Engineering Structures 79 (2014) 234243

Contents lists available at ScienceDirect

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

Tensile performance of prestressed steel elements

Jonathan Gosaye a,, Leroy Gardner a, M. Ahmer Wadee a, Murray E. Ellen b
a
Department of Civil and Environmental Engineering, Imperial College London, London, UK
b
S2, Space Solutions, Sydney, Australia

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

Article history: Prestressed steel trusses can offer efcient structural solutions for long span applications such as aircraft
Received 17 March 2014 hangars, stadia and warehouses. In the current study, the tensile behaviour of prestressed steel elements,
Revised 4 August 2014 which comprise tubular steel members with internal prestressing cables, is investigated. The stability of
Accepted 5 August 2014
the elements under prestress and the loaddeformation response of the prestressed elements to the sub-
Available online 1 September 2014
sequent application of tensile loading are examined analytically, numerically and experimentally, with
good correlation achieved between the three approaches. The benets of prestressing, in terms of
Keywords:
increased member strength and stiffness, are demonstrated, and optimal prestress levels are investigated.
Prestress
Steel
2014 Elsevier Ltd. All rights reserved.
Tension
Cable-in-tube system
Truss
Stability
Analytical
Numerical
Experimental

1. Introduction Early work on the prestressing of steel structures was described

For long-span structural systems, where self-weight becomes
an increasingly dominant component of the design loading, signif-
icant material savings can be achieved through the use of high
strength tensile steel cables in conjunction with conventional
steelwork. Additional benets can be gained by prestressing the
cables, inducing internal forces in the structure that can counteract
the applied external loads and control self-weight deections.
Cable-in-tube systems, whereby the steel cables are housed
within hollow structural sections, offer a practical means of realis-
ing the aforementioned concept, and bring further advantages such
as ensuring the stability of the tube under the prestressing forces
due to the stabilizing action of the internal tensioned cable. Exam-
ples of recent applications of prestressed cable-in-tube systems
include the reconguration of the Sydney Olympic stadium and
the Five Star Aviation hangar at Brisbane Airport, both in Australia
[1].
Corresponding author. Address: Imperial College London, Department of Civil
and Environmental Engineering, Skempton Building, Room 428, London SW7 2AZ,
UK. Tel.: +44 7907389916.
E-mail addresses: jg607@imperial.ac.uk, jonathan.gosaye-da-kaba07@imperial.
ac.uk (J. Gosaye), leroy.gardner@imperial.ac.uk (L. Gardner), a.wadee@imperial.ac.
uk (M. Ahmer Wadee), Murray.Ellen@s-squared.com.au (M.E. Ellen).
by Magnel [2] in 1950, where it was shown experimentally that
improved economy can be achieved by prestressing truss girders.
More recent studies have explored the behaviour and design of
prestressed steel beams [3,4], columns [59], trusses [10,11] and
space trusses [12,13]. Studies of the structural response of sub-
assemblies and the overall response of prestressed frames with
sliding joints have also been carried out [1417], as has a numeri-
cal investigation into the stress-erection process of such systems
[18]. Each of the above described studies identied potential econ-
omies and enhanced performance through the use of prestressing.
In the system currently being investigated, developed by the
company S2 Space Solutions [19], the prestressed cables, housed
within the bottom chord of the tubular arched trusses, apply a
compressive force to the chord members, which is opposite in
nature to the resultant forces arising from the externally applied
gravity loads. The behaviour of trussed elements in the scenario
described above, i.e. pre-compressed chord elements subsequently
subjected to external tensile loading, is the subject of investigation
in the present paper. The case of wind uplift, which would result in
additional compressive loading of the prestressed element, will be
examined in a companion paper.
Analytical and numerical models that predict the tensile
behaviour of the cable-in-tube system, including the axial response
during prestressing and under the application of tensile loading,
and the stability of the system under prestressing are developed

