Você está na página 1de 19

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Precise stiffness and damping emulation with MR dampers and its application to semi-active
tuned mass dampers of Wolgograd Bridge

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2014 Smart Mater. Struct. 23 015019
(http://iopscience.iop.org/0964-1726/23/1/015019)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 14.139.155.211
This content was downloaded on 23/06/2015 at 12:34

Please note that terms and conditions apply.

Smart Materials and Structures


Smart Mater. Struct. 23 (2014) 015019 (18pp)

doi:10.1088/0964-1726/23/1/015019

Precise stiffness and damping emulation


with MR dampers and its application to
semi-active tuned mass dampers of
Wolgograd Bridge
2

F Weber1,3 and M Maslanka


1

Empa, Swiss Federal Laboratories for Materials Science and Technology, Structural Engineering

Research Laboratory, Uberlandstrasse


129, CH-8600 Dubendorf, Switzerland
2
AGH University of Science and Technology, Faculty of Mechanical Engineering and Robotics,
Department of Process Control, aleja A. Mickiewicza 30, 30-059 Krakow, Poland
E-mail: felix.weber@empa.ch and marcin.maslanka@agh.edu.pl
Received 24 September 2013, in final form 1 November 2013
Published 10 December 2013
Abstract

This paper investigates precise stiffness and damping emulation with MR dampers when
clipping and a residual MR damper force constrain the desired control force. It is shown that
these force constraints lead to smaller equivalent stiffness and greater equivalent damping of
the constrained MR damper force than desired. Compensation methods for precise stiffness
and damping emulations are derived for harmonic excitation of the MR damper. The
numerical validation of both compensation methods confirms their efficacy. The precise
stiffness emulation approach is experimentally validated with the MR damper based
semi-active tuned mass damper (MR-STMD) concept of the Wolgograd Bridge . The
experimental results reveal that the precise stiffness emulation approach enhances the
efficiency of the MR-STMD significantly when the MR-STMD is operated at reduced desired
damping, where the impact of control force constraints becomes significant.
Keywords: control, damping, MR damper, stiffness, semi-active, TMD, Wolgograd Bridge
(Some figures may appear in colour only in the online journal)

Nomenclature

TMD

Tuned mass damper

Abbreviations

Symbols
EF
Fc-des

Fc-des
Fw
F0
V
Xd
X1
X1static
cdes
cdes
cequiv

cycle energy of friction damper


Desired friction force
Corrected desired friction force
Disturbing force amplitude
Residual force
Potential energy
Relative displacement amplitude of mass damper
Displacement amplitude of primary structure
Static deflection of primary structure
Desired viscous damping coefficient
Corrected desired viscous damping coefficient
Equivalent viscous damping coefficient

C1
C2
C3
FRF
FTE
MR
MRSTMD

control of MR-STMD without stiffness/damping


correction
control of MR-STMD with stiffness correction
control of MR-STMD with damping correction
Frequency response function
Force tracking error
Magnetorheological
Semi-active tuned mass damper based on
controlled MR damper

3 Author to whom correspondence should be addressed.

0964-1726/14/015019+18$33.00

c 2014 IOP Publishing Ltd Printed in the UK


Smart Mater. Struct. 23 (2014) 015019

cequiv
c2
fact

fact
fcl , fcu
fdes
clipped
fdes

fdes
fw
f1
iact
ides
kdes

kdes
kequiv

kequiv
k1
k2
ms
m1
m2
uw
xd
x d
x w
x1
1E

1
2

d
w
1

F Weber and M Maslanka

adaptive TMDs are seen to be the appropriate solution to the


problem [7, 8]. The large variety of adaptive TMDs may be
split into active (ATMD) and semi-active devices (STMD)a
thorough overview can be found in [10, 11]. For the mitigation
of civil engineering structures, STMDs represent a very
attractive compromise between passive TMDs and ATMDs
since they allow the adjustment of their properties to the
actual vibration state of the primary structure, to some
extent they are fail safe due to their semi-active nature and
their power requirement is low [12]. In [13], a controllable
STMD is developed based on a controllable viscous damper.
In [1416] the frequency adaptation of the STMD is realized
by actively controlled passive spring systems. Also, different
semi-active materials and devices, respectively, are used for
the development of SMTDs: shape memory alloys are adopted
to make the stiffness of passive TMDs controllable [17],
while in [18] an STMD based on piezoelectric materials
is presented, and controlled friction dampers are used for
the adaptation of STMDs [19]. Another class of STMDs
is based on real-time controlled MR dampers, which are
found to be very effective [20, 21]. Many different control
approaches are used for the control of the MR damper in the
STMD, e.g. fuzzy logic is adopted in [22] and a bangbang
approach is presented in [23]. A slightly different approach
has been developed by the authors of this paper, where the MR
damper is used to emulate the superposition of a controlled
stiffness force and a controlled damping force [24]. The
stiffness force adjusts the natural frequency of the STMD
and the friction force controls the energy dissipation in the
device as a function of the actual frequency of the primary
structure [2527]. This STMD concept, which is subsequently
abbreviated as MR-STMD, is numerically and experimentally
validated for steady state [25, 26] and transient operating
conditions [27] for single harmonic, narrow band excitation
of the primary structure. In fall 2011, the MR-STMD system
was installed by the industrial partner of this R&D project
on the Wolgograd Bridge to prevent this bridge from severe
single harmonic vibrations such as those observed in May
2010. Several videos of this very impressive vibration event
are available on YouTube.
The basic working principle of the MR-STMD is as
follows: if the primary structure vibrates at nominal resonance
frequency, the MR damper is used to emulate amplitude
proportional friction damping that dissipates the same amount
of energy as the viscous damper of a passive TMD. However,
if the primary structure vibrates at frequencies lower/higher
than the nominal resonance frequency because (i) another
mode is vibrating, (ii) environmental effects change the
resonance frequency, (iii) additional loads such as snow or
life loads change the modal mass and thereby the resonance
frequency, or (iv) forced vibrations not at resonance frequency
occur, then the MR damper emulates a negative/positive
stiffness force besides the friction force to adjust the natural
frequency of the MR-STMD to the actual frequency of
the primary structure. The control approach to combine the
controlled stiffness force with a controlled friction force and
not with a controlled viscous force is chosen because the
stiffnessfriction combination does not yield active desired

Equivalent viscous damping coefficient of corrected control force


Viscous damping coefficient of TMD
Actual MR damper force
Corrected actual MR damper force
Frequency lower, upper bounds without control
force constraints
Desired control force
Clipped desired control force
Corrected desired control force
Disturbing frequency
First resonance frequency of primary structure
Actual MR damper current
Desired MR damper current
Desired stiffness
Corrected desired stiffness
Equivalent stiffness
Equivalent stiffness of corrected control force
Stiffness coefficient of primary structure
Passive spring stiffness of mass damper
Shaking mass of shaker
Modal mass of primary structure
Mass of mass damper
Command voltage of shaker amplifier
Relative displacement of mass damper
Relative velocity of mass damper
Acceleration of shaking mass of shaker
Displacement of primary structure
Energy dissipation due to control force constraints
Damping gain
Damping ratio of primary structure
Frequency dependent damping ratio of mass
damper
Mass ratio
Frequency dependent mass ratio
Natural radial frequency of mass damper
Disturbing radial frequency
First resonance radial frequency of primary
structure.

