Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/236154556
CITATIONS READS
7 66
2 authors:
All content following this page was uploaded by Andrea Dall'Asta on 18 January 2014.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Engineering Structures 33 (2011) 786–795
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
1. Introduction isolated bridges, with isolation bearings placed both at the piers
and abutments. In PRSI bridges, two different load paths may
In the last three decades the introduction of seismic isolation be identified under transverse seismic excitation. The first path
devices between the superstructure and substructures has become involves the abutments and superstructure, which under inertia
a widely used tool for the protection and retrofit of bridges in forces shows significant bending, while the second path involves
seismic areas [1,2]. The basic aims of the seismic isolation of the piers/bearings system. The assumption of rigid displacement
bridges are (a) decoupling the movement of the substructures of the superstructure, which simplifies the design of fully isolated
from that of the superstructure, (b) increasing the fundamental bridges, is not applicable to the case of PRSI bridges. Moreover, the
period of the bridge and thus shifting the fundamental frequency of period elongation effect may at times be limited in partial isolation.
vibration to a range where the energy content of the earthquake is In fact, an upper bound to the fundamental period of vibration
reduced and (c) providing additional sources of damping. There are is given by the fundamental period of the superstructure free to
however some drawbacks to using isolation bearings on pier and move at the piers, which behaves as a beam simply supported at
abutment tops. In some situations, the period shift due to isolation the abutments. Furthermore, in the case of dissipative isolation
may lead to excessive displacements of the deck. Moreover, bearings, the global dissipation capacity of PRSI bridges may be
bridge isolation requires the use of expensive bidirectional reduced as a result of the strain energy contribution due to deck
joints at the abutments in order to accommodate superstructure elastic deformation. This structural solution is however often
displacements. These aspects may lead to the choice of restraining adopted thanks to its capability to involve the highest number of
the transverse motion of the superstructure at the abutments, in substructures in resisting the inertia forces and to avoid the use of
both seismically isolated [1,3–5] and non-isolated bridges [6–8]. expensive bidirectional joints at the abutments.
The seismic response of isolated bridges with transverse restraint To date, many simplified models and procedures have been
at abutments, also called ‘‘partially restrained seismically isolated’’ proposed for the analysis and design of fully isolated bridges
(PRSI) bridges in [5], is significantly different from that of fully [6,9,10]. These methods are mainly based on a rigid behaviour
of the superstructure in the transverse direction, which usually
yields the same isolator characteristics in the case of piers with
∗ Corresponding author. Tel.: +39 3396413360. equal height. In PRSI bridges, however, the deck shows a bending
E-mail addresses: etubaldi@libero.it (E. Tubaldi), a.asta@tin.it (A. Dall’Asta). deformation in the transverse direction so that the rigid model
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2010.12.002
E. Tubaldi, A. Dall’Asta / Engineering Structures 33 (2011) 786–795 787
The procedure can be applied both to new bridges and existing A simplified model is derived by reducing the global degree
bridges. For example, when designing a new bridge, it may be of freedoms (DOFs) of the system to the deck DOFs only. For this
necessary to obtain the same value of transverse bending moment scope, the pier-bearings system is condensed by substituting the
at the base of the piers since this would result in the same amount series arrangement of the two Kelvin-type models by an equivalent
of longitudinal reinforcement and the same foundation for all the single Kelvin-type model (subscript ‘‘c’’) with stiffness kc ,i and
piers. When retrofitting an existing bridge, it is often useful to damping constant cc ,i , so that the force fc ,i resisting the deck
design the bearing system in such a way that the bending moment motion in the transverse direction in correspondence of the i-th
demand at the base of each pier is sufficiently smaller than the pier-bearing is equal to:
bending moment capacity.