http://dx.doi.org/10.1016/j.engstruct.2014.08.009
0141-0296/ 2014 Elsevier Ltd. All rights reserved.
 J. Gosaye et al. / Engineering Structures 79 (2014) 234243
in Section 2. Experimental studies are described in Section 3, while
comparisons between the test and analytical results and discus-
sions thereof, are made in Section 4. Note that although previous
research provides an insight into the possible inuence of long-
term effects, such as creep and shrinkage, on the response of struc-
tures [20,21], these have not been addressed in the present study.
Further research is required on this topic.
2. Analytical modelling and numerical verication
In this section, the key behavioural aspects of prestressed steel
elements are described, including their response to prestressing
and under the subsequent application of tensile load. The stability
of prestressed elements during the prestressing stage, the optimal
level of prestress and tensile capacity are also investigated.
2.1. Axial response under prestress
During the prestressing stage of cable-in-tube systems, a tensile
force, P, is applied to the cable and a compression force of equal
magnitude is induced in the tube, as depicted in Fig. 1, where Lo
is the original length of the cable and tube, xc is the extension of
the cable and xt is the shortening of the tube.
When the prestressing process is complete the cable and the
tube are locked in position at both ends such that there is no rela-
tive movement between the two components, as shown in
Fig. 1(c). Since the prestressing force, P, of the same magnitude,
but opposite in direction, is applied to the cable and the tube,
the prestressed element is self-equilibrated when the tube and
cable are attached at both ends. Therefore, provided no out of plane
displacements (i.e. no buckling) are induced during prestressing
the equilibrium condition for the prestressed system can be stated
as in Eq. (1).
Pcable  Ptube K c xc  K t xt 0; 1 The numerical modelling was conducted in two stages. In the
where K c and K t are the axial stiffnesses of the cable and the tube
respectively. The respective stiffnesses are given by the following
expressions:
Ec Ac Et At
Kc ; Kt ; 2
Lo Lo
where Ec and Et are the Youngs moduli of the cable and the tube
respectively with Ac and At being the cross-sectional areas of the
cable and the tube respectively.
Lo
(a) P
xc
Lo
(b) P prior to the cable; Case II refers to where the cable yielding prior
xt
Lo
(c)
xc x t
Fig. 1. (a) Force diagram of cable during prestress; (b) force diagram of tube during
prestress; (c) tube and cable locked in position after prestress.
235
2.2. Stability under prestress
Previous analytical studies [22,23] have shown prestressing
elements (i.e. the cables, in the context of the present study) can
stabilize their compressed counterparts (i.e. the steel tubes hous-
ing the cables) provided that there are sufcient connection points
between the two elements. The connection points have been
shown to offer effective lateral restraint and thus reduce the buck-
ling length of the compressed element. If the cable is in constant
contact with the tube, cable-in-tube systems have innite connec-
tion points, which implies an innite elastic buckling load [23].
Buckling under prestress would not therefore be anticipated in
these systems.
To verify the above analytical ndings, a nite element model
was implemented using the commercial nite element (FE) analy-
sis software, ABAQUS [24]. The tube to cable cross-sectional area ratio
At =Ac and Youngs modulus ratio Et =Ec were 8.96 and 1.58,
respectively. Prestress was introduced through thermal loading
and the coefcient of linear expansion of the cable, ac , was set as
12  106 K1 .
In the numerical simulation, truss elements were used to model
the cables since they can only resist axial forces while beam
elements, which possess both axial and exural stiffness, were
used to model the tube. The cable and the tube were attached
using tie constraints at the ends and constraint equations were
applied at various connection points along the length of the tube
and cable to prevent relative out of plane motion between the
two components.
The number of connection points was varied between one (at
mid-length) and three (at quarter points). Finite intervals were
chosen since in the experimental investigation of Section 3, the
cable was signicantly smaller in diameter than the steel tube
and connections between the cable and the tube were made by
means of collars spaced regularly along the member length. This
is described in Section 3.
rst stage, an elastic buckling analysis of the tubular element
under the prestressing action of the cable was carried out and
the buckling mode shapes were obtained. In the next stage, a geo-
metrically nonlinear analysis utilising the Riks arc-length solution
technique [25], was performed with the rst eigenmode from the
elastic buckling analysis as an initial imperfection with an ampli-
tude of Lo =100000. An incremental prestress was applied to the
cable through thermal loading. As illustrated in Fig. 2, the FE
results are in close agreement with the analytical results and show
that the buckling load N cr with respect to that of the unsupported
member N E , increases in proportion to 1=L2cr , where Lcr is the spac-
ing between the connection points (or collars).
2.3. Behaviour of prestressed system under tension
Analytical expressions for the axial load-displacement equilib-
rium paths of prestressed elements under tensile loading are
derived in this section. Three cases are considered in the following
sub-sections: Case I refers to the scenario where the tube yields
to the tube; and Case III considers the two elements yielding simul-
taneously. The case which actually arises depends on the geomet-
ric and material properties of the tube and cable, and the level of
prestress, as described in the following sub-sections.
2.3.1. Case I: Tube yields prior to cable
The three stages of behaviour that occur for Case I are illus-
trated in Fig. 3, where the adopted notation is dened. The ratios
of geometric and material properties between the tube and cable,
as well as the level of applied prestress determine whether the
236 J. Gosaye et al. / Engineering Structures 79 (2014) 234243

18
Analytical: 3 collars
16
Tube
14 FE: 3 collars
N Lcr
12

10 Analytical: 2 collars
N/N E

Cable
8
FE: 2 collars
6 Analytical:1 collar Connection
point
4 (collar)
FE: Tube only FE: 1 collar Analytical:Tube only N Lcr
2

0
0 10 20 30 40 50
Axial displacement (mm)

(a) (b)
Fig. 2. Buckling response of tubular elements under the action of prestressing forces induced through cables connected to the tubular elements at regular intervals along the
length. (a) Analytical and numerical results for buckling behaviour of prestressed element. (b) Buckled conguration of tubular element with one connection point to
prestressing cable.

Start of stage 1: axial force and it will continue to store tensile strain energy, under
Lo-xt Tube increasing N. In the sign convention adopted, positive values
Cable denote tension and elongation and negative values denote com-
(a) Stage 1: N pression and shortening, with the axial displacement x1 measured
from the state of the system at the end of the prestressing process.
x1=xty+xt Stage 1 applies from the onset of loading until the tube yields in
tension at x1 xty xt , as seen in Fig. 3(a). During this stage, both
Start of stage 2: the tube and the cable remain elastic and the total potential energy
Lo+xty function V for the system can be expressed thus:
1 1
V K c x1 xc 2 K t x1  xt 2  Nx1 : 4
(b) Stage 2: N 2 2
An expression for the equilibrium path for this loading stage, Eq. (5),
x2=(xcy-xc)-(xty+xt) is then obtained by differentiating the total potential energy func-
tion V with respect to the axial generalised coordinate, x1 , and sat-
Start of stage 3: isfying the equilibrium condition by setting the resulting expression
Lo+(xcy-xc-xt) to zero:
N K c K t x1 : 5
(c) Stage 3: N
The forces in the tube Nt and cable N c can therefore be expressed as:

x3=xf -(xcy-xc) Nt K t x1  xt ; 6
Nc K c x1 xc : 7
Fig. 3. Generalised coordinate denitions for Case I. Note: displacements in each
stage are expressed with respect to tube deformation.
2.3.1.2. Stage 2: From tube yielding to cable yielding
 
