Vibrations of An Aramid Anchor Cable Subjected To Turbulent Wind

Vibrations of an aramid anchor cable subjected to turbulent wind
Michal Tomko, Stanislav Kmet
, Robert Soltys
Institute of Structural Engineering, Faculty of Civil Engineering, Technical University of Kosice, Vysokoskolska 4, Kosice, Slovakia
a r t i c l e i n f o
Article history:
Available online xxxx
Dedicated to Professor Zdenek Bittnar on
the occasion of his 70th birthday
Keywords:
Aramid anchor cable vibration
Turbulent wind velocity simulation
Cable relaxation
Nonlinear transient dynamic nite element
analysis
Time-dependent behaviour analysis
Poincar maps
a b s t r a c t
The vibration response of an initially pre-stressed anchor cable made of parallel-lay aramid bres excited
by a measured and articially simulated spatial turbulent wind eld is presented in the paper. Results of
the analyses of in situ measured wind records are described. For selected data set statistical characteris-
tics and power spectral density functions of the measured wind velocity components are calculated. The
wind stochastic velocity uctuation is modelled as a one-variate bi-dimensional random eld. Cross-
power spectral density functions, at different point locations are introduced. The combination of the
weighted amplitude wave superposition method (WAWS) with the ShinozukaDeodatis method is used
for the analyzed problem. A time-dependent behaviour of the synthetic cable is investigated which is
subjected to turbulent wind with large expected oscillations that arise as a result of slackening due to
the relaxation effects. A nonlinear transient dynamic analysis is used in conjunction with the nite ele-
ment method to determine the dynamic response of the cable subjected to turbulent wind at its initially
prestressed state and in the selected times after the relaxation effect. The constitutive equation of the
relaxation of the aramid cable follows an experimentally obtained law of the logarithmic type. To mon-
itor the dependences of the individual quantities of cable vibration in the phase space, attractors and
Poincar maps are created by sampling the cables displacement and velocity at periods of relevant fre-
quencies. Interesting ndings based on the response of the cable with rheological properties to turbulent
wind are presented.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Cables are frequently used as fundamental structural members
in tall and large span structures as well as in guyed towers, suspen-
sion or stayed roofs and bridges. Cable structures are in general
lighter and more exible than most other forms of constructions.
They are more sensitive to turbulent wind than conventional
structures. Cables are characterized by high strength and high
resistance, but also by large amplitude oscillations which are
mainly due to turbulent wind effects [1,2].
To include nonlinearities of both structural and aerodynamic
origins, a time-domain analysis for the prediction of the maximum
dynamic response of strongly nonlinear cables and cable structures
subjected to wind loads is recommended [3]. Time-domain ap-
proaches require the generation of input wind histories of multi-
dimensional wind elds which can be simulated numerically.
Wind records have shown that wind velocity can be considered
as a stationary multi-dimensional and multi-variate random pro-
cess. Consequently, the uctuating component of the wind can
be quantied by statistical functions whose spatialtemporal prop-
erties in the frequency domain are expressed by power spectral
density functions.
Large numbers of papers which deal with such problems have
been published over the last decades and signicant progress has
been made in the simulation of wind effects and in the prediction
of the dynamic behaviour of light-weight structures. It is recog-
nized that the application of different turbulence models of wind
often provides different evaluations for the dynamic response of
the structure [4,5]. Kareem [6] summarized a historical perspec-
tive, recent developments and future challenges in the probabilis-
tic numerical simulation of wind effects. Some authors present the
results of measurements in situ or wind tunnel tests while others
investigate the problem in general terms theoretically, compare
various models and data or study specic aspects through numer-
ical experiments [7]. Lazzari et al. [8] studied the modelling and
simulation of wind velocity by considering a stationary, multivar-
iate stochastic process, according to its prescribed cross-spectral
density matrix. Naprstek [9] investigated the stochastic stability
of movement inuenced by parametric noise which is generated
by the interaction of a moving bluff body and an air ow. Hanzlik
et al. [10] presented the methodology for estimating wind effects
based on a database-assisted design approach. Fu et al. [11] devel-
oped two articial neural network approaches (a backpropagation
neural network and a fuzzy neural network) for the prediction of
mean, root-mean-square pressure coefcients and time series of
wind-induced pressures acting on a large roof structure.
0965-9978/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.advengsoft.2013.08.004
Corresponding author. Tel.: +421 55 602 2121; fax: +421 55 602.

