Acta Mech 224, 525547 (2013)

DOI 10.1007/s00707-012-0771-4

Mohammed Ismail Jos Rodellar Gennaro Carusone

Marco Domaneschi Luca Martinelli

Characterization, modeling and assessment of Roll-N-Cage

isolator using the cable-stayed bridge benchmark

Received: 18 July 2012 / Revised: 13 October 2012 / Published online: 20 November 2012
Springer-Verlag Wien 2012

Abstract This paper presents the results of an extensive series of simulation tests to identify the mechanical
characteristics of an innovative isolation device known as the Roll-N-Cage (RNC) isolator. The seismic perfor-
mance of an RNC passive control scheme is subsequently investigated on a model of the cable-stayed bridge
benchmark. Starting from different configurations studied in the laboratory for a 1/10 reduced-scale proto-
type, the RNC isolator stiffness and damping properties are investigated in terms of cyclic tests with different
parameters. Tests at the ultimate level state consisting of monotonic shear and axial loading have been also
carried out as a part of the qualification process. The goal of this study is twofold: first, to examine the main
integrated mechanisms of the RNC isolator through sophisticated 3D finite element simulation models using a
multi-purpose finite element code. The main result of this step is to attempt modeling the forcedisplacement
relationship using the standard BoucWen model of smooth hysteresis. The second aim of this study is the
numerical assessment of the device efficiency through its implementation into a bridge model considering
several ground motions as external excitations. Based on these extensive studies, it was found that the RNC
isolator is promising as a reliable isotropic horizontal isolation device for bridge structures.

1 Introduction

The mechanical characterization of an innovative system or a device is often performed both experimentally
and via numerical simulation. The former approach, which is more costly than the latter one, leads directly to
the physical understanding of the system, and therefore, it can be employed for final verification. In contrast,
numerical simulation uses numerical methods such as the finite element (FE) method to quantitatively repre-
sent the evolution of a physical system and allows for more economically exploring a large number of possible
design solutions. By using accurate models, the result of such simulations can give a good representation of
the real mechanical behavior of the studied device or of the complex system which comprises the device. This
accuracy enables safe drawing of proper conclusions and getting a thorough understanding of the system.
Numerical simulation has been extensively used in the past for the performance assessment of seismic iso-
lation devices. These isolation systems are meant to enable a structure (like a building or a bridge) to survive
M. Ismail
Structural Engineering Department, Zagazig University, Zagazig, Egypt

J. Rodellar
Department of Applied Mathematics III, Universitat Politcnica de Catalunya, Barcelona, Spain

G. Carusone M. Domaneschi (B) L. Martinelli

Department of Structural Engineering, Politecnico di Milano, P.zza Leonardo da Vinci 32, 20133 Milano, Italy
E-mail: domaneschi@stru.polimi.it
Tel.: +39-02-23994267
Fax: +39-02-23994220
526 M. Ismail et al.

a potentially devastating earthquake by decoupling it from the damaging horizontal components of the input
ground motion. Key parameters of the isolation devices have been initially studied by means of analogue com-
puters (e.g., [1]) and later on with the finite element method. Indeed, the magnitude of the stiffness nonlinearity
required in the design of nonlinear isolators, with reference to resonant amplitudes and force transmissibility,
was fist predicted using analogue computers [1] and subsequently with the finite element method [2]. The
same isolation system was considered in [3] when subjected to base excitation.
When the progress of digital computers and related analysis software solutions of the discrete form of
the involved equations became standard, for example, the performance of the friction-pendulum system was
examined in [4,5] using an advanced analytical model and finite element formulation. It was found that the
local bending moment, which results from the movement of the isolator, is an important factor in the design
process. In [6], seismic isolation of liquid storage tanks and pool-type nuclear spent fuel storage tanks was
performed using the finite element method for the structure and boundary element method for the liquid under
earthquake motion. The primary and secondary stiffness values of viscoelastic isolators were studied by [7],
based on geometry changes, using finite element analysis. A hydraulic engine mount with an inertia track and
a free decoupler was examined using a developed finite element model and an available finite element package
for problem simulation in [810]. The nonlinear analysis was employed in [11] to account for the friction-
pendulum systems inherent nonlinear behavior through the implementation of the global Automatic Dynamic
Incremental Nonlinear Analysis (ADINA) [12], which was used as an analytical tool for the development of
a retrofit project for a bridge in [13] using a seismic isolation system. The influence of the shape factor on
the stress distributions and stress concentrations of a laminated elastomeric bearing subjected to vertical loads
was investigated by adopting the finite element approach in [14]. The behavior of natural rubber earthquake
isolation bearings was modeled in a finite element analysis using a low-order strain energy density function by
[15]. A seismic isolation system called the direction-optimized friction-pendulum system was proposed and
investigated to determine its mechanical behavior by a finite element formulation in [16]. Another proposed
isolation system consisting of high damping rubber bearings strengthened with glass fiber fabrics was studied
in [17], and a wide parametric numerical investigation by a finite element analysis was carried out in order to
develop and verify analytical models for that isolation device.
The ideal isolator should possess horizontal flexibility, stability and energy dissipation characteristics. The
ideal isolator for civil applications should additionally provide vertical support for relevant loads, since the
mass of civil structures is usually large. In the effort to improve and overcome the limits of existing isola-
tion devices, a new seismic isolation device, called Roll-N-Cage (RNC) isolator, has been recently proposed
[1822]. The RNC isolator conveniently provides in a single unit all the necessary functions of vertical rigid
support, horizontal flexibility with enhanced stability and energy dissipation characteristics and includes a
built-in buffer mechanism to limit the peak isolator displacement under seismic excitations stronger than the
design earthquakes. Moreover, the quasi-ellipsoidal shape of the rolling core of the RNC isolator provides a
linear built-in re-centering mechanism that prevents having residual isolators displacement to improve the
isolator stability and to prevent torsional response of the isolated structure. Such a re-centering mechanism
will be investigated in the future experimentally.
In [1922] details are found on the design principles along with numerical studies to assess the efficiency
to isolate several classes of building structures and in-house equipments. Moving toward practical realizations,
physical prototypes have been now designed. Prior to their experimental assessment, this paper presents an
extensive study of the mechanical characterization performed by means of an available finite element software.
Designing, building and testing a series of prototypes is a costly and time-consuming process; therefore, the
numerical simulation is thought of as an alternative less expensive and faster approach to gain the necessary
understanding of the details of the RNC isolator behavior, during all the stages of its development. This simu-
lation approach has to be followed by an experimental verification to validate the numerically obtained results.
For this reason the same set of future experimental tests has been numerically simulated, while the first vali-
dation for the numerical characterization of the RNC isolator is performed against experimental data already
available from the physical testing of small-scale prototypes. The results are presented and discussed in this
paper, while the experimental verification will be performed in the near future, and the results are expected
to be presented in a further paper. Finally, in order to evaluate the seismic performance of the RNC protec-
tion system for bridges, a refined version of the international ASCE (American Society of Civil Engineers)
benchmark for a controlled cable-stayed bridge is considered as case study [23].
The benchmark structure is developed in a commercial finite element code (ANSYS [24]) including
enhanced aspects in the simulation [25].
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 527