0 6 x2 6 xcy  xc  xty xt . In the second stage, as shown in
tube yields before the cable or vice versa. The tube yields prior to Fig. 3(b), the tube has yielded while the cable continues to store elas-
the cable if: tic tensile strain energy as before. This stage continues until the
f cy P f ty P cable yields at an axial displacement relative to the state of
 > ; 3 the system at the end of prestressing x1 xcy  xc or, relative to
Ec Ac Ec Et At Et
the end of Stage 1, at an axial displacement x2 xcy  xc  xty xt .
where f cy and f ty are the yield stresses of the cable and the tube The potential energy V now becomes:
respectively. 1  2 1 1
V K c xc xty xt x2 K t x2ty K t xty x2 K st x22  Nx2 8
  2 2 2
2.3.1.1. Stage 1: Onset of loading until tube yielding 0 6 x1 6 xty xt .
and the applied load N with the component forces Nt and N c are
The prestressed element, with the tube initially in compression and
given thus:
the cable in tension, both of magnitude P, is subjected to a tensile
load N as shown in Fig. 3. N K c xc xty xt  K t xty K c K st x2 ;
The initial compressive axial force in the tube unloads until a
Nt K t xty K st x2 ; 9
tensile displacement of xt is obtained, from which point the tube
is in net tension. The cable, on the other hand, has an initial tensile Nc K c xc xty xt x2 ;
 J. Gosaye et al. / Engineering Structures 79 (2014) 234243 237

where K st is the stiffness of the tube after yielding allowing, where Table 1
appropriate, for strain hardening. Analytical solution for tensile response of prestressed elements in Case II.
 
Stage Onset of loading until cable yielding: 0 6 x1 6 xcy  xc
2.3.1.3. Stage 3: From cable yielding to fracture 0 6 x3 6 xf 1
xcy  xc . This is the nal stage of loading, in which both the N K c K t x1
N t K t x1  xt
tube and the cable have yielded, and are potentially strain harden- N c K c x1 xc
ing, as shown in Fig. 3(c). The axial displacement x3 is measured
Stage From cable yielding to tube yielding:
from the position at which the cable yields, i.e. at the end of Stage 2
 
0 6 x2 6 xty xt  xcy  xc
2. Stage 3 continues until either fracture of the tube or, more likely, N K c xcy K t xt xcy  xc  K t K sc x2
fracture of the cable at a displacement relative to the state of the  
N t K t xt xcy  xc x2
system at the end of prestressing being: N c K c xcy K sc x2
     
P P Stage From tube yielding to fracture: 0 6 x3 6 xf  xty xt
xf Lo cf  or Lo tf ; 10 3
Ac Ec At E t  
N K c xcy K t xty K sc xty xt  xcy  xc K sc K st x3
where cf and tf are the fracture strain of the cable and tube respec- N t K t xty K st x3
tively. Following similar procedures to the previous loading stage,  
N c K c xcy K sc xty xt  xcy  xc x3
the expression for V is:
1 1 1 
V K c x2cy K c xcy x3 K sc x23 K t x2ty K t xty xcy  xc
2 2 2
 1  2
xty xt x3 K st xcy  xc  xty xt x3  Nx3 11 Start of stage 1:
2 Lo-xt Tube
and the applied load, together with the component forces, are given Cable
thus: (a) Stage 1: N
 
N K c xcy K t xty K st xcy  xc  xty xt K sc K st x3 ;
  x1=xcy-xc
Nt K t xty K st xcy  xc  xty xt x3 ;
Nc K c xcy K sc x3 ;
Start of stage 2:
12 Lo+(xcy-xc-xt )
where K sc is the stiffness of the cable after yielding. Note that post-
fracture of either component (cable or tube), the load is carried by (b) Stage 2: N
the remaining component, with N N c or N N t , depending on
which component fractures rst. x2=(xty+xt)-(xcy-xc)

2.3.2. Case II: Cable yields prior to tube

Start of stage 3:
Case II refers to the scenario in which the cable yields prior to Lo+xty
the tube, which occurs when:
f cy P f ty P (c) Stage 3: N
 < : 13
E c Ac E c Et At Et
x3=xf -(xty+xt)
The analytical formulations to describe the tensile response of
prestressed elements in Case II were derived in a similar manner Fig. 4. Generalised coordinate denitions for Case II. Note: displacements in each
to Case I; the resulting expressions for the three stages of loading stage are expressed with respect to tube deformation.
are summarised in Table 1; Fig. 4 illustrates the generalised
coordinate system used for this scenario. Note that the Stage 1 and
Stage 3 expressions are the same as those for Case I, since the tube Note that achieving this optimal prestress level without inducing
and cable are both elastic (Stage 1) or plastic (Stage 3), although plastic strains during the prestressing operation itself is not possible
the end points of Stages 1 and 2 and the expressions within Stage 2 for all tube and cable material and geometric properties. To prevent
differ. plastic strains during prestressing, the applied prestress must not
exceed the yield load of either the tube or cable, as stated in Eqs.
2.3.3. Tube and cable yield simultaneously (optimal prestress, P opt;t ) (16) and (17):
The optimal prestress level under tensile loading may be
dened as the one that causes the tube and the cable to yield Popt;t 6 f ty At 16
simultaneously, since this provides the most extensive elastic
and
range and hence the stiffest response.
The optimal prestress level under applied tensile load, P opt;t , can Popt;t 6 f cy Ac : 17
be derived by equating the expressions for the strain required to
yield the pretensioned cable to that of the precompressed tube,
as shown in Eq. (14): 2.3.4. Sample analytical results and nite element validation
  In this sub-section, sample analytical results are presented and
f cy Popt;t f ty Popt;t
 ; 14 comparisons are made with corresponding numerical results gen-
Ec Ac Ec Et At Et erated using the FE software ABAQUS [24]. The material and geomet-
rearranging in terms of P opt;t gives rical properties of the different components used in the analytical
and numerical models are given in Table 2. Note that both the tube
At Ac
and cable material stressstrain (r) response was modelled as
Popt;t f Et  f ty Ec : 15
At Et Ac Ec cy elastic, perfectlyplastic, with no strain hardening.
238 J. Gosaye et al. / Engineering Structures 79 (2014) 234243

