Вы находитесь на странице: 1из 14

Vol.8, No.

4 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION December, 2009

Earthq Eng & Eng Vib (2009) 8: 547-560 DOI: 10.1007/s11803-009-9115-3

Optimal design theories of tuned mass dampers with


nonlinear viscous damping
Lap-Loi Chung1, Lai-Yun Wu2, Hsu-Hui Huang2, Chung-Hsin Chang2* and Kuan-Hua Lien2+

1. Center for Research on Earthquake Engineering (NCREE), Chinese Taipei


2. Department of Civil Engineering, Taiwan University, Chinese Taipei

Abstract: Optimal design theory for linear tuned mass dampers (TMD) has been thoroughly investigated, but is still
under development for nonlinear TMDs. In this paper, optimization procedures in the time domain are proposed for
design of a TMD with nonlinear viscous damping. A dynamic analysis of a structure implemented with a nonlinear
TMD is conducted first. Optimum design parameters for the nonlinear TMD are searched using an optimization method
to minimize the performance index. The feasibility of the proposed optimization method is illustrated numerically by
using the Taipei 101 structure implemented with TMD. The sensitivity analysis shows that the performance index is less
sensitive to the damping coefficient than to the frequency ratio. Time history analysis is conducted using the Taipei 101
structure implemented with different TMDs under wind excitation. For both linear and nonlinear TMDs, the comfort
requirements for building occupants are satisfied as long as the TMD is properly designed. It was found that as the
damping exponent increases, the relative displacement of the TMD decreases but the damping force increases.

Keywords: optimal design theory; tuned mass damper (TMD); nonlinear; performance index; sensitivity analysis;
Taipei 101

1 Introduction Taipei 101 Tower (101 stories, 504 m high) (Haskett et


al., 2003), Taiwan. From field vibration measurements, it
The TMD system was first proposed in 1909 to has been proven that a TMD is an effective and feasible
reduce the mechanical vibration induced by monotonic system to use in structural vibration control against high
harmonic forces by Frahm (1911). It is found that if a wind loads.
secondary system composed of a mass, damping device, Although the basic design concept of the TMD
and spring is implemented on a primary structure and is quite simple, the parameters (mass, damping, and
its natural frequency is tuned to be very close to the stiffness) of the TMD system must be obtained through
dominant mode of the primary structure, a large reduction optimal design procedures to attain a better control
in the dynamic responses of the primary structure can be performance. Therefore, the determination of optimal
achieved. The Center-point Tower (305 m) in Sydney, design parameters of the TMD to enhance the control
Australia was the first structure in which a TMD was effectiveness has become very crucial. Den Hartog
installed to reduce the wind-induced motions of the (1956) first proposed an optimal design theory of the
structure. Since then, there have been a considerable TMD for an undamped primary system subjected to
number of successful structural implementations of the sinusoidal excitation. A single degree-of-freedom
TMD worldwide, including the Citicorp Center (279 m (SDOF) structure implemented with a linear TMD
high), New York; John Hancock Tower (457 m high), subjected to white-noise base excitation was analyzed
Boston; Funade Bridge Tower, Osaka; CN Tower (535 by random vibration (Crandall and Mark, 1973). If the
m high), Toronto (McNamara., 1977; Luft., 1979; Soong natural frequencies of a structure were well separated,
and Dargush, 1997) and the tallest building in the world, it could be considered as a SDOF structure when
determining the optimum parameters of the attached
Correspondence to: Lap-Loi Chung, Center for Research on
Earthquake Engineering(NCREE), 200, Sec. 3, HsinHai
linear TMD (Warburton and Ayorinde, 1980; Warburton,
Rd., Chinese Taipei, 106 1982). Expressions of the optimum TMD parameters
Tel: +886-2-66300800; Fax: +886-2-66300858 were derived for an undamped structure subjected to
E-mail: chung@ncree.org.tw various excitations. The effect of structural damping

