Sunteți pe pagina 1din 18

Journal of Wind Engineering

and Industrial Aerodynamics 88 (2000) 5774

Semi-active Tuned Mass Damper control strategy


for wind-excited structures
Francesco Ricciardellia, Antonio Occhiuzzib,*, Paolo Clementec
a

Dipartimento di Meccanica e Materiali,Universita` di Reggio Calabria, Italy


Dipartimento di Analisi e Progettazione Strutturale,Universita` di Napoli Federico II, via Claudio 21,
80125 Napoli, Italy
c
ENEA, Centro Ricerche della Casaccia, Rome, Italy
Received 18 October 1999; accepted 12 May 2000

Abstract
Uncertainties in the main structure dynamic properties as well as those in the characteristics
of the excitation may cause a deterioration of the performance of Tuned Mass Dampers.
This may happen because the choice of tuning and damping ratio of the auxiliary system,
based on the expected values of the dynamic properties of the main structure and on a simple
excitation pattern, may prove away from the optimum. An empirical algorithm is presented in
this paper which allows the performance of the TMD to be optimised, based on the measured
response. The algorithm relies on two assumptions: the smoothness of the spectrum of
excitation and the availability of an estimate of the dynamic properties of the main structure.
As an example the algorithm is applied to a 64-story building subjected to turbulence
buffeting. # 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Tuned mass damper (TMD); Semi-active control; Wind-excited structures

1. Introduction
Classical Tuned Mass Dampers (TMDs) made their way into practical Civil
Engineering applications in the 1970 s. The first major buildings featuring a TMD in
the USA were the John Hancock Tower in Boston, MA, completed in 1975 and the
Citicorp Center in New York, NY, completed in 1976. In both cases the devices were
aimed at mitigating the response induced by wind action. At that time the theory of
*Corresponding author. Tel.: +39-081-7683162; fax: +39-081-7683406.
E-mail address: icp@unina.it (A. Occhiuzzi).
0167-6105/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 0 0 ) 0 0 0 2 4 - 6

58

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

TMDs had already set criteria for the preliminary design of the devices. The vast
literature on the topic provides, in fact, closed-form expressions for the TMD
mechanical parameters that minimise the response of the main structure under either
an harmonic or a white noise excitation. These expressions are usually derived by
modelling the structure as a linear SDOF system (generally representing the first
mode of vibration) and assuming a linear behaviour for the TMD. The optimum
TMD parameters are a function of the modal parameters of the structure and of the
nature of excitation.
On the other hand, a constraint on the maximum allowable displacement of the
auxiliary mass has to be considered in the design of the TMD. To limit the stroke of
the TMD, an overdamping of the auxiliary system may therefore be necessary, which
results in a larger response of the primary system, and thus in a lower efficiency of
the device. A possible solution to the problem is that of controlling the TMD stroke
by applying control forces between the auxiliary mass and the primary system. This
solution was implemented on the TMD installed on the Citicorp Center, and is
described by Petersen [1]. The TMD stroke is controlled by an active servohydraulics
mechanism. An appropriate feedback algorithm for the device is shown by Lund [2].
In the last few years, a remarkable effort has been made in Japan both to design
and to build a new generation of TMDs, allowing for efficient control of the stroke
of the auxiliary mass. Sakamoto et al. [3] show significant data from theoretical and
experimental observations of the response of different buildings under actual
earthquakes and strong winds. A first structure was considered, having a PassiveActive Mass Damper (PAMD) made up of two parts: the first being a passive TMD
tuned according to the main structure dynamic properties, whereas the other is a
secondary mass damper connected to the first by a servo motor. The control activity
of the servo motor is based on a linear quadratic regulator (LQR) fed by the main
structure, TMD and AMD states. The goal is to achieve the required control effect
and to limit the stroke within an allowable range. A second structure was considered,
equipped with a V-shaped pendulum mass damper [4]. The V-shaped variable
geometry of the pendulum supporting rails allows a compact and tunable control
device to be obtained. Again, the control algorithm of the pendulum mass damper is
based on a LQR scheme, fedback by a selected number of main structure and
auxiliary mass state parameters.
A slightly different control scheme is presented by Tamura et al. [5]. In this case,
the gain matrix of the LQR is time variant to account for the effective behaviour of
the main structure. In particular, the authors refer to four different sets of matrices
selected according to the structural predominant modes in a short time period before
the activation of the control action.

2. From theory to practice: uncertainties affecting the performance of a TMD


The design of a TMD based on classical theory is quite straightforward, but the
results have to be checked out for a number of reasons.