E-mail address: stanislav.kmet@tuke.sk (S. Kmet).
Advances in Engineering Software xxx (2013) xxxxxx
Contents lists available at ScienceDirect
Advances in Engineering Software
j our nal homepage: www. el sevi er . com/ l ocat e/ advengsof t
Please cite this article in press as: Tomko M et al. Vibrations of an aramid anchor cable subjected to turbulent wind. Adv Eng Softw (2013), http://
dx.doi.org/10.1016/j.advengsoft.2013.08.004
Although considerable research related to linear and nonlinear
cable dynamics subjected to deterministic and stochastic excita-
tions has been performed in recent years, less effort has been de-
voted to the investigation of cables exposed to turbulent wind
when the cable is excited by non-periodic uctuating wind forces
with varying spatial distribution.
Most dynamic analyses of wind-excited stay or suspended
cables are focused on the investigation of basic dynamically
unstable phenomena and the modelling of aerodynamic forces
acting on a cable which may be caused by: buffeting (from turbu-
lence in the air ow), vortex shedding (from von Krmn vortexes
in the wake of the cable), galloping (self induction), wake gallop-
ing (uid-elastic interaction of neighbouring cables) and the
interaction of wind, rain and cable [1214]. Choi and Kim [15]
proposed a design procedure of aerodynamically stabilizing cables
using a FE-based buffeting analysis to stabilize the seesaw-like
motion of a free cantilevered cable-stayed bridge structure during
its construction. The most efcient positions for the hold-down
stabilizing cables were numerically investigated and the required
dimensions and pretensions of cables were also calculated. Won
et al. [16] studied a partially earth-anchored cable system in order
to reduce the dynamic wind response of cable-stayed bridges. The
employment of earth-anchored cables changes the dynamic
characteristics of cable-stayed bridges withstanding wind loads.
Liu et al. [17] compared the results of full-scale measurements
with wind tunnel test results for a long-span roof spatial struc-
ture. The time and frequency domain characteristics of full-scale
wind load measurements were analyzed based on ltered data-
acquisitions.
However, some papers were oriented toward the nonlinear
analysis of cables subjected to turbulent wind. Gattulli et al. [18]
analyzed the cable response to turbulent wind by means of a large
dimensional model by making use of the nite element method
and by applying reduced models for the wind eld and mechanical
system. The ability of reduction techniques, for loading and cable
descriptions, to accurately reproduce the dynamic response of a
suspended cable excited by an articially generated 3D turbulent
wind is discussed. Gattulli et al. [19] presented the ability of a mul-
timodal longitudinal active control to reduce non-linear cable
oscillations through analyticalnumerical models and experimen-
tal prototypes. Carassale and Solari [20] developed an analytic
expression of the proper orthogonal decomposition for a class of
processes, which include models usually adopted to represent
atmospheric turbulence. Di Paola et al. [21] used a time-domain
approach for analyzing nonlinear random vibrations of long-span
suspended cables inuenced by transversal wind. An efcient
Monte Carlo-based technique was proposed to evaluate a stochas-
tic behaviour of the cable. Carassale and Piccardo [22] investigated
the random oscillations of small-sag, small-diameter cables
induced by a turbulent wind through a reduced-order model con-
taining both mechanical and aerodynamic non-linearities. Impollo-
nia et al. [23] proposed a non-linear nite element model of
inclined cables, i.e. cables with non-levelled supports, in large dis-
placement and deformation elds in order to compute the dynamic
response to wind loads which blow in an arbitrary direction. Datta
and Jain [24] investigated the dynamic response of a modelled
articulated tower subjected rst to a random wind force and then
to the combined effect of wind, wave and current forces. Kim et al.
[25] described a geometrically nonlinear buffeting analysis of a
cable-stayed bridge in a time domain.
Vibration reduction in stay cables by means of viscous dampers
is of great interest in preventing cable damage and improving ser-
viceability of cable-stayed structural systems such as bridges,
roofs, towers and masts. Impollonia et al. [26] presented a study
on the effectiveness, as well as the limits, of rotational viscous
dampers and springs inserted at the two ends of a bending-stiff
taut cable. Zuo et al. [27] presented some characteristics of
stay-cable vibrations observed using a long-term full-scale mea-
surement system and interpreted these observations in a three-
dimensional cable-wind context. The similarities and differences
between vortex-induced vibration and rain-wind-induced vibra-
tion were investigated. Tesar and Lago [28] submitted the optimi-
zation for the number of and distance between dampers which
were adopted in high voltage conductor lines subjected to aeolic
and galloping vibrations.
Cluni et al. [29] presented a parametric investigation of the axial
fatigue performance of suspended cables subjected to lateral wind-
induced vibrations. Tesar [30] proposed the solution of a nonlinear
aeroelastic response in the ultimate range of spider-web structures
subjected to wind induced vibrations. Karmazinova et al. [31] pre-
sented some selected failure mechanisms and load-carrying capac-
ities of expansion anchors under static and dynamic tensile forces.
Oskoei and McClure [32] presented a comparison between a de-
tailed nonlinear time domain dynamic analysis and a simplied
frequency domain approach to estimate the maximum probable
response of weakly nonlinear cable roofs. This approach can be
considered as an alternative to a detailed time-domain analysis
in the preliminary design phase, or may be used to validate results
obtained from more elaborated numerical models.
Synthetic ropes are identied as an attractive material for use in
prestressed cable stayed and suspended structures. It is clear that
for these applications a good understanding of rheological proper-
ties of the structural response is highly desirable, as it relates di-
rectly to the reliability of the structural system during its
designed working life [3335].
The vibration response of an initially pre-stressed anchor cable
made of parallel-lay aramid bres excited by a measured and arti-
cially simulated spatial turbulent wind eld and the behaviour of
this cable with large expected oscillations as a result of slackening
due to relaxation effects are investigated and presented in this
paper. The proposed procedure for the nonlinear behaviour analy-
sis and response evaluation of an anchor cable subjected to wind
turbulence can be summarized by applying the following steps:
analyses of in situ measured wind records; generation of numerical
simulations of wind velocity histories; characterization of the an-
chor cable as part of a cable stayed structure; specication of the
relaxation properties of a prestressed aramid cable; denition of
the nite element model suitable for a transient dynamic analysis
subjected to turbulent wind; the assignment of simulated turbu-
lent wind histories and the determination of wind load forces;
and the evaluation of resultant vibration responses of the cable
at investigated times after relaxation effects with reference to
the initial state of the oscillation. Attractors and Poincar maps
are used to monitor the time dependences of the individual quan-
tities of the cables vibration.
2. Generation of measured and simulated wind velocity
histories
Results of the analyses of in situ measured wind records and the
generation of numerical simulations of the wind velocity histories
in the longitudinal, lateral and vertical direction are briey pre-
sented in this section. The purpose of this study is to develop a
3-D wind simulation model which is required for the transient dy-
namic analysis of the time-dependent behaviour of the cable with
rheological properties subjected to turbulent wind effects.
2.1. Measurements of wind velocities in situ
Wind speed records were measured at Bily Kriz station in
Beskydy by the Institute of Systems Biology and Ecology at the
2 M. Tomko et al. / Advances in Engineering Software xxx (2013) xxxxxx
Academy of Sciences of the Czech Republic. Since a high measure-
ment frequency for research applications was needed, wind
velocity was measured according to the propagation speed of
ultrasound signals. Records were performed by a GILL R3-100
ultrasonic anemometer, which is part of a system for measuring
the exchange of substances between the bark pocket and atmo-
sphere. A view of the mast with the attached anemometer is shown
in Fig. 1. The sampling frequency was 20 Hz and the duration of
records was 1 year.
The records comprised all three components of wind velocity
i.e. in the longitudinal, lateral and vertical direction. The device
was positioned 18 m above the height of the terrain (approxi-
mately 5 m above the pine trees) as is shown in Fig. 2.
From the annual record of wind velocities, the data from 22nd
December 2010 with the duration of 17.5 h were selected for
further statistical processing. This measurement consists of 35 re-
cords of 30 min intervals. The criterion for selecting the chosen
intervals preferably required the highest mean wind velocity for
an entire interval, i.e. approximately the same preferably highest
mean speeds with approximately coincident ow directions for a
longitudinal component of the wind during individual 30 min time
periods.
2.2. Statistical and frequency properties of the measured turbulent
wind data
Fluctuating velocity components of wind were quantied by
statistical functions. The velocity of the wind is considered to con-
sist of a mean wind velocity component and a uctuating velocity
component due to the turbulence or gusting caused by the
grounds roughness. The wind velocity components u(Z, x, t),
v(Z, y, t) and w(Z, z, t) can be expressed as the sum of the mean wind
velocity components uZ; x; vZ; y and wZ; z, and the uctuating
time-dependent velocity components u
0
(Z, x, t), v
0
(Z, y, t) and
w
0
(Z, z, t) in the form
uZ; x; t
vZ; y; t
wZ; z; t
_
_
_
uZ; x
vZ; y
wZ; z
_
_
_
u
0
Z; x; t
v
0
Z; y; t
w
0
Z; z; t
_
_
_
_
1
where x represents the longitudinal direction (the along-wind
direction), y is the lateral direction (the horizontal cross-wind direc-
tion) and z is the vertical direction (the vertical cross-wind direc-
tion) at height Z.
The mean wind velocity is dened as
u
1
T
_
T
0
utdt 2
and the standard deviation r
u
for the longitudinal direction is
r
u

r
2
u
_

1
T
_
T
0
ut u
2
dt
3
where r
2
is the variance of the wind velocity component and T is
the integration time. The longitudinal turbulence intensity is calcu-
lated as
I
u