1. Introduction

Passive tuned mass dampers, commonly abbreviated as


TMDs, are well known damping devices in the field of civil
and mechanical engineering. Their working behaviour and
design are well described in Den Hartogs book [1]. Due to
their robustness [2], TMDs are widely used for the mitigation
of the structural vibrations of, for example, tall buildings
and slender bridges [3, 4]. Passive TMDs have also been
developed based on friction dampers [5] and designed for
large tuned masses [6]. In some cases, the efficiency of
passive TMDs might not be sufficient, for example, when
modal parameters of the primary structure change during
different construction phases [7], when life loads significantly
change the modal mass of the primary structure, as can be the
case for footbridges [8], or when environmental parameters
influence the primary structure and TMD properties [9]. Then,
2

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 1. MR-STMD concept (a); MR-STMD prototype (b).

control forces as long as the maximum stiffness force


is not greater than the friction force [28, 29], while the
stiffnessviscous combination always leads to active desired
control forces [3032]. However, with a large frequency shift
between the nominal resonance frequency and the actual
frequency of the primary structure, the stiffnessfriction
approach also includes active forces which are clipped to zero
due the semi-active constraint of the MR damper. In addition,
the residual force of the MR damper at 0 A constrains the
desired stiffnessfriction force. Both control force constraints
end up in imprecise stiffness and damping emulation in the
MR damper and thereby imprecise stiffness and damping
tuning of the MR-STMD.
This paper therefore develops compensation methods
both for precise stiffness and precise damping emulations
when the desired stiffnessfriction force is constrained by
clipping and/or a residual force. Both methods are numerically
tested and the precise stiffness emulation approach is
experimentally validated with the MR-STMD concept on the
15.6 m long Empa bridge. The structure of the paper is
as follows: section 2 describes the working behaviour and
control of the MR-STMD and introduces the problem of
imprecise stiffness and damping emulations when control
force constraints are present. Section 3 develops the stiffness
and damping compensation methods, section 4 shows the
numerical verification of both methods and section 5 describes
the experimental validation. The paper is closed with a
summary and conclusions.

a controlled friction force that dissipates the same amount


of energy per cycle as the energy equivalent viscous
damper of the passive TMD; however, the viscous
coefficient is not constant as for passive TMDs but
adjusted in real-time to the actual frequency of the primary
structure.
The controlled stiffness force is combined with a
controlled friction force because this does not result in
active forces as long as the maximum stiffness force is
not greater than the friction force. This feature reduces the
amount of clipped control forces, whereby a high efficiency
of the MR-STMD is guaranteed. The MR-STMD prototype is
realized with a rotational MR damper (figure 1(b)), where a
cantilever beam is used to transmit the MR damper torque to
the MR-STMD mass. The passive spring is realized by four
compression springs. All single parts of the MR-STMD are
installed within a rack to easily attach the MR-STMD to the
primary structure.
2.2. Control of the MR-STMD

The desired control force fdes to be tracked in real-time by the


MR damper is the superposition of the desired stiffness force
and the desired friction force
fdes = kdes xd + Fcdes sgn(xd )

(1)

where kdes denotes the desired stiffness coefficient, xd is the


relative MR-STMD motion, Fc-des is the desired friction force
amplitude and x d is the relative MR-STMD velocity. kdes
is computed in real-time as a function of the actual radial
frequency w of the primary structure as follows

2. System description and problem formulation


2.1. MR-STMD concept

The MR-STMD is a semi-active tuned mass damper whose


frequency and damping are adjusted in real-time to the actual
frequency of the primary structure using a real-time controlled
MR damper that replaces the passive oil damper of passive
TMDs. As figure 1(a) sketches, the MR damper force consists
of:

kdes = k1

k2
( + 1)2

(2)

where k1 denotes the stiffness of the primary structure, k2 is


the stiffness of the passive spring of the MR-STMD derived
by Den Hartogs design [1] and is the frequency dependent
mass ratio
 2
w
=
(3)
1

a controlled stiffness force to augment or diminish the total


stiffness of the MR-STMD and thereby to adjust the natural
frequency of the MR-STMD to the actual frequency of the
primary structure
3

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 2. Force displacement (a) and force velocity (b) trajectories not constrained by clipping and/or a residual force

(F0 Fc-des |kdes |Xd ).

where = m2 /m1 is the mass ratio, m2 is the MR-STMD


mass and m1 and 1 are the target modal mass and radial
resonance frequency, respectively, of the primary structure.
Note that has its physical meaning only if w = 1 .
The desired friction force is controlled in proportion to the
MR-STMD relative motion amplitude Xd to dissipate the same
cycle energy as the viscous damper of the passive TMD with
desired viscous coefficient cdes , thus

a combined feed forward/feedback force tracking control


scheme. The feed forward is used to estimate the MR damper
current for a given desired control force adopting an inverse
model of the MR damper. For this, the forward MR damper
behaviour is modelled by the Bingham approach, which is
invertible since it does not model the pre-yield behaviour
but describes the total MR damper force as a superposition
of a strongly current dependent Coulomb friction force and
an almost current independent viscous force [34]. The force
tracking error that results from the model-based feed forward
only is further reduced by a feedback correction of the
model-based estimated current. The feedback is based on
the measured actual MR damper force using the 500 N
load cell, see figure 1(b), and only includes a proportional
feedback gain. In addition to the feed forward and feedback,
negative current is applied each half cycle when the desired
control force is far below the MR damper residual force to
reduce remanent magnetization effects that were generated by
the previous magnetization time history. The achieved force
tracking accuracy is shown and discussed in the experimental
part of the paper in sections 5.3 and 5.4. More detailed
information on the real-time force tracking scheme can be
found in [2527].

cdes w Xd
(4)
4
where cdes and the damping ratio 2 are given by the common
formula of passive TMDs but formulated as a function of the
actual frequency w of the primary structure
Fc-des =

cdes = (2 2 m2 w )
s
3
2 =
.
8 ( + 1)3

(5)
(6)

The parameter in (5) represents the damping gain, which


allows a reduction of the desired damping in the MR-STMD
to increase its relative motion amplitude. This approach
augments the efficiency of the MR-STMD because it increases
the force amplitude of the passive spring force which works
with a phase shift of almost 180 at w = 1 against the
disturbing force. Note that this approach may only be adopted
in the environment of real-time frequency adaptation of the
STMD and must not be applied to passive TMDs, which
would then result in larger responses of the structure. w is
determined from the zero crossing of xd , since the MR-STMD
vibrates at the same frequency as the primary structure
(forced vibration). This frequency estimation method works
in real-time for single harmonic vibrations of the primary
structure; in the case of multi-harmonic vibrations, frequency
estimation methods, such as described, for example, in [33],
should be used.

2.4. Resulting force displacement trajectories

In case of zero frequency shift


between the primary structure and the MR-STMD passive
massspring-system, i.e. w = 1 , the control code (1)(6)
ends up in pure friction damping with zero stiffness, since
de-tuning between w of the primary structure and the passive
elements k2 and m2 of the MR-STMD is not present. The
damping generation by controlled friction damping (4) does
not end up in nonlinear motion of m2 because the maximum
passive stiffness force k2 Xd is approximately eleven times
greater than Fc-des whereby k2 Xd dominates the motion of m2 .
2.4.1. Zero frequency shift.

