UDC 624.91 :624.071.

Tension Structures

H. A. Buchholdt
BSc PhD

Dr. Buchholdt, the author of this paper which is to

be presented for discussion at an Ordinary Meeting
of the Institutionof Structural Engineers at 6 pm on
Thursday 26 February at l l Upper Belgrave Street,
London S W I , was born and received his scholastic
education in Oslowhere he also attended Commercial
College and Military Academy. I n 1949 he came t o
Londonwhere he studied at the College of Automobile and AeronauticalEngineering.
Dr.Buchholdt then joined Tiltman and Langleys as a stress
analyst and at the same time studied for his BSc
at Kingston Technical College. I n 1956 he returned
to Sweden where he spent ayear as an ofticer on the
staff of the HQ Anit-Aircraft Artillery and another
year as project engineer with Bonde d: CO of Oslo.
Dr. Buchholdt returned toLondon in 1958 as Assistant Lecturer, then Lecturer, at Battersea College of
Advanced Technology (now
the
University
of
Surrey) and in 1967, having obtained his PhDhisthesis was 'Prestressed cable nets with finite
displacements'-from
London University, hetook
up his present appointment as Principal Lecturer in
Theory of Structures in the Department of Civil
Engineering at The
Polytechnic,
Regent Street,
London. In 1968 Dr.Buchholdt set upthe Cable
Structures Research and
Development
Centre.
Heis aMember
of theNorwegianEngineering
Society and Director of Cable Structures Ltd.

Synopsis
The paper describes different types of tension roofs and
methods by which they can be analysed. The theories
consider prestressed cable roofs as discrete systems
having significant geometrical displacements.
The methods for calculating forces and displacements
are:the
method of steepest descent; the conjugate
gradient; Newton-Raphson method. The instability of the
Newton-Raphson method and how it can be stabilized is
considered. A unified view is obtained by considering the
above methods as three ways of minimisingthe total
potential energy of structural assemblies.
The initial configuration of prestressed nets with quadrilateral or triangular boundaries in plan are given in terms
of a double Fourrier series. For nets with arbitrary boundaries theinitialconfiguration
andcorresponding preparameters on stressing forces are calculated by minimisation of the totalpotential energy.
Reference is made to experimental work, effect of design
parameters on structuralefficiency, dynamicbehaviour and
different types of anchorages.
Finally, two design examples are given,a prestressed
grid roof for a circular water reservoir and a 480 ft span
hangar for two Boeing 747 aircraft.
THE STRUCTURAL ENGlNEER

STRUCTURAL SYSTEMS
Introduction
During the last 20 years a number of large roofs have
been built in which the main load carrying elements are
subjected to tensile forces only. Since all the members
are in tension there are no stability problems, and it is
therefore natural t o use high-tensile steel rods or hightensile steel cables. If the latter is used
very large spans
canbe achieved. Roofsbuilt so far have hadspans
up t o 120 m (394 ft), andspansup
to 330 m(984ft)
have been projected.
Roofs in which the main load-carrying elements are
made from either steel cables or steel rods are inherently
cheap. For small spans and particular forms of design,
however, the total cost can beadversely affected by the
size and cost of the foundations required t o resist the
tensile forces in the cables.
Cable roofs can be divided into three categories
based
upon whether the roof cladding is supported by:
(a) simply suspended cables;
(b) pretensionedcablegirders,trussesorgirdertrusses;
(c) pretensioned cable nets.
In the following
a short description isgiven of the main
types of pretensioned cable systems which can be used
in the design of cable roofs.

Simply-suspended cable roof systems

In this form of construction the roof cladding
is supported by a series of simply-suspended cables (Fig 1).
Such systems have no stiffness. To reduce the movementscaused byany formofapplied
load, theroof
cladding must be heavy, using such materials as reinforced concrete slabs. Unless the spaces between the
slabs are filled with
concrete, thus forming a concave
shell, the roofs are easily deformed and not very satisfactory.

0 FEBRUARY 1970 0 No 2 0 VOLUME 48

Fig. 1

45

Othermethods of stiffeningfreely-suspendedcable
systems are to introduce internal ties or suspendheavy
beams from the cables.Theformermethodtends
to
reduce the internal free
volume, and as for the latter,
although satisfactory, cheaper methods exist for stiffening thesystem. These are described under the headings
Cablegirders,trussesandgirdertrusses,andPretensioned nets below.
The circular stadium in Montevideo, Uruguay, having
a diameter of
80 m (262 ft) is anexample of a simply
supported cable roof construction in which the cables
aresuspendedinplanespassingthroughthecentre
of the building and attached at the outer boundary to a
compression ring and at the centre to a tension ring
(Fig 2). The roof cladding was made by placing 50 mm
cables;
the
reinforced
thick
(1.9
concrete
on
plates
in)
a temporary 50 per cent overload in the form
of bricks
wasthenplacedontheslabs,afterwhichthespace
between the slabs was filled with concrete. When the
concrete had set, the bricks were removed, resulting in
a kind of inverted prestressed concrete
shell.