r
u
100% 4
For the lateral and vertical direction analogous expressions can be
applied.
For further analyses, two shorter records of wind velocity com-
ponents in the longitudinal, lateral and vertical direction with a
duration of 30 min and 60 s were randomly selected from the
experimentally measured 17.5 h record of wind velocity data.
Statistical characteristics for the longitudinal, lateral and verti-
cal component of the measured wind velocities obtained from the
record with a duration of 17.5 h, 30 min and 60 s are presented in
Tables 13.
In order to investigate the frequency characteristics of the wind
velocity, the power spectral density functions (PSD) for the wind
velocity components were calculated using the FlexPro software
[36]. The discrete Fourier transformation was used to transfer the
time domain into the frequency domain representation. The power
spectral density functions were determined using the fast Fourier
transformation.
Fig. 1. View of the mast with attached anemometer.
Fig. 2. View of the anemometer located on the mast at a height Z = 18 m.
Table 1
Statistical characteristics of the measured wind velocity component in the longitu-
dinal direction {u
(1)
(Z
(1)
, x
(1)
, t)} obtained from the record with a duration of 17.5 h,
30 min and 60 s and 60 s simulation at the point P
1
.
Statistical
characteristics
Record
(17.5 h)
Record
(30 min)
Record
(60 s)
Simulation
(60 s)
u
min
(ms
1
) 3.9412 3.9412 1.7837 1.7475
u ms
1
6.7988 8.83 9.5945 9.5967
u
max
(ms
1
) 20.6931 16.8189 16.1748 15.7717
r
2
u
m
2
s
2
5.0215 5.5258 7.7686 7.1745
r
u
(ms
1
) 2.2409 2.3507 2.7872 2.6785
I
u
(%) 32.96 26.62 29.05 27.9106
M. Tomko et al. / Advances in Engineering Software xxx (2013) xxxxxx 3
For each of the 35 measured records with 30 min durations,
power spectral density functions and their mean values were cal-
culated for each discrete frequency, separately for the longitudinal,
lateral and vertical component of wind. PSD values have been
smoothed in the frequency domain using the Hanning window.
Mean values of power spectral densities were approximated by
the following non-dimensional normalized power spectral density
functions
longitudinal direction
n PSD
u
u
2
297:2 f
1 73:127 f
5=3
5
transversal direction
n PSD
v
u
2
250 f
1 65 f
5=3
6
vertical direction
n PSD
w
u
2
9:137 f
1 38:331 f
5=3
7
where f Z; n nZ= uZ is a non-dimensional frequency determined
by the frequency n, the height Z above the terrain and by the mean
wind velocity uZ; u
uZK= lnZ=z
0
r
u
K is the friction veloc-
ity, where K is the von Krmns constant and z
0
= 0.3 m is the
roughness length. For the investigated area with regular forest cov-
er the height Z above the terrain, representing the vertical distance
from ground level to the position of the anemometer, is replaced
with an effective height of Z = 5.0 m. This represents the distance
between the height of the anemometer and the top of the trees.
Normalized non-dimensional average power spectral density
functions of wind velocities in the individual longitudinal, trans-
versal and vertical direction calculated from 35 measured records
with a duration of 30 min and their approximated functions in
the forms of Eqs. (5)(7) are shown in Fig. 3.
In order to compare the frequency characteristics of the wind
velocity components the approximated power spectral density
functions calculated from expressions Eqs. (5)(7) for the
experimentally measured records were compared with those
determined according to the analytical expressions proposed by
several authors.
For illustration purposes, a comparison of the mean power
spectral density function obtained from the measurement of the
wind velocity component in the longitudinal direction with those
obtained by the analytical expressions proposed by other authors
is shown in Fig. 4. From the gure it can be seen that global courses
of individual power spectral density functions are similar and
quantitative differences can be caused e.g. by the terrains proper-
ties, orography and parameters for the site of measurement. The
corresponding analytical expressions of the mean power spectral
Table 2
Statistical characteristics of the measured wind velocity component in the lateral
direction {v
(1)
(Z
(1)
, y
(1)
, t)} obtained from the record with a duration of 17.5 h, 30 min
and 60 s and 60 s simulation at the point P
1
.
Statistical
characteristics
Record
(17.5 h)
Record
(30 min)
Record
(60 s)
Simulation
(60 s)
v
min
(ms
1
) 12.1888 10.9679 5.38 8.6147
v ms
1
0.4462 0.9125 0.0144 0.0135
v
max
(ms
1
) 9.8004 9.2338 5.8121 5.9186
r
2
v
m
2
s
2
3.719 4.2348 4.5928 6.9105
r
v
(ms
1
) 1.9285 2.0579 2.1431 2.6288
I
v
(%) 23.14 23.31 22.34 27.3927
Table 3
Statistical characteristics of the measured wind velocity component in the vertical
direction {w
(1)
(Z
(1)
, z
(1)
, t)} obtained from the record with a duration of 17.5 h, 30 min
and 60 s and 60 s simulation at the point P
1
.
Statistical
characteristics
Record
(17.5 h)
Record
(30 min)
Record
(60 s)
Simulation
(60 s)
w
min
(ms
1
) 6.7436 4.1477 2.1691 2.1353
w ms
1
1.5353 2.0497 2.1245 2.1237
w
max
(ms
1
) 9.9584 8.9152 6.5567 6.8957
r
2
w
m
2
s
2
1.1835 1.492 1.3821 2.2222
r
w
(ms
1
) 1.0879 1.2215 1.1756 1.4907
I
w
(%) 16.0 13.83 12.25 15.5369
Fig. 3. Normalized non-dimensional average power spectral density functions of
wind velocities calculated from 35 measured records with a duration of 30 min and
their approximated functions in (a) the longitudinal, (b) transversal and (c) vertical
direction.
Fig. 4. A comparison of the mean power spectral density function obtained from
the measurement of the wind velocity component in the longitudinal direction with
those obtained by analytical expressions proposed by other authors.
density functions of the wind velocity component in the longitudi-
nal direction proposed by the various authors are presented in
Table 4.
Obtained statistical and frequency properties of the longitudi-
nal, lateral and vertical component of the measured turbulent wind
can nd a wider use in the analysis and design of structures in sim-
ilar areas.
2.3. Stochastic turbulent wind modelling
The wind turbulence is usually modelled as a zero-mean,
Gaussian, stationary random eld that depends on time. If P
1
and
P
2
denote two points located at {X
1
} = {x
1
, y
1
, z
1
} and
{X
2
} = {x
2
, y
2
, z
2
}, then, from a probabilistic aspect, the complete
characterization of the atmospheric turbulence is ensured by the
knowledge of the correlation function R
u
k
u
l
fX
1
g; fX
2
g; s for every
pair of turbulence components u
k
, u
l
, where s being a time lag [44].
Assuming, that the atmospheric turbulence is ergodic, the correla-
tion function is given by the Fourier transform of the cross power
spectral density function S
u
k
u
l
fX
1
g; fX
2
g; x between turbulence
components u
k
and u
l
, where x is the circular frequency.
For simulation purposes in this study, a spatial domain is dis-
cretized into 5 points (point locations) which represent signicant
nodes of the cable (Fig. 14a). The position of nodes of the simula-
tion domain is identied by the vector
fXg fx
1
; y
1
; z
1
; x
2
; y
2
; z
2
; . . . ; x
5
; y
5
; z
5
g
fx
1
; 0; z
1
; x
2
; 0; z
2
; . . . ; x
5
; 0; z
5
g 8
The velocity at a given point can be regarded either as a one-
variate four-dimensional (1V-4D) random eld or a time-depen-
dent one-variate three-dimensional (1V-3D) random eld process.
In order to simplify the concept, the wind velocity uctuation can
be characterized in a plane. Then the time-dependent one-variate
two-dimensional (1V-2D) stochastic eld process can be exam-
ined. Thus the discrete wind velocity elds can be written as the
sum of the mean velocity vectors and the time-dependent turbu-
lence vectors in the form
fu
k
Z
k
; x
k
; tg f u
k
Z
k
; x
k
g fu
0k
Z
k
; x
k
; tg 9
fv
k
Z
k
; y
k
; tg f v
k
Z
k
; y
k
g fv
0k
Z
k
; y
k
; tg 10
where k = 1, 2, . . ., 5 denotes the position of the kth node of the cable.
The mean velocity can be expressed as
uZ; x
u
K
ln
Z
z
0
_ _
11
where K is the von Krmns constant, u
is the friction velocity, Z is

the height above the ground and z
0
is the roughness length.
The WindSimul computational software was created in the
Matlab numerical computing environment to simulate the wind
velocity components in all directions [45]. The graphical user envi-
ronment of the developed WindSimul program is shown in Fig. 5.
Wind velocity elds in the longitudinal {u
(k)
(Z
(k)
, x
(k)
, t)} and lat-
eral direction {v
(k)
(Z
(k)
, y
(k)
, t)} at the individual heights are shown
in Fig. 6.
The time-dependent random turbulent elds {u
0
(Z, x, t)} and
{v
0
(Z, y, t)} are transformed into stationary random processes
fu
0k
tg and fv
0k
tg. The complete characterization of the pro-
cess is given by its power spectral density matrix [44,46]
S
fu
0k
gfu
0l
g
x
S
fu
01
gfu
01
g
x S
fu
01
gfu
02
g
x . . . S
fu
01
gfu
05
g
x
S
fu
02
gfu
01
g
x S
fu
02
gfu
02
g
x . . . S
fu
02
gfu
05
g
x
.
.
.
.
.
.
.
.
.
.
.
.
S
fu
05
gfu
01
g
x S
fu
05
gfu
02
g
x . . . S
fu
05
gfu
05
g
x
_
_
_
_
12
The matrix (12) contains power spectral density functions and cross
power spectral density functions between turbulent velocities. For
the lateral direction an analogous expression can be applied. Cross
power spectral density functions in Eq. (12) are obtained as
S
u
0k
u
0l
S
u
0k
u
0k xS
u
0l
u
0l x
_
expf
kl
x 13
where
f
kl
x
jxj
C
2
z
z
k
z
l
2
_
2p uz
k
uz
l
14
and C
z
= 10 is the attenuation coefcient [47].
2.4. Generation and numerical simulation of wind velocity histories
Several techniques were proposed in order to simulate stochas-
tic wind velocity elds to be employed in structural analyses.
Among those, the classic weighted amplitude wave superposition
method (WAWS), based on the pioneering work of Shinozuka
and Jan [48] and then modied by Deodatis [49] (in such a way
to achieve ergodic realizations and to be efciently implemented
through fast Fourier transform algorithms), has demonstrated to
guarantee the best quality of the obtained results [50]. However,
such a procedure requires the Cholesky factorization of the spectral
matrix. This leads to high computational expenses. On the other
hand, an accurate wind simulation is fundamental for the wind-in-
duced response and behaviour prediction of exible cables and
cable structures. Less demanding, however approximate, tech-
niques may be obtained by exploiting the properties of the proper
orthogonal decomposition (POD) of the spectral matrix [51,52].
The ShinozukaDeodatis method [48,49] is used in this study
and consequently, the following expression is applied
u
0
k
t
Dx
p
5
k1
Nx
j1
T
kj
x
kj
cosx
kj
t u
kj
15
where x
kj
= x
j
+ rDx/3N, T
kj
(x
kj
) are the elements of matrix
[T(x
kj
)] and u
kj
are independent phases uniformly distributed in
the interval h0, 2pi. [T(x
kj
)] is a deterministic matrix obtained
through a frequency-dependent decomposition of the power spec-
tral density matrix. In this method the Cholesky factorization algo-
rithm is adopted which provides [T(x
kj
)] in the following lower
triangular form
Tx
kj

T
11
x 0 . . . 0
T
21
x T
22
x . . . 0
.
.
.
.
.
.
.
.
.
.
.
.
T
51
x T
52
x . . . T
55
x
_
_
_
_
16
Table 4
Analytical expressions of the mean power spectral density functions of the wind
velocity component in the longitudinal direction proposed by various authors.
Author Ref. nPSDuz; n=u
2
Hino [37]
0:475k
2
x
2
=1 x
2
2
5=6
Kaimal and Simiu [38,39]
200f =1 50f
5
3
Kareem [40]
335f =1 71f
5
3
Solari [41]
2:21b
2:5
u
f =1 3:31b
2:5
u
f
5=3
Solari [42]
6:868fL
s
=z=1 10:302fL
s
=z
5=3
Teunissen [43] 105f/(0.44 + 33f)
5/3
Note: x
2
250nz
0:42
= u; b
u
r
2
u
=u
2
is the factor of turbulence intensity and

