final forces.
110
Fig. (5.24): The assumed distribution of forces in cable beams / grids for preliminary design calculations.
Fig. (5.25): The relation between the sag / span ratio and the nodal deflections of both suspension and
pretensioned cables in either concave or convex grids.
111
Fig. (5.26): The relation between the sag / span ratio and the maximum forces in both suspension and
Fig. (5.27): The relation between the rise / span ratio and the nodal deflections of both suspension and
Fig. (5.28): The relation between the rise / span ratio and the maximum forces in both suspension and
112
Fig. (5.29): The relation between (sag / span) or (rise / span) ratios and the maximum forces in struts or ties.
Fig. (5.30): The relation between the pretension of the pretensioned cable and the nodal deflections of both
suspension and pretensioned cable in either concave or convex grids.
Fig. (5.31): The relation between the pretension of the pretensioned cable and the maximum forces in both
113
Fig. (5.32): The relation between the pretension of the suspension cable and the nodal deflections of both
Fig. (5.33): The relation between the pretension of the suspension cable and the maximum forces in both
Fig. (5.34): The relation between the pretension of the suspension or the pretensioned cables and the maximum
forces in struts or ties.
114
(a) (b)
(c)
Fig. (5.35): The relation between the pretension in ties and:

(a) the nodal deflections of both suspension and pretensioned cables,
(b) the maximum forces in both suspension and pretensioned cables, and
(c) the maximum forces in ties.
(a)
(b)
Fig. (5.36): The relation between the concurrent increasing of pretension in both suspension and pretensioned
cables and: (a) their nodal deflections and (b) their maximum forces.
115
(a) (b)
Fig. (5.37): The relation between the number of segments in the diametrical member and:
(a) the nodal deflections of both suspension and pretensioned cables in either concave or convex grids;
(b) the maximum forces of both suspension and pretensioned cables in either concave or convex grids.
Fig. (5.38): The relation between the number of segments in the diametrical member and the maximum forces
in ties or struts.
(a)
(b)
Fig. (5.39): The relation between the gap that lies between both the suspension and the pretensioned cables
of a concave grid and:
(a) the nodal deflections of both the suspension and the pretensioned cables of the grid, and
116
(b) the maximum forces in both the suspension and the pretensioned cables of the grid.
Fig. (5.40): A plan of the concave part of (20  20) circular cable convex grid roof.
117
Fig. (5.41): Effect of increasing the applying load upon frequencies.
Fig. (5.42): Effect of increasing the pretension on frequencies in a circular cable net.
Fig. (5.43): Effect of increasing cable steel area on frequencies in a circular cable net.
118
Fig. (5.44): Effect of increasing sag/span ratio on frequencies in a circular cable net.
Fig. (5.45): Effect of increasing the spacing between nodes on frequencies in a circular cable net.
Fig. (5.46): The first mode-shape of a circular cable net.
119
Fig. (5.47): The second mode-shape of a circular cable net.
Fig. (5.48): Effect of the sag/span ratio on natural frequencies of circular cable grid.
Fig. (5.49): Effect of the rise/span ratio on natural frequencies of circular cable grid.
120
Fig. (5.50): Effect of the spacing between hangars and natural frequencies of circular cable grid.
Fig. (5.51): Effect of the steel area of the suspension cable on natural frequencies of circular cable grid.
Fig. (5.52): Effect of the steel area of the pretensioned cable on natural frequencies of circular cable grid.
121
Fig. (5.53): Effect of the steel area of the hangars on natural frequencies of circular cable grid.
Fig. (5.54): Effect of the pretension of the suspension cable on natural frequencies of circular cable grid.
Fig. (5.55): Effect of the pretension of the pretensioned cable on natural frequencies of circular cable grid.
122
Fig. (5.56): Effect of the pretension of the hangars on natural frequencies of circular cable grid.
Fig. (5.57): Nodal deflections during time of analysis in circular cable net.
123
Fig. (5.58): Comparison between central deflections during the time of analysis in cable net, convex grid and
concave grid.
Fig. (5.59): Nodal velocities during time of analysis in circular cable net.
Fig. (5.60): Velocity of the central joint in the suspension cable of a circular grid during the first second of the
time of analysis.
124
Fig. (5.61): Velocity of the central joint in the suspension cable of a circular grid during the last second of the
time of analysis.
Fig. (5.62): Nodal accelerations during time of analysis in circular cable net.
125
Fig. (5.63): Central deflections in circular cable net during the last second of the time of analysis.
Fig. (5.64): Central velocities in circular cable net during the last second of the time of analysis.
126
Fig. (5.65): Central accelerations in circular cable net during the last second of the time of analysis.
Fig. (5.66): Lengths of members during the time of analysis in circular cable net.
Fig. (5.67): Final tensions in members during the time of analysis in circular cable net.
(a)
127
(b)
(c)
Fig. (5.68): Comparison between lengths and tensions in suspension and pretensioned cables in circular
convex grid.
128
Chapter (6)
Summary and conclusion
6.1 Summary
Cable suspension roofs are mainly used for covering large span areas without the need for
interior columns. Generally, cable roofs are the economic solution for long span roofs and also
for temporarily roofing structures. However, because of their flexibility and lightweight, some
factors must be taken into account in analysis and design, which may be neglected for roofs
with conventional framing. Several attempts have been made for better understanding of the
general behavior of cable suspended roofs and getting simplified methods to solve the different
types of these roofs.
The main object of the thesis is to investigate the analysis of circular cable suspended
roofs. To fulfill this object many investigations were carried out. The work done in this thesis
may be summarized as follows:
1) A general review of the cables, which are the basic elements of cable suspended roofs, is
presented. This review includes the classification, pretensioning, protection and mechanical
properties of single cables. Also, the static analysis of freely hanging cable under uniformly
distributed load is included.
2) The classifications, main elements and the materials of cable roofs are outlined.
Development of suspension roofs from the view of design and construction is also
presented. Advantages and disadvantages of cable suspension roofs are included. Static and
dynamic loads, that affect the design and the response of cable roofs, are presented. Wind is
presented in more detail (i.e. its nature, wind speed, statistical properties, wind pressure and
wind loading) because it is the principal dynamic load affecting cable roofs. The methods of
the analysis of cable suspended roofs are presented with a brief review for the previous
work done in this field.
3) The development of the method of minimizing the total potential energy (TPE) using the
conjugate gradient (CG) technique to predict the static response of cable roofs is presented.
4) Methods of dynamic analysis of cable suspended roofs are presented. Time domain
analysis by expanding the method of minimizing total potential energy to include dynamic
loads and responses is completely presented. Frequency domain analysis is also presented
because it does not need any generations of wind histories. An approach to evaluate the
dynamic response due to wind is included.
5) Many circular nets and grids have been analyzed using a computer program based upon
the minimization of (TPE) by the (CG) technique. These results are used to investigate the
factors that affect the design and the response of cable suspended roofs. Also, these results
have been used in making non-dimensional tables and graphs and to infer new relationships.
These tables, graphs and relationships form a good technique for the preliminary design of
circular cable suspended roofs either nets or grids. This good preliminary technique was
examined using several examples and the results are in good agreement with these solved
using the computer program based on minimizing the (TPE) using the (CG) technique for
the analysis of cable nets. Also, computer programs for the dynamic analysis were used to
investigate the factors that affect the frequencies of circular cable suspended roofs and to
analyze circular cable net, convex grid and concave grid to express their dynamic behavior.
128
6.2 Conclusions
The investigations built on the factors that affect the behavior of the circular cable
suspended roofs presented in this thesis, have led to the following conclusions:
For cable nets

