You are on page 1of 13

Bull Earthquake Eng (2011) 9:1703–1715

DOI 10.1007/s10518-011-9259-1

ORIGINAL RESEARCH PAPER

Semi-active control of the benchmark highway bridge


based on seismic early warning systems

Giuseppe Maddaloni · Nicola Caterino ·


Antonio Occhiuzzi

Received: 7 September 2010 / Accepted: 17 March 2011 / Published online: 2 April 2011
© Springer Science+Business Media B.V. 2011

Abstract Possible interactions between seismic early warning systems and structural con-
trol are herein investigated. A SEWS can provide the knowledge of some parameters, ahead
of time, of the seismic event that is going to occur in a certain site. Current research activities
on SEWS include the estimate of the anticipated peak ground acceleration (PGA) and the
acceleration response spectrum (Sa ) of an incoming earthquake. This paper proposes the
exploitation of this information in the framework of semi-active (SA) control strategies, by
using magnetorheological dampers. The latter are time-varying properties devices able to
achieve a wide range of physical behaviours using low-power electrical currents. The main
idea of this work consists in changing the MR damper behaviour according to the forecasted
intensity of an incoming earthquake provided by the SEWS, in order to obtain the optimal
seismic response of the hosting structure. This is investigated assuming a case-study prob-
lem. It consists in the 91/5 highway bridge located in Orange County of southern California
and proposed in literature as a benchmark structure for applications in the field of structural
control. The effectiveness of the proposed control strategy is assessed and compared to more
consolidated control techniques.

Keywords Semi-active control · Seismic early warning systems ·


Magnetorheological damper · Seismic protection of bridges

1 Introduction

Seismic early warning systems (SEWS) were originally intended as a rapid-rescue tool by
means of a quick analysis of earthquake intensity and location (Kanamori et al. 1997). While
the seismic waves travel from the origin to a given site, a rapid analysis of the main properties

G. Maddaloni (B) · N. Caterino · A. Occhiuzzi


Department of Technologies (DiT), University of Naples Parthenope, Centro Direzionale, Isola C4,
80143 Naples, Italy
e-mail: giuseppe.maddaloni@uniparthenope.it

123
1704 Bull Earthquake Eng (2011) 9:1703–1715

of the striking earthquake was thought to be fast enough to leave some time for alarming or
for shutting down critical plants.
The original front-detection early warning scheme is made up of seismic stations sur-
rounding a specific site at a given distance. At the occurrence of an earthquake, the stations
are assumed to be able to detect the seismic waves and to release an alarm to the site manage-
ment. The higher speed of electromagnetic signals compared to the seismic waves leaves a
tight margin, defined as “lead time”, that can be used to stop or to put in a safe mode the plant,
as shown by Wieland et al. (2000) in the case of a nuclear power plant. A different approach
was proposed in Allen and Kanamori (2003), based on the consideration that a front-detec-
tion SEWS works only if the earthquake originates away from the site to be protected. The
approach considers the detection of the arrival, at a given site, of the P-waves, which come
before the destructive S-waves, considering that the typical ground motion sequence is as
follows: time 1, arrival of P-waves; time 2, arrival of S-waves; time 3, peak ground motion.
Therefore, the lead time between times 1–3 can be utilized for warning. The alarm can be
modulated by a rapid analysis of the P-waves which can carry enough information to estimate
the intensity of the S-waves.
Recent studies (Iervolino et al. 2006) have shown that information carried by P-waves can
be processed in real time so as to obtain probabilistic distributions of magnitude and source-
to-site distance. Subsequently, similar estimates have been proposed for intensity measures
more useful for engineering applications such as peak ground accelerations (Iervolino et al.
2008), Fourier amplitude spectra (Kuyuk and Motosaka 2008), peak ground velocities (Zollo
et al. 2009) and acceleration response spectra (Convertito et al. 2008).
Semi-active (SA) structural control systems are based on the adoption of additional devices
able to modify in a short time their stiffness and/or their dissipating capabilities. These prop-
erties can be adjusted according to the instantaneous response of the hosting structures (feed-
back) and/or to the instantaneous properties of the earthquake input (feed-forward). In both
cases, the relationships between the observed quantities (displacements and velocities of the
structure, accelerations of the ground) and the corresponding optimal values of stiffness and
damping of the adjustable devices, are provided by analytical descriptions called control
algorithms. Therefore, in conventional SA control systems, the control chain includes:

• an acquisition phase, where selected parameters of the structural response and/or of the
ground acceleration are measured;
• a decision stage, where the measured quantities are processed by the control algorithm;
• a command activity, where decisions made by the algorithm are transferred to the adjust-
able devices, typically by means of electrical signals.