2.3. Real-time force tracking control scheme of MR-STMD

2.4.2. Small frequency shifts.


If the shift between w and
1 is small, fdes requires the superposition of a controlled
friction force with a controlled stiffness force. Figure 2 plots

The real-time tracking of the desired control force with


the MR damper of the MR-STMD prototype is realized by
4

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 3. Force displacement (a) and force velocity (b) trajectories constrained by clipping and/or residual force for F0 < Fc-des and

F0 > Fcdes |kdes |Xd .

this situation for w > 1 , which uses kdes > 0 to augment


the total stiffness of the MR-STMD and thereby to adjust the
MR-STMD natural frequency to w in real-time. fdes can be
fully tracked by the MR damper and thereby the actual MR
damper force fact is equal to fdes because:

outputs dissipative forces and is not constrained by F0 . The


deviations between fact and fdes end up in:
(1) a smaller equivalent stiffness coefficient kequiv than
desired, which is caused by the horizontal parts in the
force displacement trajectory of fact versus xd with zero
equivalent stiffness

fdes is fully dissipative since the maximum desired stiffness


force is not greater than the desired friction force,
i.e. |kdes |Xd Fc-des
|fdes | is always greater than the residual force F0 of the MR
damper, which is simplified here as a pure Coulomb friction
force. The authors are aware of the fact that the residual
force of MR dampers is a combination of a predominant
hysteretic force with a small superimposed viscous force
due to the MR fluid and a friction force due to the
mechanical bearings [35, 36]; however, this simplification
and control-oriented modelling approach, respectively, will
later be used for control purposes.

(2) a higher equivalent damping coefficient cequiv than desired


due to clipping and/or F0 , which is highlighted by the
dashed areas.
Drawback (1) ends up in imprecise frequency tuning
and drawback (2) in imprecise damping tuning of the
MR-STMD. Both drawbacks will lower the efficiency of the
MR-STMD. Stiffness and damping compensation methods
are therefore developed that ensure precise stiffness and
damping emulations with MR dampers even when the desired
control force (1) is constrained by clipping and/or a residual
force. These compensation methods are presented in section 3.

2.4.3. Large frequency shifts.


In contrast to the two cases
discussed above, large frequency shifts end up in |kdes |Xd >
Fc-des which invokes active desired forces in (1). Due to
the semi-active nature of MR dampers, the active forces are
clipped to zero, which yields the clipped desired control force
as follows
(
fdes :
fdes x d > 0
clipped
fdes
=
(7)
0:
otherwise.

3. Compensation methods for precise stiffness and


damping emulations
3.1. Approach

This section derives closed-form solutions for:


(i) the precise stiffness emulation with MR dampers
(ii) the precise damping emulation with MR dampers

However, as figure 3 displays, the residual force F0 further


clipped
constrains fdes , which is seen from |xd0 | < |xdc |. The
resulting actual MR damper force then becomes

clipped

fdes
< F0 , x d > 0
F0 :

clipped
fact = F0 :
(8)
fdes
> F0 , x d < 0

clipped
fdes :
otherwise.

when the desired control force (1) is constrained by clipping


and/or a residual force. The closed-form solutions are valid
for the following assumptions:
Single harmonic motion of the MR-STMD mass, which is
a reasonable assumption considering that large amplitude
vibrations in the primary structure most likely occur
when all the external loading energy, e.g. wind load, is
transferred to one mode of vibration; exactly this situation
was observed in May 2010 on the Wolgograd Bridge.

Note that (8) represents the optimal case where the MR


damper current control does not generate any force tracking
errors in those regions where (1) can be tracked, i.e. where (1)
5

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

The residual force of the MR damper can be simplified


by the Coulomb friction model, although the residual
force is given by bearing friction of the mechanical
MR damper parts and the hysteretic force and viscous
force of the MR fluid at 0 A [35, 36]. However, the
strongly simplified, i.e. control-oriented Coulomb friction
modelling approach allows the derivation of closed-form
solutions for both compensation methods and thereby the
real-time applicability of these methods is guaranteed.

The equivalent viscous damping coefficient is derived from


the energy dissipation of fact during one vibration cycle. The
energy dissipation due to Fc-des is [38, 39]
EF = 4 Fc-des Xd .

(12)

However, fact dissipates more energy due to clipping and/or


the residual force that invoke the additional energy dissipation
1E
1E = |kdes |Xd2 +

The compensation methods are deduced in the following


two steps:

(F0 Fc-des )2
+ 2(F0 Fc-des )Xd
|kdes |
(13)

(1) Derivation of equivalent stiffness and equivalent damping


coefficients of the actual MR damper force (8) when the
desired control force (1) is constrained by clipping and/or
a residual force.
(2) Derivation of the compensation methods for the precise
stiffness and precise damping emulations when the
desired control force (1) is constrained by clipping and/or
a residual force.

which is highlighted by the two dashed areas in figure 3.


The sought equivalent viscous damping coefficient is cequiv =
(EF + 1E)/( d Xd2 ), which yields
cequiv =

|kdes |Xd + 2(F0 + Fc-des ) +

(F0 Fc-des )2
|kdes |Xd

d Xd

(14)

The case
F0 Fc-des and F0 < Fc-des + |kdes |Xd is shown in figure 4,
where the desired control force (1) is strongly constrained by
clipping and/or F0 such that xd0 0 for fact > 0. For kdes > 0

3.2.3. Desired control force constrained, case 2.


3.2. Equivalent stiffness and equivalent viscous damping
3.2.1. Desired control force not constrained.
The desired
control force fdes is not constrained by clipping and/or a
residual force if F0 Fc-des |kdes |Xd (figure 2). The
equivalent stiffness kequiv and equivalent viscous damping
coefficient cequiv generated by fact are equal to their desired
counterparts kdes and cdes , respectively.

1
(F0 Fc-des )2
V1 + V2 = Fc-des Xd + Xd2 kdes +
F0 Xd
2
2 kdes
(15)
which yields

The case
F0 < Fc-des and F0 > Fcdes |kdes |Xd is displayed in figure 3,
where the control force constraints, i.e. clipping and/or F0 ,
result in xd0 < 0 for fact > 0. The equivalent stiffness of the
constrained force displacement trajectory of fact is derived
from the potential energy that is stored and released by fact
during half a cycle [37]. The potential energy V1 that is stored
in the damper at maximum displacement is equal to the work
done from point (A) to point (B). For positive desired stiffness
kdes > 0, V1 becomes
Z Xd
V1 =
fact dxd = Fc-des Xd + 0.5 Xd2 kdes
(9)
3.2.2. Desired control force constrained, case 1.

kequiv =

and cequiv is given by (14).


The case
of F0 Fc-des + |kdes |Xd results in zero equivalent stiffness
kequiv = 0 and the equivalent damping coefficient cequiv is only
given by F0 , i.e. cequiv = 4 F0 /( d Xd ).
3.2.4. Desired control force constrained, case 3.