1) Increasing sag/span ratio, steel area of cables, cross-sectional area of the ring
beam and finally, the pretension decrease nodal deflections.
2) Decreasing the spacing between nodes causes the uniformly distributed load per
unit length of the cables to be decreased and hence both the nodal deflection and
the required steel area decrease.
3) Comparing with increasing the steel area, increasing the pretension force has a
negligible effect upon the values of final forces in the cables. This means that the
pretension has a marginal effect upon the stiffness of cable roofs and that the
steel area, not the pretension, is more decisive in carrying the applying loads.
4) It is recommended that the value of the equivalent pretension (w) is less than or
equal to the uniformly distributed applying load (w) to not allow the reflect of
the cables.
5) The relation between sag/span ratio and either steel area or nodal deflections is
quadratic. So, sag/span ratio is very effective on the value of the required steel
area for a certain allowable deflection.
For cable grids

(1) The suspension cable is the main-load carrying cable.
(2) The cross section of the suspension cable has to be greater than or equal to that
of the pretensioned cable.
(3) The steel areas of ties / struts have negligible effect on both deflections and
forces of the cables. So, it is economical to use the minimum cross-sections that
allow the distribution of forces in cable grids.
(4) The sagging of the suspension cable is the principal factor comparing with the
rise of the pretensioned cable.
(5) The sagging of the suspension cable has to be greater than or equal the rise of
the pretensioned cable.
(6) The rise of the pretensioned cable has to be taken as minimum as the grid
reaches the required stiffness (rigidity).
(7) The values of pretension in the pretensioned cables should not allow
129
compression in them.
(8) The value of the equivalent pretension in the suspension cable has to be greater
than or equal to half of the applying load, ( wS0  0.5w ).
(9) Increasing the pretension of the suspension cable is the economic and the
efficient way for decreasing the nodal deflections comparing with increasing the
pretension in both the cables in the same time.
(10) Increasing the number of segments by increasing the number of hangars