59

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

First, Civil Engineering structures are generally quite complex, and the use of a
simplified SDOF system to represent their behaviour can lead to significant errors in
the estimation of the response. In addition, the use of either an harmonic or a flat
spectrum input may lead to a wrong estimation of the optimum TMD parameters
and of the system response. The problem of the evaluation of the optimum TMD
parameters and that of the response of complex structures to a realistic wind-induced
excitation was presented by Ricciardelli [6].
Second, the knowledge of the primary system properties is not always accurate
and these in many cases have to be treated on a probabilistic basis. This is due to two
main reasons: first, the structural properties are only known with a degree of
uncertainty [7] and, second, they can vary with time. The latter variation may arise
both from the deterioration of the structure, leading to a decrease in the natural
frequency and an increase in the damping, and from the variation of the mass,
related to the variation of the carried loads. The uncertainties in the primary system
properties can be taken into account by optimising the TMD parameters in a
probabilistic sense, as suggested by Jensen et al. [8], but this approach does not allow
for the time variation of the structural properties.
Third, the excitation is generally neither harmonic nor a white noise, and its
characteristics vary with the wind direction. These characteristics are related to the
upwind terrein topography and roughness, to the aerodynamics of the structure, to
the contribution of vortex shedding to the total excitation and to the interference
with neighbour structures. In addition, if the first lateral and torsional frequencies
are close to one another, a change in the direction of incidence may result in a change
of the most excited mode.
To show the effects of the uncertainties on the structural properties and on the
excitation, an exercise was undertaken. A slender building 257 m tall, with a square,
26.50 m  26.50 m plan was designed using the wind loads prescribed by the Italian
Loading Code [9], with a reference wind speed of 28 m/s at 10 m of elevation and a
roughness z0 0:05 m. A drag coefficient CD 1:4 was used for a wind direction
orthogonal to a face of the building, and a dynamic amplification coefficient
cd 1:2. The peak tip displacement of the building was evaluated through a static
finite element analysis, and it turned out to be 0.80 m (Table 1). The modal
properties of the building were also estimated using a Finite Element model. The first

Table 1
Main structure and TMD displacements (m)
Italian
Loading
Code
x1
x^1
x~1
x^2

0.80

White noise

KaimalSimiu spectrum

o1 0:9o1;

without
TMD

with
TMD

without
TMD

with
TMD

without
TMD

with
TMD

without
TMD

with
TMD

0.36
0.81
0.15

0.36
0.62
0.07
1.63

0.36
0.70
0.12

0.36
0.62
0.08
1.48

0.44
0.97
0.16

0.44
0.94
0.14
1.49

0.36
0.69
0.11

0.36
0.69
0.09
1.55

est

o1 1:1o1;

est

60

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

modal mass and natural frequency of 14.7  106 kg and 0.92 rad/s were obtained, and
a modal damping of 1% of critical was considered in the calculations.
To mitigate the resonant part of the dynamic response of the structure a TMD was
added, located at the tip of the building. The TMD had a mass of 1% of the first
modal mass of the building, and the tuning and damping ratio were chosen using the
criterion set by Luft [10] for the case of a white noise excitation. The TMD tuning
and damping ratio were 0.99 and 0.05, respectively.
Sixty-four DOF and 65 DOF finite element models were used for the evaluation
of the response of the system without and with the TMD, by integration of
the equations of motion. The model was acted upon by 16 time-varying forces,
applied to the centre of every fourth floor (i.e., acting at elevations of
16, 32, . . . , 256 m).
In the first stage, a fully coherent white noise dynamic excitation was considered
superimposed on a mean excitation consistent with the mean wind profile used in
the design of the structure. The dynamic excitation was associated with a uniform
intensity of turbulence Iu 0:08. Each of the 16 time histories of the excitation,
synthetically generated using a standard white noise signal generator, had a
length of 1200 s and a time step of 0.25 s. The results of the simulation are presented
in Table 1, and appear to be in good agreement with the design predictions. For
the uncontrolled structure a peak tip displacement of 0.81 m was obtained, made up
of a mean and an RMS displacement of 0.36 and 0.15 m, respectively (a gust factor
of 3.0 arose from the calculations). For the controlled structure the peak and RMS
tip displacement were reduced to 0.62 and 0.07 m, respectively (a gust factor of 3.7
was calculated in this case). A TMD efficiency parameter is defined as the
complement to unity of the ratio between the RMS responses of the controlled
and the uncontrolled structure, which for the arrangement considered had a value
of 0.53.
In the second stage, a more realistic excitation was considered, having the same
static component as the previous. The dynamic part, however, was generated
consistent with the KaimalSimiu turbulence spectrum [11] and with the coherence
function suggested by Vickery [12], with an exponential decay coefficient cz 10. A
uniform intensity of turbulence equal to 0.08 was considered. Sixteen time histories
which were 55 min long were considered with a time step of 0.2 s, synthetically
generated using an autoregressive filter of order 12. In this case, the peak and RMS
tip displacement of the uncontrolled structure (Table 1) were 0.70 and 0.12 m,
respectively (giving rise to a gust factor of 2.8), while those of the controlled
structure were 0.62 and 0.08 m (with a gust factor of 3.2). The efficiency of the TMD
in this case reduced to 0.33.
In the third stage, the effect of an error in the estimation of the first natural
frequency of the building on the performance of the TMD was analysed. Two cases
were considered in which the actual frequency of the building was 10% lower or
higher than the estimated value. In the first case the error arose from overestimating
the building stiffness by 21%, and in the second it arose from overestimating the
masses by the same amount. In the first case the overestimation of the system
stiffness caused an underestimation of the building static and dynamic response, that