The proposed method is based on an approximate single degree fc ,i = kc ,i yi + cc ,i ẏi (2)
of freedom (SDOF) description of the dynamic behaviour of the where yi and ẏi are the displacement and the velocity of the
PRSI bridge. The dynamic problem is posed in a variational form superstructure in correspondence with the isolation bearings.
and the solution is sought in a reduced functional space generated The equivalence between the two visco-elastic systems in
by only one function. For this purpose, a half-cycle sinusoidal curve Fig. 1(b) may be stated for a stationary oscillatory motion with
may be conveniently assumed as displacement shape function of circular frequency ω. The motion of the top of the i-th pier-bearing
the bridge in transverse direction. Once the properties (stiffness system, yi = eiωt , is the sum of the two contributions due to the
and damping factor) of the equivalent SDOF are evaluated, the pier and the bearing. These systems vibrate at the same circular
displacement and force demands on the bridge components can frequency and their motion can be posed in the form yp,i = yp,i eiωt
be determined by comparison with the design demand spectrum.
and y0,i = y0,i eiωt . Thus, yi can be expressed as:
The bearing properties are iteratively adjusted in order to ensure
the desired performance of the piers. Only at the end, is the fp,i f0,i
performance of the superstructure and abutments checked. yi = yp,i + y0,i = + . (3)
kp,i + iωcp,i k0,i + iωc0,i
2.2. Bridge model Recalling that fp,i = f0,i in a series system, it is possible to express
these forces in function of the total displacement yi as
The partially restrained bridge is modelled as a 2-dimensional
(kp,i + iωcp,i )(k0,i + iωc0,i )
beam resting on discrete visco-elastic supports representing the fp,i = f0,i = yi . (4)
pier-bearing systems and pinned at the abutments (Fig. 1(a)). The (kp,i + k0,i ) + iω(cp,i + c0,i )
i-th pier-bearing system is idealized as a series arrangement of
On the other hand, the response of the equivalent system is:
two Kelvin-type models (Fig. 1(b)), respectively describing the pier
(subscript ‘‘p’’) and the bearing (subscript ‘‘0’’). Each Kelvin-type fc ,i = (kc ,i + iωcc ,i )yi . (5)
model is a parallel arrangement of a spring with stiffness k and
The original and equivalent systems provide the same response
a damper with damping coefficient c. The dissipative property of
only if the real and imaginary parts of the resisting forces coincide.
each component may also be described in terms of the damping
This permits obtaining the following expressions for the stiffness
factor ξi , defined as:
and damping constant of the equivalent system
ci ω Wξ ,i
ξi = = (1) kp,i k0,i (kp,i + k0,i + 4ξp2,i kp,i + 4ξ02,i k0,i )
2ki 4π WS ,i k c ,i = (6)
where Wξ is the energy dissipated by the damper per cycle and WS
(kp,i + k0,i )2 + (2ξp,i kp,i + 2ξ0,i k0,i )2
is the maximum strain energy stored in the spring, evaluated for a (ξp,i k0,i + ξ0,i kp,i − 4kp,i ξp2,i ξ0,i − 4k0,i ξp,i ξ02,i )
harmonic motion with circular frequency ω. ξ c ,i = (7)
Since the piers are assumed not to yield, their behaviour may (kp,i + k0,i + 4ξp2,i kp,i + 4ξ02,i k0,i )
be described as linear elastic with kp,i equal to their effective where ξc ,i = cc ,i ω/2kc ,i .