decreases the nodal deflections of both suspension and pretensioned cables and,
also, decreases the final forces of ties or struts.
(11) The gap between the suspension and the pretensioned cables in concave grids
has a slight effect on deflections and final forces. So, it is recommended to be as
minimum as possible to increase the stiffness of the grid.
Good Technique for the preliminary design of circular cable suspended roofs
1) The author made non-dimensional table and graphs for (10  10) circular cable
suspended net roof with infinity rigid supports. These graphs and table relate
between all factors affect the design and include:
(a) The span of the roof.
(b) The spacing between the nodes.
(c) The applying uniformly distributed loads.
(d) The steel area of cables.
(e) Modulus of elasticity for cables.
(f) The pretension in cables.
(g) The required nodal deflections.
(h) The maximum forces resulting in the cables.
(i) The sag/span and rise/span ratios of the cables.
2) Using the above table and graphs, all (10  10) circular cable nets may be
preliminarily designed.
3) For (n  n) cable net roofs, two equations (5.12 and 5.13) are used to transform
the studied roof to an equivalent (10  10) net roof and then, the previous table
and graphs will be used to complete the preliminary design.
4) For (n  n) cable grid roofs, equations (5.12 and 5.13) are used to calculate the
allowable central deflections and the maximum final forces of an equivalent
(10  10) grid roof. Then, by using both Equations from (5.20) to (5.25) and the
previous non-dimensional table and graphs, the preliminary design of circular
suspended cable grids either convex or concave can be made.
130
5) According to these non-dimensional table or graphs; the following conclusions
are investigated:
a. Circular cable nets and grids with the same arrangement ( n  n )
will act identically, i.e. ( z L ) and (  ) are invariable, where
  (Tmax / Tmin . break. )% , if the following conditions have been
fulfilled for both the suspension and pretensioned cables:
(i) The value ( EA / wL2 ) is constant.
(ii) The sag (and rise)/span ratio is constant.
(iii) The ratio of pretension to the applying load
( w w ) is constant.
b. Circular cable roofs with the same arrangement ( n  n ), that act

identically, have the same values of the quantity f (z ) in both the
suspension and the pretensioned cables, where
wLS
f (z )  .
d 2 z
EA( ) ( )
L L
c. The steel area is inversely proportional to both the required ratio

of deflection to span, the square of the sag/span ratio, and to the
spacing between nodes.
d. It is recommended to take the equivalent load of pretension less

than or equal to that of the applying load, i.e. ( w  w ).
e. For two circular roofs, nets or grids, ( n1  n1 ) and ( n2  n2 ) with

the same properties (diameter, steel area, loading, modulus of
elasticity and sag (rise)/span ratio), the ratio between either the
deflections or the tensions in the two roofs is equal to the ratio
between the spacing between nodes in these roofs, Eqs. (5.11 and
5.12). This is valid for both the pretensioned and the suspension
cables.
Conclusions of the Dynamic analysis

1. Increasing the steel areas of cables or hangars has approximately no effect upon
the dynamic stability of the structure. This is due to their very small own weight
comparing with applying loading.
2. Increasing sag/span ratio of a cable net makes the structure more dynamically
excitable, whereas increasing sag or rise to span ratios in cable grids increases,
slightly, the dynamic stability of the structure.
3. The most efficiency and economic solution to make the circular cable roof more
dynamically stable, is to increase the pretension of the cables and the hangars.
131
4. Convex grid is dynamically better than concave grid and both are better than
cable nets.
References
(1) Leonard, J. W., "Tension Structures ", Oregon, U.S.A, 1986.
(2) Subcommittee on cable-suspended structures, " Cable Suspended Roof Construction.

