Nonlinear Vibrations of Suspended Cables-Part I: Modeling and Analysis

Nonlinear vibrations of suspended cables—Part I:
Modeling and analysis

Giuseppe Rega
Dipartimento di Ingegneria Strutturale e Geotecnica, Università di Roma ‘‘La Sapienza,’’
Rome, Italy; [email protected]
This review article is the first of three parts of a Special Issue dealing with finite-amplitude
oscillations of elastic suspended cables. This part is concerned with system modeling and
methods of analysis. After shortly reporting on cable historical literature and identifying the
topic and scope of the review, the article begins with a presentation of the mechanical system
and of the ensuing mathematical models. Continuum equations of cable finite motion are for-
mulated, their linearized version is reported, and nonlinear discretized models for the analysis
of 2D or 3D vibration problems are discussed. Approximate methods for asymptotic analysis
of either single or multi-degree-of-freedom models of small-sag cables are addressed, as well
as asymptotic models operating directly on the original partial differential equations. Numeri-
cal tools and geometrical techniques from dynamical systems theory are illustrated with refer-
ence to the single-degree-of-freedom model of cable, reporting on measures for diagnosis of
nonlinear and chaotic response, as well as on techniques for local and global bifurcation
analysis. The paper ends with a discussion on the main features and problems encountered in
nonlinear experimental analysis of vibrating suspended cables. This review article cites 226
references. 关DOI: 10.1115/1.1777224兴
1 INTRODUCTION just after the failure of the Tacoma Narrows Bridge. The
Suspended cables are lighweight, flexible structural elements topic was dealt with, among other papers, in the articles of
used in numerous applications in mechanical, civil, electri- Pugsley 关8兴 and of Saxon and Cahn 关9兴, whose solutions,
cal, ocean, and space engineering due to their capability of however, still failed to reproduce the spectrum for the clas-
transmitting forces, carrying payloads, and conducting sig- sical taut string in the limit of vanishing cable sag. After-
nals across large distances. At the same time, the suspended wards, the sagged cable models developed by Simpson 关10兴
and Soler 关11兴 were capable of a correct transition to the taut
cable is a basic element of theoretical interest in applied
string, whose nonlinear vibrations were first considered by
mechanics and an archetypal model of various phenomena in
Carrier 关12兴. Nevertheless, the whole matter was clarified
dynamics.
just in the fundamental contribution of Irvine and Caughey
Cable dynamics has a long and rich history documented
关13兴—together with the basic crossover phenomenon of the
in the classic monograph by Irvine 关1兴, and summarized in
in-plane frequencies—by introducing the effect of cable elas-
the review articles by Triantafyllou 关2– 4兴 and Starossek 关5兴. ticity 共see also 关14兴兲.
The early work on cable dynamics goes back as early as the Actually, the great diversity of cable applications has led
eighteenth century, when d’Alembert, Euler, Bernoulli, and to different elastic cable theories. They refer to either small-
Lagrange 共see Routh 关6兴兲 made considerable effort in under- sag 共parabolic兲 or large-sag 共catenary兲 cables, used as over-
standing the vibrations of taut strings and of a sagged cable head transmission lines, mooring lines, hanging systems, and
with concentrated masses, as well as in the development of cableways, as well as in connection with other structural el-
the theory of partial differential equations. In the nineteenth ements in various civil engineering applications 共eg, suspen-
century, equations of motion of a cable element were derived sion bridges, cable-supported structures, etc兲. Analytical so-
by Poisson 共1820, see Routh 关6兴兲, and were then used by lutions are available for the natural frequencies and mode
Rohrs 关7兴 and Routh 关6兴 to obtain, respectively, approximate shapes of small-sag cables, wherein continuum formulations
and closed form solutions for the natural frequencies of small are adopted and the analysis is greatly simplified by the para-
oscillations of an inextensible sagged cable with distributed bolic approximation. In contrast, discrete formulations and
mass 共bare cable兲. After a rather long period in which the mostly numerical methods 共finite differences, finite elements,
topic was set aside, there was a renewal of interest towards lumped parameter techniques, multibody dynamics兲 are used
cable dynamics around the middle of the twentieth century, to deal with large-sag cables. Former numerically oriented
Transmitted by Associate Editor P Pfeiffer
Appl Mech Rev vol 57, no 6, November 2004 443 © 2004 American Society of Mechanical Engineers
444 Giuseppe Rega : Nonlinear vibrations of suspended cables—Part I Appl Mech Rev vol 57, no 6, November 2004
studies on cable nonlinear dynamics include the papers by tant part of coupled structural systems—are just mentioned,
Leonard and Recker 关15兴, West et al 关16兴, Henghold et al as well as purely discrete models and the numerical methods
关17兴, Gambhir and Batchelor 关18兴, Rosenthal 关19兴, and Fried commonly used for their dynamic analysis.
关20兴, as well as the review papers by Migliore and Webster Besides the general distinction between deterministic or
关21,22兴. More recently, finite differences have been used in stochastic features of the dynamic problem as regards both
Tjavaras et al 关23兴, Ni et al 关24兴, Chucheepsakul and the system properties and the excitation characteristics, mod-
Wongsa 关25兴, Srinil et al 关26兴, Wu et al 关27兴, and Koh and els for cable nonlinear dynamics can be classified and dis-
Rong 关28兴, whereas finite element cable modeling has been cussed with respect to different points of view:
addressed in many papers 共eg, Desai et al 关29,30兴, Zhang and i兲 The nature of the system 共bare cable/cable with
Peil 关31兴, Karoumi 关32兴, Sofi et al 关33兴, and Ni et al 关34兴兲, lumped masses兲, its geometrical configuration
and it has been used in the investigation of various aspects of 共horizontal/inclined兲, the system material 共elastic/
nonlinear dynamic behavior 共Martinelli and Perotti 关35兴, inelastic兲 and element 共axial rigidity/further rigidities兲
Gattulli et al 关36兴, and Zheng et al 关37兴兲. properties, and the low-tension/high-tension dynamic
Although often formulating the cable dynamic problem in behavior
some nonlinear regime, many of the above mentioned works ii兲 The infinite- or finite-dimensional set of continuous/
were actually concerned with the linear theory of small os- discretized/mixed equations of motion formulated to
cillations of suspended cables, a field to be considered by analyze the system dynamics
now well established, at least as regards small-sag cables. In iii兲 The number of degrees-of-freedom 共dof兲 taken into ac-
recent years, a lot of papers have actually been published on count in a finite-dimensional single-mode/multimode
both linear and nonlinear dynamic behavior of cables for model, as well as the features of free/forced, planar/
many varied applications: see the proceedings of a series of nonplanar system response to be studied with it
Cable Dynamics Symposia 共1995,1999,2001,2003兲关38兴. In iv兲 The nonresonant/resonant dynamic features of the sys-
the present article, however, the interest is in the great deal of tem 共single/multiple, internal/combination resonance兲
theoretical research work of the last two decades on the finite and of the excitation 共external/parametric/combination
amplitude oscillations of the basic suspended cable, a rich resonance兲
topic with many different, and somewhat dispersed, aspects.
In turn, methods of analysis can be distinguished with
In this respect, we borrow just from Irvine 关39兴 a statement
respect to:
of superiority of the analytical approach with respect to nu-
merical methods for being of ‘‘most help in determining the i兲 The kind of closed form/‘‘exact’’/approximate solution
physical characteristics of the problem,’’ and extend it to the to be searched for through many possible analytical,
nonlinear field. Accordingly, it is felt that in-depth under- numerical, and geometrical techniques, as well as the
standing of the fundamental features of cable nonlinear dy- features and the objectives of companion experimental
namics and proper frameworking of its rich and complex analyses
behavior require even more so using classical and modern ii兲 The weak/strong degree and the variable importance of
theoretical methods, though combined with selected numeri- geometric nonlinearities due to the elastic, inertial, and
cal tools, rather than generic computer simulations on large damping effects, as well as of possibly gyroscopic and
discrete models. Within this context and notwithstanding the material nonlinearities
apparent simplicity of the sagged cable, the relevant dynamic iii兲 The interest in performing nonlinear dynamic analyses
phenomena induced by the even and odd nonlinearities gov- in the regular/nonregular regime and in getting infor-
erning its finite amplitude equations of motion are of consid- mation on the local/global system response
erable interest and variety in themselves 关40兴. Moreover, Finally, discussion on nonlinear phenomena exhibited by
they allow us to consider the cable as a meaningful arche- suspended cables has to be concerned with:
typal continuous system to summarize and discuss general
methods and specific techniques for nonlinear dynamic i兲 Basic features of monofrequent regular dynamics in
analysis. This is thus the right time for a further, updated various resonance conditions
review article on cable dynamics, now entirely devoted to ii兲 Multimodal interaction characterizing coupled regular
system nonlinear behavior. response
Putting the many aspects of cable nonlinear dynamics iii兲 Main classes of steady regular and nonregular response
within a systematic and consistent framework is not an easy occurring in meaningful control parameter space
task, owing to the difficulty in selecting any classifying cri- iv兲 Features of response instability and of the ensuing local
terion capable of dealing with all the available knowledge bifurcations
and with the research development perspectives. So, we re- v兲 Features of system quasi-periodic and chaotic response,
strict the area of interest, and choose to report in-depth on the as well as of the associated global bifurcation mecha-
nisms and transition scenarios
most important achievements obtained in recent years on the
nonlinear dynamics of relatively small-sag cables, by focus- All of the previous topics can be addressed either with
ing on the relevant models, methods of analysis, and phe- reference to a suspended cable in air, or in accounting for the
nomena. Large-sag cables—which are also often an impor- dynamic interaction phenomena which occur between the
Appl Mech Rev vol 57, no 6, November 2004 Giuseppe Rega : Nonlinear vibrations of suspended cables—Part I 445
cable and a surrounding fluid. Further interesting topics are 2 MECHANICAL SYSTEM AND MATHEMATICAL
concerned with control of cable nonlinear vibrations. MODELING
Based on the above points, this review is organized in
2.1 Problem formulation and continuum equations of
three parts. The first two parts are devoted to the analysis of
3D finite motion
cable nonlinear dynamics under deterministic conditions. In
particular, Part I starts with a presentation of the system and A heavy cable suspended between two supports at the same
of the ensuing mathematical models 共Section 2兲. Then, it level a distance l apart is modeled as a 1D linearly elastic
aims at presenting a fairly comprehensive framework of ap- continuum with no flexural, torsional, or shear rigidities in
proximate methods used for asymptotic analysis of nonlinear the context of the dynamics of initially stressed solids out-
response 共Section 3兲, as well as of numerical and geometrical lined in 关43兴 and first applied to cable structures in Rega and
tools and techniques available for its local and global analy- Luongo 关14兴, and Luongo et al 关44兴. This needs referring to
sis 共Section 4兲. Features and problems encountered in non- three different cable configurations 共Fig. 1兲, namely the natu-
linear experimental analysis of vibrating suspended cables ral 共unstressed兲 configuration C 0 , the initial deformed con-
figuration C I of static equilibrium in the vertical plane under
are discussed in Section 5.
dead load, and the varied 共dynamic兲 configuration C V occu-
Part II 共Rega 关41兴兲 will also be concerned with determin-
pied by the cable under the action of external loads and/or
istic dynamics, and will give a thorough overview on, and a
support motions. The characteristics of the cable in C 0 are l 0c
tentative classification of, the rich and varied nonlinear phe-
length, m 0 mass per unit length, A 0 cross-sectional area, and
nomena which characterize cable regular and nonregular re-
E 0 material elastic modulus. In a finite strain model, the
sponse. Forward reference to section and figure numbers in
initial 共static兲 deformation carries the material point from the
Part II is sometimes made in this paper: the relevant labels
location P 0 (s 0 ) in C 0 to the location P I (s I ) in C I , the addi-
are denoted by a roman II symbol to distinguish them from tional 共dynamic兲 deformation carries it to its final location
section and figure numbers of the present Part I. Finally, P V (s V ) in C V , s 0 , s I , and s V being the relevant arc-length
system nonlinear dynamics in random conditions will be the coordinates measured along each configuration. By assuming
subject of Part III 共Ibrahim 关42兴兲, where specific methods of C I as the reference configuration for the cable motion, the
analysis will be discussed and the most important features of dynamic displacement vector field is expressed as u(s I ,t)
string and cable stochastic dynamics will be presented, to- ⫽RV (s I ,t)⫺RI (s I ), the Rs being position vectors of the ma-
gether with the interaction with fluid flow. terial point in C I and C V .
It is worth noticing that a number of important topics Assuming the Lagrangian strain as the strain measure, the
directly concerned with the nonlinear dynamics of the single total axial strain of the centerline in C V is
冋册
cable, or somehow involving cable dynamics, are not specifi-
cally addressed in the deterministic part of this review. They ⳵ ⳵
␧ V ⫽1/2 RV 共 s I ,t 兲 * RV 共 s I ,t 兲 ⫺1 (1)
are simply referred to when suitable or felt necessary to es- ⳵s 0
⳵s0
tablish working links with the main core of the review, by
giving characteristic references where they can be found. which can also be written, on account of the expression of u,
This is due, on one side, to the need to abridge the reported
research material and make it uniform, as far as possible, in
order to give an in-depth account of the relevant available
␧ V ⫽␧ I ⫹␧ 冉冊 ds I
ds 0
2
(2)
knowledge, as per the assumed reviewing criterion. On the where ␧ I is the initial Lagrangian strain and ␧ is the addi-
other side, it is felt that some topics are of a rather specific tional strain referred to the initial state. The expression of the
nature, some other topics largely dealt with in the literature latter depends on whether the displacement vector field is
would actually need in-depth separate treatments, whereas
research achievements on further important topics relevant to
the isolated cable do not yet allow for a relevant comprehen-
sive review. The two limiting cases of taut strings and large-
sag cables enter the group of hastily addressed topics for the
isolated cable, which also includes modeling/mechanisms of
damping, inclined cables, traveling cables, cables carrying
moving masses, wave propagation, cable loosening, flexural
and torsional stiffnesses, wind action on cables, galloping,
and control of cable dynamics. The second group of unad-
dressed topics somehow related to cable dynamics includes
coupled systems 共bundle conductors, multiple overhead
transmission lines, cable-stayed and suspended bridges,
hanging structural systems, guyed masts兲, nonlinear fluid-
cable interaction 共towing and mooring lines, underwater
cables兲, and special applications occurring, eg, in the textile Fig. 1 Cable configurations and displacement components in a
industry or in tethered space dynamics. global reference system
resolved into Cartesian components along the axes of a glo- →CI, and T I is the cable tension in the equilibrium configu-
bal 共fixed兲 orthogonal system Ox i , i⫽1, 2, 3, as in Fig. 1 ration. Moreover, the lowercase symbols u i are intended to
关44兴, or into intrinsic components aligned with the local tan- denote displacement components in either reference system,
gential II1 , normal II2 , and binormal II3 directions defining a for the sake of conciseness, p i (s I ,t) and ␮ i are correspond-
Frénet triad at any point P I 共Perkins and Mote 关45兴兲. Reso- ing components of external loads and viscous damping coef-
lution into Cartesian components u i gives the following ex- ficients per unit initial length, and ␦ f ⫽ ␦ u 2 ( ␦ f ⫽ ␦ U 1 sin ␪
pression for the additional strain ⫹␦U2 cos ␪) in the global 共intrinsic兲 system, ␪ being the
angle of inclination of the unit tangent vector II1 from the
dx I1 ⳵ u 1 dx I2 ⳵ u 2 1 ⳵ui ⳵ui
␧ 共 s ,t 兲 ⫽
I
⫹ ⫹ (3) horizontal.
ds ⳵ s I I
ds ⳵ sI I 2 ⳵sI ⳵sI The cable tension T V in the dynamic configuration C V is
which is simpler than the corresponding expression with given by
ds V
displacement components U i in the Frénet triad: T V ⫽ 共 T I ⫹E I A I ␧ 兲 I (7)
冋冉冊 2 ds
⳵U1 ⳵U1
␧ 共 s ,t 兲 ⫽
I
⫺k U 2 ⫹1/2
I
⫺k U 2
I
which, on account of the expression of ␧, provides for an
⳵sI ⳵sI equation of elastic and geometric compatibility of the cable
⫹ 冉 ⳵U2
⳵sI
⫹k U 1
I
冊冉冊册
2
⫹
⳵U3
⳵sI
2
(4)
element.
It is worth noticing how, following the largely most fre-
quent assumption made in the cable dynamics literature, ref-
where k I (s I ) is the cable curvature in C I . erence is made herein to the Lagrangian strain as the center-
Formulating the problem in either of the two reference line strain measure. However, this is a non-geometric
systems has its own peculiar features. The global system, measure accounting only approximately for system nonlin-
already considered by Irvine and Caughey 关13兴 for cable lin- earities, which in contrast would be fully described through
ear dynamics, was referred to in, among others, Takahashi the exact kinematic modeling of the axial deformation of a
and Konishi 关46,47兴, Benedettini et al 关48兴, Gattulli et al cable element based on the so called engineering strain mea-
关49兴, Chang et al 关50兴, Rega et al 关51兴, Luo and Mote 关52兴, sure ␧⫽(ds V ⫺ds I )/ds I . The latter has been actually consid-
and Nayfeh et al 关53兴. It is preferred for the analysis of ered in the formulation of the exact equations of motion
cables having small equilibrium curvature, corresponding to pursued in two recent papers by Srinil et al 关26,61兴 dealing
sag-to-span ratios of about 1:8 or less. The intrinsic system, with horizontal and inclined cables, respectively 共see also
also used in several papers 共see, eg, Lee and Perkins 关54兴, 关62兴兲, as well as in the general formulation of cable dynamics
Triantafyllou and Howell 关55兴, and Chang and Ibrahim 关56兴兲 presented by Nayfeh and Pai 关63兴 in their very recent book
allows for easier account of the transport velocity of a par- on linear and nonlinear structural mechanics.
ticle in traveling cables 共linear dynamics 关57兴; nonlinear dy- Moreover, linearly elastic material and simple viscous
namics 关45兴兲, of the tangential motion of supports 关58,59兴, of damping are considered herein. Yet, apart from the possible
aerodynamic forces on cable 关60兴, and of large deformations importance of material nonlinearities, inherent viscoelasticity
of cables with bending and torsional stiffness 关23兴. and/or internal damping should be properly accounted for in
Depending on the chosen reference system, different ex- the formulation of some synthetic cables, and would be seen
pressions will result for the nonlinear equations of motion to have meaningful effects on the system response 共see, eg,
governing cable 3D finite dynamics. They are obtained 关23兴兲.
through the extended Hamilton’s principle 2.1.1 Equations of motion in global coordinates
␦ 冕
t1
t2
共 K V ⫺⌸ V 兲 dt⫹ 冕␦
t1
t2
W V dt⫽O (5)
The nonlinear equations of motion of an arbitrarily sagged
cable in the longitudinal, vertical, and horizontal out-of-
plane direction read 共Luongo et al 关44兴, see also 关46,64兴兲:
where K V , ⌸ V , and ␦ W V are the kinetic energy, the potential
兵共 T I u ⬘i ⫹E I A I ␧ 共 x Ii ⬘ ⫹u ⬘i 兲其 ⬘ ⫺ ␮ i u̇ i ⫹ p i
energy, and the virtual work associated with gravity, external,
and damping forces, respectively, given by ⫽m I ü i , 共 i⫽1,2,3 兲 (8)
K V ⫽1/2 冕 I
lc
m I u̇ 2i ds I where ␧ has the expression 共3兲 and the prime denotes ⳵ / ⳵ s I .
Eqs. 共8兲 are approximate to the third order for small but finite
冕
dynamic displacements. The boundary conditions for fixed
⌸ V ⫽⌸ I ⫹ 共 T I ␧⫹1/2E I A I ␧ 2 兲 ds I (6) supports read u i (0,t)⫽u i (l Ic ,t)⫽0. The equations of in-
I
lc plane motion are coupled through linear terms due to the
␦ W V⫽ 冕兵I
lc
m I g ␦ f ⫹ 关 p i 共 s I ,t 兲 ⫺ ␮ i u̇ i 兴 ␦ u i 其 ds I
cable initial curvature, the in-plane and out-of-plane equa-
tions are coupled through quadratic and cubic nonlinear
terms which arise from stretching of the cable centerline. The
In Eqs. 共6兲, the dot means ⳵ / ⳵ t, m I ⫽m 0 (ds 0 /ds I ) and solution of the equilibrium equations 共under dead load兲 ob-
E A I ⫽E 0 A 0 (ds I /ds 0 ) 3 are the mass and axial rigidity of a
I
tained from Eqs. 共8兲 by setting all time derivatives and u i to
unit cable element in the reference configuration, respec- zero, and linearizing, corresponds to the classical elastic cat-
tively, which account for the nonlinear deformation C 0 enary in the vertical plane 关1,65兴.
Upon assuming: i兲 the initial static configuration de- regime 关70,71兴, and should be considered in the analysis
scribed through the parabola y⫽4d 关 x/l⫺(x/l) 2 兴 with mid- 关23兴. This also entails keeping the longitudinal equation of
span sag d 共a valid approximation for d/l⭐1/8), which en- motion without simplifying assumptions 共ie, no condensa-
tails ds I ⬇dx I and T I ⬇H, H being the horizontal component tion兲. Luo and Mote 关52兴 developed a comprehensive 3D
of the initial tension, ii兲 the initial strain negligible with re- model governing the steady response of a traveling, arbi-
trarily sagged elastic cable, and obtained exact, closed form
spect to unity, which also entails H/E I A I Ⰶ1, and iii兲 the
solutions for steady motion under various loadings. Assum-
gradient of the longitudinal 共u兲 displacement component neg- ing the engineering strain as strain measure, Chucheepsakul
ligible with respect to unity, which corresponds to moder- and Wongsa 关25兴 and Srinil et al 关26兴 included the effect of
ately large rotations in the cable motion, the equations of axial stretching in the 2D and 3D equations of motion of an
motion read 共Benedettini et al 关48兴兲: arbitrarily sagged cable, respectively, and then solved the
equations numerically for large amplitude free vibrations, by
EA 关 u ⬘ ⫹y ⬘ v ⬘ ⫹ 共 1/2兲共 v ⬘ 2 ⫹w ⬘ 2 兲兴 ⬘ ⫺ ␮ u u̇⫹p u ⫽mü
also analyzing the ensuing spatial variability of the dynamic
兵 H v ⬘ ⫹EA 共 y ⬘ ⫹ v ⬘ 兲关 u ⬘ ⫹y ⬘ v ⬘ ⫹ 共 1/2兲共 v ⬘ 2 ⫹w ⬘ 2 兲兴其 ⬘ tension. Proper computation of the latter was also given at-
tention in 关28兴.
⫺ ␮ v v̇ ⫹p v ⫽m v̈ (9)
2.1.2 Equations of motion in intrinsic coordinates
兵 Hw ⬘ ⫹EAw ⬘ 关 u ⬘ ⫹y ⬘ v ⬘ ⫹ 共 1/2兲共 v ⬘ 2 ⫹w ⬘ 2 兲兴其 ⬘ ⫺ ␮ w ẇ In the absence of damping, the nondimensional equations of
motion of an arbitrarily sagged cable in the tangential, nor-
⫹p w ⫽mẅ
mal, and binormal direction read, respectively, by omitting
In Eqs. 共9兲, terms of order up to (d/l) 3 are retained, the the I superscript 共Lee and Perkins 关54兴兲:
superscript I is omitted, and engineering notations are used
关v 2t T 共 U ⬘1 ⫺kU 2 兲 ⫹ v 2l 共 1⫹U 1⬘ ⫺kU 2 兲 ␧ 兴 ⬘ ⫺k 关共 v 2t T⫹ v 2l ␧ 兲
for coordinates (x,y,z) and displacement components
(u, v ,w). In 关48兴, prescribed motions of supports in the glo- ⫻ 共 U ⬘2 ⫹kU 1 兲兴 ⫹ p 1 ⫽Ü 1
bal reference system enter the boundary conditions.
In the absence of external actions in the longitudinal di- 关共 v 2t T⫹ v 2l ␧ 兲共 U 2⬘ ⫹kU 1 兲兴 ⬘ ⫹k 关v 2l 共 1⫹U 1⬘ ⫺kU 2 兲 ␧
rection, and neglecting the corresponding inertia and viscous
forces, the u displacement component can be obtained from ⫹ v 2t T 共 U 1⬘ ⫺kU 2 兲兴 ⫹ p 2 ⫽Ü 2 (12)
the first of Eqs. 共9兲 through a static condensation procedure.
This is frequently used in both linear and nonlinear dynamics 关共 v 2t T⫹ v 2l ␧ 兲 U ⬘3 兴 ⬘ ⫹ p 3 ⫽Ü 3
of elastic shallow structures, starting from Kirchhoff’s work In Eqs. 共12兲, ␧ has the expression 共4兲, the constant coeffi-
共关66兴; see also, eg, 关67,68,63兴兲, and corresponds to the as- cients v l ⫽(EA/mgl) 1/2 and v t ⫽(H/mgl) 1/2 represent the
sumption that the cable stretches in a quasi-static manner nondimensional propagation speed of longitudinal and trans-
关13兴. Actually, the accuracy of this assumption deteriorates in verse waves along the cable, respectively, and the noncon-
the linear theory 关45兴 for a response involving high-order stant coefficients T(s) and k(s) represent the nondimen-
elastic modes which exhibit substantial longitudinal motion sional equilibrium tension and the curvature of the classical
关69兴. Thus, Eqs. 共9兲 are reduced to two partial integro- catenary. Perkins and Mote 关45兴 also accounted for the effect
differential equations in the planar v and nonplanar w dis- of transport speed of a particle, while small prescribed oscil-
placement components lations of one support in the direction tangential to the equi-
兵 H v ⬘ ⫹ 共 EA/l 兲共 y ⬘ ⫹ v ⬘ 兲 e 共 t 兲其 ⬘ ⫺ ␮ v v̇ ⫹p v ⫽m v̈ librium cable entered the boundary conditions in Perkins
(10) 关58兴.
兵 Hw ⬘ ⫹ 共 EA/l 兲 w ⬘ e 共 t 兲其 ⬘ ⫺ ␮ w ẇ⫹ p w ⫽mẅ When introducing the parabolic approximation to the cat-
where enary, which entails d/l⬍1/8, the equilibrium tension and
curvature are constant to first order in the small curvature
e共 t 兲⫽ 冕
l
关 y ⬘ v ⬘ ⫹1/2共 v ⬘ 2 ⫹w ⬘ 2 兲兴 dx (11)
parameter k⬇1/( v t ) 2 ⫽(8d)/l, and the equations of motion
共in U, V, W displacement notations兲 of an asymptotic model
for small equilibrium curvature are obtained:
with h(t)⫽EAe(t) representing the spatially uniform addi-
tional 共dynamic兲 cable tension which appears in Eq. 共7兲. v 2l 关 U ⬘ ⫺kV⫹1/2共 V ⬘ 2 ⫹W ⬘ 2 兲兴 ⬘ ⫽Ü
Such equations 共which can be suitably nondimensional-
ized兲 are accurate for studying the low-frequency nonlinear 共兵 v 2t ⫹ v 2l 关 U ⬘ ⫺kV⫹1/2共 V ⬘ 2 ⫹W ⬘ 2 兲兴其 V ⬘ 兲 ⬘ ⫹k v 2l 共 U ⬘
forced dynamics of suspended cables for which H/EA
⫺kV 兲 ⫽V̈ (13)
⫽O 关 (d/l) 2 兴 and the dynamic displacement components u,
v , and w are, respectively, u⫽O 关 ␧d 2 /l 兴 , v and w 共兵 v 2t ⫹ v 2l 关 U ⬘ ⫺kV⫹1/2共 V ⬘ 2 ⫹W ⬘ 2 兲兴其 W ⬘ 兲 ⬘ ⫽Ẅ
⫽O 关 ␧d 兴 , ␧ herein denoting a small parameter of the order
of the amplitude. In technical applications, such situations Neglecting the longitudinal inertia term can be given an
occur for sag-to-span ratios up to about 1/20. interpretation just in terms of propagation speed of longitu-
It is worth noticing how, depending on cable elasto- dinal and transverse waves, the former greatly exceeding the
geometric properties, the effect of axial extensibility on the latter. Indeed, the ratio v 2t / v 2l , which coincides with the ini-
dynamic behavior can be significant already in the linear tial tension-to-axial rigidity ratio H/EA, is of order k or
smaller. Consequently, on the time scale of 共lower frequency兲 H v ⬙ ⫹hy ⬙ ⫽m v̈