Mechanical characterization and efficiency assessment of the RNC control devices have been first evaluated
by 3D FE numerical simulations, and a suitable BoucWen model arrangement is identified for further seismic
analyses on the aforementioned benchmark problem of cable-stayed bridges. The target response of a prototype
isolator has been addressed by 3D FE analyses. The BoucWen model is then employed in an appropriate
arrangement to model the forcedisplacement relationship resulting from the simulated 3D response of the
RNC isolator FE prototypes. This approach would provide a convenient tool that replaces the RNC isolator in
a more general parametric oriented structural analysis of the bridge.
Subsequently, the refined bridge model is developed in the ANSYS multi-purpose finite element environ-
ment including new modeling aspects, in the simulation of the stay cables dynamics, in the implementation of
the seismic excitation and in the soilstructure interaction. The RNC devices are implemented in the bridge
model between the deck and the piers and bents. The model comprises also soilstructure interaction by the use
of impedance functions; the foundation is simulated by lumped masses with equivalent springs and dampers.
Focusing the attention on the simulation of the structural dynamics, the cable model is refined moving from
the single rod type representation, used in the original benchmark, to a description with six rope elements for
each cable [25].
Analyses are carried out in time domain, and the vertical component of the earthquake is also considered.
The seismic input, which is not the same on all the bridge supports, satisfies a fixed coherence function between
the stations and the seismic components at the same station. The soil type regulates the correlation degree.

2 RNC bearing design

Three different configurations of a 1/10 reduced-scale prototype of the RNC isolator have been designed
based on the maximum allowed displacement. These prototypes will be used for experimental mechanical
characterization of the RNC isolator, in addition to the subsequent experimental assessment of its efficiency
through implementation into a scale building model and equipment isolation. The RNC isolator, Fig. 1, is
composed of deformable and non-deformable components. The non-deformable components are the upper
and lower bearing plates, the rolling body and the horizontal ring holders of the metallic yield dampers. The
ring holders are made of steel while two copies of the other non-deformable components are fabricated using
stiff aluminum and Delrin. Regarding the deformable components of the RNC isolator, they are rubber plates
and the vertical metallic yield dampers. Neoprene is selected for the rubber plates while several materials are
to be tried as dampers such as mild steel, lead, copper and aluminum alloys in order to select the damper
material that gives the RNC isolator the required characteristics. The manufacturing process is carried out with
a minimum precision of 0.01 mm.
The three RNC isolator prototypes allow for maximum shear displacements of 50, 60 and 70 mm, respec-
tively, for the first, second and third configuration. The prototypes can have up to 16 metallic yield dampers, of
3 mm diameter, arrayed around the rolling body at equal in-plan angles. The second and third configurations
are provided with rubber plates, two plates for each configuration having the same thickness, of 5 and 10 mm,
respectively, whereas the first configuration has no rubber plates. The outer horizontal and vertical dimensions
of the prototypes are 241.71 mm 156.20 mm, 251.71 mm 166.20 mm, 261.71 mm 176.20 mm for first,
second and third configuration, respectively. A quoted sample design drawing of the second configuration is
shown in Fig. 2.
It is worth noting that this version of the RNC device is recognized as the RNC-c one. The alternative
form of the RNC isolator is designed for supporting heavy masses and has an elastomeric ring surrounding
the rolling body. The RNC with elastomer is classified as RNC-a if it presents curved bars surrounding the
devices core or as RNC-b if it has straight bars embedded into the elastomeric ring. In the first part of the
paper results, mechanical characterization concerning the RNC-c isolator is presented. RNC-c has been chosen
for easily decoupling the different components of the control force. This same typology has been adopted to
run seismic performance evaluations of the RNC device with the bridge FE model. A complete description of
RNC devices and their features can be found in [1822].

3 Types of tests

A carefully selected large number of tests are performed, considering the designed reduced-scale proto-
types, by means of an available finite element package to characterize the prototypes mechanical properties.
A typical set of performed tests is listed in Table 1. These tests have been repeated considering different damper
materials.
528 M. Ismail et al.

Fig. 1 The 1/10 reduced-scale RNC isolator

3.1 Unidirectional characteristic tests

3.1.1 Basic test

The performed basic type of test has a horizontal displacement with sinusoidal shape, conducted at variable
axial loads and for three cycles of loading at shear strain (displacement vs. design displacement) amplitudes
of 5, 25, 50, 75 and 100 %. A 100 % shear strain amplitude brings the RNC isolator to a position at which the
buffer mechanism starts working. The loading frequency is 1 Hz for all of these tests.

3.1.2 Variable axial vertical load

The basic test is performed at a number of different axial loads. The vertical load is constant during each test,
and tests are performed at loads of 0.50Pdes , 1.0Pdes and 1.50 Pdes , where Pdes is the vertical design load
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 529

Fig. 2 RNC isolator prototype with 5 mm rubber pads

of the small-scale prototype. The rate of loading and the strain increments are the same as for the basic test
described above in Sect. 3.1.1.

3.1.3 Variable frequency

The basic test is also performed at loading frequencies of 0.25, 0.50, 1.0, 2.0 and 4.0 Hz. The other test variables
are the same as those for the basic test.
530 M. Ismail et al.

Table 1 Summary of the performed numerical simulation tests

Rubber plates thickness [mm] Shear strain amplitude [%] Variable axial load
Variable loading frequency [Hz]

0.25 0.5 1.0 2.0 4.0

0.5 Pdes
0.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
5.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
10.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
1.0 Pdes
0.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
5.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
10.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
1.5 Pdes
0.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
5.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v
10.0 5 v v v v v
25 v v v v v
50 v v v v v
75 v v v v v
100 v v v v v

3.2 Unidirectional ultimate test

3.2.1 Monotonic shear test

This test is performed to examine the performance of the built-in buffer mechanism of the RNC isolator. A
100 % of shear strain amplitude brings the RNC isolator to a position at which the buffer mechanism starts
working. Therefore, the RNC isolator is subjected to a monotonic shear loading up to 140 % shear strain to
study the forcedisplacement relationship with the buffer engaged.
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 531

Table 2 Data points on the uniaxial stressstrain curve for the rubber material

Strain (m/m) Stress (MPa)

0.5 0.2383
0.3 0.1035
0.1 0.0275
0.0 0.0000
0.1 0.0111
0.3 0.0280
0.5 0.0409
0.7 0.0516
0.9 0.0610

3.2.2 Monotonic tension test

This test is performed to study the behavior of the metallic yield dampers under tensile loading. The RNC
isolator is subjected to a gradually increasing tension, at the same rate of loading of axial compression, until
a separation between the upper bearing plate and the rolling body is detected. This test quantifies the amount
of tension resistance, if any, provided by metallic yield dampers.