The control phases are usually designed to be continuously operated throughout the duration
of a seismic event, so as to continuously adjust the mechanical properties of the SA device in
order to optimize the structural response. This implies that an uninterruptible power source
be available during the event to feed the hardware involved in the control chain. Even though
the power requirements of SA control systems (in the order of magnitude of kW, considering
sensors, computers and power supplies) are smaller compared to those of fully active con-
trol systems (in the order of several MW), the probability of power outages during severe
earthquakes is obviously high.
The combination of the concepts behind SEWS and SA control suggests different
approaches to enhance the seismic structural response. The possibility of exploiting the
estimate of the anticipated intensity measures of an incoming earthquake to enhance the per-
formance of a structural control system was firstly envisioned by Kanda et al. (1994). Their

123
Bull Earthquake Eng (2011) 9:1703–1715 1705

proposal was intended to re-shape a structural control algorithm based on the estimate of the
spectral characteristics of an incoming event.
Pnevmatikos et al. (2004) proposed to escape structural resonance by changing the struc-
tural stiffness according to the dynamic characteristics of the incoming earthquake, measured
and processed in real time near-field or at the site. Qualification of the ground motion is based
on the fast Fourier transform (FFT) of the 1st part of the ground motion at the site (up to the
arrival of the S-waves) or, alternatively, on the FFT of the whole signal recorded near-field.
Although FFT is conceptually associated to steady-state signals and, therefore, its meaning
could be somewhat obscure in the case of seismic accelerograms, the paper introduces the
interesting idea of modifying the structural properties according to the frequency content of
an incoming earthquake. A similar proposal (De Iuliis et al. 2008) suggested to modify the
structural stiffness according to the FFT of the incoming earthquake based on a complex
performance index in a multi-choice scenario.
The authors of this paper have been working for some years on a simplified approach to
combine the exploitation of SA control and SEWS (Occhiuzzi et al. 2004, 2006, 2008a,b).
The basic idea is to add to a structure some passive but smart dissipating devices able to
adjust their mechanical properties according to some intensity measures of the incoming
earthquake forecasted by a SEWS. The adjustment is supposed to happen only once, just
before the quake actually strikes, and according to an estimate of either the anticipated peak
ground acceleration (PGA) or the anticipated spectral acceleration at the natural period of
structural vibration. In this case, provided that a SEWS be available, the structural control
framework is quite simple and is based on the possibility of re-shaping the energy dissipa-
tion properties of passive, but smart, additional damping devices shortly before the arrival
of the seismic event at a given location. The present paper describes the application of this
protection technique to the highway bridge presented in the “highway bridge benchmark”
framework, aimed to directly compare, on a given infrastructure, the amount of seismic pro-
tection corresponding to different strategies (Agrawal et al. 2009; Tan and Agrawal 2009;
Nagarajaiah et al. 2009). Thus, the proposed control strategy is numerically tested on a real
structure and its effectiveness is directly compared to many other techniques for seismic
protection of bridges.

2 The benchmark bridge

The structure used for the benchmark study is the 91/5 highway bridge, located in Orange
County, Southern California. A brief description of the bridge, its model and the operations
of the control system adopted is herein presented, whereas a detailed description can be found
in Agrawal et al. (2009). The superstructure of the bridge consists of two-span continuous
cast-in situ pre-stressed concrete. Each span is 58.5 m long, spanning a four-lane highway,
with two skewed abutments. Central support is provided by a 31.4 m long prestressed beam,
which rests on two columns approximately 6.9 m high. Pile groups, each consisting of 49
driven concrete friction piles, represent the column foundation. The width of the deck is
12.95 m. The total mass of the bridge is about 4,200 tons; the mass of the deck is about 3,200
tons (see Fig. 1).
In the actual bridge, the deck is isolated using four traditional non-seismic elastomeric
pads at each abutment, and a total of eight passive fluid dampers are installed between the
end abutments and the deck (four dampers at each end) to reduce seismic response.
The structural model of the bridge adopted as a reference structure is slightly different.
In the 3D FEM model developed in ABAQUS and proposed in the benchmark problem,

123
1706 Bull Earthquake Eng (2011) 9:1703–1715

Fig. 1 The benchmark bridge

lead rubber bearings (LRBs) are used in place of the elastomeric bearings and the passive
fluid dampers are not considered. Consequently, the uncontrolled structure, used as a basis
for comparison to quantify the effectiveness of various control systems, corresponds to the
model of the bridge isolated by four LRBs at each deck-end.
A linear behaviour is assumed for the superstructure of the bridge, including the deck and
the bent beam. Structural member nonlinearities are included to capture the inelastic moment–
curvature behaviour of the 2 columns. The 8 bearings are modelled as shear elements, i.e.
the vertical axial stiffness of the bearings is taken as infinite and their torsional rigidity and
bending stiffness are assumed to be zero. The shear force–displacement relationship in two
horizontal directions is considered to be bilinear.
From the 3D FEM model, the dynamic characterization of the bridge in the linear range
is obtained as follows. The first mode of the uncontrolled bridge is torsional with a natural
period of T1 = 0.813 s (frequency 1.23 Hz). The second mode is torsional coupled with
vertical (period T2 = 0.781 s). The third and the fourth modes are vertical and transverse,
respectively with periods T3 = 0.645 s and T4 = 0.592 s; the fifth and sixth modes are
second vertical and transverse (respectively T5 = 0.565 s and T6 = 0.307 s).
The seismic response of the benchmark bridge is investigated by nonlinear time-history
analyses by adopting six real earthquake ground excitations, namely (i) North Palm Springs
(1986), (ii) TUC084 component of Chi-Chi earthquake, Taiwan (1999), (iii) El Centro compo-
nent of Imperial Valley earthquake (1940), (iv) Rinaldi component of Northridge earthquake
(1994), (v) Bolu component of Duzce, Turkey earthquake (1999) and (vi) Nishi-Akashi com-
ponent of Kobe earthquake (1995). The 6 accelerograms cover a wide variety of magnitudes,
distances to fault and soil types, as reported in Agrawal et al. (2009). For the excitation of the
longitudinal (EW) and transverse (NS) directions of the bridge, both components are simul-
taneously used. The peak ground accelerations (PGA) of the earthquake ground motions are
shown in Table 1.
The acceleration response spectra at a 5% damping ratio (Fig. 2) shows that the ground
motions are characterized by a low homogeneity in terms of both energy and frequency
content.
In the benchmark paper, the bridge is supposed to be upgraded to reduce the seismic
response. Three types of sample control strategy, namely passive, active and semi-active,
including devices, control algorithms and sensors, are designed and presented for compari-
son in the relevant paper (Tan and Agrawal 2009).