:c:

C?)

Fig 3

Fig 2

Cable girders, trusses and girder-trusses

A lighter and stiffer
system than the heavily loaded cable
is achieved ifasecondcable
ofreverse curvatureis
connected to,the suspended cable, as shown inLFig3(a),
(b) and (c), and then tensioned to such a degree that
under any form of applied load the cables always remain
i n tension.Ifthetwocables
are inter-connected by
verticalmembersonly,relativelylargedisplacements
may takeplaceundernon-symmetricalloading.
It is
therefore better for larger spans to introduce diagonal
membersasshowninFig
3(b)as
thesewillnullify
themechanicalfreedomoftheassembliesshown.
The type of truss shown in Fig
3(b), but without the
vertical members included, has been used in a number
of buildings, one of which is the ice-hockey stadium in
Stockholm which has a span of 83 m (272 ft).
T wo different
systems
for
circular
buildings
are
shown, Fig 4 and 5; theyare a sports stadium and a
hangar or airterminallargeenough
to accommodate
five Boeing 747. In the stadium half-girders are attached
to an outer compression ring and an inner tension ring.
In the hangar a
two-waypretensionedgridsystemis
supported byan outercompressionring.Oneofthe
advantages
of
circular
buildings
for housinglarge
aircraft, beside the economy of space, is that the compression ring can also provide supports for the hangar
doors.

Fig 4

suspension cables and the convex ones as prestressing

cables.
achieved
Stiffness
is
cables
by tensioning
the
Thethirdclass
of cableroofscontainthoseinwhichbeforeplacingtheroofcladding.
To preventrelative
the cables form large saddle-shaped nets
as shown in
movement of the cables they are clamped at their points
Fig 6. The concavecablesareusuallyreferred
to as
ofintersection.Forhighlycurvednetstheclampingmust

Pretensioned
nets

46

THE STRUCTURAL ENGiNEER

0 FEBRUARY

1970

0 No 2 0 VOLUME

48

THE STRUCTURAL ENGINEER

FEBRUARY 1970 0 No 2

0 VOLUME 48

47

suitable for circular roofshaving spans up to30 m (98 ft)

diameter. Several roofs of this type have been built in
countries in Eastern Europe, where they have proved to
be a very economical form of roof construction9.
Roofs built as tension structures are inherently cheap
andcan cover largerspansthan
is possible by conventional methods of design. The lack of stiff ness which
many engineers and architects have considered as their
main disadvantage can largely be overcome by choosing
suitable geometries, intensities of prestressing and
dimensions of the main load-carrying elements. In the
past, increased stiffness toprevent flutter of cable roofs
has been sought by increasing the weight of the roof
claddingand/or
by attachinginternalties.Thefirst
method increases the costof the project; the lattergives
an untidy appearance.

THEORY

Fig. 6

take place before tensioning the cables

as it is otherwise
extremely difficulttopositionthecables
accurately.
Pretensioned net roofs have been built with both rectilinear and round boundaries in plan.A n example of the
first type of plan geometry is the ice hockey stadium at
Tammerfors in Finland, covering
an area of 92 x 63 m
(302 x 207ft). A diagram of thestadiumisshownin
Fig 6(a). This building is of additional interest
as high
tensile steel rods were used instead of cables. To the
best of the author's knowledge this is the first time high
tensile steel rods have been used for a roof with such
a large span.
In Fig 6(b) is shown a roof in which the arches
are
inclined to the horizontal. One of the first cable roofs
net
built, the Raleigh Arena in USA, is constructed in this
way. In this building the arches
are inclined at 22" to
the horizontal, and the building has a maximum span
of 92 m(302ft).The
advantageof inclinedarchconor completely eliminate the
structions is that they reduce
necessity of tension anchorages.
The shapeof the roof shown in Fig(c)6resembles that
ofahyperbolicparaboloid,but
is,as thepreviously
described nets, a minimum-energy surface. Because of
thestraight-edge
members, thisformofdesignis
suitable for only the smaller typesof roofs having sides
not greater than 20 to 30 m (66 to 98ft).
If the roofs are only of a semi-permanent nature and
it is possible to replace
stiffness is a minor consideration
the edge members shown in Fig 6(c) by cables. This is
alsopossibleforcablenets
of differentconfigurationsg~~6.
For small roofs it is possible to support the
roof cladding on completely flatnets. This is particularly
48