4 Numerical simulations

Using the nonlinear analysis software ADINA [12], numerical simulation and analysis are carried out with the
aim of finding the characteristics of the RNC isolator. That commercial software is selected to benefit from its
powerful solver as the RNC isolator is highly nonlinear. The main sources of nonlinearity encountered in the
RNC isolator model are as follows:
Contact nonlinearity, due to the direct contact between the elements in rolling contact.
Geometric nonlinearity, which is a result of large strains.
Material nonlinearity, which is a result of using hyperelastic (for neoprene plates) and plastic (for dampers)
materials.
The nonlinear stressstrain behavior of the hyperelastic neoprene plates is described by the Ogden material
model [26]. It is a hyperelastic material model used to describe the nonlinear stressstrain behavior of com-
plex materials such as rubbers, polymers and biological tissue at high strains. The Ogden model, like other
hyperelastic material models, assumes that the material behavior can be described by means of a strain energy
density function, from which the stressstrain relationships can be derived. These materials can generally be
considered to be isotropic, incompressible and strain rate independent. In the Ogden material model, the strain
energy density W is expressed in terms of the principal stretches j , j = 1; 2; 3 as:


N
p p
W (1 , 2 , 3 ) = (1 + 2 p + 3 p 3), (1)
p
p=1

where N, p , p , are material constants. The shear modulus results from


N
2 = pp. (2)
p=1

With N = 3 and by fitting the material parameters, the material behavior of rubbers can be described
accurately. In this paper, the curve fitting feature of ADINA is used to generate the rubber material constants
in the Ogden material model using the (given) data points listed in Table 2, on the stressstrain curve.
The plastic behavior of the mild steel and lead bars is represented using the bilinear model with a kinematic
strain hardening type. The material behavior is described by a bilinear total stresstotal strain curve starting
at the origin and with positive stress and strain values. The initial slope of the curve is taken as the elastic
modulus of the material E. At the specified yield stress C1 , the curve continues along the second slope defined
by the tangent modulus C2 . In this study, the mechanical characteristics of mild steel are as follows: E =
532 M. Ismail et al.

Fig. 3 Shear forcedisplacement hysteresis loop

2.1 108 kN/m2 , C1 = 240,000 kN/m2 , C2 = 1.0353 106 kN/m2 and Poissons ratio = 0.29. The lead
material has the following characteristics: E = 1.41 107 kN/m2 , C1 = 7,031 kN/m2 , C2 = 11,380 kN/m2
and Poissons ratio = 0.42. The remaining components of the RNC isolator model are represented by an
isotropic linear model.
Regarding the simulation limitations, they receive enough attention to obtain reasonably accurate results.
The computational efforts have been effectively addressed by the powerful solver of the used finite element
software ADINA; furthermore, the small size of the 1/10 reduced-scale prototypes produced a reasonably low
number of elements and nodes after meshing compared to what obtained from meshing the real scale model.
The boundary conditions are chosen to approach the real behavior of the RNC isolator under real loading
conditions. The lower bearing plate of the RNC isolator is assumed always fixed against rotation and displace-
ment. The upper bearing plate is allowed to move only vertically under gravity loads without rotation and then
released for only horizontal motion under cyclic horizontal displacement with no possibility to move again
vertically after the application of horizontal displacement nor to rotate. The problem of stress concentration
is avoided to improve the solution convergence by means of avoiding sharp edges in these highly nonlinear
simulation tests.

5 Test results

Due to the huge amount of outputs and space limitations, only a carefully selected set of results are presented
and discussed in this paper: they are relative to the mild-steel material for the dissipative bars. These selected
results are supposed to be adequate enough to enable drawing of proper conclusions and getting a reasonable
understanding of the RNC isolation system.
The forcedisplacement relationship of the RNC isolator is found to be of hysteretic nature. Typical hyster-
esis loops obtained under sinusoidal shear displacement loading are presented in Fig. 3. The loops are shown
for the first configuration RNC prototype at shear strains of 5, 25, 50, 75 and 100 % and at a loading frequency
of 4 Hz. At 5 % shear strain, the shear displacement is too small for the RNC isolator to be in the nonlinear
zone. It is also noticed that the post-yield stiffness slightly decreases as the shear strain amplitude increases.
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 533

Fig. 4 Effective stiffness versus shear strain

Fig. 5 Damping versus shear strain

5.1 Characteristic tests

The effective stiffness ke ff and damping of the RNC isolator are calculated from the obtained hysteresis
loops using the following relationships:

Fmax Fmin
ke ff = , (3)
dmax dmin
Aloop
= , (4)
2 Fmax dmax
where Fmax , Fmin are the peak values of the shear force; dmax , dmin are the peak values of the shear displace-
ment; Aloop is the hysteresis loop area.
These parameters are plotted as functions of the shear strain in Figs. 4 and 5. Figure 4 shows the variation
of the effective stiffness considering three sets of mild steel dampers against shear strain amplitudes up to
100 % and at a loading frequency of 1.0 Hz. From the figure, it seems that the significant influence of changing
the number of metallic yield dampers is limited to relatively small shear strains lower than 30 %. This means
534 M. Ismail et al.

Fig. 6 Effect of axial loads on the forcedisplacement relationship

Table 3 Effect of loading frequency on effective stiffness at different shear strains

Shear strain (%) Variable loading frequency (Hz)

0.25 0.5 1.0 2.0 4.0
5 1.8610 1.8608 1.8610 1.8611 1.8610
25 0.4693 0.4693 0.4693 0.4693 0.4693
50 0.2524 0.2524 0.2524 0.2524 0.2524
75 0.1775 0.1775 0.1775 0.1775 0.1775
100 0.1391 0.1391 0.1391 0.1391 0.1391

that increasing or reducing the number of dampers mainly raises or lowers, respectively, the pre-yield elastic
stiffness of the RNC isolator which is the main source of resistance to minor vibrations. On the other hand, it
seems evident from Eq. (4) that the damping ratio of the RNC isolator, at the same shear strain amplitude, is
linearly proportional to the number of metallic yield dampers . However, in Fig. 5 the damping ratio goes up
rapidly to a relatively high value of 47 % and then remains almost constant as the shear strain increases. This
less-expensive higher damping ratio provided by the low-cost metallic yield dampers has the advantage of
avoiding excessive bearing displacements under strong ground motions and to quickly damp down the vibra-
tions after the excitation comes to an end. A question about the source damping may arise when the vibration
amplitude decays after earthquake, making the dampers deform within their elastic range. The answer is that
damping is an inherent material characteristic that exists even if it vibrates elastically. Moreover, if the metallic
yield dampers are plastically stressed, their behavior will not be purely elastic because of the memory effect
of their past large deformation.
The influence of vertical axial load variation on the mechanical characteristics of the RNC isolator is inves-
tigated by subjecting the RNC prototypes to axial loads followed by a sinusoidal horizontal displacement for
three cycles. The axial loads are applied in the form of vertical displacement as percentages of 0.0, 50, 100 and
200 % of the design vertical displacement at shear strain of 50 % and a loading frequency of 2.0 Hz as shown in
Fig. 6. This figure shows that the case subjected to zero axial load before horizontal shear displacement, which
is plotted in solid black line, has higher pre-yield elastic stiffness, while the rest of its behavior is identical
to all the other cases subjected to variable vertical load before shear displacement as the hysteresis loops are
coincident. This means that except the pre-yield elastic stiffness, all the other mechanical characteristics of the
RNC isolator remain unaffected, under the same loading conditions, by pre-vertical loading prior to horizontal
shear displacement. The basic tests performed at variable loading frequencies and at different shear strains do
not show variations in the device effective stiffness or damping (Tables 3, 4).
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 535