123
Bull Earthquake Eng (2011) 9:1703–1715 1707

Table 1 Peak ground accelerations of six selected earthquake ground motions

Earthquake (year) Type PGA [g] PGA [g]


EW component NS component

North Palm Springs (1986) Far field 0.492 0.612


Chi-Chi (1999) Near fault 1.157 0.417
El Centro (1940) Far field 0.313 0.215
Northridge (1994) Near fault 0.838 0.472
Turkey (1999) Near fault 0.728 0.822
Kobe (1995) Near fault 0.509 0.503

3.50
3.25
3.00
2.75
2.50
2.25
2.00
Sa [g]

1.75
1.50
1.25
1.00
0.75
0.50
0.25
0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
T [s]

El Centro North Palm Springs Kobe Turkey Northridge Chi-Chi

Fig. 2 Elastic response 5% damping spectra of the EW component of the earthquakes used in the analyses

The passive strategy is based on 16 nonlinear viscous dampers, placed between the deck
and the abutments. At each side of the bridge, 4 dampers are placed in the longitudinal direc-
tion whereas the other 4 are in the transversal direction. The active strategy is based on 16
hydraulic actuators, placed as before, working according to the H2 /LQG control algorithm.
The actuators are assumed to be ideal, i.e., to precisely supply the force commanded by the
control algorithm. Actuator dynamics are neglected and no actuator–structure interaction is
considered, although these may occur in the physical control system. The semi-active strategy
is based on 16 magnetorheological (MR) dampers working according to a clipped H2 /LQG
control algorithm, used as a primary controller to calculate the desired control force. The
mechanical model of MR dampers is based on the Bouc–Wen hysteretic model (mimicking
a smoothed friction element) in parallel with a linear viscous dashpot.
With the purpose of evaluating the effectiveness of different control strategies, in the
benchmark paper, a set of evaluation criteria has been developed. Criteria J1–J8 are defined
to measure the reduction in peak response quantities of the benchmark highway bridge,
evaluated by normalizing the response quantities by the corresponding ones for the uncon-
trolled reference bridge; criteria J9–J14 are quantities based on normed responses over the
entire time duration of each earthquake; criteria J15–J16 are related to the control system.

123
1708 Bull Earthquake Eng (2011) 9:1703–1715

30 30

20 20

10 10

F [kN]
F [kN]

0 0

-10 -10

-20 -20

-30 -30
-20 -10 0 10 20 -200 -100 0 100 200
x [mm] v [mm/s]
i =0.0 A i =0.9 A i =1.8 A i =2.7 A

Fig. 3 Experimental force-displacement and force-velocity loops (Caterino et al. 2010) shown by a prototype
MR damper for imposed harmonic cycles at different current levels

Each criterion is organized so that a value less than 1 indicates a better performance of
controlled systems compared to the reference bridge.

3 Semi-active structural control based on a SEWS

Typically, SA control systems are based on smart additional devices able to modify their
mechanical properties so as to modify the stiffness and/or the damping properties of the
whole hosting structure. Modifications of their properties are made according to a control
algorithm which usually makes decisions based on the instantaneous values of some param-
eters of the structural response. MR dampers are particularly suitable to be adopted as smart
devices because their mechanical properties can be strongly modified (Fig. 3) in milliseconds
through the application of small electric currents (Occhiuzzi et al. 2003).
In order to work, a typical SA control system based on MR dampers needs a complex
electronic equipment, because: (1) the structure has to be properly instrumented in order
to measure quantities such as displacements, accelerations and devices’ forces during the
earthquake; (2) these signals have to be acquired and processed in real time by a high speed
CPU according to purposely written control algorithms, leading to the identification of the
instantaneous optimal value of the current feeding the devices; (3) the CPU has to command
special power supplies able to give, in almost real-time, the desired value of current to be
applied to each device. The effectiveness of such a system generally depends on the par-
ticular control algorithm chosen, as well as on the magnitude of unavoidable delays in the
control chain (acquisition-processing-command) and requires that all the involved tools (i.e.,
transducers, load cells, computer, power supplies) work sinchronously and flawlessly.
In the present study, two additional control strategies are implemented and compared
to the benchmark control ones. The control systems adopted are designed according to the
idea of combining SEWS and SA control techniques, in order to explore the potentialities
of feed-forward schemes in structural control. Classical, feed-back control schemes require
on-line measurement and processing of some parameters describing the structural response,
such as the relative displacement between the deck and the abutment. In both the proposed
cases, instead, these issues are simplified and/or eliminated. The proposed control systems do
not need on-line acquisition and processing of the structural response and can be conducted
relying on simple control electronics.