Ls 300z=200
0:670:05lnz0
.
Power spectral density functions given by Eqs. (5)(7) were
used for the simulation of the wind velocity component in the lon-
gitudinal, lateral and vertical direction by the developed program.
Eq. (15) was used analogously for the simulation of wind velocity
components in the lateral and vertical direction.
On the basis of the random selected records from experimen-
tally measured wind velocities with durations of 60 s, numerical
simulations of the measured wind velocity component in the lon-
gitudinal, lateral and vertical direction were determined using the
created program. For the best portrayal of the experimental
Fig. 5. The graphical user environment of the developed WindSimul program.
Fig. 6. Wind velocity elds: (a) in the longitudinal {u
(k)
(Z
(k)
, x
(k)
, t)} and (b) lateral direction {v
(k)
(Z
(k)
, y
(k)
, t)} at the individual heights.
measurements by simulations the following inputs were chosen in
accordance with the measured record: the duration of the record
t = 60 s, the mean wind velocity component in the longitudinal
direction u 9:5945 ms
1
, in the lateral direction v 0:0144
ms
1
and in the vertical direction w 2:1245 ms
1
, an effective
height above the zero level Z = 5.0 m, the terrain category III
according to EN 1991-1-4 [14] with the corresponding roughness
length z
0
= 0.3 m, the frequency of the recording 20 Hz and the fre-
quency range from 0 to 10 Hz (n
max
= 10 Hz). From these input data
the number of recorded time steps s = 1200 and the number of dis-
crete frequencies N
x
= 600 were considered for the simulations.
The simulation with a duration of 60 s represents one period.
For the number of discrete frequencies N to be chosen with re-
spect to zero mean velocity of the uctuating wind component, a
random function must consequently include one period. To achieve
one period of the uctuating component, the number of discrete
frequencies is N
x
= tn
max
.
Time courses of the wind velocity components in the longitudi-
nal, lateral and vertical direction of the randomly selected mea-
sured record with a duration of 60 s and their simulations at the
point P
1
are shown in Figs. 7 and 8.
Statistical characteristics of the measured and simulated wind
velocity components in the longitudinal, lateral and vertical direc-
tion are introduced in Tables 13. The relative rate of the appear-
ance of the measured and simulated wind ow direction is
shown through a wind rosette in Fig. 9.
Relative rates and distribution functions of the occurrence of
the measured and simulated wind velocity components in the lon-
gitudinal direction at the point P
1
and the corresponding power
spectral density functions are shown in Fig. 10.
A comparison of the results obtained from the measured and
simulated data was presented. These results conrmed the
required correctness of the simulated uctuating wind velocity
histories and the possibility of their further use in dynamic analy-
ses. The physical importance of the mathematical approach and
the functionality of the developed program were demonstrated
and proved.
Longitudinal and lateral wind velocity power spectra such as
power spectral density functions at the points P
1
and P
5
, and cross
power spectral density functions between the points P
1
and P
5
,
and/or P
3
and P
5
are shown in Figs. 11 and 12.
3. Description of an investigated anchor cable
The vibration response of an anchor cable for a television tower
with a height of 200.0 m is analyzed in this study. A global axono-
metric view of the tower with the anchor cables, detail of the
spatial steel structure and a section of the structure with the deno-
tation of the analyzed anchor cable are shown in Fig. 13. The tower
is anchored by means of three cables at each of the four nodes sit-
uated in the same level. The cables are anchored to massive
unmovable concrete foundations at the other end.
Anchor cables are made of synthetic aramid bres with a high-
strength and high-modulus. The parallel-lay aramid cables have
several advantages including low weight and therefore also less
initial sag compared to traditional steel ropes and in addition,
can achieve better fatigue properties. Earlier works [33] indicated
that a stress rupture of aramid bres can be described using a
linear relationship between the tensile load and the logarithm of
Fig. 7. Time course of the experimentally measured and simulated wind velocity components in the longitudinal {u
(1)
(Z
(1)
, x
(1)
, t)} and lateral direction {v
(1)
(Z
(1)
, y
(1)
, t)} at the
point P
1
.
Fig. 8. Time course of the experimentally measured and simulated wind velocity component in the vertical direction {w
(1)
(Z
(1)
, z
(1)
, t)} at the point P
1
.
Fig. 9. Rosette of the relative rate of the occurrence of a wind ow direction
obtained from the measurement and simulation with durations of 60 s at the point
P
1
.
time to breaking point; especially when subjected to loading forces
equivalent to 50% of the nominal breaking strength, resulting in a
material lifetime of up to a 100 years.
Optimization studies and static analyses of the complex spatial
tower structure with anchor cables, under relevant load combina-
tions, using the mean wind velocity components were performed
to stipulate the appropriate initial pre-stressing forces in the
cables. The initial pre-stressing force of P(s
0
) = 285 kN of the an-
chor cable at initial time s
0
is considered and the cable cross sec-
tion is such as to operate in the subcritical range.
One of the cables of the cable stayed tower structure was se-
lected which was then further analyzed. A scheme of the investi-
gated anchor cable and wind load directions is shown in Fig. 14.
The distance between the supporting nodes of the analyzed anchor
cable is l
s
= 73.55 m. The selected diameter of the cable is
d = 80 mm, with a cross sectional area of A = 5026.55 mm
2
, having
a modulus of elasticity of E = 126500 MPa and a weight per unit
length of g = 0.04734 N mm
1
.
4. Prediction of the stress relaxation of the cable
In order to study a time-dependent dynamic behaviour of the
aramid anchor cable subjected to a turbulent wind eld due to
its rheological properties, our main attention in this section is fo-
cused on the time-dependent losses of the cables initial prestress
as a consequence of stress relaxation.
The relaxation tests of aramid-based composite cables con-
rmed [53] that the relaxation versus log(s) curve is a straight line
and therefore the following expression can be written for the stress
losses Dr(s)
Drs Alogs B 17
where Dr(s) = r(s
0
) r(s), r(s
0
) is the initial stress due to the ini-
tial prestressing force P(s
0
) at time s
0
, r(s) is the stress at time s
and A and B are constants. The relaxation behaviour of the aramid
bres follows an experimental law of the logarithmic type and its
rate follows a hyperbolic law. At the end of eight hours the losses
had already reached 3/4 of their maximum values.
The relaxation percentage p
Dr
(s) at time s can be expressed in
the form
Fig. 10. Comparisons of (a) relative rates and (b) distribution functions of the
occurrence of the measured and simulated wind velocity components in the
longitudinal direction and (c) the corresponding power spectral density functions at
the point P
1
.
Fig. 11. Longitudinal wind velocity power spectra: (a) power spectral density functions at (a) the point P
1
, (b) the point P
5
, and cross power spectral density functions
between the points (c) P
1
and P
5
, and (d) P
3
and P
5
.
p
Dr
s
Drs
rs
0
100% 18
In our case, the initial stress r(s
0
) in the anchor cable due to the ini-
tial prestressing force, self weight and stress acting under the mean
wind load is less than 10% (r(s
0
) = 56.69 MPa at the initial pre-
stressing force P(s
0
) = 285.10
3
N) of the ultimate strength of the
cable r
ult
:
2200 MPa so that the following expression for the
relaxation can be used Dr(s) = 6.6log(s) + 13.3. This expression also
includes additional elongations of the cable after installation which
can inuence the stress relaxation due to the setting of attach-
ments, deformation after initial loading by wind actuating on the
whole structure of the tower and others. The relaxation quantities
(relaxation percentage), for this illustrative example, at the end of
the two selected characteristic times s
1
= 3.8 min and
s
2
= 3801.6 min (s
2
= 2.64 day) obtained, according to Eq. (18), the
following values p
Dr
(s
1
) = (17.2/56.69)100% = 30.34% and p
Dr
(s
2
) =
(37.2/56.69)100% = 65.62%, respectively.
For the time-dependent rheological analysis of the cable
behaviour, the time domain was divided into a discrete number
of times s
0
, s
1
, s
2
, . . .. At each time, the cable was analyzed acting
under the applied turbulent wind load with the imposed current
geometry and prestress state originating from the previous
relaxation effect. The current geometry of the cable corresponding
to the actual prestress state was determined by nonlinear static
analysis. In this approach, the dynamic response of the cable with
the initial geometry and prestressing force was taken as the refer-
ence state.
Fig. 12. Lateral wind velocity power spectra: (a) power spectral density functions at (a) the point P
1
, (b) the point P
5
, and cross power spectral density functions between the
points (c) P
1
and P
5
, and (d) P
3
and P
5
.
Fig. 13. The tower with anchor cables: (a) global axonometric view, (b) detail of the spatial steel structure and (c) a section of the structure with the denotation of the
analyzed anchor cable.
In an effort to assess the inuence that the decreased initial pre-