Research Fellow and Adjunct Professor; Professor; PhD; was also investigated. For the case of a damped
*
Master; +PhD Candidate primary structure, it is difficult to obtain the closed-
Supported by: Science Council(NSC), Chinese Taipei form solutions for the optimum TMD parameters. The
Received August 8, 2009; Accepted September 1, 2009 design parameters of a linear TMD for a linear damped
548 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

primary structure were optimized based on minimum- confirmed by stochastic simulation.


maximum amplitude criteria (Randall et al., 1981). In addition to passive TMD, active TMD (Chang
Computational graphs were presented to determine the and Soong, 1980; Soong, 1990) and semi-active TMD
optimal linear TMD. The tuning ratio of the linear TMD (Hrovat et al., 1982; Abe, 1996; Abe and Igusa, 1996;
attached to a viscously damped structure was optimized Ricciardellia et al., 2000; Yalla et al., 2001; Nagarajaiah
by the frequency locus method (Thompson, 1981). The and Varadarajan, 2005) have been proposed. Although
optimum damping ratio of the linear TMD could be found active and semi active TMDs possess higher
directly once the optimum tuning ratio was obtained. performance in vibration suppression, they must have
Using the perturbation technique, design formulas of control algorithms, sensors, external power and higher
optimum parameters were derived for a linear TMD maintenance cost.
under various types of loadings (Fujino and Abe, 1993). In this paper, an optimal design of a TMD with
Using a numerical searching technique, the optimum nonlinear viscous dampers implemented on a SDOF
damping and tuning frequency ratio of the linear TMD structure is investigated. The dynamic equation of the
attached to a viscously damped SDOF main system system is expressed in discrete-time state space. The
under external force and base acceleration modeled as sufficient and necessary conditions for the optimal
Gaussian white-noise random process was obtained design parameters of the nonlinear TMD are derived
to minimize various mean square responses. Explicit from the minimization of a performance index. Since
formulae for damper damping, tuning frequency and the the sufficient and necessary conditions are nonlinear,
corresponding minimized response were derived using a the method of steepest gradient is adopted. The
curve-fitting technique (Bakre and Jangid, 2007). design parameters are updated with the optimal step
In the papers mentioned above, the damping of obtained from a golden section search. The design
TMD was assumed to be linear. However, nonlinear parameters keep updating until the difference between
dampers may be cost effective in fabrication and the performance indices of two consecutive iterations
maintenance. Friction dampers acting transversely to is small enough. The feasibility of the proposed
the direction of the motion were proposed as a means optimization method is illustrated numerically by
for energy dissipation of TMD (Inaudi and Kelly, 1995). using the Taipei 101 structure implemented with
Subjected to white-noise support excitation, the mean TMD. By varying the design parameters around the
square structural responses, estimated by statistical optimal values, the sensitivity of design parameters to
linearization, were minimized to obtain the optimum the performance index is studied. The performance of
parameters of TMD. The results were justified by Monte TMDs with different damping power law exponents
Carlo simulation. TMD with dry friction damping implemented on Taipei 101 are compared.
were investigated based on an understanding of the
behavior of the system (Ricciardlli and Vickery, 1999).
2 Equation of motion of structure with
Under harmonic excitation, closed-form expressions
nonlinear TMD
for optimum TMD parameters and system responses
were derived by the balance of work and energy.
The primary structure is modeled as a SDOF
Optimum design parameters were determined from
structure with ms, cs, ks and w(t) being the generalized
the level of excitation and the target system response.
mass, damping, stiffness and external disturbance,
Design parameters of a TMD with a nonlinear power
respectively. Its equation of motion can be described as
law viscous damper under white noise excitation were
optimized by statistical linearization such that the xs (t ) + cs x&s (t ) + ks xs (t ) = w(t )
ms && (1)
variance of structural displacement was minimized
(Rudinger, 2006). Contrary to the linear TMD, the After a nonlinear TMD is attached to the SDOF
optimal parameters for the nonlinear TMD depended structure, as shown in Fig. 1, it becomes a two-degree-
on the structural damping and excitation intensity. The of-freedom (2DOF) system. The equation of motion
approximate solution from statistical linearization was of the nonlinear system can be expressed as

w(t)
ks

fr
xs xd
cs
md
fd
ms

Fig. 1 System model of nonlinear TMD attached to SDOF structure.