Previous work and preliminary

Historically the earliest work in this field was done in
the 19th century in connection with the design
of suspensionbridges.
It was graduallyappreciated
thatthe
analysis of these
major
engineering
achievements
differed from those of frameworks inasmuch as it was
necessary to consider the equilibrium of system
the
in its
deformed state. The reason for this is that the system
behaves i n certain respects like the linkage in a machine,
rather than a framework whosegeometry is fixed except
for infinitesimal displacements. When the concept of a
structural mechanism, as distinct from a fixed
framework, is extended t o threedimensions,the
analysis
becomes complicated. One method of approach, widely
adopted, and applicable to nets, is to replace the network of cables by a membrane and use a shell theory to
approximate the solution99 159 18, 19, 20. This, however, is
a retrograde step since an exact solution of the membrane problem is further away from the original problemthanthesolutionofthecorrespondingdiscrete
system inwhich
derivatives are replaced by finite
differences and the resulting finite difference problem
solved numerically over a grid of points.
In this paper methods of solving these typesof structures without losing sight
of their essentially discrete
nature are reviewed. Thetheoryis
divided intotwo
distinct parts. The first part consists of a general treatment of pin-jointed link assemblies by energy methods
which,whilstlimitedto
linear elasticdeformation of
the
members
themselves, allows
for
geometrically
significant displacements of the joints, and thus
enables
the entire field of pretensioned cable assemblies to be
treated with complete generality. This is referred to as
the deformation theory. The second part is referred to
as the Configuration theory and is concerned only with
thestaticalproblemofdeterminingthegeometrical
configuration of nets consistent with initial pretension
inthe cables. Itsprimaryapplicationistoadoublythreaded network of cables whose plan projection consists of two families of cable lines, though it can also
beextended t o certain casesof triply-andquadruplythreaded nets.
In
the
following
references
cable
structures are
treated as discrete systems: 2,3,4,5,6,8, 11, 15,17.

Deformation theories
Assumptions
(a) The link elements between the joints are straight
lines.
(b)Theapplied
loads, includingtheself-weight
of
the links, act at the joints.
(c) Bending stiffness of the cables is negligible and
each joint is considered to be pin-jointed.
A cable structure is therefore considered as a pin-

THE STRUCTURAL ENGINEER c] FEBRUARY 1970 D No 2 0 VOLUME 48

brium position of the structure is calculated when the

right-hand side of equation 5 is equal to zero, i.e., when

jointedassemblyconsisting
ofelastic
rodsinwhich
internal stresses are due to axial forces only.

[aw/ax] = 0.

T h e gradient vector of the total potential energy

A unified view of relevant methods of analysing loaded

pretensioned link assemblies is obtained if the
attainment of equilibrium is regarded as the process of minimization of the total potential energy with respect to all
possible displacements.
Thetotalpotential
energyof astructural system is
given by:

. . . (1)

W=U+V

where U = the elastic potential or strain energy of the

system and
V = the potential energy of the loads applied to
the system.
For constant loading equation 1 may be written as:

m= I

where [ F ] = column vector of all the externally applied

forces,
1x1 = column vector of all joint displacements,
Urn = strain energy stored in any link or member
m and
M = total number of links.
Differentiating equation 2 with respect to x j r we get

aW/JXji = -

Pjn/Ljn((Xni - - X j t )

n= 1

+ (Xni

. .-. Fit
(4)

Xii))

where

number of links meeting at joint j ,

P n j = force in member n j ,
L n j = length of member n j ,
X n i and X i i = co-ordinates of joints n and j in
the i-direction and
Xni and Xji = displacements of joints n and j in
the i-direction.
=

if Thus

tin =

2
N

tjn((Xn

4-( x ,

- x j ) ) -- ~j~

n- 1

.7W/Jyi =

n= I

tjn((Yn

vi) + (yn

Vi))

Fjy

From the right-hand side of equation 5 we see that the

gradient vector [aW/ax] represents
the
unbalanced
forces at the joints for a displacementvector [ x ] .
Since the attainment of equilibrium is considered as
the process of minimization of the total potential energy
withrespecttoallpossibledisplacementstheequiliTHE STRUCTURAL ENGINEER

(a) the method of steepest descent;

(b) the method of conjugate gradients; and
(c) the Newton Raphson method22.
The method of steepest descent
Suppose that at the k'h stageof the calculations the
displacement vector is [ x l k . The Euclidian norm of the
out-of-balance force is then given by :

R=

(pw/ax];[awpxlk)*/2

. . . (6)

Dividing each element in the gradient vector by - R

gives a unit vector [ v ] k in the direction of the steepest
descent on the energy surface. To minimize W we move
[v]k adistance S untiltheminimum
inthedirection
value of W is reached in that direction. The
next improved
displacement vector is then given by:
M k + 1 = CXIk

-+ S b l k

. . . (7)

Since W is reduced at each stage, successive iterations will eventually produce a displacement vector for
which R is zero or negligible. When this isso the forces
at all joints will be in equilibrium and the correct displaced form due to the applied loading calculated.
The method of steepest descent is inherently astable
process but tends to converge only slowly. The rate of
convergency
is,
however, remarkably improvedif
S
is replaced by a reduced step-length,
lS where 0<1<1.0.
It is not easy to estimate the best value of ;1 t o use. It
appears that for 0.3<11<0-9 the number of iterations is
reduced to anything between 5 per cent and 25 per cent
of those required using the full step-length4.
The method of conjugate gradients
For the mathematical theory of the method of conjugate
gradients the reader is referred to the paper t y Fletcher
and Reeves? Here it is sufficient to point out that this
method is a descent methodlike the method of steepest
descent, but one in which the direction
of descent is
given by