123
Bull Earthquake Eng (2011) 9:1703–1715 1709

SEWS like those described before can provide an estimate of the intensity of an incom-
ing ground motion at the site of the construction a few seconds before the quake actually
strikes. This estimate, at first tentative expressed in terms of PGA and subsequently in terms
of spectral acceleration (at the natural period T1 of the bridge), can be exploited to drive the
proposed SA control system. The deck is assumed to be connected to the abutments through
16 MR devices having the possibility to modulate their behaviour according to a command
voltage which, in turn, generates a current inside the dampers. The MR devices described in
the benchmark paper and adopted herein can be fed by voltages uc ranging from 0 to 10 V,
corresponding to current ranging from 0 to 5 A.
According to the first proposed strategy, the relationship:

uc = 0 if PGA < 0.45


uc = 13.33 · (PGA − 0.45) if 0.45 ≤ PGA < 1.2 (1)
uc = 10 if PGA ≥ 1.2

has been used to establish, given a PGA value (expressed in g) predicted by the SEWS, which
intensity of voltage uc (in V) should be supplied to the MR damper in order to achieve an
optimal response. Therefore, starting from the PGA estimate forecasted by the SEWS, the
voltage feeding the MR devices is set to a fixed value shortly before the quake actually strikes
at the site and kept constant throughout the event.
Similarly, the second proposed strategy considers, for each earthquake, the spectral accel-
eration Sa (T1 ), expressed in g, evaluated at the fundamental period of the bridge (T1 =
0.813 s) on the elastic 5% damped spectra. In this case, the voltage applied to the MR
damper is set as:

uc = 0 if Sa (T1 ) < 0.40


uc = 6.25 · (Sa (T1 ) − 0.40) if 0.40 ≤ Sa (T1 ) < 2.0 (2)
uc = 10 if Sa (T1 ) ≥ 2.0

Equations 1 and 2 have been derived according to the results of many numerical analyses
aimed to find the value of uc corresponding to optimal response, in terms of J1 and J2 criteria,
for any given earthquake. The piecewise linear relationships expressed by Eqs.1 and 2 have
been found by fitting these optimal solutions.
The nonlinear analyses of the MR controlled bridge have been performed using the SIMU-
LINK model proposed in the benchmark paper and adopting for each earthquake the voltage
value obtained by the above defined relationship for the PGA and Sa (T1 ) values of the EW
component (preliminarily chosen as the most significant for the present case). Table 2 shows
the voltages assumed, according to the two algorithms, for the accelerograms considered.

Table 2 Voltages assumed according to the two algorithms represented in Eqs. 1 and 2

Earthquake name PGA [g] Sa (T1 )[g] uc (Eq.1) [V] uc (Eq.2) [V]

El Centro 0.31 0.42 0.00 0.91


North Palm Springs 0.49 0.32 0.56 0.53
Kobe 0.51 0.55 0.79 1.10
Turkey 0.73 0.85 3.71 6.05
Northridge 0.84 0.91 5.17 10.00
Chi-Chi 1.16 0.90 9.43 10.00

123
1710 Bull Earthquake Eng (2011) 9:1703–1715

4 Numerical results

The peak values of different response quantities of interest under the described earthquake
motions are presented in Tables 3, 4, 5, 6 (NPS and NR denote North Palm Springs and North-
ridge earthquakes, respectively). The results corresponding to the 3 strategies adopted in the
benchmark paper are referred to as passive, semi-active and active. Madhekar and Jangid
(2009) proposed, on the same benchmark structure, 3 additional control strategies based on
linear viscous dampers, friction-type devices and 2-step SA viscous dampers. These results
are also compared in the following tables. The strategies introduced in the present paper are
referred to as SEWS-PGA and SEWS-Sa , respectively associated to the adoption of a control
scheme based on Eqs. 1 and 2. Bold numbers indicate the minimum values on each column
and for each criterium.
In Table 3, evaluation criteria J1 to J4 are compared for all six earthquakes. The two pro-
posed strategies lead to good results in comparison with those of other systems. In particular,
for the SEWS-PGA strategy, all the peak quantities assume values lower or in few cases
slightly higher than one. For strong earthquakes, such as Chi-Chi and Northridge, the control

Table 3 Evaluation criteria J1 to J4 for passive, semi-active, active, viscous damper, friction-type, two-step
viscous control strategies and the proposed ones SEWS-PGA and SEWS-Sa

