ArticlePDF Available

Concave sliding isolator’s performance under multi-directional excitation

Authors:

Abstract and Figures

Due to their large displacement capability and stable energy dissipation associated with a compact shape and new highly performing materials, the use of concave sliding isolators have been continuously increasing for application in buildings and bridges. In this paper the results of dynamic tests on full scale devices are presented. Their response was studied in a wide velocity range, for bi-directional patterns under different compressive loads. In this range of loading characteristics, which is typical of design for earthquake excitation, the behavior of these isolators appears significantly affected by the multi-directionality of the motion, and more specifically by the degradation of the coefficient of friction due to heating phenomena at the sliding surface. An analytical model, applicable to the prediction of bi-directional sliding behavior of friction-based isolators has been experimentally validated. Results of this study suggest that these phenomena should be considered in the design of structures equipped with these popular anti-seismic devices.
Content may be subject to copyright.
17
Anno XXX – N. 3 – luglio-settembre 2013
Concave sliding isolator’s performance under multi-directional
excitation
Giuseppe Lomiento*, Noemi Bonessio**, Gianmario Benzoni***
SUMMARY – Due to their large displacement capability and stable energy dissipation associated with a compact
shape and new highly performing materials, the use of concave sliding isolators have been continuously increasing
for application in buildings and bridges. In this paper the results of dynamic tests on full scale devices are pre-
sented. Their response was studied in a wide velocity range, for bi-directional patterns under different compressive
loads. In this range of loading characteristics, which is typical of design for earthquake excitation, the behavior
of these isolators appears significantly affected by the multi-directionality of the motion, and more specifically by
the degradation of the coefficient of friction due to heating phenomena at the sliding surface. An analytical model,
applicable to the prediction of bi-directional sliding behavior of friction-based isolators has been experimentally
validated. Results of this study suggest that these phenomena should be considered in the design of structures
equipped with these popular anti-seismic devices.
Keywords: Friction Pendulum System (FPS), seismic isolation, analytical model, multi-directional excitation.
of the isolator and vertical force variation. The model
implemented by Almazan et al. (1998) is capable of
simulating the uplift and impact of the slider in the
vertical direction. An analytical formulation including
the P-D effects was presented in Almazan et al. (2002),
which showed that for structures subjected to impulsive
ground motions the small-displacement models may
lead to discrepancies up to 20% in global response and
over 50% in local response such as the normal force in
the isolators or the inter-story deformations.
Specific studies were focused on the frictional be-
havior at the interface of the stainless steel surface
and sliding material. Most of the early studies, which
deal with fluoropolymers such as the PTFE, clearly
show the dependency of friction forces on the contact
pressure and the sliding velocity (Constantinou et al.
1990, Chang et al. 1990, Mokha et al. 1993; Soong
and Constantinou 1994; Bondonet and Filiatrault 1997).
Similar dependency was observed also for recently
manufactured sliding isolators that use hydrocarbon
high strength polymers as sliding materials (Lomiento
et al. 2011).
Recent development of testing facilities allowed test-
ing full scale isolators across a range of realistic ver-
tical loads and sliding velocities for multi-directional
excitations (Benzoni et al. 2011, Lomiento et al. 2012).
When compared with experimental results, the predic-
tion based on commonly accepted models of concave
sliding isolators appears sometimes inadequate. Based
on a large set of experimental results, Lomiento et al.
(2013) indicated that accounting for the friction deg-
radation induced by heat generated during the slid-
ing motion can be crucial for a correct assessment of
forces, displacements and energy dissipation induced by
earthquake excitations. The availability of experimental
results from bi-directional tests allows validating, under
bi-directional excitations, the friction model proposed
by Lomiento et al. (2013). With this goal the results
Introduction
The Friction Pendulum System (FPS) is currently
among the most widely used technology for seismic
isolation of bridges, buildings, and industrial facilities
due to its appealing features. The compact shape of
the sliding concave isolators, with considerably lower
height with respect to elastomeric bearings of simi-
lar capabilities, makes them extremely convenient for
retrofit applications. Their most relevant characteris-
tics include the large displacement capacity, which is
limited only by the in-plane geometric dimensions of
the device, and the capability of imposing an isolated
period to the structure that is primarily controlled by
the radius of curvature of the concave sliding surface.
Since the initial development and testing on a two-
story frame structure (Zayas et al. 1987), many experi-
mental and numerical studies have been conducted on
sliding concave isolators. Mokha et al. (1991) presented
the results of shake table tests on a ~1/3 downscaled
six-story steel moment-resisting frame in which the
ratio of height to maximum distance between bearings
was 2.25. Experimental results showed the effective-
ness of the FPS bearings in reducing inter-story drifts
and residual displacements. Force and displacement of
the isolators induced by mono-directional earthquake
excitations were predicted with an accuracy of ±20%
by the numerical model. Significant improvements to
the model were proposed by Tsai et al. (1997), which
developed a general finite element formulation for 3D
motion including local bending moments at the bottom
* Department of Structural Engineering, University of California San
Diego, La Jolla, California, USA, glomiento@ucsd.edu
** Department of Structural Engineering, University of California San
Diego, La Jolla, California, USA, nbonessio@ucsd.edu
*** Department of Structural Engineering, University of California San
Diego, La Jolla, California, USA, benzoni@ucsd.edu
18 Anno XXX – N. 3 – luglio-settembre 2013
Tab. 1. Geometric characteristics of the isolator.
Geometric dimension length (mm)
A435
a180
R2235
r375
T40
t120
h160
d = r – h 215
Tab. 2. Testing summary.
Test
#
Test
type
Vertical
load N
(kN)
Contact
pressure
pc
(MPa)
Peak
Long.
Displ.
Dlong
(mm)
Peak
Long.
Vel. Vlong
(mm/s)
Peak Lat.
Displ.
Dlat (mm)
Peak
Lat.
Vel. Vlat
(mm/s)
01 CL 3263 15 200 90 100 45
02 CL 6525 30 200 90 100 45
03 CL 13050 60 200 90 100 45
04 CL long.
comp. 6525 30 200 90 – –
05 CL lat.
comp. 6525 30 100 45
Legend: CL = cloverleaf test; pc = N/ra2, with N > 0 vertical compres-
sion load.
Fig. 1. (a) Schematic of concave sliding bearing (Lomiento et al. 2013); (b) typical test setup at the Caltrans SRMD shake table.
(a) Schema di un dispositivo a pendolo scorrevole (Lomiento et al. 2013); (b) tipica installazione di un dispositivo sulla tavola vibrante del laboratorio
Caltrans SRMD.
b
a
SLIDING CONCAVE SPHERICAL
SURFACE IN STAINLESS STEEL
STEEL SLIDER
SLIDING POLYMER
COMPOSITE LINER
STEEL PLATE WITH
HOUSING FOR THE
SLIDER
of experimental tests on a concave sliding isolator are
presented hereafter. The analysis of the experimental
results allows also an accurate study of the pendulum
mechanism as well as of the directionality of the fric-
tion forces.
Experimental campaign
The friction isolator under investigation is a single
pendulum device, consisting of an upper steel plate with
housing cap for the slider, a bottom plate with a concave
semi-spherical stainless steel surface, and a lentil-shaped
articulated slider lined by a high bearing capacity poly-
mer, as shown in Fig. 1(a). The full scale isolator was
tested at the Caltrans Seismic Response Modification
Device (SRMD) Laboratory at the University of Cali-
fornia San Diego, equipped with a 6 DOFs shake table
specifically designed for full scale testing of isolators
and energy dissipators (Benzoni and Seible 1998). The
displacement range of the table in longitudinal direction
is +/– 1.22 m with a maximum horizontal capacity of
9000 kN and a vertical load capacity of 53400 kN. The
peak velocity of the table longitudinal motion is 1.8 m/s.
The installation procedure of the device on the testing
machine was consistent with the standard installation
of isolation devices. The device bottom plate was con-
nected to the table and the top portion, above the slider,
was bolted to the vertical reaction beam that represents
a fixed reference, as shown in Fig. 1(b). In each test,
the table was raised to impose the vertical load to the
device and then commanded to the requested 3D motion.
The geometric characteristics of the isolator are re-
ported in Table 1, with all the labels graphically pre-
sented in Fig. 2.
Cloverleaf bi-directional tests were performed on
the isolator under three different vertical loads. The
main characteristics of the tests are specified in Table
2. Cloverleaf tests were used to check the accuracy of
Fig. 2. Geometric dimensions.
Dimensioni geometriche del dispositivo.
19
Anno XXX – N. 3 – luglio-settembre 2013
Fig. 4. (a) Deformed geometry; (b) kinematics of a friction pendulum
isolator.
(a) Configurazione deformata; (b) cinematica di un isolatore a pendolo
scorrevole.
a b
Fig. 3. Cloverleaf test: (a) in plane trajectory and (b) longitudinal and
lateral components of motion.
Test Cloverleaf (quadrifoglio); (a) percorso nel piano orizzontale;
(b)componenti di moto longitudinale e laterale.
a
the prediction model for bi-directional motion. Mono-
directional tests, with single cloverleaf components in
longitudinal and lateral direction, were also completed
for comparison with the related bi-directional tests.
The cloverleaf displacement pattern in the horizontal
plane is shown in Fig. 3(a). The applied displacement
time histories in the two perpendicular directions are
reported in Fig. 3(b). The loops of limited amplitude
applied at the beginning and the end of the run are
introduced to avoid excessive levels of acceleration.
Friction pendulum behavior
The geometry of the isolator (e.g. curvature and size
of the siding surface and the slider) and the character-
istics of the materials (e.g. pressure strength) as well
as characteristics of the system like the coefficient of
friction, concur together in controlling the displace-
ments and the forces experienced by the superstructure
under seismic loads. The friction pendulum behaviour
is commonly separated into two components (Zayas et
al. 1987): the restoring force Fr and the friction force
Ff. The resultant horizontal force Fh across the isolator
can be expressed in the general format as:
FiFiF
hrrff
$$=+
(1)
where ir and if are unit vectors expressing the direc-
tions of the restoring force Fr and the friction force
Ff. The restoring force is directly associated with the
pendulum behaviour of the isolator due to the curva-
ture of the sliding surface, while the friction force is
generated at the interface between the sliding surface
and the slider.
Pendulum behavior
The pendulum behavior is generated by the semi-
spherical shape of the sliding surface and aimed at pro-
ducing a lateral restoring force related to the vertical
load acting on the isolator. The lateral stiffness associ-
ated with the restoring force is a function of the vertical
load and the isolator’s geometry. The computation of
the restoring force requires the study of the kinematics
of the three components of the device, presented in Fig.
4(a) and Fig. 4(b). Horizontal radial displacement u and
vertical displacement w can be expressed as a function
of the position of the slider, identified by the angle i,
and the device’s geometry. Under the hypotheses of
null deformations of the isolator components, negligible
rotations of the top and bottom plates compared to the
angle i, and frictionless contact, closed form expression
can be obtained for the displacements u and w.
As shown in Fig. 4(b), the displacement pattern of
the upper plate of the isolator lies on a spherical sur-
face of effective radius (Fenz and Constantinou 2008):
RRd
eff
=+
(2)
In the original formulation of the pendulum be-
haviour proposed by Zayas (1987) and used in early
studies (Tsai 1997, Almazan et al. 1998), the physi-
cal dimensions of the slider have been neglected and
the effective radius has been approximated by the
curvature radius of the sliding surface R. In many
b
20 Anno XXX – N. 3 – luglio-settembre 2013
approximate expressions are generally higher than the
errors achieved when predicting the kinematics of the
isolator. However, the errors are significantly reduced
as the R/A ratio increases. For the device under study,
characterized by A R/5, the error at maximum dis-
placement is <1%.
As shown in Fig. 5(b), the lateral restoring force Fr
is always directed towards the geometric center of the
sliding concave surface. For a reference axes system
with origin at the center of the isolator, the vector of
the restoring force directions is given by:
iu
u
u
u
uu
u
11
ry
x
y
x
xy
22
$$
=
+
=
::
DD
(9)
In this model negligible imperfections of the con-
cave surface and eccentricities in the positioning of the
slider are assumed.
If Eqs. 8(b) and 9 are substituted into Eq. 1, the
contribution of the restoring force Fr is linear under
constant vertical loads with the slope given by the re-
storing stiffness Kr.
The frictional force contribution Ff is added to the
restoring force Fr to produce the generic performance
loop of Fig. 6(a). In Fig. 6(b), the restoring force con-
tribution is plotted over the experimental loops for the
mono-directional test 04.
recently manufactured isolators, the effective radius
Reff is noticeably different from the curvature radius
R, mainly due to the reduced thickness of the isola-
tors. For the isolator object of the present study, Reff
is ~10% higher than R.