transverse motions, the stretching of the cable occurs nearly (16)
instantaneously or quasi-statically 关13兴. This allows to deter- Hw ⬙ ⫽mẅ
mine the tangential displacement component U, while the In Eqs. 共16兲, h(t)⫽EAe(t) represents the first-order part of
ensuing two partial integro-differential equations 共corre- the spatially uniform dynamic tension 共see Eq. 共11兲兲.
sponding to Eqs. 10兲 in the two 共nondimensionalized with The horizontal out-of-plane motion w is uncoupled from
respect to span兲 transverse displacement components V and the vertical in-plane motion v because, to first order, it in-
W read: volves no additional cable tension. The solutions of the in-
plane and out-of-plane eigenvalue problems provide the rel-
关v 2t ⫹ v 2l g 共 t 兲兴 V ⬙ ⫹k v 2l g 共 t 兲 ⫽V̈
(14) evant natural frequencies and linear normal modes 关13兴.
关v 2t ⫹ v 2l g 共 t 兲兴 W ⬙ ⫽Ẅ 2.2.1 Out-of-plane vibration
The eigenvalue problem for the out-of-plane horizontal
where motion is identical to that of a taut string and admits the
冕 1
eigenmodes
g共 t 兲⫽ 关 ⫺kV⫹1/2共 V ⬘ 2 ⫹W ⬘ 2 兲兴 dx (15)
w i 共 x 兲 ⫽A wi sin共 i ␲ x/l 兲 , i⫽1,2,3, . . . (17)
0
and the associated natural frequencies

Herein, the sum 关v 2t ⫹ v 2l g(t) 兴 represents the nondimensional
total cable tension composed of the static component v 2t and
the dynamic component v 2l g(t) 共see Eq. 共7兲兲. ␻ i⫽
i␲
l
冑 H
m
, i⫽1,2,3, . . . (18)
The low sag-to-span 共or small equilibrium curvature兲
cable model, formulated in either reference system, repre- The frequency of the first 共symmetric兲 out-of-plane horizon-
sents a nonlinear extension of the corresponding linear tal mode is the lowest natural frequency of any given para-
theory 关13兴共see also 关57兴兲. Moreover, provided displacement bolic cable.
is nondimensionalized with respect to cable span 关44兴, the
2.2.2 In-plane vibration
nonlinear equations of motion of a taut string 关68,72兴 are
Antisymmetric and symmetric modes with respect to cable
obtained in the limit of zero sag-to-span ratio 共equilibrium
midspan must be distinguished. The former 共the latter兲 con-
curvature兲, due to vanishing of quadratic nonlinearities.
sist of antisymmetric 共symmetric兲 vertical components and
symmetric 共antisymmetric兲 longitudinal components. No ad-
2.2 Linearized dynamics of small-sag continuous cable ditional cable tension h(t) is induced by the motion in the
The in-plane linear mode of a 共shallow兲 cable in the non- antisymmetric modes, to first order, whereas additional cable
condensed formulation always combines transverse and lon- tension is induced in the symmetric modes 关77兴.
gitudinal motion. However, transverse 共elastic兲 modes are The natural frequencies of the antisymmetric modes are
those which involve substantial transverse/vertical given by
共tangential/longitudinal兲 motion 共Burgess and Triantafyllou
关69兴兲, the behavior of cables at low frequency being domi-
nated by transverse modes, while that at high frequencies
␻ i⫽
2i ␲
l
冑 H
m
, i⫽1,2,3, . . . (19)
involving meaningful elastic modes. Since we are dealing and are the same as of taut string modes 关6兴. The correspond-
with low-frequency nonlinear dynamics, we report only on ing vertical modal components read
the corresponding transverse modes 共Irvine and Caughey
v i 共 x 兲 ⫽A v i sin共 2i ␲ x/l 兲 , i⫽1,2,3, . . . (20)
关13兴兲, whereas elastic modes of high-frequency are dealt with
in 关69兴. The effect of axial stretching on the linear frequen- The vertical component of the ith-symmetric mode is
cies of arbitrarily sagged cables is accounted for in 关73兴 and, given by
recently, in the numerical model of Srinil et al 关26兴, who also
investigate the nonlinear interaction effects of low-frequency C vi
v i共 x 兲 ⫽ 兵 1⫺tan共 1/2␤ i l 兲 sin ␤ i x
elastic modes occurring for highly extensible cables 关74兴. In 共 ␤ il 兲2
turn, accounting for the longitudinal inertia within a wavelet-
Galerkin procedure has allowed Al-Qassab and Nair 关75,76兴 ⫺cos ␤ i x 其 , i⫽1,3,5 . . . (21)
to highlight the occurrence of high-order modes with hori- where C v i is an arbitrary constant and ␤ i ⫽ ␻ i (m/H) , ␻ i 1/2
zontal components larger than the vertical ones, called re- being the corresponding natural frequency. The value of ␤ i l
verting partially swapping modes. follows from the transcendental equation
Upon eliminating the longitudinal displacement compo-
nent u from the linearized version of Eqs. 共9兲 through the tan共 1/2␤ l 兲 ⫽ 共 1/2␤ l 兲 ⫺ 共 4/␭ 2 兲共 1/2␤ l 兲 3 (22)
condensation procedure, the equations for linearized un-
where
damped free dynamics of a parabolic cable in global coordi-
nates are obtained from Eqs. 共10兲 by considering that H ␭ 2 ⫽64共 EA/H 兲共 d/l 兲 2 ⫽512共 EA/mgl 兲共 d/l 兲 3 ⫽ v 2l / 共 v 2t 兲 3
⫽const. and by neglecting second-order terms: (23)
is a nondimensional parameter which accounts for the elastic the amplitudes of the corresponding vertical components,
and geometric cable properties, and governs the nature of the which also justifies condensing them in the treatment of non-
roots of the equation. linear dynamics 共see Section 2.1兲.
When ␭ 2 is very large 共ie, large E, being d/l of about 1:8 The spectrum of natural frequencies of the parabolic cable
or less兲, Eq. 共22兲 reduces to tan(␤l/2)⫽ ␤ l/2, that is the tran- is shown in Fig. 3 in terms of the parameter ␭/␲. Increasing
scendental equation of inextensible cables 关6,7兴. When ␭ 2 is ␭/␲ from the taut string value 共␭/␲→0兲 beyond the first
very small 共ie, d/l→0 for technical cables, unless consider- crossover value 共␭/␲⫽2兲, the frequency of the first planar
ing very soft—eg, rubber—cable material兲, the transcenden- mode, the symmetric one 共one half-wave on the span, Fig.
tal equation is reduced to tan(␤l/2)⫽⫺⬁, whose roots cor- 2a兲, grows and the relevant modal shape undergoes a modi-
respond to those of the symmetric modes of the taut string. fication towards a shape with three half-waves on the span
Three basic cases are considered as regards the value of 共Fig. 2c兲. In contrast, the frequency and shape of the second
␭ 2 . According to whether ␭ 2 ⬍4 ␲ 2 (␭ 2 ⬎4 ␲ 2 ), the planar mode, the antisymmetric one 共two half-waves on the
frequency of the first symmetric in-plane mode is lower span兲, keep practically constant. Increasing ␭/␲ further on,
共greater兲 than the frequency of the first antisymmetric the natural sequence of planar frequencies is inverted: odd
in-plane mode. In the former case, the first symmetric verti- 共even兲 modes are antisymmetric 共symmetric兲. In contrast, all
cal modal component has no internal nodes along the span of the out-of-plane modes are independent of ␭/␲. Crossover
共Fig. 2a兲; in the latter case, it has two nodes along the span points characterized by tuning of in-plane symmetric and
共Fig. 2c兲. If ␭ 2 ⫽4 ␲ 2 , the first symmetric and antisymmetric antisymmetric frequencies are of high interest from a nonlin-
frequencies are equal, which corresponds to the first cross- ear dynamics viewpoint, since they correspond to realization
over point: correspondingly, the first symmetric vertical of internal resonance conditions. At these points, being the
modal component is tangential to the static profile at each m-th nonplanar frequency, m⫽1,2,3, . . . , in the m/p ratio
support 共Fig. 2b兲. A similar situation occurs for pairs of with the p-th crossover frequency, p⫽2,4,6, . . . , the pos-
symmetric/antisymmetric modes of higher order at succes- sible occurrence of multiple 共primary and secondary兲 internal
sive crossover points characterized by higher parameter val- resonances appears evident. Further internal resonance con-
ues ␭ 2 ⫽4 j 2 ␲ 2 , j⫽2,3 . . . : each of them marks the transi- ditions are also highlighted in Fig. 3.
tion from a relevant symmetric mode of a category of
‘‘elastic and taut cables’’ (␭ 2 →0) to that of a category of 2.3 Linearized dynamics of sagged cableÕmass suspen-
‘‘inextensible and sagged cables’’ (␭ 2 →⬁). sions
For both antisymmetric and symmetric modes, the longi- It is worth reporting shortly on the dynamics of suspended
tudinal components of motion are found 共through the con- elastic cables which support an array of attached masses
densation procedure兲 from the cable compatibility equation 共cable/mass suspension systems兲. The relevant nonlinear
e(t)⫽u ⬘ ⫹y ⬘ v ⬘ , obtained by neglecting the longitudinal in- equations of 3D finite motion can be derived through the
ertia forces in the linearized version of Eq. 共9a兲. The relevant extended Hamilton’s principle 共Eq. 共5兲兲 by simply including
amplitudes are found to be small of order (d/l) relative to the terms due to concentrated masses in the kinetic energy
and in the virtual work done by gravity. The general solution
of the ensuing linearized equations governing the free re-
sponse of a system with arbitrarily large equilibrium curva-
ture and nonuniform equilibrium tension is described by a
piecewise smooth catenary 关1兴.
Fig. 3 Spectrum of natural frequencies of parabolic cable. Dashed

Fig. 2 Possible forms of the first symmetric vertical modal com- 共horizontal continuous兲 lines denote frequencies of out-of-plane
ponent. a兲 ␭ 2 ⬍4 ␲ 2 ; b兲 ␭ 2 ⫽4 ␲ 2 ; c兲 ␭ 2 ⬎4 ␲ 2 关13兴. symmetric 共in-plane and out-of-plane antisymmetric兲 modes.
For small sag-to-span ratios, the parabolic approximation 2.4 Discrete models of continuous cable for analysis of
to the catenary is made, and an asymptotic model of cable/ reduced problems
mass suspensions is derived 共Cheng and Perkins 关78兴兲 along Discretized versions of the equations of motion of a continu-
the same line as for small-sag bare cables. The relevant ei- ous cable in global coordinates are obtained in the frame-
genvalue problem of free in-plane vibration is governed by work of the search for a solution for the unknown displace-
the same elastogeometric parameter ␭ which governs the dy- ment components u i through separate variables:
namics of the bare cable, and the frequency spectrum of a ⬁
symmetric mass array exhibits the same crossover phenom-
enon between symmetric and antisymmetric frequencies. The
u i 共 x,t 兲 ⫽ 兺
j⫽1
f i j共 x 兲q i j共 t 兲, i⫽1,2,3 (24)
parameter ␭ influences only the symmetric modes, which
gain two additional nodes after each crossover and induce Then, one assumes the spatial functions, f i j (x), space dis-
dynamic tension in the system, whereas the antisymmetric cretization, or the temporal functions, q i j (t), time discretiza-
modes are independent of ␭. However, within this general tion. With time discretization, the q i j (t) are usually taken to
pattern, differences occur in the eigenstructure of the spec- be harmonic and the method of harmonic balance is used to
trum depending on the number of masses. With few masses, obtain an infinite set of nonlinear boundary-value problems
the natural frequencies are substantially clustered into pass- for the f i j (x). With space discretization, the f i j (x)—and
bands separated by large stop bands, an eigenstructure cor- thus the spatial dependence—are assumed a priori 共assumed
responding to that of the string/mass system 关79兴 which the mode technique兲 and are considered constant during the mo-
cable/mass one converges to for ␭ 2 →0. Moreover, the cross- tion. If the boundary conditions are homogeneous, the f i j (x)
over phenomenon disrupts the stop band/pass band eigen- are usually taken to be the eigenfunctions of the linearized
structure as the natural frequency associated with the problem. The method of weighted residuals or variational
highest-order symmetric mode in each pass band migrates principles are then used to determine an infinite set of non-
through the adjacent stop band to the next pass band. In linear ordinary differential equations governing the q i j (t).
contrast, with a large number of evenly spaced masses, the Nearly all of the cable discrete models developed in the lit-
eigenstructure of the spectrum is nearly identical in appear- erature have been formulated in the context of a space dis-
ance to that of a bare cable, which the asymptotic model cretization approach through the Galerkin procedure. More-
over, the summation in Eq. 共24兲 is usually truncated to a
converges to.
finite integer n.
The situation differs both qualitatively and quantitatively
Accordingly, one main aspect is concerned with the num-
for asymmetric arrays. Indeed, repeated frequencies at cross-
ber of discretizing terms to be considered in an approximate
over split with each other, giving rise to a curve veering
finite-degree-of-freedom model. It depends on the features of
phenomenon 关78,80兴 characterized by close but distinct fre-
the original continuum problem we are trying to preserve and
quencies, whose degree is mode dependent. Moreover, all of
on the method we will apply to get the solution of the dis-
the corresponding in-plane modes are asymmetric 共ie, ‘‘hy-
crete system. Low-order models must be considered when
brid’’兲 and, therefore, they may all induce cable dynamic
searching for a solution through analytical or semianalytical
tension. These phenomena associated with asymmetric ar-
methods, which often allow to highlight basic dynamical fea-
rangements of masses also occur for the bare cable when
tures of the nonlinear problem.
asymmetry is induced by supports at different levels 共Trian-
The selection of proper dof to be taken into account for
tafyllou 关81兴兲 or cable translation speed 共Perkins and Mote
the elastic cable depends on whether the attention is devoted
关45兴兲. A unifying interpretation of the crossover/veering phe-
to system 2D 共planar兲 or 3D 共spatial兲 dynamics, on possible
nomena occurring in symmetric/asymmetric suspended
nearness to conditions of internal resonance involving sys-
cables is given in 关82兴 in the background of the equally rich
tem natural frequencies, and on the dynamical features of the
phenomena of mode localization.
external actions. In the sequel, reduced models for 2D and
Generally speaking, three generic transitions can be ob-
3D finite dynamics are distinguished, making reference to
served in the frequency spectrum of cable/mass suspensions
the equations in the global reference system.
as ␭ 2 is varied towards large values 关83兴. i兲 The elastic mode
Of course, purely numerical approaches based on either
transition, corresponding to the one in 关69兴, which influences
finite differences or finite elements, and possibly combined
only very high-order cable modes unless the cable material is with other techniques 共perturbation, harmonic balance, hy-
extremely soft. ii兲 The string/inextensible cable transition, brid pseudo-force/Laplace transform兲, are implemented
corresponding to the crossover 共veering兲 phenomenon ac- when one is interested in developing multi-dof models of
cording to whether the system is symmetric 共asymmetric兲, cables 关24,26 –28,36,37,84兴. These alternative approaches
which occurs in the small-sag regime. iii兲 The decoupling will not be addressed in the following.
transition, characterized by localized vibration modes, which
occurs for cable/mass systems with significant sag 共ie, rela- 2.4.1 Reduced models for 2D finite dynamics
tively low static tension兲. The first two transitions also occur To study the cable monofrequent planar response to a har-
for the bare cable, and they even coexist in the case of sym- monic forcing of frequency ⍀ with given spatial distribution
metric 共ie, horizontal兲 configurations of highly extensible ␸ (x), namely p(x,t)⫽ ␸ (x) P cos ⍀t, reference is made to
cables, the latter one is typical of 共large-sag兲 cable/mass sus- the simple single-dof model obtained by describing the dis-
pensions. placement v (x,t) through one eigenfunction of the linearized
equation of free motion 共16a兲: v (x,t)⫽ f (x)q(t). By apply- not only a parametric excitation of transverse response
ing the Galerkin method to Eq. 共10a兲共with w⬅0), one di- 关90,91兴 but also an external excitation 关58,92兴.
mensionless ordinary differential equation is obtained 共Lu- Within a more general finite-dimensional context, the con-
ongo et al 关44兴, Benedettini and Rega 关85兴兲: tribution of an increasing number of planar modes not di-
rectly excited has recently been considered 关93兴 in the analy-
q̈⫹ ␮ q̇⫹q⫹c 2 q 2 ⫹c 3 q 3 ⫽ P cos ⍀t (25) sis of primary resonance of the first symmetric or
antisymmetric mode of a noninternally resonant cable.
The coefficient of the linear term in Eq. 共25兲 is equal to unity 2.4.2 Reduced models for 3D finite dynamics
since time is nondimensionalized with respect to a cable lin- Due to their overall flexibility, cables are strongly susceptible
ear frequency, and the other coefficients depend on cable to undergoing spatial finite oscillations. Coupled in-plane
properties and on the assumed shape functions. and out-of-plane motion has been investigated through vari-
Equation 共25兲 corresponds to a Helmholtz-Duffing oscil- ous low-order models. The first two-dof model was the
lator in the theory of dynamical systems 共see Section 4.1兲. It spring/pendulum-like system formulated by Luongo et al
contains all of the main features of the original planar con- 关94兴 directly in a discrete framework. Besides the two-dof
tinuum model, namely the quadratic 共Helmholtz兲 and cubic model of Al-Noury and Ali 关95兴, who used sine functions 共ie,
共Duffing兲 nonlinearities associated with the initial curvature the eigenmodes of the taut string兲 as shape functions, the
and stretching of the cable axis, respectively. The relevant basic two-dof model of a parabolic cable—obtained from
coefficients, c 2 and c 3 , account for cable properties just Eqs. 共10兲 or 共14兲 through a Galerkin procedure—accounts
through the same unique elastogeometric parameter ␭ 2 共Eq. for the relevant first planar and nonplanar symmetric modes.
共23兲 that governs the linear dynamics, when the displacement It allows to analyze the enhanced nonlinear interaction phe-
is nondimensionalized with respect to the sag 关44兴. In con- nomena due to the 2:1 internal resonance occurring at first
trast, they depend separately on the mechanical and geo- crossover. It was first considered in Rao and Iyengar 关96兴 and
metrical cable properties, when the displacement is nondi- Lee and Perkins 关54兴, who considered an in-plane harmonic
mensionalized with respect to the span. This is, however, a excitation 共the former authors also included a uniform lateral
more suitable assumption for the effective description of component to simulate forces occurring during vortex-
resonant motions of cables with only different sag-to-span induced oscillations, at least to a first approximation兲. More-
ratios 共see Section II.2.1兲. over, it was used in Perkins 关58兴, whose discrete equations
The longitudinal displacement component u is accounted include both parametric and external excitation terms en-
for in the single-dof cable model through the above men- tailed by the tangential motion of one support. Parametric
tioned condensation procedure. An alternative procedure to and external terms due to in-plane supports motion also oc-
take it into account would consist in searching for a separate cur in a two-dof model recently developed by Nielsen and
variable solution of Eqs. 共9a, b兲共with w⬅0) in the form Kirkegaard 关97兴 to analyze the 1:1 internal resonance be-
u(x,t)⫽g(x)q(t), v (x,t)⫽ f (x)q(t), where g(x), f (x) are tween in-plane and out-of-plane modes of an inclined cable
the two components of the assumed eigenfunction of the with very small sag 共nearly a taut string兲, see also 关98兴.
linearized problem associated with Eqs. 共9a, b兲 with no lon- Moreover, a two-dof model is considered in Xu and Yu 关99兴
gitudinal inertia forces, and then applying the Galerkin pro- and Zhao et al 关100兴 to study the same resonance problem
cedure 关86,87兴. The same sdof model 共Eq. 共25兲兲 is obtained, for a small sag inclined cable externally excited at in-plane
but the coefficient of the cubic term is different, due to non- and out-of-plane primary resonance, respectively. A three-dof
exact satisfaction of the nonlinear integral relation between model using sine functions as the functional basis was con-
the longitudinal and vertical dispacement components deriv- sidered by Takahashi and Konishi 关46兴 to analyze coupled
able from Eq. 共9a兲关44兴. in-plane and out-of-plane eigenvibrations of cables with
With proper choice of the cable vertical eigenfunction, large sag. In turn, Lee and Perkins 关101兴 added the first non-
Eq. 共25兲 can be used for studying either symmetric or anti- planar antisymmetric mode to their two-dof model including
symmetric cable planar response under the corresponding the first planar and nonplanar symmetric modes to account
harmonically varying distributed forces. It is worth noticing for the simultaneous 2:1:2 internal resonances occurring at
that the quadratic nonlinearity in Eq. 共25兲 vanishes if the first crossover. A four-dof model including the planar and
assumed mode is antisymmetric. nonplanar, symmetric and antisymmetric, modes was first
To study the coupled nonlinear response involving both considered in Benedettini et al 关48兴 to account for the mul-
symmetric and antisymmetric components—the latter aris- tiple 2:2:1:2 internal resonance which actually characterizes
ing, eg, in the presence of symmetric external forces due to a cable at first crossover.
parametric excitation terms—at least a two-degree-of- The latter is obtained from Eqs. 共10兲 by describing each
freedom model must be developed, by considering in Eq. of the two 共in-plane and out-of-plane兲 displacement compo-
共24兲 a two-term expansion with vertical components of the nents through a two-term truncated series of eigenfunctions,
corresponding eigenfunctions 关88兴. A model with higher one symmetric ( f 1 and f 3 ) and one antisymmetric ( f 2 and
共five兲 dof number was considered in 关89兴 to analyze the cou- f 4 ):
pling effects of parametric and external resonances induced
v共 x,t 兲 ⫽q 1 共 t 兲 f 1 共 x 兲 ⫹q 2 共 t 兲 f 2 共 x 兲
by axial and transverse in-plane motions of one support in an (26)
inclined cable. Indeed, support motion induces, in general, w 共 x,t 兲 ⫽q 3 共 t 兲 f 3 共 x 兲 ⫹q 4 共 t 兲 f 4 共 x 兲
Correspondingly, the external 共in-plane and out-of-plane兲共27兲. They produce both external excitations 共terms with s
forcing are decoupled through harmonic loads p i (t), i⫽1, 4, superscripts兲, which add to directly external loads 共terms
with assigned, symmetric and antisymmetric, spatial distri- with e superscripts兲, and linear and nonlinear parametric ex-
butions ␸ i (x). The following system of ordinary differential citations 共terms with p superscripts兲 associated only to anti-
equations in the unknown amplitudes q i (t) of the four dis- symmetric support motions. For comparable, though slightly
different, nonlinear term, and external/parametric excitations
cretizing modes is obtained
4 4
due to supports motion of nearly taut inclined cables, see the
multi-dof equations of 关97,104兴. It is worth noticing the iden-
q̈ 1 ⫹ ␮ v q̇ 1 ⫹␭ 21 q 1 ⫹ 兺
i⫽1
c i q 2i ⫹ 兺
i⫽2,4
c ip q i0 q i ⫹ 兺 d 1i q 1 q 2i
i⫽1
tical structure of the second and fourth equations in system
共27兲, which depends on the eigenfunctions describing the
nth-order planar and nonplanar antisymmetric modes being
⫹ 兺
i⫽2,4
p 2
共 e 1i q i0 q 1 ⫹g 1i
p
q i0 q i q 1 兲 equal: this circumstance explains the need to account for
both of them in a reduced-order model.
⫽c e1 p 1 共 t 兲 ⫹c s1 共 q̈ 10⫹ ␮ v q̇ 10兲 ⫹ 兺
i⫽2,4
s 2
h 1i q i0
4
3 APPROXIMATE ASYMPTOTIC ANALYSIS
q̈ 2 ⫹ ␮ v q̇ 2 ⫹␭ 22 q 2 ⫹c 21q 2 q 1 ⫹c 21
p
q 20q 1 ⫹ 兺
i⫽1
d 2i q 2 q 2i Approximate solutions to the PDEs 共ODEs兲 governing the
weakly nonlinear dynamics of the continuous 共discretized兲
4 system can be obtained with any reduction method, such as
⫹ 兺 d 2ip q 20q 2i ⫹ i⫽2,4
i⫽1
兺 e 2ip q i02 q 2 ⫹g 24p q 40q 2 q 4 the method of multiple scales, the method of averaging, or
the method of normal forms 共see, eg, 关105–107兴兲.
Most recent studies of forced vibrations of suspended
⫹m 24
p
q 40q 20q 4
cables have been developed within the framework of the
space discretization approach 共see, eg, 关108兴兲, by applying
⫽c e2 p 2 共 t 兲 ⫹c s2 共 q̈ 20⫹ ␮ v q̇ 20兲 ⫹n s2 q 20⫹ 兺
i⫽2,4
s 2
r 2i q i0 q 20 the asymptotic procedure to the discretized ODEs. Single-
mode models are considered when the system is driven near
4
the natural frequency of a given mode that is not involved in
q̈ 3 ⫹ ␮ w q̇ 3 ⫹␭ 23 q 3 ⫹c 31q 3 q 1 ⫹ 兺 d 3i q 3 q 2i
i⫽1 an internal resonance with other modes, namely the response
is assumed to consist of only the excited mode. In contrast,
⫹ 兺
i⫽2,4
p 2
共 e 3i q i0 q 3 ⫹g 3i
p
q i0 q i q 3 兲
multimode models are considered when there is need to take
into account all of the modes that are either directly excited
or indirectly excited through some nonlinear coupling
⫽c e3 p 3 共 t 兲 ⫹c s3 共 q̈ 30⫹ ␮ w q̇ 30兲 mechanisms—possibly enhanced by the occurrence of inter-
4 nal resonance conditions—embedded in the equations of mo-
q̈ 4 ⫹ ␮ w q̇ 4 ⫹␭ 24 q 4 ⫹c 41q 4 q 1 ⫹c 41
p
q 40q 1 ⫹ 兺 d 4i q 4 q 2i
i⫽1
tion. In any case, when considering reduced-order models,
the problem arises of properly selecting the number and
4 kinds of modes to be retained in the discretization in such a
⫹ 兺 d 4ip q 40q 2i ⫹ i⫽2,4
兺 e 4ip q i02 q 4 ⫹g 42p q 20q 4 q 2 way to preserve all of the main features of the response of
i⫽1
the actual infinite-dimensional system 共see Section II.6.2兲.
⫹m 42
p
q 20q 40q 2 These problems are overcome when using the direct ap-
proach, where no a priori assumptions are made regarding
兺
the spatial distribution of the response and the asymptotic
⫽c e4 p 4 共 t 兲 ⫹c s4 共 q̈ 40⫹ ␮ w q̇ 40兲 ⫹n s4 q 40⫹ s 2
r 4i q i0 q 40
i⫽2,4 procedure is applied directly to the original PDEs 共see, eg,
Nayfeh et al 关109兴兲. This approach is considerably more in-
(27) volved than the discretization approach from the analytical
whose coefficients depend on the cable properties and on the point of view, but it allows us to take into account, in prin-
assumed shape functions. An analogous, though simpler, ciple, the contributions of all of the modes from the pertinent
four-dof model has been derived for the cable in intrinsic eigenspectrum to the nonlinear motions arising in the ab-
coordinates 共see 关102兴兲. Up to eight dof 共four in-plane and sence or in conditions of internal resonance. Direct attack of
four out-of-plane兲 have been dealt with in the Galerkin dis- the PDEs with a given reduction method and application of
cretization for specific purposes 关103兴. the same method to the associated infinite-dimensional
Quadratic and cubic nonlinearities occur in all four Eqs. Galerkin-discretized model lead to the same asymptotic dy-
共27兲. The former are, however, of pure nature—denoting that namics 关110,111兴. Accordingly, it becomes possible to extract
the symmetric in-plane component is always forced by the appropriate finite-dof models from the original system while
other components—in Eq. 共27a兲, and of a mixed nature in- still preserving the main features of the nonlinear response
volving q 1 —due to the initial curvature—in Eqs. 共27b–d兲. 关112兴.
The effect produced by symmetric and antisymmetric in- In the next sections, approximate solutions for single-
plane (q 10 and q 20) and out-of-plane (q 30 and q 40) pre- mode/multimode discretized models of suspended cable, as
scribed motions of the supports are also included in Eqs. well as for the relevant direct models, will be discussed.
Table 1. Ordering of problem parameters and external resonance con- Table 2. Order of first occurrence of contributions in the perturbation
ditions „single-dof model… equations „P: primary; S2: 1Õ2-sub, 2-super; S3: 1Õ3-sub, 3-super…
1/2-sub 1/3-sub excitation damping quadratic cubic
primary 2-super 3-super
␧ S2 S3
damping ␮ ␧2 ␧ ␧2 ␧2 S2 P S2 S3
excitation P ␧3 ␧ ␧ ␧3 P P S3 P S2 S3
frequency ⍀ 1⫹␧ 2 ␴ 1 ⫹␧ 4 ␴ 2 2⫹␧␴ 3⫹␧ 2 ␴
共1⫹␧␴兲/2 (1⫹␧ 2 ␴ )/3
perharmonic兲 occur with values of the excitation amplitudes