Response quantity Control strategy NPS Chi-Chi El Centro NR Turkey Kobe Average

Peak base shear, J1 Passive 1.22 0.76 0.64 0.78 0.78 0.86 0.84
Semi-active 0.96 0.84 0.78 0.89 0.90 0.82 0.87
Active 0.95 0.88 0.79 0.90 0.91 0.79 0.87
Viscous 0.98 0.81 0.79 0.83 0.89 0.84 0.85
Friction-type 1.53 0.64 0.66 0.68 0.91 1.00 0.90
2-step viscous 1.00 0.82 0.86 0.84 0.91 0.88 0.88
SEWS-PGA 1.00 0.63 0.98 0.80 0.90 0.90 0.87
SEWS-Sa 0.99 0.61 0.91 0.67 0.88 0.88 0.82
Peak base moment, J2 Passive 0.63 0.96 0.58 0.96 0.88 0.55 0.76
Semi-active 0.75 0.98 0.71 0.98 0.98 0.67 0.84
Active 0.77 0.97 0.74 0.98 0.98 0.70 0.86
Viscous 0.73 0.97 0.72 0.97 0.98 0.69 0.84
Friction-type 0.79 0.95 0.36 0.94 0.56 0.50 0.68
2-step viscous 0.78 0.97 0.78 0.97 0.98 0.78 0.88
SEWS-PGA 0.86 0.94 0.95 0.96 0.98 0.84 0.92
SEWS-Sa 0.86 0.94 0.84 0.93 0.75 0.79 0.85
Peak mid-span displacement, J3 Passive 0.64 0.71 0.65 0.70 0.59 0.63 0.66
Semi-active 0.80 0.79 0.78 0.86 0.72 0.66 0.77
Active 0.82 0.80 0.78 0.87 0.75 0.70 0.79
Viscous 0.79 0.76 0.80 0.78 0.72 0.70 0.76
Friction-type 0.53 0.58 0.37 0.57 0.38 0.51 0.49
2-step viscous 0.82 0.84 0.87 0.81 0.80 0.79 0.82
SEWS-PGA 0.89 0.56 0.99 0.73 0.69 0.84 0.78
SEWS-Sa 0.89 0.53 0.92 0.54 0.50 0.79 0.69
Peak mid-span acceleration, J4 Passive 1.30 0.95 0.94 0.90 0.94 1.07 1.02
Semi-active 0.98 0.88 0.90 0.90 0.80 0.99 0.91
Active 0.79 0.88 0.88 0.84 0.80 0.90 0.85
Viscous 1.02 0.97 0.96 0.92 0.86 0.94 0.95
Friction-type 1.56 0.91 0.96 0.86 1.14 1.23 1.11
2-step viscous 1.04 1.03 1.02 0.96 0.95 0.88 0.98
SEWS-PGA 0.92 0.89 1.01 0.91 0.99 1.03 0.96
SEWS-Sa 0.92 0.87 1.03 0.84 1.18 1.03 0.98

123
Bull Earthquake Eng (2011) 9:1703–1715 1711

Table 4 Evaluation criteria J5 to J8 for passive, semi-active, active, viscous damper, friction-type, two-step
viscous control strategies and the proposed ones SEWS-PGA and SEWS-Sa

Response quantity Control strategy NPS Chi-Chi El Centro NR Turkey Kobe Average

Peak bearing displacement, J5 Passive 0.40 0.67 0.34 0.66 0.54 0.28 0.48
Semi-active 0.81 0.77 0.57 0.85 0.67 0.51 0.70
Active 0.94 0.80 0.64 0.88 0.71 0.59 0.76
Viscous 0.60 0.73 0.45 0.75 0.68 0.34 0.59
Friction-type 0.17 0.53 0.13 0.51 0.32 0.17 0.30
2-step viscous 0.65 0.80 0.49 0.78 0.75 0.42 0.65
SEWS-PGA 0.91 0.51 0.78 0.69 0.64 0.65 0.70
SEWS-Sa 0.92 0.47 0.53 0.48 0.45 0.58 0.57
Peak ductility, J6 Passive 0.63 0.60 0.58 0.59 0.19 0.55 0.52
Semi-active 0.75 0.70 0.71 0.83 0.37 0.67 0.67
Active 0.77 0.74 0.74 0.85 0.46 0.70 0.71
Viscous 0.73 0.66 0.72 0.69 0.36 0.69 0.64
Friction-type 0.79 0.41 0.36 0.37 0.12 0.50 0.43
2-step viscous 0.78 0.78 0.78 0.74 0.56 0.78 0.74
SEWS-PGA 0.86 0.38 0.95 0.62 0.29 0.84 0.66
SEWS-Sa 0.86 0.32 0.84 0.32 0.17 0.79 0.55
Peak dissipated energy, J7 Passive 0.00 0.23 0.00 0.33 0.00 0.00 0.09
Semi-active 0.00 0.47 0.00 0.57 0.24 0.00 0.21
Active 0.00 0.51 0.00 0.62 0.33 0.00 0.24
Viscous 0.00 0.41 0.00 0.44 0.13 0.00 0.16
Friction-type 0.00 0.03 0.00 0.12 0.00 0.00 0.03
2-step viscous 0.00 0.55 0.00 0.56 0.29 0.00 0.23
SEWS-PGA 0.00 0.03 0.00 0.38 0.04 0.00 0.08
SEWS-Sa 0.00 0.02 0.00 0.08 0.00 0.00 0.02
Maximum plastic connections, J8 Passive 0.00 0.67 0.00 0.75 0.00 0.00 0.24
Semi-active 0.00 0.67 0.00 1.00 0.33 0.00 0.33
Active 0.00 0.67 0.00 1.00 0.33 0.00 0.33
Viscous 0.00 0.67 0.00 1.00 0.33 0.00 0.33
Friction-type 0.00 0.33 0.00 0.50 0.00 0.00 0.14
2-step viscous 0.00 0.67 0.00 1.00 0.33 0.00 0.33
SEWS-PGA 0.00 0.33 0.00 1.00 0.33 0.00 0.28
SEWS-Sa 0.00 0.33 0.00 0.50 0.00 0.00 0.14