No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 549

&&(t ) + Cx& (t ) + Kx (t ) = bc f d (t ) + bc f r (t ) + ec w(t )


Mx (2) The damping force fd[k] obeys a nonlinear viscous
damping law and the restoring force fd[k] obeys a
linear law. They are, respectively, given by
x d (t )
where x 21 (t ) = is the displacement vector
x s (t ) f d [k ] = c d x& d [k ] x& s [k ] sgn( x& d [k ] x& s [k ]) =
m 0 0 0
of the 2DOF system; M 22 = d , C 22 = c d D 2 z[k ] sgn( D 2 z[k ]) (7a)
0 ms 0 c s
0 0
and K 22 = are the mass, damping and
0 k s f r [k ] = k d ( x d [k ] x s [k ]) = k d D1 z[k ] (7b)
stiffness matrices of the system, respectively; xd(t)
and xs(t) are displacements of the TMD and the SDOF where cd is the damping coefficient of the TMD with
structure, respectively; md is mass of the TMD; fr(t) dimension of mass.length1-.time-2; is the damping
and fd(t) are, respectively, the restoring force and power law exponent of the TMD; kd is the stiffness
the nonlinear damping force provided by the TMD; of the TMD; (D1)14 = [1 1 0 0] and (D2)14 =
( bc ) 21 = [ 1 1] and ( e c ) 21 = [0 1] are the TMD
T T
[0 0 1 1] are, respectively, the output matrices for
internal loading and the primary structural external the displacement and velocity of the TMD relative to
loading vectors, respectively. the structure. After Eqs. (7a) and (7b) are substituted
The equation of motion (Eq. (2)) is conveniently into Eq. (6), the discrete-time state equation can be
expressed as a first-order differential equation in state rewritten as
space,

z&(t ) = Az (t ) + bf d (t ) + bf r (t ) + ew(t ) (3) z[k + 1] = Ad z[k ] + cd bd D2 z[k ] sgn( D2 z[k ]) +
kd bd D1 z[k ] + ed w[k ] (8)
x (t ) 0 I
where z 41 (t ) = , A44 = ,

x& (t )
1
M K M 1C
3 Derivation of optimal design theory for
0 0 nonlinear TMD
b41 = 1 and e 41 = 1 .
M bc M ec
The optimal design parameters for nonlinear
The solution of Eq. (3) is
TMD, cd and kd, can be obtained such that a
certain performance index is minimized. The most
z (t1 ) = e A(t1 t 0 ) z (t 0 ) + appropriate performance index J is the sum of square
t1 of an interested structural response y[k],
e A(t1 ) {bf d ( ) + bf r ( ) + ew( )} d (4)
t0 k1
J = y 2 [k ] (9)
In the interval between two consecutive sampling k =0
instants, t 0 = k t and t1 = (k + 1)t , it is assumed
that the nonlinear damping force f d ( ) , the restoring where k1 is the instant at which external loading
force f r ( ) and the external force w( ) are piecewise is terminated. The structural response y[k] is a
constant, combination of structural states and external force,

f d ( ) = f d (k t ) y[k ] = Dz[k ] + Ew[k ] (10)



f r ( ) = f r (k t ) for k t < (k + 1)t (5)
w( ) = w(k t ) where D14 and E are constant with appropriate

dimensions. When the above equation is substituted
where t is the sampling period. into Eq. (9), the performance index J can be rewritten
Under the assumption mentioned above, the as
solution (Eq. (4)) can be expressed in discrete-time
fashion as a first-order difference equation, k1
J = z T [k ]D T Dz[k ] + 2 z T [k ]D T Ew[k ] + E 2 w2 [k ] (11)
k =0

z[k + 1] = Ad z[k ] + bd f d [k ] + bd f r [k ] + ed w[k ] (6)