3.2.5. General expressions for equivalent stiffness and


equivalent damping.
Comparing the expressions (11) and

and the potential energy release V2 from point (B) to point (D)
is
Z 0
(Fc-des F0 )2
fact dxd = F0 Xd
V2 =
.
(10)
2 kdes
Xd
The sought equivalent stiffness is kequiv = (V1 +
which yields
kequiv =

1
Fc-des F0
(Fc-des F0 )2
kdes + sgn(kdes )
+
2
Xd
2 kdes Xd2
(16)

(16), the only difference is the sign in front of the last


term. The equivalent stiffness and the equivalent damping
coefficients of fact (8) in their general forms therefore become

V2 )/Xd2 ,

1
Fc-des F0
(Fc-des F0 )2
kdes + sgn(kdes )

2
Xd
2 kdes Xd2
(11)

kequiv

where the signum function is introduced such that (11) is


valid for positive and negative desired stiffness coefficients.
6

kdes :
F0 Fc-des |kdes |Xd

Fc-des F0
1

kdes + sgn(kdes )

2
Xd

(Fcdes F0 )2 :
= sgn(Fc-des F0 )
2 kdes Xd2

Fc-des |kdes | Xd < F0

< Fc-des + |kdes |Xd

0:
F0 Fc-des + |kdes |Xd

(17)

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 4. Force displacement (a) and force velocity (b) trajectories constrained by clipping and/or a residual force for F0 Fc-des and
F0 < Fcdes + |kdes |Xd .

Figure 5. Example of force displacement trajectories before (a) and after (b) stiffness correction.

cequiv

trajectory parts. The basic idea therefore is to replace the


desired stiffness kdes in (1) by the corrected desired stiffness
which must be greater than k
kdes
des (for positive kdes ) to
compensate for the horizontal trajectory parts with zero

equivalent stiffness such that the equivalent stiffness kequiv


of the entire force displacement trajectory of the corrected
is equal to the originally desired
actual MR damper force fact
stiffness kdes . This procedure is illustrated in figure 5. The
corrected desired control force becomes

c :
F0 Fc-des |kdes |Xd
des

c-des )

|kdes |Xd + 2(F0 + Fc-des ) + (F0|kF

des |Xd

d Xd
=
(18)
Fc-des |kdes | Xd < F0

< Fc-des + |kdes |Xd

4F0
:
F0 Fc-des + |kdes | Xd .
d Xd

The expressions (17) and (18) demonstrate that the control


force constraints of clipping and/or F0 result in |kequiv | <
|kdes | and cequiv > cdes , which ends up in imprecise
frequency and damping tunings of the MR-STMD. The
following subsection therefore derives stiffness and damping
compensation methods which are formulated as closed-form
solutions to ensure precise stiffness and damping emulations
in MR dampers in real-time.

fdes
= kdes
xd + Fc-des sgn(xd )

(19)

that is constrained by clipping and/or F . The


which yields fact
0
is
equivalent stiffness of fact

kequiv
=

1
Fc-des F0

k + sgn(kdes
)
2 des
Xd
sgn(Fc-des F0 )

3.3. Compensation method for precise stiffness emulation

(Fc-des F0 )2
X2
2 kdes
d

(20)

which must be equal to the originally desired stiffness kdes ,

i.e. kequiv
= kdes . Solving (20) with kequiv
= kdes yields a
quadratic equation whose first root is not feasible. The second

As seen from figures 3 and 4, the control force constraints


of clipping and/or the residual force result in reduced
equivalent stiffness, i.e. |kequiv | < |kdes |, due to the horizontal
7

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

3.4. Compensation method for precise damping emulation

The control force constraints of clipping and/or a residual


force yield higher energy dissipation in the MR damper than
desired, i.e. cequiv > cdes , which can be seen from the dashed
areas in figures 3 and 4. To emulate the correct energy
dissipation in the MR damper, cdes in (4) must be replaced
by the corrected value cdes , which must be smaller than cdes
to compensate for the too high energy dissipation of fact
due to clipping and/or F0 such that the equivalent damping
coefficient cequiv of the force displacement trajectory of the
is equal to the originally
corrected actual MR damper force fact
desired damping coefficient cdes . Thus, the corrected desired
control force is computed with the corrected friction force

amplitude Fc-des

Figure 6. Corrected desired friction force versus desired friction

force.

root yields the sought corrected desired stiffness

b + b2 4 a c

(kdes ) root2 = sgn(kdes )


2a

fdes
= kdes xd + Fc-des
sgn(xd )

(23)

which yields the equivalent viscous damping coefficient of fact


as follows

(21)

Xd2 , b

with a =
= 2 Xd (Fcdes F0 Xd |kdes |) and c =
sgn(Fc-des F0 )(Fc-des F0 )2 . Since (21) is only needed
is constrained by clipping and/or F , i.e. if F
when fact
0
cdes <
|kdes |Xd + F0 , the general formulation of the corrected desired
stiffness becomes

kdes :
|kdes | (Fc-des F0 )/Xd

sgn(kdes )
(Fc-des

Xd

F0 Xd |kdes |) +

(22)
kdes
=
2

{(F

c-des F0 Xd |kdes |)

2 1/2

+
sgn(F

F
)(F

F
)
}
:
c-des
0
c-des
0

|kdes | > (Fc-des F0 )/Xd .

cequiv

)+
|kdes |Xd + 2(F0 + Fc-des

d Xd

(F0 Fcdes
)2
|kdes |Xd

(24)

which must be equal to the originally desired viscous


damping coefficient cdes , i.e. cequiv = cdes . Solving (24) for

the corrected friction force amplitude Fc-des


with cequiv = cdes
leads to a quadratic equation with the sought solution as
follows (first root not feasible)

b + b2 4 a c

(25)
(Fc-des ) root2 =
2a
with a = 1, b = 2(|kdes |Xd F0 ), c = |kdes |2 Xd2
|kdes | d Xd2 cdes + 2|kdes |Xd F0 + F02 and for F0 Fc-des <
F0 + |kdes |Xd . The solution of (25) is given by (26) and
graphically shown in figure 6:

(i) If F0 Fc-des < F0 +|kdes |Xd then Fc-des


is given by (26).
(ii) If Fc-des F0 + |kdes |Xd , the desired force is not

constrained and therefore Fc-des


= Fc-des .
(iii) If Fc-des < F0 the precise energy dissipation control is
not possible due to the dominant residual force. Then,

Fc-des
should be equal to the smallest value possible, thus
F0 |kdes | Xd given by b2 4 a c = 0 in (25). This solution
yields a negative desired friction force such that the error
between cdes and cequiv is minimized for given F0 , given
desired stiffness kdes and given Xd .

Fc-des :
Fc-des F0 + |kdes | Xd

F |k | X

0 q des d

Fc-des =
+ |kdes | d Xd2 cdes 4 |kdes | Xd F0 :

F0 Fc-des < F0 + |kdes | Xd

F0 |kdes | Xd :
Fc-des < F0 .

Figure 5 depicts the force displacement trajectories


before and after stiffness correction for the case of F0 =
10 N and d = 19.85 rad s1 . It is seen that the stiffness
correction method (22) increases the slope of the force
| > |k |, and thereby
displacement trajectory, due to |kdes
des
compensates for the horizontal trajectory parts with zero

equivalent stiffness such that the equivalent stiffness kequiv


of
the entire corrected force displacement trajectory is equal to
the originally desired stiffness kdes . In contrast, the equivalent
stiffness due to (1) with kdes , i.e. without stiffness correction,
results in an error of (kdes kequiv )/kdes = 22.3%! Whereas
the stiffness correction generates precise stiffness emulation
as targeted, the amplification of the corrected stiffness results
in augmented cequiv , which is observed from the larger area
and
surrounded by the force displacement trajectory of fact

from the displayed numbers of cequiv and cequiv , respectively.