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

61

the simulation revealed to be higher than expected. Mean, peak and RMS tip
displacements of the uncontrolled structure were calculated as 0.44, 0.97 and 0.16 m,
respectively (Table 1), corresponding to a gust factor of 3.3. On the other hand, in
both cases the efficiency of the TMD dramatically dropped, to 0.12 in the first case
and to a bare 0.08 in the second case. In addition, the peak displacement of
the controlled structure was only 3% lower than that of the uncontrolled structure in
the first case, while the two values even coincided in the second case.
The fact that the modal parameters of the building are known only with a degree
of uncertainty can be accounted for by using a probabilistic approach for the choice
of the mechanical parameters of the TMD. This technique was introduced by Jensen
et al. [8], and is based on the evaluation of the optimum TMD parameters as those
that minimise the expected value of the displacement of the primary system subjected
to a white noise input. A different solution is that suggested by Kareem et al. [13],
who introduce the idea of the multiple mass damper (MMD). The proposed device is
made of a number of auxiliary masses, each tuned to a different frequency in a range
around the expected resonant frequency of the structure. Being a range of
frequencies controlled by the device, the efficiency in the case in which there are
uncertainties in the frequency of the main structure is higher than that of a TMD
with equal total mass. The drawback of the proposed device is the higher complexity,
and thus higher costs with respect to the TMD.
The alternative proposed in this paper is a semi-active TMD, whose frequency and
damping can be varied with time to keep the performance of the device always close
to its optimum. Such a device would accomodate not only the uncertainties on the
parameters of the main structure, but also their variation, as well as the variation of
the excitation. The advantage of a semi-active device with respect to an active device
is the lower amount of energy required for the control [14]. External energy is in this
case required only to change the parameters of the device, but no control forces are
applied to the system. The semi-active device can be obtained from a passive TMD
equipped with pneumatic springs and hydraulic dampers, when the stiffness and
damping can be automatically varied with time.
The properties of the semi-active TMD have to be changed only on the basis
of the system measured response, since there are uncertainties in both the main
structure parameters and the excitation. An optimisation criterion based only on the
measured response is thus needed. In the remainder of this paper an empirical
criterion is proposed, linked to the hypothesis of a white noise excitation. The
criterion is then applied to the case of a real wind excitation to check its
performance.

3. Optimising the TMD properties from the measured response


In the hypotheses that the structural mode of vibration to be controlled is the first,
and that the higher modes give a negligible contribution to the total response, the
structure with the TMD can be modelled as the 2DOF system of Fig. 1. The

62

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

Fig. 1. Model of structure with the TMD.

equations of motions can be conveniently expressed in the form


8
< x 2x o x_ 2x o mx_ x_ o2 x O2 o2 mx x 1 qt
1
1
2
1
2
1 1 1
2 1
1 1
1
;
m1
:
2 2
x2 2x2 o1 x_ 2 x_ 1 O o1 x2 x1 0;

where o1 and x1 are the first natural frequency and damping ratio of structure and x2
is the damping ratio of the TMD, and where:
m

m2
;
m1

o2
o1

are, respectively, the mass and tuning ratios.