Since the state vector z[k] in the above equation
should satisfy the state equation (Eq. (8)), Eq. (8)
where z[k ] = z (k t ) , ( Ad )44 = e , ( bd )41 = A ( Ad I )b ,
At 1
is the equality constraint to the performance index.
( ed )41 = A1 ( Ad I )e , f d [k ] = f d (k t ) , f r [k ] = f r (k t ) After incorporation of the equality constraint, the
and w[k ] = w(k t ) . performance index becomes
550 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

k1
J = z T [k ]D T Dz[k ] + 2 z T [k ]D T Ew[k ] + E 2 w2 [k ] converged. If the step s is too small, convergence is
k =0 guaranteed but the number of iterations may be very
large. Therefore, a golden section search method is
{ }
k1
+ T [k ] Ad z[k ] + cd bd D2 z[k ] sgn( D2 z[k ]) + kd bd D1z[k ] + ed w[k ] z[k + 1]

k =0
adopted to find the optimal step of increment (Lewis.,
(12) 1986; Lee et al., 2006). The iteration procedures are
repeated until the tolerance between two consecutive
where 41 [ k ] is the Lagrange multiplier vector. Since iterations is acceptable (Fig. 2).
the performance index J' is a function of the state
vector z[k], the co-state vector [k], the TMD damping 4 Numerical verifications and sensitivity
coefficient cd and the TMD stiffness kd, the sufficient analysis
and necessary conditions for the minimization of the
performance index J' are In this paper, the mass md and the stiffness kd
of the TMD are normalized to be dimensionless as,
J respectively,
= Ad z[k ] + cd bd D2 z[k ] sgn( D2 z[k ]) + kd bd D1 z[k ] +
[k ] md
Rm = (15a)
ed w[k ] z[k + 1] = 0, z[0] = z0 (13a) ms

J d kd
= 2 D T Dz[k ] + 2 D T Ew[k ] + AdT [k ] + Rf = = (15b)
z[k ] s k s Rm
1
cd D2 z[k ] bdT [k ] + kd D1T bdT [k ]
where Rm is the ratio of the TMD mass md to the first
[k 1] = 0 , [k1 ] = 0 (13b) structural modal mass mS; Rf is the ratio of the TMD
frequency d to the first structural modal frequency
s.
k The structural response y[k] reduced by the
J 1



= T [k ]bd D 2 z[k ] sgn( D 2 z[k ]) = 0 (13c) application of TMD may be the displacement xs[k],
c d k =0 the velocity x& s [k ] and the acceleration &&
x s [k ] of the
k structure,
J 1

=
k d k =0
T [k ]bd D1 z[k ] = 0 (13d)
y[k ] = Dz[k ] + Ew[k ] =
xs [k ], D = [ 0 1 0 0] , E = 0.
where z[0]=z0 is the initial condition for the state (16)
equation (Eq. (13a)) and [k1]=0 is the terminal x&s [k ], D = [ 0 0 0 1] , E = 0.
condition for the co-state equation (Eq. (13b)).
xs [k ], D = 0 s2 0 2ss , E = 1 ms .
&&
The sufficient and necessary conditions
represented by Eq. (13a) to (13d) for the optimal In order to evaluate the effectiveness of the TMD,
design parameters of TMD are so complex that the the performance indices are, respectively, normalized
closed-form solution cannot be obtained. In this as the displacement ratio Rd, the velocity ratio Rv, and
paper, the steepest gradient method is adopted. the acceleration ratio Ra,
The initial guess for the TMD damping coefficient
and the TMD stiffness are assigned as cd(0) = 0 and
( J xs ) with TMD
kd(0) = md (2fs ) 2 , respectively. The state vector z[k] Rd = (17a)
and co-state vector [k] are, respectively, solved from ( J xs ) without TMD
Eqs. (13a) and (13b) in sequence. The gradients of the
TMD damping coefficient and the TMD stiffness in
Eqs. (13c) and (13d) usually will not vanish and they ( J x& s ) with TMD
are used to update the parameters, Rv = (17b)
( J x& s ) without TMD