Pjn/Ljn

we may write the expressions of the gradients of the

total potentialenergy at jointj in theX , Y and Z directions
as :
JWJzj = -

In the following aregiven three different processes

forlocatingthelocalminimumonthetotalpotential
energy surface. They are:

0 FEBRUARY 1970 0 No 2

For linear structures with f degrees of freedom, the

process will locate the minimum position on the total
potential energy surface in not more thanf steps except
forroundingofferrors.
For non-linearstructuresthe
method may diverge unlessprecautions aretaken to
ensure convergency. A check that the methodconverges
is that the total potential
energydecreases with each
step.
The process is best started by initially moving in the
direction of steepest descent. The convergency rate of
the methodof conjugate gradients is considerably higher
than that of theprevious method. This is particularly
noticeable in the vicinity
of the minimum which indicates
that in this region theenergy surface is nearly quadratic
in the displacements.
>

The Newton-Raphson method

The most powerful iterative process in terms of iteration

is the Newton-Raphson
approach. The methodhas, how-

VOLUME 48

49

ever, certain limitations which are mentioned later on i n

this section.
Let the gradient vector for any position x in displacement space be:

r~w/axi~z

i 1= , 2 , .

. .,f

Using Taylor's series we get:

x+dx

in which repetition of a suffix within a term implies summation on that suffix over all values from 1 to f.
If dxi, dxh, . . . are taken to be so small that their higher
powers are negligible and also noting that V is linear in
x equation 9 may be written as:

Since the vectors on the left-hand side of equation 10

represent the change in residual forces due to a change
[ d x ] in displacement space the Hessian matrix on the
right-hand side of the same equation can be recognized
as the stiffness matrix of the assembly.
Inthe
Newton-Raphsonprocedurelet,as
inthe
previous two methods, the approximation to the solution
afterthe k f h iterationbe [ x ] ~ , andafterthe
( k + 1)th
iteration
[xlk

+1

= [x]k

+ [dx]k

..

(11)

where [ d x ] / +~ 1 is calculated by putting [ J W / a x t ] k + 1 = 0

in equation 10 which then can be written as
[d~]k
+ 1 = - Kk-1

. . . (12)

[aW/Jxt]k

K, + d z and ( a W / a x i ] z + dz can then be computed

for
use in the next cycle of iteration.
If K is well conditioned convergence is rapid. If, however, the condition of K is poor, which is often thecase
for cable structure systems, the method may diverge or
locate a local minimum on the
energy surface other than
thatrepresentingthetrueequilibriumposition
of the
structure.
TheinstabilityintheNewton-Raphsonprocessis
best overcome by adopting the following procedure.
Calculate a unit vector in the Newton-Raphson direction by normalizing [dxlk

. . . (13)
and then move along
[ v ] k a distance S as in the two
previousdescentsmethodsuntilthetotalpotential
energy is a minimum in this direction.
Finally,reverting
toequation
10 we see thatthis
requires an expression fora'U/axjax, for each link in the
assembly. For a pin-jointed pretensioned link
it can be
shown that:

GGT
GGT
GGT - GGz']

'L

PO

I -l
[-ll]

Configuration theory
The maintheoriesdevelopedforthecalculation
of
initialconfiguration ofpretensionedcablenetswhen
these are considered as discretesystems
aresummarized in this section. Explicit solutions are
given in
reference 5 for nets with arbitrarily vertical boundaries
which in plan have:
(a) atwo-directionalnon-parallelcablesystemand
any quadrilateral or triangular boundary;
(b) two-directional
a
diagonal
cable
system
and
parallelogrammicboundary;
(c) athree-directionalcablesystemandparallelogrammic boundary;
(d)afour-directionalcablesystemandaparallelogrammic boundary.
Only a summary is given in this paper of the theory for
the two-directional types of nets since these are those
most commonly used so far.
Theconfiguration
of two-directionalnon-parallelcable
netsystemswithquadrilateral
or triangularboundaries.

Assumptions:
(a) theprojections of thecablesonthehorizontal
plane are straight lines;
(b) the bending stiffness of the cables is negligible;
(c) theweight of thecablesactattheirpoints
of
intersection.
Referring to Fig 7 the equilibrium in the Z-directionof
joint n , m canbeexpressedby
theellipticdifference
equation:
(tm 6

+ tn 8m2) Zn,m

untm

. . (19)
*

In reference 5 it is shown that this will yield an expression for Zn,, in the form of a Fourier series where.