This drawback cannot be avoided because the stiffness force
cannot be controlled independently of the friction force in MR
dampers when control force constraints such as clipping and
F0 are present.

(26)
Figure 7 displays the change of the force displacement
trajectory due to the friction force correction for F0 =
8

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 7. Example of force displacement trajectories before (a) and after (b) damping correction.

10 N and d = 19.85 rad s1 . It is seen that the damping


correction method yields the correct equivalent viscous
damping coefficient by commanding a smaller corrected
viscous damping coefficient than originally desired, i.e. cdes <

cdes , but thereby generates |kequiv


| < |kdes |, which cannot be
avoided since friction and stiffness forces cannot be controlled
independently in MR dampers when control force constraints
such as clipping and/or F0 are present.

4.2. System parameters

The modal mass, resonance frequency, stiffness and damping


ratio of the primary structure are: m1 = 1681 kg, f1 =
3.145 Hz, k1 = 656.48 kN m1 , 1 = 0.4%. The mass and
spring stiffness of the passive elements of the MR-STMD are:
m2 = 26.325 kg, k2 = 9.966 kN m1 . These properties are
equal to those identified from the Empa bridge, which is used
in section 5 for the experimental validation of the MR-STMD
controlled by C1 and C2. The values of m2 and k2 are equal to
those of the prototype MR-STMD, as depicted in figure 1(b).
m2 = 26.325 kg corresponds to a mass ratio of approximately
1.57%, which is a typical value for footbridges [40].

4. Numerical validation
4.1. Control laws under consideration

Both the precise stiffness and the precise damping emulation


approaches are numerically validated with the MR-STMD
concept. The following three control laws of the MR-STMD
control approach are investigated:

4.3. Dynamic simulation

(1) Control law #1 (C1): the desired control force for the MR
damper of the MR-STMD is given by (1)(7) without
stiffness and damping corrections.
(2) Control law #2 (C2): the desired control force for the
MR damper of the MR-STMD is given by (1)(7) with
stiffness correction (22).
(3) Control law #3 (C3): the desired control force for the
MR damper of the MR-STMD is given by (1)(7) with
damping correction (26).

The primary structure with MR-STMD is computed in


R
R
MATLAB
/Simulink
using the solver ode5 (DormandPrince) at a fixed step size of 1 ms. The interaction between
the primary structure and the MR-STMD is given by the
passive spring force and the actual MR damper force (8),
which is computed by C1, C2 and C3, respectively. The
steady state responses of the primary structure and the
MR-STMD, respectively, are obtained from harmonic, single
frequency, narrow band excitation of the primary structure.
The non-dimensional responses of the primary structure with
MR-STMD controlled by C1, C2 and C3, respectively, which
are subsequently denoted as frequency response functions
(FRFs), are determined as follows

All three control laws are computed for nominal desired


damping in the MR-STMD, i.e. damping gain = 1 in
(5), and for reduced desired damping in the MR-STMD,
i.e. damping gain < 1, which increases the relative motion
amplitude of the MR-STMD and thereby the passive spring
force amplitude which eventually augments the efficiency of
the mass damper. This method to enhance the efficiency of
a mass damper can be applied only if the frequency of the
mass damper is adjusted in real-time to the actual frequency of
the primary structureas is realized by the MR-STMDand
only if the maximum tolerable relative motion amplitude in
the mass damper is not exceeded.

X1
X1static

X1 k1
Fw

(27)

where X1 denotes the displacement amplitude of the primary


structure at the anti-node position and Fw is the disturbing
force amplitude. The evaluation criterion (27) corresponds to
the assessment method of the passive TMD connected to the
vibrating primary structure according to Den Hartog [1].
9

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 8. Simulated FRFs of the nominal bridge with MR-STMD with F0 = 2 N and passive TMD for = 1 (a) and = 0.7 (b).

Figure 9. Simulated desired and resulting equivalent stiffness (a) and damping (b) coefficients for C1, C2 and C3 with = 0.7 of

MR-STMD with F0 = 2 N and for the nominal bridge.

at large frequency shifts when the disturbing frequency fw


is below fcl and above fcu , which triggers large negative
and positive desired stiffness coefficients in (2). Figures 8(a)
and (b) demonstrate that both for = 1 and = 0.7 the
precise damping approach is not preferable for controlled
mass dampers since the maxima of the FRFs due to C3 are
higher than or equal to those due to C1 and C2, respectively. In
contrast to the precise damping approach, the precise stiffness
emulation method improves C1 significantly, especially at
reduced damping with = 0.7, since then the control force
constraints more strongly constrain the desired control force
(1) than at nominal damping with = 1. C2 leads to precise
frequency tuning of the MR-STMD at any , whereas C1
generates larger frequency tuning errors at = 0.7 than at
= 1. The improvement due to C2 compared to the passive
TMD with nominal damping, i.e. the TMD with classical
design [1], is up to 35%.
Figures 9(a) and (b) depict the equivalent stiffness
and damping coefficients as functions of the disturbing
frequency fw . It is seen that all three control laws yield
the same equivalent values if fw is within the frequency
band [fcl , fcu ], where control force constraints do not exist.

4.4. Simulation results for F0 = 2 N

The FRFs resulting from the MR-STMD that is controlled


by C1, C2 and C3, respectively, at nominal damping with
= 1 and reduced damping with = 0.7 and for the residual
force F0 = 2 N are displayed in figures 8(a) and (b). The
residual force is assumed as 2 N, because this is the value of
the MR damper of the prototype MR-STMD when remanent
magnetization effects do not exist which invoke values of up to
5 N [2527]. As benchmark, the FRFs due to the passive TMD
according to Den Hartogs design [1] are also plotted. Notice
that the passive TMD is computed without any residual force,
i.e. under ideal conditions. The figures are entitled as nominal
bridge because the primary structure is assumed to be the
Empa bridge with nominal modal parameters m1 = 1681 kg
and f1 = 3.145 Hz. In addition to the FRFs, the disturbing
frequencies fcl and fcu are also plotted below and above which
the desired control force fdes (1) due to C1 is constrained by
clipping and/or a residual force.
As expected, the compensation methods (22) and (26),
respectively, show an impact on the resulting FRFs only when
fdes (1) is constrained by clipping and/or F0 . This is observed
10

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 10. Simulated force displacement trajectories at fw = fA = 2.86 Hz ((a)(c)) and fw = fB = 3.5 Hz ((d)(f)) of the MR-STMD with
F0 = 2 N, = 0.7 and for the nominal bridge.

the MR-STMD gets lost, see displayed numbers of kequiv


in
figures 10(c), (f). Therefore, C3 is investigated no further here.