Table 4 Effect of loading frequency on damping (%) at different shear strains

Shear strain (%) Variable loading frequency (Hz)

0.25 0.5 1.0 2.0 4.0
5 2.7718 2.7726 2.7718 2.7710 2.7716
25 42.7928 42.7928 42.7285 42.7276 42.8001
50 47.2598 47.2431 47.2444 47.1809 47.2501
75 47.7047 47.7047 47.7017 47.7223 47.7116
100 47.3897 47.3734 47.3734 47.3922 47.3922

Fig. 7 Ultimate shear displacement test result

5.2 Ultimate tests

A monotonic shear displacement up to 140 % of shear strain is applied to the second and third configuration of
the prototypes, those having neoprene plates, to examine the performance of the integrated buffer mechanism
of the RNC isolator. This mechanism forces the RNC isolator to stop its horizontal motion stroke at a predeter-
mined design displacement limit to avoid excessive displacements under low-probability strong earthquakes
as shown in Fig. 7b. The forcedisplacement relationship shown in Fig. 7a is studied at two different positions:
position A, Fig. 7c, where the rolling body becomes in just-in-touch position with the neoprene plates that
cover the inner faces of the vertical side walls of bearing plates; and position B, Fig. 7d, where the rubber plates
are extremely deformed under compression. Figure 7a shows a smooth variation of the shear force against
the normalized thickness of neoprene plates. It is found that the rate of compressibility of the neoprene plates
decreases under continuous application of compressive load. This develops a gradually increasing resistance
536 M. Ismail et al.

Fig. 8 Behavior of RNC isolator under vertical axial loading, tension and compression, considering dampers of different number
and material

to the applied shear displacement making the shear force to increase monotonically against a significantly
decreasing shear displacement.
The tensile capacity of the RNC isolator (if any) is provided by the slim bent metallic yield dampers. The
RNC isolator prototypes are subjected to a gradually increasing tension, at the same rate of loading of axial
compression, until a separation between the upper bearing plate and the rolling body is detected. The resulting
forcedisplacement curve is plotted against the case subjected to 1.50 Pdes axial compression considering two
sets of dampers and their material in Fig. 8. The figure shows that the bearing resistance to tension is completely
negligible. However, this does not represent a major drawback because only the shear-type structures are valid
for seismic isolation. This type of structures almost exhibits no overturning, and consequently, the isolation
bearings are not susceptible to tension.

6 Mathematical modeling

The restoring forcedisplacement relationship of the RNC isolator is hysteretic. Therefore, the versatile Bouc
Wen model [27,28] of smooth hysteresis is used to provide a simplified model that can predict the hysteretic
restoring force of the RNC isolator in a manageable form for design purposes. The standard form of the model
is expressed as:
FRNC (t) = kx(t) + (1 )D y kz(t), (5)
z (t) = D 1
y (A x(t)
|x(t)|
|z(t)|n1
|z(t)| ),
z(t) x(t) n
(6)
where x is the displacement, z is an auxiliary variable, FR N C is the isolator restoring force, kx(t) is the elastic
force component, the dot symbol on the variable denotes the time derivative, n > 1 is a parameter that governs
the smoothness of the transition from elastic to plastic response, D y > 0 is the yield constant displacement,
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 537

Fig. 9 Measured (FEM) restoring shear force versus the predicted one by the mathematical BoucWen model under a scaled real
ground motion

k > 0 and 0 < < 1 represents the post-yielding-to-pre-yielding stiffness ratio (kb /ke ), while A, and
are non-dimensional parameters that govern the shape and size of the hysteresis loop.
The BoucWen model can match a hysteretic behavior by properly tuning its parameters. It is a black-box
semi-physical model, that is, the model parameters may not have physical meanings. For this reason, the
BoucWen model has to fulfill some general physical properties. It was asserted by [29] that, for n 1, the
BoucWen model is bounded-input bounded-output, passive and consistent with physical asymptotic motion
if and only if
A > 0, + > 0, 0. (7)
In the case of having a physical system to be identified, actual measurements are normally used to perform
the model identification. When physical systems are not available, it is a common practice to use sophisticated
computer codes able to describe with accuracy the dynamic behavior as the true system in order to identify
simplified models that capture essential inputoutput properties for specific purposes. In this study the parame-
ters of the BoucWen model are identified by comparing the model output to the obtained forcedisplacement
from the ADINA finite element model. A constrained nonlinear least squares optimization algorithm, available
in Matlab [30], is used to obtain the seven model parameters in Eqs. (5) and (6).
Then the identified parameters are validated under a scaled real random seismic excitation (Kern Country
earthquake, Taft Lincoln school tunnel, July 21, 1952). Both outputs of the mathematical BoucWen model
and the finite element model are plotted in Fig. 9 using time and displacement scales considering the Kern
earthquake. A close agreement between both outputs makes the hysteretic BoucWen model an accurate
replacement of the RNC isolator for further applications on the bridge model.

7 Simulation of the RNC isolator on the bridge model

Testing of isolation devices is traditionally uniaxial because of the complexity of loading in two directions.
However, the earthquake response of an isolated bridge deck is two dimensional; hence, bidirectional charac-
538 M. Ismail et al.

Master node

Fixed constraints
Fig. 10 3D FE model of the dissipative bars

y y Deck (master node)

D y

A B F C A-H B-D E-G Tower

E F x x

C Tower
x

(a) (b) (c)

Fig. 11 Displacement paths in the horizontal plane (a, b), BoucWen analytical scheme (c)

terization of the devices is crucial in order to understand the behavior and evaluate the efficacy of the assumed
dissipative models [31]. In light of this observation, this section is devoted to deepen this aspect for the RNC
isolator device.
A 3D numerical model of the dissipative bars system has been developed to evaluate the response to bidi-
rectional excitations in the horizontal plane. The presence of the rolling body and related contact problems
between surfaces have been disregarded. The FE model has been subjected to several kinematic histories to
obtain displacementsforces curves. In turn, these have been assumed as the target behaviors to be reproduced
by the BoucWen law for further implementation on the bridge structural system.