Simplified closed-form expressions for the displace-
ments can be obtained by assuming the small-angles
approximation. For isolators with an in-plane radius
of the concave surface A < R/3, the errors associated
with this approximation are very limited (less than 3%
of the closed-form exact values), with maximum errors
achieved at maximum displacement. The exact and sim-
plified equations for the maximum allowable angle i,
the radial displacement u, and the vertical displacement
w are proposed in Table 3.
From the force balance of Fig. 5(a) the lateral re-
storing force Fr, the contact force Fc, the top Mrt and
bottom Mrb moments generated by the restoring force
can be expressed as functions of the vertical load N,
the displacement projection on the horizontal plane u
and the isolator’s geometry. Exact and simplified ex-
pressions for Fr, Fc , and the average contact pressure
pc are reported in Table 4. The lateral restoring stiffness
Kr is defined in Eq. 8.
It should be noted that the errors associated with the
Tab. 3. Exact and approximated equations for basic performance parameters.
Variable Exact expression Simplified expression Error for A<R/3
Maximum angle si
ns
in
R
A
R
a
max
11––
i=
``
jj
(3a)
max
i=<2% (3b)
Radial disp. sinuR
effi=(4a) uR
effi=<2% (4b)
Vertical disp.
()
coswR 1–
effi=(5a) /wR 2
2
effi=<3% (5b)
Tab. 4. Exact and approximated expressions of FPS forces.
Variable Exact expression Simplified expression Error for A<R/3
Contact force
cos
F
N
c
i
=(6a) Fc = N%
A
Aa
6
1(6b)
Contact pressure (/ )
cos
pN
a
aR
2
11
2
2
cir
=+(7a) p
a
N
2
2
c
r
= %
A
Aa
6
1(7b)
Restoring force cos
FKuR
Nu
rr
eff
$$
i
== (8a) FKuR
N
u
rr eff
$$
== %A
Aa
6
1(8b)
Fig. 5. (a) Force balance of a friction pendulum isolator; (b) in-plane
direction of the restoring force along a generic sliding path.
(a) Equilibrio delle forze; (b) direzione in pianta della forza di richiamo
per un moto generico dell’isolatore.
a b Fig. 6. Restoring force contribution in a mono-directional force-displa-
cement loop under constant vertical load: (a) schematic; (b) experimen-
tal data from test 04.
Ciclo forza-spostamento per una prova mono-direzionale a carico ver-
ticale costante: (a) forma schematica; (b) risultati sperimentali per il
test 04.
a b
21
Anno XXX – N. 3 – luglio-settembre 2013
Dz 4 · sin–1(h/Reff) (12)
where h is the height of the slider and Reff the effec-
tive curvature radius. For flat sliding surfaces (i.e.
Reff ), the angle shift becomes null and the fric-
tional force is parallel to the velocity vector. For the
isolator under investigation, an angular shift Dz 14°
was identified.
Due to the angular shift, the directional vector of the
frictional force is given by:
cos
sin
i
f
f
fz
z
=
^
^
h
h
=
G
(13)
where (/)
(/)
arctan cos
sin
2
f
vvv
vvv
zzz
zz
zz
zz
D
D
=+
+
oo
oo
eo
is the angle
of the friction force and arctan 2
x
y
o
o
z=
o
cm
is the angle
of the velocity with ox and oy the orthogonal compo-
nents of the velocity vector. Arctan2 is the arctangent
function with two arguments, which can distinguish
between diametrically opposite directions.
According to the simplified Coulomb theory for slid-
ing bodies, the frictional force is directly proportional
to the applied load and opposed to the sliding motion.
The sliding coefficient of friction is considered inde-
pendent by the effective contact area and the sliding
velocity. It is however experimentally demonstrated that
this theory cannot be applied to concave sliding devices
It must be noted that the stiffness based on the ra-
dius R is higher than the actual stiffness obtained by
using the effective radius Reff (Eq. 8 (b)). The restoring
force based on Reff appears in good agreement with the
experimental evidence as documented by the upper and
lower portion of the loop of Fig. 6(b).
Associated with the restoring force, moments Mrt
and Mrb are generated at the top and the bottom of
the isolator, respectively. These moments are origi-
nated by the eccentricity of the contact force with
respect to both the top and bottom plates. From the
rotational equilibrium, the following equations are
obtained:
MFeFrt
nrrt r
$$==+
^h
(10)
MFeFRT
rb rrbr
$$
==+
^h
(11)
where ert and erb are the lever arms of Fr from the top
and the bottom plate, respectively (Fig. 4). It should be
noted that generally erb > ert (2275 mm and 345 mm for
the current device). Even if moments at the top plate
can sometimes be neglected, the moment at the bottom
plate could be significantly high and should be included
in the analysis of the structure supporting the isolation
system. The maximum moments reached during test 04
are as high as Mrb = 1192 kNm and Mrt = 180 kNm.
This type of isolators are also quite often installed in a
reverse configuration, with the housing cap at the bot-
tom. In this occurrence also the moments are reversed,
with the higher moment at the top of the isolator.
Frictional behaviour
The top and bottom frictional forces acting on the
slider create an overturning moment, which is balanced
by a shift of the contact forces Fc with respect to the
theoretical position of Fig. 5(a). In the present study,
this misalignment is neglected for simplicity.
The in-plane eccentricity of the frictional forces
generates a torsional moment that can results in a
rotation of the slider about its normal axis acting si-
multaneously with the translational sliding motion.
Due to the concurrence of rotational and translational
movements, the frictional contribution to the resisting
force is not parallel to the projection of the veloc-
ity vector on the horizontal plane. In Fig. 7(a), the
angular shift of the frictional force with respect to
the velocity vector is labeled as Dz. In Fig. 7(b) the
direction of the frictional force, the velocity vector
and the restoring force are plotted for a given instant
of the cloverleaf test 02.
In Fig. 8 the angles z of both the velocity and the
force vector, from the reference direction of motion, are
compared. The angle shift at each instant is given by
the difference between the angle of the frictional force
and the angle of the velocity vector and was found
approximately constant during all the cloverleaf tests.
Based on experimental results the angular shift can be
calculated as:
Fig. 7. Angular shift of the frictional force with respect to the velocity
vector: (a) schematic; (b) experimental data from test 04.
Differenza angolare tra il vettore della forza di attrito e il vettore velo-
cità: (a) schema; (b) risultati sperimentali per il test 04.
a b
Fig. 8. Directional angle z of friction force and velocity for test 04.
Angoli z dei vettori velocità e forza, misurati rispetto alla direzione
longitudinale del moto per il test 04.
22 Anno XXX – N. 3 – luglio-settembre 2013
4. “velocity effect”, i.e. the variation of the coef-
ficient of friction with the sliding velocity.
By removing the restoring force contribution from
the experimental data, these effects are clearly visible
in the trends of the coefficient of friction from both
mono-directional and bi-directional tests. For instance,
the variation of the coefficient of friction n for the clo-
verleaf tests 01 and 03 is plotted versus time in Fig. 9
where the dashed lines indicate the mean and standard
deviation trends. Velocity trends are plotted right below
the coefficient of friction.
From the comparison of Fig. 9(a) and 9(b), the
load effect is recognized as a global reduction of n
due to the increased contact pressure from test 01
(pc = 15 MPa) to test 03 (pc = 60 MPa). An aspect that
appears remarkably important in both tests is the con-
tinuous degradation of the coefficient of friction. The
average reduction is due to the cycling effect and is
proportionally higher in test 03 than in test 01 (56%
versus 34% of the initial value) because of the higher
contact pressure and consequent more intense heating.
In addition to this continuous degradation, a fluctuation
of n between relative maximum and minimum values
is also occurring. This fluctuation of the coefficient of
friction is due to a combination of the velocity effect
(increment of n as the sliding velocity increases), and
a heating effect (reduction of n due to local high tem-
perature values). The reduction due to the local heating
overcomes the increment due to the velocity effect. For
this reason, the coefficient of friction is reduced even
at peak sliding velocity occurring at zero displacement
for cloverleaf tests. An experimental datum that con-
firms this statement is the standard deviation v of the
fluctuation, which is higher in the high pressure test
(~15% of the average values) than in the low pressure
test (~12%) because of the higher heat flux. A similar
fluctuation was noticed also for mono-directional tests
(Lomiento et al. 2013), but appears more significant for
bi-directional tests.
where the coefficient of friction is affected by breaka-
way transition phases, contact pressure, sliding veloc-
ity, and temperature rise due to the heat flux generated
across the sliding interface. These factors directly af-
fect the shape of the friction coefficient-displacement
loops, accordingly to four effects described in detail in
Lomiento et al. (2013):
1. “breakaway effect”, i.e. the sudden increase of
coefficient of friction n at the beginning of the motion
or at each motion reversal;
2. “load effect”, i.e. the reduction of the coefficient
of friction for increasing contact pressure;
3. “cycling effect”, i.e. the continue reduction of
the coefficient of friction with the repetition of cycles,
more pronounced for high-pressure and high-velocity
sliding motion, due to the temperature rise at the slid-
ing surfaces;
Fig. 10. Coefficient of friction and velocity for mono-directional and
bi-directional tests: test 02, 04, and 05: pc = 30 MPa.
Andamento del coefficiente d’attrito e della velocità nel tempo per prove
mono-direzionali e bi-direzionali: test 02, 04, and 05: pc = 30 MPa.
Fig. 9. Coefficient of friction and velocity trends for (a) CL test 01: pc = 15 MPa; (b) CL test 03 pc = 60 MPa.
Andamento del coefficiente d’attrito e della velocità nel tempo per (a) test 01 CL: pc = 15 MPa; (b) CL test 03 pc = 60 MPa.
a b
23
Anno XXX – N. 3 – luglio-settembre 2013
The breakaway effect appears more pronounced in
mono-directional tests where full motion reversal oc-
curs frequently, rather than in bi-directional tests. This
is visible in Fig. 10, where the coefficient of friction
from mono-directional tests 04 and 05 (longitudinal
and lateral component of cloverleaf tests) are compared
with the results from the related bi-directional test 02.
Spikes due to breakaway effects in the coefficient of
friction are only visible for mono-directional tests.
In this study, the friction force is obtained through
the phenomenological model proposed in Lomiento et
al. (2013) which considers the friction coefficient of
n as the product of three independent contributions:
(, ,) () () ()NC fN fC f
NC
$$no o=
o (14)
where fN, fC, fo are functions that take into account the
load, cycling and velocity effects, respectively, N > 0
is the compressive vertical load, o the sliding velocity
and C the cycling variable defined as:
()Ct A
Nd
t
2
22
2
t
t
0
ar
o=# (15)
The cyclic variable has the dimensions of a heat
flux, i.e. the rate of heat transfer per unit cross-sec-
tional area. It represents the temperature effects under
the hypothesis of uniform distribution of the heat flux
on the sliding surface. This assumption, introduced here
for simplicity, disregards the existence of higher heat
fluxes in areas interested by more frequent and faster
sliding activity. A more complete description of the
phenomenon can be found in Lomiento et al. (2013).
The three components of Eq. 14 can be modeled as
reported in Table 5.
The experimental parameters listed in Table 5 can
be identified from laboratory tests under different ver-
tical loads and varying velocity. For the specific de-
Tab. 5. Components of the friction model.
Effect Contribution Experimental parameters
Load fN(N) = ns0 · eN/Nref (16) ns0 = null-load coefficient of friction
Nref = load associated to a 63% friction reduction due to load effects
Cycling fC(C) = e–(C/Cref)b(17)
Cref = C associated to a 63% friction reduction due to cycling effects (reference heat flux,
dimensions of power per unit surface)
b = exponential rate of the friction degradation
Velocity fo(o) = c + (1 – c) · e|o|/ore f (18)
c = fast/slow motion friction coefficient ratio
oref = o associated to a 63% variation of the slow motion coefficient of friction due to ve-
locity increment
Tab. 6. Values of the experimental parameters for the isolator under
investigation.
Effect Parameter Value
Load
ns0 0.103
Nref 12300 kN
Cycling
Cref 9700 kN/ms
b0.5
Velocity
c1.40
oref 10 mm/s
Fig. 11. Coefficient of friction in mono-directional tests: (a) test 04 CL
longitudinal component; (b) test 05 CL lateral component.