. . . (14)

where G = [ I , m , n I T and l, m and n are the direction

cosines of the member.

M-

4-

N-l

M-l

T h e position of minimumtotalpotentialenergyinthe
direction of descent

The total potential energy of a pin-jointed assembly can

be expressed as a fourth order polynomial in the steplength S as givenbelow.
50

. nnp . mnp
sln -- sm

ENGINEER THE STRUCTURAL

. . . (20)

[7

FEBRUARY 1970 0 No 2

VOLUME 48

where
ZP

= 2 (cos --

ApIV

The influence of the weight of the cables on the configuration is best taken intoaccount by an iterative
process in which initially thenet is assumed to
be weightless. At the end of each iteration the current values of
LO are computed and used in thenext iteration.
It should be noticed that when the solution is in an
explicit form this presupposes assigned
values of the
tension co-emcients.

l ) , ( p = 1 , 2,

Zn,m and

are internal
and
boundary
coordinates respectively.
This form of solution has the advantage from a computationalpointof
view that the vertical co-ordinates
can be expressed as a product of matrices.

m+l

rn
m- 1
n m-1

Configuration of cable nets with arbitrary boundaries

the gradient method of analysis.

by

It was shown in the previous section how the initial configuration ofcablenets
couldbe determined by considering the equilibrium of the forces at each joint. T o
solve the systemof
equationsthus
evolved it was
necessary toassume values forthehorizontalcomponents of the tensile forces in the
cables. Since the
geometryof a net is a function
of the relative magnitude of the prestressing forces, a given configuration
can be obtained only by trial and error.
It wouldtherefore be useful t o have analternative
method of approach by which the required prestressing
forces could be calculated from the knowledge of the
geometry of the nets and which makes no assumptions
with regards to the geometry of the boundaries.
The most obvious method of analysis is by assuming
somevalues
fortheforcesinthe
cables andthen
minimisingthetotalpotential
energy until each joint
is in equilibrium. The initiallyunbalanced forces at each
joint will give rise to small changes in the co-ordinates,
which, if they are too large t o beacceptable, can be
reduced by a second calculation replacing the assumed
forces by those calculatedand setting the displacements
t o zero. Examples of this form of analysis are given In
reference 4.

GUYED M A S T S
In view of the increasing demand and usage of guyed
masts it is worth noticing that the minimization theories
given are eminently well-suited for the analysis of these
types of structures. This is particularlyso when the guy
ropes are subjected t o severe loading conditions in the
form of ice andwind. Computerprogrammesforthe
analysis ofthese
typesof
structures based onthe
conceptofminimumtotalpotential
energy have been
developed.

EXPERIMENTAL WORK
Fig 7

The matrices are defined as follows:

r

S1

L
L EJ
sm -

S2 =
W

, (symmetricoforder
x (N -

(N

1)

-l

sm

, (symmetric, of order (M - 1 )
x ( M - 1))

= [ W , , ] , (of order

(N - 1 ) x (M - 1 )

B = [bpq] = SlWS2

THE EFFECTOFDESIGNPARAMETERS
ON STRUCTUAL EFFICIENCY A N D C O S T

whence

Z =

The deformation theories given in this paper have all

been programmed and the methods tested for stability
and convergencyratesby
application to the analyses
of all the main types of structures previously described
in thispaper. Also all the theoriesand programmes have
been verified by experimentalwork on curved and flat
nets, girders,trusses and girder-trusses in which the
cable elements have been made from high tensile steel
cablesof
7 X 7 strand varying from 1.6 mm (hin)
diameter for the nets to 2.4 mm ( A in) diameter for the
two-dimensional systems.

[zn,,,z]
= -S1AS2
NM

. . (21)

A considerable amount of computational work has been

carried out to enable a study of
(a) the stiffness,
(b) the magnitude of support reactions and
(c) the load-bearing capacities