7.1 Bidirectional FE tests

The only components of the RNC device (Fig. 1) which exhibit a dissipative hysteretic behavior are the metal-
lic bars surrounding the quasi-elliptical rotating body [1822]. The 3D model (Fig. 10) of the system of steel
damping bars has been developed in the commercially available FE ANSYS program [24]. To assess the effects
(if any) of mesh dependency, different finite element sizes and element types have been used.
These solutions have been tested by applying to the system the bidirectional displacement time histories
shown in Fig. 11a, b, with reference to [32]. The first branches of the displacement paths are unidirectional
to allow us to calibrate the BoucWen model parameters for uniaxial hysteretic cycles. The displacements
history has been assigned to the master node positioned in the center of the bars system (Fig. 10). The degrees
of freedom for the nodes of the damping bars are restrained with a body constraint to the master node, while
the base of each bar has been completely restrained. Rotations of the top surface can be neglected, and the
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 539

0.4
(a)
0.3

Fx [kN]
0.2

0.1

0
-25 -20 -15 -10 -5 0 5 10 15 20 25
F
-0.1

-0.2
3d ANSYS
-0.3
2BWElem.exe
-0.4
Ux [mm]

0.4
(b)
0.3
Fx [kN]

0.2

0.1

0
-25 -20 -15 -10 -5 0 5 10 15 20 25
F
-0.1

-0.2
3d ANSYS
-0.3
2BWElem.exe
-0.4
Ux [mm]

Fig. 12 x (a) and y (b) RNC characteristic, for displacement path in Fig. 11a, compared to results from two orthogonal BoucWen
elements

distance between the top and bottom surface of the system is constant. This implies that the master node
only translates in an horizontal plane. The nonlinear behavior of steel is assumed as bilinear with kinematic
hardening. The different finite element sizes and element types used to asses the eventual presence of mesh
dependency, triggered by nonlinear behavior of the materials, make reference: the first to Timoshenko beam
elements (ANSYS beam188 element) and the others to hexahedral solid elements (ANSYS solid45 element).
A reasonable equivalence in the results has been reached.

7.2 Calibration of the bidirectional BoucWen arrangement

The target behaviors resulting from the 3D FE characterizations are reproduced following the proposal by two
independent, orthogonal, elasto-plastic springs simulated by the BoucWen model (Fig. 11c). This has been
used in its classical formulation [27] ignoring the yielding displacement contribution (D y = 1 in Eq. 6). To
identify the BoucWen model parameters, a MATLAB routine has been used, considering the response to the
initial unidirectional branch of the displacement path and the following equation which relates the yielding
force [33,34]:
y K
FR N C = 1
. (7)
( + ) n
Figure 12 shows the fitting of the ANSYS 3D FE numerical simulations, for the initial (plane trajectory)
displacement path in Fig. 11a, with those of the BoucWen model characterized by the following identified
parameters: A = 1, = = 651605000 m1 , = 0.16, n = 4, k = 38 kNm1 . It is worth noting that
540 M. Ismail et al.

400
(a)
300

Fx [N]
200

100

0
-30 -20 -10 0 10 20 30
F
-100

-200
3d ANSYS
-300
2BWElem.exe
-400
Ux [mm]

400
(b)
300
Fy [N]

200

100

0
-30 -20 -10 0F 10 20 30
-100

-200
3d ANSYS
-300
2BWElem.exe
-400
Uy [mm]

Fig. 13 x (a) and y (b) RNC characteristic, for displacement path in Fig. 11b, compared to results from two orthogonal BoucWen
elements

all the simulations have been run in ANSYS framework, using a BoucWen user-element implemented as an
external executable program from within the main ANSYS analysis procedure. More details about the descrip-
tion of the procedure can be found in [25]. Figure 13 is devoted to the displacement trajectory in Fig. 11b. The
good agreement in matching the numerical and analytical results on the scaled prototype gives confidence in
using the bidirectional BoucWen model configuration to simulate the full-scale RNC behavior.

8 The bridge structure

The Bill Emerson Memorial Bridge is a 1,206-m-long, fan-type, cable-stayed structure (the main span length
is 350.6 m), and its main geometry is depicted in Fig. 14ac with some details.
The structure consists of 128 cables, two longitudinal stiffened steel girders and two towers in the cable-
stayed spans, and 12 additional piers in the Illinois approach span. The 128 cables are made of high-strength,
low-relaxation steel. The smallest cable area is 28.5 cm2 , and the largest cable area is 76.3 cm2 . All cables
behave elastically under a design earthquake. Their factor of safety is larger than 2.35 at all times [35]. On
the other hand, the cable subjected to least stress is always in tension, ensuring no slack occurrence during
the earthquake. Therefore, cable elements can be simplified with a linear mechanical behavior for seismic
analysis. Cables and bearings significantly influence the stiffness of the bridge system. The sagging of cables
should be considered in the modeling of the cable-stayed bridge to account for geometric nonlinear effects.
The towers have a solid section as the lower portion, below the cap beam, which is generally more critical
than the hollow section of the upper portion. For out-of-plane behavior, the upper portion of the towers above
the cap beams remains nearly elastic with a significant margin of safety. The lower portion of the towers,
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 541

Fig. 14 Bridge and main geometry (a) with deck (b) and piers (c) details. The ANSYS finite element mesh (d)

however, likely experiences moderate yielding out of plane during the design earthquake though the safety of
the bridge is not a concern. The in-plane behavior of the two towers is always in the elastic range under the
design earthquake, with a wide margin of safety.
The 3-D response and behavior of the cable-stayed bridge is described by the vibration modes which are
characterized by coupled shapes. The dynamic characteristics of the bridge also indicate that the cable-stayed
structure is more flexible in vertical direction and less in longitudinal direction [35]. More details about the
description of the structural members and the modal characteristics can also be found in [23,25].

8.1 The bridge FE model

The FE model is that of the Bill Emerson Memorial bridge, which has been the object of an international
control benchmark. The original statement [23] was focused on the bidirectional horizontal nature of the
seismic excitation, soilstructure interaction was not considered, and the same ground motion was applied
at each support, considering only the time delay due to the finite velocity of propagation. In the improved
FE model, developed in the ANSYS framework [25], the seismic input consists in five synthetic realizations
which have as target [36] the Eurocode 8 elastic response spectrum for soil of type B [37]. The vertical
component of the earthquake has been introduced, the delayed seismic input is not the same at all the supports,
but a coherence function of literature is introduced so that the signals satisfy a given correlation function. The
soil type regulates the correlation degree. The choice of five seismic realizations for investigating the seismic
behavior of the structure was justified following the minimum standard requirements from Eurocode 8 (Part
2) [37], which fixes to a minimum of three the suite of artificial accelerograms for structural seismic analysis.
The model (Fig. 14d) comprises soilstructure interaction by the use of impedance functions. So, the foun-
dations are simulated by lumped masses with equivalent springs and dampers; consequently, ground displace-
ments and velocities are required. Additional details on the simulation of the structural model by impedance
functions can be found in [38,39], while details on the seismic loading and the soilstructure interaction can
be found in [25]. Finally, the bridge is connected to the foundations by a master-slave procedure.
542 M. Ismail et al.