stressing force of the cable has on its dynamic behaviour, numeri-
cal solutions were performed for the following three states: an
initial state with a cable prestressing force of P(s
0
) = 285 kN, a state
at time s
1
= 3.8 min with a prestressing force of P(s
1
) = 189 kN and
a state at time s
2
= 3801.6 min with a prestressing force of
P(s
2
) = 93.3 kN.
The initial equilibriumof the anchor cable, assumed as the static
initial conguration acting under its self-weight and the current
prestressing force at the individual states was dened by a static
nonlinear nite element analysis. Consequently, the corresponding
sags in the middle of the cable span at the individual states
s(s
0
) = 33.0 mm, s(s
1
) = 52.1 mm and s(s
2
) = 105.2 mm were ob-
tained by means of a static nonlinear analysis. All remaining cable
parameters were kept constant such as were described in Section 3.
5. Finite element model
The nite element method is applied to simulate the behaviour
of the investigated cable subjected to dynamic effects of turbulent
wind [54,55]. Finite element ANSYS software was used for its non-
linear transient dynamic analysis in the time domain [56].
5.1. Finite elements
The spatial element marked as SOLID 186 in the ANSYS 12
Classic library was used for the geometrically nonlinear analysis
of the anchor cable. SOLID186 is a higher order 3-D 20-node solid
element that exhibits a quadratic displacement behaviour. The
element is dened by 20 nodes having three degrees of freedom
per node: translations in the nodal x, y and z directions. SOLID
186 is used for the three dimensional modelling of the cable. The
element supports plasticity, hyper-elasticity, creep, stress stiffen-
ing, large deection, and large strain capabilities.
5.2. Finite element mesh generation
SOLID 186 elements were applied for the structural discretiza-
tion of the investigated cable. Convergence studies were conducted
to decide on an appropriate mesh density, with the aim of achiev-
ing suitably accurate results whilst minimizing computational
time. The model of the anchor cable consists of 210 nite elements
and 1108 nodes. Elements with a length of 350 mm were used for
the analysis.
5.3. Boundary conditions
Boundary conditions were set for the end supporting nodes of
the anchor cable. The hinged ends at the places of supports were
considered. These supporting nodes are xed in all directions
(displacements in the direction of x, y and z-axis are restrained)
but can rotate in all directions. The degrees of freedom of the
supporting nodes of the cable nite element models were dened
according to the prescribed boundary conditions.
5.4. Material properties
The results obtained from uniaxial tensile tests of aramid bres
showed that the stressstrain relationship is linear up to rupture
[53]. Consequently, the material model of the cable is character-
ized by a linear relation between stresses and strains and only
the elastic region of the cable behaviour is considered. Hence,
Hookes law is valid for its behaviour in the elastic range. Displace-
ments are supposed to be large, therefore a geometrically nonlin-
ear behaviour of the cable is assumed.
6. Nonlinear transient dynamic analysis
A transient dynamic analysis (sometimes called a time-history
analysis) is used to determine the dynamic vibration response of
an anchor cable made of aramid bres subjected to a turbulent
wind load. As a starting point, vibration characteristics such as nat-
ural frequencies and mode shapes of the anchor cable were deter-
mined by modal analysis. The rst natural frequencies of the cable
obtained at individual states of prestress at times s
0
, s
1
and s
2
are
f
1
(s
0
) = 1.71 Hz, f
1
(s
1
) = 1.41 Hz and f
1
(s
2
) = 1.02 Hz, respectively.
6.1. Description of basic equations and applied techniques
The basic equation of motion solved by a transient dynamic
analysis at time t is
Mtf wtg Ctf _ wtg K
T
tfwtg fF
appl
tg 19
where [K
T
(t)] = [K
E
(t)] + [K
G
(t)] is the tangential stiffness matrix,
[C(t)] is the damping matrix, [M(t)] is the mass matrix, {w(t)} is the
nodal displacement vector, f _ wtg is the nodal velocity vector,
f wtg is the nodal acceleration vector and {F
appl
(t)} is the applied
load vector. To solve these equations at discrete time points the
Newmark time integration method including an improved Hilber,
Hughes and Taylor algorithm called HHT [57] for controlling
numerical damping of high-frequency oscillations is applied. The
adopted procedure is described in Appendix A.
Geometric nonlinear behaviour with a large displacement range
was assumed in the analysis. Incremental and NewtonRaphson
iterative solution strategies with the updated Lagrangian formula-
tion were used to solve the geometrically non-linear problem.
Within an increment of load, the geometry and stress update were
performed once the iterative process had converged. The conver-
gence was veried with the displacements and forces criteria for-
mulated independently. Computational time can be signicantly
shortened by applying parallel computations [58].
6.2. Modelling of turbulent wind loading
The investigated anchor cable was loaded by the experimentally
measured and theoretically simulated components of turbulent
wind velocities in the longitudinal and lateral directions. The
Fig. 14. Scheme of the anchor cable: (a) geometry with a location of points P
1
P
5
and (b) directions of the wind load.
orientation of the anchor cable is nearly vertical (due to the large
inclination of the cable) and hence the vertical component of the
wind load was neglected.
The analyzed aramid anchor cable was loaded by the measured
and simulated wind velocity in the longitudinal u(t) and lateral v(t)
direction as is shown in Fig. 7. Wind velocities u(t) and v(t) were
transformed into wind load forces F
appl
u
t and F
appl
v
t (in a general
form) by means of the following relationships
F
appl
u
t
1
2
qcu
2
tA
F
appl
v
t
1
2
qcv
2
tA 20
where q is the air density, A is the reference area of the struc-
tural component and c = 0.8 is the aerodynamic coefcient. Wind
load forces were applied during the period of 60 s with a step of
0.05 s.
Fig. 15. The trajectories of the motion of the midpoint of the cable loaded by (a) the measured and (b) simulated wind velocities at the initial state (at initial time s
0
without
the inuence of relaxation) and trajectories obtained by the simulated wind velocities (c) at time s
1
= 3.8 min and (d) s
2
= 3801.6 min (states after the relaxation process).
Fig. 16. Responses of the midpoint of the anchor cable under (a) the measured and (b) simulated turbulent wind at the initial state (at time s
0
) and simulated responses (c) at
time s
1
= 3.8 min and (d) s
2
= 3801.6 min where both displacement components in the longitudinal (x-displacement) and lateral (y-displacement) direction are presented.
7. Analysis of resultant vibration responses at the investigated
states and discussion
A nonlinear transient dynamic analysis of the anchor cable sub-
jected to turbulent wind at initial time s
0
, and at the times
s
1
= 3.8 min and s
2
= 3801.6 min after the relaxation process, was
performed in order to compare the time-dependent dynamic
behaviour of the initially pre-stressed anchor cable and the behav-
iour of the cable after slackening due to the relaxation effect. The
starting conditions of the cable at the individual times differ in
value of the actual prestressing force due to relaxation and in the
current geometry dened by the sag in the mid-span of the cable.
The goal is to point out the differences in the cables behaviour and
in its response to turbulent wind at the individual times due to the
relaxation effects.
In all analyses, the vibration response of the cable during 60 s
time intervals was investigated. The cables behaviour is described
through the displacement components in the longitudinal (along-
wind) and lateral (across-wind) local reference axes that coincide
in the investigated cases with horizontal and transversal global ref-
erence axes x and y (see Fig. 14).
The trajectories of motion for the midpoint of the cable loaded
by the measured and simulated wind velocities during the time
interval of 60 s at the initial state (results of the analysis at the ini-
tial time s
0
without the inuence of relaxation) and trajectories ob-
tained by the simulated turbulent wind velocities at the times
s
1
= 3.8 min and s
2
= 3801.6 min (states after the relaxation pro-
cess) are shown in Fig. 15.
Fig. 16 shows the response of the midpoint of the anchor cable
subjected to the measured and simulated turbulent wind at the ini-
tial state (at time s
0
) and at the times s
1
= 3.8 min and
s
2
= 3801.6 min where both displacement components in the lon-
gitudinal (displacement x) and lateral (displacement y) direction
are present. In this respect a good agreement is found between
the cable responses to the measured and articially simulated tur-
bulent wind at initial time s
0
(Fig. 16a and b). The sizes of the two
components are quite different and as expected the longitudinal
displacement component is prevailing. The changes of the initial
response of the mid-point of the cable acting under the simulated
turbulent wind due to relaxation at the investigated times s
1
and
s
2
are shown in Fig. 16c and d.
The cables response to the simulated turbulent wind effects
was compared at individual states of the prestressing force. Exam-
ination of the direct comparison of the time histories (trajectories)
shown in Fig. 16 it is clearly visible that the overall aspect and