When the desired control force (1) is constrained by clipping


and/or F0 , the precise stiffness approach yields the correct
equivalent stiffness within the entire disturbing frequency
range. This cannot be achieved by the precise damping
emulation approach within the entire disturbing frequency
range because large frequency shifts trigger case (iii) in (26).
The force displacement trajectories of all three control
laws are plotted in figure 10 for the two selected disturbing
frequencies fw = fA = 2.86 Hz and fw = fB = 3.50 Hz. At
fw = fA , all control laws yield negative desired stiffness, due
to fw < f1 , in order to adjust the natural frequency of the
MR-STMD to fA . At fw = fB , all control laws yield positive
desired stiffness, due to fw > f1 , in order to augment the
natural frequency of the MR-STMD to fB . As figures 10(b),
(e) show, the precise stiffness approach magnifies the desired
stiffness such that the equivalent stiffness of the corrected
is equal to the originally desired stiffness and
actual force fact
thereby generates precise frequency tuning of the MR-STMD.
From figures 10(c), (f), it can be observed that C3 generates
the correct equivalent damping coefficient of the corrected
, although the areas that are surrounded by
actual force fact
the force displacement trajectories are larger than those in
figures 10(a), (b) and (d), (e), respectively. The reason for
this is that the term d Xd stands in the denominator of
expression (24) with which cequiv is determined.
The discussion of the simulation results shows that the
precise damping approach (26) is not preferable for the
MR-STMD because it ends up in very imprecise stiffness
emulation whereby the precise real-time frequency tuning of

4.5. Impact of residual force on the performances of C1 and


C2

The impact of the magnitude of F0 on the FRFs due to C1 and


C2 at reduced desired damping in the MR-STMD is displayed
in figures 11(a)(c) for = 0.7 and in figures 11(d)(f) for
= 0.5. This investigation is not made for = 1 because
then C1 and C2 almost yield the same responses in the primary
structure, as seen from figure 8(a). Figure 11 demonstrates that
C2 outperforms C1 in almost all cases, except for disturbing
frequencies between approximately 2.95 Hz and fcl , where
the increased damping due to the stiffness correction of C2
reduces the motion amplitude in the MR-STMD such that the
response of the primary structure becomes slightly larger than
for C1 despite the slightly imprecise frequency tuning of the
MR-STMD by C1. However, the FRFs due to C2 between
2.95 Hz and fcl are only a little higher than those due to C1.
Thus, it can be concluded that the precise stiffness emulation
approach increases the efficiency of the MR-STMD within the
entire narrow band disturbing frequency range compared to
C1 for F0 = 0, 2 and 4 N. These values correspond to 0%,
approximately 5% and approximately 10% of the maximum
MR damper force of approximately 40 N for the system under
consideration. Thus, the impact analysis of F0 on the resulting
FRFs is made with reasonable values of F0 .
11

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 11. Simulated FRFs of the nominal bridge with MR-STMD with different F0 and for = 0.7 ((a)(c)) and = 0.5 ((d)(f)).

5. Experimental validation
5.1. Empa bridge with prototype MR-STMD

The Empa bridge is a stay cable bridge with a main span


of 15.6 m (figure 12). The prototype MR-STMD that is
depicted in figure 1(b) is attached to the bridge deck at
the anti-node position. The modal parameters m1 = 1681 kg
and f1 = 3.145 Hz of the Empa bridge with nominal
masses on the bridge deck are identified from mode shape
measurements [25, 26]. The damping ratio 1 = 0.4% is
obtained from free vibration decay tests [25, 26]. The
mass m2 = 26.325 kg of the MR-STMD corresponds to a
mass ratio of 1.57%, which represents a typical value for
footbridges [40]. The four passive springs have a total stiffness
of k2 = 9.966 kN m1 such that the natural frequency of
the passive springmass-system of 3.10 Hz corresponds to
that according to Den Hartogs design [1] for the mass ratio
of 1.57%. The rotational MR damper produces a maximum
torque of 45 N m at a maximum tolerated constant current of
2 A.

Figure 12. Empa bridge with prototype MR-STMD.

R
uw of the TIRA
amplifier. The FRFs in the experiments are
derived according to (27) from x1 , x w and ms as follows

X1
5.2. Measurement and control hardware

X1static

rms(x1 ) k1
rms(xw ) ms

(28)

where rms denotes the root mean square of the corresponding


measured quantity over 5 s during steady state conditions.
The MR-STMD relative motion xd is measured by a 50 mm
linear variable displacement transducer and the actual MR
damper force fact is measured with a 500 N load cell, see
figure 1(b). The control algorithms C1 and C2, respectively,
R
R
are implemented in the MATLAB
/dSPACE
real-time

Figure 13 shows a schematic of the entire test set-up. A


laser triangulation sensor is used to measure the anti-node
displacement x1 of the bridge, the disturbing force is measured
by the measured acceleration x w of the shaking mass ms =
R
32.5 kg of the electrodynamic shaker of type TIRA
. The
R

signal generator of type Tektronix is used to define the
frequency and amplitude of the sinusoidal command voltage
12

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

demonstrates that C2 outperforms the C1 significantly,


which agrees well with the numerical finding presented in
section 4.4. C2 with < 0.7 is not tested because the
simulation results at = 0.5, see figures 11(d)(f), indicate
that the maxima of the FRFs cannot be further reduced relative
to the maxima of the FRFs due to C2 with = 0.7. C1 and C2
at = 0.7 reduce the steady state vibration amplitudes in the
primary structure by 19%41% compared to those resulting
from the passive TMD at 1.0c2 , which represents the classical
TMD design.
The real-time force tracking
in the MR damper, which is realized by the combined
model-based feed forward/feedback control scheme, as
described in section 2.3, is depicted in figure 15 for fw = fA =
2.92 Hz and in figure 16 for fw = fB = 3.28 Hz. Two main
sources of force tracking errors (FTE) are visible. The FTE(a)
occurs when the actual MR damper force should be rapidly
increased from zero to a large value at the displacement
extreme, which is the case during the emulation of negative
stiffness (figure 15). When the desired control force fdes
steps up, the inverse MR damper model of the feed forward
estimates a sudden increase of the desired current. The desired
current step is amplified by the feedback due to the FTE
|fdes fact | > 0. However, the maximum current constraint
of 2 A and the limited power of the current driver make it
impossible to achieve zero FTE during this force step. Also,
the pre-yield stiffness of the MR damper under consideration
limits how fast the actual force can be increased at the
displacement extreme. This is indicated by the approximate
slope of the pre-yield stiffness in figure 15(a). The pre-yield
stiffness thereby explains that fact cannot track the vertical
step of fdes when MR dampers are operated in the pre-yield
regime and when the maximum current is limited. Further
information on the force tracking accuracy limitation due to
the pre-yield stiffness can be found in [41].
The second main source of FTE is of type FTE(b), when
the desired control force decreases from a large value to
zero, i.e. it comes close to or below the residual force. In
5.3.2. Force tracking accuracy.

Figure 13. Schematic of test set-up.

R
control environment. The amplifier of type KEPCO
is
used as current driver and thereby ensures that the actual MR
damper current iact tracks precisely the desired current ides .

5.3. Test results for nominal bridge

The
experimental validation of C1 and C2 at = 1 and 0.7
is plotted in figure 14; C3 is not validated since it is not
appropriate for the MR-STMD, as explained in section 4.4.
At nominal desired damping ( = 1), C1 and C2 almost
perform equally well, as observed in the simulations. Both
approaches perform better than the passive TMD due to the
adaptation of frequency and damping in the MR-STMD to the
actual disturbing frequency. For the experiments, the TMD
according to Den Hartogs design [1] is generated by the
emulation of constant viscous damping in the MR-STMD
and is therefore very precisely realized, as can be seen from
the two equal peaks of the FRF and from the measured
force displacement trajectories displayed in figures 19(c),
(f). At reduced damping in the MR-STMD, figure 14(b)
5.3.1. Bridge response due to harmonic excitation.