State of the Art ", ASCE, Journal of the Structural Division, Vol. 97, No. ST6, June
1971, Paper 8190, pp.1715-1761.
(3) Krishna, P.," Cable-Suspended Roofs ", McGraw-Hill, New York, 1978.
(4) Foster, E. P.," Experimental and Finite Element Analysis of Cable Roof Structures
Including Precast Panels ", Ph. D. Thesis, University of Vanderbilt, Tennessee, U.S.A.,
1974.
(5) Buchholdt, H. A., " An Introduction to Cable Roof Structures ", Cambridge University
Press, 1985.
(6) Urelius, D. E. and Fowler, D. W., "Behavior of Prestressed Cable Truss Structures",
ASCE, Journal of Structural Division, vol. 100, No. ST8, August 1974, Paper 10747,
pp. 1627 -1641.
(7) Scalzi, J. B. and McGrath, W. K., "Mechanical Properties of Structural Cables ",
ASCE, Journal of the Structural Division, Vol. 97, No. STI2, December 1971, Paper
8604, pp. 2837 -2844.
(8) Greenberg, D. P., " Inelastic Analysis of Suspension Roof Structures ", ASCE, Journal
of the Structural Division, Vol. 96, No. ST5, May 1970, pp. 905 -930.
(9) Morales, R. C.," Shear-Volume Method of Solving Tensions in Cables ", ASCE,
Journal of the Structural Division, Vol. 94, No. ST1, January 1964, Paper 5718, pp.
111-118.
(10) O'Brien, W. T.," Behavior of Loaded Cable Systems ", ASCE, Journal of the
Structural Division, Vol. 94, No. STI0, October 1968, Paper 6162, pp.2281- 2302.
(11) Buchanan, G. R.," Two-Dimensional Cable Analysis ", ASCE, Journal of the
Structural Division, Vol. 96, No. ST7, July 1970, Paper 7436, pp. 1581-1586.
(12) Wilson, A. J. and Wheen, R. J.," Direct Design of Taut Cables under Uniform
Loading" ASCE, Journal of the Structural Division, Vol. 100, No. ST3, March 1974,
132
Paper 10427, pp. 565-578.
(13) Wilson, A. J. and Wheen, R. J.," Inclined Cables under Load-Design Expressions",
ASCE, Journal of the Structural Division, Vol. 103, No. ST5, May 1977, Paper 12922,
pp. 1061- 1078.
(14) Desai, S. B.," Dip Calculations for a Simple Cable ", The Structural Engineer, Vol.
54, No.5, May 1976, pp. 183-185.
(15) Judd, B. J. and Wheen, R. J., "Nonlinear Cable Behavior ", ASCE, Journal of the
Structural Division, Vol. 104, No. ST3, March 1978, Paper 13628, pp.567-575.
(16) Huddleston, J. V., "Computer Analysis of Extensible Cables ", ASCE, Journal of the
Engineering Mechanics Division, Vol. 107, No. EM1, February 1981, paper 16024,
pp. 27-37.
(17) Jayaraman, H. B. and Knudson, W. C., " A Curved Element for the Analysis of
Cable Structures ", Computers & Structures, Vol. 14, No. 3-4, 1981, pp. 325-333.
(18) Levinson, D. A. and Kane, T. R., "A Usable Solution of the Hanging Cable Problem",
Computers & Structures, Vol. 46, No.5, 1993, pp. 821-844.
(19) EI-Sherbeny, Y. H.," Non-Linear Analysis of Cable Roof Structures ", M. Sc.
Thesis, Ain Shams University, 1995.
(20) Buchholdt, H. A.," Tension Structures ", The Structural Engineer, Vol. 48, No. 2,
February 1970, pp. 45-54.
(21) Allam, M.G., "Nonlinear Analysis of Three Dimensional Cable Structures", Ph. D.
Thesis, Ain Shams University, 1997.
(22) EI-Ibiary, S. A. N., "Suspended Roofs ", Ph. D. Thesis, Ain Shams University,
1982.
(23) Buchholdt, H. A., " Structural Dynamics for Engineers ", Redwood Books, Great
Britain, 1997.
(24) Buchholdt, H. A., Moossavinejad, S. and Iannuzzi, A., "Non-Linear dynamic Analysis
of Guyed Masts Subjected to Wind and Guy Ruptures " , Proceedings, The Institution of
Civil Engineers, Part. 2, Sept. 1986, Paper 9015, pp. 353-395.
(25) Davenport, A. G., "The Application of Statistical Concepts to Wind Loading on

Structures ", Proceeding, Institution of Civil Engineering, Vol. 19, August 1961.
(26) EI-Rakabawy, M. M., "Dynamic Response of Cable Roofs Subjected to Turbulent
133
Wind", Ph. D. Thesis, Ain Shams University, 1989.
(27) “The Egyptian Code for calculating the loads and forces in the structural and
building works.”, Ministry of Housing, Egypt, 1994.
(28) Abu-El-Farh, E. M. T.," Dynamic Relaxation Analysis of Suspension Roofs ", M.