Only the case of external excitation 共ie, no parametric exci- larger than those at primary resonance. This justifies the as-
tation兲, mostly addressed in the literature, will be considered. sumed ␧-order for the excitation and its appearance in the
lowest-order perturbation equation at the same time as the
3.1 Single-mode discretized models
free oscillation part of the solution. Damping is ordered in
Approximate solutions to the single-dof ODE 共25兲 describing such a way to first appear in the same equation that generates
cable 2D forced dynamics are pursued by the method of the resonance, that is associated with the quadratic and with
multiple time scales 关105兴. Introducing a small but finite pa- the cubic nonlinearity in the order-1/2, order-2 cases and in
rameter ␧ of the order of the motion amplitude 共to denote a the order-1/3, order-3 cases, respectively. A second-order ap-
weakly nonlinear system兲 and new independent time vari- proximation to the solution q(t;␧) was considered for the
ables T h ⫽␧ h t, h⫽0,1, . . . ,n, the solution q(t) is repre- sdof model in all four cases 关113,114兴. Recently, order-2 su-
sented by an uniform expansion in power series of ␧, namely perharmonic resonance of a two-dof cable model has been
n
dealt with through the averaging method 关97兴.
q 共 t;␧ 兲 ⫽ 兺
h⫽0
␧ h⫹1 q h⫹1 共 T 0 ,T 1 , . . . ,T n 兲 (28) By using expansion 共28兲 and the positions in Table 1, (n
⫹1) recursive linear perturbation equations are obtained up
In Eq. 共28兲, T 0 is the fast time scale governing the dynamic to the ␧ n⫹1 order. Equations of first occurrence of the vari-
phenomenon, T 1 . . . ,T n are slow scales which the ampli- ous contributions are schematically indicated in Table 2. The
tudes and phases of q are modulated on by the nonlinearity, generating solution 共␧-order equation兲 is
damping, and external excitation. To account for the pres- q 1 ⫽A 共 T 1 , . . . ,T n 兲 e iT 0 ⫹⌳e i⍀T 0 ⫹cc (29)
ence of both quadratic and cubic nonlinearities, at least a
second nonlinear order approximation to the solution where ⌳⫽ P/2(1⫺⍀ 2 ) is the amplitude of the forced term
共namely, an unimproved ␧ 3 -order multiple scale approxima- 共missing in the primary resonance case兲, A is the complex
tion, where the solution is not obtained, see the terminology amplitude of the free term—whose dependence on the slow
in 关94兴兲 has to be sought. Ordering of damping and excitation time scales is determined by imposing the solvability condi-
depends on the considered resonance case, and must be suit- tions at subsequent orders—and cc stands for complex
able to balance with the nonlinearities in the perturbation conjugate.
equations. By sequentially eliminating terms producing secular terms
Five resonance cases with respect to a cable natural fre- in the solution, due to the resonance condition, and solving
quency are summarized in Table 1, namely: i兲 primary reso- the perturbation equation at each order, n first-order differen-
nance, ⍀⬵1 关85兴; ii兲 order-1/2, 共⍀⬵2兲, and order-1/3, 共⍀⬵ tial equations are obtained. Each of them describes the
3兲, subharmonic resonances 关113兴; iii兲 order-2, 共⍀⬇1/2兲, and modulation D k A, with D k ⫽ ⳵ / ⳵ T k , of the amplitude A on
order-3, 共⍀⬵1/3兲, superharmonic resonances 关114兴. Ordering the T k slow time scale (k⫽1, . . . ,n). A is found to be inde-
of problem parameters is reported, the ␴’s being 0共1兲 detun- pendent of the odd scales, namely A⫽A(T 2 ,T 4 ), in the pri-
ing parameters which describe quantitatively the nearness of mary resonance case as well as in the secondary resonances
the excitation frequency to the nominal resonance condition. associated with the cubic nonlinearities. To the performed
The ordering considered for damping and forcing at pri- nth-order approximation, the modulation on the actual non-
mary resonance is suitable for studying oscillations of am- dimensional time scale t is obtained through a reconstitution
plitude much greater than the static deflection of the cable procedure 关115兴
关1兴, such as they occur near resonance with small values of n
P, as well. Moreover, it allows for their effect to first appear Ȧ⫽ 兺

1k
␧ kD kA (30)
in the frequency-response relationship at the same stage as
the effects of the nonlinearities. A fourth nonlinear order ap- Performing a reconstitution is needed whenever a high-
proximation to the solution q(t;␧) 共namely, an unimproved order asymptotic solution is accomplished. Different proce-
␧ 5 -order multiple scale solution兲 is needed to correctly dures have been proposed and used in this respect even in the
describe the cable nonlinear behavior at rather large vibration larger context of internally resonant multimode models
amplitudes, and the lowest term in the expansion for ⍀ is 关51,53–54,116 –120兴, see Section 3.4.1.
assumed 0(␧ 2 ) for consistency with the order of the By expressing the complex amplitude A in the polar form
excitation 关85兴. A⫽1/2ae ib , with a and b real amplitude and phase, Eq. 共30兲
The weakly nonlinear oscillations in the neighborhood of gives rises to the amplitude and phase modulation equations
the main secondary resonances 共either subharmonic or su- 共APMEs兲:
a ␥˙ ⫽⌺ k ␧ k aD k ␥ , ȧ⫽⌺ k ␧ k D k a (31) cording to whether the effective nonlinearity coefficient

␣—which depends on both quadratic and cubic nonlineari-
where ␥ is a new phase introduced to obtain an autonomous ties and governs the backbone curve of the free oscillations
system. of the system—is positive or negative 共see Section II.2兲.
The time law of the steady response occurring to the
3.1.1 Steady state solutions
␧ 4 (␧ 2 ) order for the case of primary 共secondary兲 resonance
Steady state periodic solutions of Eq. 共25兲 correspond to the is obtained by means of Eq. 共28兲. Tuning of the frequency of
singular points of the APMEs: they are thus determined by the free term in the generating solution 共29兲 to subharmonic
imposing vanishing of the reconstituted slow time scale or superharmonic values of the excitation frequency is pro-
variation of a and ␥, namely ȧ⫽0, ␥˙ ⫽0. In the various duced by the nonlinearities in the cases of secondary reso-
resonant cases, both trivial (a⫽0) and nontrivial solutions— nance. To orders higher than 0共␧兲, superharmonic terms add
possibly coexisting—can occur in the neighborhood of sub- in all cases, whereas ultrasubharmonic terms add in the sub-
harmonic resonances, and they are alternatively attained de- harmonic resonance cases. Drift terms due to the quadratic
pending on the initial conditions. In contrast, only nontrivial nonlinearity in Eq. 共25兲 also occur in all cases, thus indicat-
solutions occur in the neighborhood of primary and super- ing that the oscillatory motion is not centered at q⫽0.
harmonic resonances. The actual deflection v (x,t) is obtained by means of the
A reconstitution procedure was followed in 关113,114兴 for assumed singlemode discretization v (x,t)⫽ f (x)q(t). To
the secondary resonance cases, thus obtaining the frequency- second approximation of primary resonance, it reads:
response equation f (a, ␴ ; ␮ , P)⫽0 for the nontrivial solu-
tion, where damping and forcing amplitude appear as param-
eters. Within a more general framework, a comparable 再
v共 x,t 兲 ⫽ a cos共 ⍀t⫺ ␥ 兲 ⫹
1
2 冋 1
c 2 a 2 cos 2 共 ⍀t⫺ ␥ 兲 ⫺1
3 册
冎
reconstitution procedure was adopted to study the in-plane
oscillations of a cable at nearly primary resonance in 关49兴: ⫹ . . . f 共x兲 (33)
therein, the excitation frequency ⍀ is expanded with a single
detuning parameter ␴ at ␧ 2 -order, instead of the two, ␴ 1 and
␴ 2 , reported in Table 1. 3.1.2 Stability of steady-state solutions
Considering distinct detuning parameters, as in 关85兴, Local stability of steady solutions is determined by the na-
makes sense when using an alternative procedure to obtain ture of the singular points of the APMEs. This analysis can
steady periodic solutions of Eq. 共25兲, namely, when imposing be made with reference to the relevant polar form 共Eqs. 共31兲兲
vanishing of the variations D k A at each ␧-order. Providing if there exist just nontrivial steady solutions, as occurs at
sequential frequency corrections to the linear oscillation fre- primary and superharmonic resonances. Small perturbations
quency, it allows us to evaluate and compare the effects of ␦ a and ␦␥ to the nontrivial solutions are introduced into Eqs.
the accomplished approximate nonlinear analyses to what- 共31兲, and the stability is determined by the eigenvalues of the
ever order. The formal equivalence of this procedure to the Jacobian matrix 共evaluated at the singular point兲 of the en-
former one depends on the system at hand and, in any case, suing system of linearized variational ODEs. Typically, loci
it relies on the ␧-power expansion of a further problem pa- of points separating stable and unstable solutions are deter-
rameter, eg, the damping 关117,121兴共see also the discussion mined in the frequency-response plane.
in 关122兴兲. This is aimed at introducing a second set of un- In contrast, whether both trivial and nontrivial steady so-
known coefficients 共in addition to the detuning coefficients兲 lutions exist, as happens at subharmonic resonances, their
needed to make vanishing of both real (aD k ␥ ⫽0) and stability analysis has to be made with reference to the com-
imaginary (D k a⫽0) conditions—ensuing from the imposed plex 共cartesian兲 form 共Eq. 共30兲兲 of the APMEs 关105兴. Alter-
vanishing of complex amplitude variation natively, separate analyses have to be made for the trivial and
(D k A⫽0)—possible at each ␧-order. This allows consistent nontrivial solutions. 共See, however, the discussion on the
reconstitution of all of the ␧-power expanded quantities. If it proper form of the amplitude modulation equations of reso-
is not done, inconsistencies can arise in the asymptotic solu- nant systems made in 关123兴兲. In such a case, regions of
tions, though the steady response of the system is not af- existence/nonexistence of nontrivial stable subharmonic os-
fected, as a rule, by the reconstitution procedure, as for the cillations in the frequency-amplitude space of the excitation
cable at primary resonance in 关85兴. The whole matter of re- are obtained. At both 1/2- and 1/3-subharmonic resonances,
constitution is addressed in 关120兴 with a worthy discussion critical values of the excitation occur below which only
on how correctly accomplishing it in terms of both ‘‘consis- trivial stable solutions exist and a finite-amplitude subhar-
tency’’ and ‘‘completeness.’’ monic is never excited. Above such thresholds, either regions
By way of example, we report the expression of the of existence of only nontrivial stable solutions or regions of
frequency-response relationship at primary resonance to the coexistence of both trivial and nontrivial stable solutions oc-
␧ 3 -order 共which does not imply reconstitution兲: cur 关113兴: which kind of oscillation actually develops in the
冉冊 1/2
latter case depends on the considered initial conditions. Do-
1 p2 mains of attraction of trivial or nontrivial solutions can be
␴ ⫽⫺ ␣ a ⫾ 2
⫺␮2 (32)
represented in the (a, ␥ ) plane of the response modulation
2 a2
parameters, wherein trajectories of nonstationary motions are
It exhibits a softening- or hardening-type behavior ac- obtained through numerical integration of the APMEs 共31兲.
Table 3. Ordering of problem parameters and resonance conditions for in-planeÕout-of-plane interaction at first crossover
primary external resonance 1/2-sub. res
Rao Lee Lee Benedettini Benedettini
Iyengar Perkins Perkins et al Rega
(q1 ,q3 ) (q1 ,q3 ) (q1 ,q3 ,q4 ) (q1 ,q2 ,q3 ,q4 ) (q1 ,q2 ,q3 ,q4 )
␮ ␧2 ␧, ␧ 2 ␧
pe ␧3 ␧ 2, ␧ 3 ␧
⍀ ␻ 1 ⫹␧ 2 ␴ ␻ 1 ⫹␧ ␴ 1 ⫹␧ 2 ␴ 2 2 ␻ 1 ⫹␧ ␴
internal resonances 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 3 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 31⫹␧ 2 ␳ 32 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 3
␻ k⫽ ␻ 1 ⫹␧ ␳ k1 ⫹␧ ␳ k2
2 ␻ k⫽ ␻ 1 ⫹␧ ␳ k
共k⫽4兲共k⫽4兲
same with k⫽2 same with k⫽2
2:1 2:1:2 2:2:1:2
Whether only nontrivial solutions exist, all of the nonstation- in the absence of internal resonances. For example, an out-
ary trajectories in the (a, ␥ ) plane are attracted to the stable of-plane mode can be excited as a consequence of a principal
foci—possibly more than just one—corresponding to the parametric resonance induced by a nonresonantly excited in-
steady solutions. plane mode, when there is a 2:1 resonance of the external
forcing with respect to the out-of-plane mode 关47,49兴. Of
3.2 Multimode discretized models course, coupling is enhanced by the occurrence of an ir con-
Approximate solutions to a multi-dof system of ODEs 共eg, dition between the two modes 关58兴.
the four-mode model in Eqs. 共27兲兲 describing cable 3D According to previous points, major classifications in the
forced dynamics are sought by the multiple time scale multimodal analysis of finite dynamics are concerned with
method with the same approach as that followed for the sdof the occurrence of external/parametric resonances and,
model, though much more involved from an algebraic point mostly, of one or more—and possibly simultaneous—
of view. Thus, using a symbolic manipulator 共see, eg, 关124兴兲 internal resonance conditions 关125,126兴. From the latter
is highly recommended for developing the perturbation pro- viewpoint, the elastic parabolic cable is a very natural can-
cedure and obtaining the APMEs. Nonetheless, cumbersome- didate to strong nonlinear interaction. Indeed, its spectrum of
ness of analytical computations usually suggests to seek for natural frequencies in terms of ␭ 2 共Fig. 3兲 shows overall
just a second-order approximation to the solution 共up to ␧ 3 ): perfect tuning of the n-th order planar/nonplanar antisym-
this is the lowest approximation that allows to reliably ac- metric modes, a variety of two-mode resonances, and several
count for the various dynamical phenomena possibly occur- possible simultaneous internal resonances between different
ring in a system with quadratic and cubic nonlinearities. The modes at the crossover points. Based on the occurrence of
solution q j (t) is expanded in a power series of ␧ as: orthogonality conditions of the involved nonlinear normal
2
modes 共Lacarbonara et al 关127兴兲, it can be shown that not all
兺
of the potential internal resonances are actually activable;
q j 共 t;␧ 兲 ⫽ ␧ h⫹1 q j 共 h⫹1 兲共 T 0 ,T 1 ,T 2 兲 , j⫽1,m (34)
h⫽0 nevertheless, most of them are 共Lacarbonara and Rega
关128兴兲.
The picture of resonance conditions of possible interest is
much richer than in the single-dof model. It depends on the 3.2.1 Analysis of nonlinear interaction at first crossover
cable characteristics summarized by the elastogeometrical A cable at nearly first crossover is in a multiple internal
parameter ␭ 2 , on the number m and the kind of displacement resonance condition involving all four first modes. Indeed,
modal components accounted for in the discretization, and both the symmetric (q 1 ) and antisymmetric (q 2 ) planar
on the frequency content of the considered components of modes and the antisymmetric nonplanar mode (q 4 ) have the
time-varying excitation. Many meaningful combinations of same frequency 共1:1 ir兲, whereas the symmetric nonplanar
internal/external resonances can be considered in practical mode (q 3 ) has half frequency 共2:1 ir兲. According to how
applications. Internal resonances 共ir兲 may be responsible for many modes are included in the discretized model, up to a
enhanced nonlinear coupling between the involved modes, complete 2:2:1:2 simultaneous resonance can be realized.
whose importance is directly correlated with the degree of A summary of problem parameters and resonance condi-
tuning of resonant frequencies. Strong modal interaction as- tions for the two-dof 共Rao and Iyengar 关96兴, Lee and Perkins
sociated with energy transfer from a mode directly excited at 关54兴兲, three-dof 共Lee and Perkins 关101兴兲, and four-dof 共Bene-
some 共primary or secondary兲 resonance to another mode in- dettini et al 关48兴兲 models proposed to study the in-plane/out-
volved in a nearly ir with the former one, often characterizes of-plane interaction at first crossover is reported in Table 3.
forced oscillations. By contrast, nearly independent nonlin- For each of them, a second-order multiple scale solution to
ear single-mode oscillations occur far from ir. Coupling phe- the relevant ODEs of motion 共all derivable from Eqs.
nomena also take place in the case of free oscillations under 共27兲—or from the equivalent intrinsic equations—after
ir conditions and proper disturbances of a given mode. How- eliminating the dofs not included in the discretization兲 was
ever, in forced oscillations, these phenomena may occur even performed under some external resonance condition of the
symmetric planar mode directly excited with a symmetrically Table 4. Summary of contributions to perturbation equation in the
case of support motion
distributed harmonic vertical forcing. In the table we report:
i兲 the orderings of the modal damping parameters and of the Order
external excitation, which depend on the chosen order of first Terms ␧ ␧2 ␧3
occurrence of the relevant contributions in the perturbation symmetric and inertial harmonic excitation x x x
equations to balance with the nonlinearities, and, mostly, on antisymmetric dissipative harmonic excitation x x
the occurrence of a primary or secondary external resonance; support motion
antisymmetric direct 2-superharmonic x
ii兲 the external and the various ir conditions respectively con- support motion excitation 共*兲
trolled by the detunings ␴ and ␳ k (k⫽2⫼4), possibly split at direct constant excitation 共*兲 x
parametric 2-superharmonic x
different ␧-orders as already discussed for the single-dof excitation
model. parametric constant excitation x
damping x x
By using expansion 共34兲 and the positions in Table 3, a quadratic nonlinearities x x
sequence of recursive systems of linear perturbation equa- cubic nonlinearities x
tions is obtained up to the ␧ 3 -order. The solution to the
␧-order system 共that generalizes Eq. 共29兲兲 is described by ( )
* only in equation of first symmetric mode.
i␻ jT0
q j1 ⫽A j 共 T 1 ,T 2 兲 e ⫹ ␦ j1 ⌳ j e i⍀T 0
⫹cc, j⫽1,4 (35)
with the excitation term arising 共in the equation for the first turbation analysis, the modulation of the complex amplitudes
dof, see the Kronecker ␦兲 only in the cases of secondary A j on the actual nondimensional time-scale t is obtained
external resonance 共eg, the 1/2-subharmonic in Table 3兲. By through the reconstitution procedure as
sequentially eliminating the secular producing terms due to
Ȧ j ⫽␧D 1 A j ⫹␧ 2 D 2 A j , j⫽1,4 (36)
the resonance conditions and solving the inhomogeneous
perturbation equations, two systems of equations governing Introducing the polar expressions A j ⫽1/2a j exp(ib j) and new
the modulation of the complex amplitudes A j on the T 1 and phases ␥, ␪ k , k⫽2⫼4, needed to reduce system 共36兲 to an
T 2 slow time scales are obtained. autonomous form, and separating real and imaginary parts,
The scale T 1 accounts for the 2:1 ir between the symmet- lead to eight nonlinear differential equations 共APMEs兲 gov-
ric in-plane and out-of-plane components activated by the erning the modulation of the four real amplitudes a j and of
quadratic nonlinearities. Accordingly, no difference occurs the four phases ␥ and ␪ k of the solution. The explicit APMEs
up to the ␧ 2 -order among the predictions of coupled response are very involved, so they are given here only in implicit
furnished by the two-, three-, or four-dof models—all ac- form:
counting for those two components—apart from the contri-
a j ␥˙ ⫽g j 共 a 1 ,a 2 ,a 3 ,a 4 ; ␥ , ␪ 2 , ␪ 3 , ␪ 4 兲
butions given by damping and external excitation terms,
共 j⫽1,4 兲
which enter the relevant modulation equations only when ȧ j ⫽ f j 共 a 1 ,a 2 ,a 3 ,a 4 ; ␥ , ␪ 2 , ␪ 3 , ␪ 4 兲 (37)
lowerly ordered 共see Table 3兲. Whether accounted for by the
model, the amplitudes A 2 and A 4 of the antisymmetric com- on the understanding that the generic function on the right-
ponents are either independent of T 1 or, in any case, decou- hand sides of Eqs. 共37兲 does not necessarily depend on all of
pled from the symmetric ones A 1 and A 3 . Indeed, they come the reported arguments. They contain many different terms
into play just due to the 1:1 ir with the symmetric in-plane according to whether primary 关48兴 or 1/2-subharmonic 关130兴
component which is linked with the cubic nonlinearities. external resonance of the first symmetric mode is considered,
This latter resonance is accounted for by the T 2 time scale, as well as to what kind of 共planar/nonplanar, symmetric/
together with the higher-order corrections due to the qua- antisymmetric兲 harmonic external excitation is taken into ac-
dratic nonlinearities. Of course, different predictions about count. Generally speaking, the coefficients of the nonlinear
the composition of the coupled response on the T 2 scale are and/or coupling terms in the APMEs depend on the cable
obtained when considering models with a lower or higher properties, on the assumed modes, and on the considered
number of modes. excitation. Their number is strongly reduced with respect to
A second-order multiple scale solution to the equations of that ensuing from the analysis by accounting for the orthogo-
motion of the four-dof model was also accomplished to ac- nality conditions of the considered eigenfunctions, as well as
count for the effects produced by symmetric and antisym- for a number of relationships existing between them. The
metric planar/nonplanar support motions with an order-1/2 latter highlight the symmetry 关131兴 of the matrices of coef-
subharmonic resonant frequency 关123兴. As observed in Sec- ficients occurring in the APMEs, which in turn reflects the
tion 2.4.2, support motions give rise to both external and self-adjoint nature of the original problem. They ensue from
parametric excitations of different kinds in Eqs. 共27兲. The the requirement for the APMEs to be derivable from a La-
relevant contributions to the various orders of the perturba- grangian 共as regards their unforced conservative part兲 in or-
tion procedure are summarized in Table 4, together with der to have a physical meaning 关51兴.
those of damping and of the nonlinearities. As regards the case commonly considered of only vertical
symmetric excitation, it is worth noticing how 关130兴 the non-
3.2.2 Amplitude and phase modulation equations linear terms depending on the modal amplitudes of response
With reference to the 2:2:1:2 internally resonant four-d.o.f. components are the same in the APMEs at primary or 1/2-
model of cable at first crossover and to a second-order per- subharmonic resonances. In contrast, the terms accounting
for the external excitation differ considerably in the two of motion—likely containing just the dof responsible for in-
cases. Indeed, at 1/2-subharmonic resonance, occurrence of stability of the former class—that is established at the bifur-
the excitation in the lowest-order system of perturbation cation point. For a general qualitative analysis of the existing
equations 共right column in Table 3兲—at the same time as the classes of motion of a multiresonant system, see 关132,133兴.
free oscillation part of the solution 共see Eq. 共35兲兲—gives rise With reference to a symmetric in-plane excitation, several
to many new terms of interaction between excitation and classes of steady solutions can occur for the four-d.o.f model.
response amplitudes in the eight modulation equations. Referring to just the nontrivial components, and keeping
It is also worth noticing that the terms in the APMEs apart special solutions 共including also the four-mode solu-
accounting for the effect of a vertical forcing symmetrically tion兲 which occur only for specific values of some modal
distributed along the cable can also represent the effect of an damping parameters, one can basically distinguish the fol-
in-phase harmonic vertical motion of the supports, provided lowing classes of solutions. i) Unimodal, which include the
equivalent amplitudes of a fictitious distributed loading are symmetric planar solution (a 1 ), or the antisymmetric planar
identified 关129兴. (a 2 ) or nonplanar (a 4 ) solution, depending on the consid-
ered external resonance condition. ii兲 Bimodal, either includ-
3.2.3 Steady-state solutions and stability analysis ing an antisymmetric component—planar (a 1 a 2 ) or nonpla-
For a given set of system parameters, the singular points of nar (a 1 a 4 )—or being totally symmetric and nonplanar
the APMEs 共37兲 give the approximate steady-state periodic (a 1 a 3 ). The former originate from 1:1 internal resonances
solutions to system 共27兲. Qualitatively different classes of and compete with each other by virtue of system parameters
motion, partially trivial 共involving one to three active dof兲 or values, whereas the latter originates from 2:1 internal reso-
nontrivial 共four dof兲 can occur in the neighborhood of pri- nance. iii兲 Trimodal, including two symmetric components
mary or 1/2-subharmonic resonance. and one antisymmetric component—either the planar
Because of the existence of several classes containing (a 1 a 2 a 3 ) or nonplanar (a 1 a 3 a 4 ) solution. They account
trivial components, their stability analysis has to be per- for simultaneous 1:1 and 2:1 internal resonances and com-
formed on the cartesian form 共Eq. 共36兲兲 of the APMEs. Small pete with each other. The various classes and their respective
perturbations ␦ A j to the steady-state, trivial or nontrivial, importance in cable finite dynamics will be discussed in Sec-
complex amplitudes corresponding to the singular points are tion II.4.2 from a mechanical point of view.
considered, and linearized variational ODEs are derived. The The time law of steady response to the ␧ 2 -order is ob-
time evolution of each perturbation depends on the sign of tained by means of Eq. 共34兲. Depending on the considered
the eigenvalues ␭ n , n⫽1⫼8, of the Jacobian matrix associ- excitation and external resonance condition, as well as on the
ated with the real and imaginary parts of the complex ampli- specific modal component of the response, rather involved
tude of the perturbations. They depend on the steady solution expressions of q j (t), j⫽1, 4 occur. They contain harmonic,
whose local stability is analyzed. When varying a control subharmonic, superharmonic, ultra-subharmonic, and drift
parameter, zeroing of the real part of one eigenvalue gives a terms, of either free or forced nature. In particular, due to the
bifurcation condition of the solution between the stable occurrence of all possible pure quadratic terms in the first
(Re ␭n⬍0) and unstable (Re ␭n⬎0) branches along an equi- discretized equation of motion 共27a兲, the second-order con-
librium path. At the relevant point, either the same class of tributions to the symmetric planar response depend on the
motion with different amplitude and/or period or a new class amplitudes of all of the dofs and/or of all of the components
of motion sets in. of support motion. In turn, the occurrence of a symmetric
The stability of each class of motion has to be evaluated planar component 共corresponding to either a prescribed sup-
with respect to the complete set of possible perturbations. port motion or an excited dof amplitude兲 in all of the second-
However, from both a mathematical and a physical point of order contributions to the other response components, high-
view, it is worth distinguishing between perturbations on ac- lights the dominant role played by the former in the
tive or nonactive modes and also considering separately the nonlinear modal interaction 关121兴.
single perturbations on the nonactive modes. Indeed, the
3.3 Direct models
composition in active modes of a considered class governs
the degree of coupling of the eigenvalue problem to be Within the framework of a direct asymptotic approach to
solved in the relevant stability analysis. For a class of steady nonlinear oscillations 关109兴, the method of multiple scales is
motion characterized by m active and p nonactive modes 共for applied directly to the PDEs governing cable finite dynamics
Eqs. 共36兲, m⫹p⫽4), it is never necessary to solve an 8⫻8 共Pakdemirli et al 关134兴, Rega et al 关51兴, Nayfeh et al 关53兴,
coupled eigenvalue problem in all four perturbations, but just Yu and Pan 关135兴, and Nayfeh and Pai 关63兴兲.
one coupled subproblem 共of order 2m⫻2m) involving the m Referring to Eqs. 共10兲 in global coordinates, solution for
active modes and p uncoupled subproblems 共of order 2⫻2兲 planar response (w⫽0) in the absence of any internal reso-
involving the p nonactive modes. Besides reducing the algo- nance is sought in the form
rithmic effort of the numerical stability analysis to be made n
in wide ranges of variation of the control parameters, this v共 x,t;␧ 兲 ⫽ 兺