amplitudes of the cable motion are different at the individual
investigated times. Amplitudes increase over a time due to the
stress relaxation. The most signicant differences and increments
are in the displacements of the cable subjected to wind turbulence
in the longitudinal (along-wind) direction as is shown in Fig. 16.
Results conrmed that the relation between the dynamic
behaviour of the cable at the initial state and the next states is
strongly nonlinear.
Dependency between the displacement and velocity of the mid-
point of the cable in the longitudinal and lateral direction of the
measured and simulated wind turbulence at the initial state and
at the investigated times after relaxation are shown in Figs. 17
and 18. Figs. 17 and 18 demonstrate that the relaxation effect pro-
duces important nonlinear redistributions of anchor cable dis-
placements and introduces additional deformations to the
system. The maximum amplitudes of the response depend on the
orientation of cable vibration, and consequently signicant oscilla-
tions of the cable occur in the longitudinal direction. Fig. 17a and b
demonstrate that the results of the eld data analysis show consis-
tent similarities with the simulated values.
Indeed, signicant differences are observed amongst the simu-
lated cable responses to turbulent wind elds at different levels of
prestressing. To explain these differences it may be useful to eval-
uate a power spectral density function of the displacement re-
sponse associated to each state.
Power spectral density functions for the displacement of the
midpoint of the cable loaded by simulated wind velocities in the
longitudinal x direction at the initial state and at the investigated
times after relaxation are shown in Fig. 19. This gure describes
how the power of the displacement is distributed with the fre-
quency at individual prestressing forces of the anchor cable. From
the comparison shown in Fig. 19, it is clear that this power is in-
creased when the prestressing forces are decreased.
Courses of the reaction components in the x- and y-direction at
the supporting node of the cable acting under measured and sim-
ulated turbulent wind at the initial state s
0
and the changes in the
courses of the reaction components acting under the simulated
Fig. 17. Dependences of the displacement versus velocity of the midpoint of the cable in the longitudinal direction of (a) the measured and (b) simulated wind turbulence at
the initial state (at initial time s
0
without the inuence of relaxation with the prestressing force P(s
0
) = 285 kN) and simulated dependences (c) at the investigated time
s
1
= 3.8 min with the prestressing force P(s
1
) = 189 kN and (d) s
2
2
) = 93.3 kN after relaxation.
wind due to relaxation at the investigated times s
1
= 3.8 min and
s
2
= 3801.6 min are shown in Figs. 20 and 21.
Several parameters that may inuence the cables vibration
under turbulent wind include amplitude, frequency and direction
of external excitation, current properties of the structure, current
level of cable prestressing forces, etc. DTAttractor (Data Transfor-
mation to the Attractor) software has been created to monitor
the dependences of individual quantities of the cables vibration
in the multidimensional phase space and to describe a system to-
ward which the system tends to evolve over time. Attractors and
Poincar maps were created by sampling the displacement and
velocity of the cables midpoint at periods of relevant excitation
frequencies. Each representation of the Poincar map for former
vibrations is simply one xed point coherent with one-period mo-
tions as is shown in the following gures. All Poincar sections
were constructed with a phase that captured amplitudes of the
cables oscillation at the investigated states.
Poincar maps are used to analyse the numerical results. The
Poincar map is a sequence of points (displacements at the mid-
point of the cable) which characterize the dynamic response of
the cable subjected to turbulent wind actions. Poincar maps allow
one to look at the time-dependent vibration behaviour of the cable
in a straightforward manner. They show how displacements can
lead to sparse arrays at individual times due to slackening of the
cable after the relaxation process.
Fig. 18. Dependences of the displacement versus velocity of the midpoint of the cable in the lateral direction of (a) the measured and (b) simulated wind turbulence at the
initial state (at initial time s
0
without the inuence of relaxation with the prestressing force P(s
0
) = 285 kN) and simulated dependences (c) at the investigated time
s
1
1
) = 189 kN and (d) s
2
2
) = 93.3 kN after relaxation.
Fig. 19. Power spectral densities (PSD) of the midpoint displacement of the cable
loaded by the simulated wind velocities in the longitudinal x-direction at the initial
state (at initial time s
0
with a prestressing force P(s
0
) = 285 kN without the
inuence of relaxation) and at the investigated times s
1
= 3.8 min (P(s
1
) = 189 kN)
and s
2
= 3801.6 min (P(s
2
) = 93.3 kN) after relaxation.
Fig. 20. Courses of the reaction components in the x-direction at the supporting node of the cable under (a) the measured and (b) simulated turbulent wind at initial state s
0
and the changes in the courses of the reaction components under the simulated wind due to relaxation at investigated times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min.
The attractor represents a geometrical object toward which the
trajectory of a dynamical system, represented by a curve in a phase
space, converges in the course of time. The vibration response of
the cable as a dynamical system can be described geometrically
as a rotational motion on an attractor. Displacements of the cables
midpoint in the longitudinal and lateral direction at selected peri-
ods and relevant velocities are transformed to the rotational mo-
tions expressed in the polar coordinate system for the 3-
dimensional space. Mutual comparisons of the attractors at indi-
vidual times show changes in the vibration response of the cable
before and after the relaxation process. When the non-linearity
of the system becomes more important, more complicated attrac-
tors may appear. The nal trajectory in which the system settles
may have a very irregular shape, without any apparent periodicity.
Dependence between the displacement and velocity of the mid-
point of the cable in the longitudinal direction of the measured and
simulated wind turbulence is studied at the initial states using
attractors as is shown in Fig. 22a and b. The time axis is cycled
at a period T
1
(s
0
) with a frequency of 1.71 Hz (T
1
(s
0
) = 1/
1.71 = 0.58 s), which corresponds to the rst natural frequency
f
1
(s
0
), of the cable vibration at initial time s
0
and enables the
displacements of the cable under the unilateral oscillation to be
Fig. 21. Courses of the reaction components in the y-direction at the supporting node of the cable under (a) the measured and (b) simulated turbulent wind at initial state s
0
and the changes in the courses of the reaction components under the simulated wind due to relaxation at investigated times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min.
Fig. 22. Attractors of the dependence of the displacement and velocity of the midpoint of the cable in the longitudinal direction for (a) the measured and (b) simulated wind
turbulence at initial state s
0
and the simulated dependences at the times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min after relaxation (the time axes are cycled at periods T
1
(s
1
)
and T
1
(s
2
) with frequencies of 1.41 Hz and 1.02 Hz (T
1
(s
1
) = 1/1.41 = 0.71 s and T
1
(s
2
) = 1/1.02 = 0.98 s), which correspond to the rst natural frequencies f
1
(s
1
) and f
1
(s
2
) of
the cable vibration at the times s
1
= 3.8 min and s
2
= 3801.6 min after relaxation, respectively.
Fig. 23. Poincar maps of the dependence of the displacement and velocity of the midpoint of the cable in the longitudinal direction for (a) the measured and (b) simulated
wind turbulence at the initial state s
0
and the simulated dependences at the times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min after relaxation (the time axes are cycled at periods
T
1
(s
1
) and T
1
(s
2
) with frequencies of 1.41 Hz and 1.02 Hz (T
1
(s
1
) = 1/1.41 = 0.71 s and T
1
(s
2
) = 1/1.02 = 0.98 s), which correspond to the rst natural frequencies f
1
(s
1
) and
f
1
(s
2
) of the cable vibration at the times s
1
= 3.8 min and s
2
= 3801.6 min after relaxation, respectively).
Fig. 24. Attractors of the dependence of the displacement and velocity of the midpoint of the cable in the longitudinal direction for (a) the measured and (b) simulated wind
turbulence at initial state s
0
and the simulated dependences at times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min after relaxation (the time axes are cycled at half periods T
2
(s
0
),
T
2
(s
1
) and T
2
(s
2
) with frequencies of 3.42 Hz, 2.82 Hz and 2.04 Hz (T
2
(s
0
)=1/3.42 = 0.29 s = 0.5T
1
(s
0
) and analogously for the other periods T
2
(s
1
) = 0.5T
1
(s
1
) and
T
2
(s
2
) = 0.5T
1
(s
2
)), which correspond to the rst natural double frequencies f
1
(s
0
), f
1
(s
1
) and f
1
(s
2
) of the cable vibration at the initial state (at time s
0
) and at the times
s
1
= 3.8 min and s
2
captured. Intersections of the Poincar section with the trajectory
in the attractors are marked by points at the areas with the ex-
pected amplitudes of displacements. These corresponding Poincar
sections are shown separately in Fig. 23a and b.