In the case in which the excitation is a white noise, the power spectrum of the
displacement x1 of the structure coincides with the squared modulus of the first row
first column term of the transfer matrix of the system of Eq. (1):
 jH11 o
 j2
Sx1 o

2
2
2
 2 4x22 O2 o
O o

2



2
2
 2 O2 4 x1 O2 x2 O x2 O x1 mx2 Oo
 4 O 4x1 x2 O 1 mO2 o
2
2 o
o
3

 o=o1 .
where o
For values of the tuning ratio close to 1, Eq. (3) can be represented as shown in
Fig. 2. The addition of a TMD, in fact, transforms the lowly damped first mode of
the uncontrolled structure into two coupled and highly damped modes of the 2DOF
system. Different criteria can be found in the literature [6] to define the optimum
TMD parameters Oopt and xopt
2 . Generally speaking the goal is that of minimising the
RMS of the structural response, i.e., minimising the area under the curve of Fig. 2.
By integration of Eq. (3) it is possible to estabilish a relation between the TMD
parameters and the variance of the response [15]. This, however, does not allow
expressions to be easily derived for the optimum tuning and TMD damping ratio.
On the other hand, even if that were possible, the results would depend on the

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

63

Fig. 2. PSDF of the main structure displacement.

knowledge of the main structure parameters and to the hypothesis of white noise
excitation.
The semi-active control strategy proposed in this paper is based on a trial-anderror procedure. Tentative parameters are given to the TMD, calculated on
the basis of any optimum criterion in the literature, applied to the expected values
of the structure modal parameters. The measured response of the structure is then
used to determine the behaviour of the device and to change its parameters in order
to improve its performances. However, a different criterion from that of minimising
the RMS of the response has to be established in this case, since knowledge of
the RMS of the response does not contain any information on the performance
of the device, nor on how to improve it. On the other hand, it was observed that the
knowledge of the spectral values y1, y2 and y3 in Fig. 2 can be used to understand
whether the device is close to its optimum performance or not, and in the latter
case to decide how to change its tuning and damping ratio to improve its
performances.
A first parametric analysis was carried out to assess the influence of tuning on the
RMS of the response of the main structure and on the spectral values y1 and y3. By
considering a number of combinations of the main structure damping x1 , of the mass
ratio m and of the TMD damping ratio x2 , the system response was evaluated for
different values of the tuning O . This allowed the tuning to be found to which the
minimum RMS response is associated and to relate the response of main structure to
the spectral values y1 and y3. This was done for values of the damping ratio of the
main structure x 1 in the range 0.0020.05, values of the mass ratio m in the range
0.0050.10 and values of the damping ratio of the TMD x2 in the range 0.030.15. It
was found that if the criterion of the minimum RMS response is substituted by the
following:
Oopt O: y1 y3

the RMS displacement of the structure with a TMD with the optimum tuning
exceeds the minimum value by no more than 0.6%.

64

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

The criterion of Eq. (4) allows the optimum tuning to be reached by correcting the
actual tuning through a factor q, which is a function of the spectral values y1 and y3:
Oopt qy1 ; y3 O:

The function qy1 ; y3 , on the other hand, depends on the system parameters x1 , m
and x2 , i.e., for every choice of the system parameters x1, m e x2 a function qy1 ; y3
can be found that satisfies Eq. (5). The dependance of q on y1 and y3 is well expressed
through the dependance on the parameter
y1 y3
;
6
t
y1 y3
which is a measure of the unbalance between the peaks P1 and P3 .
The parameter t can span the open interval ( 1, 1), and the function qt has to
satisfy the conditions
8
< t ! 1 ) q ! 1;
qt:
t 0 ) q 1;
7
:
t ! 1 ) q ! 1:
The following expression will be used for q:
qt 1 k tanat

in which a and k account for the dependance on the system parameters x1 , m


and x2 .1
The parametric analysis showed that the dependance of a and k on x2
is quite weak and can be neglected therefore. The dependance on x1 and m is
expressed as:
k k0 ma
a a0 mb

where k0, a 0, a and b are functions of x 1. The values of k0, a 0, a and b are given in
Table 2, and for k0 the following interpolating expression can be used:
k0 1:07x0:22
1 :

10

As an example, Fig. 3 shows the function q(t) for x1 0:002 and m =0.02, and for
different values of the TMD damping. In the figure, the results of the parametric
analysis are shown, together with Eq. (8).
Once the correction factor is evaluated from Eq. (8), the optimum tuning is given
to the system by multiplying the stiffness of the TMD by the square of q.
As a second step, a similar procedure is used to find the optimum TMD damping
ratio x 2. It was found that the minimum response of the main structure is generally
achieved when y2=12 y1=12 y3. In Fig. 4, for x1 0:01 and for different values
of m , the ratio between the RMS response and its minimum value is plotted as a
function of y2 =y1 . As anticipated, the minimum response is always obtained for
y1 =y2 2.
1

In order to satisfy Eq. (7) it should be a = p /2, however different values will be used to better
interpolate the results of the parametric analysis.