h⫽0
␧ h⫹1 v h⫹1 共 x,T 0 ,T 1 , . . . ,T n 兲 (38)
circumstance allows one to identify a specific nonactive
mode as the key perturbation for instability of a given class that replaces Eq. 共28兲, and perturbation equations are ob-
of motion. This is also useful to the search for the new class tained up to a given order. Because, in the presence of damp-
ing, all of the modes that are not directly or indirectly excited highest-order (␧ 3 ) of the solution. This simplified ordering
decay with time 关68兴, the generating solution in primary with respect to that producing effects split on two different
resonance conditions is assumed to consist of the directly levels (␧ 2 ,␧ 3 ) 共see the four-mode discrete model in Table 3兲,
excited mode only entails the same results at the last considered order of the
asymptotic solution.
v 1 ⫽ 关 A 共 T 1 , . . . 兲 e i ␻ T 0 ⫹cc兴 f 共 x 兲 (39)
A sequence of recursive systems of linear perturbation
where ␻ and f (x) are its natural frequency and spatial shape, PDEs up to the ␧ 3 -order is obtained. The solution of the
respectively. ␧-order system is described as:
A second-order approximation to the deflection is ob-
tained in the form 关93,112兴: u k1 ⫽A i 共 T 1 ,T 2 兲 e i ␻ i T 0 f i 共 x 兲
1 2 ⫹A i⫹1 共 T 1 ,T 2 兲 e i ␻ i⫹1 T 0 f i⫹1 共 x 兲 ⫹cc, k⫽1,2 (42)