As described in Section 2.2, truss elements were used to model 600

the cable and beam elements were used to model the tube. The
two components were tied (all displacement degrees of freedom)
500 System response
along the member length using tie constraints. In the rst step,
prestressing was applied through thermal loading of the cable
which caused the cable to contract, resulting in compression in 400
Tube

Force (kN)
the tube and tension in the cable. In the next step, the load was
applied incrementally through displacement control until the ulti- 300
mate load of the prestressed system was reached.
Two levels of preload were considered, both to compare the
200
analytical and numerical results as well as to illustrate the funda- Cable
mental mechanics of the prestressed element under different pre-
stress levels. Fig. 5 shows the results obtained for a system with no 100 Analytical
prestress. Numerical
The numerical and analytical models may be seen to be in very 0
close agreement. For the chosen section properties, the rst reduc- 0 5 10 15 20 25 30
tion in stiffness occurs when the tube yields, while the ultimate Axial displacement (mm)
load is reached once both components have yielded. The tube
yields prior to the cable since the chosen cable has a lower Youngs Fig. 5. Tensile performance of a prestressed element with zero initial prestress.

modulus and a higher yield strength than the tube. The drop in
stiffness of the system due to the yielding of the tube is greater
550 Prestressed system
than the case when the cable yields since the axial stiffness ratio
K t =K c 9:2. Analytical (prestressing )

The results for the second preload condition, where the optimal Analytical (loading ) 450
preload P opt;t determined from Eq. (15) was applied, are shown in Numerical (prestressing)
Tube
Fig. 6. The material and geometrical properties are again as given Numerical (loading) 350
in Table 2, and close agreement between the analytical and numer-
Force (kN)

ical results may be observed. During the prestressing process the
cable is subjected to tensile forces and the tube is subjected to
compressive forces of the same magnitude. These preload forces,
along with the corresponding displacements, are plotted in Fig. 6
following the sign conventions dened in Section 2.3.1.
For this prestress level, Fig. 6 shows that the initial stiffness of
the system is maintained until the ultimate load is reached, at
which point both the cable and the tube yield simultaneously. As
a result of maintaining the initial stiffness, the displacements
required to reach the ultimate load are minimised. Increasing pre-
stress levels beyond Popt;t would have the effect of yielding the
cable prior to that of the tube and thereby increase the displace-
ments at ultimate load, while a lower prestress also results in
higher displacements at ultimate load, but now due to earlier
yielding of the tube.
3. Experimental study
An experimental study was performed in the Structures
Laboratory at Imperial College London to verify the analytical
and numerical ndings for the tensile performance of prestressed
cable-in-tube systems. Material testing, prestressing operations
and the tensile member tests are described.
3.1. Introduction
A total of 7 cable-in-tube specimens were tested: 4 being
non-grouted and 3 being grouted. The key variables of the test
Table 2
Properties of prestressed element in the analytical and numerical model.
Yield strength of tube, f ty 355 N=mm2
Yield strength of cable, f cy 1860 N=mm2
Youngs modulus of tube, Et 210 kN=mm2
Youngs modulus of cable, Ec 200 kN=mm2
Cross-sectional area of tube, At 879 mm2
Cross-sectional area of cable, Ac 100 mm2
Coefcient of linear expansion of cable, ac 12  106 K1
Initial length of tube and cable, Lo 2000 mm
250
150
Cable
50
-15 -10 -5 -50 0 5 10 15
-150
Axial displacement (mm)
Fig. 6. Tensile performance of a prestressed element with P opt;t initial prestress.
programme were the prestress level, the presence of grout and
the tube to cable cross-sectional area ratio, At =Ac . The cross-section
used for all specimens was a cold-formed and seam welded
60  60  4 square hollow section (SHS) of grade S355 steel. The
cables were 7 wire prestressing strands of grade Y1860S7 steel.
The adopted test specimen labelling system is explained in Fig. 7
and a full list of the test specimens and their properties is given
in Table 3. Note that three target prestress levels were considered:
Pnom ; 12 Popt;t and P opt;t where Pnom is a nominal prestress of 5 kN to
ensure that the cables were not slack during grouting and Popt;t is
the optimum prestress described in Section 2.3.3. The achieved
(measured) levels of prestress are given in Table 3 and differ
slightly from the target values.
3.2. Material tests
3.2.1. Tube
Tensile coupon tests were carried out to determine the basic
engineering stressstrain response of the SHS tubes used to fabri-
cate test specimens. Coupons were cut from the at and corner
regions of the cross-section. All at coupons had a parallel necked
region of length 150 mm and width 20 mm. The curved corner cou-
pons were parallel with a nominal length of 320 mm.
An Instron 8802 250 kN hydraulic machine was used to carry
out the tensile coupon tests in accordance with EN10002-1 [26].
 J. Gosaye et al. / Engineering Structures 79 (2014) 234243 239

Applied force: Prestress

T = tensile level (kN) Machine Cable Wedge
head Barrel
T-1-NG-5
Threaded
No. of Grouted (G) or hole
cables not grouted (NG)
Fig. 7. Test specimen labelling system adopted. Bolt
End plate
Stiffener
Table 3 Tube
Summary of prestressed cable-in-tube test specimens.