Sc. Thesis, Ain Shams university, 1997.
(29) Singh, B.P. and Dhoopar, B. L., "Membrane Analogy for Anisotropic Cable
Networks ", ASCE, Journal of the Structural Division, Vol. 100, No. ST5, May1974,
Paper 10544, pp.1053-1066.
(30) Kwan, A.S.K., "A New Approach to Geometric Nonlinearity of Cable Structures",
Computers & Structures, Vol. 67, 1998, pp. 243-252.
(31) Saafan, S. A., " Theoretical Analysis of Suspension Bridges ", ASCE, Journal of the
Structural Division, Vol.89, No. ST4, August 1966, Paper 4885, pp.1-11.
(32) Saafan, S. A.," Theoretical Analysis of Suspension Roofs ", ASCE, Journal of the
Structural Division, Vol.96, No. ST2, February 1970, Paper 7107, pp.393-405.
(33) Jonatowski, J. J. and Birnstiel, C., "Inelastic Stiffened Suspension Space Structures
", ASCE, Journal of the Structural Division, Vol. 96, No. ST6, June 1970, Paper 7364,
pp.1143-1166.
(34) Mollmann, H., " Analysis of Plane Prestressed Cable Structures ", ASCE, Journal
of the Structural Division, Vol. 96, No. STI0, October 1970, Paper 7598, pp. 2059-2082.
(35) Baron, F. and Venkatesan, M. S., "Nonlinear Analysis of Cable and Truss Structures
", ASCE, Journal of the Structural Division, Vol.97, No. ST2, February 1971, Paper
7937, pp. 679-710.
(36) Hood, C. G.," A General Stiffness Method for the Solution of Nonlinear Cable
Networks With Arbitrary Loading ", Computers & Structures, Vol. 6, 1976, pp. 391-
396.
(37) Valente, G., "Hanging Roof With Stiffening Beams in the Second Order Theory ",
Computers & Structures, Vol. 11,1980, pp.113-125.
(38) Godbole, P. N., Krishna, P. and Jan, J. K.,"Boundary Effects in Suspended Cable
Roofs”, ASCE, Journal of Structural Engineering, Vol. 110, No.5, May 1984,
pp. 1099-1113.
(39) Abo-Hamad, M. and Aly, F. F.," Analytical study of Pretensioned cable Roofs", The
Scientific Engineering Magazine, Cairo University, 1986.
134
(40) Kassimali, A. and Parsi-Feraidoonian, H., "Strength of Cable Trusses under
Combined Loads ", ASCE, Journal of Structural Engineering, Vol. 113, No.5, May
1987, pp. 907-924.
(41) Desai, Y. M., Popplewell, N., Shah, A. H. and Buragohain, D. N., " Geometric
Nonlinear Static Analysis of Cable Supported Structures ", Computers & Structures,
Vol. 29, No.6, 1988, pp.l00l-l009.
(42) Swaddiwudhipong, S., Wang, C. M., Liew, K. M. and Lee, S. L., "Optimal
Pretensioned Forces for Cable Networks ", Computers & Structures, Vol. 33, No.6,
1989, pp. 1349-1354.
(43) Saafan, S. A. and EI-Rakabawy, M. M.," On the Economical Study of Olympic
Games Stadium-Egypt ", The First International Conference for Building and
Construction, Cairo, Egypt, June 23-26, 1994.
(44) Shan, W., Yamamoto, C. and Oda, K., "Analysis of Frame-Cable Structures ",
Computers & Structures, Vol. 47, No. 4/5, 1993, pp.673-682.
(45) EL-Sadawy, M. M., "Parametric Study of Space Cable Roofs Subjected to Support
Flexibility and Optimum Analysis ", M. Sc. Thesis, Cairo university, 1994.
(46) Eisenloffel, K. and Adeli, H, “Interactive Microcomputer-Aided Analysis of Tensile

Network Structures”, Computers & Structures, Vol. 50, No.5, 1994, pp. 665-675.
(47) Peyrot, A. H. and Goulois, A. M., " Analysis of Cable Structures”, Computers &
Structures, Vol. 10, 1979, pp. 805-813.
(48) Tabarrok, B. and Qin, Z., "Nonlinear Analysis of Tension Structures ", Computers &
Structures, Vol. 45, No.5/6, 1992, pp. 973-984.
(49) Mitsugi, J., "Static Analysis of Cable Networks and Their Supporting Structures",
Computers & Structures, Vol. 51, No.1, 1994, pp. 47-56.
(50) Buchholdt, H. A., "A Nonlinear Deformation Theory Applied to Two Dimensional
Pretensioned Cable Assemblies ", Proceedings, The Institution of Civil Engineers, Vol.
42, 1969, pp. 129-141.
(51) Murray, T. M. and Willems, N., "Analysis of Inelastic Suspension Structures",