v共 x,t 兲 ⫽a cos共 ⍀t⫺ ␥ 兲 f 共 x 兲 ⫹ a 关 cos 2 共 ⍀t⫺ ␥ 兲 ␺ 1 共 x 兲
2
which replaces Eq. 共35兲 written for the case of primary reso-
⫹ ␺ 2 共 x 兲兴 ⫹ . . . (40) nance. In Eq. 共42兲, i⫽1(3) when k⫽1(2), and f 1 (x),
f 2 (x), ( f 3 (x), and f 4 (x)) are the lowest symmetric and an-
where the ␺ i (x) are solutions of boundary-value problems. tisymmetric in-plane, Eqs. 共21兲 and 共20兲, 共out-of-plane, Eq.
Comparing Eqs. 共33兲 and 共40兲, we note that, in the single-dof 共17兲兲 eigenmodes. Obtaining a second-order solution implies
model, the second-order correction affects the amplitude and the following steps. i兲 Sequentially accounting for internal
the phase of the motion with the drift and the superharmonic resonance conditions and determining solvability conditions
term, whereas in the direct model, the second-order correc- at ␧ 2 and ␧ 3 perturbation orders. Due to self-adjointness of
tions are both temporal and spatial. In particular, the full the associated homogeneous problems, this demands that the
spectrum of the system eigenmodes is taken into account in right-hand side of the in-plane 共out-of-plane兲 perturbation
the spatial functions ␺ i (x). equation at each order be orthogonal to e ⫺i ␻ 1 T 0 f 1 (x),
The amplitude a and the frequency detuning ␴ are still e ⫺i ␻ 2 T 0 f 2 (x) (e ⫺i ␻ 3 T 0 f 3 (x), and e ⫺i ␻ 4 T 0 f 4 (x)). Two suc-
related by the frequency-response equation 共32兲 where, how- cessive systems of ODEs governing the modulation of com-
ever, the effective nonlinearity coefficient ␣ has a different plex amplitudes A j , j⫽1, 4, on the T 1 and T 2 time scales are
expression. The following key points occur with respect to obtained. ii兲 Solving a number of boundary-value problems
the discretized single-mode model. i兲 The term generated by at ␧ 2 order, which in general also entails accounting for the
the cubic nonlinearity—responsible for hardening homogeneous solutions of the associated perturbation equa-
behavior—is the same to this order of approximation. ii兲 The tions. iii兲 Obtaining the general solution of the perturbation
term due to the quadratic nonlinearities—responsible for equations at ␧ 2 order. iv兲 Using solvability conditions at ␧ 2
softening behavior—is expressed in terms of the excited order to obtain the modulation equations at ␧ 3 order in terms
mode shape only in the discretization approach, while it cap- of the T 2 scale only.
tures contributions from all of the modes through the ␺ i (x) The modulation equations governing the time evolution of
functions in the direct approach. As a consequence, there is a the complex amplitudes A j on the actual nondimensional
possibility of a sign difference in the effective nonlinearity time scale t up to second order are obtained through the same
coefficient, depending on whether the discretization or direct reconstitution procedure as that leading to the modulation
approach is used 共Section II.2.1.2; see also 关93兴兲. equations 共37兲 for the four-mode discrete model. Indeed,
The multimodal direct approach has been applied to ob- there is a perfect correspondence between the coefficients of
tain asymptotic solutions in conditions of internal resonance. the nonlinear terms in the two sets of reconstituted equations
A four-mode Hamiltonian model of orbiting string with only for the discrete and direct models. Their number is strongly
quadratic nonlinearities associated to gyroscopic forces was reduced—in the latter case, too—by taking into account or-
considered in 关136兴. Two-mode and four-mode models of thogonality of the eigenmodes, as well as the requirement of
cable were developed to analyze, respectively, 1:1 ir between having the modulation equations derivable from a Lagrang-
the third 共second symmetric兲 in-plane mode and the third ian. The remarkable difference is that the coefficients depend
out-of-plane mode 共Pakdemirli et al 关110兴兲, and simulta- now not only on the cable geometrical/mechanical properties
neous 2:2:1:2 ir at first crossover 共Rega et al 关51兴, Nayfeh and on the assumed eigenmodes, but also on the solutions of
et al 关53兴兲. In the latter case, a second-order approximation the boundary-value problems characterizing the in-plane and
to the solution of Eqs. 共10兲 for nonplanar response at primary out-of-plane responses. This entails different values of the
external resonance was sought in the form coefficients of corresponding nonlinear terms in the two
2 models 关51兴. Accordingly, the possible relevant classes of
u k 共 x,t;␧ 兲 ⫽ 兺
h⫽0
␧ h⫹1 u k 共 h⫹1 兲共 x,T 0 ,T 1 ,T 2 兲 , k⫽1,2 steady solutions are the same, but qualitative as well as quan-
titative differences can occur as regards stability features and
(41)
bifurcations of some of them 共Section II.4.4.2兲.
with u k ⫽ v (w) when k⫽1(2), that replaces Eq. 共34兲. Order- The displacement field up to second order is obtained
ing of problem parameters and detunings of external/internal from Eqs. 共41兲. The two in-plane and out-of-plane generating
resonance conditions were chosen so as to have the effects of solutions 共␧ order兲 only account for the relevant symmetric
damping, excitation, and 1:1 ir directly appearing in the and antisymmetric eigenfunctions. On the other hand, the
second-order contributions to each displacement component problem 关122兴 or, more generally, by working on the equiva-
are solutions of some pertinent boundary-value problems, lent system of first-order 共ordinary or partial兲 differential
whose inhomogeneous terms depend on either the corre- equations of motion 关53,122兴. It is worth noticing how per-
sponding mode shapes or shape interaction terms between forming an incomplete reconstitution would entail signifi-
the fundamental modes and with other modes. Generally cantly reducing the effects of the 2:1 ir in the coupled system
response with respect to the 1:1 resonance, which is likely to
speaking, the effects of the infinite number of modes from
be a wrong mechanical effect.
the spectrum of the eigenvalue problems are included in the
solution, and the spatial dependence of the cable motion is 3.4.2 Analyzing strongly nonlinear dynamics
captured. These are remarkable differences with respect to Nonlocal 共ie, global兲 dynamical regimes of a shallow cable
the discrete model. are associated with its reaching and possibly keeping an in-
verted equilibrium position 共see Section 4.1兲. In this case,
3.4 Further topics cable dynamics are described by differential equations such
3.4.1 On the reconstitution in high-order solutions and that the orders of magnitude of linear and nonlinear terms are
mixed internal resonances essentially the same, so that linearization may give qualita-
Within the above discussed asymptotic approaches to cable tively incorrect information about the dynamic behavior. The
finite dynamics at first crossover, the simultaneous occur- ensuing strongly nonlinear regime cannot be predicted using
rence of different 共1:1 and 2:1兲 internal resonances— perturbation techniques starting from the original static equi-
accounted for at different nonlinear orders of the perturbation librium position.
solution—results in the modulation of nonlinear response If considering global dynamics, an ill-posed problem 关55兴
with two dominant slow time scales, namely ␧t and ␧ 2 t. arises when the cable tension becomes negative 关137兴. Pilip-
They account for the 2:1 and 1:1 resonance, respectively. As chuk and Ibrahim 关138,139兴 have recently addressed the mat-
already observed, the dependence on two time scales neces- ter showing that a qualitatively reasonable description of the
sitates performing a reconstitution of the modulation equa- nonlocal dynamical regime can be provided based on at least
tions governing the two nonlinear orders. Furthermore, the a two-mode approximation, with the two modal components
effects associated with the slow scale ␧t also appear in the being strongly coupled. They have introduced a special type
second-order nonlinear problem. of coordinate transformation which provides a description of
This is a general matter arising for all asymptotic solu- the in-plane and out-of-plane motions in terms of a set of
tions governed by at least two time scales, irrespective of coordinates constructed along and perpendicular to the un-
their being obtained with a discretization or direct approach. stretched manifold of cable configurations. It proves useful
It has received increasing attention in the recent literature in reducing the number of nonlinear equations of motion by
关118,119兴. The occurrence of spurious solutions due to re- one, and in dealing with different regimes of nonlinear modal
constitution was shown in 关117兴 and is again addressed in a interactions which involve stretching 共axial兲, transverse 共in-
forthcoming paper aimed at resolving the relevant controver- plane兲, and/or swinging 共out-of-plane兲 components.
sies 关122兴. Complete and consistent reconstitution proce-
dures were discussed and compared in 关120兴. A procedure 4 NUMERICAL AND GEOMETRICAL ANALYSIS
working on the set of first-order differential equations in time „SINGLE-DOF MODEL…
共see Section 4.1兲 equivalent to the second-order set describ- The asymptotic analyses discussed in the previous section
ing the original mechanical system 共ie, state-space formula- allow us to detect approximate steady regular solutions of the
tion兲 has been proposed in 关126兴. equations for cable finite dynamics and to identify some rel-
Both incomplete and complete reconstituted modulation evant instability regions corresponding to the onset of other
equations have been obtained in cable dynamics, the former regular or possibly nonregular solutions. Major merits of
共the latter兲 being based on requiring vanishing of 共accounting those analyses consist in making available—at least for
for兲 the terms in the second-order problem depending on handable low-order models—closed-form relationships for
derivatives with respect to the ␧t scale. Most phenomena such basic phenomena as the frequency-response dependence
highlighted through discrete models are based on the use of or the time law of motion components. At the same time,
the former procedure. The latter has been used in 关51,53兴. In they give a general framework of regular nonlinear behavior
the former, a comparison of the effects of the two procedures in control space of problem parameters against which dis-
within the direct approach is also given, showing how the cussing and understanding the many complicated, regular,
homogeneous solutions must be included in the solution of and mostly quasiperiodic and chaotic, kinds of motion which
the first-order problem accounting for the 2:1 resonance in actually characterize the finite amplitude dynamics of sus-
order to make the complete reconstituted modulation equa- pended cables.
tions derivable from a Lagrangian 共as dictated by the inher- Nevertheless, for going in-depth into the description of
ent symmetric nature of the mechanical problem兲. However, dynamics, a decisive role is rather played by numerical and
this results in determining the coefficients governing this in- geometrical analyses to be accomplished using the many the-
teraction to within an arbitrary constant. This indicates a oretical and computational tools available within the context
mathematical drawback to be overcome either by choosing of dynamical systems theory. In turn, even for relatively
just the amplitudes of the homogeneous solutions which pro- simple systems, such involved analyses are actually appli-
duce vanishing of terms depending on ␧t in the second-order cable just in limited zones of the control parameter space
with the aim of detecting specific bifurcation mechanisms frequency stretching oscillations provide the cable with an
and/or nonlinear responses. In contrast, they do not allow for effective bending rigidity which can result in a stabilization
a general overview of the variable dynamic phenomena of its inverted position even in the absence of a cable trans-
which may occur for whatever combination of problem pa-
rameter values. lation speed 关137兴. However, initially prestressed single-dof
In the next sections, a summary of techniques and tools cable models never undergoing compression 关87兴 have
for bifurcation and chaos analysis, as applied in the literature mostly been studied. The corresponding nonlinear oscillator
to describe cable finite dynamics, will be given. Reference is has a negative (c 22 ⫺4c 3 ) quantity and, accordingly, a single
made to the single-dof cable model described by Eq. 共25兲. equilibrium point and corresponding one-well asymmetric
Actually, numerical techniques have been used for low-order potential.
multi-dof models, too, however, either considering them as This is the case of the nonlinear oscillator (c 2 ⫽35.952,
complementary tools for locally comparing and validating c 3 ⫽534.53) widely considered in the literature, which de-
results obtained through approximate asymptotic techniques, scribes a technical cable used in overhead transmission
or focusing on specific investigation purposes: see, eg, Cai lines—with axial rigidity-to-initial tension ratio (EA/H) and
and Chen 关89兴, who analyzed numerically the interaction of sag-to-span ratio (d/l) respectively equal to 500 and 1/45—
external and parametric resonances in the planar response of vibrating with the first symmetric mode under uniform forc-
a five-dof system. In contrast, systematic use of various com- ing. Due to the presence of both even and odd nonlinearities,
putational tools and of geometric interpretation has been already this simple single-dof model exhibits great richness
made for the single-dof model with the aim of understanding and variety of dynamic regimes with many fundamental
the variable aspects of its rich nonlinear dynamic behavior. characteristics of behavior of nonlinear dynamical systems,
as it will be discussed in Section II.3. Indeed, it has been
4.1 The Helmholtz-Duffing archetypal model for bifur- used to describe and understand bifurcation and chaos in
cation and chaos terms of: i兲 diagnosis, namely implementing and correctly
using measures for characterization of nonlinear dynamics
The Helmholtz/Duffing-type oscillator describing the single-
and chaos; ii兲 scenario, ie, obtaining response charts in con-
dof cable dynamics 共Eq. 25兲 can be considered a further—
trol parameter space and basins of attraction in initial condi-
simple but rich—archetypal model for bifurcation and chaos
tion space; and iii兲 prediction, ie, determining critical bounds
in mechanics 共Rega 关140兴兲 besides the classical models
of technical interest for response predictability and possible
widely considered in the literature 关141–143兴. Indeed, it de-
occurrence of bifurcations and chaos in regions of regular
scribes the planar finite dynamics of a large class of 1D
nonlinear response 关140,145兴.
structural/mechanical systems with initial curvature, which
Within this context, reference has been made to the
also includes shallow arches, initially buckled beams, and
autonomous system of first-order equations 共state-space
pinended two-bar systems 共see, eg, 关144兴兲. One key feature
formulation兲
of the model is the two-well asymmetric potential V(q) of
the associated Hamiltonian system, which has three equilib-
ẋ⫽y
rium points, q⫽0 and q⫽(⫺c 2 ⫾(c 22 ⫺4c 3 ) 1/2)/2c 3 , ie, a
nonlocal structure of the phase portrait, for generic values of
ẏ⫽⫺ 关 ␮ y⫹x⫹c 2 x 2 ⫹c 3 x 3 ⫺ P cos共 ␾ 兲兴 (43)
the coefficients. For the suspended cable 共Fig. 4兲, the first
point corresponds to the initial static equilibrium stable con-
figuration 共S兲 of the system close to the unloaded natural ␾˙ ⫽⍀
configuration. The other two, whether real, correspond to
two further equilibrium configurations, one unstable 共U兲— which Eq. 共25兲 can be transformed into through the positions
located just above the horizontal line through the supports— x⫽q, y⫽q̇, ␾ ⫽⍀t, as common in dynamical systems
and the other stable (S̄)—the upward one nearly symmetric theory. The search for relevant 共stable and unstable兲 steady
to the first configuration. Since the cable can only resist ten- solutions in the main regions of interest in control parameter
sile forces, it is not capable of maintaining its geometrical space has been accomplished through combined use of sev-
shape in the two configurations 共U兲 and (S̄), which are thus eral numerical techniques 关143,146兴. Namely: i兲 point-by-
not physically admissible. Actually, in the context of a non- point computer simulations with prescribed initial condi-
local analysis of multimode cable dynamics 共Section 3.4.2兲, tions; ii兲 algorithms for computing specific quantitative
Pilipchuk and Ibrahim 关139兴 have shown that the high- measures of nonlinear/chaotic dynamics, needed for a reli-
able characterization of the attractors; iii兲 continuation algo-
rithms giving the system discretized response under variation
of a primary control parameter; iv兲 cell mapping procedures
showing basins of attraction of coexisting solutions; and v兲
numerical integrations aimed at obtaining global invariant
manifolds of saddle points. Use of these different tools is
essential to both the search for and the analysis of the evo-
lution of dominant attractors of Eqs. 共43兲, as well as to the
deep understanding of various topological phenomena which
Fig. 4 Cable potentials a兲 and static equilibrium configurations b兲 characterize system global dynamics.
4.2 Computer simulations for characterization of algorithm based on the simultaneous integration of the third-
nonlinearÕchaotic dynamics order system 共43兲 and of the corresponding linearized equa-
Direct integrations of nonlinear dynamic equations are usu- tions 关152兴. Attention must be paid to the selection of both
ally accomplished to validate results of regular response ob- the step of reorthonormalization of the vector identifying a
tained through analytical techniques. In addition, they are nearby trajectory with respect to a reference trajectory, and
used to get primary hints about possible chaotic responses, the number of forcing periods to be used for obtaining a
though more elegant predictions for chaos are possibly based reliable characterization of the response. In 关150兴, the former
on theoretical approaches and specific dynamic features of was given a value of one, though just values no greater than
the system 关147,148兴. ten were suggested in the literature for other systems 关153兴,
More generally, computer simulations 共based, eg, on a while the latter was shown to depend considerably on the
standard fourth-order Runge-Kutta algorithm兲 allow us, on level of chaoticity. On the other hand, the correlation dimen-
one side, to directly obtain local attractors corresponding to sion of the attractor in the Poincaré map 关154兴:
given sets of initial conditions 共ic兲. These are then quantita-
log C 共 r 兲
tively characterized through specific measures of nonlinear D c ⫽ lim (45)
dynamic response whose calculation relies just on those nu- r→0 log r
merical results. On the other side, if performed systemati-
cally, computer simulations allow us to obtain overall re- r being the radius of a 2D box centered at each point xi
sponse charts in control space and specific bifurcation ⫽ 兵 q i ,q̇ i 其 of the set and C(r) the correlation function
diagrams associated with variations of control parameters. N N
1
4.2.1 Measures for diagnosis of response C共 r 兲⫽ 兺 j⫽1
兺 H 共 r⫺ 兩 xi ⫺xj 兩兲
N 2 i⫽1
(46)
To identify the type of system response with reasonable ac-
curacy, various qualitative and quantitative measures 共see, is calculated by counting the number N of points in each box
eg, 关142,143兴兲 were used in Benedettini and Rega 关149兴 and 共with H(s)⫽1 if s⭓0 and H(s)⫽0 if s⬍0). For a reliable
Rega et al 关150兴 whenever there was a doubt of chaotic be- calculation, we must pay attention to the size of the boxes
havior. As a rule, first diagnosis of a chaotic motion is made with which to cover the attractor. Indeed, the dimension has
through qualitative measures such as phase plane portraits, to be properly correlated with the density of points in the
Poincaré maps, and frequency-power spectra, with special map, since it must be calculated by considering only the r
attention payed to the latter which furnish reliable quantita- domain where the log C(r) versus log r ratio is linear
tive indications for chaos, too. Nevertheless, precise charac- 关149,155兴.
terization of a persistent chaotic response is based on the Generally speaking, careful computation of dynamical
calculation of further global quantitative measures, such as measures is demanded by the need to obtain consistent indi-
Lyapunov exponents and dimension of the attractor. cations through a number of different measures before mak-
In the numerical analysis of potentially chaotic systems, it ing a reliable statement of ‘‘chaoticity’’ of the response. As a
is essential to distinguish between their intrinsic sensitivity result of the investigation made for the single-dof cable
to variations of ic and spurious results produced by the im- model in the 1/2-subharmonic, 1/3-subharmonic, and super-
proper selection of computational quantities. Accordingly, harmonic resonance regions 关150兴, notably variable shapes
considerable care has to be paid to the choice of both general of chaotic strange attractor were observed. Meaningful topo-
parameters of the computer simulations and algorithmic pa- logical differences are concerned with both the ‘‘thickness’’
rameters playing a role in the calculation of specific mea- of the attractors and their single- or multi-piece structure, the
sures 关149兴. Indeed, obtaining reliable results requires cali- latter highlighting the presence of a dominant periodic solu-
bration of the length of the response transient, of the total tion which the motion fluctuates around. These variable fea-
integration length, and of the time step increment in order to tures also entail different levels of chaoticity 共see Section
be reasonably sure that the response obtained is really a II.3.2兲.
steady one and that chaos, whether occurring, actually per-
4.2.2 Response scenarios
tains to the ODE of motion and not to a difference approxi-
Response scenarios obtained through point-by-point com-
mation of it 关151兴. Useful hints for a proper selection of
puter simulations include response charts in control space,
these general parameters are given just by the requirement to
basins of attraction in initial conditions space, and bifurca-
obtain well stabilized values of some global measures of the
tion diagrams associated with variations of specific control
dynamics. In turn, calculation of each of these requires a
parameters.
specific algorithmic care.
Response charts giving a general overview of regions of
The spectrum of Lyapunov exponents:
periodic/chaotic responses exhibited by the cable model in
1 d i共 t 兲 the excitation parameter space ⍀, P, with fixed values of
␭ i ⫽ lim log2 (44) damping and ic, were built in 关150兴. Meaningful regions of
t→⬁ t d0
chaotic response were identified and analyzed against the
where d 0 and d i (t) are the radius of an infinitesimal sphere background of the nonlinear phenomena obtained through
of ic and the corresponding principal axes of an ensuing approximate asymptotic solutions 共Section 3兲, in particular
ellipsoid, respectively, is calculated by a nowadays classical of the most important zones of external resonance. The pres-
ence of both quadratic and cubic nonlinearities produces a 4.3.1 Predicting local bifurcations and possible onset of
notably varied and involved overall response picture. Re- chaos
gions of different periodic or chaotic response are obtained Point-by-point computer simulations are an essential step for
in the three main resonant frequency ranges—1/2- detecting bifurcation and chaos. Nevertheless, there is also
subharmonic, 1/3-subharmonic, and superharmonic—into strong interest in a priori evaluations of possible nonregular/
which the excitation parameter space can be subdivided 共see chaotic regimes. From a theoretical point of view, the interest
Section II.2.2.2兲. is mostly in formulating ‘‘exact’’ criteria for loss of regular
Basins 共or domains兲 of attraction in the ic space, which behavior and onset of persistent chaos. Unfortunately, such
are also a very special kind of response chart, were system- an objective is often hard to obtain, so practically all of the
atically built in Benedettini et al 关156兴, Rega et al 关157兴, and developed criteria 共关142兴, see also Rega 关145兴兲 lead just to
Rega and Salvatori 关158兴, in various resonant regions, using necessary conditions for chaos. The most firmly founded cri-
more advanced—while numerically simple—techniques than terion for single-dof systems—the Melnikov criterion based
‘‘brute’’ computer simulations. on bifurcation of the homo/heteroclinic orbits—gives just
lower forcing bounds for onset of 共transient兲 chaos. 共Other
All of these charts were aimed at i兲 clarifying overall
criteria occur for multi-dof systems, eg, the Shilnikov crite-
aspects of system dynamic behavior, ii兲 verifying robustness
rion: see 关59,148兴 and Section II.5.2.2.兲 Actually, such a cir-
and coexistence of attractors, and iii兲 analyzing response
cumstance is not a great drawback from the engineering
sensitivity—and the associated fractal features—not only in
point of view, since the simple assessment of response
ic space but also in mechanical parameter space. unpredictability—or the onset of a dynamic regime different
from the reference one—is often more important than its
precise characterization as chaotic. Indeed, in technical ap-
4.3 Local bifurcation analysis plications, it can be sufficient to reliably predict the condi-
tions leading a system designed to work in a regular regime
Local bifurcation diagrams were first built based on the re- to undergo also irregular/chaotic—and possibly just
sults of systematic computer simulations accomplished in transient—dynamics, which have to be controlled, quenched
specific zones of control parameter space with the aim of or even exploited.
identifying mechanisms of transition between regular and Within this context, it is already worth reliably predicting
complex responses 关150兴. disappearance of steady periodic solutions, and onset of non-
The periodic or chaotic nature of the steady solutions was stationary 共modulated兲 solutions. Predictive criteria associ-
identified based on the analysis either of the corresponding ated with stability analysis of regular nonlinear oscillations
Poincaré maps and power spectra, or of the associated values are of special interest in this respect, since they allow us to
of the nonzero Lyapunov exponents, namely the first two recover the huge patrimony of the relevant knowledge re-
exponents for the third-order system 共43兲, the third one along quired for understanding complex dynamic phenomena. It is
the flow being zero. The spectrum of Lyapunov exponents thus possible to establish relationships between irregular so-
obtained by varying a control parameter proves to be a pow- lutions and dominant regular solutions in various resonance
erful tool for obtaining bifurcation diagrams 关159,160兴 and, zones, and to locate regions of possible chaotic response ver-
more generally, for highlighting the variable pattern of a cha- sus the well-known regions of primary/secondary resonant
otic region in control space 关149兴. Nevertheless, the algorith- solutions and their stability. The goals are in predicting bi-
mic aspects make construction of bifurcation diagrams furcations with meaning of possible precursors to chaos, and
through Lyapunov exponents more time consuming than in bounding limited regions in control parameter space
through simple observation of the Poincaré map. Moreover, where searching for transient or persistent chaos through
the former measure simply allows us to identify chaotic variable and accurate—but localized—computational tools.
zones as those corresponding to positive values of the first Simple search for approximate steady solutions through
one of the available analytical methods 共classical perturba-
exponent, and transition between different periodic solutions
tions, averaging, harmonic balance兲 and analysis of their sta-
共eg, through a period doubling bifurcation兲 as occurring
bility are already useful to the purpose. Being the onset of
when the first negative exponent closely approaches the zero
chaos in the single-dof cable model associated with large
value. In contrast, the latter measure shows to be more com- nonlinearities, Benedettini et al 关156兴 employed the har-
plete, for allowing us to precisely identify the response peri- monic balance method—also used by Takahashi and Konishi
odicity and effectively detect the route from periodic to cha- 关46,47兴 for a three-dof model, and in 关161兴—since it does not
otic motions. involve small parameter assumptions. Low- or high-order
Construction of bifurcation diagrams in the 1/2- approximations of the method can be developed 共see, eg,
subharmonic, 1/3-subharmonic, and superharmonic resonant 关162,163兴兲, differing from each other as to the number of
frequency ranges led to the observation of smooth as well as harmonic terms to be considered in order to obtain accurate
sharp 共or sudden change兲 mechanisms of transition from pe- solutions and stability analysis. In principle, the latter are
riodicity to chaos. They include complete and incomplete able to furnish better results in the study of complex systems
period-doubling cascades, period multiplications, jumping, or phenomena, mostly as far as reliable stability evaluations
and intermittency bifurcations 共see Section II.3.1兲. are concerned. Recently, Cardona et al 关164兴 have tested on a
nearly taut cable a multiharmonic method based on a system- stability 共the usual hysteresis branch兲 and second-order insta-
atic use of the FFT alogorithm, and Yu and Xu 关165兴 have bility 共along nominally stable branches, see, eg, 关168兴兲 of
applied the harmonic balance within a general formulation of each frequency-response curve. Hints were obtained about
the nonlinear dynamics of cable-dampler systems. Zheng the need to construct higher subharmonic approximations of
et al 关37兴 have used a multiharmonic approximation, com- the solution aimed at highlighting the possibility of further
bined with a finite element approach, to deal with different period doubling bifurcations. However, successive predic-
kinds of 共superharmonic, subharmonic, and internal兲 reso- tions were not deemed necessary given the quite small re-
nances in the suspended cable. Yet, in the context of analyti- gions in which bifurcations occur, while choosing to resort
cally based criteria, the kind of considered harmonic directly to computer simulations to investigate the nature of
terms—to be properly chosen depending on the investigated motions established in the unstable zones.
regular solutions—appears even more important. The obtained results show the bifurcation predictive ca-
Benedettini et al 关156兴 examined the bifurcation predic- pability of stability analysis of simple approximate periodic
tive capability of low-order approximations properly selected solutions. In this framework, irregular motions of a system
in various resonance regions for the cable. Therein, the in- can be detected: i兲 by obtaining stability boundaries of domi-
terest was in identifying zones of response unpredictability nant approximate regular solutions—even of low order—in
rather than strictly chaotic motions, and the use of too rich meaningful ranges of control parameter values; ii兲 by look-
approximations seemed to be inconsistent with the simple ing for possible chaos within unstable regions through vari-
single-dof model referred to. Solution to Eq. 共25兲 was sought ous dynamic measures based on localized point-by-point and
through the following expansions: cell mapping computer simulations.
q 共 t 兲 ⫽a 0 ⫹a 1 cos共 ⍀t⫹ ␪ 兲 4.3.2 Bifurcation analysis through continuation
q 共 t 兲 ⫽a 0 ⫹a 1 cos共 ⍀t⫹ ␪ 兲 ⫹a 1/2 cos共 ⍀t/2⫹ ␸ 兲 Within a more modern context, a complete scenario of re-
sponse curves of both stable and unstable periodic solutions
q 共 t 兲 ⫽a 0 ⫹a 1 cos共 ⍀t⫹ ␪ 兲 ⫹a 1/3 cos共 ⍀t/3⫹ ␸ 兲 (47) is obtained with a path follower 共or continuation兲 algorithm
q 共 t 兲 ⫽a 0 ⫹a 1 cos共 ⍀t⫹ ␪ 兲 ⫹a 2 cos共 2⍀t⫹ ␸ 兲 giving the fixed points of the Poincaré map as one control
parameter is varied 共see, eg, 关143,169–171兴兲. It is based on
⫹a 3 cos共 3⍀t⫹ ␺ 兲 the description of the flow 共Eq. 共25兲兲 in terms of discrete
all of them being asymmetric, consistent with the asymmetry dynamical systems theory 共see, eg, Guckenheimer and
of the ODE. This is of great concern as regards possible Holmes 关147兴 and Wiggins 关148兴兲. Accordingly, the fixed
development of chaos. Indeed, according to Floquet theory point p̄(x,ẋ) of the mapping f: R2 →R2 of the phase plane
关143,166,167兴, both jump 共or saddle-node兲 and period dou- (x,ẋ) into itself, under prescribed initial conditions (x 0 ,ẋ 0 )
bling bifurcations can occur for an asymmetric solution at t 0 , is equivalent to a T⫽2 ␲ /⍀ periodic solution x
when varying a control parameter, the latter being the start- ⫽x(t;x 0 ,ẋ 0 ), ẋ⫽ẋ(t;x 0 ,ẋ 0 ) to the set of first-order Eqs.
ing point of a possibly complete sequence of bifurcations to 共43兲 equivalent to Eq. 共25兲.
chaos. From this viewpoint, it is worth noticing the differ- An mT⫽2m ␲ /⍀ periodic solution of system 共43兲共de-
ence with respect to a symmetric solution typical for noted as P-m solution兲 is represented by an m-periodic point
Duffing-type oscillators, which must undergo a symmetry of f, which is also a fixed point of the mth iterate ( f m ) of f to
breaking prior to the possibility of a period doubling be calculated by solving the equation p̄⫽ f m (p̄) with a
bifurcation. Newton-Raphson procedure. A linearized stability analysis
The simplest period T expansion 共47a兲 containing just the with the Floquet theory made at any fixed point p̄⫽(x,ẋ) of
drift and the fundamental harmonic was used to describe the the mapping f, or its mth iterate, allows us to detect the local
reference dominant solution. In turn, the simplest period 2T bifurcations of a path of steady periodic solutions. The type
and 3T expansions 共47b兲 and 共47c兲 were used in the neigh- of fixed point is determined through calculation of the eigen-
borhood of the 1/2- and 1/3-subharmonic resonances, respec- values, (␭ 1 ,␭ 2 ), of the Jacobian of f m . It can be either stable
tively 共for the former, see also 关88兴兲. The period T expansion 共sink, S兲 or unstable 共saddle兲 depending on whether both
improved by the presence of order-2 and order-3 harmonics eigenvalues, or only one of them, stay inside the unit circle
共Eq. 共47d兲, see also 关47兴兲 was used in the range of superhar- in the complex plane. By varying a control parameter, di-
monic resonances. rectly 共D-saddle cycle兲 or inversely 共I-saddle cycle兲 unstable
A set of algebraic nonlinear equations was obtained for periodic solutions are established through saddle-node (SN)
each of the four distinct solutions and was then solved nu- or period-doubling ( PD) bifurcation, respectively, when an
merically for the unknown amplitudes and phases to obtain eigenvalue becomes larger than 1 or smaller than ⫺1. They
the relevant frequency-response resonance curves 共see Fig. correspond to first- and second-order instability of the re-
II.7兲. Local stability analysis of each steady-state periodic sponse curves. The notations S mj , D mj , and I mj will be used in
solution was carried out repeatedly with varying frequency Section II.3.3 to denote P-m sinks, direct saddles and in-
by considering a small disturbance ␦ q(t) and by calculating verse saddles, respectively, the subscript j⫽1,2, . . . ,m refer-
the eigenvalues 共Floquet multipliers兲 of the monodromy ma- ring to different image points of the P-m solution under the
trix associated with the linearized Hill’s type variational mapping f m .
equation. It allowed the authors to detect both first-order in- A path follower algorithm was used to obtain the
frequency-response curves 共in terms of maximum response mapping, with initial conditions started along the ingoing
amplitude兲 of various periodic solutions playing meaningful 共outgoing兲 eigenvector of the relevant Jacobian. They are
role in the 1/2-subharmonic 关157兴 and 1/3-subharmonic referred to as W sk (D mj ), W sk (I mj ) (W uk (D mj ), and W uk (I mj )),
关158兴 resonance regions. For each frequency value, several with the subscript k⫽1, 2 denoting the two branches of the
executions of the algorithm with different number of itera- stable 共unstable兲 manifold of the direct or inverse saddle cor-
tions of the mapping and/or triggering values 共ic兲 of the responding to the P-m unstable solution and having j
curves were made to identify both the main coexisting solu- ⫽1,2, . . . m images in the Poincaré map.
tions with different period, and the distinct branches of a When obtained systematically with a varying control
possibly multivalued response of given periodicity. The val- parameter, stable and unstable manifolds of saddles allow
ues obtained for the coexisting stable and/or unstable fixed for a deep geometrical understanding of numerically
points through the path follower algorithm were also used as observed attractor and basin metamorphoses. Indeed, the
different initial conditions in the direct integrations made to main global bifurcations are associated with a homoclinic
obtain rich and complex pictures of periodic and chaotic re- 共heteroclinic兲 tangency between the stable and unstable
sponse in numerical bifurcation diagrams. Paths of the sink manifolds of a given D or I saddle 共of two different D
S mj corresponding to a P-m stable solution are sometimes to and/or I saddles兲. Conditions like, eg, W s (D)艚W u (D)
be distinguished also with respect to coexisting, resonant and ⫽0 (W s (D m )艚W u (D n )⫽0) are first satisfied at a ho-
nonresonant, branches of the frequency-response diagram moclinic 共heteroclinic, involving two direct saddles of differ-
共see the l S 2j and u S 2j branches in Fig. II.18兲. The paths of ent, m and n, periodicity兲 tangency. The closures
direct (D mj ) and inverse (I mj ) saddles corresponding to P-m (W s (D mj ),W s (I mj )) of the stable manifold of a direct saddle
unstable solutions are usually reported in the diagrams, too, and of an inverse saddle coincide with the border between
since they are of notable help for understanding of the two basins of attraction of different solutions and, respec-
geometrical mechanisms through which global bifurcations tively, between two distinct subdomains identifiable in the
occur. basin of a given solution under a mapping f 2m . When vary-
ing a control parameter, occurrence of manifold tangling is a
4.4 Global analysis of response sufficient condition for fractal basin boundary 共and ensuing
unpredictability of response兲 and a necessary condition for
Cell to cell mapping algorithms 关172兴 like the one based on
onset of a chaotic attractor in one basin. Whether existing,
the simple technique developed by Foale and Thompson
关169兴 are used to determine, for a given grid of starting con- this attractor is contained within the closure (W uk (D mj )) of
ditions, both fixed points and their basins of attraction in the the kth branch of the unstable manifold of the direct saddle
Poincaré plane. The periodic points located inside the basins located on the relevant basin boundary. Moreover, it is con-
are sinks 共S兲 whereas those on the basin boundaries are direct tained within the closures (W u (I mj )) of the unstable mani-
saddles 共D兲. An inverse saddle I m ensuing from an S m sink folds of all of the inverse saddles from which it was likely to
through PD bifurcation, S m →I m ⫹S 2m , is located inside the originate through successive PD bifurcations. The sequence
domain of attraction of the P-2m solution and on the bound- of homoclinic/heteroclinic tangencies and the intersections
ary separating the two subdomains of different P-1 solutions of the saddle manifolds corresponding to coexisting unstable
that can be identified by applying the algorithm under a map- periodic solutions govern the sequence of attractor-basin bi-
ping f 2m . When there is a suspicion that a chaotic attractor furcations and determine the variable occurrence of sudden