Coefficiente d’attrito per prove mono-direzionali: (a) test 04 CL com-
ponente longitudinale; (b) test 05 CL componente laterale.
a
b
Fig. 12. Force-displacement loops in mono-directional test 04 CL lon-
gitudinal component: (a) prediction without cycling; (b) prediction with
cycling.
Cicli forza-spostamento per il test mono-direzionale 04 CL in direzione
longitudinale: (a) risposta calcolata senza l’effetto di ripetizione ciclica;
(b) risposta calcolata includendo l’effetto di ripetizione ciclica.
a
b
24 Anno XXX – N. 3 – luglio-settembre 2013
vice under study the values obtained are summarized
in Table 6.
Comparison of Numerical And Experimental Results
Mono-Directional Tests
The experimental variation of the coefficient of fric-
tion during mono-directional tests 04 and 05 is pre-
sented in Fig. 11. In order to highlight the contribu-
tion of the cycling effect two prediction models are
proposed. Neglecting the phenomena activated by the
continuation of the motion (prediction-no cycling) the
expected values of n are in substantial disagreement
with the experimental results.
The degradation of n with time, introduced by cy-
cling, is particularly noticeable along the longitudinal
component of motion (Fig. 11(a)) due to higher veloc-
ity and displacements.
Force-displacement loops predicted with and without
cycling are presented in Fig. 12(a) and 12(b) for the
longitudinal component, and in Fig. 13(a) and 13(b)
for the lateral component.
The application of the model to mono-directional
tests confirms the good agreement between experi-
mental evidence and model results when including the
cycling effects.
Bi-Directional Tests
The coefficient of friction predicted with and without
cycling, for bi-directional tests 01 and 03, is presented
in Fig. 14(a) and 14(b), respectively.
The high pressure test 03 appears more affected
by the friction variation than the low pressure test 01.
The model appears able to describe the cycling effect
but a more sophisticated model is required to simulate
the friction fluctuations.
For the cloverleaf test 02 the degradation and the
fluctuation of n (Fig. 15) appeared more distinct than
in the related longitudinal and lateral mono-directional
test of Fig. 11. This could be ascribed to an increased
significance of the cycling effect for multi-directional
motions.
The numerical and experimental force-displacement
loops for the two mono-directional components of test
02 are compared in Fig. 16 and Fig. 17.
Forces in the tests at the maximum displacement
are overestimated by 48% if the cycling effect is not
considered in the prediction. Even though a compre-
hensive model should include also the variation of the
coefficient of friction due to local heating, the inclu-
sion of the cycling effect appears to allow reasonable
estimates of forces in concave sliding devices.
The force-displacement loops for bi-directional mo-
tion of Fig. 18 appears significantly narrower than the
corresponding loops from mono-directional tests. This
is due to the augmented reduction of the coefficient
of friction due to thermal effects and the coexistence
of sliding motion components in two directions. The
Fig. 13. Force-displacement loops in mono-directional test 05 CL lateral
component: (a) prediction without cycling; (b) prediction with cycling.
Cicli forza-spostamento per il test mono-direzionale 05 CL in direzione
laterale: (a) risposta calcolata senza l’effetto di ripetizione ciclica; (b)ri-
sposta calcolata includendo l’effetto di ripetizione ciclica.
a
b
Fig. 14. Coefficient of friction in (a) CL test 01; (b) CL test 03.
Coefficiente d’attrito (a) test 01 CL; (b) test 03 CL.
a
b
Fig. 15. Coefficient of friction for CL test 02.
Coefficiente d’attrito per test 02 CL.
25
Anno XXX – N. 3 – luglio-settembre 2013
top plate (housing side) and should be taken into ac-
count for design purposes. For the isolator under in-
vestigation, the maximum moments observed were as
high as Mrb = 1192 kNm and Mrt = 180 kNm.
change in the loop shape is associated with a signifi-
cant reduction, for bi-directional tests, in the energy
dissipation. The energy dissipated in longitudinal direc-
tion reduces from 562 kNm for the mono-directional
test to 430 kNm for the same component of the bi-
directional test (reduction of 23%). In the lateral di-
rection a reduction of 41% was observed. The device
response obtained from mono-directional tests can thus
over-estimate the real capability of dissipation with the
consequence of potential under-estimate of peak dis-
placements during a seismic event.
Conclusions
An analytical model applicable to the prediction of
bi-directional sliding behavior of friction pendulum
isolators was experimentally validated. Bi-directional
cloverleaf tests and mono-directional tests representing
the longitudinal and lateral component of cloverleaf
tests were performed on a full scale single pendulum
concave sliding isolator under constant vertical loads.
The analytical model predicted force-displacement
loops that were in good agreement with the experi-
mental ones.
The experimental results confirmed the accuracy of
simplified expression to predict the restoring stiffness
of the isolator if the effective radius Reff is used instead
of the actual radius of curvature of the sliding surface
R. Associated with the restoring force, moments Mrt
and Mrb are generated at the top and the bottom of the
isolator, respectively. Moments at the bottom (sliding
surface side) are generally higher than moments at the
Fig. 16. Force-displacement loops in longitudinal direction from test 02
CL: (a) prediction without cycling; (b)prediction with cycling.
Cicli forza-spostamento per il test 02 CL in direzione longitudinale:
(a) risposta calcolata senza l’effetto di ripetizione ciclica; (b) risposta
calcolata includendo l’effetto di ripetizione ciclica.
a
b
Fig. 17. Force-displacement loops in lateral direction from test 02 CL:
(a) prediction without cycling; (b) prediction with cycling.
Cicli forza-spostamento per il test 02 CL in direzione laterale: (a) rispo-
sta calcolata senza l’effetto di ripetizione ciclica; (b)risposta calcolata
includendo l’effetto di ripetizione ciclica.
a
b
Fig. 18. Experimental force-displacement loops from bi-directional and
mono-directional tests: (a) longitudinal direction; (b) lateral direction.
Cicli sperimentali forza-spostamento per test mono e bi-direzionali:
(a) componente longitudinale; (b) componente laterale.
a
b
26 Anno XXX – N. 3 – luglio-settembre 2013
quencies”. Journal of Bridge Engineering. 2:4,
139-148.
Chang J.C., Hwang J.S. and Lee G.C. (1990). “Analyti-
cal model for sliding behaviour of Teflon-stainless
steel interfaces”. Journal of Engineering Mecha-
nics, 116(12), 2749-2763.
Constantinou M.C., Mokha A. and Reinhorn A. (1990).
“Teflon Bearings in Base Isolation, Part II: Mo-
deling”. Journal of Structural Engineering. 116:2
455-474.
Fenz D.M. and Constantinou M.C. (2008). “Mechani-
cal behavior of multi-spherical sliding bearings”.
Report No. MCEER 08-07, Multidisciplinary
Center for Earthquake Engineering Research,
Buffalo, NY.
Lomiento G., Bonessio N. and Benzoni G. (2011). “Ex-
perimental Performance and Modeling of Sliding
Anti-Seismic Devices”. 7th World Congress on
Joints, Bearings, and Seismic Systems for Concrete
Structures, American Concrete Institute.
Lomiento G., Bonessio N., Benzoni G. (2012). “Effects
of Loading Characteristics on the Performance of
Sliding Isolation Devices”. 15th World Conference
on Earthquake Engineering, Lisbon, Portugal.
Lomiento G., Bonessio N. and Benzoni G. (2013).
“Friction Model for Sliding Bearings Under Sei-
smic Excitation”. Journal of Earthquake Enginee-
ring. (DOI:10.1080/13632469.2013.814611).
Mokha A., Constantinou M.C., Reinhorn A.M. and
Zayas V. (1991). “Experimental study of friction
pendulum isolation system”. Journal of Structural
Engineering. 117:4, 1201-1217.
Mokha A., Constantinou M.C. and Reinhorn A.M.
(1993). “Verification of friction model of teflon
bearings under triaxial load”. Journal of Structural
Engineering. 119:1, 240-261.
Soong T.T. and Constantinou M.C. (1994). Passive and
active structural vibration control in civil enginee-
ring, Springer, New York.
Tsai C.S. (1997). “Finite element formulations for fric-
tion pendulum seismic isolation bearings”. Inter-
national Journal for Numerical Methods In Engi-
neering. 40:1, 29-49.
Zayas V., Low S. and Mahin S. (1987). The FPS
earthquake resisting system. Report No. CB/EERC-
87/01, Earthquake Engineering Research Center,
University of California, Berkeley, California.
The tests evidenced the significant variations of the
coefficient of friction generated by heating during the
sliding motion. These variations are more pronounced
for high contact pressure and peak velocity of the mo-
tion. The continue reduction of the coefficient of fric-
tion with the repetition of cycles (cycling effect) and
fluctuations due to the sliding motion over areas at
different temperatures are visible for all the tests. A
reduction up to 56% of the initial coefficient of friction
was observed for the highest pressure test. Bi-direc-
tional tests, characterized by higher velocity values,
produced a higher reduction of the friction coefficient
than mono-directional tests. Neglecting the cycling ef-
fects and the bidirectional interaction could result in
significant over-estimate of the frictional characteristic
and capacity of energy dissipation. The energy dissi-
pated in cloverleaf bi-directional tests was 23% and
41% lower than the energy dissipated in longitudinal
and lateral mono-directional tests, respectively. For the
above mentioned reasons a comprehensive design of
structures equipped with this technology should be sup-
ported by a complete set of experimental data.
References
Almazan J.L. and De la Llera J.C. (2002). “Analyti-
cal Model of Structures with Frictional Pendulum
Isolators”. Earthquake Engineering and Structural
Dynamics, 31: 305-332.
Almazan J.L., De la Llera J.C. and Inaudi J.A. (1998).
“Modeling Aspects of Structures Isolated with the
Frictional Pendulum System”. Earthquake Engine-
ering and Structural Dynamics, 27: 845-867.
Benzoni G. and Seible F. (1998). “Design of The
Caltrans Seismic Response Modification De-
vice (SRMD) Test Facility”. USA – ITALY
Workshop on Protective Systems. Report No.
MCEER-98-0015, Multidisciplinary Center for
Earthquake Engineering Research, Buffalo, New
York City.
Benzoni G., Bonessio N., Lomiento G. (2011). “Testing
protocols for seismic isolation systems” in Procee-
dings of 14th Conference of Seismic Engineering,
ANIDIS 2011, Bari.
Bondonet G. and Filiatrault A. (1997). “Frictional Re-
sponse of PTFE Sliding bearings at Higher Fre-
27
Anno XXX – N. 3 – luglio-settembre 2013
Comportamento di isolatori a pendolo scorrevole soggetti
ad eccitazione multi-direzionale
G. Lomiento, N. Bonessio, G. Benzoni
SOMMARIO – L’utilizzo degli isolatori a pendolo scorrevole è in continua crescita in ragione della loro elevata ca-
pacità di spostamento e della loro stabile dissipazione di energia che, unitamente al loro ingombro limitato anche
dovuto al recente impiego di materiali ad alta prestazione, li rendondo particolarmente adatti per applicazioni sia
su edici sia su ponti. Nel presente lavoro si presentano i risultati di prove dinamiche su un isolatore a pendolo
scorrevole in scala reale. La risposta del dispositivo è stata studiata tramite prove bi-direzionali a velocità variabile
sotto diversi carichi di compressione. Nell’ambito delle prove effettuate, rappresentative di situazioni comunemente
riscontrabili in eccitazioni sismiche assimilabili a terremoti di progetto, il comportamento di questi isolatori appare
signicativamente inuenzato dalla multi-direzionalità del moto ed, in particolare, dalla degradazione del coefficiente
di attrito conseguente all’incremento di temperatura delle superici di scorrimento per effetto del moto. Un modello
analitico per la previsione del comportamento in caso di moto bi-direzionale è stato validato sulla base dei risultati
sperimentali. I risultati di questo studio suggeriscono l’importanza dell’utilizzo di specici modelli, in grado di tener
conto della multi-direzionalità del moto, nella progettazione di strutture munite di tali dispositivi antisismici.
Parole chiave: Isolatori ad attrito, isolamento sismico, modello analitico, eccitazione multi-direzionale.
Introduzione
I sistemi di isolatori a pendolo scorrevole sono at-
tualmente tra le tecnologie più ampiamente usate per
l’isolamento sismico di ponti, edici e costruzioni indu-
striali in ragione di alcune caratteristiche che li rendono
estremamente convenienti. In particolare, il limitato in-
gombro di tali dispositivi, la cui altezza è notevolmente
inferiore a quella di isolatori in gomma con simili carat-
teristiche prestazionali, li rende particolarmente adatti
per l’adeguamento sismico di strutture esistenti. Tra le
caratteristiche più importanti degli isolatori a pendolo
scorrevole si possono annoverare l’elevata capacità di
spostamento, limitata solo dalle dimensioni in pianta del
dispositvo, e la capacità di imporre un determinato pe-
riodo di isolamento alla struttura sovrastante, funzione
principalmente del raggio di curvatura della supercie
concava di scorrimento.