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

65

Table 2
Values of the parameters of Eq. (9)
x1

k0

a0

0.002
0.005
0.010
0.020
0.050

0.29
0.33
0.38
0.46
0.58

0.86
0.91
0.73
0.61
0.36

0.53
0.56
0.53
0.52
0.42

0.073
0.071
0.116
0.168
0.273

Fig. 3. Function q(t). x1 0:002 and m =0.02.

Fig. 4. RMS structural response vs. the ratio between the spectral values y1 and y2.

The following optimum criterion is then adopted:


1
xopt
2 x2 : y2 4y1 y3 :

11

As for the tuning, the optimum TMD damping ratio is obtained from the actual
value through the correction
xopt
2 px2 ;

12

66

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

in which the correction factor p depends on the parameter


s

y1 y3
:
2y2

13

The parameter s has to be a positive number and the dependancy of p on s such that
p2 1. The following expression is considered:
p p0 s c ;

14

where p0 and c are functions only of the main structure damping x1, being the
dependancy on the mass ratio m negligible. The following expressions were found
for p0 and c:
;
p0 0:74x0:0084
1

15

c px1 0:1:

As an example, in Fig. 5 the function ps is shown for x1 0:01 and for different
values of the mass ratio m. In the figure, the results of the parametric analysis are
shown, together with Eq. (14).
The procedure shown is effective for a system subjected to a white noise excitation,
and allows for the direct calculation of the optimum values of the TMD tuning and
damping ratio from the knowledge of the actual values of these parameters, based on
the system measured response. For a system subjected to a broad-band excitation, it
is assumed that the corrected values of the TMD tuning and damping ratio obtained
from Eqs. (5), (8)(10), (12), (14) and (15), though not necessarily the optimum, are
better than those before the correction, i.e., bring about a lower response of the
main structure. If this is true, the procedure shown can be used in an iterative way, to
improve the performance of a non-optimum TMD. Moreover, if changes arise in the
excitation or in the main structure properties, the procedure can accomodate for the
variations by updating the TMD parameters.
The possibility of implementing the iterative procedure in the case of a system with
varying parameters, subjected to a broad-band excitation is considered in the next
section.

Fig. 5. Function p(s) for z1 0:01.

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

67

Fig. 6. Building elevation.

4. Numerical application
The control scheme shown in the previous sections was applied in a numerical
example. A slender structure with a braced frame was designed and its behaviour
in the along-wind direction analysed. The design was based on the Italian Loading Code [9]. The 64-story building was 257 m tall and had a square plan
(26.5 m  26.5 m). The total mass was equal to 57.9  103 kg. Both dead and live
loads were the same at all floors, but the structural mass of columns and bracings
decreases with the height. Figs. 6 and 7 show, respectively, a schematic elevation and
a structural plan. Table 3 shows the cross-sections adopted for columns and braces.
At the corners, compound columns made up by four welded beams were used. The

68

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

Fig. 7. Building plan.

Table 3
Cross-sections of structural members
Floors

Columns cross-section

Braces cross-section

112
1324
2536
3748
4960
6164

HSH 600/828
HSH 500/631
HSH 500/640
HSH 400/297
HSH 400/150
HE 360 B

HSH 400/407
HSH 400/407
HSH 400/191
HSH 400/191
HE 360 B

dynamic properties of the structure without the damper were evaluated using a FEM
model and are shown in Table 4 for the first three modes. The FEM model was made
up by 3748 beam elements connected in 1664 nodes. The SAP90 programme was
used to compute eigenvalues and eigenvectors. The damping ratio corresponding to
each mode was set to 1%.