flexural stiffness. The values of kp,i depend on the pier height, pier The isolator damping factor ξ0,i is generally larger than the
cross-section shape and the amount of longitudinal reinforcement. pier damping factor ξp,i and the global damping ξc ,i assumes
The effective value of kp,i accounting for concrete cracking may intermediate values controlled by the stiffness ratio k0,i /kp,i . If the
be evaluated based on moment–curvature analysis. The damping damping provided by the isolation system is not very high with
factor ξp,i associated to piers is usually in the range 2%–5% [6]. respect to the pier damping and if the ratio between the bearing
Commonly used isolation bearings are characterized by differ- and the pier stiffness is low, Eqs. (6) and (7) may be reduced to the
ent behaviours, which may be one or a combination of hysteretic, simplified formula already used in [5,10]:
viscous or friction behaviour [6,12]. In preliminary design and of-
ten in current practice, equivalent linear visco-elastic models are 1
usually adopted to represent the isolation devices. The stiffness of k c ,i ∼
= 1 1
(8)
kp,i
+ k0,i
the equivalent model k0,i is usually posed equal to the secant stiff-
ness at maximum displacement while the damping factor ξ0,i de- ξp,i k0,i + ξ0,i kp,i
scribes the dissipative properties and is usually derived from cyclic ξ c ,i ∼
= . (9)
kp,i + k0,i
tests at a given oscillatory frequency. Different formula may be
used [9,10,13–15] to estimate k0,i and ξ0,i , depending on the de- Fig. 2 shows the normalized values of damping factor ξc ,i /ξ0,i
vice used. (Fig. 2(a)) and of stiffness kc ,i /k0,i (Fig. 2(b)) of the pier-bearing
E. Tubaldi, A. Dall’Asta / Engineering Structures 33 (2011) 786–795 789
Fig. 2. Damping ratio ξc ,i /ξ0,i (a) and stiffness ratio kc ,i /k0,i (b) vs. ratio k0,i /kp,i for different ξ0,i /ξp,i ratios and ξp,i = 5%.
system, versus the ratio of bearing to pier stiffness, for different slender and heavy piers), this contribution cannot be neglected and
values of bearing damping factor ξ0,i and assuming ξp,i = 5%. can be accounted for by adding the relevant term in Eq. (12).
The simplified formulas of Eqs. (8) and (9) (dashed line in The internal and external actions are expressed in terms of the
Fig. 2) overestimate the system damping and underestimate the displacements based on the following constitutive relations:
system stiffness with respect to Eqs. (6) and (7). The error in
Md (x, t ) = b(x)y′′ (x, t )
estimating ξc ,i may be up to 15%–20% for high values of bearing (13)
Mv (x, t ) = a(x)ẏ′′ (x, t )
damping factor and k0,i /kp,i ratios. The error in the approximation
of kc ,i introduced using Eq. (8) is usually smaller, and increases where b(x) represents the stiffness of the deck section for trans-
for increasing values of k0,i /kp,i ratios and for increasing values verse bending and a(x) is the deck damping function.
of ξ0,i . A relative error of 10% is observed for k0,i /kp,i = 0.4 and By substituting Eqs. (2) and (11)–(13) into Eq. (10), the following
ξ0,i /ξp,i = 6. expression is obtained:
∫ L ∫ L
2.3. Generalized SDOF for approximating structural behaviour b(x)y′′ (x, t )η′′ (x)dx + a(x)ẏ′′ (x, t )η′′ (x)dx
0 0
N
The properties of the equivalent SDOF system are obtained −
according to the Lagrange–D’Alembert principle [16]. Let U be a
+ [kc ,i y(xi , t ) + cc ,i ẏ(xi , t )]η(xi )
suitable space of real-valued displacement functions u(x):(0, L) →
i=1
∫ L
R defined along the bridge length L and satisfying the kinematic
=− m(x)[ÿ(x, t ) + ÿg (x, t )]η(x)dx ∀η ∈ U ; t ∈ [t0 , t1 ].