ASCE, Journal of the Structural Division, Vol. 97, No. STI2, December 1971, Paper
8574, pp.2791-2806.
(52) Buchholdt, H. A., Bhattacharya, S. R.," A Gradient method for the Analysis of Cable
Structures with Flexible Boundaries and Laterally Loaded Cables ", The Structural
Engineer, Vol. 56B, No.2, June 1978, pp. 28.
(53) Monforton, G. R. and EI-Hakim, N. M., " Analysis of Truss-Cable Structures ",
135
Computers & Structures, Vol. 11, 1980, pp. 327-335.
(54) Fried, I., "Large Deformation Static and Dynamic Finite Element Analysis of
Extensible Cable ", Computers & Structures, Vol. 15, No.3, 1982, pp. 315-319.
(55) Stefanou, G. D., Moossavi, E., Bishop, S. and Koliopoulos, P., "Conjugate Gradients
for Calculating the Response of Large Cable Nets to Static Loads ", Computers &
Structures, Vol. 49, No.5, 1993, pp. 843-848.
(56) Stefanou, G. D. and Nejad, S. E. M., "A General Method for the Analysis of Cable
Assemblies With Fixed and Flexible Elastic Boundaries ", Computers & Structures, Vol.
55, No.5, 1995, pp. 897-905.
(57) Lewis, W. J., "The Efficiency of Numerical Methods for the Analysis of Prestressed
Nets and Pin-Jointed Frame Structures ", Computers & Structures, Vol. 33, No.3, 1989,
pp. 791-800.
(58) Lewis, W. J. and Shan, J., "Numerical Modeling of Non-Linear Static Response of
Clad Cable Net Structures ", Computers & Structures, Vol. 35, No.1, 1990, pp. 15-22.
(59) Stefanou, G. D., Moossavi, E., Bishop, S. and Koliopoulos, P.," Dynamic Analysis
Approach of Nonlinear Structures by Minimizing the Total Potential Dynamic Work",
Computers & Structures, Vol. 44, No.6, 1992, pp.1197-1203.
(60) Naguib, M. “ Natural Frequencies and Modes of Vibrations for Cable Roofs”, Mansoura
Engineering Journal (MEJ), Vol. 21, No. 2, June 1996.
(61) Naguib, M. “ Buckling strength and dynamic response of guyed towers”, PH. D. Thesis,
Struct. Engrg. Dept., Mansoura University, Mansoura, Egypt, 1989.
(62) Naguib, M. “ Optimum Shape of Concave Cable Roof Beams”, 6th Arab Conference on
Structural Engineering, Faculty of Civil Engineering, Damascus University, Syria,
October 21-24, 1995.
(63) Naguib, M. “ Optimum Shape of Convex Cable Roof Beams”, 6th Arab Conference on
Structural Engineering, Faculty of Civil Engineering, Damascus University, Syria,
October 21-24, 1995.
(64) Naguib, M. “ Optimum Shape of Concave-Convex Cable Roof Beams”, 6th Arab
Conference on Structural Engineering, Faculty of Civil Engineering, Damascus University,
Syria, October 21-24, 1995.
136
‫جامع ال نصور‬
‫ك ي ال ن س‬
‫قسم ال ن س اإنشائي‬
‫ا الكابا‬ ‫التح يل ااستاتيكي ال يناميكي لش كا اأسقف ال ائري‬
‫س ل مق م من‬
‫ع ا‬ ‫م ن س ‪ /‬مح محيي ال ين مح‬
‫لح و عي‬
‫ج ال جس ي في ال س اإنش ئي‬
‫ال ش فو‬
‫أستا كتور ‪ /‬ن يل سي مح و‬

‫ئيس قسم ال س اإنش ئي‬
‫ك ي ال س س ج مع ال و‬
‫أستا كتور ‪ /‬اح ب ر‬

‫أس ال س اإنش ئي‬
‫كتور ‪ /‬مح نجيب أبو السع‬ ‫أستا‬

‫أس مس ع ال س اإنش ئي‬
‫ال و‬
‫‪2003‬‬
‫ال وافق ع ي ال س ل‬
‫ا الكابا‬ ‫التح يل ااستاتيكي ال يناميكي لش كا اأسقف ال ائري‬
‫ج‬ ‫‪ /‬مح محيي ال ين مح عث ل ح و ع ي‬ ‫ال س ل مق م من ال‬

‫ق ت ت م قش ال س ل ال وافق ع ي من ‪-:‬‬ ‫ال جس ي في ال س اإنش ئي‬
‫‪ .1‬أستا كتور ‪ /‬سعفا ع الجوا سعفا‬

‫أس ال س اإنش ئي ‪ -‬ك ي ال س س ج مع عين ش س‬
‫‪ .2‬أستا كتور ‪ /‬ن يل سي مح و‬
‫و‬ ‫ئيس قسم ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال‬
‫‪ .3‬أستا كتور ‪ /‬يوسف إبراهيم حجا‬
‫أس ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال و‬
‫نجيب أبو السع‬ ‫كتور ‪ /‬مح‬ ‫‪ .4‬أستا‬
‫و‬ ‫مس ع ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال‬ ‫أس‬
‫توقيع ال شرفين‬
‫ال وقيع‬ ‫ااسم‬
‫‪-------------------------‬‬ ‫‪ .1‬أستا كتور ‪ /‬ن يل سي مح و‬

‫ئيس قسم ال س اإنش ئي‬
‫‪-------------------------‬‬ ‫بــــ ر‬ ‫‪ .2‬أستا كتور ‪ /‬اح‬

‫أس ال س اإنش ئي‬
‫و‬ ‫ك ي ال س س ج مع ال‬
‫‪-------------------------‬‬ ‫كتور ‪ /‬مح نجيب أبو السع‬ ‫‪ .3‬أستا‬

‫أس مس ع ال س اإنش ئي‬
‫ت يخ ال قش‬
‫‪2003 / 5 / 5‬‬
‫تعريف بال احث‬
‫‪ :‬مح محيي ال ين مح عث‬ ‫ااســـــــــــم‬
‫‪1970 / 9 / 21 :‬‬ ‫تاريخ ال يا‬
‫ي‬ ‫‪ :‬اإس‬ ‫محل ال يــا‬
‫ال رج الجامعي اأ لى ‪ :‬ب لو يو ال س ال ني‬