exists in some basin, long term 共thousands of forcing peri- change bifurcations and crises 共Grebogi et al 关175兴兲.
ods兲 computer simulations with starting conditions therein
4.4.2 Attractor-basin-manifold phase portraits and global
are made to check about its actual nature. Chaotic attractors
bifurcational mechanisms
made of a finite number of disconnected pieces are obtained
Within the global analysis of sytem dynamics, combining i兲
in several situations. They are denoted as C nj attractors, the
attractors resulting from numerical integrations, ii兲 basin por-
subscript j⫽1,2, . . . ,n referring to the different pieces con-
traits obtained with cell-to-cell mapping, and iii兲 invariant
tained in subdomains of the whole basin.
manifold descriptions, leads to construction of powerful and
4.4.1 Global invariant manifolds of saddles synthetic attractor-basin-manifold phase portraits 共AB-
Getting a deep understanding of the attractor global structure MPPs兲.
and of the attractor and basin bifurcations requires examin- For the cable single-dof model, they have been discussed
ing the numerical results obtained with path following, cell in-depth in the 1/2-subharmonic 关157兴 and 1/3-subharmonic
mapping, and computer simulation algorithms in the light of 关158兴 resonance regions, greatly contributing to understand-
a geometrical description of system dynamics ing of the relevant bifurcation structures 共see Section II.3.3兲.
关147,148,173,174兴. It is based on the construction of the glo- Indeed, performing both local and global analysis, and com-
bal invariant manifolds of the saddle points of the mappings, bining bifurcation analyses of attractor-basin structure and of
corresponding to various unstable periodic solutions. They manifold structure, showed to be a powerful mixing of tools
play a fundamental role in basin organization. The global for deep insight into complex bifurcation behavior. Among
stable 共unstable兲 invariant manifold of a saddle is obtained the goals reached with those analyses we mention:
through backward 共forward兲 time numerical integrations • understanding mechanisms of onset, evolution, and disap-
from either a direct (D mj ) or an inverse (I mj ) saddle of a pearance of regular attractors
• describing main and secondary evolutions to chaos peting multimode solutions, comprehensive local bifurcation
• identifying sudden bifurcational events and crises, and analyses through continuation are already considerably in-
characterizing them through dynamical systems signatures volved 关53兴. But it is mostly the global analysis of the AB-
• highlighting features of basin metamorphosis and erosion, MPP evolution which becomes practically unaffordable from
of attractor-basin boundary accessibility/unaccessibility, a computational point of view and unreadable from a geo-
and of topological structure attractivity metrical point of view, due to the enormous amount of data
• analyzing regular windows occurrence to be processed, stored, and graphically represented in a mul-
• highlighting the role played by saddles of high periodicity tidimensional state space. Tentative positive answers about
in the evolution of system global dynamics the possibility of describing bifurcational and topological be-
havior of multi-dof systems are likely to rely just on the use
In particular, a nearly complete scenario of sudden-type of powerful parallel computing procedures and of suitable
mechanisms of global bifurcation of chaotic attractors was geometrical representations. Meaningful proposals and expe-
detected. They are usually associated with collision of attrac- riences have been made in this respect 关180兴, but nothing
tors with unstable fixed points, and depend on topological really decisive is yet available, not only as regards multi-dof
behavior of invariant manifolds of coexisting direct and/or cable models.
inverse saddles. They include: i兲 boundary crisis, which en-
tails sudden appearance 共destruction兲 of a chaotic attractor;
ii兲 interior crisis of switching-type, entailing merging 共split- 5 EXPERIMENTAL ANALYSIS
ting兲 of bands of an existing chaotic attractor and its sudden 5.1 Experimental nonlinear dynamics of flexible con-
enlargement 共reduction兲 in size; iii兲 interior crisis of tinuous systems
bursting-type, entailing sharp widening 共shrinking兲 of a nar-
Analyzing in-depth nonlinear dynamic regimes means: i兲
row 共large兲 chaotic attractor; iv兲 subduction, with the associ-
finding all of the main solutions attained by the system when
ated sudden metamorphosis from a periodic 共chaotic兲 to a
varying a meaningful control parameter, ii兲 studying the rel-
chaotic 共periodic兲 attractor in one given basin.
evant stability, iii兲 constructing basins of attraction, iv兲 un-
Even deeper understanding of sudden enlargements of a
derstanding bifurcation mechanisms between contiguous
chaotic attractor requires their characterization in terms of
and/or competing solutions, and finally v兲 getting a synthetic
chaotic saddle. At a certain parameter value, this is the non-
description of system global dynamics via consistent mea-
attractive topological set containing the infinite direct and
sures of possible strangeness and chaoticity. This pattern can
inverse saddles ensuing from formerly stable periodic solu-
be pursued with relative ease for very simple theoretical
tions and delimiting an inaccessible area of the phase plane
models with one 共or very few兲 dof, as discussed in Section 4.
with the closure of its stable manifold 关176兴. A process of
Instead, even given the recent or expected advances in par-
nearly simultaneous tangencies of a narrow chaotic attractor allel computation, it is a hard or nearly impossible task for
with all of the stable manifolds of the saddles composing the multi-dof or infinite-dimensional models. Their direct simu-
nonattractive set entails sudden enlargement of the attractor, lation using, eg, multibody or finite element codes are virtu-
which englobes all of the saddles existing in its basin of ally incapable of answering certain fundamental dynamical
attraction and accessible to it. A partial numerical description questions which may even be obscured by excessive detail in
of such phenomenon was achieved for the single-dof cable the ‘‘exact’’ simulation. This is one main reason for the ex-
model in a case involving a single-piece chaotic attractor and tensive search for minimal models preserving most of the
an identifiable low-periodicity saddle initiating its enlarge- main features of nonlinear dynamics of the actual system. In
ment 共关177兴; see Fig. II.24兲. The chaotic saddle contributing such cases, experimental tests on even small-scale models
to the structure of the widened attractor was determined play an essential role to the aim of getting a reliable, possibly
through complex and time-consuming computations accom- low-cost, understanding of system dynamics as well as hints
plished by means of the PIM-triple algorithm 关178兴 imple- for refinement of its analytical modeling.
mented for the flow 共Eqs. 43兲. Generally speaking, experimental analysis of nonlinear
In principle, numerical and geometrical analysis of multi- dynamic behavior of structural elements is important both
dof cable dynamics can be accomplished along the same for verification of theoretical predictions, and for detection of
lines as discussed for the single-dof model. According to the complex or new dynamic phenomena associated with system
theory of averaging 共关106兴, see also, eg, Johnson and Bajaj actual nonlinearities but unmodelled in the theoretical analy-
关179兴兲, quasiperiodic or chaotic 共namely, periodically or cha- sis. In particular, characterization of a possibly complex sys-
otically amplitude modulated兲 solutions of the ODEs 共27兲 tem response has to be done in terms of both time and space
correspond to nonstationary solutions of the APMEs 共37兲. features. From this viewpoint, experimental tests on highly
They are constructed through numerical integrations of the flexible systems like strings and suspended cables pose spe-
latter upon identifying Hopf bifurcations of the relevant cial measurement problems, since contact tools 共eg, acceler-
steady solutions which correspond to the onset of limit ometers兲 cannot be used because their masses may signifi-
cycles in amplitude and phase. Alternatively, quasiperiodic cantly influence the system dynamics. In contrast, no-contact
and chaotic solutions are obtained through direct computer measurements require rather expensive optical follower sys-
simulations of the ODEs 共27兲. tems 共eg, displacement probes兲 or laser equipment often re-
Nevertheless, owing to the much richer scenario of com- sulting in a very low 共1 or 2兲 number of synchronous mea-
sures. This small number is not so limiting in regular chaotic attractors and reconstructing dynamic properties and
regimes, where using nonsimultaneous measures still allows spatial features of the response from the analysis of experi-
us to reconstruct the actual phase space with the wanted pre- mental data.
cision since they refer to a common 共excitation兲 phase. In-
stead, acquiring nonsimultaneous time series makes no sense 5.2 Mechanical model and experimental apparatus
in the case of a seemingly chaotic regime, because of the
Designing an experimental apparatus for extensive analysis
random-like continuous variation of both amplitude and
of finite dynamics of a suspended cable entails i兲 realizing a
phase at different acquisition points: this entails using proper
suitable mechanical model, ii兲 choosing a suitable procedure
procedures for reconstructing the global time/space proper- to measure large displacements, and iii兲 dealing with a con-
ties of the dynamics when starting from a single 共or few兲 siderable response sensitivity to environmental/operating
time series. conditions due to system lightweight and low-damping char-
Thus, for relatively stiff structural systems like beams and acteristics.
arches, a considerable number of experimental papers were
devoted to general/specific aspects of nonlinear dynamic re- 5.2.1 Mechanical models
sponse 共eg, 关181–183兴兲. In contrast, relatively few experi- The mechanical model has to be designed according to
mental studies have addressed the finite dynamics of flexible whether the interest is in investigating a specific cable under
systems. Investigations on nonregular vibrations of strings well identified excitation conditions, or different cables. Let
us keep in mind the expression of the elastogeometrical pa-
include the papers by Molteno and Tufillaro 关184兴, O’Reilly
rameter ␭ 2 ⫽(EA/mgl)(8d/l) 3 共Eq. 共23兲兲 and the shape
and Holmes 关185兴, and Nayfeh et al 关186兴.
modification of planar symmetric modes 共with an increasing
Starting from the earlier work by Yamaguchi et al 关187兴
odd number of half waves on the cable span兲 occurring at
who observed coupled vibrations between in-plane and out-
crossover points when ␭ 2 is increased 共see Fig. 2兲. We re-
of-plane modes under sinusoidally time-varying loads, recent mind that each crossover marks a sort of transition from
experimental works on nonlinear dynamics of suspended dynamic features of ‘‘elastic 共flexible兲 and taut’’ cable to fea-
cables have been concerned with the following main topics. tures of ‘‘inextensible 共stiff兲 and sagged’’ cable 共see also
i兲 Analysis of modal coupling in the nonlinear response of a 关83兴兲, both kinds of cable being actually employed in such
continuous model of hanging cable subjected either to para- applications as overhead transmission lines.
metric excitation at one support 共Perkins 关58兴兲 or to external When investigating variable dynamic behaviors of a
harmonic excitation applied at a specific cable point in the small-scale model around first crossover by simply varying
normal direction 共Lee and Perkins 关188兴兲. ii兲 Active stiffness its sag-to-span ratio—which is easily implemented from a
or active sag-induced force control of in-plane vibration of practical viewpoint through an adjustable anchorage—we
an inclined cable by axial support motion 共Fujino et al have to properly balance material elastic properties, cross-
关189兴兲, Fujino and Susumpow 关190兴兲. iii兲 Planar and nonpla- sectional area, and mass per unit length of the model. Indeed,
nar complex responses of a parametrically excited string car- we have to realize an acceptable axial rigidity-to-weight ratio
rying two concentrated masses, to be considered as a first EA/mgl, whose high values could give rise to too ‘‘stiff’’
approximate discrete model of a sagged cable 共Benedettini cables. At the same time, undesidered flexural stiffnesses
and Moon 关191兴兲. iv兲 Analysis of local and overall nonlinear must be avoided. These points suggested using a sand-filled
response of a hanging cable/mass under different harmonic surgical tubing 关197兴, a woven nylon cord 关188兴, different
motions of supports and external/internal resonance condi- kinds of steel wire 关190,198, 98, 196兴, or a 共fishing-like兲
tions 共Rega et al 关192兴, Benedettini and Rega 关193兴兲. Very nylon wire 关192,199兴, possibly with additional lumped
recent investigations have been concerned with in-depth masses at regular intervals aimed at increasing the weight-
characterization of various experimental scenarios to chaos to-tension ratio towards values typically existing in a bare
cable.
exhibited by the same cable/mass system 共Alaggio and Rega
A two-mass string model intended as a first approximation
关194兴, Rega and Alaggio 关195兴兲, and with experimental veri-
of sagged cable was experimented in 关191兴. Cable models
fication of numerical dynamic tensions in a rubber cable
with a larger number of equally spaced concentrated masses
关28兴. Experimental analyses of the nonlinear vibrations of a
and one adjustable support were experimented by Fujino and
nearly taut inclined cable owed to a concentrated force 关99兴 Susumpow 关190兴 and by Rega et al 关192兴, in the context
or associated with supports motion 关104,196兴 have also been of the active control of cable nonlinear vibrations and of a
performed in the background of the dynamic behaviour of systematic investigation of nonlinear/chaotic responses,
cable-stayed bridges. respectively.
Of course, with respect to the constrained modal shapes of Of coarse, attaching masses to the cable—even possibly
discrete theoretical models, experimental models have the well-centered and equally spaced—can induce ‘‘spurious’’
great advantage of being capable to account for highly vari- torsional modes or undesidered coupling between in-plane
able modal contributions to the system response. and out-of-plane motions. Moreover, the dynamic behavior
In the next sections, we first discuss some main experi- of the ensuing discrete model is expected to be rather close
mental setups for cable nonlinear dynamics. Then, we ad- to that of a sagged cable/mass suspension, which can exhibit
dress the main problems encountered in detecting regular/ further peculiar dynamic phenomena depending on the dis-
tribution of masses and the considered modes 共Section 2.3, Table 5. Experimental, theoretical and finite element frequencies of
cableÕmass
see 关78兴兲. Nevertheless, in the case of a symmetric array of
equal masses and considering only lower modes, the se- Mode
quence of the first natural frequencies and mode shapes re- ␻ 共Hz兲 H1 V2 H2 V1 H3 V3 V4 H4 H5 V5
mains practically unchanged with respect to the bare cable. Exp. 3.49 6.68 6.77 8.82 9.97 11.80 12.95 12.99 15.72 15.87
The main difference consists in a shift of crossover points Theor. 3.48 6.85 6.85 8.74 10.02 11.38 12.89 12.89 15.36 15.47
Num. 3.48 6.63 6.70 8.80 9.96 11.65 12.95 13.04 15.56 15.88
towards lower values of ␭ 2 .
Noncontacting two-axis optical follower cameras allow us
to avoid feedback in the system from the measurement de-
vice and to cover a fairly large and properly enlightened experimental setups share some typical features. A schematic
sensing area. They are currently used to measure finite am- layout of the experimental apparatus used by Rega et al
plitude displacements in two orthogonal directions 共vertical, 关192兴 is shown in Fig. 5. The cable-mass model is hung
or along the normal, in-plane; horizontal, ie, along the binor- between two points which are given a vertical, sinusoidal
mal, out-of-plane兲 either of the cross section 共in the bare motion by means of two shakers driven by a function gen-
cable兲 or of a mass 共in the cable-mass兲, at possibly different erator. A biaxial optical camera, whose control circuit gives a
locations. signal proportional to two 共vertical and horizontal兲 displace-
The main natural frequencies of experimental models are ment components of motion of one targeted mass—or, alter-
usually determined by means of a swept-sine spectral analy- natively, to the displacement and velocity of one
sis. After roughly identifying a particular mode, a narrow- component—is used. The output signals are real-time ana-
banded sinusoidal sweep around each frequency allows us to lyzed simultaneously by an FFT analyzer and a digital stor-
locate it more accurately—during the decaying free motion age oscilloscope used also as a waveform recorder. Motion
which follows turning off the excitation—at the lowest re- recordings are directly acquired on a PC and postprocessed
sidual amplitudes corresponding to an acceptable signal-to- to give quantitative measures of system response regarding
noise ratio. Pugsley 关8兴 was among the few to obtain earlier both its time evolution and the attractor structure. A control
experimental frequencies of cables with small to large 共up to on the actual imposed motion, used as a feedback control, is
1/4兲 sag-to-span ratio. Recently, extended frequency mea- obtained by measuring the displacement of the shakers head
surements of horizontal/inclined bare cables and cable-mass with an LVDT.
suspensions have been given in 关83兴 in a large ␭ 2 -range. One single mass was the target of the follower camera in
Other independent techniques have to be used to estimate the the overall analysis of nonlinear response made in 关192,193兴.
natural frequencies of cables exhibiting closely spaced fre- Of course, in the steady regime, the optical device was prop-
quency ratios, which highlight possible simultaneous internal erly switched between masses symmetrically located on the
resonances involving planar and nonplanar, symmetric and cable in order to distinguish between symmetric and anti-
antisymmetric, modes. This typically occurs, eg, at first symmetric modes which are both present in the in-plane/out-
crossover. These techniques include exciting the cable with a of-plane motion. Instead, two cameras movable along the
low amplitude random noise, and locating the frequencies cable were used in Alaggio and Rega 关194兴 and Rega and
through the analysis of the relevant transfer function; or even Alaggio 关195兴 for in-depth characterization of nonregular re-
applying manual excitations proportional to various modal sponse and bifurcational scenarios to chaos. They are needed
shapes, and locating again the relevant frequencies during to simultaneously acquire time laws of variable pairs of
the decaying free motions. Moreover, as a rule, a filtering masses, as requested by a spatial coherence analysis of the
section is used to cut the high frequencies mainly related to response 共see Section 5.5兲.
environmental noise and model imperfections. Averaged values of the relevant experimental frequencies
The modal damping coefficients 共see, eg, 关200,201兴兲 are are reported in Table 5 for the first ten modes, either hori-
usually evaluated with two different techniques: i兲 by the zontal Hn 共out-of-plane兲 or vertical Vn 共in-plane兲. Odd
logarithmic decrement of the time law during the decaying 共even兲 mode numbers n correspond to symmetric 共antisym-
free motion of the system after turning off a low-amplitude metric兲 spatial shapes. Corresponding theoretical predictions
excitation around the nominal frequency of the relevant
mode; ii兲 by a nonlinear regression analysis of the experi-
mental transfer function obtained by forcing the system with
a low-amplitude in-plane/out-of-plane random noise. Diffi-
culties are often encountered in determining reliable damp-
ing coefficients. They are due both to the unavoidable occur-
rence of some modal coupling in system response, and to
their strong variability resulting from the low involved val-
ues and from variations of environmental conditions.
5.2.2 Experimental setup and dynamic features of cable/
mass model
Notwithstanding possible differences in the mechanical
model and excitation/acquisition sections, all of the realized Fig. 5 Schematic layout of experimental apparatus
are also reported in Table 5. They were obtained through the 2:2:1:2 internal resonance, just as in the theoretical four-dof
asymptotic model of Cheng and Perkins 关78兴 for free linear model of Section 3.2.
response of cable/mass suspensions with small equilibrium
curvature 共Section 2.3兲. A general satisfactory agreement
with the experimental results is observed. The natural fre-
5.3 DetectingÕclassifying regular and complex dynamics
quencies obtained with a nonlinear finite element code are
also reported in Table 5: their pattern looks again quite sat- Several problems are usually encountered in interpreting/
isfactory and confirms the formerly reported mode sequence. classifying results of experimental investigations on cable
The purely numerical model shows the occurrence of a par- models. They are due to availability of data for very few
allel sequence of out-of-plane torsional modes induced by measurement points, to system sensitivity to variations of
the rotary inertia of concentrated masses and not reproduced initial/environmental conditions, to occurrence of long tran-
by the theoretical model. Corresponding ‘‘spurious’’ modes sients due to lightness of the model, and to local dynamic
with comparable frequencies have been observed in the ex- effects due to possible presence of concentrated, though
perimental model, likely due to imperfections in mounting
small, masses. Various operational measures are taken to re-
the masses perfectly aligned and equally spaced. They are
seen to also contribute to the system nonlinear response in duce such difficulties, and check about the actual nature and
the relevant frequency ranges. robustness of response. i兲 Control parameters are systemati-
The first eight dimensionless in-plane frequencies ␻/␲ of cally swept up and down. ii兲 Long experimental sessions
the theoretical cable-mass are given in Fig. 6 in terms of the strongly influenced by viscoelasticity effects are avoided. iii兲
bare cable parameter ␭/␲, the first eight out-of-plane fre- Several repetitions of short sessions around critical points,
quencies also being indicated. The substantial similarity with aimed at detecting structurally stable bifurcations occurring
the bare cable frequency spectrum 共Fig. 3兲 is worthily no- at averaged values of some relevant parameter, are usually
ticed. For easier identification of the various labels, the pla- performed. Finally, iv兲 different dynamic tools are jointly
nar mode shapes of the reference experimental cable are also used to reliably identify the kind of response.
shown in the figure. Real time information on the dynamical nature of regular
Indeed, experimental results were obtained mostly for a attractors are obtained, during the experiments, from the
cable corresponding to a ␭-value slightly higher than first analysis of frequency power spectra and of other qualitative
crossover, in the range where two different three half-waves dynamic tools: time laws, orbit shapes in phase space, Poin-
on the span planar symmetric modes (V1 and V3) do exist. caré map projections, probability density functions, and au-
The model naturally exhibits nearly simultaneous 2:2:1 inter- tocorrelation functions 关142,143兴. Based on all observation
nal resonances involving the first planar (V2) and nonplanar results 共sometimes, on visual observations, too兲, a mechani-
(H2) antisymmetric modes, and the first nonplanar symmet- cal classification of cable responses is made with relative
ric mode (H1). Moreover, an experimental cable/mass at ease in terms of modal components present in different regu-
nearly perfect crossover was considered, where the first in- lar classes of motion. However, for strongly coupled mo-
plane symmetric mode (V1) is also tuned in a simultaneous tions, it is complicated by the large number of contributing
modes not easily discernible through few measurement
points.
The situation is much harder for complex dynamic re-
gimes developing in large unknown state spaces. The above
mentioned measures of dynamics can still furnish first hints
about the actual nature of nonregular response, from either a
geometrical or mechanical viewpoint. By way of example,
time laws 共a兲, power spectra 共b兲, autocorrelation functions
共c兲, orbit shapes 共upper兲 and probability density functions
共lower兲共d兲, of the vertical 共upper兲 and horizontal 共lower兲
displacement components of one observed mass are reported
in Figs. 7 and 8 for a seemingly quasiperiodic and chaotic
motion, respectively. The power spectra are typically broad-
banded in the chaotic case, while they exhibit several spikes
in the quasiperiodic case. Likewise, the autocorrelation func-
tions describing the time memory of the solution decay to-
wards zero in few seconds for the former, whereas they are
simply modulated with loss and renewal of correlation for
the latter. Analogously, in the chaotic case, the probability
density functions of the displacement components show an
inversion of curvature resembling a Gaussian curve, accord-
ing to the probability to find the system state normally dis-
tributed between the minimum and maximum values.
Fig. 6 Spectrum of natural frequencies of theoretical cable/mass At the same time, identifying, eg, in the phase portrait of
model, and experimental planar mode shapes 关192兴 a nonregular motion, a prevailing timely modulated modifi-
cation of a regular orbit occurring in adjacent regions of the series 共one motion component兲 relevant to one cable point.
control parameter space, can help the mechanical interpreta- But of course more reliable multivariate reconstructions
tion of the motion, as can observations about different levels are obtained when using the two scalar series provided at
of chaoticity occurring in various regions. one single point 共two motion components, vertical and
Nevertheless, all of these basically qualitative tools are horizontal兲 by a biaxial transducer, or simultaneously at
not suitable for an actually reliable characterization of the two different points 共the same motion component兲 by two
dynamical/mechanical nature of nonregular motions, which transducers.
has to be done by means of proper quantitative measures. • Characterizing the response in terms of spatial properties,
Indeed, classifying the complex dynamics of high- which implies identification of the number and shape of
dimensional systems requires two main steps. mechanical configuration variables mostly contributing to
• Characterizing the response in terms of global topological nonregular dynamics. Different from phase space recon-
and dynamical properties of the underlying attractor: di- struction, configuration variables can be identified only
mensionality, strangeness, and possible chaoticity. This im- based on at least two contemporary measures of the same
plies i兲 proper phase space reconstruction of the attractor motion components at two different cable points, which
starting from the available measurements, ii兲 evaluation of requires at least two biaxial transducers.
such invariant measures of the dynamics as attractor di-
mension, Lyapunov exponents, and local topological di- Statements of ‘‘strangeness’’ and/or ‘‘chaoticity’’ must be
mension of the attractor, iii兲 characterization of other made very carefully, since different measures can give quite
meaningful invariant sets. These measures are commonly different indications in terms of chaos ‘‘strength.’’ More pre-
calculated for theoretical models 共Section 4.2.1兲. Reliably cise classification of nonregular motions, as well as
calculating them in companion experimental models is es- detection/prediction of regions in control space where they
sential to distinguish strangeness and chaoticity actually are likely to occur, involve 共see Sections II.6.1, II.5.2兲: i兲
pertaining to the real system from those sometimes ob- comparison of different dimensionality estimates made with
served in theoretical models merely as a result of discreti- independent procedures and ii兲 understanding of system evo-
zation, approximations, and truncation. Phase space recon- lutionary dynamics with variation of a control parameter. In
struction can be made starting from even a single time turn, such topics rely on local and global characterization of
Fig. 7 Quasiperiodic motion: a兲 time laws, b兲 power spectra, c兲 autocorrelation functions, d兲 orbit shapes 共upper兲 and probability density
functions 共lower兲, of the vertical 共upper兲 and horizontal 共lower兲 displacement components of one observed mass 关193兴
the flow structure in phase space, and on reliable description attractor onto the 1D space in which the experimental mea-
of bifurcational mechanisms leading from regular to non- sure is evaluated, by using a proper time delay. The theoret-
regular response. All of this information has to be obtained ical bases for reconstructing a pseudophase space starting
through systematic use of rather sophisticated techniques of from only one or few experimental time series, and obtaining
dynamical systems theory 共Sections 5.4 and 5.5兲, which re- an embedding of the actual, high-dimensional, but inacces-
quire considerable experimental and computational effort. sible phasespace of the system, are by now well established
关202–206兴. Under certain conditions, there is a one-to-one
5.4 Reconstructing dynamic properties from experi- correspondence between the real and reconstructed phase-
mental measurements spaces. The latter preserves all topological properties of the
For analytical models, invariant measures of chaoticity and manifold containing the attractor of the real system, such as
strangeness of underlying attractors are directly computed the possible fractal dimension of the attractor and the diverg-
using the solution vectors evolving in state space, which is ing time evolution of neighboring trajectories which ac-
usually low-dimensional after applying some discretization/ counts for possible chaoticity of the response.
reduction procedure. In contrast, for intrinsically infinite- From a practical point of view, the delay embedding pro-
dimensional experimental models, they can be computed just cedure consists of obtaining reconstructed m-dimensional
after a coherent reconstruction of an image of the real un- pseudovectors in the pseudophase space of delayed coordi-
known state space, in which such global dynamic properties nates. Having a scalar time series 关 x 1 ,x 2 , . . . ,x N 兴 of N dis-
are preserved. crete measures sampled with a time interval ⌬t, after choos-
Reconstruction procedures are based on a time series ing a proper time delay ␶ 共expressed in number of ⌬t), it is
analysis approach, upon which good estimation of strange- possible to construct m-dimensional pseudovectors j v whose
ness, chaoticity, and other invariant measures becomes pos- components are delayed measures of the time series:
sible. In particular, it allows us to reliably estimate them
even if some level of noise is present, as always occurs in
j
v⫽ 共 j v 1 , j v 2 , . . . j v m 兲 , with j v i ⫽x j⫹ 共 i⫺1 兲 ␶ ,
(48)
real measurements. 1⭐i⭐m, 1⭐ j⭐N⫺ 共 m⫺1 兲 ␶
The basic procedure for characterizing time evolution
properties of an asymptotic response is the delay embedding For a known phase space dimension n, Takens’ theorem
technique. It consists basically in projecting the unknown states that, for a generic ␶ 共provided ⌬t is not an integer
Fig. 8 Chaotic motion: a兲 time laws, b兲 power spectra, c兲 autocorrelation functions, d兲 orbit shapes 共upper兲 and probability density
functions 共lower兲, of the vertical 共upper兲 and horizontal 共lower兲 displacement components of one observed mass 关193兴
multiple of the frequency of any periodic component of the sample covariance matrix of delayed coordinates 关211兴 was
motion兲, there is an invertible map preserving the attractor used, and then the invariance of the correlation dimension
structure between the actual and the delay-reconstructed for m⬎m min was checked to control the assumption. Time
phase space, as long as the embedding dimension m reaches series of up to 150,000 points 共corresponding on average to
a ‘‘sufficient’’ value, namely m⭓2n⫹1. However, this refers 3750 forcing periods兲 were used, controlling the invariance
to an ideal situation of scalar series of infinite length and no of the computed measure for variations of the algorithmic
noise. In contrast, in a real finite-length experimental time parameters. Still working on measurements from the cable-
series, the true phase space dimension n is obviously un- mass, Benedettini 关212兴 successfully used the eigenanalysis
known and the data are affected by noise and finite precision of the correlation matrix to perform a base rotation aimed at
due to truncation in acquisition. obtaining a representation of data which guarantees the best
Generally speaking, effectiveness of reconstruction de- linear independence of pseudovectors and thus a good ex-
pends on the resolution obtainable from data at a given pre- pansion of the attractor.
cision of the acquisition board: the acquisition rate must be When more than one time series is used, the efficiency of
as high as possible in order to better ‘‘resolve’’ the structure reconstruction depends on the number of synchronous mea-
of the attractor. Also, acceptable values of signal-to-noise sures, on their nature 共ie, equivalent variables at different
ratio must be achieved through effective noise reduction al- positions, or variables with different physical meaning兲, and
gorithms 关207兴 and/or proper data manipulation. Anyway, the on the spatial coherence of measures taken from different
most important point for reliable reconstruction is a correct points.
choice of the two main parameters m and ␶. After a good embedding, ad hoc algorithms 共Section
A satisfactory choice of the time delay ␶ is governed by 4.2.1兲 are applied to estimate the attractor dimension 共as a
the need to realize a correlation between the new coordinates rule, the correlation dimension兲, the chaoticity of time laws
sufficiently low to unfold the system dynamics on the main 共the maximum Lyapunov exponent or the whole spectrum of
dimensions, but not too low to highlight unreliable nearly exponents, see, eg, 关213,214兴兲, and/or further invariants. The
infinite-dimensional reconstructed dynamics. This can be ob- correlation dimension algorithm 关154兴 allows us calculation
tained in correspondence of the first zero of the autocorrela- of the attractor dimension. The main steps of the procedure
tion function, or of the first minimum of the mutual average are illustrated in Parts a – c of Figs. 9 and 10, respectively,
information. The former criterion allows us to make the de- for a quasiperiodic and a chaotic time series 共see the relevant
lay coordinates x j⫹ ␶ and x j as reliably uncorrelated as pos- Poincarè sections in Part e兲 of the cable-mass in 关193兴. The
sible in a linear sense—and is thus questioned by a number box-counting algorithm, for each embedding dimension, fur-
of researchers 关208兴. The latter one 关209,210兴 accounts for nishes the log(C(r))⫺log(r) curves 共Part a兲, whose slope es-
both linear and nonlinear dependence between the two vari- timation in the scaling region around values of r approaching
ables, and gives a measure of the information that x j can zero 共Part b兲 gives the requested correlation dimension D c .
provide about x j⫹ ␶ . The trend of D c for increasing embedding dimension m is
In turn, the embedding dimension m must be chosen in reported in Part c of the figures, confirming the measure
such a way to avoid both fictitious orbit intersections, which saturation as the right m is reached. The fractal nature and
can arise when projecting the attractor onto a too low- the strangeness of the attractor in Fig. 10e are associated
dimensional space, and possible contamination of the dy- with a noninteger value of D c . The maximum positive
namics produced by noise in a too high-dimensional space Lyapunov exponent extracted from each time series using the
共higher dimensions contain only noise兲. This can be obtained Wolf et al 关152兴 algorithm is also reported in the two figures
based on different criteria 共see, eg, 关143兴兲: i兲 Singular-value 共Part d兲: a positive value was consistently found in the cha-
decomposition of the covariance matrix of delay coordinates otic case, while in the quasiperiodic case the computation
allows us to identify in the eigenvalue spectrum plotted for furnished just the zero value predicted by the dynamical sys-
increasing embedding dimension—above a constant low tems theory.
level representative of experimental noise—all the real mea- These quantitative measures effectively allow us to char-
surable coordinates giving a measure of the embedding di- acterize the actual nature of experimentally observed non-
mension m sufficient for suitable reconstruction. ii兲 False regular motions. Properly fitting them in a general classifica-
nearest neighbors analysis identifies a correct minimum dition scheme based on system dimensionality evaluations is a
mension to ‘‘resolve’’ the attractor as the one at which ficti- more difficult and lengthy task. In this respect, first meaning-
tious crossings of orbits due to a too drastic projection of the ful hints are already obtained from attractor dimension cal-
attractor are avoided. iii兲 Calculation of system invariants culations. Indeed, the number of phase space dimensions 共ie,
like correlation dimension or Lyapunov exponents for a the embedding dimension m兲 needed to contain an attractor
growing embedding dimension allows us to indirectly esti- with fractal dimension D c is bounded by 关 D c 兴 ⭐m⭐ 关 2D c
mate its correct value as the one corresponding to saturation ⫹1 兴共关•兴 being the next greatest integer to 共•兲兲, where the
of the invariant. upper bound is due to Mané 关204兴. Accordingly, the values of
For the cable-mass in Benedettini and Rega 关193兴, a right D c and m can be checked with each other for internal con-
␶ for each considered case was selected using the mutual sistency.
average information criterion. Regarding the minimum Eventually, based on the same reconstructions used to
length m min of pseudovectors, the eigenanalysis of the classify time laws by means of invariant measures, it is pos-
sible to model the data to forecast future states of the system the embedding procedure is of limited value from the view-
关215兴, as well as to control its dynamics even in the chaotic point of system theoretical modeling, where it is essential to
regime 共through, eg, the OGY method and its variants, see know the appropriate configuration variables to be accounted
关216,217兴兲. for in the model formulation. A further independent evalua-
tion of system dimensionality is thus needed.
5.5 Determining spatial properties of nonlinear re- The spatial content of the response and its possible com-
sponse plexity are characterized just by identifying the configura-
The embedding procedure permits a reliable estimation of tions 共spatial shapes兲 that are most visited, on average, dur-
system dimensionality, for giving indications on the actual ing a spatiotemporal evolution. This is usually done through
number of degrees-of-freedom taking meaningful part in the the proper orthogonal decomposition 共POD兲 of the covari-
system nonlinear response. Nevertheless, it provides no spa- ance of a random function, also called Karhunen-Loève de-
tial information concerned with the most important involved composition 共Holmes et al 关218兴兲. It allows us to character-
modes, which have usually great engineering relevance. So, ize the response both in terms of number of involved dof 共as
Fig. 9 Quasiperiodic motion:

correlation dimension a–c兲,
Lyapunov exponent d兲, and Poin-
caré section e兲关193兴
per the embedding procedure兲 and with respect to their me- The object of the study in the POD procedure is the spa-
chanical meaning, associated with the spatial shape. The tial correlation matrix
POD is based on the spatial coherence analysis of experi- R i j ⫽ 具共 x i ⫺x̄ i 兲共 x j ⫺x̄ j 兲典 (49)
mental 共or numerical兲 time series data measured simulta-
neously at different positions throughout the system. This of the response, where x i (t) and x j (t) are displacements si-
procedure, applied originally in mechanics to the study of multaneously measured at the ith and jth points of the sys-
turbulent fluid flows, has recently also been applied to ex- tem, the angle brackets denote the time average, and the
perimental systems in solid and structural mechanics overbar indicates the mean value. The eigenvectors of R
共Cusumano et al 关219兴, Georgiou et al 关220兴, Benedettini form a complete orthogonal set and are called proper or-
and Rega 关221兴, Kappagantu and Feeny 关222兴, Alaggio and thogonal modes 共POMs, or K-L modes兲. The corresponding
Rega 关194兴兲; for a general overview, see the forthcoming eigenvalues ␭ i are the associated mean-square amplitudes,
paper by Kerschen et al 关223兴. and stand for the amount of energy captured by the eigen-
Fig. 10 Chaotic motion: correlation dimension a–c兲, Lyapunov exponent d兲, and Poincaré section e兲关193兴
vectors. Since R i j is positive definite, all of the ␭ i should be has been discussed, focusing on periodic solutions occurring
positive, and the occurrence of negative eigenvalues is a use- in either resonant or nonresonant excitation and system con-
ful indicator of measurement errors. ditions. Techniques and tools for numerical and geometrical
For the suspended cable with eight masses in 关194,221兴, analysis of nonlinear and complex dynamics have been illus-
using two follower cameras to simultaneously measure the trated with reference to the single-dof cable model, reporting
displacements of two masses in a given stationary regime, on measures for response diagnosis as well as on local and
the 共8⫻8兲 response correlation matrix was obtained by global bifurcation analyses. Finally, models and features con-
evaluating all possible combinations of two-mass spatial cor- cerned with the experimental investigation of cable nonlinear
relation measurements 共Eq. 共49兲兲. A quite good estimation of dynamics have been discussed. Special emphasis has been
R was obtained by considering time series with a minimum placed on time series analysis and the ensuing techniques for
of 60,000 points 共corresponding on average to 1500 forcing reconstructing and characterizing dynamic and spatial prop-
periods兲, and averaging several times over the 10⫼20 mea- erties of system nonlinear response from experimental
sures made for each pair of points during the whole obser- measurements.
vation time. Computation of POMs shows to be of great The bundle of analytical, numerical, geometrical, and ex-
interest to the aim of classifying nonregular attractors, and perimental models and methods of analysis summarized in
understanding relevant transition scenarios 共see Section Part I of this review article constitutes the theoretical frame-
II.5.2兲. In addition, it allows indirect evaluation of system work against which discussing the main deterministic non-
dimensionality, to be compared for consistency with the pre- linear phenomena that characterize the finite-amplitude
dictions obtained through the embedding procedure 共Section regular/nonregular dynamics of suspended cables. This will
II.6.1兲. be the subject of the companion Part II 共Rega 关41兴兲.
Indeed, the POMs give optimal mode shapes in a least
squares sense, namely, they belong to the basis capturing
more power per mode than any other set of basis functions. ACKNOWLEDGMENTS
The sum of the ␭ i is equal to the total mean square amplitude Figure 2 is reprinted from the indicted original source with
共the ‘‘power’’兲 of the response at the measured points, and it kind permission of Royal Society. Figures 6 –10 are reprinted
approaches the true mean square amplitude as the number of from the indicated original sources with kind permission of
measured points increases. Thus, one gets indications, on one Kluwer Academic Publishers.
side, on the minimum number of dof needed to capture the
observed dynamics, estimated as the number of POMs
REFERENCES
needed to exceed a fixed fraction 共say 99%兲 of the total
关1兴 Irvine HM 共1981兲, Cable Structures, MIT Press 共1992, Dover Publ,
power. On the other side, on the configurations most visited, New York兲.
on average, by the system—proportionally to the ratio 关2兴 Triantafyllou MS 共1984a兲, Linear dynamics of cables and chains,
␭ i /⌺␭ j —during the spatiotemporal evolution of its re- Shock Vib. Dig. 16, 9–17.
关3兴 Triantafyllou MS 共1987兲, Dynamics of cables and chains, Shock Vib.
sponse: they correspond to the mechanical variables taking Dig. 19, 3–5.
most meaningful part in it. All of this information becomes 关4兴 Triantafyllou MS 共1991兲, Dynamics of cables, towing cables and
more important as more involved 共namely, nonregular兲 the mooring systems, Shock Vib. Dig. 23, 3– 8.
关5兴 Starossek U 共1994兲, Cable dynamics: A review, Int J Struct Eng Mar,
response is. 171–176.
The identified optimal basis and the corresponding main 关6兴 Routh EJ 共1868兲, The Advanced Part of a Treatise on the Dynamics of
a System of Rigid Bodies 共1955, 6th edition, Dover Publ, New York兲.
POMs are thus of great interest also from the modeling view- 关7兴 Rohrs JH 共1851兲, On the oscillations of a suspended cable, Trans.
point 共Section II.6.2兲. They can be suitably referred to for Cambridge Philos. Soc. 9, 379–398.
‘‘best’’ projecting a continuous theoretical model of the sys- 关8兴 Pugsley AG 共1949兲, On the natural frequencies of suspension chains,
Q. J. Mech. Appl. Math. II, 412– 418.
tem through some analytical reduction technique 共see, eg, 关9兴 Saxon DS, and Cahn AS 共1953兲, Modes of vibration of a suspended
Kappagantu and Feeny 关224兴, Alaggio and Rega 关225兴, and chain, Q. J. Mech. Appl. Math. VI共3兲, 273–285.
Georjiou 关226兴兲. 关10兴 Simpson A 共1966兲, Determination of the in-plane natural frequencies
of multispan transmission lines by a transfer matrix method, Proc.
Inst. Electr. Eng. 113, 870– 878.
6 CONCLUDING REMARKS 关11兴 Soler AI 共1970兲, Dynamic response of single cables with initial sag, J.
Franklin Inst. 290, 377–387.
Mechanical and mathematical modeling of elastic suspended 关12兴 Carrier GF 共1945兲, On the non-linear vibration problem of the elastic
cables, and methods of analysis to be used for investigating string, J. Q. Appl. Math. 3, 157–165.
关13兴 Irvine HM, and Caughey TK 共1974兲, The linear theory of free vibra-
the relevant finite-amplitude dynamics, have been addressed tions of a suspended cable, Proc. R. Soc. London, Ser. A 341, 299–
in this paper. 315.
关14兴 Rega G, and Luongo A 共1978兲, Free oscillations of suspended cables
Working in a deterministic framework, the partial and or- 共in Italian兲, Rep II-236, Istituto di Scienza delle Costruzioni, Univ of
dinary differential equations governing the 3D nonlinear os- Rome.
cillations of a continuous cable and, respectively, of 关15兴 Leonard JW, and Recker WW 共1972兲, Nonlinear dynamics of cables
with low initial tension, J. Eng. Mech. Div. 98, 293–309.
Galerkin-reduced cable models have been reported. Attention 关16兴 West HH, Geshwindner LF, and Suhoski JE 共1975兲, Natural vibra-
has been focused on small-sag bare cables and cable/mass tions of suspension cables, J. Struct. Div. ASCE 101, 2277–2291.
suspensions, whose main linearized dynamics features have 关17兴 Henghold WM, Russell JJ, and Morgan JD 共1977兲, Free vibrations of
cable in three dimensions, J. Struct. Div. ASCE 103, 1127–1136.
been summarized. Approximate asymptotic analysis of 关18兴 Gambhir ML, and Batchelor B de V 共1978兲, Parametric study of free
single-mode and multimode, discrete and direct cable models vibration of sagged cable, Comput. Struct. 8, 641– 648.
关19兴 Rosenthal F 共1981兲, Vibrations of slack cables with discrete masses, 关49兴 Gattulli V, Pasca M, and Vestroni F 共1997兲, Nonlinear oscillations of
J. Sound Vib. 78, 573–583. a nonresonant cable under in-plane excitation with a longitudinal con-
关20兴 Fried I 共1982兲, Large deformation static and dynamic finite element trol, Nonlinear Dyn. 14, 139–156.
analysis of extensible cables, Comput. Struct. 15, 315–319. 关50兴 Chang WK, Pilipchuk V, and Ibrahim RA 共1997兲, Fluid flow-induced
关21兴 Migliore HJ, and Webster RL 共1979兲, Current methods for analyzing nonlinear vibration of suspended cables, Nonlinear Dyn. 14, 377–
dynamic cable response, Shock Vib. Dig. 11, 3–16. 406.
关22兴 Migliore HJ, and Webster RL 共1982兲, Current methods for analyzing 关51兴 Rega G, Lacarbonara W, Nayfeh AH, and Chin CM 共1999兲, Multiple
dynamic cable response—1979 to the present, Shock Vib. Dig. 14, resonances in suspended cables: Direct versus reduced-order models,
19–24. Int. J. Non-Linear Mech. 34, 901–924.
关23兴 Tjavaras AA, Zhu Q, Liu Y, Triantafyllou MS, and Yue DKP 共1998兲, 关52兴 Luo ACJ, and Mote Jr CD 共2000兲, Equilibrium solutions and exis-
The mechanics of highly-extensible cables, J. Sound Vib. 213, 709– tence for traveling, arbitrarily sagged elastic cables, ASME J. Appl.
737. Mech. 67, 148 –154.
关24兴 Ni YQ, Lou WJ, and Ko JM 共2000兲, A hybrid pseudo-force/Laplace 关53兴 Nayfeh AH, Arafat HN, Chin CM, and Lacarbonara W 共2002兲, Mul-
transform method for non-linear transient response of a suspended timode interactions in suspended cables, J. Vib. Control 8, 337–387.
cable, J. Sound Vib. 238, 189–214. 关54兴 Lee CL, and Perkins NC 共1992兲, Nonlinear oscillations of suspended
关25兴 Chucheepsakul S, and Wongsa S 共2001兲, Effect of axial stretching on cables containing a two-to-one internal resonance, Nonlinear Dyn. 3,
large amplitude free vibration of a suspended cable, Struct. Eng. 465– 490.
Mech. 11, 185–197. 关55兴 Triantafyllou MS, and Howell CT 共1994兲, Dynamic response of
关26兴 Srinil N, Rega G, and Chucheepsakul S 共2004兲, Three-dimensional cables under negative tension: An ill-posed problem, J. Sound Vib.
nonlinear coupling and dynamic tension in the large amplitude free 173, 433– 447.
vibrations of arbitrarily sagged cables, J. Sound Vib. 269, 823– 852. 关56兴 Chang WK, and Ibrahim RA 共1997兲, Multiple internal resonance in
关27兴 Wu Q, Takahashi K, and Nakamura S 共2003兲, Non-linear vibrations of suspended cables under random in-plane loading, Nonlinear Dyn. 12,
cables considering loosening, J. Sound Vib. 261, 385– 402. 275–303.
关28兴 Koh CG, and Rong Y 共2004兲, Dynamic analysis of large displacement 关57兴 Triantafyllou MS 共1985兲, The dynamics of translating cables, J.
cable motion with experimental verification, J. Sound Vib. 272, 187– Sound Vib. 103, 171–182.
206. 关58兴 Perkins NC 共1992a兲, Modal interactions in the non-linear response of
关29兴 Desai YM, Popplewell N, Shah AH, and Buragohain DN 共1988兲, elastic cables under parametric/external excitation, Int. J. Non-Linear
Geometric nonlinear analysis of cable supported structures, Comput. Mech. 27, 233–250.
Struct. 29, 1001–1009. 关59兴 Zhang W, and Tang Y 共2002兲, Global dynamics of the cable under
关30兴 Desai YM, Yu P, Popplewell N, and Shah AH 共1995兲, Finite element combined parametrical and external excitations, Int. J. Non-Linear
modelling of transmission line galloping, Comput. Struct. 57, 407– Mech. 37, 505–526.
420. 关60兴 Luongo A, and Piccardo G 共1998兲, Nonlinear galloping of sagged
关31兴 Zhang QL, and Peil U 共1999兲, Dynamic behaviors of cables in para- cables in 1:2 internal resonance, J. Sound Vib. 214, 915–940.
metrically unstable zones, Comput. Struct. 73, 437– 443. 关61兴 Srinil N, Rega G, and Chucheepsakul S 共2003兲, Large-amplitude
关32兴 Karoumi R 共1999兲, Some modeling aspects in the nonlinear finite three-dimensional free vibrations of inclined sagged elastic cables,
element analysis of cable supported bridges, Comput. Struct. 71, Nonlin Dyn 33, 129–154.
397– 412. 关62兴 Srinil N 共2004兲, Large-amplitude 3-D dynamic analysis of arbitrarily
关33兴 Sofi A, Borino G, and Muscolino G 共2002兲, Dynamic analysis of sagged inclined extensible cables, PhD dissertation, King Mongkut’s
prestressed cables with uncertain pretension, Meccanica 37, 67– 84. University of Technology Thonburi, Bangkok, Thailand.
关34兴 Ni YQ, Ko JM, and Zheng G 共2002兲, Dynamic analysis of large- 关63兴 Nayfeh AH, and Pai PF 共2004兲, Linear and Nonliner Structural Me-
diameter sagged cables taking into account flexural rigidity, J. Sound chanics, Wiley Series in Nonlinear Science, New York.
Vib. 257, 301–319.
关64兴 Yamaguchi H, Miyata T, and Ito M 共1982兲, Dynamics of a suspended
关35兴 Martinelli L, and Perotti F 共2001兲, Numerical analysis of the nonlin-
cable in three dimensions, Trans. JSCE 14, 424 – 427.
ear dynamic behavior of suspended cables under turbulent wind ex-
关65兴 Simpson A 共1972兲, On the oscillatory motions of translating elastic
citation, Int J Struct. Stab. Dyn. 1, 207–233.
cables, J. Sound Vib. 20, 177–189.
关36兴 Gattulli V, Martinelli L, Perotti F, and Vestroni F 共2004兲, Nonlinear
oscillations of cables under harmonic loading using analytical and 关66兴 Kauderer H 共1958兲, Nichtlinear Mechanik, Springer-Verlag, Berlin.
finite element models, Comput. Methods Appl. Mech. Eng. 193, 69– 关67兴 Rehfield LW 共1974兲, Nonlinear flexural oscillations of shallow arches,
85. AIAA J. 12, 91–93.
关37兴 Zheng G, Ko JM, and Ni YQ 共2002兲, Super-harmonic and internal 关68兴 Nayfeh AH, and Mook DT 共1979兲, Nonlinear Oscillations, Wiley,
resonances of a suspended cable with nearly commensurable natural New York.
frequencies, Nonlinear Dyn. 30, 55–70. 关69兴 Burgess JJ, and Triantafyllou MS 共1988兲, The elastic frequencies of
关38兴 Proc Int Symp on Cable Dynamics 共1995兲, Liège 共1999, Trondheim; cables, J. Sound Vib. 120, 153–165.
2001, Montréal; 2003, Santa Margherita Ligure兲, AIM, Liège. 关70兴 Triantafyllou MS, and Yue DKP 共1995兲, Damping amplification in
关39兴 Irvine HM 共1976兲, Discussion on ‘‘Natural vibrations of suspension highly extensible hysteretic cables, J. Sound Vib. 186, 355–368.
cables,’’ by HH West et al, J. Struct. Div. ASCE 102, 1286 –1288. 关71兴 Behbahani-Nejad M, and Perkins NC 共1996兲, Freely propagating
关40兴 Rega G 共1996兲, Non-linearity, blfurcation and chaos in the finite dy- waves in elastic cables, J. Sound Vib. 196, 189–202.
namics of different cable models, Chaos, Solitons Fractals 7, 1507– 关72兴 Antman SAS 共1995兲, Nonlinear Problems of Elasticity, Springer-
1536. Verlag, New York.
关41兴 Rega G 共2004兲, Nonlinear vibrations of suspended cables, Part II: 关73兴 Shih B, and Tadjbakhsh IG 共1984兲, Small-amplitude vibrations of
Deterministic phenomena, Appl. Mech. Rev. 57, 479–514. extensible cables, J. Eng. Mech. Div. 110, 569–576.
关42兴 Ibrahim RA 共2004兲, Nonlinear vibrations of suspended cables, Part 关74兴 Srinil N, Rega G, and Chucheepsakul S 共2003兲, Nonlinear interaction
III: Random excitation and interaction with fluid flow, Appl. Mech. in the 3D free vibrations of horizontal and inclined sagged cables,
Rev. 57, 515–549. Proc. 5th Int. Symp. Cable Dynamics, Santa Margherita Ligure, AIM,
关43兴 Reismann H, and Pawlik P 共1977兲, Dynamics of initially stressed Liège, 77– 84.
hyperelastic solids, Solid Mech. Arch. 2, 129–185. 关75兴 Al-Qassab M, and Nair S 共2003兲, Wavelet-Galerkin method for free
关44兴 Luongo A, Rega G, and Vestroni F 共1984兲, Planar non-linear free vibrations of elastic cable, J. Eng. Mech. Div. 129, 350–357.
vibrations of an elastic cable, Int. J. Non-Linear Mech. 19, 39–52. 关76兴 Al-Qassab M, and Nair S 共2004兲, Wavelet-Galerkin method for the
关45兴 Perkins NC, and Mote Jr CD 共1987兲, Three-dimensional vibration of free vibrations of an elastic cable carrying an attached mass, J. Sound
travelling elastic cables, J. Sound Vib. 114, 325–340. Vib. 270, 191–206.
关46兴 Takahashi K, and Konishi Y 共1987a兲, Non-linear vibrations of cables 关77兴 Irvine HM, and Griffin JH 共1976兲, On the dynamic response of a
in three-dimensions, Part I: Nonlinear free vibrations, J. Sound Vib. suspended cable, Earthquake Eng. Struct. Dyn. 4, 389– 402.
118, 69– 84. 关78兴 Cheng SP, and Perkins NC 共1992兲, Closed-form vibration analysis of
关47兴 Takahashi K, and Konishi Y 共1987b兲, Non-linear vibrations of cables sagged cable/mass suspensions, ASME J. Appl. Mech. 59, 923–928.
in three-dimensions, Part II: Out-of-plane vibration under in-plane 关79兴 Sergev SS, and Iwan WD 共1981兲, The natural frequencies and mode
sinusoidally time-varying load, J. Sound Vib. 118, 85–97. shapes of cables with attached masses, J. Energy Resour. Technol.
关48兴 Benedettini F, Rega G, and Alaggio R 共1995兲, Nonlinear oscillations 103, 237–242.
of a four-degree-of freedom model of a suspended cable under mul- 关80兴 Leissa AW 共1974兲, On a curve veering aberration, Z. Angew. Math.
tiple internal resonance conditions, J. Sound Vib. 182, 775–798. Phys. 25, 99–111.
关81兴 Triantafyllou MS 共1984b兲, The dynamics of taut inclined cables, Q. J. tions of Mechanical Systems, E Kreuzer and N van Dao 共eds兲, Solid
Mech. Appl. Math. 37, 421– 440. Mechanics and Its Applications 77, Kluwer, Dordrecht, 235–246.
关82兴 Triantafyllou MS, and Triantafyllou GS 共1991兲, Frequency coales- 关113兴 Rega G, and Benedettini F 共1989兲, Planar non-linear oscillations of
cence and mode localization phenomena: A geometric theory, J. elastic cables under subharmonic resonance conditions, J. Sound Vib.
Sound Vib. 150, 485–500. 132, 367–381.
关83兴 Lin HP, and Perkins NC 共1995兲, Free vibration of complex cable/mass 关114兴 Benedettini F, and Rega G 共1989兲, Planar non-linear oscillations of
systems: Theory and experiment, J. Sound Vib. 179, 131–149. elastic cables under superharmonic resonance conditions, J. Sound
关84兴 Desai YM, Yu P, Shah AH, and Popplewell N 共1996兲, Perturbation- Vib. 132, 353–366.
based finite element analyses of transmission line galloping, J. Sound 关115兴 Nayfeh AH 共1985兲, Perturbation methods in nonlinear dynamics,
Vib. 191, 469– 489. Nonlinear Dynamics Aspects of Particle Accelerators, JM Jowett, M
关85兴 Benedettini F, and Rega G 共1987兲, Nonlinear dynamics of an elastic Month, and S Turner 共eds兲, Springer-Verlag, New York.
cable under planar excitation, Int. J. Non-Linear Mech. 22, 497–509. 关116兴 Rahman Z, and Burton TD 共1986兲, Large amplitude primary and su-
关86兴 Hagedorn P, and Schäfer B 共1980兲, On non-linear free vibrations of an perharmonic resonances in the Duffing oscillator, J. Sound Vib. 110,
elastic cable, Int. J. Non-Linear Mech. 15, 333–339. 363–380.
关87兴 Rega G, Vestroni F, and Benedettini F 共1984兲, Parametric analysis of 关117兴 Rahman Z, and Burton TD 共1989兲, On higher order methods of mul-
large-amplitude free vibrations of a suspended cable, Int. J. Solids tiple scales in non-linear oscillations periodic steady state response, J.
Struct. 20, 95–105. Sound Vib. 133, 369–379.
关88兴 Takahashi K, Wu Q, and Nakamura S 共2003兲, In-plane antisymmetric 关118兴 Hassan A 共1994a兲, Use of transformations with the higher-order
response of cables through bifurcation under symmetric sinusoidally method of multiple scales to determine the steady state periodic re-
time-varying load, J. Sound Vib. 268, 1–14. sponse of harmonically excited nonlinear oscillators, Part I: Transfor-
关89兴 Cai Y, and Chen SS 共1994兲, Dynamics of elastic cable under paramet- mation of derivative, J. Sound Vib. 178, 1–19.
ric and external resonances, J. Eng. Mech. Div. 120, 1786 –1802. 关119兴 Hassan A 共1994b兲, Use of transformations with the higher-order
关90兴 Takahashi K 共1991兲, Dynamic stability of cables subjected to an axial method of multiple scales to determine the steady state periodic re-
periodic load, J. Sound Vib. 144, 323–330. sponse of harmonically excited nonlinear oscillators, Part II: Trans-
关91兴 Lilien JL, and Pinto da Costa A 共1994兲, Vibration amplitudes caused formation of detuning, J. Sound Vib. 178, 21– 40.
by parametric excitation of cable stayed structures, J. Sound Vib. 174, 关120兴 Luongo A, and Paolone A 共1999兲, On the reconstitution problem in
69–90. the multiple time-scale method, Nonlinear Dyn. 19, 135–158.
关92兴 Perkins NC 共1992b兲, Discussion of ‘‘Dynamic stability of cables sub- 关121兴 Luongo A, Rega G, and Vestroni F 共1986兲, On nonlinear dynamics of
jected to an axial periodic load,’’ J. Sound Vib. 156, 361–365. planar shear indeformable beams, ASME J. Appl. Mech. 53, 619– 624.
关93兴 Arafat HN, and Nayfeh AH 共2003兲, Nonlinear responses of suspended 关122兴 Nayfeh AH 共2004兲, Resolving controversies in the application of the
cables to primary resonance excitations, J. Sound Vib. 266, 325–354. method of multiple scales and the generalized method of averaging,
关94兴 Luongo A, Rega G, and Vestroni F 共1982兲, Monofrequent oscillations Nonlinear Dyn. 共submitted兲.
of a nonlinear model of suspended cable, J. Sound Vib. 82, 247–259. 关123兴 Luongo A, Di Egidio A, and Paolone A 共2002兲, On the proper form of
关95兴 Al-Noury SI, and Ali SA 共1985兲, Large-amplitude vibrations of para- the amplitude modulation equations for resonant systems, Nonlinear
bolic cables, J. Sound Vib. 101, 451– 462. Dyn. 27, 237–254.
关96兴 Rao GV, and Iyengar R 共1991兲, Internal resonance and nonlinear re- 关124兴 Nayfeh AH, and Chin CM 共1999兲, Perturbation Methods with Math-
sponse of a cable under periodic excitation, J. Sound Vib. 149, 25– 41. ematica™, Dynamic Press Inc, Blacksburg, VA.
关97兴 Nielsen SRK, and Kirkegaard PH 共2002兲, Super and combinatorial 关125兴 Nayfeh AH, and Balachandran B 共1989兲, Modal interactions in dy-
harmonic response of flexible elastic cables with small sag, J. Sound namical and structural systems, Appl. Mech. Rev. 42, S175–S201.
Vib. 251, 79–102. 关126兴 Nayfeh AH 共2000兲, Nonlinear Interactions: Analytical, Computa-
关98兴 Berlioz A, and Lamarque CH 共2004兲, A non-linear model for the tional, and Experimental Methods, Wiley Series in Nonlinear Science,
dynamics of an inclined cable, J. Sound Vib. 共in press兲. New York.
关99兴 Xu YL, and Yu Z 共1999兲, Non-linear vibration of cable-damper sys- 关127兴 Lacarbonara W, Rega G, and Nayfeh AH 共2003兲, Resonant nonlinear
tems. Part II: Application and verification, J. Sound Vib. 225, 465– normal modes of shallow one-dimensional structural systems, Part I:
481. Analytical treatment, Int. J. Non-Linear Mech. 38, 873– 887.
关100兴 Zhao YY, Wang LH, Chen DL, and Jiang LZ 共2002兲, Non-linear 关128兴 Lacarbonara W, and Rega G 共2003兲, Resonant nonlinear normal
dynamic analysis of the two-dimensional simplified model of an elas- modes of shallow one-dimensional structural systems, Part II:
tic cable, J. Sound Vib. 255, 43–59. Activation/orthogonality conditions, Int. J. Non-Linear Mech. 38,
关101兴 Lee CL, and Perkins NC 共1995兲, Three-dimensional oscillations of 851– 872.
suspended cables involving simultaneous internal resonance, Nonlin- 关129兴 Rossi I 共1997兲, Nonlinear oscillations of an elastic suspended cable in
ear Dyn. 8, 45– 63. multiple internal resonance conditions, PhD Thesis, Univ of Rome,
关102兴 Ibrahim RA, and Chang WK 共1999兲, Stochastic excitation of sus- Italy 共in Italian兲.
pended cables involving three simultaneous internal resonances using 关130兴 Benedettini F, and Rega G 共1994兲, Analysis of finite oscillations of
Monte Carlo simulation, Comput. Methods Appl. Mech. Eng. 168, elastic cables under internal/external resonance conditions, Nonlinear
285–304. and Stochastic Dynamics ASME, AMD-192, 39– 46.
关103兴 Gattulli V, and Vestroni F 共2000兲, Nonlinear strategies for longitudi- 关131兴 Feng ZC, and Leal LG 共1994兲, Symmetries of the amplitude equa-
nal control in the stabilization of an oscillating suspended cable, Dyn. tions of an inextensional beam with internal resonance, ASME J.
Control 10, 359–374. Appl. Mech. 62, 235–238.
关104兴 Berlioz A, and Lamarque CH 共2005兲, Nonlinear vibrations of an in- 关132兴 Luongo A, Paolone A, and Di Egidio A 共2003兲, Qualitative analysis of
clined cable, J. Vibr. Acoust. 共accepted兲. classes of motion of multiresonant systems, I: An algebraic method,
关105兴 Nayfeh AH 共1981兲, Introduction to Perturbation Techniques, Wiley, Int. J. Non-Linear Mech. 共submitted兲.
New York. 关133兴 Luongo A, Paolone A, and Di Egidio A 共2003兲, Qualitative analysis of
关106兴 Sanders JA, and Verhulst F 共1985兲, Averaging Methods in Nonlinear classes of motion of multiresonant systems, I: A geometrical method,
Dynamical Systems, Springer-Verlag, New York. Int. J. Non-Linear Mech. 共submitted兲.
关107兴 Rand RH, and Armbruster D 共1987兲, Perturbation Methods, Bifurca- 关134兴 Pakdemirli M, Nayfeh SA, and Nayfeh AH 共1995兲, Analysis of one-
tion Theory, and Computer Algebra, Springer-Verlag, New York. to-one autoparametric resonances in cables—Discretization vs direct
关108兴 Finlayson BA 共1972兲, The Method of Weighted Residuals and Varia- treatment, Nonlinear Dyn. 8, 65– 83.
tional Principles, Academic Press, New York. 关135兴 Yu P, and Pan K 共2003兲, Direct perturbation analysis on cables with
关109兴 Nayfeh AH, Nayfeh JF, and Mook DT 共1992兲, On methods for con- concentrated loads, Comm. Nonlin. Sci. Num. Sim. 8, 97–112.
tinuous systems with quadratic and cubic nonlinearities, Nonlinear 关136兴 Di Egidio A, Luongo A, and Vestroni F 共1996兲, Nonstationary non-
Dyn. 3, 145–162. planar free motions of an orbiting string with multiple internal reso-
关110兴 Pakdemirli M, and Boyaci H 共1995兲, Comparison of direct- nances, Meccanica 31, 363–381.
perturbation methods with discretization-perturbation methods for 关137兴 Perkins NC, and Mote Jr CD 共1988兲, Theoretical and experimental
non-linear vibrations, J. Sound Vib. 186, 837– 845. stability of two translating cable equilibria, J. Sound Vib. 128, 397–
关111兴 Lacarbonara W 共1999兲, Direct treatment and discretizations of non- 410.
linear spatially continuous systems, J. Sound Vib. 221, 849– 866. 关138兴 Pilipchuk VN, and Ibrahim RA 共1997兲, Strong nonlinear modal inter-
关112兴 Rega G, Lacarbonara W, and Nayfeh AH 共2000兲, Reduction methods action in shallow suspended cables with oscillating ends, Chaos, Soli-
for nonlinear vibrations of spatially continuous systems with initial tons Fractals 8, 637– 657.
curvature, IUTAM Symp Recent Developments in Nonlinear Oscilla- 关139兴 Pilipchuk VN, and Ibrahim RA 共1999兲, Nonlinear modal interactions
in shallow suspended cables, J. Sound Vib. 227, 1–28. software for ordinary differential equations 共available by anonymous
关140兴 Rega G 共1995兲, Bifurcation and chaos in the Helmholtz-Duffing os- ftp from ftp.cs.concordia.ca/pub/doedel/auto兲.
cillator, Bifurcation and Chaos, J Awrejcewicz 共ed兲, Springer Verlag, 关172兴 Hsu CS 共1992兲, Global analysis by cell mapping, Int. J. Bifurcation
Berlin, 191–215. Chaos Appl. Sci. Eng. 2, 727–771.
关141兴 Thompson JMT, and Stewart HB 共1986兲, Nonlinear Dynamics and 关173兴 Ueda Y 共1991兲, Survey of regular and chaotic phenomena in the
Chaos, Wiley, Chichester, England. forced Duffing oscillator, Chaos, Solitons Fractals 1, 199–231.
关142兴 Moon FC 共1992兲, Chaotic and Fractal Dynamics, Wiley, New York. 关174兴 Abraham RH, and Shaw CD 共1992兲, Dynamics: The Geometry of
关143兴 Nayfeh AH, and Balachandran B 共1995兲, Applied Nonlinear Dynam- Behavior, Addison-Wesley, Redwood City.
ics Wiley, New York. 关175兴 Grebogi C, Ott E, and Yorke JA 共1983兲, Crises, sudden changes in
关144兴 Szemplinska-Stupnicka W, Plaut RH, and Hsieh JC 共1989兲, Period chaotic attractors, and transient chaos, Physica D 7, 181–200.
doubling and chaos in unsymmetric structures under parametric exci- 关176兴 Grebogi C, Ott E, and Yorke JA 共1987兲, Basin boundary metamor-
tation, ASME J. Appl. Mech. 56, 947–952. phoses: Changes in accessible boundary orbits, Physica D 24, 243–
关145兴 Rega G 共1991兲, From regular nonlinear oscillations to chaotic mo- 262.
tions in structural dynamics, Nonlinear Problems in Engineering, C 关177兴 Salvatori A, and Rega G 共1996兲, Computational characterization of
Carmignani and G Maino 共eds兲, ENEA Workshops on Nonlinear Dy- attractor-basin structure in the nonlinear dynamics of multidegree-of-
namics 4, World Scientific, Singapore, 195–209. freedom structural models, Structural Dynamics, G Augusti, C Borri,
关146兴 Parker TS, and Chua LO 共1989兲, Practical Numerical Algorithms for and P Spinelli 共eds兲, Balkema, Rotterdam, 341–348.
Chaotic Systems, Springer-Verlag, New York. 关178兴 Lai YC, Grebogi C, and Yorke JA 共1992兲, Sudden change in the size
关147兴 Guckenheimer J, and Holmes P 共1983兲, Nonlinear Oscillations, Dy- of chaotic attractors: How does it occur? Applied Chaos, JH Kim and
namical Systems, and Bifurcation of Vector Fields, Springer-Verlag, J Stringer 共eds兲, Wiley, New York, 441– 455.
New York. 关179兴 Johnson JM, and Bajaj AK 共1989兲, Amplitude modulated and chaotic
关148兴 Wiggins S 共1990兲, Introduction to Applied Nonlinear Dynamical Sys- dynamics in resonant motion of strings, J. Sound Vib. 128, 87–107.
tems and Chaos, Springer-Verlag, New York. 关180兴 Kreuzer E, and Lagemann B 共1996兲, Cell mapping for multi-degree-
关149兴 Benedettini F, and Rega G 共1990兲, Numerical simulations of chaotic of-freedom systems parallel computing in nonlinear dynamics,
dynamics in a model of elastic cable, Nonlinear Dyn. 1, 23–38. Chaos, Solitons Fractals 7, 1683–1691.
关150兴 Rega G, Benedettini F, and Salvatori A 共1991兲, Periodic and chaotic 关181兴 Zaretzky CL, and Crespo da Silva MRM 共1994兲, Experimental inves-
motions of an unsymmetric oscillator in nonlinear structural dynam- tigation of non-linear modal coupling in the response of cantilever
ics, Chaos, Solitons Fractals 1, 39–54. beams, J. Sound Vib. 174, 145–167.
关151兴 Tongue BH 共1987兲, Characteristics of numerical simulations of cha- 关182兴 Nayfeh TA, Nayfeh AH, and Mook DT 共1994兲, A theoretical and
otic systems, ASME J. Appl. Mech. 54, 695– 699. experimental investigation of a three-degree-of-freedom structure,
关152兴 Wolf A, Swift J, Swinney HL, and Vastano JA 共1985兲, Determining Nonlinear Dyn. 6, 353–374.
Lyapunov exponents from a time series, Physica D 16, 285–317. 关183兴 Cusumano JP, and Sharkady MT 共1995兲, An experimental study of
关153兴 Stoop R, and Meier PF 共1988兲, Evaluation of Lyapunov exponents bifurcation, chaos and dimensionality in a system forced through a
and scaling functions from time series, J. Opt. Soc. Am. 5, 1037– bifurcation parameter, Nonlinear Dyn. 8, 467– 489.
1045. 关184兴 Molteno TCA, and Tufillaro NB 共1990兲, Torus doubling and chaotic
关154兴 Grassberger P, and Procaccia I 共1983兲, Characterization of strange string vibrations: Experimental results, J. Sound Vib. 137, 327–330.
attractors, Phys. Rev. Lett. 50, 346 –349. 关185兴 O’Reilly O, and Holmes PJ 共1992兲, Non-linear, non-planar and non-
关155兴 Moon FC, and Li GX 共1985兲, The fractal dimension of the two-well periodic vibrations of a string, J. Sound Vib. 153, 413– 435.
potential strange attractor, Physica D 17, 99–108. 关186兴 Nayfeh SA, Nayfeh AH, and Mook DT 共1995兲, Nonlinear response of
关156兴 Benedettini F, Rega G, and Salvatori A 共1992兲, Prediction of bifurca- a taut string to longitudinal and transverse end excitation, J. Vib.
tion and chaos for an asymmetric elastic oscillator, Chaos, Solitons Control 1, 307–334.
Fractals 2, 303–321. 关187兴 Yamaguchi H, Miyata T, and Ito M 共1978兲, A behavior on nonlinear
关157兴 Rega G, Salvatori A, and Benedettini F 共1995兲, Numerical and geo- dynamic response of cable systems, Proc 24th Symp Structural Engi-
metrical analysis of bifurcation and chaos for an asymmetric elastic neering, 55– 61.
nonlinear oscillator, Nonlinear Dyn. 7, 249–272. 关188兴 Lee CL, and Perkins NC 共1993兲, Experimental investigation of iso-
关158兴 Rega G, and Salvatori A 共1996兲, Bifurcation structure at 1/3- lated and simultaneous internal resonances in suspended cables, Non-
subharmonic resonance in an asymmetric nonlinear elastic oscillator, linear Vibrations ASME, DE-54, 21–31.
Int. J. Bifurcation Chaos Appl. Sci. Eng. 6, 1529–1546. 关189兴 Fujino Y, Warnitchai P, and Pacheco BM 共1993兲, Active stiffness con-
关159兴 Kreuzer E 共1987兲, On the numerical study of bifurcation problems, trol of cable vibration, ASME J. Appl. Mech. 60, 948 –953.
Bifurcations: Analysis, Algorithm, Applications, T Küpper, R Seydel, 关190兴 Fujino Y, and Susumpow T 共1994兲, An experimental study on active
and H Troger 共eds兲, Birkhäuser, Basel, 161–171. control of in-plane cable vibration by axial support motion, Earth-
关160兴 Pezeshki C, and Dowell EH 共1989兲, Generation and analysis of quake Eng. Struct. Dyn. 23, 1283–1297.
Lyapunov exponents for the buckled beam, Int. J. Non-Linear Mech. 关191兴 Benedettini F, and Moon FC 共1995兲, Experimental dynamics of a
24, 79–97. hanging cable carrying two concentrated masses, Int. J. Bifurcation
关161兴 Pinto da Costa A, Martins JAC, Branco F, and Lilien JL 共1996兲, Chaos Appl. Sci. Eng. 5, 145–157.
Oscillations of bridge stay cables induced by periodic motions of 关192兴 Rega G, Alaggio R, and Benedettini F 共1997兲, Experimental investi-
deck and/or towers, J. Eng. Mech. Div. 122, 613– 622. gation of the nonlinear response of a hanging cable, Part I: Local
关162兴 van Dooren R 共1988兲, On the transition from regular to chaotic be- analysis, Nonlinear Dyn. 14, 89–117.
havior in the Duffing oscillator, J. Sound Vib. 121, 327–339. 关193兴 Benedettini F, and Rega G 共1997兲, Experimental investigation of the
关163兴 Leung AYT, and Fung TC 共1989兲, Construction of chaotic regions, J. nonlinear response of a hanging cable, Part II: Global analysis, Non-
Sound Vib. 131, 445– 455. linear Dyn. 14, 119–138.
关164兴 Cardona A, Lerusse A, and Géradin M 共1998兲, Fast Fourier nonlinear 关194兴 Alaggio R, and Rega G 共2000兲, Characterizing bifurcations and
vibration analysis, Comput. Mechanics 22, 128 –142. classes of motion in the transition to chaos through 3D-tori of a con-
关165兴 Yu Z, and Xu YL 共1999兲, Non-linear vibration of cable-damper sys- tinuous experimental system in solid mechanics, Physica D 137, 70–
tems. Part I: Formulation, J. Sound Vib. 225, 447– 463. 93.
关166兴 Hayashi C 共1964兲, Nonlinear Oscillations in Physical Systems, 关195兴 Rega G, and Alaggio R 共2001兲, Spatio-temporal dimensionality in the
McGraw-Hill, New York. overall complex dynamics of an experimental cable/mass system, Int.
关167兴 Szemplinska-Stupnicka W 共1990兲, The Behaviour of Nonlinear Vi- J. Solids Struct. 38, 2049–2068.
brating Systems, Vols I and II, Kluwer, Dordrecht. 关196兴 Campbell JE, Baker JR, Gagel JC, and Smith SW 共2004兲, Response
关168兴 Szemplinska-Stupnicka W, and Bajkowski J 共1986兲, The 1/2- of a sagging, inclined cable to base excitation: Laboratory experi-
subharmonic resonance and its transition to chaotic motion in a non- ments and nonlinear transient simulations, Tenth Conf. on Nonlinear
linear oscillator, Int. J. Non-Linear Mech. 21, 401– 419. Vibrations, Stability, and Dynamics of Structures 共CDRom of Ab-
关169兴 Foale S, and Thompson JMT 共1991兲, Geometrical concepts and com- stracts兲, Blacksburg, VA.
putational techniques of nonlinear dynamics, Comput. Methods Appl. 关197兴 Cheng SP, and Perkins NC 共1994兲, Theoretical and experimental
Mech. Eng. 89, 381–394. analysis of the forced response of sagged cable/mass suspensions,
关170兴 Seydel R 共1994兲, Practical Bifurcation and Stability Analysis, ASME J. Appl. Mech. 61, 944 –948.
Springer-Verlag, New York. 关198兴 Russell JC, and Lardner TJ 共1998兲, Experimental determination of
关171兴 Doedel EJ, Champneys AR, Fairgrieve TR, Kuznetsov YuA, Sandst- frequencies and tension for elastic cables, J. Eng. Mech. Div. 124,
ede B, and Wang XJ 共1997兲, AUTO97: Continuation and bifurcation 1067–1072.
关199兴 Gattulli V, and Benedettini F 共1999兲, Nonlinear feedback control for 关214兴 Rosenstein MT, Collins JJ, and De Luca CJ 共1993兲, A practical
the stabilization of cable oscillations: Analytical and experimental method for calculating largest Lyapunov exponents from small data
models, Proc XIII ASCE Engng Mech Div Conf, CD Rom, Baltimore. sets, Physica D 65, 117–134.
关200兴 Yamaguchi H, and Fujino Y 共1987兲, Modal damping of flexural os- 关215兴 Kantz, and Schreider 共1997兲, Nonlinear Time Series Analysis, Non-
cillation in suspended cables, Struct. Eng./Earthquake Eng. 4, 413s– linear Science Series 7, Cambridge Univ Press, Cambridge, UK.
421s. 关216兴 Ott E, Grebogi C, and Yorke JA 共1990兲, Controlling chaos, Phys. Rev.
关201兴 Barbieri N, de Souza Junior OH, and Barbieri R 共2004兲, Dynamical Lett. 64, 1196 –1199.
analysis of transmission line cables. Part 2—damping estimation, 关217兴 Ott E 共1993兲, Chaos in Dynamical Systems, Cambridge Univ Press,
Mech. Systems Sign. Process. 18, 671– 681. Cambridge.
关202兴 Packard NH, Crutchfield JP, Farmer JD, and Shaw RS 共1980兲, Geom- 关218兴 Holmes P, Lumley JL, and Berkooz G 共1996兲, Turbulence, Coherent
etry from a time series, Phys. Rev. Lett. 45, 712–716. Structures, Dynamical Systems and Symmetry, Cambridge Mono-
关203兴 Takens F 共1981兲, Detecting strange attractors in turbulence, Dynami- graphs on Mechanics, Cambridge Univ Press, New York.
cal Systems and Turbulence, DA Rand and LS Young 共eds兲, Springer 关219兴 Cusumano JP, Sharkady MT, and Kimble BW 共1994兲, Experimental
Lecture Notes in Mathematics 898, New York, 266 –281. measurements of dimensionality and spatial coherence in the dynam-
关204兴 Mané R 共1981兲, On the dimension of the compact invariant sets of ics of a flexible-beam impact oscillator, Proc. Phys. Soc., London,
certain nonlinear maps, Dynamical Systems and Turbulence, DA Sect. A 347, 421– 434.
Rand and LS Young 共eds兲, Springer Lecture Notes in Mathematics 关220兴 Georgiou IT, Schwartz I, Emaci E, and Vakakis A 共1999兲, Interaction
898, New York, 230–242. between slow and fast oscillations in an infinite degree-of-freedom
关205兴 Sauer T, Yorke JA, and Casdagli M 共1991兲, Embedology, J. Stat. linear system coupled to a nonlinear subsystem: Theory and experi-
Phys. 65, 579– 616.
ments, ASME J. Appl. Mech. 66, 448 – 459.
关206兴 Abarbanel HDI, Brown R, Sidorowich JJ, and Tsimring LS 共1993兲,
关221兴 Benedettini F, and Rega G 共1999兲, Nonregular regimes of monodi-
The analysis of observed chaotic data in physical systems, Rev. Mod.
mensional mechanical systems with initial curvature: Experiments
Phys. 65, 1331–1391.
and time series analysis, New Applications of Nonlinear and Chaotic
关207兴 Kostelich EJ, and Schreiber T 共1993兲, Noise reduction in chaotic
time-series data: A survey of common methods, Phys. Rev. E 48, Dynamics in Mechanics, FC Moon 共ed兲, Solid Mechanics and Its
1752–1763. Applications 63, Kluwer, Dordrecht, 139–148.
关208兴 Abarbanel HDI 共1995兲, Tools for analyzing observed chaotic data, 关222兴 Kappagantu RV, and Feeny BF 共2000a兲, Part I: Dynamical character-
Stability, Vibration and Control of Structures, A Guran and DJ Inman ization of a frictionally excited beam, Nonlinear Dyn. 22, 317–333.
共eds兲, World Scientific, Singapore. 关223兴 Kerschen G, Golinval JC, Vakakis AF, and Bergman LA 共2005兲, The
关209兴 Fraser AM, and Swinney HL 共1986兲, Independent coordinates for method of proper orthogonal decomposition for dynamical character-
strange attractors from mutual information, Phys. Rev. A 33, 1134 – ization and order reduction of mechanical systems: An overview,
1140. Nonlinear Dyn. 共accepted兲.
关210兴 Fraser AM 共1989兲, Reconstructing attractors from scalar time series: 关224兴 Kappagantu RV, and Feeny BF 共2000b兲, Part II: Proper orthogonal
A comparison of singular system and redundancy criteria, Physica D modal modeling of a frictionally excited beam, Nonlinear Dyn. 23,
34, 391– 404. 1–11.
关211兴 Broomhead DS, and King GP 共1986兲, Extracting qualitative dynamics 关225兴 Alaggio R, and Rega G 共2001兲, Exploiting results of experimental
from experimental data, Physica D 20, 217–236. nonlinear dynamics for reduced-order modeling of a suspended cable,
关212兴 Benedettini F 共1996兲, An experimental time series analysis approach ASME Des Engng Techn Conf, 18th Bienn Conf Mechanical Vibration
in the classification of non periodic motions in nonlinear structural and Noise, DETC01/VIB-21554, CD-Rom.
dynamics, Structural Dynamics, G Augusti, C Borri, and P Spinelli 关226兴 Georgiou IT 共2005兲, Advanced proper orthogonal decompostion tools:
共eds兲, 1, 415– 421, Balkema, Rotterdam. Using reduced order models to identify normal modes of oscillation
关213兴 Holzfuss J, and Lauterborn W 共1989兲, Lyapunov exponents from a and slow invariant manifolds in the dynamics of planar nonlinear
time series of acoustic chaos, Phys. Rev. A 39, 2146 –2152. rods, Nonlinear Dyn. 共accepted兲.
Giuseppe Rega is Professor of Solid and Structural Mechanics at the Department of Structural and
Geotechnical Engineering of the University of Rome ‘‘La Sapienza,’’ where he received his ‘‘Laurea
cum laude’’ in civil engineering (1970), and was formerly appointed as a research assistant. From
1978 to 1995, he was assistant professor, associate professor, and full professor at the University of
L’Aquila, where he was also Head of the Department of Structure, Water and Soil Engineering. Since
its establishment, he was a member of the Secretariat of Italian Professors of Structural Mechanics
for ten years. Currently, he is Chairman of the PhD Program in Structural Engineering of the
University of Rome ‘‘La Sapienza.’’ Previous scientific interests were concerned with various topics
in solid and structural mechanics, earthquake engineering, and nonlinear oscillations in mechanical
and structural engineering. Current research interests are concerned with nonlinear dynamics, bi-
furcation, chaos, and their control in applied mechanics and engineering, addressed through ana-
lytical, computational, geometrical, and experimental approaches. He has published nearly 150
papers in refereed journals, edited volumes, and refereed conference proceedings, and has given plenary and invited lectures
in international conferences. He chaired an International Symposium, an Euromech Colloquium, and an IUTAM Symposium
on various topics of nonlinear dynamics, bifurcation, chaos, and control in mechanical systems and processes, as well as
sessions and minisymposia within several international conferences. He was a member of the EUROMECH Nonlinear
Oscillations Conference Committee (1992–2002) and edited Special Issues of Nonlinear Dynamics, Meccanica, and Chaos
Solitons & Fractals. Editor-in-Chief of Meccanica (1998–2004), he is currently Associate Editor of Chaos Solitons &
Fractals for Chaos in Civil Engineering and a member of the editorial boards of Nonlinear Dynamics, Journal of Vibration
and Control, International Journal of Nonlinear Science and Numerical Simulations, and Mathematical Problems in
Engineering.