(essential) boundary conditions. Let y(x, t ) : [t0 , t1 ] → U 0
be the motion, defined in the time interval considered [t0 , t1 ]. (14)
The variational principle states that the external virtual work
δ WE (y, η) resulting from external and inertia forces acting through A series decomposition may be used to express the motion
every virtual displacement η ∈ U is equal to the internal virtual y(x, t ) as the product between the generalized coordinates Zn (t )
work δ WI (y, η) resulting from the stresses acting through the (depending only on time) and the associated functions ψn (x)
corresponding virtual strains: which serve as a base for the space U (repeated index denotes
summation):
δ WE (y, η) = δ WI (y, η) ∀η ∈ U ; t ∈ [t0 , t1 ]. (10)
y(x, t ) = Zn (t )ψn (x). (15)
It is assumed that the internal virtual work δ WI has the following
In order to approximate the system to a SDOF system, a reduced
expression:
space is considered which consists only of the first function of
L L the series, ϕ(x) = ψ1 (x). The balance condition is expressed as
∫ ∫
δ WI (t ) = Md (x, t )η′′ (x)dx + Mv (x, t )η̇′′ (x)dx follows:
0 0
N mZ̈ (t ) + (c d + c c )Ż (t ) + (kd + kc )Z (t ) = −meff ÿg (t ) (16)
−
+ fc ,i η(xi ) (11) in which:
i =1 ∫ L
where N is the number of piers, Md (x, t ) is the superstructure m= m(x)ϕ 2 (x)dx (17)
elastic transverse bending moment and Mv is the superstructure 0
L
Based on Eq. (29), the damping factor ξ of the equivalent
∫
meff = m(x)ϕ(x)dx. (22) system, evaluated according to Eq. (24), may be also expressed as
0
The circular frequency of the equivalent SDOF system is expressed N π xi
ωd2 ξd + 2ωc2,i ξc ,i sin2
∑
as ω2 = (kd + kc )/m. L
i=1
Since the expression for the damping function a(x) is often not ξ= . (30)
known, the contribution of the deck to the generalized damping ω 2
may be expressed by assuming a prefixed ratio between the energy The SDOF approximation generally provides a satisfactory local
dissipated by the deck and the maximum energy stored in a cycle description of the displacement field, y(x, t ), and of relevant
at the circular frequency ω. This ratio can be related to the damping derived quantities, as the elastic reaction and viscous force at the
factor ξd and the relative damping coefficient c d may be coherently i-th pier-bearing system. It is however evident that the quantities
expressed as c d = 2kd ξd /ω. Estimated values of ξd based on related to higher order spatial derivatives cannot be described
dynamic tests may be found in [17]. Usually, ξd assumes values in locally with the same accuracy and a less satisfactory estimation
the range 2%–5% [6]. is obtained by considering the first term of the series only. In
The SDOF problem is usually expressed by dividing Eq. (16) by particular, it can be interesting to evaluate in a more refined way
m, thus obtaining the deck’s transverse curvatures (bending moments), related to the
second order derivative of the displacements, and the abutment
Z̈ (t ) + 2ξ ωŻ (t ) + ω2 Z (t ) = −ρ ÿg (t ) (23) reactions, which can be obtained by the deck shear at the end
where ρ is the participation factor, defined as ρ = meff /m. points and, therefore, are related to the third order derivative of the
In Eq. (23), the global dissipative properties of the SDOF system displacements. In order to estimate these quantities with a better
are described in terms of the generalized mass m and stiffness degree of accuracy, the contribution of higher terms of the series
kd + kc as may be included, based on the derivations reported in Appendix.
N
2.4. Design algorithm
ξd kd + kc ,i ξc ,i ϕ 2 (xi )
∑
(c d + c c )ω i =1
ξ= = (24) The bridge geometry and material characteristics are assumed
2(kd + kc ) kd + kc to be known, so that the deck stiffness can be evaluated. Suitable
where the global damping factor ξ results from the weighted sum values for structural damping ξd and ξp,i are assumed for the deck
of the contribution of the deck and pier-bearing system. and piers. The damping properties of the isolation bearings are
Different choices may be made for the base series ψn (x) and its known and are described by ξ0,i , which eventually depends on the
first term ϕ(x) = ψ1 (x). A convenient choice for the problem is the displacement amplitude and on the bridge oscillatory frequency.
sine-only Fourier series: The target values of internal actions in the piers are chosen by
nπ x assigning the values of bending moment Mp,i or shear Vp,i =
ψn (x) = sin . (25) Mp,i /hp,i that must be transmitted to each pier, whose height is
L denoted by Hi . The pier stiffness is evaluated based on the amount
The functions of this Fourier series coincide with the vibration of reinforcement provided at the base of the pier following the
shapes of a simply-supported beam with a uniformly distributed prescriptions of EC8-part 2 [18]. The values of displacement at the
mass and uniformly distributed stiffness. If the superstructure’s pier top are equal to dp,i = Vp,i /kp,i .