Specimen Number of cables Grouted (Y/N) Prestress (kN)
T-1-NG-5 1 N 5P nom
T-1-NG-68 1 N 680:44P opt;t
T-1-NG-126 1 N 1260:82P opt;t
T-2-NG-108 2 N 1080:38P opt;t
T-0-G-0 0 Y
T-1-G-68 1 Y 680:44P opt;t
T-1-G-131 1 Y 1310:86P opt;t
The machine was driven under strain control at a rate of 0.001%/s
up to the 0.2% proof stress and at a rate of 0.002%/s beyond the 0.2%
proof stress and up to the ultimate tensile stress, from which point
a strain rate of 0.04%/s was applied until fracture. Two linear elec-
trical resistance strain gauges placed on either side of the coupons
and used to measure the strain. The applied load and axial dis-
placement and other relevant were recorded at 1-s intervals. The
measured weighted average (based on the cross-sectional areas
of the at and corner regions) tensile material properties of the
tube are given in Table 4 in which E is the Youngs modulus, r0:2
is the 0.2% proof stress, ru is the ultimate stress, u is the percent-
age strain at the ultimate tensile stress, f is the percentage strain
at fracture over the standard gauge length and n and n00:2;u are the
RambergOsgood parameters [27].
3.2.2. Cable
The nonlinear stressstrain curve for the 7-wire strand cable
was determined by testing a representative two metre long sam-
ple. The cable was passed through end plates at either end and
anchored using barrels and wedges. The end plates were bolted
onto the testing machine using M30 socket head cap screws.
Fig. 8 shows the end connection for the test specimens, although
the same end plate bolting arrangement was also used in testing
the cable alone. The test was performed using an Instron
2000 kN machine. The machine was driven under displacement
control with a constant displacement rate of 0.5 mm/min until
fracture. The diameter of each strand was measured and the
cross-sectional area of the 7-wire strands cable was determined
by summing up the cross-sectional areas of the individual strands.
The recorded axial load and displacement were used together with
the measured length of sample and its cross-sectional area to
determine the experimental stressstrain curve. The measured
material properties of the cable are also reported in Table 4.
3.2.3. Grout
Some of the test specimens were grouted to reect common
construction practice. The grout provides a bond between the tube
At =Ac Fig. 8. End plate bolting arrangement for testing of cable-in-tube systems and
cables alone.
8.8
8.8
8.8
4.4 and the cable, which reduces reliance on the end anchorages and
serves to protect the cables from corrosion. Note, however, that
8.8 long-term behaviour of the grout has not been examined herein.
8.8 In this study, a Portland cement-based grout, with plasticiser used
to enhance owability, was used. The water to cement ratio was
0:35. Samples were taken from each grout mixture to test strength
in accordance with EN 12390-3 [28]. The grout was poured into
100 mm cube moulds and were left to dry for one day. The samples
were then removed from the mould and cured in water at room
temperature. The cubes were tested at various time intervals using
an Automax5 testing machine under load-control at a constant rate
of 3 kN/s. Two samples were tested at each time interval and the
average cube strength f cu is reported in Table 5. The cable-in-tube
specimens were tested 7 days after prestressing and grouting.
3.3. Preparation of specimens and experimental setup
The cold-formed 60  60  4 SHS tubes were cut to the required
length of two metres, which is a typical length for a bottom chord
element used in the construction of prestressed truss structures.
The tubes were welded onto 20 mm thick end plates and stiffeners
were added, as shown in Fig. 11, to prevent premature failure at
the ends. The end plates had been drilled to allow the specimens
to be bolted on to the testing machine directly, as illustrated in
Fig. 8.
To prevent buckling of the tubes during the prestressing pro-
cess, specimens with initial prestress levels higher than the nomi-
nal prestress level, Pnom (5 kN), had circular connecting collars
tted to the cables at mid-length and at quarter points, as shown
in Fig. 9.
The circular connecting collars were machined to an external
diameter equal to the inside dimension of the SHS and had a cen-
tral hole to accommodate the cable. The cable, with connecting col-
lars attached at Lo =4; Lo =2 and 3Lo =4 from the member ends, was
then inserted into the tube. The buckling length of the test speci-
mens under preload were therefore approximately one quarter of
their full length, as discussed in Section 2.2.
Prior to the prestressing of the specimens, the xed end anchor
was seated by applying an initial prestress of 5 kN. The prestress
was then released to t a wedge at the stressing end of the speci-
mens. Once both anchors were in place, the specimens were pre-
stressed using a T25 stressing jack. After each specimen was
prestressed to the appropriate level, the wedge at the stressing

Table 4
Measured weighted average tensile material properties of the tube and cable.

Component E kN=mm2 r0:2 N=mm2 ru N=mm2 n n00:2;u u % f %

Tube 193 354 408 12.0 2.0 3.2 4.1

Cable 125 1775 1880 10.2 1.5 3.6 3.6
240 J. Gosaye et al. / Engineering Structures 79 (2014) 234243

Table 5
Strength of grout with time.

1 day 3 days 7 days 28 days

2 15 38 53 66
Cube strength, f cu (N=mm )

Fig. 10. Prestressing conguration.

Fig. 9. Test specimen, also showing cable and connecting collar arrangement.