J ( 0)
c d(1) = c d( 0) s (14a)
c d ( J &&xs ) with TMD
Ra = (17c)
( J &&xs ) without TMD
J ( 0)
k d(1) = k d( 0) s (14b)
k d where Rd, Rv and Ra are, respectively, the ratios of the
root mean square of the structural displacement xs,
where s is the step of increment for design parameters. velocity x& s and acceleration &&
x s with TMD to those
If the step s is too large, the solution may not be without TMD.
No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 551

Fig. 2 Flow-chart of iteration procedures

4.1 Numerical verifications parameters with the lowest performance index from the
3-D graphs are close to the optimal design parameters
The external wind force, w(t), is assumed to be from the optimization procedure. Therefore, the
white noise as shown in Fig. 3. With mass ratio Rm proposed method for the optimal design parameters of
1.25%, the optimal design parameters c dopt and Rfopt TMD is feasible. Table 1 shows that: (1) the higher the
are computed by the proposed optimization procedures damping exponent , the larger the optimal damping
for damping exponent, = 0.5, 1.0 and 2.0 such that coefficient c dopt ; (2) the optimal frequency ratio Rfopt
the performance indices, Rd, Rv and Ra are minimized changes slightly for different damping exponents ;
(Table 1). Furthermore, the 3-D variation of the and (3) the optimal displacement ratio R dopt and
performance indices, Rd, Rv and Ra against the TMD acceleration ratio Raopt remain almost the same for
design parameters, cd and Rf, are plotted as shown different damping exponents .
in Figs. 4-6. As listed in Table 1, the TMD design
552 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

1
w (103kN)

-1

-2
0 200 400 600
Time (s)
Fig. 3 External white-noise wind force

Table1 Numerical verifications for Rm = 1.25%


0.5 1.0 2.0
opt
R Rf 0.9961 0.9916 0.9873
d

cd (kN ( s m ) )

13.62 59.34 988.96
Rd 0.5915 0.5721 0.5600
opt
R Rf 0.9983 0.9942 0.9893
v

cd (kN ( s m ) )

13.70 59.74 999.70
Rv 0.5857 0.5671 0.5561
opt
R Rf 0.9936 0.9886 0.9842
a

cd (kN ( s m ) )

13.60 59.09 980.43
Ra 0.8271 0.82059 0.8164

0.75 0.75
0.70 0.70
0.65 0.65
Rd
Rd

0.60 0.60
0.55 0.55
0.50 0.50
1.06 1.06
1.02 25 30 1.02 100 120
0.98 20 0.98 80
60
Rf
0.94 5 10 15 Rf
0.94 20 40
cd (kN ( s m ) ) cd (kN ( s m ) )

(a) = 0.5 (b) = 1.0

0.75
0.70
0.65
Rd

0.60
0.55
0.50
1.06
1.02 2000
1600
Rf 0.98 1200
0.94 400 800
cd (kN ( s m ) )

(c) = 2.0
Fig. 4 3-D variation of displacement ratio Rd
No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 553

0.75 0.75

0.65 0.65

Rv
Rv

0.55 0.55
1.06 1.06
1.02 25 30 1.02 100 120
0.98 20 0.98 80
Rf 15 Rf 60
0.94 5 10 0.94 20 40
cd (kN ( s m ) ) cd (kN ( s m ) )

(a) = 0.5 (b) = 1.0

0.75

0.65
Rv

0.55
1.06
1.02 2000
1600
Rf 0.98 1200
0.94 400 800
cd (kN ( s m ) )

(c) = 2.0
Fig. 5 3-D variation of velocity ratio Rv

0.90 0.90

0.86 0.86
Ra
Ra

0.82 0.82
1.06 1.06
1.02 25 30 1.02 100 120
0.98 20 0.98 80
60
Rf
0.94 5 10 15 Rf
0.94 20 40
cd (kN ( s m ) ) cd (kN ( s m ) )

(a) = 0.5 (b) = 1.0

0.90

0.86
Ra

0.82
1.06
1.02 2000
1600
Rf 0.98 1200
0.94 400 800
cd (kN ( s m ) )