THE STRUCTURAL ENGINEER 0 FEBRUARY 1970 0 No 2 0 VOLUME 48

51

appropriate. This is attributed to the opening

of fissures
of different types of assemblies.
In doingso the following
and a subsequent softening of the clay around the pile.
design parameters have been varied in order to find out
how they will influence the size and hence the cost of
No results are available fornormallyconsolidated
the boundary structure and foundations:
clays so that this cannot be checked. Armishaws tests
with driven piles in chalk gave similar results to those
(a) cable curvature;
inLondon
clay. Thelong-termeffect
has not been
(b) intensity of initial pretensioning;
investigated.
(c) cable dimensions; and
When tension piles are used in groups, a reduction in
(d) intensity and form of loading.
theload per pileshouldbeallowedforinthe
same
For both two- and three-dimensional systems
it has
manner as for compression pile groups. The reduction
been found that increased cable curvature considerably
factors that are generally used were not obtained from
increases the stiffness of the assemblies.
field tests or large groups of piles but from model tests.
A high degree of pretension increases the stiffness of
These indicate that for single rows
of piles at normal
nets, but only marginally affects the stiffness
of the twodimensional types of cable structures shown in Fig. 3.
spacings the reduction in the carrying capacity
due to
For these types of constructions therefore the pretensiongroup action is small. Some rather unusual methods of
should be kept to a minimum, and not be
greater than
forminganchorsin
clay are described by Littlejohn25,
that required to keep the cables in tension under
any
who also gives some valuable information on the fixedform and intensity of loading.
anchor type of pile ingeneral. He quotesan ultimate load
Increasing cable sizes increases the stiffness of all
of the order of 203 Mg (200 ton) for a 3.6 m (12 ft) long
types of assemblies. Since, however, the types of roof
fixed anchor at a depth of 15.3 m (50ft) in a gravel with
coveringusedfortheseformsofconstructionsare
an angle of internal friction of 40.
usuallysufficientlyflexibleto
take upthe necessary
For anchors in sand it is considered that anchor plates
deformation, it would appear that cable sizes considerare probablythemostsuitableform
ofanchorage.
Several investigators have carried out model test studies
ablyabove
thoserequiredfromaconsiderationof
strength are required only when the roof supports crane
and have proposed formulae for calculating the ultimate
systems.
load of anchorsinsandwith
vertical pulls. Ballas26
method of analysis isgenerally considered tobe the best
but it overestimates the anchor capacity in loose sands.
D Y N A M I C BEHAVIOUR
Forwell-designedtensionroofsflutterandvibration
Baker and Kondners model tests results suggest that
isnotaproblem;suchstructures
are self-damping.
forloose sand, stripanchorage at adepth of 6.1 m
(20ft) and approximately 4.6 m (15 ft) wide would carry
Experimental
work
onmodelshasshownthat
any
aworkingload
of about 203 Mg (200 ton) above the
difficultiesinthisrespectforagivendesigncan
be
overcome by altering membersizes, geometry or prewatertable.
Withthe water table at ground level, a
tension,whicheveristhemost
suitable21. The vibra6.1 m (20 ft) wide strip anchor at a depth of 9.2 m (30 ft)
tionarybehaviour
of tensionstructuresiscurrently
would probably be required for the same working load.
beinginvestigatedandacomputerprogramforthe
For the same working load of 203 Mg (200 ton) it is concalculation of the natural frequencies of tension strucsidered that for dense sand with the ground water table
tures developed.
below the anchor, a strip anchor about3.6 m (12 ft) wide
at a depth of 7.6 m (25 ft)would be suitable, and with the
water table at ground level, the strip anchor would need
GROUND ANCHORS
Research into the bearing capacity of ground anchors is
to be 6.0 m (19.7 ft) wide at a depth of 6.0 m (19.7 ft).
still in itsearly stages. There is, however, a great deal of
practical experience in
installing
different
types
of
DESIGN EXAMPLES
ground anchors and load testshave been made on these
Engineers and architects concerned with the design of
anchors.
The
design
is
still
semi-empirical,
using
cable structures have often pointed out the economical
equations based on model test results and on a
fewer
use of materials in tension structuresand in reference 9
number of field test results.
a number of different designs are described.
Gravityanchors may, of course,beusedformost
In the following descriptions
are given of two different
situations but these would be very costly.
designs developedat the Cable Structures Research
In cohesive soils, some type of tension pile is probably andDevelopmentCentreinconnectionwithindustry.
the most convenient and the most economical solution.
The first is a circular roof for a water reservoir and the
In chalk, too, piles are likely to be suitable. The pilesmay
second an aircraft hangar.
either extend to ground level and terminate in a pile cap,
or consist of a
fairly
small-diameter
fixed
anchor
Pretensioned rooffor 61 m (200ft) diameter
at some
depth
below
ground
level. A cable,
circular water reservoir
A diagram of the above roof is shown in Fig 8. The roof
homed
in
the
fixed
anchor,
is
prestressed
to
the
structureconsistsofa
two-way pretensionedgirder
requiredworkingloadagainstafreeanchorblock
grid whose maximum depth is 4.6 m (15 ft) and in which
at ground level. Theanchors may be under-reamed.
Using
the
analogy with
compression
a
pile, the
the main load-carrying elements are made from 35 mm
loadwillbecarriedmainly
by adhesionalongthe
(13 in) diameter high-tensile steel rods spaced at 6.1 m
(20 ft) centres in plan top and bottom. The forces in the
shaft of the pile, with
some
load
carried
by
rods are resisted by aconcretering
having acrossend bearingwiththefixed-anchortype
ofpile.
The
problem to be determined is the value of the adhesion
section 1-83 m (6 ft) wide x 0.6m (2 ft) deep. The ring
betweenthepileshaftandthesoil.Forboredcast
beam is held rigidly to the top of the reservoir wall. The
struts aremade fromaluminiumtubingandtheroof
in-situ compression piles in London
clay, which is an
claddingfrom deep troughaluminium
sheeting. The
over-consolidated stiff fissured
clay, Whitaker andCook23
shape of the roof is thatof a paraboloid. Theadvantage
givea valueof Ca = 0.45 x undrainedcohesion Cu.
of this form of construction is that the snap-through
On the basis of tests with bored piles in London
clay,
problem,whichisthemaindesignconsiderationin
by ArmishawoftheCableStructure
Research and
certain types of flat domes, is non-existent. Excessive
Development Centre, it appears that the adhesion for
snow-loading will causeonly increased displacements
tension piles is considerably less than this
value and
without causing structural failure due to instability.
it is considered that a value of Ca = 0.25 Cu is more
52