strategy SEWS-Sa yielded the best results. For medium intensity earthquakes, such as North
Palm Springs, Kobe and El Centro, the performance of SEWS-Sa strategy is comparable to
that of other control schemes and also to the response reduction obtained with more sophis-
ticated systems such as semi-active and active controllers. On average, the peak base shear
response is best controlled by the SEWS-Sa strategy.
Similar observations can be made for evaluation criteria J5 to J8 presented in Table 4. The
SEWS-Sa control strategy reduces the peak bearing displacement for Chi-Chi, Northridge
and Turkey earthquakes to half that of the reference bridge. Furthermore, the minimum dissi-
pated energy of curvatures at the end of members and the minimum number of plastic hinges
formed indicate that the SEWS-Sa strategy leads to the minimum amount of post-elastic
demand.
Comparison of evaluation criteria J9 to J12 is presented in Table 5. The SEWS-Sa control
strategy works well in reducing the normed base shear: the average value obtained from
all earthquakes is the lowest one with this control strategy. Similarly, for the norm base
moment, the proposed strategy shows improved response over the active controller in the
cases of Chi-Chi, El Centro, Northridge, Turkey and Kobe earthquakes. As for normed mid-
span displacement, the performance of the SEWS-Sa control strategy is more effective than

123
1712 Bull Earthquake Eng (2011) 9:1703–1715

Table 5 Evaluation criteria J9 to J12 for passive, semi-active, active, viscous damper, friction-type, two-step
viscous control strategies and the proposed ones SEWS-PGA and SEWS-Sa

Response quantity Control strategy NPS Chi-Chi El Centro NR Turkey Kobe Average

Norm base shear, J9 Passive 1.03 0.75 0.52 0.71 0.74 0.75 0.75
Semi-active 0.78 0.85 0.60 0.83 0.84 0.69 0.76
Active 0.74 0.89 0.68 0.87 0.89 0.74 0.80
Viscous 0.85 0.84 0.66 0.80 0.84 0.73 0.78
Friction-type 1.25 0.53 0.54 0.58 0.83 0.83 0.76
2-step viscous 0.82 0.88 0.73 0.83 0.90 0.80 0.83
SEWS-PGA 0.85 0.49 0.90 0.69 0.80 0.78 0.75
SEWS-Sa 0.86 0.47 0.64 0.55 0.84 0.74 0.68
Norm base moment, J10 Passive 0.53 0.75 0.39 0.72 0.37 0.52 0.54
Semi-active 0.66 0.80 0.56 0.84 0.50 0.66 0.67
Active 0.70 0.83 0.64 0.88 0.53 0.71 0.72
Viscous 0.68 0.79 0.62 0.81 0.51 0.69 0.68
Friction-type 0.62 0.57 0.27 0.75 0.23 0.44 0.48
2-step viscous 0.77 0.84 0.70 0.85 0.69 0.76 0.77
SEWS-PGA 0.81 0.54 0.88 0.69 0.41 0.74 0.68
SEWS-Sa 0.81 0.48 0.60 0.72 0.30 0.70 0.60
Norm mid-span displacement, J11 Passive 0.56 0.65 0.41 0.65 0.44 0.54 0.54
Semi-active 0.68 0.75 0.58 0.78 0.57 0.68 0.67
Active 0.70 0.78 0.66 0.81 0.61 0.73 0.71
Viscous 0.71 0.73 0.64 0.72 0.57 0.72 0.68
Friction-type 0.41 0.42 0.24 0.52 0.27 0.36 0.37
2-step viscous 0.79 0.79 0.72 0.78 0.64 0.79 0.75
SEWS-PGA 0.83 0.39 0.89 0.64 0.46 0.77 0.66
SEWS-Sa 0.83 0.35 0.62 0.47 0.36 0.72 0.56
Norm mid-span acceleration, J12 Passive 1.02 0.77 0.74 0.78 1.01 1.06 0.89
Semi-active 0.79 0.81 0.69 0.82 0.81 0.84 0.79
Active 0.72 0.79 0.69 0.80 0.80 0.80 0.76
Viscous 0.89 0.84 0.73 0.83 0.90 0.90 0.85
Friction-type 1.23 0.63 0.79 0.73 1.15 1.21 0.96
2-step viscous 0.87 0.92 0.82 0.90 0.90 0.93 0.89
SEWS-PGA 0.96 0.64 0.93 0.83 1.11 0.90 0.89
SEWS-Sa 0.96 0.63 0.80 0.78 1.17 0.91 0.88

passive, semi-active and active control schemes, yielding approximately 50% ∼ 75% response
reduction for the strongest earthquakes adopted in the analyses.
Table 6 presents evaluation criteria J13 to J16. The amount of normed displacement is
substantially smaller than the active strategy. From the results obtained according to criterion
J15, it can be observed that the proposed strategies are able to involve the minimum amount
of control forces, resulting in less expensive additional dampers.
In general, the results of the numerical analyses show that the proposed control strate-
gies can yield response improvement comparable to, or even better than, more sophisticated
semi-active or active systems. Furthermore, the response reduction provided by the pro-
posed techniques is less dependent on the characteristics of the ground motion than any other
passive system considered. The passive device referred to as friction, for instance, yields
generally good performances but causes an increase in the peak base shear of 53% compared
to the reference, uncontrolled bridge (i.e., in the case of North Palm Springs earthquake).
Being relatively simple, but smart enough to adjust itself to the properties of the incoming
earthquake, the proposed control system seldom performs worse than the reference bridge.