Figure 14. Measured FRFs of the nominal bridge with the MR-STMD and passive TMD at = 1 (a) and = 0.7 (b).

13

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 15. Measured real-time force tracking in the MR damper at 2.92 Hz due to C1 ((a)(c)) and C2 ((d)(f)) at = 0.7 for the nominal

bridge.

Figure 16. Measured real-time force tracking in the MR damper at 3.28 Hz due to C1 ((a)(c)) and C2 ((d)(f)) at = 0.7 for the nominal

bridge.

14

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 17. Measured FRFs of lighter bridge with MR-STMD and passive TMD at = 1 (a) and = 0.7 (b).

of the previous amplitude, and fw which includes a time delay


of approximately 5 s due to the low pass filtering to make C2
and C3, respectively, insensitive to measurement noise [27].
Consequently, the emulation of the desired stiffness and
damping values during transients include small errors, but
the errors vanish rapidly when steady state vibrations are
reached. Large amplitude vibrations of real civil structures
are often predominantly single harmonic vibrations with
slowly changing frequencies and amplitudes; as observed, for
example, on the Franjo Tudjman Bridge, where stay cables
oscillated at their fundamental frequencies with mid-span
amplitudes of up to 1 m [42], or the Wolgograd Bridge,
where the bridge deck vibrated at 0.45 Hz with amplitudes
of up to 40 cm [43]. Thus, C2 improves the frequency tuning
of the MR-STMD for most vibration scenarios of real civil
structures, where both frequency and amplitude of the primary
structure will change slowly over time.

this case, the FTE(b) is caused by remanent magnetization


effects due to the previous magnetization time history when
fact was increased from small values to its maximum and
then was lowered again. The remanent magnetization, which
is caused by the magnetic hysteresis, then invokes larger
MR damper forces for a certain current level than without
remanent magnetization. The approach chosen here to reduce
this nonlinear and undesirable effect is the application of a
negative current during a fairly short time interval and with
limited magnitude, with the goal to cancel the remanent magnetization as far as possible. The negative current application
must be limited in time and magnitude only to cancel the
remanent magnetization but not to magnetize the MR fluid and
damper housing in the opposite direction, which would result
in the same dissipative MR damper force as with a positive
current. The application of negative current is also modulated
by the proportional feedback, as seen in figures 15(c), (f)
and 16(c), (f). In the case of negative stiffness with friction,
a negative current can be applied during a fairly large time
interval, which then results in very low FTE(b), see figure 15,
since fdes is slowly reduced. In the case of positive stiffness
with friction (figure 16), the desired force steps down to zero
at the displacement extreme. Due to the large magnetization
at maximum force, a force overshoot in fact results when fdes
changes its sign. This FTE is also reduced as far as possible
by the application of a negative current. However, the force
overshoot is not fully removed because the available time
interval of negative current application between the previous
force maximum and step down is simply very short.
Despite the existence of FTEs, the displayed numbers of
the equivalent stiffness show that C2 generates more precise
stiffness emulation than C1, which explains the improved
performance of C2 compared to C1, as seen from figure 14(b).

5.4. Test results for lighter and heavier bridges

The control laws C1 and C2 at nominal ( = 1) and reduced


( = 0.7) damping in the MR-STMD are also tested for a
lighter bridge and a heavier bridge, which is achieved by
changing the additional local masses on the bridge deck. The
resonance frequency of the lighter bridge is 3.487 Hz, which
corresponds to a frequency shift of +10.87% relative to the
nominal resonance frequency of f1 = 3.145 Hz for which the
passive massspring-system of the MR-STMD is designed.
The resonance frequency of the heavier bridge is 2.760 Hz,
which corresponds to a frequency shift of 12.24% relative
to f1 .
The FRFs of the lighter and heavier bridges are depicted
in figures 17 and 18; the real-time force tracking at the
selected disturbing frequencies 3.56 and 2.70 Hz is plotted
in figure 19. As for the nominal bridge, C2 outperforms C1 by
more when the MR-STMD is operated at reduced damping
with = 0.7. The measured improvements of C1 and C2
relative to the passive TMD at 1.0c2 are 42% and 52% for
the lighter bridge and 40% and 50% for the heavier bridge.
The improvements between C1 and C2 with = 0.7 are
larger than those obtained for the nominal bridge because

When the disturbing frequency and amplitude are changed to measure the
steady state response of the bridge at the distinct frequencies
of the plotted FRFs (figure 14), both the precise stiffness and
damping approaches run stable during these transients despite
the fact that both methods rely on the real-time detection of
Xd , which is delayed by half a period because Xd is the value

5.3.3. Behaviour during transient excitation.

15

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

Figure 18. Measured FRFs of heavier bridge with MR-STMD and passive TMD at = 1 (a) and = 0.7 (b).

Figure 19. Measured real-time force tracking in the MR damper at 3.56 and 2.70 Hz due to C1 ((a), (d)), C2 ((b), (e)) and emulation of

viscous damping for tests with passive TMD ((c), (f)) for lighter and heavier bridges.

precise frequency tuning becomes especially important when


de-tuning between the actual resonance frequency of the
bridge and the nominal resonance frequency is present, as
for the lighter and heavier bridges. The measured real-time
force tracking shows the same two main sources of FTEs as
observed in figures 15 and 16 for the nominal bridge.

adjust in real-time both the natural frequency and damping


of the MR-STMD to the actual bridge frequency by the
emulation of a combined stiffnessfriction control force.
Control force constraints such as clipping and a residual
MR damper force lead to a lower equivalent stiffness and
higher equivalent damping of the actual MR damper force
than desired, and thereby to imprecise frequency and damping
tunings of the MR-STMD.
In order to overcome this drawback, two independent
compensation methods that yield precise stiffness and
damping emulations with MR dampers in the presence of

6. Summary and conclusions

The MR damper in the semi-active tuned mass damper


concept of the Wolgograd Bridge (MR-STMD) is used to
16

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

control force constraints are presented. Both methods are


formulated as closed-form solutions to ensure their real-time
applicability, and are valid for single harmonic excitation
of the MR damper. The numerical verification of both
methods by the MR-STMD concept demonstrates that the
precise stiffness approach enhances the efficiency of the
MR-STMD due to the precise frequency tuning of the
MR-STMD, especially when the MR-STMD is operated at
reduced damping, where control force constraints become
more relevant. The precise damping emulation approach turns
out to be inappropriate for the control of the MR-STMD due
to the resulting imprecise frequency tuning of the MR-STMD.
The experimental validation of the precise stiffness emulation
approach with the prototype MR-STMD on the 15.6 m long
Empa bridge confirms the numerical findings. It is therefore
concluded that the precise stiffness emulation approach is a
powerful tool to make the MR-STMD more efficient.