THE STRUCTURAL ENGlNEER

FEBRUARY 1970 0 No 2 U VOLUME 48

-I

l
200'-0-

Fig. 8

Aircraft hangar for two Boeing 747

For the purposes of this design study it was assumed
that the site conditions and the client's requirements
were such that two Boeing
747 aircraft were to beaccommodated wing-tip to wing-tip in the
same hangar. Allowing for an independent crane system on rubber tyres,
it was found that the aircraft-bearing
pavement would
occupy a clear
area of 146 m (480ft) by 67 m(220ft),
assuming that the nose of
each aircraft couldbe accommodated in extensions to the main building at the rear.

The 146 m (480ft) opening in the front of the building

was to be closed by eight folding doors supported on a
bottom rail, each of 18.3 m (60ft) length folding to 3 m
(
O
I ft). Allowance was to be made so that each door
could run anywhere within its own half of the building.
Atraditionalsolutiontothisproblemwouldalmost
certainly
have
utilised
heavily
braced
'gun-frames'
cantilevering in the direction of shortest span. While
it is
appreciatedthatthisformofdesigndoesprovidea
practical answer to the problem, it is suggested that a
more competitive solution is embodied in the structure
described below.
The structural form used is best seen in Fig 9 and 10
which illustrate a scale model of the hangar. The main
structural element is a 146 m (480ft) span cable-girder
truss fabricated from 48 mm (I Qin) diameter galvanised
locked coil cable and high-yield steel rectangular hollow
sections.Twenty-threegirder-trussesat3m
(IO ft)
centres are supported on 12 steel columns, 0.6 m (2ft)
CHS, on either side at 6 m (20 ft) centres along the side
of the
building,
intermediate
girder-trusses
being
fastened t o edge beams spanning between the columns.
The roof slopes from a height of
32 m (105 ft) at the front
to 26 m (85 ft) at the rear. The mid-span depth of each
girder-truss is 7.6 m (25 ft). Bending in the main columns
is eliminated by the provision of braced trestles at the
endsof eachgirder-truss.Thesetransmitthetensile
forces in the cables to the ground, absorb wind forces
across the roof and support side
sheeting.
The roof cladding, troughed steel decking, is clipped
directly on to the main tendons and in conjunction with
thelongitudinalbracing,whichconnectthegirdertrusses, gives a high
degree of lateral stiffness to the
roof.
All wind forces are conveyed to the foundations by
counter bracing between the main columns at the end
and centre bays.
Thetotalamountof
materialused inthestructure
itselfincludingthenose-boxesandthewindgirder,

Fig. 10
THE STRUCTURAL ENGlNEER 0 FEBRUARY 1970

0 No 2 0 VOLUME

48

53

which takes the horizontal wind reaction at the top

of
the doors, is approximately
5480 kN (550 ton) of hightensile steel and 797 kN (80 ton) of cables. Whilst an
independent
crane
system obviously
is
the
most
economicalsolution,theuse
of roofcranesisnot
impossible in a building fo this
nature. The problem is to
limit the deflections. This can be achievzdby increasing
the size andcurvature
of the cables. A preliminary
design study using 64 mm (24 in) diameter cables has
shown that this only marginally affects the
total weight
of the steel in the boundary structure but considerably
increases the weight of the roof itself.

ACKNOWLEDGMENTS
Theauthorwishesto
express hisgratitudeto
Dr.
M. J. L. Hussey, Kings College, for many fruitful discussionsinconnectionwiththetheoretical
work, to
Mr. J. A. Boyle, British Ropes Ltd., and Mr. H. B. Walker,
StewartsandLloyds
Ltd., fortheir help,advice
and
assistancewiththedesign
of thehangarandto
Mr.
Frank Kubertski for making the hangar model.