Fig. 11. Tension specimen test set-up.

end was hammered to reduce prestress loss due to anchorage slip
upon jack removal.
A summary of the geometric properties of the test specimens is
shown in Table 6, where Lo is the specimen length, b and d are the
to ow straight into the tube. Specimens to be grouted were set
outer width and height of the SHS specimens, t is the material
upright and grout was poured into the specimens at the top end
thickness, ri is the internal corner radius with At and Ac being the
using a funnel. The specimens were tapped, using a metal rod,
cross-sectional areas of the tube and cable, respectively.
along the length immediately after the grout was poured to ensure
that the grout had lled the tube.
3.4. Prestress application and grouting

During prestressing, a load cell was used to monitor the applied
force, while strain gauges, afxed to the four faces of the steel tube,
were used to verify the force in the tube and to monitor any ex-
ure, as shown in Fig. 10.
The tension in the cables was used to stabilize the compressed
tubular specimens during prestressing, with the collars used to
transfer the lateral resistance provided by the cable to the tube,
as discussed in Section 2.2. Fig. 12 conrms the stabilizing effect
of the cables during the prestressing process. The strain measure-
ments (on the primary vertical axis) and the prestress level (on the
secondary vertical axis) of specimen T-1-G-131, during the pre-
stressing stage, are plotted against the elapsed time in Fig. 12
where a and b are the axial and bending strains respectively.
Fig. 12(a) shows that at an applied prestress level of 96 kN, the
bending strain was only 13% of the axial strain whereas the bend-
ing strain was 27% of the axial strain when a tube of the same prop-
erties was tested under an externally applied compression of the
same magnitude. A prestress greater than the target level was ini-
tially applied to the specimens to take account of the unavoidable
loss of a small proportion of the prestress upon jack removal. In
some instances, the jack was removed during the prestressing
operation resulting in a temporary drop in load; an example of this
can be seen in Fig. 12 at approximately 0.04 h.
Grouting was carried out after prestressing. Specimens that
required grouting had holes in the end-plates to allow the grout
3.5. Experimental results
The full set of experimental results are shown in Fig. 13, which
gives the tensile loadextension response of the seven specimens.
The gure illustrates that the test specimens exhibit the character-
istic behaviour predicted by the analytical model discussed in Sec-
tion 2, whereby the addition of a cable increases the member
capacity and increasing the level of prestress up to Popt;t reduces
the displacements required to reach the ultimate load, i.e. the ini-
tial stiffness is maintained over a greater range. Specimen T-0-G-0
may be considered as a control specimen, against which the bene-
cial inuence of the inclusion of cables and prestressing can be
assessed. As discussed in Section 4.2, the presence of grout has a
relatively small inuence on tensile structural behaviour, though
a less ductile response was typically displayed by these specimens
(see Fig. 13) and the ultimate displacements were generally over-
predicted by the numerical and analytical models; as shown in
Figs. 18 and 19. This may be attributed to non-uniformity of strain
along the cable length, with high concentrations of strain at crack
locations in the grout prompting premature failure. For the non-
grouted specimens, the cable capacity was lost through the pro-
gressive fracture of individual strands followed by fracture of the
tube, whereas for the grouted specimens simultaneous fracture
of all cable strands was observed, which was followed closely by
fracture of the tube.

Table 6
Summary of measured properties of cable-in-tube test specimens.

Specimen Lo (mm) b (mm) d (mm) t (mm) ri (mm) At mm2 Ac mm2

T-1-NG-5 2000.0 60.1 60.0 3.9 2.0 876 101.7
T-1-NG-68 2000.0 60.0 60.0 3.9 2.0 875 101.7
T-1-NG-126 2000.0 59.9 60.1 3.9 2.0 875 101.7
T-2-NG-108 2000.0 60.0 60.2 3.9 2.0 877 101.7
T-0-G-0 2000.0 60.0 60.1 3.9 2.0 876
T-1-G-68 2000.0 60.0 60.0 3.9 2.0 875 101.7
T-1-G-131 2000.0 60.0 59.8 3.9 2.0 874 101.7
 J. Gosaye et al. / Engineering Structures 79 (2014) 234243 241

SG1
0.0007 120
0.0006 100

Prestress level (kN)

0.0005
80 Weld

Strain,
0.0004
60 SG4 SG2
0.0003
SG 1
SG 3 40
0.0002
SG 2
0.0001 SG 4 20
Prestress
0 0
0.00 0.02 0.04 0.06 0.08 0.10
Elapsed time (hrs) SG3
(a) Strain variations for specimen T-1-G-68 (b) Strain gauge locations (SG).
Fig. 12. Prestress level and exural behaviour of specimens during prestressing stage.

700 600
T-2-NG-108 Symbol FE: applied load
ddd definitions: Analytical:
Cable strand
T-1-G-131 fractures applied load
600 T-1-NG-68
Tube fractures 500
Test: applied load
500
400
Axial force (kN)

T-1-G-68
Analytical: force in tube
T-1-NG-126

Axial force (kN)

400
T-1-NG-5
300
FE: force in tube
300
T-0-G-0 Analytical: force in cable
200
200
FE: force in cable
100
100

0 0
0 20 40 60 80 100 120 140 160 -20 0 20 40 60 80 100
Axial displacement (mm) -100
Axial displacement (mm)
Fig. 13. Loaddisplacement response for all grouted and non-grouted test
specimens. Fig. 14. Experimental results with theoretical and numerical model comparisons
for specimen T-1-NG-5.