Dal loro sviluppo iniziale e dalla sperimentazione
preliminare condotta su una struttura a telaio a due
piani (Zayas et al. 1987), sono stati condotti numerosi
studi sperimentali e numerici sugli isolatori a pendolo
scorrevole. Mokha et al. (1991) hanno presentato i ri-
sultati di prove su tavola vibrante condotte su un telaio
a sei piani in acciaio in scala ~ 1/3 in cui il rapporto
tra altezza dei pilastri e massima luce tra i dispositivi
era 2.25. I risultati di questo caso studio hanno dimo-
strato l’efficacia degli isolatori a pendolo nel ridurre
spostamenti interpiano e nel controllare gli spostamenti
residui. Una precisione di ± 20% rispetto al modello
numerico è stata riscontrata nella previsione di forze
e spostamenti indotti sugli isolatori da eccitazioni si-
smiche mono-direzionali. Signicativi miglioramenti dal
punto di vista della modellazione sono stati proposti da
Tsai et al. (1997), che hanno sviluppato una formula-
zione generale agli elementi niti per strutture soggette
a movimenti tridimensionali nella quale sono inclusi gli
effetti dei momenti ettenti agenti sulla porzione infe-
riore dell’isolatore e della variazione di forza verticale. Il
modello proposto da Almazan et al. (1998) è in grado di
simulare il sollevamento ed il successivo impatto in di-
rezione verticale. Una formulazione analitica per tenere
conto degli effetti P-D è stata successivamente presen-
tata da Almazan et al. (2002), che hanno dimostrato
come, per strutture soggette a movimenti impulsivi del
terreno, i modelli che assumono l’equilibrio valutato su
piccoli spostamenti possono condurre a differenze del
20% sulle caratteristiche globali di risposta del sistema
di isolamento e di oltre il 50% sulle caratteristiche di
risposta locale, quali la forza normale negli isolatori o
le deformazioni inter-piano.
Numerosi studi specici sono stati condotti sul com-
portamento attritivo all’interfaccia tra la supercie in ac-
ciaio lucidato a specchio ed il materiale a basso attrito
responsabile dello scivolamento. Nella gran parte, gli
studi sviluppati nei primi anni di utilizzo della tecnologia
si sono occupati sostanzialmenti di uoropolimeri come
il PTFE e hanno mostrato chiaramente la dipendenza
delle forze di attrito dalla pressione di contatto e dalla
velocità di scorrimento (Constantinou et al. 1990, Chang
et al. 1990, Mokha et al. 1993, Soong e Constantinou
1994, Bondonet e Filiatrault 1997). Una dipendenza da-
gli stessi parametri è stata osservata anche per mate-
riali di più recente utilizzo, costituiti da polimeri ad alta
resistenza (Lomiento et al. 2011).
I più moderni laboratori di prova consentono ora-
mai di testare isolatori in scala reale soggetti ad una
vasta gamma di carichi verticali e velocità di scorri-
mento, in condizioni rappresentative di condizioni di
progetto, per eccitazioni multidirezionali (Benzoni et al.
2011, Lomiento et al. 2012). Dal confronto con i risultati
sperimentali, la previsione basata su modelli comune-
mente accettati per gli isolatori a pendolo scorrevole
risulta talvolta inadeguata. Sulla base di un ampio nu-
mero di risultati sperimentali, Lomiento et al. (2013)
hanno mostrato che tener conto della degradazione
dell’attrito per incrementi di temperatura generati dal
moto di scivolamento può risultare cruciale per una cor-
retta valutazione di forze, spostamenti e dissipazione
di energia in isolatori soggetti ad eccitazioni sismiche.
28 Anno XXX – N. 3 – luglio-settembre 2013
Comportamento dell’isolatore a pendolo scorrevole
La geometria dell’isolatore, ovvero la curvatura e le
dimensioni delle superci di scorrimento e dell’articola-
zione interna, le caratteristiche di resistenza e rigidezza
dei materiali e le caratteristiche di sistema, quali il va-
lore del coefficiente di attrito, concorrono insieme al
controllo della risposta del dispositivo e della struttura
sovrastante alle azioni sismiche, in termini di forze e
spostamenti. Il comportamento a pendolo scorrevole è
comumente considerato suddiviso in due componenti
(Zayas et al. 1987): la componente legata alla forza
di richiamo Fr e la componente associata alla forza di
attrito Ff. La forza orizzontale risultante Fh sull’isolatore
nel suo complesso può essere espressa nella forma
generale:
Fh = ir · Fr + if · Ff (1)
dove ir ed if sono vettori unitari che esprimono la di-
rezione della forza di richiamo Fr e della forza di at-
trito Ff. La forza di richiamo è direttamente associata
al comportamento a pendolo dell’isolatore dovuto alla
curvatura della supercie di scivolamento, mentre la
forza di attrito è generata all’interfaccia tra la supercie
di scivolamento e l’articolazione interna.
Comportamento a pendolo
Il comportamento a pendolo è originato dalla forma
semi-sferica della supercie di scorrimento e la sua
funzione è di produrre una forza di richiamo laterale
che è funzione dell’entità del carico verticale applicato
all’isolatore. La rigidezza laterale associata a questa
forza di richiamo è quindi funzione del carico verticale,
L’attuale disponibilità di risultati sperimentali consente
di convalidare il modello di attrito proposto da Lomiento
et al. (2013), inizialmente calibrato sulla base di prove
mono-direzionali, per eccitazioni bi-direzionali. A que-
sto scopo, nel presente lavoro si presentano i risultati
di prove sperimentali su un isolatore scorrevole, la cui
analisi ha consentito un accurato studio del funziona-
mento a pendolo del dispositivo nonché della direzio-
nalità delle forze di attrito.
Campagna sperimentale
La tipologia di isolatore oggetto di studio è costitu-
ita dall’isolatore a pendolo scorrevole a singola curva-
tura, composto da un piastra d’acciaio superiore, sede
dell’alloggiamento per l’articolazione interna, una pia-
stra inferiore, all’interno della quale è realizzata una
supercie concava semi-sferica ricoperta da acciaio lu-
cidato a specchio, ed un’articolazione di forma lentico-
lare ricoperta da un polimero ad alta capacità portante,
come mostrato in Fig. 1(a).
Un isolatore a scala reale è stato soggetto a prove
sperimentali nel laboratorio Caltrans Seismic Response
Modication Device (SRMD) presso l’università della
California di San Diego, dotato di una tavola vibrante
a 6 gradi di libertà specicatamente progettata per test
in scala reale su isolatori e dissipatori (Benzoni and
Seible 1998). Il campo di spostamenti consentito dalla
tavola in direzione longitudinale è di ±1.22 m con una
capacità orizzontale in termine di forza pari a 9000 kN
ed una capacità di carico verticale di 53400 kN. La
velocità di picco che può essere raggiunta dalla tavola
in direzione longitudinale è pari a 1.8 m/s. La procedura
di installazione del dispositivo nella macchina di prova
è consistente con le procedure standard di installazione
dei dispositivi di isolamento. La piastra inferiore dell’i-
solare è stata connessa alla tavola vibrante, mentre la
parte superiore, con l’alloggiamento per l’articolazione
interna, è stata ancorata alla trave trasversale di rea-
zione che rappresenta il riferimento sso per il moto,
come mostrato in Fig. 1(b). Il carico verticale è stato
applicato sollevando la tavola, che dopo l’applicazione
del carico, mantenuto costante durante la prove, è stata
mossa in controllo di spostamenti.
Le caratteristiche geometriche dell’isolatore, descritte
in Fig. 2, sono state riportate in Tabella 1. Test bi-dire-
zionali di tipo cloverleaf (a quadrifoglio) sono stati con-
dotti sull’isolatore per tre differenti valori del carico ver-
ticale applicato. Le caratteristiche principali delle prove
effettuate sono specicate in Tabella 2. Le prove clover-
leaf sono state effettuate per vericare l’accuratezza del
modello di predizione proposta per moti bi-direzionali,
mentre le prove mono-direzionali, rappresentative delle
singole componenti di moto in direzione longitudinale
e laterale di una prova cloverleaf, sono state condotte
per confronto con le prove bi-direzionali, in modo da
evidenziare le differenze di comportamento dell’isolatore
dovute alla multi-direzionalità del moto. Il percorso di
spostamenti di una prova cloverleaf è rappresentato in
Fig. 3(a). Le componenti di spostamenti in direzione
longitudinali e lateriali per la suddetta prova sono in-
vece riportate in Fig. 3(b). I cicli di ampiezza limitata
all’inizio ed alla ne della prova sono stati introdotti per
ridurre gli effetti impulsivi di applicazione della forzante
sul dispositivo.
Tab. 1. Caratteristiche geometriche dell’isolatore.
Dimensione Valore (mm)
A435
a180
R2235
r375
T40
t120
h160
d = r – h 215
Tab. 2. Protocollo di prova.
Prova # Tipo
di prova
Carico
verticale
N
(kN)
Pres-
sione di
contatto
pc
(MPa)
Spost.
long. di
picco
Dlong
(mm)
Vel. long.
di picco
Vlong
(mm/s)
Spost.
lat. di
picco
Dlat
(mm)
Vel. lat.
di picco
Vlat
(mm/s)
01 CL 3263 15 200 90 100 45
02 CL 6525 30 200 90 100 45
03 CL 13050 60 200 90 100 45
04 CL long.
comp. 6525 30 200 90 – –
05 CL lat.
comp. 6525 30 100 45
Legenda: CL = prova tipo “cloverleaf”; pc = N/ra2, in cui N > 0 è il
carico verticale di compressione.
29
Anno XXX – N. 3 – luglio-settembre 2013
Come mostrato in Fig. 5(b), la forza laterale di ri-
chiamo Fr è sempre diretta verso il centro geometrico
della supercie concava di scorrimento. In un sistema
di assi cartesiani di riferimento con origine nel centro
dell’isolatore, il vettore che esprime la direzione della
forza di richiamo è quindi espresso come
iu
u
uu u
u
u
11
ry
x
xy y
x
22
$$
=
+
=
::
DD
(9)
Nel modello presentato, si assumono nulle le im-
perfezioni della supercie concava di scorrimento e le
eccentricità dovuti a possibili difetti di posizionalmento
dell’articolazione interna.
Se le Eq. 8(b) e 9 sono sostituite nell’Eq. 1, in condi-
zioni di carico verticale costante, il contributo della forza
di richiamo Fr cresce linearmente con lo spostamento
radiale, proporzionalmente alla rigidezza di richiamo Kr.
Il contributo della forza di attrito Ff si aggiunge a quello
della forza di richiamo Fr per produrre il generico ciclo
forza-spostamento di Fig. 6(a). In Fig. 6(b), il contributo
della forza di richiamo è stato rappresentato sui cicli
sperimentali ottenuti nella prova mono-direzionale 04.
Si noti che la rigidezza calcolata sulla base del rag-
gio di curvatura R è più elevata della reale rigidezza
ottenuta usando il raggio effettivo Reff (Eq. 8 (b)). La
forza di richiamo calcolata sulla base di Reff appare in
buon accordo con i dati sperimentali, come evidenziato
dal parallelismo tra la porzione superiore ed inferiore
del ciclo di Fig. 6(b) con la rigidezza di richiamo.
In concomitanza con la forza di richiamo, i momenti
Mrt e Mrb agiscono rispettivamente alla sommità ed
alla base dell’isolatore. Questi momenti sono originati
dall’eccentricità della forza di contatto rispetto al cen-
tro della piastra superiore ed inferiore del dispositivo.
Dall’equilibrio rotazionale, si ottengono le seguenti
espressioni per i momenti:
Mrt = Fr · ert = Fr · (r + t) (10)
Mrb = Fr · erb = Fr · (R + T) (11)
in cui ert ed erb sono i bracci della forza Fr dalla pia-
stra superiore ed inferiore rispettivamente. Si noti che,
generalmente, erb > ert (2275 mm e 345 mm per il di-
spositivo oggetto di studio). Sebbene i momenti alla
sommità dell’isolatore possano talvolta essere trascu-
rati, i momenti alla base dell’isolatore possono risul-
tare particolarmente elevati ed i loro effetti possono
essere signicativi, in particolar modo sulla struttura al
di sotto del sistema d’isolamento. I massimi momenti
raggiunti durante il test 04 di Fig. 6(b) sono risultati
Mrb = 1192 kNm e Mrt = 180 kNm. Nei frequenti casi in
cui l’isolatore è installato in congurazione rovesciata,
con l’alloggiamento per l’articolazione interna alla base
e la supericie di scivolamento in sommità, i momenti
ovviamente sono rovesciati anche loro, con valori mag-
giori in sommità piuttosto che alla base.