REFERENCES
l.Brotton, D. M., The application of digital computers to sfructural engineering problems, Spon, London, 1962.
2. Buchholdt, H. A., Deformation of prestressed cable nets.
Acta
Polyfechnica
Scandinavica,
Ci 38, Norges
tekniske
videnskaps akademi, Trondheim, Norway, 1966.
3. Buchholdt, H. A., The behaviour of small prestressed cable
roofssubjectedtouniformlydistributed
loading,Internationalconferenceonspacestructures,London,
1966.
4. Buchholdt, H. A., Prestressedcablenetswithfinitedisplacements, P h D Thesis,University of London, 1967.
5. Buchholdt, H. A., The configuration of prestressed cable
nets, Acta Polyfechnica Scandinavica, Ci 54, Norges tekniske
videnskaps akademi, Trondheim, Norway, 1968.
6. Buchholdt, H. A., A non-linear deformation theory applied
to two dimensional pretensioned cable assemblies,
Proc.
ICE, Vol. 42, January 1969, pp. 129-141.

7. Buckingham, R. A., Numerical methods, Pitman, London.

8. Dean, D. L. and Ugarte, C. P., Analysis of structuralnets,
IABSE publications, 1963, pp. 71-90.
9. Hanging Roofs, Proceedings of /ASS colloquium in Paris,
1962, Amsterdam, 1963.
10. Hoff, N. J., The analysis of structures, Chapman and Hall
London, 1962.
11. Hussey, M. J. L., Skeletalassembliessubjecttogeometricallysignificantdisplacements,2ndPan-American
Conference, Lima, Peru, 1962.
12. Fletcher, R. and Reeves, C. M., Functionminimization by
conjugate gradients, The ComputerJournal, No. 2, Vol. 7,
p. 149, 1964.
13. Levey, H. andLessman,
F., Finite Difference Equations,
Pitman, London, 1959.
14. Livesley, R. K., Matrix methods of structural analysis, Pergamon Press, London, 1964.
15. Mnrllman, H., Astudy in the theory of suspended sfrucfures,
Akademisk Forlag, Denmark, 1965.
16. Otto, F., Das hdngende Dach, Bauwelt Verlag, Berlin, 1954.
17. Poskitt, T.J., Numerical solution of non-linear structures.
Proc. Am. Soc. C,1967, 93 (Aug.), ST4, 69-94.
18. Schleyer, F. K., DieBerechnungvonSeilnetzen,paper
in (13).
19. Shore, S. and Bathish, G. N., Membrane analysis of cable
roofs,
International
conference
on
space
structures,
London, 1966.
20. Siev, A.andEidelman,
J., Shapesofsuspendedroofs,
paper in (13).
21. Jawerth, D. andSchulz, H., Ein BeitragzurFrageder
Eigenschwingungen Krafte und aerodynamischen Stabllitat
bei
hangenden
Dachern,
Der Stahlbau 35, Jahrgnan,
Heft l/ Januar 1966.
22. Livesley, R. K., Matrix methods of structuralanalysis, Pergamon Press, 1964, pp. 241-258.
23. Whitaker,T.andCooke,
R. W., An investigationofthe
shaft and base resistances of large bored piles in London
Clay, Symposiumonlargediameterboredpiles,London
1966.
24. Armishaw, W., personal communication.
25. Littlejohn, G. S., Recentdevelopment ingroundanchor
construction, Ground Engineering, Vol. 1, No. 3, May 1968.
26. Balla, A., Theresistancetobreakingoutofmushroom
foundations for pylons. Proc.4th lnt. Conf. onSoil Mechanics and Foundation Engineering, Vol. 1.

correspondence
UDC 666.928:624.072.2:624.046.5

Limit design of beams for two-way reinforced concrete

slabs by R. Park, published in The Structural Engineer,
Vol. 46, No. 9, September 1968, pp. 269-274.

Mr. John M. Rolfe: It is pleasing to see the use of

Johansens yield line theory being given some publicity,
since its power and versatility are not as widely appreciated as they deserve. However, there a few points on
which I would like to comment.
With
regard
to
proportioning
the
reinforcement
I feelthatthe
betweenthespans
of atwo-wayslab
proportions (but not the quantities) laid downby the old
Rankine coefficients have much to commend them.
(a) They give areasonableapproximation
tothe
elastic
distribution
of moment,
reducing
the
possibility of yield, in either direction, occurring
under working loads.
(b) Using the valuesof ,U so obtained and inverting
L

the L! ratio which determines the

Lx

54

shapeof

the

slab, results in a single set of coefficients applicable to both directions with the qualification that
the bending moment in each direction is obtained
fromthe
other span: i.e., M z = cwLy2 and
MY/ = CWL$.
(c) A safe side approximation to this coefficient in
1 Lx
thenormal
range of slabshapes
is - - (L,
20 L!/
being the shorter dimension) which avoids a lot
of reference to tables.
The authors difficulty with panels with unsupported
exterior edges would appear to be overcome if the nodal
forces are included in the external loads acting on the
segment. If this is done then
MIYdetermined by equation
19 isequaltothe
value obtained by consideringthe
external forces.
I do not agree that a similar difficulty ariseswhen the
yield lines from the corners
meet and run out to the
edge
as a single yieldline. Referring toFig Cl :

THE STRUCTURAL ENGlNEER U FEBRUARY 1970 c1 N o 2 U VOLUME 48