4. Comparison between experimental, analytical and numerical

results predicted by the analytical model is less distinct in the experimen-
tal results. This is attributed largely to the slightly rounded stress-
A detailed description of the experimental results is given in strain response of the tested cold-formed steel tubes, the presence
this section and comparisons are made with the analytical and of residual stresses, variation of yield stress between the four faces
numerical results. The measured geometrical, material and pre- of the tubes, and small possible deformations or slippage in the test
stress levels were implemented in the numerical and analytical set-up.
models. The two stage RambergOsgood model using the mea- Fig. 14 shows the results for specimen T-1-NG-5. There is a good
sured r response of the cable and tube in Table 4 was adopted correlation between the experimental, numerical and analytical
in the numerical model, whereas in the analytical model a bilinear results, but the premature fracture of one strand, in the 7-wire pre-
representation of the material was assumed. The latter used the stressing cable, results in the analytical and numerical models
measured 0.2% proof stress, r0:2 , as f y and the measured ultimate slightly overpredicting the experimental data. From the axial force
stress and strain values (ru and u ) to determine Est and Esc , which variation plotted for the tube and cable in Fig. 14, it is observed
are the linear strain hardening moduli of the tube and cable respec- that the addition of a prestressed cable increased the tensile capac-
tively. The fracture strains reported in Table 4 were used in both ity of the system by more than 50%. The experimental results for
the numerical and analytical models. In the analytical model, the specimen T-1-NG-68 are shown in Fig. 15 and are in very close
post-fracture capacity of the system is determined by the proper- agreement with the analytical and numerical results. As the gure
ties of the remaining component (cable or tube), as discussed in illustrates, the increased level of prestress in specimen T-1-NG-68
Section 2.3.1. had the effect of increasing the elastic range of the system response
which reduced the axial deection required to reach ultimate load
by 30% in comparison with that of specimen T-1-NG-5.
4.1. Non-grouted specimens The highest prestress level for the non-grouted elements was
applied to specimen T-1-NG-126, the behaviour of which is shown
The experimental loaddisplacement curves for the non-grout- in Fig. 16. The gure again demonstrates good agreement between
ed specimens, together with the corresponding analytical and experimental, numerical and analytical results, and shows that the
numerical results, are shown in Figs. 1417. In general, there is desired simultaneous yield of the tube and cable was almost
close agreement between the theoretical and experimental results, achieved in the test; compared to specimen T-1-NG-5 the deection
though the sharp change in slope at the yield point of the tube required to reach ultimate load was reduced by approximately 60%.
242 J. Gosaye et al. / Engineering Structures 79 (2014) 234243

600 600
Analytical: applied load Analytical: applied load
FE: applied load FE: applied load
500 500
Test: applied load
Test: applied load
400 FE: force in tube 400 FE: force in tube
Analytical: force in tube Analytical: force in tube

Axial force (kN)

Axial force (kN)

300 Test: force in tube 300 Test: force in tube

Analytical: force in cable FE: force in cable

Analytical: force in cable
200 200
Prestressing Prestressing
stage Test: force in cable stage
100 FE: force in cable
100 Test: force in cable

Prestressing stage Prestressing stage

0 0
-20 0 20 40 60 80 100 -20 0 20 40 60 80 100

-100 -100
Axial displacement (mm) Axial displacement (mm)

Fig. 15. Experimental results with theoretical and numerical model comparisons Fig. 18. Experimental results with theoretical and numerical model comparisons
for specimen T-1-NG-68. for specimen T-1-G-68.

600
Analytical: applied load 600
FE: applied load Test: applied load Analytical: applied load
500
Test: applied load 500 FE: applied load

400 Analytical: force in tube

FE: force in tube
Test: force in tube 400 Test: force in tube
Axial force (kN)

300
Axial force (kN)

Analytical: force in tube

300 FE: force in tube
Analytical: force in cable
200 Analytical: force in cable
FE: force in cable 200 FE: force in cable
Prestressing Prestressing
stage stage Test: force in cable
100 Test: force in cable 100

0 0
-20 0 20 40 60 80 100 -20 0 20 40 60 80 100
-100 Prestressing stage -100
Prestressing stage

-200 -200
Axial displacement (mm) Axial displacement (mm)

Fig. 16. Experimental results with theoretical and numerical model comparisons Fig. 19. Experimental results with theoretical and numerical model comparisons
for specimen T-1-NG-126. for specimen T-1-G-131.

700 load carrying capacity than the single cable specimens, are again
Analytical:
applied load well predicted by the analytical and numerical models.
600 Test: applied load FE: applied load

500
4.2. Grouted specimens
Analytical: force in tube Test: force in cable
400 Analytical: force in cable The loaddeection response of the grouted specimens are
Axial force (kN)