(c) = 2.0
Fig. 6 3-D variation of acceleration ratio Ra

4.2 Sensitivity analysis


acceleration ratio Raopt is about 0.82 for damping
When the mass ratio Rm is 1.25%, the optimal exponent = 0.5, 1.0 and 2.0 (Figs. 7 and 8). When
displacement ratio R dopt is about 0.58 and the optimal the optimal damping coefficient c dopt varies by 10%,
554 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

the displacement ratio Rd changes by 0.4205%, 25.5797% for damping exponent = 0.5, 1.0 and 2.0 ,
0.2796% and 0.1257% for damping exponent respectively, and the acceleration ratio Ra changes
= 0.5, 1.0 and 2.0 , respectively, and the acceleration by 5.4726%, 5.3299% and 5.6728% for damping
ratio Ra changes by 0.098%, 0.059% and 0.026% for exponent = 0.5, 1.0 and 2.0 , respectively. The
damping exponent = 0.5, 1.0 and 2.0 , respectively. performance indices Rd and Ra are very sensitive to the
The performance indices Rd and Ra are less sensitive frequency ratio Rf . In other words, the performance
to the damping coefficient cd. When the optimal index is more sensitive to the frequency ratio Rf than
frequency ratio Rfopt varies by 10%, the displacement to the damping coefficient cd.
ratio Rd changes by 21.7986%, 22.7732% and

0.50% 0.10%
= 0.5 = 0.5
= 1.0 0.08% = 1.0
0.40%
= 2.0 = 2.0
[Rd(Rd)opt]/(Rd)opt

[Ra(Ra)opt]/(Ra)opt
0.30% 0.06%

0.20% 0.04%

0.10% 0.02%

0 0
0.90 0.95 1.00 1.05 1.10 0.90 0.95 1.00 1.05 1.10
Cd/(Cd)opt Cd/(Cd)opt
Fig. 7 Sensitivity of displacement ratio Rd (left) and acceleration ratio Ra (right) to damping coefficient (cd)

30% 6%
= 2.0 = 2.0
= 1.0 = 1.0
= 0.5 = 0.5
[Rd(Rd)opt]/(Rd)opt

[Ra(Ra)opt]/(Ra)opt

20% 4%

10% 2%

0 0
0.90 0.95 1.00 1.05 1.10 0.90 0.95 1.00 1.05 1.10
Rf /(Rf)opt Rf /(Rf)opt

Fig. 8 Sensitivity of displacement ratio Rd (left) and acceleration ratio Ra (right) to frequency ratio (Rf)

5 Analysis example 33 m in length, down to the bed rock. The total floor
area is about 374,000 m2 including the floor area
The Taipei 101 structure holds the title of world's about 30,277 m2 of the basement under the ground.
tallest building (for now) at 508 m, shown in Fig. 9. The underground structure is made of RC-SRC. There
The superstructure is made of steel and the boxed are eight mega columns which are filled with concrete
cross section columns in the lower stories are filled of strength 68.9 MPa.
with concrete. The excavation depth is 22.25 to 22.95 Because the TMD is hung in the 92th floors, in order
m. The foundation constitutes a huge reinforced to turn the structure into SDOF structure, the first mode
concrete slab 3.0 to 4.7 m in thickness. Underneath shape is normalized such that the component at the 92th
the concrete slab, there are 380 friction piles, 15 to floor is unity (Wu et al., 2005; Johnston, 2006). After
No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 555

model reduction, the first modal mass ms is 52696.5 t, is 1.25% (md = 658.71 t). The spherical mass is
the first modal frequency fs is 0.14251 Hz, and the supported by a sling of eight steel cables. Eight
first modal stiffness ks = ms (2fs ) 2 is 42250.0 kN/m. viscous dampers act like shock absorbers when the
It is assumed that the first modal damping ratio s is sphere shifts as shown in Fig. 10. The viscous damper
2% so that the first modal damping coefficient can be is nonlinear of damping power law exponent = 2.0
obtained as cs = 2ms (2fs ) = 1513.2 kN s/m . with damping coefficient cd = 1,000 kN.(s/m)2. Each
The mass of TMD for Taipei 101 is 660 t, the damper is 3.1m in length and inclines 60 degrees with
heaviest in the world, so that the mass ratio Rm the floor. It can provide at most 1,000 kN of damping
force. The mass is able to move 1.5 m in any direction.
This gold-colored orb is on view from restaurants,
bars and observation decks between the 88th and
92nd stories. A bumper ring prevents the ball from
10000 7200 45800=23200 53400=17000