(a) EC8 Soil type B

Average
3

[m/s 2]
2

0
0 1 2 3
[s]
4

(b)
3
[m/s 2]

0
0 1 2 3
[s]
Fig. 15 Average pseudo-acceleration spectrum and Eurocode 8 elastic spectrum for the horizontal (a) and vertical (b) component

The original benchmark statement does not consider degrees of freedom for the stays, beside those of the
extreme nodes, neglecting their modal and dynamic description. Focusing the attention on the simulation of
the structural dynamics, the model for the stays is refined moving from the single rod type representation to a
description with six rope elements for each cable. The adopted ANSYS cable element was previously tested
in a tensioned setting, similarly to that of the bridge stay cable, and the results were in good agreement with
an analytical approach based on the transfer functions and Irvine theories [25].

8.2 The seismic input

The free-field ground displacements and velocities, needed by the lumped approach of SSI, have been obtained
from synthetic acceleration time histories generated following the approach presented in [40], considering the
spatial variability of the seismic ground motion. At all stations, the generated accelerations satisfy the well-
known Kanai-Tajimi Power Spectral Density, as modified in [41]. The statistical differences between the
motions at different stations satisfy the coherency function proposed in [42]. A velocity of the shear waves
vs = 3,000 m/s and an incoherency factor = 0.2 has been adopted.
The parameters of the Clough and Penzien power spectral density have been carefully chosen as described in
[36] in order to minimize the difference between the value of the median response pseudo-acceleration spectrum
with the Eurocode 8 elastic response spectrum for soil of type B [37]. Figure 15 shows the average pseudo-
acceleration spectrum, computed from 10 realizations, for the horizontal and vertical component, respectively.

8.3 Passive control strategy

In recent years, efforts have been made to enhance the technology of structural control for cable-stayed bridges.
Based on the benchmark results, the type of connection employed between the bridge deck and tower is one
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 543

Table 5 BoucWen model parameters

Set A Dy [m] n k[k N /m]

1 1.696 0.068 0.0008212 0.00001 0.01477 4.392 2226.2
2 1.696 0.068 0.0008212 0.00001 0.00738 4.392 1113.1
3 1.696 0.068 0.0008212 0.00001 0.00369 4.392 556.54
4 1.696 0.068 0.0008212 0.00001 0.00185 4.392 278.27
5 4.582 0.369 0.022 0.00001 0.00764 1.515 9001.2
6 4.582 0.369 0.022 0.00001 0.01529 1.515 18002
7 4.582 0.369 0.022 0.00001 0.01 1.515 5000
8 1.696 0.068 0.0008212 0.00001 0.01 4.392 5000
9 1.696 0.068 0.0008212 0.00001 0.01 4.392 50000
10 1.696 0.068 0.0008212 0.00001 0.01 4.392 100000
11 4.582 0.068 0.0008212 0.00001 0.01 4.392 100000

Table 6 Evaluation criteria. Percentage reduction of the proposed parameters

Criterion Response
H1 Maximum longitudinal mid-deck acceleration
H2 Maximum tower base bending moment in longitudinal direction
H3 Maximum tower base shear in longitudinal direction
H4 Maximum mid-deck displacement in longitudinal direction
H5 Maximum transversal mid-deck acceleration
H6 Maximum tower base bending moment in transversal direction
H7 Maximum tower base shear in transversal direction
H8 Maximum mid-deck displacement in transversal direction
H9 Maximum tower base torque

of the most important factors that affect the seismic performance of cable-stayed bridges [23]. A strong con-
nection will indeed reduce the displacements, but the shear force and bending moment at the base of the tower
will be increased significantly [43,44].
The proposed protection strategy decouples horizontally the bridge deck motion from that of the towers
and bent, while guaranteeing energy dissipation in every direction thanks to the hysteresis capacity and the
axial symmetry of the RNC control devices. In the vertical direction the link is supposed to be rigid. The RNC
passive devices are symmetrically located, with respect to the longitudinal axis under the bridge deck. Eight
devices connect the deck with Bent 1, Pier 2, Pier 3 and Pier 4.
The control capability, and eventually the limitations, of the numerically simulated RNC passive devices
has been investigated with the aim of identifying optimal values and checking the technical feasibility of
the RNC as a bridge control device. The results provided under five different seismic excitation histories
have been used, each one characterized by three components, spatial correlation and time delay into a multi-
support excitation arrangement of the bridge. RNC devices are simulated by resorting to an external executable
program (BWelem.exe) which integrates the equations of the BoucWen model at each step of the analysis.
This program is called from the main list file of ANSYS, and APDL routines are responsible for passing it to
the state of the structure and read back in, and introduce in the FE model the control forces [25].
Due attention is paid to the influence of introducing the RNC devices into the bridge structure by this
numerical tool. Being the connections by RNC devices between the deck and substructures, at the towers and
abutment zones, we control that the distance between piers and deck is bigger than the RNC devices struc-
tural buffer to detect hammering phenomena. Furthermore, the analysis time step has been chosen to avoid
introducing numerical errors in the simulation.

9 Selection of optimal parameters and seismic analyses on the ANSYS bridge model

Different sets of parameters, listed in Table 5, have been considered in a multi-parametric analysis of the
bridge for the BoucWen model summarized by Eqs. (5) and (6). The performance of the selected passive
control systems is evaluated by normalizing the response to that of the uncontrolled structure. Depending on
the structural response variable, the evaluation criteria in Table 6 are considered as percentage reduction in
extreme (worst) structural responses.
 544 M. Ismail et al.