123
Bull Earthquake Eng (2011) 9:1703–1715 1713

Table 6 Evaluation criteria J13 to J16 for passive, semi-active, active, viscous damper, friction-type, two-step
viscous control strategies and the proposed ones SEWS-PGA and SEWS-Sa

Response quantity Control strategy NPS Chi-Chi El Centro NR Turkey Kobe Average

Norm bearing displacement, J13 Passive 0.25 0.62 0.25 0.62 0.29 0.20 0.37
Semi-active 0.45 0.75 0.39 0.77 0.41 0.36 0.52
Active 0.48 0.78 0.48 0.82 0.52 0.47 0.59
Viscous 0.33 0.71 0.42 0.70 0.38 0.28 0.47
Friction-type 0.10 0.39 0.11 0.47 0.17 0.10 0.22
2-step viscous 0.36 0.77 0.47 0.76 0.44 0.32 0.52
SEWS-PGA 0.47 0.35 0.60 0.60 0.32 0.42 0.46
SEWS-Sa 0.47 0.32 0.40 0.43 0.25 0.34 0.37
Norm ductility, J14 Passive 0.53 0.66 0.39 0.99 0.04 0.52 0.52
Semi-active 0.66 0.69 0.56 0.77 0.22 0.66 0.59
Active 0.70 0.65 0.64 0.83 0.24 0.71 0.63
Viscous 0.68 0.62 0.62 0.62 0.22 0.69 0.58
Friction-type 0.62 0.68 0.27 1.22 0.02 0.44 0.54
2-step viscous 0.77 0.51 0.70 0.72 0.52 0.76 0.66
SEWS-PGA 0.81 0.64 0.88 0.76 0.12 0.74 0.66
SEWS-Sa 0.81 0.50 0.60 1.17 0.03 0.70 0.64
Maximum control force, J15 Passive 0.012 0.024 0.010 0.022 0.017 0.013 0.016
Semi-active 0.011 0.023 0.008 0.023 0.016 0.010 0.015
Active 0.010 0.024 0.006 0.023 0.015 0.008 0.014
Viscous 0.009 0.022 0.006 0.020 0.014 0.009 0.013
Friction-type 0.024 0.024 0.021 0.024 0.024 0.024 0.024
2-step viscous 0.011 0.024 0.007 0.021 0.016 0.010 0.015
SEWS-PGA 0.002 0.024 0.001 0.014 0.010 0.002 0.009
SEWS-Sa 0.002 0.026 0.003 0.025 0.015 0.003 0.012
Maximum device stroke, J16 Passive 0.38 0.64 0.32 0.60 0.54 0.27 0.46
Semi-active 0.78 0.73 0.52 0.78 0.67 0.51 0.66
Active 0.90 0.77 0.59 0.80 0.71 0.58 0.73
Viscous 0.58 0.69 0.42 0.68 0.67 0.33 0.56
Friction-type 0.16 0.50 0.12 0.47 0.31 0.17 0.29
2-step viscous 0.62 0.76 0.45 0.71 0.75 0.41 0.62
SEWS-PGA 0.88 0.49 0.72 0.63 0.64 0.64 0.67
SEWS-Sa 0.89 0.45 0.48 0.44 0.45 0.57 0.55

5 Conclusions

The present paper describes the possibility of exploiting earthquake information obtained
by a seismic early warning system in the framework of semi-active control strategies by
using magnetorheological dampers. The proposed system differs from traditional semi-active
strategies because it does not require the complex hardware and software needed to evaluate
on-line the structural response and to drive in real-time the MR dampers. It also differs from
active control systems because it does not need MegaWatts of electrical powers to feed actu-
ators. It only needs dozens of Watts, which can be provided by small accumulators whose
charge can be guaranteed, for instance, by photovoltaic panels. The proposed system is based
on the idea that MR dampers can be adjusted once, shortly before the incoming earthquake,
according to a forecasted intensity measure provided by a SEWS. Therefore, the devices,
calibrated just before the incoming earthquake, remain unaltered during the subsequent seis-
mic excitation. The intensity measures considered in the analysis are alternatively the peak
ground acceleration and the spectral acceleration at the fundamental structural period.