Attractors for the dependence of the displacement and velocity
of the midpoint of the cable in the longitudinal direction of the
simulated wind turbulence at the investigated times after relaxa-
tion are shown in Fig. 22c and d, and the corresponding Poincar
sections at the same times are shown in Fig. 23c and d.
Attractors for the rst natural double frequency of the cables
vibration of dependence between the displacement and velocity
of the midpoint of the cable in the longitudinal direction of the
experimentally measured and theoretically simulated wind turbu-
lence at the initial state (results of the analysis at initial time s
0
without the inuence of relaxation) and at the investigated times
s
1
= 3.8 min and s
2
= 3801.6 min after the relaxation are shown
in Fig. 24. The corresponding Poincar sections at the same times
are shown in Fig. 25. The time axes are cycled at half periods. Thus,
the selected periods allow the bilateral displacements at the mid-
point of the anchor cable to be captured. A colour spectrum for the
differentiation of the time dependence of monitored variables in
the Poincar sections is used (see the time axis on the right side
of the gures). Consequently, the blue points correspond to the
beginning of the time interval with a duration of 60 s. Figs. 23d
and 25d show that a lower density of points is seen at the times
after the relaxation process, which corresponds to larger displace-
ments at these states. Vibrations in a variety of modes and high-le-
vel dynamic responses of the cable to random excitations can
occur. The overall characteristics of damping are different at the
individual times, and are inuenced by the current properties of
the cable. From a dynamic perspective, the properties of the cables
are modied by the initial value of prestress.
Poincar maps at the individually investigated times show that
the displacements of the cable due to turbulent wind are more
dispersed after relaxation when the initial prestressing force is suc-
cessively decreased in comparison with those at the initial state. In
this state when the cable is prestressed by the required prestress-
ing force, at the end of the monitored time section, the displace-
ments begin to become more clustered in comparison with those
at the states after relaxation. This clustering can be attributed to
the consolidation of the oscillation in certain limits due to a suf-
cient level of prestressing. In further investigations, it would be
useful to monitor vibration during a longer time interval.
Results of the time-dependent response analysis of the anchor
cable loaded by the experimentally measured and simulated tur-
bulent wind velocities at the initially prestressed state and at the
states after slackening due to the relaxation effects were presented.
The inuence of the level of actual prestressing force on the cable
response was evaluated. The performed comparisons between
cable vibration responses at individual states conrmed that the
level of cable prestress has a signicant effect on its excited modes
namely from both quantitative and qualitative aspects.
8. Conclusions
The vibration response of the initially pre-stressed anchor cable
made of parallel-lay aramid bres excited by a measured and arti-
cially simulated spatial turbulent wind eld and the behaviour of
this cable with large expected oscillations after slackening due to
the relaxation effects were investigated and presented in the
paper.
Synthetic aramid ropes have been identied as an attractive
material for use in prestressed cable stayed and suspended struc-
tures. It is clear that for these applications a good understanding
of the rheological properties of the structural response is highly
desirable, as it relates directly to the reliability of the structural
system during its design working life.
Fig. 25. Poincar maps of the dependence of the displacement and velocity of the midpoint of the cable in the longitudinal direction for (a) the measured and (b) simulated
wind turbulence at initial state s
0
and the simulated dependences at times (c) s
1
= 3.8 min and (d) s
2
= 3801.6 min after relaxation (the time axes are cycled at half periods
T
2
(s
0
), T
2
(s
1
) and T
2
(s
2
) with frequencies of 3.42 Hz, 2.82 Hz and 2.04 Hz (T
2
(s
0
) = 1/3.42 = 0.29 s = 0.5T
1
(s
0
) and analogously for the other periods T
2
(s
1
) = 0.5T
1
(s
1
) and
T
2
(s
2
) = 0.5T
1
(s
2
)), which correspond to the rst natural double frequencies f
1
(s
0
), f
1
(s
1
) and f
1
(s
2
) of the cable vibration at the initial state (at time s
0
) and at the times
s
1
= 3.8 min and s
2
Vibrations of aramid anchor cables, their stresses and deforma-
tions can be signicantly inuenced by time-dependent processes
such as creep and relaxation. Since the stresses in the anchor cables
in service conditions are limited to approximately 45% of their ulti-
mate tensile strengths (UTS) the effect of the relaxation phenome-
non can be reduced. The considerable relaxation effects occur for
stresses higher than 50% of UTS. In these cases, the dynamic re-
sponse along the lifespan of the structure should be evaluated by
taking time dependent effects into account. However, time effects
always introduce a signicant redistribution of forces from the an-
chor cables to the main structure (tower).
The relaxation effect produces important nonlinear redistribu-
tions of anchor cable forces and displacements as well as introduc-
ing additional deformations to the main system. The nonlinear
dynamic analysis of the aramid cable evidenced, in particular, the
sensitivity of the structure to relaxation effects.
If proper attention is not devoted to these effects, structural
reliability in terms of serviceability and, in some instances, of ulti-
mate safety may be adversely affected. An appropriate evaluation
of such effects for designing durable and safe structures requires
the establishment of reliable methods for predicting stress relax-
ations of cables (a material properties problem), and for determin-
ing the consequent time-dependent response of cable structures
with an adequate degree of accuracy (a structural analysis prob-
lem). The proposed approach can be used for these purposes since
it provides time-dependent control of the dynamic behaviour of
the polymer cable structures. The procedure provides the designer
with effective ways of guiding the evolution of the dynamic system
by an appropriate and rational choice of its initial parameters.
Further investigations, aimed at a deeper comprehension of the
nonlinear dynamic response of wind-excited aramid cables, should
consider a uidsolid coupled simulation of the cable interaction
with an environment under the turbulent wind eld and represent
a future task for the authors.
Acknowledgements
This work is part of Research Projects No. 1/0321/12 and No.
APVV-0179-10, partially founded by the Scientic Grant Agency
and The Slovak Research and Development Agency of the Ministry
of Education, Science, Research and Sport of Slovak Republic and
the Slovak Academy of Sciences. The present research has been car-
ried out within the project Centre of excellent integrated research
for progressive building structures, materials and technologies,
supported by European Union Structural funds. Special thanks be-
longs to the staff of the Institute of Systems Biology and Ecology
from Academy of Sciences of the Czech Republic for providing
the measured wind records.
Appendix A. The Newmark method
The Newmark method uses nite difference expansions in the
time interval Dt = t
n+1
t
n
, in which it is assumed that
f _ w
n1
g f _ w
n
g 1 df w
n
g df w
n1
gDt A:1
fw
n1
g fw
n
g Dtf _ w
n
g
1
2
a
_ _
f w
n
g af w
n1
g
_ _
Dt
2
A:2
where a = 0.25(1 + c)
2
and d = 0.5 + c are the Newmark integration
parameters and c is the amplitude decay factor. The Newmark
method becomes the constant average acceleration method when
c = 0 which consequently in turns means a = 0.25 and d = 0.5. Re-
sults from the constant acceleration method do not show any
numerical damping in terms of displacement amplitude errors. If
other sources of damping are not present (an anchor cable without
damping devices) the lack of numerical damping can be undesirable
in that the higher frequencies of the cable structure can produce
unacceptable levels of numerical noise. A certain level of numerical
damping is usually desirable and is achieved by degrading the New-
mark approximation by setting c > 0. In order to obtain a stable
solution, the length of the time step Dt = 0.01 s was selected.
In particular, it is desirable to have controllable numerical
damping in the higher frequency modes, since using nite
elements to discretize the spatial domain results of these higher
frequency modes are less accurate. However, the addition of high
frequency numerical damping should not incur a loss of accuracy
nor introduce excessive numerical damping in important low
frequency modes. In the full transient analysis, the HHT time inte-
gration method has the desired property for the numerical damp-
ing. The basic form of the HHT method is given by
Mf w
n1am
g Cf _ w
n1a
f
g K
T
fw
n1a
f
g fF
appl
n1a
f
g A:3
where a
m
and a
f
are the two extra integration parameters for the
interpolation of the acceleration and the displacement, velocity
and loads.
Introducing the Newmark assumption as given in Eqs. (A.1) and
(A.2) into Eq. (A.3), the displacement {w
n+1
} at time step n + 1 can
be obtained as
a
0
M a
1
C 1 a
f
K
T
fw
n1
g
1 a
f
fF
appl
n1
g a
f
fF
appl
n
g a
f
fF
int
n
g Ma
0
fw
n
g
a
2
f _ w
n
g a
3
f w
n
g Ca
1
fw
n
g a
4
f _ w
n
g a
5
f w