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

69

Table 4
Calculated dynamic properties of the building
Mode

Period (s)

Frequency (Hz)

Frequency (rad/s)

1
2
3

6.822
1.724
0.986

0.1466
0.5802
1.014

0.9211
3.645
6.369

A reduced model of the controlled building was created including only the first
three longitudinal sway modes of the uncontrolled frame and the TMD as shown in
Ref. [6]. The TMD is characterised by a constant mass m2 and variable stiffness k2 and
damping c2. The tuning and TMD damping were chosen following the criterion
presented by Luft [10], derived for an undamped system subjected to a white noise
excitation:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
;
16
Oopt
1 1:5m
xopt
2

1pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m1 0:75m;
2

17

where m was set to 1% of the estimated structural mass associated with the first
mode. However, in order to test the control scheme, we positively introduced
estimation errors in computing Eqs. (16) and (17). In particular, the building stiffness
was underestimated by 10% and the mass overestimated by 10%. Therefore, the
initial parameters of the TMD, which were affected by these errors, are
m2 162103 kg;
k2 109:16 kN m1 ;
c2 13:24 kN m1 s:

18

The stiffness and the damping of the TMD were subsequently updated to optimise
the structural response, as will be shown in the following.
As a first step, the natural frequency f0 of the structure with the TMD locked to
the main structure was evaluated from the PSDF of the top floor displacement.
Subsequently the TMD was released and the PSDF of the top floor displacement
plotted. In the latter case the PSDF had two peaks, corresponding to the frequencies
f1 and f3 , around f0 f1 5f0 5f3 and a local minimum located between them. In
order to identify the peaks, the following procedure was used. The spectrum was
scanned in order to locate its maximum value at the frequency fmax . If fmax 5f0 , then
f1 fmax , otherwise f3 fmax . Then the local maximum point greater than f1 (less
than f3) was found, which located f3 ( f1). The frequency f2 was then set equal to the
minimum point in the open interval [ f1, f3 ]. If a local maximum point greater than f1
(less than f3) could not be found, it was set f3=f0( f1=f0). However, this would
happen only if the tuning were quite away from its optimum value, i.e., if the initial

70

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

estimate of the structural properties were quite inaccurate. Furthermore, in order to


avoid identifying a peak in the background range, the spectrum was scanned only in
a selected window around f0.
The procedure provided the spectral coordinates y1, y2 and y3 needed to use
the empirical optimisation criterion introduced in the previous section. The
numerical procedure, however, needed a high resolution in the spectrum and,
therefore, a relatively long duration of the displacement records. In the subsequent
application, 55 min long records, with a sampling frequency of 5 Hz were
adopted. The spectra were calculated by dividing the time histories into subrecords 820 s long, with an overlap of 205 s, to obtain a spectrum resolution of
0.00122 Hz.
The numerical procedure started by calculating the structural response for about
55 min. In this phase, the TMD was not active, i.e., it was locked to the main
structure. At the end of this stage, it was possible to identify the natural frequency f0 .
In the second phase the TMD was unlocked and the structural response was
calculated for the same time interval. From the spectrum of the top floor
displacement, the frequencies f1 and f3 were identified, as well as the corresponding
spectral values. By applying Eqs. (8)(10) the stiffness of the TMD could finally
be updated. The procedure was then iterated as many times as needed to satisfy
Eq. (4).
Once the stiffness update was completed, the spectrum of the top floor displacement was then used to identify the frequency f2 and the associated spectral
value y2. By applying Eqs. (12), (14) and (15), the damping coefficient of the TMD
was updated and then the new structural response measured. The procedure was
then again iterated, as many times as needed to satisfy Eq. (11).
At the end of the damping update, however, Eq. (4) might not be satisfied
anymore. Therefore it might be necessary to alternatively update the stiffness and the
damping of the TMD.
Table 5 shows the results of the iterative procedure described above. Looking
at the first two rows, we notice that even a TMD designed according to an erroneous
estimation of stiffness and mass of the main structure has an efficiency of about
24%. Fig. 8 shows the PSDFs corresponding to the along-wind displacement of
the upper floor. Iterations from 1 to 8 represent the structural responses during
the stiffness updating cycle. Eq. (4) is practically satisfied at iterations 7 and 8.
Fig. 9 shows the PSDFs corresponding to the initial set up of the TMD and
to iterations 1 and 8. During the stiffness update phase, the RMS of the top
floor displacement decreases and is 34% lower compared to the case of building
with the TMD locked. Iterations 9 to 13 refer to the damping update cycle.
Eq. (11) is practically satisfied at iteration 13, where the standard deviation of the top
floor displacement is 2% higher than that of iteration 8. However, the standard
deviation of the TMD displacement is 19% lower than that of iteration 8. Once the
damping cycle was completed, the stiffness update procedure was repeated. Iteration
14, however, shows no significant variations of the response either for the main
structure or for the TMD. Fig. 10 shows the PSDFs corresponding to iterations
8 and 14.