Comportamento attritivo
Le forze attritive al di sopra ed al di sotto dell’artico-
lazione interna creano un momento ribaltante che viene
bilanciato da una traslazione delle forze di contatto Fc
rispetto alla posizione teorica di Fig. 5(a). Questa tra-
slazione prova un disallineamento delle forze di contatto
che nel presente studio viene per semplicità trascurato.
oltre che della geometria dell’isolatore. La determina-
zione della forza di richiamo richiede lo studio della ci-
nematica dei tre componenti del dispositivo, rappresen-
tata in Fig. 4(a) e Fig. 4(b). La componente orizzontale
dello spostamento radiale u e l’associata componente
verticale w possono essere espresse in termini della
posizione dell’articolazione interna, identicata tramite
l’angolo i, e della geometria del dispositivo. Nell’ipo-
tesi di deformazioni nulle nei componenti dell’isolatore,
rotazioni trascurabili della piastra superiore ed inferiore
e di contatto non attritivo, si possono derivare espres-
sioni in forma chiusa per gli spostamenti u e w, come
riportato in Fig. 4(b).
È importante notare che, come mostrato in Fig. 4(b),
il percorso di spostamento della piastra superiore dell’i-
solatore, coincidente con lo spostamento della base
della struttura sovrastante, avviene su una supercie
sferica di raggio effettivo (Fenz and Constantinou 2008):
Reff = R + d (2)
Nella formulazione originale del comportamento a
pendolo proposta da Zayas (1987) e generalmente uti-
lizzata nei primi studi sul dispositivo (Tsai 1997, Alma-
zan et al. 1998), le dimensioni siche dell’articolazione
interna erano state trascurate ed il raggio effettivo era
stato approssimato dal raggio di curvatura della su-
percie di scorrimento R. È da notare, tuttavia, che in
numerosi isolatori di recente fabbricazione, il raggio
effettivo Reff differisce signicativamente dal raggio di
curvatura R, nella maggior parte dei casi in ragione del
ridotto spessore degli isolatori. Per l’isolatore in oggetto,
il raggio effettivo Reff risulta ~10% maggiore del raggio
di curvatura R.
Le espressioni in forma chiusa per gli spostamenti
del dispositvo possono essere semplificate assumendo
valida l’approssimazione di piccoli angoli. Per isola-
tori con un raggio in pianta della superficie concava
A > R/3, gli errori associati a questa approssimazione
sono estremamente ridotti (meno del 3% rispetto ai
valori delle espressioni esatte in forma chiusa), con
errori massimi in corrispondenza del massimo sposta-
mento consentito dall’isolatore. Le espressione esatte
e semplificate delle espressioni per il massimo valore
consentito dell’angolo i, dello spostamento radiale u
e dello spostamento verticale w sono riportate in Ta-
bella 3.
La forza di richiamo laterale Fr, la forza di contatto
Fc, i momenti in sommità Mrt ed alla base Mrb asso-
ciati alla forza di richiamo sono determinati sulla base
dell’equilibrio delle forze interne ed esterne agenti
sull’isolatore, rappresentato in Fig. 5(a), e possono
essere espressi come funzioni del carico verticale N,
della proiezione orizzontale dello spostamento radiale
u e delle dimensioni geometriche dell’isolatore. Le
espressioni esatte e semplificate per Fr, Fc e per la
pressione di contatto media pc sono riportate in Ta-
bella 4. La rigidezza di richiamo laterale Kr è definita
in Eq. 8.
Si noti che i massimi errori associati alle espressioni
approssimate per le forze sono generalmente più elevati
dei massimi errori riscontrati per la predizione degli
spostamenti dell’isolatore. Tuttavia, si consideri che que-
sti errori si riducono signicativamente al crescere del
rapporto R/A. Per il dispositivo oggetto di studio, per
esempio, per cui A R/5, il massimo errore commesso
nella predizione delle forze, ottenuto al raggiungimento
del massimo spostamento radiale, risulta <1%.
30 Anno XXX – N. 3 – luglio-settembre 2013
In accordo con la teoria semplicato di Coulomb per
corpi soggetti a scivolamento, la forza di attrito è diret-
tamente proporzionale al carico normale applicato e si
oppone alla direzione del moto. Il coefficiente di attrito
dinamico è considerato indipendente dall’area effettiva
di contatto e dalla velocità di scivolamento. Tuttavia,
l’esperienza sperimentale dimostra che questa teoria
semplicata non può essere applicata a dispositivi a
pendolo scorrevole peri quali le forze attritive risultano
signicativamente condizionate da fasi di transizioni del
moto da statico a dinamico (breakaway), pressioni di
contatto, velocità di scivolamento e incrementi di tem-
peratura generati all’interfaccia di scivolamento. Questi
fattori condizionano signicativamente la forma dei cicli
forza-spostamento di questi isolatori, secondo i quattro
effetti descrittinin dettaglio in (Lomiento et al. 2013):
1. effetto di breakaway, ovvero il repentino incre-
mento del coefficiente di attrito n all’inizio del moto ed
ad ogni inversione del moto;
2. “effetto del carico”, ovvero la riduzione del coeffi-
ciente di attrito al crescere della pressione di contatto;
3. “effetto di ripetizione ciclica, ovvero la continua
riduzione del coefficiente di attrito con la ripetizioni di
cicli, maggiormente pronunciata per alte pressioni e alte
velocità del moto di scivolamento, dovuta all’incremento
di temperature delle superci di scorrimento.
4. “effetto della velocità, ovvero la variazione del
coefficiente di attrito con la velocità di scivolamento.
Rimuovendo il contributo della forza di richiamo dai
dati sperimentali, questi effetti sono chiaramente visibili
nell’andamento del coefficiente di attrito n col tempo
durante le prove, sia mono-direzionali sia bi-direzionali.
La variazione di n nelle prove 01 e 03 è rappresentata
in Fig. 9, in cui le linee tratteggiate indicano la varia-
zione media ± la deviazione standard. Immediatamente
al di sotto, si riportano gli andamenti delle velocità di
scivolamento.
Dal confronto tra la Fig. 9(a) e 9(b), l’effetto del ca-
rico è riconoscibile in forma di riduzione globale del
coefficiente di attrito dalla prova 01 alla prova 03 per
effetto della maggior pressione di contatto, che varia
da pc = 15 MPa a pc = 60 MPa. Un aspetto particolar-
mente evidente è la continua riduazione del coefficiente
L’eccentricità in pianta delle forze attritive genera un
momento torsionale che produce un moto di rotazione
dell’articolazione interna attorno al suo asse normale
che avviene contemporaneamente al moto traslazionale
di scivolamento. Per la concomitanza di moti rotazionali
e traslazionali, il contributo resistente dovuto alla forza
attritiva non è necessariamente parallelo alla proiezione
orizzontale del vettore velocità. La diversa orientazione
del vettore delle forze attritive dal vettore velocità è indi-
cata con l’angolo Dz in Fig. 7(a). In Fig. 7(b) le direzioni
di forza attritiva, vettore velocità e forza di richiamo
sono rappresentate in un punto generico del percorso
di spostamenti della prova cloverleaf 02.
In Fig. 8 si confrontano gli angoli z dei vettori velo-
cità e forza, misurati rispetto alla direzione longitudinale
del moto assunta a riferimento. La differenza in angolo
tra i due vettori ad ogni istante è risultata approssima-
tivamente costante durante tutti le prove cloverleaf, in-
dipendentemente dal carico verticale applicato. In base
ai risultati sperimentali, la differenza angolare è stata
valutata come:
Dz 4 · sin–1(h/Reff) (12)
in cui h è l’altezza dell’articolazione interna ed Reff il
raggio effettivo di curvatura. Per superci di scorrimento
piatte (Reff ), la differenza angolare diventa nulla
e la forza attritiva è diretta come la velocità, ovvero
nella direzione del moto. Per l’isolatore in oggetto si
è identicato una differenza angolare tra i due vettori
pari a Dz 14°.
Come conseguenza della differenza angolare, la di-
rezione del vettore delle forze attritive è data da:
cos
sin
if
f
f
z
z
=
^
^
h
h
=
G
(13)
dove (/)
(/)
arctan cos
sin
2
f
vvv
vvv
zzz
zz
zz
zz
D
D
=+
+
oo
oo
eo
è l’angolo della
forza attritiva e 2arctan
x
y
zo
o
=
o
cm
è l’angolo della dire-
zione del moto, calcolata sulla base delle componenti
ortogonali ox e oy del vettore velocità. Arctan2 è la fun-
zione arcotangente a due argomenti che permette di
distinguere tra direzioni diametricalmente opposte.
Tab. 3. Formule esatte ed approssimate per i fondamentali parametri di risposta.
Variabile Espressione esatta Espressione semplicata Errore per A < R/3
Angolo massimo si
ns
in
R
A
R
a
max
11––
i=
``
jj
(3a) R
Aa
max
i=<2% (3b)
Spost. radiale u = Reff sin i(4a) u = Reff
i<2% (4b)
Spost. verticale w = Reff(1 – cosi)(5a) w = Reff i2/2 <3% (5b)
Tab. 4. Formule esatte ed approssimate per le forze dell’isolatore.
Variabile Espressione esatta Espressione semplicata Errore per A < R/3
Forze di contatto cos
F
N
c
θ
=(6a) Fc = N %
A
Aa
6
1(6b)
Pressione di contatto cos
pN
a
aR
2
11
c2
2
θ
π
=
+
_i
(7a) pa
N
2
c2
π
=%
A
Aa
6
1(7b)
Forza di richiamo cos
FKuR
N
u
rr
eff
$$
θ
== (8a) FKuR
N
u
rr
eff
$$
== %A
Aa
6
1(8b)
31
Anno XXX – N. 3 – luglio-settembre 2013
condotte per diversi carichi verticali applicati e velocità
variabili. Per il dispositivo in oggetto, i valori determi-
nati sulla base delle prove effettuate sono riportati in
Tabella 6.
Confronto tra risultati numerici e sperimentali
Prove mono-direzionali
La validazione sperimentale del coefficiente di attrito
durante le prove mono-direzionali 04 e 05 è presen-
tata in Fig. 11. Per sottolineare il contributo dell’effetto
di ripetizione dei cicli, sono stati utilizzati due modelli,
uno che includesse il degrado del coefficiente d’attrito
e l’altro che includesse solo gli effetti del carico e della
velocità. Come evidente dalla Fig. 11, trascurare i fe-
nomeni di degrado dovuti al continuo moto di scivola-
mento comporta signicative differenze tra la predizione
ed i risultati sperimentali.
La riduzione di n col tempo per effetto della ripeti-
zione ciclica è particolarmente evidente per la compo-
nente longitudinale del moto di Fig .11(a) in ragione di
valori più elevati di velocità e spostamenti.
I cicli forza-spostamento predetti considerando e non
considerando l’effetto di ripetizione ciclica sono confron-
tati con i corrispondenti cicli sperimentali in Fig. 12(a) e
12(b) per la componente longitudinale ed in Fig. 13(a)
e 13(b) per la componente laterale.
L’applicazione del modello di attrito conferma il buon
accordo tra risultati sperimentali e predizione del com-
portamento a condizione che gli effetti di ripetizione
ciclica siano inclusi nel modello.
Prove bi-direzionali
In Fig. 14(a) e 14(b) si confronta l’andamento speri-
mentale del coefficitente di attrito nel tempo con i risul-
tati del modello con e senza effetti di ripetizione ciclica
per le prove bi-direzionali 01 e 03.
La variazione di attrito appare signicativamente
maggiore nella prova 03 ad alta pressione che nella
prova 01, in cui la pressione di contatto era signica-
tivamente più bassa. Come già evidenziato, il modello
proposto appare adeguato a descrivere l’effetto di con-
tinua riduzione del coefficiente di attrito ma non le con-
tinue uttuazioni dovuti a fenomeni di riscaldamento
locale, per i quali sarebbe necessario un modello più
sosticato
Per la prova cloverleaf 02, sia gli effetti di degrado
che le uttuazioni sono più evidenti che nelle corrispet-
tive prove mono-direzionali di Fig. 11 in ragione dell’in-
tuibile maggiore effetto di riscaldamento associato al
moto multi-direzionale rispetto al moto mono-direzionale.