swaying too far, should that much swaying ever need


to occur. The length of the pendulum for the TMD is
determined from the TMD stiffness, which is in turn is
determined by the frequency ratio. The frequency ratio
Rf is 0.9855, computed from the equation proposed by
Sadek (Sadek et al., 1997).
The Taipei 101 structure is excited by external
wind force with a return period of half a year as
shown in Fig. 11 (Wu et al., 2005). In addition to
the original TMD implemented on the Taipei 101
structure, optimal designs of TMD with a minimum
acceleration ratio Ra for damping exponents =
0.5, 1.0 and 2.0 are also investigated. The design
parameters are listed in Table 2. The frequency ratio
Rf for the original TMD and the optimal frequency
ratio Rfopt for the TMD with = 2.0 are 0.9855 and
0.9768, respectively. They are different by 0.89%. The
damping coefficient cd for the original TMD and the
844200=352800

opt
448000

optimal damping coefficient c d for the TMD with


= 2.0 are 1000.0 and 903.17 kN(s/m)2, respectively.
They are different by 10.72%. Since the performance
is less sensitive to the damping coefficient cd than to
the frequency ratio Rf , the acceleration ratios Ra for
46300=25200
8400
4200 4500
5600
19400
33100=9300

(a) Side view (b) Plane view


Fig. 9 Structure of Taipei 101 (Wu et al., 2005)
556 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

the two cases are close to each other. No matter what


the damping exponents are, the acceleration ratios
Ra are similar and lie between 0.6662 and 0.6992.
When no TMD is implemented on the Taipei 101
structure, the peak structural displacement xs max
and acceleration && xs max are 7.850 cm and 6.501
cm/s2, respectively, (Table 2 and Figs. 12-13). With
the implementation of TMDs, the peak structural
displacement xs max is reduced to the range from
5.555 cm (optimal TMD with = 2.0) to 5.775
cm (optimal TMD with =1.0) (Fig. 12) and the
peak structural acceleration && xs max is reduced to
the range from 4.423 cm/s2 (optimal TMD with
= 1.0) to 4.575 cm/s2 (optimal TMD with = 2.0)
(Fig. 13). All the TMDs match the requirement of
comfort. In order to realize the reduction in structural
Fig. 10 Tuned mass damper (TMD) attached to Taipei 101 response, the peak TMD relative displacement
(Wu et al., 2005) xs xd max is in the range from 22.72 cm (original

1.5

1.0

0.5
w (103kN)

-0.5

-1.0

-1.5
0 200 400 600
Time (s)
Fig. 11 External wind force with return period of half a year

Table 2 Analysis results for Taipei 101

Without Original TMD


0.5 1.0 2.0
TMD (2.0)
Rfopt 0.9855 1.0042 0.9964 0.9768

cdopt (kN ( s m ) )

1000.0 16.73 65.69 903.17

Raopt 0.6992 0.6662 0.6794 0.6980

xs max
(cm) 7.850 5.710 5.708 5.775 5.555

xs
&& max
(cm / s 2 ) 6.501 4.592 4.285 4.423 4.575

xs xd max
(cm) 22.72 26.95 25.61 23.37

fd max
(kN) 42.24 23.59 30.64 40.31

TMD) to 25.61 cm (optimal TMD with = 1.0) hysteretic loops of the TMD damping forces fd versus
(Fig. 14) and the peak TMD damping force f d max the corresponding TMD relative displacements (xs xd)
is in the range from 30.64 kN (optimal TMD with are plotted as shown in Fig. 14. When the damping
= 1.0) to 42.24 kN (original TMD) (Fig. 14). exponent is = 1.0, the hysteretic loop is a circle
The higher the damping exponent , the smaller or ellipse. It is close to a rectangle and an eye shape
the peak TMD relative displacement xs xd max but when = 0.5 and = 2.0, respectively.
the larger the peak TMD damping force f d max . The
No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 557