Parameter sets 567 in Table 5 refer to RNC-b configuration, which present straight lead bars embedded
into the elastomeric ring. The remaining sets refer to RNC-a with curved steel bars surrounding the devices
core. RNC-a sets 14 are characterized by low stiffness values, while sets 811 by the highest ones. It is
worth noting that the elastomeric ring surrounding the rolling body, specifically designed for supporting heavy
masses, as those exerted by the bridge deck, contributes negligibly to the isolator damping. Instead, it is
implemented in the device essentially by the metallic bars, modeled by the BoucWen law [1822,33,34].
For aim of conciseness, only a carefully selected set of results (tower base internal actions) are presented
in the following. However, the others from Table 6 have been discussed. Additional quantitative details can be
found in [45].
Parameter H1 (maximum longitudinal mid-deck acceleration) is uniformly mitigated by all the RNC sets
under the five seismic realizations. Focusing now the attention on the transversal acceleration (H5 parameter),
the best performances are achieved by the lower stiffness sets (15). On the contrary, the deck displacements
(H4 and H8 ) results are amplified by the RNC implementation. This result should not surprise, being in good
agreement with the isolation control theory where the seismic effect mitigation is reached by paying the toll
of higher displacements with respect to the structural fixed configuration.
The tower base internal actions highlight the most interesting outcomes, and they are depicted as percent-
age of reduction by the histograms in Fig. 16. All the parameter sets associated with low stiffness and curved
bars perform satisfactorily under the five considered seismic realizations. In general, straight bars sets seem to
perform in a less satisfactorily way, particularly set 6, characterized by a medium/high pre-yielding stiffness
and a high post-yielding stiffness.
Figure 17 depicts the longitudinal and transversal tower base shear variation in time (17a, c) and frequency
domain (17b, d), relative to the static configuration of the bridge in the uncontrolled and controlled config-
uration under the seismic realization 1. The efficiency of the RNC isolation can be clearly appreciated, it
consists in the reduction in the response amplitude and in a decoupling effect between the superstructure and
the support. This last aspect is mainly in the longitudinal direction.
Finally, a significant reduction in internal actions at the base of the tower is then achieved; on the other
hand, an increasing deck displacements value is obtained due to the change of the bridges static scheme by
introduction of the RNC devices. Coming now to the different sets of RNC parameters, the lower horizontal
stiffness of the control system has provided good performances.

10 Conclusions

The performed numerical characterization studies in this paper lead to the following specific conclusions:
(i) The forcedisplacement relationship of the RNC isolator is of hysteretic nature.
(ii) Only the pre-yield elastic stiffness of the RNC isolator is affected by the axial vertical loading, while all the
other mechanical characteristics remain unchanged. This is important since axial load variation is always
expected in a bridge.
(iii) Changing the number of metallic yield dampers only slightly affects the effective stiffness at relatively
small shear strains but has no influence on the damping ratio at the same shear displacement amplitude.
(iv) The metallic yield dampers are a less-expensive source of a relatively higher damping for the RNC isolator.
(v) The integrated buffer mechanism of the RNC isolator helps avoiding excessive bearing displacement under
the low-probability strong earthquake.
(vi) The neoprene plates within the integrated buffer mechanism allow for a gradual and smooth decrease in
the shear displacement.
(vii) The tensile resistance of the RNC isolator is completely negligible without representing a major drawback
since only the shear-type structures, which exhibit no overturning failure, are valid for seismic isolation.
(viii) The BoucWen model of smooth hysteresis is an accurate replacement of the RNC isolator for future
analytical studies.
These conclusions give interesting physical insight after performing an extensive campaign of numerical tests.
This insight will be very useful in planning and conducting experimental tests. If the numerically and exper-
imentally obtained results are close enough, then numerical simulation will be adopted as an efficient and
cheaper, in terms of time and cost, approach to perform more future studies to optimize the design of RNC
isolators.
A multi-directional study of the hysteretic behavior for the RNC passive device has also been conducted
to find a way to reproduce such device by numerical tools. The RNC isolator has been first characterized
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 545

90
(a)
70

50

30

10

-10 1 2 3 4 5 6 7 8 9 10 11

-30

-50
70
(b)
60

50

40

30

20

10

0
1 2 3 4 5 6 7 8 9 10 11
-10

-20
70
(c)
60

50

40

30

20

10

0
1 2 3 4 5 6 7 8 9 10 11

20
(d)
15

10

0
1 2 3 4 5 6 7 8 9 10 11

-5
Paramenters sets
Excit. 1 Excit. 2 Excit. 3 Excit. 4 Excit. 5

Fig. 16 Percentage reduction with respect to the uncontrolled case at the tower base. H9 (a), H6 (b), H3 (c) and H7 (d).
Gray scale histograms show, respectively, realization from 1 to 5
546 M. Ismail et al.

x 10
4 Base shear - longitudinal x 10
7 Base shear - longitudinal
1 5
(a) (b)
0.5 4

0 3

|FFT|
[KN]

-0.5 2

-1 1

-1.5 0
0 5 10 15 20 25 0 0.5 1 1.5 2 2.5 3
[s] [Hz]

x 10
4 Base shear - transversal x 10
7 Base shear - transversal
1.5 5
(c) (d)
1
4

0.5
3
|FFT|
[KN]

0
2
-0.5

1
-1

-1.5 0
0 5 10 15 20 25 0 0.5 1 1.5 2 2.5 3
[s] [Hz]
Fig. 17 Tower base shear. Longitudinal and transversal directions, respectively (a) and (c), in time domain; (b) and (d), in
frequency domain. Continuous line: control set 1. Dashed line: uncontrolled bridge (seismic realization 1)

by numerical 3D analyses and subsequently reproduced by analytical simulation by two external nonlinear
springs, according to the BoucWen law, with satisfactory outcomes.
To assess the RNC performance in improving the seismic response of bridge structures, a refined version
of the ASCE bridge benchmark model has been studied inside the ANSYS commercial finite element code
framework. Multi-support dynamic excitation has been considered, which consists in five seismic records
applied to the bridge, and the soilstructure interaction has been modeled also. The positive contribution of the
new dissipative devices in the mitigation of the seismic effects has been assessed in terms of internal actions
and displacements of the structural components.

Acknowledgements This work has been partially supported by Generalitat de Catalunya under the technological project
Seismic Isolation (VALTEC-09-2-0022) and by MIUR (Italian Ministry of Education, University and Research) under the
project Dynamic response of linear and nonlinear structures: modeling, testing and identification (PRIN 2009).

References

1. Dange, Y., Gore, S.: Analysis of a nonlinear vibration isolator with the help of an analog computer. Indian J. Technol. 16,
282285 (1978)
2. Jiang, L., Stredulinsky, D., Szabo, J., Chernuka, M.: Numerical characterization of nonlinear stiffness properties of
pre-stressed vibration isolation mounts. Can. Acoust. Acoustique Canadienne 30, 7071 (2002)
3. Hundal, M., Parnes, P.: Response of a base excitation system with Coulomb and viscous friction. J. Sound Vib. 64,
371378 (1979)
4. Tsai, C.: Seismic behavior of buildings with fps isolators. In: ASCE Proceedings of the Second Congress on Computing in
Civil Engineering, Atlanta, GA, USA, vol. 2, pp. 12031211 (1995)
5. Tsai, C.: Finite element formulation for friction-pendulum seismic isolation bearings. Int. J. Numer. Methods Eng. 40,
2949 (1997)
6. Kim, M., Lim, Y., Cho, S., Cho, K., Lee, K.: Seismic analysis of base-isolated liquid storage tanks using the befebe
coupling technique. Soil Dyn. Earthq. Eng. 22, 11511158 (2002)
Characterization, modeling and assessment of Roll-N-Cage isolator using the cable-stayed bridge benchmark 547