shown in Figs. 18 and 19. In general, the grouted specimens exhibit

300 similar behaviour to their non-grouted counterparts due to the
Test: force in tube anticipated limited contribution of the grout in tension. Clearly, a
200
Prestressing FE: force in tube greater contribution would be expected in compression. The small
stage
100 enhancement in capacity of specimen T-1-G-131 over specimen
FE: force in cable
T-1-NG-126 could be attributed to the grout providing restraint
0
against necking of the steel tube and thereby inducing a biaxial
-20 0 20 40 60 80 100
-100 tension state, in which the yield strength in the longitudinal direc-
Prestressing stage
tion is improved, as discussed in previous studies on concrete lled
-200 steel tubes [29].
Axial displacement (mm)
5. Conclusions
Fig. 17. Experimental results with theoretical and numerical model comparisons
for specimen T-2-NG-108.
The tensile performance of prestressed steel cable-in-tube sys-
tems has been examined in this paper. The stability of the system
Specimen T-2-NG-108 included two cables, thus altering the under prestress as well as the response of the elements to the sub-
tube to cable cross-sectional area ratio. The experimental load sequent tensile loading were investigated. An analytical prediction
deection results, which show a stiffer initial response and higher of the behaviour under preload was veried both numerically and
 J. Gosaye et al. / Engineering Structures 79 (2014) 234243
experimentally. For the latter, connecting collars between the
cable and the tube located at intervals along the member length
were shown to be able to transfer the restraining forces provided
by the tensioned cable to the tube and restrict overall buckling.
Hence, provided the cable is in continuous contact with the tube,
or a sufcient number of connecting collars are utilised, buckling
under preload is effectively prevented.
An analytical model to predict the tensile loaddeection
response of the prestressed elements was then established, with
the key controlling parameters being the yield loads of the tube
and cable, the pre- and post-yield stiffnesses of the tube and cable,
and the level of preload. An optimum prestress level, correspond-
ing to simultaneous yield of the tube and cable was also dened;
at this prestress level, the axial deformation corresponding to the
yielding of the system was minimised. The analytical predictions
were veried by comparison with experimental and numerical
results. The full loaddeection curves determined by means of
the three approaches (experimental, numerical and analytical)
were shown to be in close agreement, and the benets of the inclu-
sion of the cables and their prestressing, in terms of both enhanced
strength and stiffness, were demonstrated. The behaviour of pre-
stressed elements in compression as well as the inuence of
long-term effects such as creep and shrinkage, are currently being
investigated.
Acknowledgements
The authors are grateful to S2 Space Solutions for their nancial
and technical contributions to the project, and would also like to
thank the Structures Laboratory technicians at Imperial College
London.
References
[1] Ellen ME, Gosaye J, Gardner L, Wadee MA. Design and construction of long-
span post-tensioned tubular steel structures. In: Gardner L, editor. Tubular
structures XIV; 2012. p. 68793.
[2] Magnel G. Prestressed steel structures. Struct Eng 1950;28:28595.
[3] Belletti B, Gasperi A. Behaviour of prestressed steel beams. J Struct Eng ASCE
2010;136(9):11319.
243
[4] Madrazo-Aguirre F, Ruiz-Teran AM, Wadee MA. Dynamic behaviour of steel-
concrete composite under-deck cable-stayed bridges under the action of
moving loads. Eng Struct 2014 [submitted for publication].
[5] Saito D, Wadee MA. Post-buckling behaviour of prestressed steel stayed
columns. Eng Struct 2008;30:122439.
[6] Saito D, Wadee MA. Buckling behaviour of prestressed steel stayed columns
with imperfections and stress limitation. Eng Struct 2009;31:115.
[7] Saito D, Wadee MA. Numerical studies of interactive buckling in prestressed
steel stayed columns. Eng Struct 2009;31:43243.
[8] Saito D, Wadee MA. Optimal prestressing and conguration of stayed columns.
Proc Inst Civil Eng Struct Build 2010;163:34355.
[9] Wadee MA, Gardner L, Osofero AI. Design of prestressed stayed columns. J
Construct Steel Res 2013;80:8290.
[10] Ayyub BM, Ibrahim A. Post-tensioned trusses: reliability and redundancy. J
Struct Eng, ASCE 1990;116(6):150721.
[11] Han KB, Park SK. Parametric study of truss bridges by the post-tensioning
method. Can J Civil Eng 2005;32:4209.
[12] Li H, Schmidt L. Post-tensioned and shaped hypar space trusses. J Struct Eng,
ASCE 1997;123(2):1307.
[13] Schmidt LC, Li H. Studies on post-tensioned and shaped space-truss domes.
Struct Eng Mech 1998;6:693710.
[14] Hancock GJ, Key P, Olsen CJ. Structural behaviour of a stressed arch structural
system. In: Ninth international specialty conference on cold-formed steel
structures; 1988. p. 55769.
[15] Clarke MJ, Hancock GJ. Finite-element nonlinear analysis of stressed-arch
frames. J Struct Eng, ASCE 1991;117(10):281937.
[16] Clarke MJ, Hancock GJ. Simple design procedure for cold-formed tubular top
chord of stressed-arch frames. Eng Struct 1994;16(5):37785.
[17] Clarke MJ, Hancock GJ. Tests and nonlinear analysis of small-scale stressed-
arch frames. J Struct Eng, ASCE 1995;121(2):187200.
[18] Kyoungsoo L, Ziaul H, SangEul H. Analysis of stabilizing process for stress-
erection of strarch frame. Eng Struct 2013;59(2014):4967.
[19] Ellen PE. Post-tensioned steel structure; 1987.
[20] Bazant ZP. Prediction of concrete creep and shrinkage: past, present and
future. Nucl Eng Des 2001;203:2738.
[21] Gilbert RI, Ranzi G. Time-dependent behaviour of concrete structures. London;
New York: Spon Press; 2011.
[22] Magnel G. Prestressed concrete. London concrete publications limited; 1948.
[23] Belenya E. Prestressed load-bearing metal structures. Mir; 1977.
[24] Dassault Systemes Simulia Corp. ABAQUS/standard: users manual version 6.6.
Hibbitt, Karlsson & Sorensen, Inc.: Pawtucket, USA; 2006.
[25] Riks E. An incremental approach to the solution of snapping and buckling
problems. Int J Solids Struct 1979;15(7):52951.
[26] EN 10002-1. Metallic material tensile testing Part 1: Method of test at
ambient temperature. European Committee for Standardization (CEN):
Brussels; 2001.
[27] Afshan S, Rossi B, Gardner L. Strength enhancements in cold-formed structural
sections part I: Material testing. J Construct Steel Res 2013;83:17788.
[28] EN 12390-3. Testing hardened concrete Part 3: Compressive strength of test
specimens. European Committee for Standardization (CEN): Brussels; 2001.
[29] Han LH, He SH, Liao FY. Performance and calculations of concrete lled steel
tubes (CFST) under axial tension. J Construct Steel Res 2011;67:1699709.