Di conseguenza, trascurare l’effetto di ripetizione ciclica
comporta errori più signicativi nella modellazione del
comportamento bi-direzionale dell’isolatore che nel com-
portamento mono-direzionale.
I cicli sperimentali forza-spostamento ottenuti nella
prova bi-direzionale 02 in direzione longitudinale e la-
terale sono confrontati con i cicli predetti da modello
in Fig. 16 e Fig. 17.
Le forze predette in corrispondenza del massimo
spostamento risultano sovrastimate no al 48% se gli
effetti di ripetizione ciclica non sono inclusi nel mo-
dello. Sebbene un modello completo dovrebbe includere
di attrito durante le prove. La riduzione media, dovuta
all’effetto di ripetizione dei cicli, è maggiore nella prova
03 che nella prova 01 (55% contro 34% del valore ini-
ziale) in ragione della più alta pressione di contatto
che è causa di un riscaldamento più severo durante lo
scivolamento. Le continue uttuazioni del coefficiente di
attrito tra minimi e massimi relativi, che si aggiunge alla
continua riduzione nel tempo, è dovuta alla combina-
zione dell’effetto di velocità (incremento di n al crescere
della velocità) con un effetto di riscaldamento locale
(riduzione di n in corrispondenza di alti valori locali di
temperatura). Nelle prove effettute, la riduzione dovuta
al riscaldamento locale prevale sull’effetto di velocità e,
per questa ragione, il coefficiente d’attrito mostra valori
minimi in corrispondenza dei picchi di velocità, che si
vericano nella zona centrale più calda dell’isolatore.
Un dato sperimentale a conferma di quanto appena
evidenziato è la deviazione standard v della uttua-
zione, che è maggiore nella prova a pressione più alta
(~15% del valor medio) rispetto alla prova a pressione
inferiore (~12%) in ragione di un maggiore usso di ca-
lore. Fluttuazioni simili sono state notate anche in prove
mono-direzionali (Lomiento et al. 2013) ma appaiono
signicativamente più pronunciate in prove bi-direzionali.
L’effetto di breakaway è più signifcativo nelle prove
mono-direzionali, per effetto delle frequenti inversioni
del moto, che nelle prove bi-direzionali, come evidente
in Fig. 10 dove l’andamento del coefficiente di attrito
dalle prove mono-direzionali 04 e 05 (componente lon-
gitudinale e laterale della prova cloverleaf) è confrontato
con l’andamento riscontrato nella corrispondente prova
bi-direzionale. I repentini incrementi di attrito dovuti
all’effetto di breakaway sono visibili solo nelle prove
mono-direzionali.
Nel presente studio, il modello fenomenologico pro-
posto in (Lomiento et al. 2013) è stato usato per predire
la variazione del coefficiente di attrito n, che è espresso
come prodotto di tre contributi indipendenti:
(, ,) () ()
()
NC fNfCf
NC
$$
no o=o (14)
dove fN, fC, fo sono funzioni che tengono conto rispet-
tivamente dell’effetto di carico, ripetizione ciclica e ve-
locità, in funzione del carico verticale di compressione
N ≥ 0, della velocità di scivolamento o e della variabile
di ripetizione ciclica C denita come:
()Ct A
Nd
t
2
22 2
t
t
0
ar
o=# (15)
La variabile di ripetizione ciclica ha le dimensioni
di un usso di calore, ovvero del tasso di calore tra-
sferito per unità di superficie, ed è rappresentiva
dell’incremento di temperatura nell’ipotesi di di stri bu-
zio ne uniforme del flusso di calore sulla superficie di
scivolamento. Questa ipotesi semplicativa è in disac-
cordo con la presenza di maggiori temperature nelle
porzioni della supercie di scivolamento interessate
da più frequenti e più veloci moti di scivolamento ed
è quindi non adeguata a descrivere le uttuazioni del
coefficiente d’attrito per effetti termici locali. Una de-
scrizione più completa del fenomeno è riportata in (Lo-
miento et al. 2013).
Le tre componenti dell’Eq. 14 possono essere
espresse tramite le funzioni riportate in Tabella 5. I pa-
rametri sperimentali elencati in Tabella 5 sono inden-
ticati sulla base di prove sperimentali di laboratorio
32 Anno XXX – N. 3 – luglio-settembre 2013
I risultati delle prove hanno confermato l’accuratezza
delle espressioni semplificate per la modellazione della
rigidezza di richiamo dell’isolatore, a condizione che
il raggio effettivo di curvatura Reff venga usato al po-
sto del raggio di curvatura della superficie concava di
scorrimento R. Per effetto dell’eccentricità del carico ri-
spetto al centro dell’isolatore, nella generica configura-
zione deformata si sviluppano momenti flettenti sia alla
base (Mrb) che alla sommità (Mrt) dell’isolatore. Detti
momenti sono proporzionali alla forza di richiamo e
possono risultare tanto elevati da non poter essere tra-
scurati nelle analisi strutturali, in particolare alla base
dell’isolatore, ovvero dal lato della superficie concava
di scorrimento. I massimi momenti durante le prove
sull’isolatore in oggetto sono risultati Mrb = 1192 kNm
e Mrt = 180 kNm.
Le prove hanno evidenziato signicative variazioni
del coefficiente di attrito dovute al calore generato all’in-
terfaccia di scorrimento durante il moto. Tali variazioni
sono risultate maggiornamente signifcative nelle prove
a più elevato carico verticale e maggiore velocità di
scorrimento. In tutte le prove sono state evidenziate una
continua riduzione del coefficiente d’attrito al procedere
dei cicli di deformazione (effetto di ripetizione ciclica)
e uttuazioni dovuti al passaggio dell’articolazione in-
terna su porzioni a diversa temperatura della supercie
di scorrimento. Riduzioni del coefficiente di attrito no
al 56% del valore iniziale sono state riscontrate nelle
prove a più alto carico verticale. Le uttuazioni sono
risultate maggiori nelle prove bi-direzionali che nelle
prove mono-direzionali, in ragione delle maggiori ve-
locità di scorrimento raggiunte durante queste prove e
di una maggiore disuniformità nella distribuzione delle
temperatura sulla supercie di scorrimento. Nelle prove
bi-direzionali si sono riscontrati valori del coefficiente
d’attrito no al 25% inferiori a quelli evidenziati nelle
corrispondenti prove mono-direzionali come conse-
guenza della multi-direzionalità del moto e del maggior
degrado del coefficiente di attrito per effetti termici. In
generale, trascurare gli effetti ciclici e l’interazione bi-
direzionale nella predizione del comportamento degli
isolatori a pendolo scorrevole può comportare una si-
gnicativa sovrastima delle forze di attrito e della con-
seguente capacità dissipativa. L’energia dissipata nelle
prove bi-direzionali è risultata circa il 70% dell’energia
totale dissipata nelle corrispondenti prove-monodirezio-
nali. Per questo motivo, una completa caratterizzazione
meccanica del comportamento di tali dispositivi è un
prerequisito essenziale per la progettazione di strutture
protette tramite questa tecnologia di isolamento.
anche gli effetti di uttuazione dovuti a riscaldamenti
locali, l’inclusione degli effetti di ripetizione ciclica già
consente una ragionevole stima delle forze negli isola-
tori a pendolo scorrevole.
Un ulteriore aspetto da sottolineare è che i cicli
forza-spostamento da prove bi-direzionali sono più
stretti dei corrispondenti cicli ottenuti da prove mono-
direzionali, come mostrato in Fig. 18. Ciò è dovuto alla
maggiore riduzione del coefficiente di attrito per effetti
termici ed alla coesistenza del moto di scivolamento in
due direzioni. Il cambiamento della forma del ciclo è
associato ad una signicativa riduzione dell’energia dis-
sipata in ciascuna direzione per le prove bi-direzionali.
L’energia dissipata in direzione longitudinale si riduce
infatti da 562 kNm della prova mono-direzionale a 430
kNm per la corrispondente componente della prova bi-
direzionale (riduzione del 23%). In direzione laterale si
osserva una ancor più signicativa riduzione del 41%.
La risposta del dispositivo ottenuta in prove mono-
direzionali può quindi sovrastimare la reale capacità
di dissipazione energetica dell’isolatore ed assumere
una aumentata capacità di dissipazione può provocare
una potenzialmente signifcativa sottostima dei massimi
spostamenti attesi durante eccitazioni sismiche.
Conclusioni
Un modello analitico per la predizione del comporta-
mento bi-direzionale di isolatori a pendolo scorrevole è
stato validato sulla base di risultati sperimentali. Prove
bi-direzionali di tipo cloverleaf (quadrifoglio) e prove
mono-direzionali riproducenti le singole componenti,
longitudinale e laterale, delle prove bi-direzionali sono
state condotte su un dispositivo in scala reale soggetto
a diversi livelli di carico verticale. I cicli forza-sposta-
mento predetti tramite il modello analitico proposto sono
risultati in un buon accordo con i cicli sperimentali.
Tab. 5. Componenti del modello numerico
Effetto Contributo Parametro sperimentale
Carico fN(N) = ns0 · e–N/Nref (17) ns0 = coefficiente d’attrito a carico nullo
Nref = carico corrispondente ad una riduzione di attrito del 63% per l’effetto del carico
Ripetizione ciclica fC(C) = e–(C/Cref)B(18) Cref = C corrispondente ad una riduzione di attrito del 63% per l’effetto della ripetizione
ciclica (usso di calore di riferimento, con unità di potenza per supercie unitaria)
b = tasso esponenziale di degrado dell’attrito
Velocità fo(o) = c + (1 – c) · e|o|/oref (19) c = rapporto tra coefficienti d’attrito veloce / lento
oref = o corrispondente ad una variazione del 63% del coefficiente d’attrito per effetto
dell’incremento di velocità
Tab. 6. Valori sperimentali per l’isolatore oggetto di studio
Effetto Parametro Valore
Carico ns0 0.103
Nref 12300 kN
Ripetizione ciclica Cref 9700 kN/ms
b0.5
Velocità c1.40
oref 10 mm/s
... Considerable bidirectional experiments of FPBs were conducted to study the bidirectional efects of friction force on FPBs [13], the coupling efect between two orthogonal components on the behavior of FPBs [14], and the bidirectional behavior of FPBs across a range of realistic vertical loads and sliding velocities [15]. It was generally believed that the bidirectional efect and coupling efect of bidirectional GMs on the maximum displacement of FPBs cannot be ignored [14][15][16]. ...
... Considerable bidirectional experiments of FPBs were conducted to study the bidirectional efects of friction force on FPBs [13], the coupling efect between two orthogonal components on the behavior of FPBs [14], and the bidirectional behavior of FPBs across a range of realistic vertical loads and sliding velocities [15]. It was generally believed that the bidirectional efect and coupling efect of bidirectional GMs on the maximum displacement of FPBs cannot be ignored [14][15][16]. However, in previous studies, the frequency and intensity characteristics of GM were mostly discussed because of their efect on the selection of μ and R S , leaving the bidirectional behavior of GMs, whose efect on determining the displacement capacity (Φ) can be signifcant, not thoroughly discussed. ...
... Te real-time TC increase at the contact interface of DCFPB is simulated by the friction heating model displayed in Figure 8, where one temperature monitor point was applied [28]. Using this model, the temperature incremental (ΔTC) history of the contact interface can be obtained using equation (14), where the heat fux history at the monitor point (q) can be calculated by using equation (15); k t and D t are the thermal conductivity and difusivity of stainless steel, which are 0.016 W/(mm°C) and 4.07 mm 2 /s, respectively [32]. For future work, it will be possible to simplify the modeling of the biaxial behavior of FPB by introducing a rate-independent phenomenological model, which has an algebraic nature, and an explicit structure-dependent time integration method, which do not require iterative procedures [33]. ...
Article
Full-text available
Recent destructions of structures due to insufficient isolator deformation capacity have led to demands for greater seismic redundancy in seismic isolation design. For a friction pendulum system (FPS), the effect of bidirectional behavior of earthquakes on the maximum response and its effect on friction heating, temperature, and in turn on the maximum response can be significant. However, the extent of these effects under different FPS design parameters and different types of ground motions (GMs) is still not clear. In this study, an analytical model of double concave FPS considering the coupling effect of friction heating and bidirectional behavior was proposed and validated by bidirectional earthquake response orbits, which reflect the characteristics of both GMs and FPSs. Then, the effects of bidirectional GM and corresponding bidirectional temperature change on the response were investigated under different types of strong GMs. Finally, a performance-based design method with a bidirectional-effect-compensation mechanism was proposed. For double concave friction pendulum bearings with PTFE-related layers, it was found that the bidirectional behavior of earthquakes will amplify the maximum isolator displacement by an average of 110–210% (60 MPa) and the maximum superstructure acceleration by an average of 100–140% (60 MPa) under strong GMs (PGV-C1 > 0.2 m/s) and optimum design parameters. The amplification ratio is not only influenced by GM characteristics but also highly related to the design parameters and friction-heating effect of DCFPS.