Without TMD
With TMD
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Iteration 6
Iteration 7
Iteration 8
Iteration 9
Iteration 10
Iteration 11
Iteration 12
Iteration 13
Iteration 14

Table 5
Numerical results

13.24
13.24
13.24
13.24
13.24
13.24
13.24
13.24
13.24
19.22
22.84
25.50
27.30
28.45
28.45

(KN/(m/s))

(KN/m)

109.16
115.16
116.67
118.86
122.04
124.90
126.44
127.40
128.03
128.03
128.03
128.03
128.03
128.03
126.75

c2

k2

0.891
0.915
0.921
0.930
0.942
0.953
0.959
0.963
0.965
0.965
0.965
0.965
0.965
0.965
0.960

4.98%
4.85%
4.82%
4.77%
4.71%
4.65%
4.63%
4.61%
4.60%
6.67%
7.93%
8.85%
9.48%
9.88%
9.93%

x2

0.36
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37

(m)

x1

0.80
0.71
0.69
0.68
0.68
0.66
0.66
0.65
0.65
0.65
0.66
0.66
0.66
0.66
0.67
0.67

(m)

x^1

0.129
0.098
0.093
0.091
0.090
0.087
0.086
0.086
0.085
0.085
0.086
0.086
0.087
0.088
0.088
0.088

(m)

x~1

1.29
1.30
1.31
1.38
1.48
1.53
1.55
1.56
1.56
1.28
1.15
1.06
1.01
0.98
0.98

(m)

x^TMD

0.378
0.400
0.405
0.411
0.417
0.422
0.425
0.426
0.427
0.356
0.325
0.306
0.295
0.288
0.288

(m)

x~TMD

0.148
0.144
0.146
0.146
0.146
0.133
0.133
0.133
0.133
0.133
0.140
0.140
0.140
0.140
0.140
0.140

(Hz)

f1
2

5.991
0.309
0.561
0.499
0.417
0.401
0.438
0.447
0.450
0.451
0.466
0.520
0.553
0.573
0.585
0.556

(m /t)

y1

0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150
0.150

(Hz)

f3
2

1.910
1.423
1.288
1.092
0.825
0.618
0.522
0.469
0.436
0.457
0.477
0.492
0.503
0.510
0.542

(m /t)

y3

0.145
0.148
0.148
0.148
0.145
0.145
0.145
0.145
0.145
0.145
0.145
0.145
0.145
0.145
0.145

(Hz)

f2

0.255
0.536
0.469
0.383
0.106
0.086
0.078
0.074
0.072
0.133
0.177
0.212
0.238
0.255
0.257

(m2/t)

y2

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774


71

72

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

Fig. 8. PSDF of the tip displacement.

Fig. 9. PSDF of the tip displacement. Initial values of TMD parameters and iterations 1 and 8.

Fig. 10. PSDF of the tip displacement. Iterations 8 and 14.

5. Feasibility considerations
The control scheme shown in the previous sections is based on the measure of the
structural response for time intervals lasting about 1 h and on the assumption that
stiffness and damping of the TMD can be varied in a reasonably short time. In

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

73

Refs. [16,1] a comprehensive description of the technology used to design and build
the TMD applied to the Citicorp Building in New York in 1976 can be found. Even
though it is a classical, passive TMD, the manufacturer had to design a device
with time varying properties. The elastic connection between the TMD and the
building is made up by a series of pneumatic springs with a geometric compensation
of the inherent non-linearity of adiabatic gas compression in a closed vessel. The
pressure can be varied  6% around the nominal value. The damping ratio of the
hydraulic damper between the TMD and the main structure can be changed by a
special servovalve from 8% to 14% of the critical. In addition, a locked position
of the servovalve allows a damping ratio of about 70%. The time needed to update
stiffness and damping is below 5 min. From Table 5 it appears that the control
scheme proposed in this paper needs a stiffness variation of about 8% around the
initial value and a damping ratio variable in the range 410% of critical. Therefore,
it can be concluded that the technology needed to implement the Semi-Active TMD
proposed is practically already available, and that the adoption of a semi-active
control scheme adds no major costs in buildings already designed to host a Tuned
Mass Damper.