Nonlinear Vibrations of Suspended Cables-Part I: Modeling and Analysis

Uploaded by

Copyright:

Available Formats

Nonlinear Vibrations of Suspended Cables-Part I: Modeling and Analysis

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

Nonlinear Vibrations of Suspended Cables-Part I: Modeling and Analysis

Uploaded by

Copyright:

Available Formats

Nonlinear vibrations of suspended cables—Part I:

Modeling and analysis

Transmitted by Associate Editor P Pfeiffer

smaller. Consequently, on the time scale of 共lower frequency兲 H v ⬙ ⫹hy ⬙ ⫽m v̈

and the associated natural frequencies

Fig. 3 Spectrum of natural frequencies of parabolic cable. Dashed

perharmonic兲 occur with values of the excitation amplitudes

P, as well. Moreover, it allows for their effect to first appear Ȧ⫽ 兺

a ␥˙ ⫽⌺ k ␧ k aD k ␥ , ȧ⫽⌺ k ␧ k D k a (31) cording to whether the effective nonlinearity coefficient

internal resonances 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 3 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 31⫹␧ 2 ␳ 32 2 ␻ 3 ⫽ ␻ 1 ⫹␧ ␳ 3

in wide ranges of variation of the control parameters, this v共 x,t;␧ 兲 ⫽ 兺

1 2 ⫹A i⫹1 共 T 1 ,T 2 兲 e i ␻ i⫹1 T 0 f i⫹1 共 x 兲 ⫹cc, k⫽1,2 (42)

Fig. 9 Quasiperiodic motion:

You might also like