stiffness is significantly higher than the pier-isolation bearing In accordance with a displacement based design framework [6],
system’s stiffness, then the influence of the restraint exerted by the seismic demand is described by means of the displacement
the pier-bearing system on the dynamic transverse deformed response spectrum Sd (ω, ξ ), where ω is the circular frequency of
shape is small and the scatter between the vibration shapes of the SDOF system and ξ is the damping coefficient. For the case
the bridge and the functions of the Fourier series is quite small. considered, given a reference input spectrum, the maximum value
As a consequence, the first term of the series ϕ(x) = sin(π x/L) of the displacement experienced by the bridge’s midspan point
is adopted to represent the transverse deformed shape and the during the design seismic action can be evaluated as
parameters describing the equivalent SDOF system can be obtained
according to Eqs. (17)–(22). These can be made explicit in the case ymax = max(Z (t )) = ρ Sd (ω, ξ ) (31)
t
of constant stiffness b(x) = EId and constant mass m(x) = md :
where ρ = 4/π for stiffness and mass constant along the deck.
md L The stiffness and dissipative properties of the isolation system
m= (26) can be obtained by means of the following steps:
2
0. The maximum displacement ymax at midspan is assumed
π 4 EId N
− πx
i as the parameter describing the displacement field. An initial
kd + kc = + kc ,i sin2 (27)
2L3 L value may be chosen by considering the deck contribution only
in the evaluation of the SDOF system properties, kd and ξ d ,
i=1
3.1. Design examples Vp,1 = 740.74 kN kp,1 = 14863.25 kN/m dp,1 = 0.0498 m
Vp,2 = 594.06 kN kp,2 = 7352.94 kN/m dp,2 = 0.0808 m (40)
Two twin-girder steel–concrete composite (SCC) bridges are Vp,3 = 495.87 kN kp,3 = 4198.15 kN/m dp,3 = 0.118 m.
chosen intentionally as a case study in order to illustrate the
application of the procedure and investigate the limits of the The superstructure has a variable effective stiffness b(x). The
proposed design method. The cases considered are a regular bridge values corresponding to the different cross-sections (distributed
configuration (Fig. 3(a)), where all the piers have the same height, symmetrically with respect to deck midspan) are reported in
and an irregular bridge configuration (Fig. 3(b)), where the piers Table 1.
have different heights. The second configuration is selected in For design purposes, the weighted mean value of deck’s
order to highlight the ability of the procedure to regularize the effective stiffness EId = 1.1E + 09 kNm2 is considered. The deck’s
792 E. Tubaldi, A. Dall’Asta / Engineering Structures 33 (2011) 786–795
Fig. 3. Bridge configurations analyzed: (a) regular bridge and (b) irregular bridge.
Table 2
Bearing properties at the different pier-bearing locations.
i-th pier Regular bridge Irregular bridge
1 2 3 1 2 3
Fig. 5. Design response spectrum vs. response spectra of (a) artificially generated and (b) natural records.
Table 3
Comparison of design vs. analysis results for the regular configuration.
Analysis method ymax (m) Mp,1,max (kNm) Mp,2,max (kNm) Mp,3,max (kNm) Rab,max (kN) d0,1,max (m) d0,2,max (m) d0,3,max (m) Md,max (kNm)
Table 4
Comparison of design vs. analysis results for the irregular configuration.
Analysis method ymax (m) Mp,1,max (kNm) Mp,2,max (kNm) Mp,3,max (kNm) Rab,max (kN) d0,1,max (m) d0,2,max (m) d0,3,max (m) Md,max (kNm)
equivalent system. In order to evaluate the level of approximation The seismic response obtained from the different models and
obtained, the response of the simplified model (SM) is compared analysis methods are reported in Table 3 (regular bridge) and in
with the results provided by a 3-dimensional finite element (FE) Table 4 (irregular bridge). The following quantities are compared:
model with a more refined description of the displacement field. the maximum value of the deck midspan displacement ymax , the
The dynamic response under seismic input of the FE model is pier base moment Mp,i,max , the abutment reactions Rab,max , the
evaluated by means of modal dynamic analysis (MDA), time relative bearing displacements d0,i,max and the transverse bending
history analysis with artificial accelerograms (THA) and time moment demand Md,max at the deck’s midspan.