... A number of experimental campaigns under different loading scenarios were presented in the literature, especially for comparing FREIs with SREIs [9,[22][23][24], based on which a higher damping capacity of the former than the latter has been acknowledged. However, the majority of literature studies were focused on small-scale prototypes [25,26] and were typically conducted under unidirectional loading only, i.e., monodirectional and quasi-static tests [27][28][29], although traditional isolation devices can exhibit significant bidirectional interaction [30][31][32][33]. Among others, Abe et al. [34] and Kim et al. [35] studied the multi-axial behavior of SREIs and found remarkable coupling effects in the comparison between triaxial and biaxial experiments. ...
... The same trend was found by previous studies on SREIs for both stiffness and damping [34,35]. The stiffness reduction effect was also observed for friction-based devices [30,31,33]. These findings are confirmed in Fig. 11b where 1D and 2D tests are compared in a higher velocity range (200-260 mm/s). ...
Article
Unbonded fiber reinforced elastomeric isolators (UFREIs) can be used for the seismic protection of structures as a lower-cost alternative to conventional laminated rubber bearings, by replacing internal steel shims with fiber reinforcement and exploiting the frictional mechanism at the rubber-concrete interface to avoid anchorage bolts. In this paper, the hysteretic behavior of full-scale (diameter 620 mm) UFREIs is investigated in both an experimental and a numerical framework when subjected to triaxial loading (i.e., simultaneous imposed displacement along two horizontal directions with concurrently applied vertical load), to account for the inherent multi-directional nature of a real earthquake scenario. Experimental results of UFREIs tested under two bidirectional orbits involving different velocities and amplitudes up to 100% shear strain are presented, and peculiar effects ascribed to the lateral coupling of the devices are identified by comparison with monodirectional test results. Then, an efficient nonlinear phenomenological model is proposed to simulate the isotropically coupled biaxial hysteretic behavior of UFREIs detected from the experiments. This model consists of a set of nonlinear springs arranged in a circular configuration and governed by just three parameters with clear mechanical significance. Novel analytical (closed-form) expressions for the model calibration are ad-hoc developed in this paper. This modeling approach is used to quantify the impact of biaxial coupling of UFREIs on the structural performance of base-isolated structures under bidirectional seismic excitation. To this aim, the seismic response of a three-dimensional reinforced concrete building isolated with UFREIs is numerically simulated. Numerical results show that isolator displacements and superstructure accelerations tend to be higher by using an uncoupled model, calibrated upon monodirectional tests and that neglects bidirectional interaction, as usually performed in practice.
... At the generic time instant t, the instantaneous values of the longitudinal (F x ) and transversal (F y ) components of the horizontal force resisted by a sliding isolator in a bi-directional motion in the x-y plane (see Fig. 2) can be calculated as [6,30]: ...
... Besides the degradation effect due to frictional heating, captured from the estimator as already illustrated in Section 4, the CUKF also identifies the variation of the friction coefficient induced from changes in the instantaneous sliding velocity. In agreement with available literature friction models [30], the maximum value μ B = 0.24 of the coefficient of friction is estimated at the motion breakaway when the velocity is null (v = 0mm/s), while local maxima are observed at = 1.7s, and t = 12.0s when the sliding velocity has local minima ( Fig. 15-c). The mean percentage error (err % (t) = 100⋅(μ EXP (t) − μ CUKF (t) )/μ EXP (t)) on the estimate of the effective friction coefficient μ eff is 7.9%. ...
Article
One of the most challenging task for earthquake engineers is the accurate prediction of the dynamic response of structures implementing sliding seismic isolators, like the curved surface sliders. Indeed, the force–displacement behaviour of these devices strictly depends on the coefficiens of friction developed at the two sliding surfaces whose values vary instantaneously depending on variable compression load, sliding velocity, and contact temperature developed during the seismic motion. However, only the overall (or effective) friction coefficient of the isolator can be estimated, as weighted average of the values at the two sliding surfaces, through tipycal displacement controlled prototype experimental tests. This information is not suitable for the calibration of predictive isolator models (e.g. FEM analyses) or to characterize the tribological behaviour of potentially different friction pads at the two sliding surfaces. In this paper, an estimation approach, based on a Constrained Unscented Kalman Filter (CUKF) integrated with a Random Walk Model technique (RWM), that is capable to identify the two distinct friction coefficients, and their time-variations during displacement-controlled test is presented. The proposed tool allows the identification of the distinct frictional properties without any a-priori knowledge about the design properties at the two sliding surfaces that can widely differ in terms of adopted sliding materials, presence of lubrificant, and radius of curvature. The developed tool is firstly validated through the comparison with the results of FEM analyses and then applied on experimental tests carried out on a full-scale isolator prototype demonstrating its suitability for the assessment of the actual friction coefficients.
... Additionally, Warn et al. proposed that the influence of bidirectional excitation and coupling on the maximum displacement cannot be ignored [26]. If the bidirectional coupling of the responses of FPB is ignored, it will result in an underestimation of the maximum isolator displacement by approximately 20% [27,28]. Despite of numerous existing studies on the effect of GM characteristics, design parameters, friction dependencies, friction heating, and bidirectional coupling on the seismic behavior of FPSs, these effects have rarely been considered when designing an FPS because the mutual influence among these parameters is complicated. ...
Article
Full-text available
Double concave friction pendulum system (DCFPS) performs well in the vibration control of building structures, while its effectiveness is limited due to the absence of considering the interrelation between design parameters, ground motion (GM) characteristics, friction heating and seismic responses. By systematically evaluating the optimum design parameters for DCFPS under different strong GM classifications and taking into account the effect of friction heating, this paper contributes a comprehensive understanding of the optimum design solutions for DCFPS. To provide a precise characterization of the effect of friction heating and loading conditions on the responses, an analytical model for DCFPS in the SDOF structure is established, and its accuracy in characterizing the friction heating effect is validated by full-scale bearings under different dynamic loading conditions. Based on this, the friction coefficient considering the influence of friction heating under strong GMs is applied in the seismic design. Subsequently, from the perspective of GM generation, strong GM records are categorized based on earthquake magnitude, distance-to-fault and soil conditions to indicate their magnitudes and spectral attributes. Then, the optimal design parameters are investigated for different GM classifications considering the effect of friction heating and response restrictions using response magnitude as criterion, and the effect of GM characteristics on the response and design parameters of DCFPS is investigated. It was found that, the consideration of residual displacement restriction and friction heating shows critical influence on the optimum selection of design parameters. Furthermore, the optimum friction coefficient is highly related to distance-to-fault, while the optimum range of isolation period is minimally influenced by GM characteristics. On top of these, a performance-based design method of FPS considering three performance criteria corresponding to three GM intensity levels is proposed, providing an effective method for mitigating the vibration of FPS under GMs.
... Additionally, Warn et al. proposed that the in uence of bidirectional excitation and coupling on the maximum displacement cannot be ignored [19]. If the bidirectional coupling of the responses of FPB is ignored, it will result in an underestimation of the maximum isolator displacement by approximately 20% [20,21]. Despite of numerous existing studies on the effect of GM characteristics, design parameters, friction dependencies, friction heating, and bidirectional coupling on the seismic behavior of FPSs, these effects have rarely been considered when designing an FPS because the mutual in uence among these parameters is complicated. ...
Preprint
Full-text available
This paper addresses effectiveness limitations in the design of double concave friction pendulum systems (DCFPS) due to the absence of considering the interrelation between design parameters, ground motions (GM) characteristics, friction heating and seismic responses. By systematically evaluating the optimum design parameters for DCFPS under different strong GM classifications considering the effect of friction heating, it contributes a comprehensive understanding of the efficacy of DCFPS in vibration control. In this study, an analytical model for DCFPS in the SDOF structure is established, providing a precise characterization of the effect of friction heating and loading conditions on the responses. Based on this, the friction coefficient considering the influence of friction heating under strong GMs is applied in the seismic design. Subsequently, from the perspective of GM generation, strong GM records are categorized based on earthquake magnitude, distance-to-fault and soil conditions to indicate their magnitudes and spectral attributes. Then, the optimal design parameters are investigated for different GM classifications considering the effect of friction heating and response restrictions using response magnitude as criterion, and the effect of GM characteristics on the response and design parameters of DCFPS is investigated. It was found that, the consideration of residual displacement restriction and friction heating shows critical influence on the optimum selection of design parameters. Furthermore, the optimum friction coefficient is highly related to distance-to-fault, while the optimum range of isolation period is minimally influenced by GM characteristics. On top of these, a performance-based design method of FPS considering three performance criteria corresponding to three GM intensity levels is proposed, providing an effective method for mitigating the vibration of FPS under GMs.
... kN/mm) are lower than those obtained from 1D tests at comparable amplitude 140 mm (0.72 kN/mm shown in Fig. 3). Similar trends were obtained for other tests, here not shown for the sake of brevity, with an increase of damping ratio and a substantial reduction of the effective stiffness in 2D tests compared to 1D tests; the latter trend is in line with test results reported in the literature for SREIs (Abe et al. 2004;Kim et al. 2019) and for friction-based devices (Lomiento et al. 2013;Furinghetti et al. 2019). ...
... In the left part of Fig. 9, the force-displacement cycles in the y direction for uni-directional and bi-directional tests having identical amplitude and comparable peak velocity (in the range 100 mm/s) are shown: the peak-to-peak secant slope, characterizing the equivalent stiffness of the isolator, was found to be lower in bi-directional tests compared to uni-directional tests. This stiffness reduction effect was likely caused by the bidirectional nature of the motion and was also observed in other literature studies in the field of friction-based devices (Lomiento et al., 2013;De Domenico et al., 2018, Furinghetti et al., 2019. To confirm this outcome for a wider range of excitations, the equivalent stiffness for various amplitudes is compared in the right part of Fig. 9. ...
Conference Paper
Seismic base isolation is nowadays a mature technology to mitigate seismic risk in earthquake-prone regions. Nevertheless, it has not been implemented extensively in developing countries because of the relatively high cost of isolation devices. To reduce manufacturing and assembly cost of elastomeric bearings, fiber-reinforced elastomeric isolators (FREIs) can be used as an alternative to traditional steel reinforced elastomeric isolators (SREIs). Steel shims representing the reinforcement and offering the confinement action in traditional SREIs can be replaced by bidirectional fiber layers in FREIs, and installation can be accomplished by exploiting friction mechanisms and roughness at the rubber-concrete interface, thus avoiding any anchorage bolts (unbonded configuration). In this contribution, experimental dynamic tests on two full-scale circular (diameter 620 mm) FREIs realized with polyester fabric and tested in unbonded configuration are presented. The tests were performed at the EUROLAB of the CERISI, University of Messina, Italy, to investigate the hysteretic behavior under a range of frequencies, axial loads, and amplitudes. The testing protocol, inspired by initial type tests of elastomeric bearings as per European Standards EN15129, included additional 2D tests to investigate the variability of stiffness and damping properties under bidirectional excitation. It has been found that the tested full-scale FREIs exhibit a stable hysteretic behavior and satisfy the main prescriptions of EN15129 standards. However, roll-over phenomena and resulting reduction of contact area under increasing lateral displacements limit the use of these FREIs to within shear strain level of 100%.
... The program not only included the type tests required by the standards [26] but also tri-directional time-history protocols. Lomiento et al. [32] tested CSSs in order to investigate the influence of multi-directionality of motion and degradation of the coefficient of friction due to heating phenomena at the sliding surface on full scale devices. Cloverleaf bidirectional tests were performed for three different levels of vertical loads. ...