123
1714 Bull Earthquake Eng (2011) 9:1703–1715

The effectiveness of the proposed system has been checked through its application to the
case-study of the 91/5 highway bridge located in Orange County, Southern California and
proposed in literature as a benchmark structure for applications in the field of structural con-
trol (Agrawal et al. 2009; Tan and Agrawal 2009). Therefore, the seismic response reductions
associated with the application of the proposed systems, declined in the 2 different versions,
have been directly compared to many different control strategies.
The analyses have demonstrated that the proposed control strategies are capable of reduc-
ing the response of the benchmark highway bridge for a wide variety of earthquake records.
The amount of response reduction is comparable to that obtained with more complex and
expensive semi-active and active systems, whereas it is less scattered compared to any passive
control strategy.
In addition, the proposed control strategy based on the estimate of the anticipated spectral
acceleration yields the overall best response reduction in the case of strong earthquakes, out-
performing even the active control system considered. On the contrary, in some cases it leads
to worse results compared to passive systems. This behaviour corresponds to cases in which
the acceleration response spectra show a significant excitations of higher modes, which are
not considered by the control algorithm. A modified version of the control algorithm is being
currently developed.
In the numerical simulations, the MR dampers have been adjusted based on the actual
values of the intensity measures considered. In other words, their estimates operated by the
SEWS have been assumed to be quite accurate. However, methods for dealing with unavoid-
able uncertainties are currently being developed. Possible errors in estimates have not been
considered in the described analysis and their effect on the proposed control system will be
the next step of the present research.

Acknowledgments This research has been partially funded by the Italian Department of Civil Protection in
the frame of the national project ReLUIS 2005–2008—Theme 9.

References

Agrawal A, Tan P, Nagarajaiah S, Zhang J (2009) Benchmark structural control problem for a seismically
excited highway bridge—part I: phase I problem definition. Struct Control Health Monit 16(5):503–508
Allen RM, Kanamori H (2003) The potential for earthquake early warning in Southern California. Science
300:786–789
Caterino N, Occhiuzzi A, Spizzuoco M (2010) Physical behavior and numerical modeling of magnetorhe-
ological dampers for semi-active control. 14th European conference on earthquake engineering
Convertito V, Iervolino I, Zollo A, Manfredi G (2008) Prediction of response spectra via real-time measure-
ments. Soil Dyn Earthq Eng 28:492–505
De Iuliis M, Petti L, Palazzo B (2008) Semi-active control of structures by using early warning seismic network
information. In: Proceedings 4th European conference of structural control. pp 367–374
Iervolino I, Convertito V, Giorgio M, Manfredi G, Zollo A (2006) Real-time risk analysis for hybrid earthquake
early warning systems. J Earthq Eng 10(6):867–885
Iervolino I, Galasso C, Manfredi G (2008) Information-dependent lead-time maps for earthquake early warning
in the Campania region. In: Proceedings 14 world conference on earthquake Engineering
Kanamori H, Hauksson E, Heaton T (1997) Real-time seismology and earthquake hazard mitigation. Nature
390:461–464
Kanda K, Kobori T, Ikeda Y, Koshida H (1994) The development of a “Pre-arrival transmission system for
earthquake information” applied to seismic response controlled structures. In: Proceedings 1st world
conference on structural control TA-3. pp 23–32
Kuyuk HS, Motosaka M (2008) Spectral forecasting of earthquake ground motion using regional and national
earthquake early warning systems for advanced engineering application against approaching Miyagi-Ken
Oki earthquakes. In: Proceedings 14 world conference on earthquake Engineering

123
Bull Earthquake Eng (2011) 9:1703–1715 1715

Madhekar SN, Jangid RS (2009) Variable dampers for earthquake protection of benchmark highway bridges.
Smart Mater Struct 18:1–18
Nagarajaiah S, Narasimhan S, Agrawal A, Tan P (2009) Benchmark structural control problem for a seismi-
cally excited highway bridge—part III: phase II sample controller for the fully base-isolated case. Struct
Control Health Monit 16(5):549–563
Occhiuzzi A, Spizzuoco M, Serino G (2003) Experimental analysis of magnetorheological dampers for struc-
tural control. Smart Mater Struct 12(5):703–711
Occhiuzzi A, Grasso VF, Manfredi G (2004) Early warning systems from a structural control perspective. 3rd
European conference on structural control
Occhiuzzi A, Iervolino I, Manfredi G (2006) Feedforward control algorithms for seismic early warning sys-
tems. 4th world conference on structural control and monitoring, 4WCSCM-266
Occhiuzzi A, Caterino N, Maddaloni G (2008a) Exploitation of seismic early warning networks for structural
control. 4th European conference on structural control
Occhiuzzi A, Caterino N, Maddaloni G (2008b) Structural control strategies for seismic early warning systems.
14th world conference on earthquake engineering
Pnevmatikos NG, Kallivokas LF, Gantes CJ (2004) Feed-forward control of active variable stiffness systems
for mitigating seismic hazard in structures. Eng Struct 26:471–483
Tan P, Agrawal A (2009) Benchmark structural control problem for a seismically excited highway bridge—
part II: phase I sample control designs. Struct Control Health Monit 16(5):530–548
Wieland M, Griesser L, Kuendig C (2000) Seismic early warning system for a nuclear power plant. In:
Proceedings 12 world conference on earthquake engineering
Zollo A, Iannaccone G, Lancieri M, Cantore L, Convertito V, Emolo A, Festa G, Gallovic F, Vassallo M, Mar-
tino C, Satriano C, Gasparini P (2009) Earthquake early warning system in southern Italy: methodologies
and performance evaluation. Geophys Res Lett 36:L00B07

123

You might also like