‫ي س ‪1992‬‬ ‫ك ي ال س س ج مع اإس‬
‫نش ئي‬ ‫‪:‬م‬ ‫الوظيف‬
‫ال وقيع ‪:‬‬

‫ال يخ ‪2003 / 5 / 5 :‬‬
‫شـكـر‬
‫هم‪:‬‬ ‫ال س ل‬ ‫ال ش فين ع‬ ‫ل الس‬ ‫اق خ لص ش‬
‫ك و ‪ /‬ن يل سي مح و‬ ‫‪ )1‬أس‬
‫و‬ ‫ئيس قسم ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال‬
‫ك و ‪ /‬اح ب‬ ‫‪ )2‬أس‬
‫و‬ ‫أس ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال‬
‫ك و ‪ /‬مح ن يب أبو السع‬ ‫‪ )3‬أس‬

‫و‬ ‫أس مس ع ال س اإنش ئي ‪ -‬ك ي ال س س ج مع ال‬
‫إه اء‬
‫أ ك ال س ي‬ ‫وا‬ ‫بن ع ه ع يه اف ل ال‬ ‫‪ :‬مح‬ ‫ل خي خ ق ه‬
‫‪:‬‬ ‫ل اص ي م ي‬
‫ل أبي أمي الغ ليين اخوتي اأع اء‬
‫ل من انعم ه ب م ع ي ‪:‬‬
‫نو ا‬ ‫ل ال ائع ين ي‬ ‫ج ي الغ لي‬ ‫ل‬
‫‪:‬‬ ‫ل ال ع ءين ائ‬

‫ال ه اء‬ ‫ل اح م ه‬
‫‪:‬‬ ‫مع ي‬ ‫ل أس‬

‫ال ك و مح ن يب‬ ‫ل اأس‬
‫ال شو‬ ‫ل ع ي ب لع م العو‬ ‫ل كل من ت‬

‫جامع ال نصور‬
‫ك ي ال ن س‬
‫قسم ال ن س اإنشائي‬
‫م ص رسال ال اجستير ال ق م من ‪:‬‬

‫‪ /‬مح محيي ال ين مح عث‬ ‫م‬
‫عنوا الرسال ‪:‬‬

‫اأسقف ال ائ ي ا ال با "‬ ‫" ال ح يل ااس تي ي ال ي مي ي لش‬
‫أس اء ال شرفين ‪:‬‬

‫ك و ‪ /‬ن يل سي مح و‬ ‫أس‬
‫ك و ‪ /‬اح ب‬ ‫أس‬
‫ك و ‪ /‬مح ن يب أبو السع‬ ‫أس‬
‫تاريخ ال ناقش ‪2003 / 5 / 5 :‬‬ ‫تاريخ التسجيل ‪1996/10/ 24 :‬‬
‫ص‬ ‫ال‬
‫ال ف ال ئيسي ل ال س ل هو بحث ال ح يل ااس تي ي ال ي مي ي لأسقف ال ائ ي ال ع ق‬
‫ا ال با مع اخ اخ ي ه ال ش في ااع ‪ .‬في ال اي ت ت اس ع ا أنوا ال با‬
‫كا‬ ‫س ئل س ق اإج‬ ‫ح ي‬ ‫خواص ال ي ني ي‬ ‫ال س م في مثل ه ال ش‬
‫ال ح يل ااس تي ي ل بل بسيط ال ع يق تحت ح ل م م‪ .‬أي ت ت اس اأسقف ال ع ق ا‬
‫ال وا ال س م في ك ا م ي ات عيوب ‪ .‬تم ال ع‬ ‫م ون ت‬ ‫ال با من حيث أنواع‬
‫لأح ااس تي ي ال ي مي ي ال ي ت ث ع ه ال ش مع ت يل اك أح ال ي ب ع ه‬
‫ل ح يل ه‬ ‫ال‬ ‫ال ث ال ي مي ي ال ئيسي ع ي اأسقف ال ع ق ا ال با ثم اس ع ا ال‬
‫افي ل يق ال ق‬ ‫‪ .‬تم تق يم ش‬ ‫ب ال‬ ‫لأبح الس بق ال‬ ‫م‬ ‫ال ش مع ع‬
‫يق ال ح ا ال ل كي ي‬ ‫ق الوضع ل يع الع ص اإنش ئي ب س ا‬ ‫ال ي ع ت غي‬
‫اس ام في ج اء ال ح يل الاخ ي ل ال ش سواء ك اس تي ي أ ي مي ي ‪ .‬أي تم‬
‫ك ا يق م س لحس‬ ‫(م ) ال ا‬ ‫يق ن‬ ‫اس ع ا ال ح يل ال ي مي ي ب س ا‬
‫ال ش ا الح ث ب عل ال ي ‪.‬‬
‫ت ت اس تح يل الع ي من اأسقف ال ائ ي ال ع ق ا ال با ب س ا ب ن مج ح سب آلي ق ئم‬
‫يق ال ح ا ال ل ثم اس ا ن ئج ال ح يل في‬ ‫ق الوضع ب س ا‬ ‫يق ت غي‬ ‫ع‬
‫ل عوامل ال ث ع مثل ه ال ش ‪ .‬أي تم اس ا ه ال ئج في‬ ‫ال أثي ا ال‬ ‫اس‬
‫ال سوم الا‬ ‫عاق ج ي ‪ .‬ه ال ا‬ ‫سوم بي ني ا بع ي ك ا اس‬ ‫ع لج ا‬
‫يم ااب ائي لأسقف ال ائ ي ا‬ ‫فع ل ل‬ ‫ش ت سي جي‬ ‫بع ي ك ا العاق ال س‬
‫توافق ك ي ا مع ت ك ال ي تم الح و‬ ‫ال با ‪ ،‬ع ب ن ئج ال ح يل ب س ا ه ال يق ق أ‬
‫ع ي ب س ا ب ن مج الح سب اآلي ال س قت اإش ليه‪.‬‬
‫ا ال با في اس‬ ‫في ال ي تم اس ا ب امج ح سب آلي خ ص ب ل ح يل ال ي مي ي ل ش‬
‫ك ا ال ع ي عن س وك ال ي مي ي‪.‬‬ ‫العوامل ال ث ع ت ا اأسقف ال ع ق ا ال با‬
‫‪View publication stats‬‬