7. Morman, K., Kao, B., Nagragaal, J.: Finite element analysis of viscoelastic elastomeric structures vibrating about nonlinear
statically stressed configurations. In: Proceedings of the 1981 society of automotive engineers international congress and
exposition. Detroit, MI, USA. Paper No. 811309 (1981)
8. Shangguan, W., Lu, Z.: Nonlinear modeling of hydraulic engine mounts of a car powertrain with computational fluid structure
interaction finite element analysis models. Chin. J. Mechan. Eng. 40, 8086 (2004)
9. Zhang, Y., Shangguan, W.: A novel approach for lower frequency performance design of hydraulic engine mounts. Comput.
Struct. 84, 572584 (2006)
10. Shangguan, W., Song, Z., Yun, Y., Jiang, K., Xu, C.: Experimental study and simulation analysis of hydraulic engine mounts
with multiple inertia tracks. J. Vib. Eng. 18, 318323 (2005)
11. Mutobe, R., Cooper, T.: Nonlinear analysis of a large bridge with isolation bearings. Comput. Struct. 72, 279292 (1999)
12. ADINA release 8.73 documentation. ADINA R & D, Inc., 71 Elton Avenuev Watertown, MA 02472, USA (2011)
13. Savage, I., Eddy, J., Orsolini, G.: Seismic analysis and base isolation retrofit design of a steel truss vertical lift bridge. Comput.
Struct. 72, 317327 (1999)
14. Imbimbo, M., De Luca, A.: F.e. stress analysis of rubber bearings under axial loads. Comput. Struct. 68, 3139 (1998)
15. Jerrams, S., Sanders, K., Goo, K.: Realistic modelling of earthquake-isolation bearings. J. Mater. Process. Technol. 118,
158164 (2001)
16. Tsai, C., Lu, P., Chen, W., Chiang, T., Yang, C., Lin, Y.: Finite element formulation and shaking table tests of direction-
optimized-friction-pendulum system. Eng. Struct. 30, 23212329 (2008)
17. Mordini, A., Strauss, A.: An innovative earthquake isolation system using fibre reinforced rubber bearings. Eng.
Struct. 30, 27392751 (2008)
18. Ismail, M., Rodellar, J., Ikhouane, F.: A seismic isolation system for supported objects. Spanish Patent No. P200802043,
Spanish Office of Patents and Marks (2008)
19. Ismail, M., Rodellar, J., Ikhouane, F.: A novel isolation bearing for light- to moderate-weight structures. J. Struct. Control
Health Monit. doi:10.1002/stc.421 (2009)
20. Ismail, M., Rodellar, J., Ikhouane, F.: An innovative isolation device for a seismic design. J. Eng. Struct. 32, 11681183 (2010)
21. Ismail, M., Rodellar, J., Ikhouane, F.: Performance of structure-equipment systems with a novel Roll-N-Cage isolation
bearing. Comput. Struct. 87, 16311646 (2009)
22. Ismail, M., Rodellar, J., Ikhouane, F.: An innovative isolation bearing for motion-sensitive equipment. J. Sound Vib. 326,
503521 (2009)
23. Caceido, J.M., Dike, S.J., Moon, S.J., Bergman, L., Turan, G., Hague, S.: Phase II benchmark control problem for seismic
response of cable stayed bridges. J. Struct. Control 10, 137168 (2003)
24. Ansys, Release 12.1, Ansys Inc, USA
25. Domaneschi, M., Martinelli, L.: Performance comparison of passive control schemes for the numerically improved ASCE
cable-stayed bridge model. Earthq. Struct. 3, 181201 (2012)
26. Ogden, R.: Large deformation isotropic elasticityon the correlation of theory and experiment for incompressible rubberlike
solids. Proc. R. Soc. Lond. Series A. Math. Phys. Sci. 326, 565584 (1972)
27. Wen, Y.: Method for random vibration of hysteretic systems. J. Eng. Mechan. Div. 102(EM2), 246263 (1976)
28. Ismail, M., Ikhouane, F., Rodellar, J.: The hysteresis Bouc-Wen model, a survey. J. Arch. Comput. Methods Eng. 16,
161188 (2009)
29. Ikhouane, F., Manosa, V., Rodellar, J.: Dynamic properties of the hysteretic Bouc-Wen model. Syst. Control Lett. 56,
197205 (2007)
30. The MathWorks, Inc. 3 Apple Hill Drive, Natick, Massachusetts. Release R2007b (2008)
31. Fenves, G.L., Huang, W.H., Whittaker, A.S., Clark, P.W., Mahin, S.A.: Modeling and characterization of seismic isolation
bearings. U.S. Italy Workshop on Seismic Protective Systems for Bridges (1998)
32. Abe, M., Yoshida, J., Fujino, Y.: Multiaxial behaviors of laminated bearings and their modeling. I: experimental study. J.
Struct. Eng. ASCE 130, 11191132 (2004)
33. Casciati, F.: Stochastic dynamics of hysteretic media. Struct. Saf. 6, 259269 (1989)
34. Domaneschi, M.: Simulation of controlled hysteresis by the semi-active Bouc-Wen model. Comput. Struct. 106-107,
245257 (2012)
35. Chen, G., Yan, D., Wang, W., Zheng, M., Ge, L., Liu, F.: Assessment of the Bill Emerson memorial cable-stayed bridge
based on seismic instrumentation data. Report No. OR08-003 Control. University of Missouri-Rolla (2007)
36. Martinelli, L., Barbella, G., Feriani, A.: A numerical procedure for simulating the multi-support seismic response of
submerged floating tunnels anchored by cables. Eng. Struct. 33, 28502860 (2011)
37. Eurocode 8: UNI ENV 1998, Design of structures for earthquake resistance (1998)
38. Fogazzi, P., Perotti, F.: On the seismic response of seabed anchored floating tunnels. Proceedings of XI European Conference
on Earthquake Engineering., Paris F. (1998)
39. Sieffert, J.G., Cevaer, F.: Handbook of impedance function. Ouest Editions, Presses Academiques 1992. ISBN 2-908261-32-4
40. Monti, G., Nuti, C., Pinto, P.E.: Nonlinear response of bridges under multisupport excitation. J. Struct. Eng. ASCE 122,
11471159 (1996)
41. Clough, R.W., Penzien, J.: Dynamics of structures. McGraw-Hill, New York (1975)
42. Luco, J.E., Wong, H.L.: Response of a rigid foundation to a spatially random ground motion. Earthq. Eng. Struct.
Dyn. 14, 891908 (1986)
43. Domaneschi, M.: Feasible control solutions of the ASCE benchmark cable-stayed bridge. Struct. Cont. Heath Monit. 17,
675693 (2010)
44. Li, H., Liu, J., Ou, J.: Seismic response control of a cable-stayed bridge using negative stiffness dampers. Struct. Cont. Heath
Monit. 18, 265288 (2011)
45. Carusone, G.: Controllo strutturale di ponti strallati mediante sistema RNC, (in Italian). MSc Thesis 2011; Politecnico di
Milano, Milan I (2011)