Article
Fiber reinforced elastomeric isolators (FREIs) have shown to be a promising option as an alternative to classical steel reinforced elastomeric isolators (SREIs). Previous investigations were limited to scaled-geometries without any attempt to test such devices under code-compliant protocols as per international standards. In the present study the authors investigated two circular full-scale (diameter 620 mm) FREIs manufactured with a non-standard process adopting a soft rubber compound and polyester fibers. Experimental tests were performed in unbounded configuration through the large anti-seismic device test facility at the EUROLAB of the University of Messina, Italy. A significant number of protocols were imposed to the prototypes in order to demonstrate the effect of different loading conditions, i.e., strain level, frequency of the excitation, axial load and repeated loading. The tests confirmed the significant dependency of mechanical behavior on axial load which tends to increase damping (i.e., higher friction mechanisms) while reducing stiffness (i.e., lower stability limits). Due to internal slippage at the fiber-rubber layers interface, damping capacity of FREIs achieved 20% of critical, i.e., significantly higher than that commonly achieved in SREIs with the same rubber compound. Even if adequate capacity was reached in compression, the run-in effect would limit the axial stiffness of FREIs. A significant roll-over phenomenon was detected under lateral loading up to a maximum shear strain of 100% without damage and permanent deformation after unloading. A numerical study finally demonstrated the effectiveness of FREIs when compared to SREIs in a base isolation system designed for a 3-storey reinforced concrete frame located in a high seismicity region in Italy. Lower axial stiffness of FREIs did not affect the seismic performance of the building due to limited rocking motion component and beneficial higher damping mechanism. This paper provides a significant contribution to the standardization of FREIs to be adopted in base isolation of conventional buildings.
... The design of the proper period elongation, internal forces in the superstructure elements can be significantly reduced; on the other hand a large displacement demand may occur at the isolation level, even though lower values can be achieved, by means of the hysteretic characteristics of the implemented devices. Curved Surface Slider devices, among the others, represent one of the most suitable solutions for baseisolation ( [2], [3], [10], [12], [13], [14]). Such isolators can actually accommodate large displacement demands, with potentially high dissipative capacities, which are provided by the frictional response of the induced sliding motion. ...
Conference Paper
Full-text available
Base isolation represents one of the most efficient strategy for the reduction of the structural vulnerability of buildings and bridges. Design procedures generally aim to provide the proper period shift, in order to reduce spectral acceleration values and, consequently, the base shear and internal forces. On the other hand, high displacement demands can be achieved, which can be partially limited by providing dissipative capacity through hysteretic behaviors. Although design procedures allow to fairly estimate the design displacement of the adopted devices , extreme seismic event can occur, and displacement higher than the design value can be experienced. Especially for Curved Surface Slider devices, if the displacement demand exceeds a certain geometrical limit, non-negligible damage can occur at the sliding pad, and variations in the force response are consequently noticed. In this work modeling strategies for the computation of the seismic response of base-isolated buildings are presented, by considering extreme earthquake loading conditions. Analytical models are reported for Curved Surface Slider devices, calibrated through the experimental outcomes of tests performed at the Laboratory of EUCENTRE Foundation in Pavia (Italy). In addition, simplified dynamic systems are defined, which allow fast assessments of the global response of a base-isolated structure, even though extreme seismic events are applied. Results have been compared to the response returned by an experimental hybrid simulation, in order to evaluate the accuracy of the presented dynamic systems.
... In addition, the growing interest in investigating the experimental response of such devices have highlighted a number of important aspects to account for, such as some dependencies of the friction coefficient with respect to specific response parameters, in terms of sliding velocity, vertical load/contact pressure and cyclic effect ( [6] , [11] ). These dependencies have been detected, by analyzing the force response of both flat and curved sliding motions, and consequently a certain correlation between these different loading conditions is expected, regardless the applied loading conditions ( [7] , [9] , [2] ). From a theoretical perspective, the lateral force of a Curved slider can be computed as the summation of a recentering behavior, modeled as a linear spring with respect to displacements and a frictional response ( [3] , [10] ). ...
Conference Paper
Full-text available
Curved Surface Slider devices have been widely used in last years for the protection of both building and structural systems. The spherical shape of the implemented sliding surfaces provide a certain recentering capability, which is generally combined to significant amount of energy dissipation, due to the frictional characteristics of the adopted sliding material. Since both behaviors act simultaneously during motion, experimental tests could return significantly high force values, especially if large bearings are considered. In some of those cases, the maximum force capacity of the testing equipment can be even overcome, and consequently experimental tests can not be performed. The scope of the present work is to provide experimental evidence of the comparison between flat and curved sliding motions. Precisely, the outcomes of bi-directional tests performed on on full-scale Double Curved Surface Slider and Flat Slider devices have been analyzed. On the former typology the frictional and the recentering behaviors have been numerically de-coupled, in order to compare the obtained results to the frictional response of the latter device. Results have shown a good agreement between the considered sliding motions, which seems to suggest that the experimental evaluation of flat sliding characteristics could be representative of curved sliding motions.
Conference Paper
Full-text available
Recent seismic events like Chile in 2010 and Japan in 2011 confirmed the seismic isolation technology as capable to provide structures with a high level of protection in case of earthquakes of exceptional intensity. This relatively young protection technique continues to experience a steady increase of applications beyond particular structures like bridges and buildings of strategic importance. Even though the use of seismic isolators and energy dissipators is still surrounded, in some countries, by a certain level of diffidence, the confidence level in its potential and reliability was significantly improved, in the last decade, by extensive experimental campaigns completed at newly designed and dedicated testing facilities. The availability of testing rigs of large capabilities, able to perform testing programs of full scale devices under realistic loading conditions allowed both the certification of unprecedented devices as well as the improvement of the knowledge of the behavior of existing solutions. The experimental activity contributed as well to the development of ad hoc codes that regulate the testing protocols, procedures and acceptance criteria for all the device’s typologies. In this paper, recent observations generated by the completion of many tests on a large variety of seismic isolators and energy dissipators, are presented and discussed. In particular the effects of applied load and velocity on the performance parameters of lead-rubber bearings and friction pendulum systems, the complex behavior of double concave sliding devices, the urgent issue of the assessment of the durability of material and assemblies, will be discussed. The evidence from past experience points towards the need for a more pro-active role for the structural engineer A deeper understanding of the device’s performance, beyond the specific design requirements established by the codes, is indeed a new crucial duty of the designer that is supported in this effort by the industry, the academia as well as the experimental facilities.
Article
Full-text available
A shake-table study of the friction-pendulum isolation system, installed in a six-story, quarter-scale, 52-kip model structure, is presented. Two bearing materials are studied, one with a peak friction coefficient of 0.075 and another of 0.095. In both cases, the isolation system has a rigid-body mode period of 1 sec. The isolated structure is found to be capable of withstanding strong earthquake forces of different frequency content. In tests with the El Centro motion, the isolated structure sustains, without any damage, a peak ground acceleration six times greater than what it could under fixed-based conditions. It is found that the bearing displacements are low and that the permanent bearing displacements at the end of free vibration are very small, in general, not exceeding 6% of the bearing design displacement. The system is shown to have quantifiable properties, and analytical techniques are presented that provide accurate prediction of the response.
Article
In this paper, an experimentally validated model is proposed in order to take into account main sources of performance degradation that could be experienced by friction-based devices during a seismic event. Particular attention is dedicated to the degradation of friction characteristics due to repetition of cycles and consequent temperature rise. This effect can be responsible for overestimate of the dissipation capacity of the device. The proposed model of frictional behavior is suitable for immediate implementation in generalized structural analysis codes and provides an important design tool for realistic assessment of the seismic response of structures equipped with friction-based isolators.
Article
The writers presented in an earlier paper a mathematical model of friction of sliding bearings under conditions of compression and bidirectional motion. The model could account for the dependence of the coefficient of friction on the velocity and direction of sliding. The model was capable of reproducing the behavior observed in experiments with unidirectional motion, however, no experimental validation of the model was provided for bidirectional motion. The present paper presents experimental results on Teflon sliding bearings under simultaneous compression and high velocity bidirectional motion that provide a complete verification of the presented model. Furthermore, the experimental results show the existence of bidirectional interaction between the orthogonal components of the frictional force at the bearing interface. The importance of these effects is studied in the analysis of an isolated structure. The structure is subjected to a number of pairs of horizontal earthquake excitation components that are representative of seismic zone 4, soil type S2seismic motion. It was found that neglect of these effects leads to underprediction of the isolation system displacement and overprediction of the structural shear.
Article
A mathematical model of the factional behavior of Teflon sliding bearings for conditions of interest in base isolation is developed. The calibration of the model is based on extensive experimental data that were presented in an accompanying paper. The model is capable of accounting for: (1) Unidirectional and multidirectional motion at the Teflon-steel interface; (2) velocity and pressure dependence of the coefficient of sliding friction; and (3) breakaway (or static) friction effects. The model is characterized by four parameters. These are the minimum and maximum values of the sliding coefficient of friction, the ratio of breakaway to sliding coefficient of friction at initiation of sliding and a parameter that describes the variation of the sliding coefficient of friction with velocity. Values of these parameters are presented for sixteen combinations of type of Teflon, bearing pressure and condition (surface roughness) of mating steel surface. Applications of the model in the analysis /design of a sliding isolation system are presented and the effects of bearing pressure and breakaway friction are evaluated. Furthermore, an assessment of the implications of using Coulomb's constant friction model rather than the developed model is presented.
Article
An analytical model is proposed to describe the interfacial sliding characteristics of Teflon and stainless steel based on experimentally observed quasistatic and dynamic sliding characteristics. The effects of normal pressure, sliding distance, normal pressure history, sliding velocity, sliding velocity history, normal pressure rate, sliding work, etc., are included. The dependence of the dynamic friction force on both the normal pressure and the sliding velocity is uncoupled in this formulation. The dynamic friction force is determined by multiplying the quasistatic friction force by an amplification factor. The amplification factor is a pure function of sliding velocity. The proposed model is validated by a quasi-static test and a dynamic sliding test. In the quasi-static verification test, the applied normal contact pressures are changing during sliding. For the dynamic validation test, the Teflon-stainless-steel interfaces are subjected to varied normal pressures and sliding velocities. Very good agreement between the predicted and the experimental results is obtained.
Article
Dynamic tests were done on polytetrafluoroethylene (PTFE) sliding bridge bearings to evaluate their frictional characteristics at frequencies above 1 Hz. High frequency seismic responses of PTFE bearings are expected to occur in bridge structures in eastern North America, where the estimated ground motion is well above 1 Hz. Three different PTFE-steel interfaces were tested at frequencies ranging from 0.02 Hz to 5 Hz, with displacement amplitudes of up to ±70 mm and under confining pressures ranging from 5 MPa to 45 MPa. The experimental results showed a significant initial transient frictional response at frequencies above 1 Hz before the dynamic, steady-state, frictional behavior was achieved. This transient response is characterized by an initial high static coefficient of friction that slowly decreases to a dynamic steady-state coefficient of friction after several response cycles. Modifications to an existing mathematical friction model are proposed to take into account this initial transient frictional response.
Article
Different modelling aspects of structures isolated using the frictional pendulum system and subjected to earthquake ground motions are studied herein. Although the vertical dynamics of these structures is given special emphasis, other effects such as large isolator deformations and bidirectional input motion are also considered. Different structural models of the FPS are developed and tested for single-storey structures and a real four-storey building frame; among them, an ‘exact’ formulation of the FPS force–deformation constitutive relationship is presented. Results show that global building responses can be computed within 20 per cent error in the mean using a simplified model that ignores the vertical motion of the building; however, structural member deformations and forces need to be computed using a model that considers such motion. This is of particular importance when there exist correlation between the horizontal and vertical components of ground motion. Further, a physical model of the FPS is introduced and used to determine the response of a real four-storey frame, including uplift and downward impact. Results from this analysis show that local column responses may vary substantially depending on the stiffness of the isolation storey and the presence of a mass at the isolation level. Such mass is capable of filtering the large increase in column shear that results from the impact of the structure after uplift. Uplift occurs at several instants of the response of the structure considered, leading to an increase in column base shear as large as 3 times the shear obtained by ignoring the vertical dynamics of the building. © 1998 John Wiley & Sons, Ltd.
Article
In this paper, an advanced analytical model and finite element formulations including local bending moment effects for the Friction Pendulum System (FPS) are presented. Some important findings in these investigations reveal that local bending moments which result from the movements of the isolator, should be accounted for in the engineering design. The results presented in this paper show that if the effects of local bending moments are excluded from design considerations, it may endanger structures during earthquakes. The new formulations considering local effects provide an effective tool to simulate the behaviour of FPS-isolated structures subjected to earthquake loadings. The proposed formulations can be applied directly for both two- and three-dimensional analyses without any further assumption. It is easy to implement the proposed techniques in existing finite element programs. The results obtained from the numerical analyses suggest that the local bending moment effects are of significant importance, and should be taken into account to assure the safety of isolated structures during earthquakes. © 1997 by John Wiley & Sons, Ltd.