6. Conclusions
In this paper an empirical algorithm for the optmisation of the properties of
Tuned Mass Dampers based on the measured response of the main structure is
presented. The algorithm does not need the exact knowledge of the main structure
mechanical properties, nor it is bound to any particular form of excitation. For its
implementation only an estimate of the main structure first frequency is required,
together with the smoothness of the spectrum of the excitation. The procedure allows
the properties of the TMD to be up-dated in order to improve its performance, and
is thus effective not only in the case in which there are uncertainties in the main
structure properties and in the excitation, but also in the case in which either the
structural properties or the excitation vary with time.
A numerial example showed the effectiveness of the procedure in optimising
the TMD tuning for a structure for which there is no exact knowledge of the
natural frequencies. On the other hand the optimisation of the TMD damping
did not bring any reduction of the main structure response, but rather to a
reduction of the displacement of the added mass. This can be explained by the low
sensitivity of the response of the main system to the TMD damping. The
possibility, however, of reducing the response of the TMD without decreasing its
effectiveness seems attractive. As a future development, an optimisation
criterion accounting for both the structure and the TMD responses seems to be of
interest.
Finally, the semi-active control strategy presented in the paper does not need any
technology significantly more sophisticated, thus more expensive, than that used for
many of the existing Tuned Mass Dampers.

74

F. Ricciardelli et al. / J. Wind Eng. Ind. Aerodyn. 88 (2000) 5774

Acknowledgements
This paper is dedicated to the memory of the late Prof. Mario DApuzzo, under
whose supervision the work was started.
The authors also wish to express their gratitude to Prof. Gianni Bartoli of the
University of Florence, who provided them with the wind speed time series used in
the simulations.

References
[1] N.R. Petersen, Using servohydraulics to control hi-rise building motion the citicorp TMD system,
Proceedings of National Conference on Fluid Power, Chicago, IL, USA, 1981.
[2] R.A. Lund, Active damping of large structures in winds, in: H.H.E. Leipholz (Ed.), Structural
Control, North-Holland, Amsterdam, 1980.
[3] M. Sakamoto, T. Kobori, T. Yamada, M. Takahashi, Practical applications of active and hybrid
response control systems and their verifications by earthquake and strong wind observations,
Proceedings of First World Conference on Structural Control, Los Angeles, CA, USA, 1996.
[4] Y. Koike, T. Murata, K. Tanida, M. Mutaguchi, T. Kobori, K. Ishii, Y. Takenaka, T. Arita.
Development of V-shaped hybrid mass damper and its application to high-rise buildings, Proceedings
of First World Conference on Structural Control, Los Angeles, CA, USA, 1996.
[5] K. Tamura, K. Shiba, Y. Inada, A. Wada. Control gain scheduling of a hybrid mass damper system
against wind response of a tall building, Proceedings of First World Conference on Structural
Control, Los Angeles, CA, USA, 1996.
[6] F. Ricciardelli, A linear model for structures with tuned mass dampers, Wind Struct. 2 (3) (1999)
151172.
[7] G. Solari, Gust-excited vibrations, in: H. Sockel (Ed.), Wind-Excited Vibrations of Structures,
Springer, Berlin, 1996.
[8] H. Jensen, M. Setereh, R. Peek, TMDs for vibration control of systems with uncertain properties,
J. Struct. Div. ASCE 118 (1992) 32853296.
[9] D.M. LL. PP. 16.01.1996, Norme tecniche relative ai criteri generali per la verifica di sicurezza delle
costruzioni e dei carichi e sovraccarichi (in Italian).
[10] R.W. Luft, Optimal tuned mass dampers for buildings, J. Struct. Div. ASCE 105 (1979) 27662772.
[11] E. Simiu, R.H. Scanlan, Wind Effects on Structures: an Introduction to Wind Engineering, Wiley,
New York, 1986.
[12] B.J. Vickery, On the reliability of gust loading factors, C.E. Trans., Inst. Aust. Soc. Civ. Eng. (1)
(1971) 93104.
[13] A. Kareem, S. Kline, Performance of multiple mass dampers under randon loading, J. Struct. Eng.
ASCE 121 (2) (1995) 348361.
[14] G.W. Housner, L.A. Bergman, T.K. Caughey, A.G. Chassiakos, R.O. Claus, S.F. Masri, R.E.
Skelton, T.T. Soong, B.F. Spencer, J.T.P. Yao, Structural control: past, present and future, J. Eng.
Mech. 123 (9) (1997) 897971.
[15] S.H. Crandall, W.D. Mark, Random Vibration in Mechanical Systems, Academic Press, New York,
1963.
[16] N.R. Petersen, Design of large scale tuned mass dampers, Proceedings of ASCE National
Convention, Boston, MA, 1979.

S-ar putea să vă placă și