history analysis with natural accelerograms (THN). While MDA is Globally, satisfactory agreement between the design values and
performed to highlight the contribution of higher vibration modes, the calculated values is observed. The relative error between the
THA and THN are employed to highlight the effects related to the measured and design quantities is quite low, especially when SM
non-classical nature of the damping. In MDA, the design spectrum and MDA are considered. The estimate of the abutment reactions
is defined in such a way that the damping factor corresponding is quite accurate. Significant scatter between the design and
to the first mode of vibration is ξ while the damping factor obtained values is experienced only for the values of Mp,i . The
corresponding to the higher modes is equal to ξ d = 0.02. In maximum relative difference between pier moment demand and
the corresponding design value is about 10%–15%. It is however
the case of THA and THN, the results are deduced from a set of 7
noteworthy that pier base moment demands is influenced by
accelerograms and the mean values are reported. The 7 artificial
superior modes of vibration of the piers which are not taken into
spectrum-compatible accelerograms are generated according to
account in the design procedure, although simplified procedures
Eurocode 8 (ECS 2005) so that no value of the mean 5% damping
to include their influence may be considered [6].
elastic spectrum, calculated from all time histories, is less than
In order to evaluate the approximation in the response evalu-
90% of the corresponding value of the 5% damped elastic response
ation due to the assumed displacement field, the normalized di-
spectrum (ECS 2005). Natural accelerograms have been selected agrams of displacements and transverse bending moments along
from the European strong motion Database [19] in order to satisfy the deck due to MDA, THA and THN are reported in Fig. 6 and com-
the requirements of Eurocode 8 [18] and minimise the scatter from pared with the normalized corresponding design values. Only the
the code spectrum. The spectra of the ground motion record sets results for the irregular bridge configuration are shown, due to lack
are compared with the design spectrum in Fig. 5. of space.
The open source finite element program OpenSees [20] is used Again, satisfactory agreement is observed between the shape
to perform the time history analyses while SAP2000 [21] is used for according to SM and according to THA and THN (Fig. 6(a)). Bending
dynamic modal analysis. In both the simulation platforms a three- moment demands (Fig. 6(b)) according to MDA, THN and THA
dimensional model of the bridges is employed. In the time history analysis show some deviation from the SM shape, denoted by SM
analyses the bearings are represented by means of springs and (1). After including the contribution of the third vibration mode,
dampers acting in parallel while the dissipative properties of the the shape according to the SM, denoted by SM (1+3), becomes very
piers and deck are described by the Rayleigh damping matrix [16]. close to the shape according to MDA. Thus, the variation of deck
794 E. Tubaldi, A. Dall’Asta / Engineering Structures 33 (2011) 786–795
Fig. 6. (a) Normalized deformed shape and (b) normalized transverse bending moments along the deck according to the various analysis methods.
∫ L
stiffness does not affect significantly the bending moment shape.
The scatter between MDA and THA or THN may be attributed to [M]ij = m(x)ψi ψj dx (A.2)
0
the non-classically damped nature of the problem, not accounted
N
by MDA, and by the simplified rule used to combine the modal −
[Kc ]ij = kc ,p ψi (xp )ψj (xp ) (A.3)
responses. Finally, it is observed that the pier-bearing system does
p=1
not significantly influence the deck moment shape, as expected.
∫ L
It is noteworthy that the design procedure leads to a more
[Kd ]ij = b(x)ψi′′ ψj′′ dx (A.4)
regular bridge behaviour, where the deformed shape and the
0
transverse bending moments have an approximately symmetric
N
shape, despite the non regular piers configuration. −
[Cc ]ij = cc ,p ψi (xp )ψj (xp ) (A.5)
p=1