[12] Spencer B F and Nagarajaiah S 2003 State of the art of


structural control J. Struct. Eng. 129 84556
[13] Pinkaew T and Fujino Y 2001 Effectiveness of semi-active
tuned mass dampers under harmonic excitation Eng. Struct.
23 8506
[14] Nagarajaiah S and Sonmez E 2007 Structures with semiactive
variable stiffness single/multiple tuned mass dampers
J. Struct. Eng. 133 6777
[15] Eason R P, Sun C, Dick A J and Nagarajaiah S 2013
Attenuation of a linear oscillator using a nonlinear and a
semi-active tuned mass damper in series J. Sound Vib.
332 15466
[16] Xu Z B, Gong X L and Chen X M 2011 Development of a
mechanical semi-active vibration absorber Adv. Vib. Eng.
10 22938
[17] Mishra S K, Gur S and Chakraborty S 2013 An improved
tuned mass damper (SMA-TMD) assisted by a shape
memory alloy spring Smart Mater. Struct. 22 095016
[18] Lallart M, Yan L, Wu Y-C and Guyomar D 2013
Electromechanical semi-passive nonlinear tuned mass
damper for efficient vibration damping J. Sound Vib.
332 5696709
[19] Lin G L, Lin C C, Lu L Y and Ho Y B 2012 Experimental
verification of seismic vibration control using a semi-active
friction tuned mass damper Earthq. Eng. Struct. Dyn.
41 81330
[20] Kang J and Kim H S 2010 Fuzzy hybrid control of a
wind-excited tall building Struct. Eng. Mech. 36 38199
[21] Kang J, Kim H S and Lee D G 2011 Mitigation of wind
response of a tall building using semi-active tuned mass
dampers Struct. Des. Tall Special Build. 20 55265
[22] Kim H S and Kang J W 2012 Semi-active fuzzy control of a
wind-excited tall building using multi-objective genetic
algorithm Eng. Struct. 41 24257
[23] Yan A-Z, Kong Q and Lu J-J 2012 Research on control
effectiveness of semi-active MR-TMD to some coal gas
manufacturing plant Appl. Mech. Mater. 117119 38
[24] Weber F, Boston C and Maslanka M 2011 Adaptive TMD
based on the emulation of positive and negative stiffness
with MR damper Smart Mater. Struct. 20 015012
[25] Weber F and Maslanka M 2012 Frequency and damping
adaptation of a TMD with controlled MR damper Smart
Mater. Struct. 21 055011
[26] Weber F and Distl H 2013 Real-time controlled tuned mass
dampers for Wolgograd Bridge Beton-und Stahlbetonbau
108 36272 doi:10.1002/best.201300013
[27] Weber F 2013 Dynamic characteristics of controlled
MR-STMDs of Wolgograd Bridge Smart Mater. Struct.
22 095008
[28] Boston C, Weber F and Guzzella L 2009 Optimal semi-active
damping of cables: evolutionary algorithms and
closed-form solutions Smart Mater. Struct. 18 055006
[29] Boston C, Weber F and Guzzella L 2011 Optimal semi-active
damping of cables with bending stiffness Smart Mater.
Struct. 20 055005
[30] Li H, Liu J L and Ou J P 2011 Seismic response control of a
cable-stayed bridge using negative stiffness dampers Struct.
Control Health Monit. 18 26588
[31] Weber F and Boston C 2011 Clipped viscous damping with
negative stiffness for semi-active cable damping Smart
Mater. Struct. 20 045007
[32] Wu B, Shi P F and Ou J P 2013 Seismic performance of
structures incorporating magnetorheological dampers with
pseudo-negative stiffness Struct. Control Health Monit.
20 40521
[33] Nagarajaiah S and Varadarajan N 2005 Short time Fourier
transform algorithm for wind response control of buildings
with variable stiffness TMD Eng. Struct. 27 43141

Acknowledgments

The authors gratefully acknowledge the financial support of


Empa, Swiss Federal Laboratories for Materials Science and
Technology, Dubendorf, Switzerland, the financial support of
AGH University of Science and Technology, Department of
Process Control, Krakow, Poland (statutory research funds
No. 11.11.130.560), and the technical support of the industrial
partner Maurer Sohne GmbH & Co. KG, Munich, Germany.
References
[1] Den Hartog J P 1934 Mechanical Vibrations (York, PA:
McGraw-Hill Book Company, The Maple Press Company)
[2] Yu H, Gillot F and Ichchou M 2013 Reliability based robust
design optimization for tuned mass damper in passive
vibration control of deterministic/uncertain structures
J. Sound Vib. 332 222238
[3] Matta E 2011 Performance of tuned mass dampers against
near-field earthquakes Struct. Eng. Mech. 39 62142
[4] Casado C M, Diaz I M, de Sebastian J, Poncela A V and
Lorenzana A 2013 Implementation of passive and active
vibration control on an in-service footbridge Struct. Control
Health Monit. 20 7087
[5] Qin L, Yan W and Li Y 2009 Design of frictional pendulum
TMD and its wind control effectiveness Earthq. Eng. Eng.
Vib. 29 1537
[6] Moutinho C 2012 An alternative methodology for designing
tuned mass dampers to reduce seismic vibrations in
building structures Earthq. Eng. Struct. Dyn. 41 205973
[7] Casado C M, Poncela A V and Lorenzana A 2007 Adaptive
tuned mass damper for the construction of concrete piers
Struct. Eng. Int. 17 2525
[8] Occhiuzzi A, Spizzuoc M and Ricciardelli F 2008 Loading
models and response control of footbridges excited by
running pedestrians Struct. Control Health Monit.
15 34968
[9] Hazra B, Sadhu A, Lourenco R and Narasimhan S 2010
Re-tuning tuned mass dampers using ambient vibration
measurements Smart Mater. Struct. 19 115002
[10] Fisco N R and Adeli H 2011 Smart structures: part I-active
and semi-active control Sci. Iran. 18 27584
[11] Fisco N R and Adeli H 2011 Smart structures: part IIhybrid
control systems and control strategies Sci. Iran. 18 28595
17

Smart Mater. Struct. 23 (2014) 015019

F Weber and M Maslanka

[39] Weber F, Hgsberg J and Krenk S 2010 Optimal tuning of


amplitude proportional Coulomb friction damper for
maximum cable damping J. Struct. Eng. 136 12334
[40] Bachmann H et al 1995 Vibration Problems in Structures:
Practical Guidelines (Basel: Birkhauser)
[41] Weber F 2013 BoucWen model-based real-time force
tracking scheme for MR dampers Smart Mater. Struct.
22 045012
[42] Savor Z, Radic J and Hrelja G 2006 Cable vibrations at
Dubrovnik bridge Bridge Struct. 2 97106
[43] Dancing Bridge of Wolgograd Tamed, Press Release 2013
Maurer Sohne Structural Protection Systems Division
Available at www.maurer-soehne.com/files/
bauwerkschutzsysteme/pdf/en/press release/MAU PR
Wolgograd engl 0312.pdf

[34] Maslanka M, Sapinski B and Snamina J 2007 Experimental


study of vibration control of a cable with an attached MR
damper J. Theor. Appl. Mech. 45 893917
[35] Weber F, Feltrin G and Distl H 2008 Detailed analysis and
modelling of MR dampers at zero current J. Struct. Eng.
Mech. 30 78790
[36] Boston C, Weber F and Guzzella L 2010 Modeling of a
disk-type magnetorheological damper Smart Mater. Struct.
19 045005
[37] Meirovitch L 2001 Fundamentals of Vibrations (New York:
McGraw-Hill)
[38] Weber F and Boston C 2010 Energy based optimization of
viscous-friction dampers on cables Smart Mater. Struct.
19 045025

18

Você também pode gostar