Static and Dynamic Analysis of Circular Cable Roof Nets

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Static and Dynamic Analysis of Circular Cable Roof Nets

Uploaded by

Copyright:

Available Formats

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

Static and Dynamic Analysis of Circular Cable Roof Nets

Research · July 2016

Mohamed Mohie Eldin

The user has requested enhancement of the downloaded file.

STATIC AND DYNAMIC ANALYSIS OF

Professor Dr. Associate Prof. Dr.

Name, Title and Affiliation Signature

2- Prof. Dr. Ahmed Badr ----------------------------------

3- Prof. Dr. Mohamed Naguib Abou El-Saad ----------------------------------

This thesis is submitted to Mansoura University for the Degree of Master of

No part of this thesis has been submitted for a degree or a qualification at

Name : Mohammad Mohie Eldin Mohammad Ossman

Date of Birth : 21 - 9 – 1970

Place of Birth: Alexandria

Qualification : B. Sc. Degree in Civil Engineering - Faculty of Engineering -

Job : Structural Engineer

Abstract of the M. Sc. Thesis

Registration Date: 24/ 10 / 1996 Examination Date: 5 / 5 / 2003

My appreciation, thanks and grateful acknowledgement are due to Dr.

Finally, special acknowledgement to my parents, my wife, my daughters ,

Chapter (2) Literature Review

Chapter (4) Dynamic analysis of cable roofs

Fig. (1.2): Types of cables: a) Strand ; b) Rope. 9

Fig. (2.1) Simply suspended roofs. 30

Fig. (2.4) Multi-span cable beam structures. 33

Fig. (2.5): Saddle-shaped cable net with edge cables. 33

Fig. (2.8): Discrete and continuous representation of suspension structures. 35

Fig. (2.9): Roof cladding. 36

Fig. (2.11): Diagrammatic representation of the stiffness method. 36

Fig. (3.4): Construction of two conjugate directions: The directions v k and v k 1

Fig. (3.7): Coordinates and force for link ab in:

Fig. (5.15): (10  10) Circular cable roof. 107

Fig. (5.35): The relation between the pretension in ties and:

Fig. (5.46): The first modeshape of a circular cable net. 119

Fig. (5.47): The second modeshape of a circular cable net. 119

Table (2.1) Experimental values for  and z g . 29

Table (2.3) The principal wind pressure, q, in (kg/m2). 29

Table (2.4) The exposing factor, k. 29

Table (5.1) Data assumptions for case study (1). 95

Table (5.5) Data assumptions for case study (2). 97

Table (5.6) Data assumptions for case study (3). 97

Table (5.7) Example (6):

Table (5.8) Example (6):

Table (5.9) Example (8):

Table (5.10) Example (8):

Table (5.11) Example (10):

Table (5.12) Example (11):

1.2 Classification of cables

Pretensioning is achieved either by (3):

1.4 Protective coating

1.5 Analysis of freely hanging cables

Greenberg (8) presented a mathematical representation of cable stress-strain curves.

Huddleston (16) presented an iterative method. It consists of numerically integrating

1.5.2 Static analysis (3)

The approximation of cable segment as straight element is valid only if:

The approximation of a catenary as a parabola is justified in the following two cases:

Freely hanging cable under uniformly distributed loading (5)

Values of H and d will change to ( H  H ) and (d  d ) if there are changes in:

For horizontal movements of supports

For change in cable length, l , caused by either (a), (b), or (c)

For (a), change in applied load:

For (b), change in temperature: l   t l (1.17b)

For (c), change in cable length: l  (1.17c)