n
g A:4
where a
0
, a
1
, a
2
, a
3
, a
4
and a
5
are the coefcients that depend on the
four parameters a, d, a
f
and a
m
used in the HHT method. These
parameters are related to the input as follows: a = 0.25(1 + c)
2
,
d = 0.5 + c, a
f
= c and a
m
= 0. Numerical damping stabilizes the
numerical integration scheme by damping out the unwanted high
frequency modes. For the Newmark method, numerical damping
also affects the lower modes and reduces the accuracy of the
integration scheme from second order to rst order. For the HHT
method, numerical damping affects only the higher modes and
always maintains second-order accuracy.
For the nonlinear vibration analysis of cables, it is often difcult
to accurately determine the appropriate value of numerical
damping. With very small damping values, numerical instability
appeared in the computation and the dominant frequency of
oscillation of the anchor cable was 13.3 Hz.
With regard to the nature of the load and the natural frequency
of the cable, it was assumed that the dominant frequency of the
vibration response will be the rst natural frequencies of the cable.
For this reason, the numerical value of damping in the analysis was
set to c = 0.025. This value ensured the reduction of vibrations in
the unstable area of the cable. The dominant frequency became
the rst natural frequency and the higher frequencies appeared
in a negligible measure.
References
[1] Irvine HM. Cable structures. Cambridge (MA): The MIT Press; 1981.
[2] Buchholdt HA. Introduction to cable roof structures. 2nd
ed. Cambridge: Cambridge University Press; 1998.
[3] Chen X, Matsumoto M, Kareem A. Time domain utter and buffeting response
analysis of bridges. J Eng Mech ASCE 2000;126:716.
[4] Solari G, Piccardo G. Probabilistic 3-D turbulence modelling for gust buffeting
of structures. Probabilist Eng Mech 2001;16:7386.
[5] Holmes DJ. Wind loading of structures. London, New York: Spon Press; 2001.
[6] Kareem A. Numerical simulation of wind effects: a probabilistic perspective. J
Wind Eng Ind Aerodynam 2008;96:147297.
[7] Li QS, Zhi LH, Hu F. Field monitoring of boundary layer wind characteristics in
urban area. Wind Struct Int J 2009;12:55374.
[8] Lazzari M, Saetta AV, Vitaliani RV. Non-linear dynamic analysis of cable-
suspended structures subjected to wind actions. Comput Struct
2001;79:95369.
[9] Naprstek J. Stability domains of wind-excited random nonlinear systems
through Lyapunov function. J Wind Eng Ind Aerodynam 2001;89:1499512.
[10] Hanzlik P, Diniz S, Grazini A, Grigoriu M, Simiu E. Building orientation and
wind effects estimation. J Eng Mech ASCE 2005;131:2548.
[11] Fu JY, Liang SG, Li QS. Prediction of wind-induced pressures on a large
gymnasium roof using articial neural networks. Comput Struct
2007;85:17992.
[12] Kolousek V, Pirner M, Fischer O, Naprstek J. Wind effects on civil engineering
structures. Amsterdam: Elsevier; 1984.
[13] EN 1993-1-11. Design of steel structures. Part 1.11: design of structures with
tension components. CEN; 2006.
[14] EN 1991-1-4. Actions on structures. Part 1.4: general actions wind actions.
CEN; 2005.
[15] Choi SW, Kim HK. Design of aerodynamic stabilizing cables for a cable-stayed
bridge during construction. Wind Struct Int J 2008;11:391411.
[16] Won JH, Yoon JH, Park SJ, Kim SH. Effects of partially earth-anchored cable
system on dynamic wind response of cable-stayed bridges. Wind Struct Int J
2008;11:44153.
[17] Liu H, Qu WL, Li QS. Comparison between wind load by wind tunnel test and
in-site measurement of long-span spatial structure. Wind Struct Int J
2011;14:30119.
[18] Gattulli V, Martinelli L, Perotti F, Vestroni F. Dynamics of suspended cables
under turbulence loading: reduced models of wind eld and mechanical
system. J Wind Eng Ind Aerodynam 2007;95:183207.
[19] Gattulli V, Rocco A, Potenza F. Analytical prediction and experimental
validation for longitudinal control of cable oscillations. Int J Nonlinear Mech
2008;43:3652.
[20] Carassale L, Solari G. Wind modes for structural dynamics: a continuous
approach. Probabilist Eng Mech 2002;17:15766.
[21] Di Paola M, Muscolino G, So A. Monte Carlo simulation for the response
analysis of long-span suspended cables under wind loads. Wind Struct Int J
2004;7:10730.
[22] Carassale L, Piccardo G. Non-linear discrete models for the stochastic analysis
of cables in turbulent wind. Int J Nonlinear Mech 2010;45:21931.
[23] Impollonia N, Ricciardi G, Saitta F. Vibrations of inclined cables under skew
wind. Int J Nonlinear Mech 2011;46:90718.
[24] Datta TK, Jain AK. Response of articulated tower platforms to random wind and
wave forces. Comput Struct 1990;34:13744.
[25] Kim HK, Shinozuka M, Chang SP. Geometrically nonlinear buffeting response
of a cable-stayed bridge. J Eng Mech ASCE 2004;130:84857.
[26] Impollonia N, Ricciardi G, Saitta F. Dynamic behavior of stay cables with
rotational dampers. J Eng Mech ASCE 2010;136:697709.
[27] Zuo D, Jones NP, Main JA. Field observation of vortex- and rain-wind-induced
stay-cable vibrations in a three-dimensional environment. J Wind Eng Ind
Aerodynam 2008;96:112433.
[28] Tesar A, Lago J. Aeolic and galloping vibrations of overhead line conductors.
Build Res J 2009;57:16983.
[29] Cluni F, Gusella V, Ubertini F. A parametric investigation of wind-induced
cable fatigue. Eng Struct 2007;29:3094105.
[30] Tesar A. Aeroelastic response of spider-web structures. Build Res J
2009;57:7792.
[31] Karmazinova M, Melcher J, Kala Z. Design of expansion anchors to concrete
based on results of experimental verication. Adv Steel Constr
2009;5:390405.
[32] Oskoei SAG, McClure G. Dynamic analysis of cable roofs under transient wind:
a comparison between time domain and frequency domain approaches.
Tsighua Sci Technol 2008;13:537.
[33] Guimares GB, Burgoyne CJ. Creep behaviour of a parallel-lay aramid rope. J
Mater Sci 1992;27:247389.
[34] Kmet S. Non-linear rheology of tension structural element under single and
variable loading history. Part I: theoretical derivations. Struct Eng Mech
2004;18:56589.
[35] Kmet S, Holickova L. Non-linear rheology of tension structural element under
single and variable loading history. Part II: creep of steel rope examples and
parametrical study. Struct Eng Mech 2004;18:591607.
[36] FlexPro. Documentation. Version 7.0.24, Weisang GmbH, St. Ingbert, Germany;
2005.
[37] Hino M. Spectrum of gusty wind. In: Proceedings of the 3rd international
conference on wind effects on buildings and structures, Tokyo; 1971. p. 6977.
[38] Kaimal JC, Wyngaard JC, Izumi Y, Cote OR. Spectral characteristics of surface-
layer turbulence. Quart J Roy Meteorol Soc 1972;98:56389.
[39] Simiu E. Wind spectra and dynamic along-wind response. J Struct Div ASCE
1974;100:1897910.
[40] Kareem A. Wind induced response analysis of tension leg platforms. J Struct
Eng ASCE 1985;111:3755.
[41] Solari G. Turbulence modeling for gust loading. J Struct Eng ASCE
1987;113:155069.
[42] Solari G. Gust buffeting I: peak wind velocity and equivalent pressure. J Struct
Eng ASCE 1993;119:36582.
[43] Teunissen HW. Structure of mean winds and turbulence in the planetary
boundary layer over rural terrain. Bound Layer Meteorol 1980;19:187221.
[44] Ubertini F, Giuliano F. Computer simulation of stochastic wind velocity elds
for structural response analysis: comparisons and applications. Adv Civil Eng
2010:120. http://dx.doi.org/10.1155/2010/74957.
[45] MATLAB. Documentation. Optimization toolbox users guide. Version 3.1. The
MathWorks, Inc., Natick, MA, USA; 2006.
[46] Di Paola M. Digital simulation of wind eld velocity. J Wind Eng Ind
Aerodynam 1998;7476:91109.
[47] Li J, Chen J. Stochastic dynamics of structures. Singapore: J. Wiley & Sons;
2009.
[48] Shinozuka M, Jan CM. Digital simulation of random processes and its
applications. J Sound Vib 1972;25:11128.
[49] Deodatis G. Simulation of ergodic multivariate stochastic processes. J Eng
Mech ASCE 1996;122:77887.
[50] Rossi R, Lazzari M, Vitaliani R. Wind eld simulation for structural engineering
purposes. Int J Numer Meth Eng 2004;61:73863.
[51] Love M. Probability theory. Princeton: D. Van Nostrand Company; 1955.
[52] Papoulis A. Probability, random variables and stochastic processes. New
York: McGraw Hill; 1965.
[53] Kaci S. Experimental study of mechanical behaviour of composite cables for
prestress. J Eng Mech ASCE 1995;121:70916.
[54] Bathe KJ. Finite element procedures. Englewood Cliffs: Prentice-Hall; 1996.
[55] Bittnar Z, Sejnoha J. Numerical methods in structural mechanics. New York,
London: ASCE Press and Thomas Telford; 1996.
[56] ANSYS, Inc. Release documentation of Ansys 11.0. Southpointe, 275
Technology Drive, Canonsburg; 2005.
[57] Hilber HM, Hughes TJR, Taylor RL. Improved numerical dissipation for time
integration algorithm in structural dynamics. Earthq Eng Struct Dynam
1977;5:28292.
[58] Bittnar Z, Kruis J, Nemecek J, Patzak B, Rypl D. Parallel and distributed
computations for structural mechanics: a review. In: Topping BHV, editor. Civil
and structural engineering computing. Edinburg: Saxe-Coburg Publications;
2001. p. 21133.

Vibrations of An Aramid Anchor Cable Subjected To Turbulent Wind

Uploaded by

Document Information

Original Title

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

Vibrations of An Aramid Anchor Cable Subjected To Turbulent Wind

Uploaded by

Copyright:

Available Formats

Vibrations of an aramid anchor cable subjected to turbulent wind

Michal Tomko, Stanislav Kmet

Corresponding author. Tel.: +421 55 602 2121; fax: +421 55 602.

is the friction velocity, Z is

is the factor of turbulence intensity and

You might also like