0.09
Displ. of main system (m)

Without TMD
0.06
0.03
0
-0.03
-0.06
-0.09 0 200 400 600

0.09 Original TMD


Displ. of main system (m)

0.06
0.03
0
-0.03
-0.06
-0.09
0 200 400 600

0.09
Displ. of main system (m)

Optimal TMD with = 0.5


0.06
0.03
0
-0.03
-0.06
-0.09
0 200 400 600

0.09
Displ. of main system (m)

Optimal TMD with = 1.0


0.06
0.03
0
-0.03
-0.06
-0.09
0 200 400 600

0.09
Displ. of main system (m)

Optimal TMD with = 2.0


0.06
0.03
0
-0.03
-0.06
-0.09
0 200 400 600
Time (s)

Fig. 12 Structural displacement time histories of Taipei 101


558 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8
Acc. of main system (m/s2)

0.08
Without TMD
0.04

-0.04

-0.08
0 200 400 600
Acc. of main system (m/s2)

0.08
Original TMD
0.04

-0.04

-0.08
0 200 400 600

0.08
Acc. of main system (m/s2)

Optimal TMD with = 0.5


0.04

-0.04

-0.08
0 200 400 600
Acc. of main system (m/s2)

0.08
Optimal TMD with = 1.0
0.04

-0.04

-0.08
0 200 400 600

0.08
Acc. of main system (m/s2)

Optimal TMD with = 2.0


0.04

-0.04

-0.08
0 200 400 600
Time (s)

Fig. 13 Structural acceleration time histories of Taipei 101


No.4 Lap-Loi Chung et al.: Optimal design theories of tuned mass dampers with nonlinear viscous damping 559

45 Original TMD 45
Optimal TMD with = 0.5
30 30
Damping force (kN)

Damping force (kN)


15 15

0 0

-15 -15

-30 -30

-45 -45
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
Rel. displ. of main system (m) Rel. displ. of main system (m)

45 Optimal TMD with = 2.0


45
Optimal TMD with = 1.0
30 Damping force (kN) 30
Damping force (kN)

15 15

0 0

-15 -15

-30 -30

-45 -45
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
Rel. displ. of main system (m) Rel. displ. of main system (m)

Fig. 14 Hysteretic loops of TMDs for Taipei 101

of comfort for building occupants is satisfied as


6 Conclusions long as the TMD is properly designed. It was found
that as the damping exponent increases, the relative
In this paper, an optimal design for a TMD with displacement of the TMD decreases but the damping
nonlinear viscous damping is developed. Numerical force increases.
procedures are proposed to systematically and
efficiently determine the optimal design parameters
for nonlinear TMD such that a certain performance References
index is minimized. The feasibility of the proposed
optimization method is illustrated numerically by Abe M (1996), Semi-active Tuned Mass Dampers for
using the Taipei 101 structure implemented with Seismic Protection of Civil Structures, Earthquake
TMD. From the results of the sensitivity analysis, the Engineering and Structural Dynamics, 25: 743749.
performance index is less sensitive to the damping Abe M and Igusa T (1996), Semi-active Dynamic
coefficient than to the frequency ratio. The higher the Vibration Absorbers for Controlling Transient
damping power law exponent, the performance index Response, Journal of Sound and Vibration, 198:
becomes more sensitive to the frequency ratio, and 547569.
less sensitive to the damping coefficient. Time history Bakre SV and Jangid RS (2007), Optimum Parameters
analysis was conducted for the Taipei 101 structure of Tuned Mass Damper for Damped Main System,
implemented with different TMDs and subjected to Structural Control and Health Monitoring, 14: 448470.
wind loads with a return period of half a year. For
Chang JCH and Soong TT (1980), Structural Control
both linear and nonlinear TMDs, the requirement
Using Active Tuned Mass Dampers, Journal of
560 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.8

Engineering Mechanics, 106(6): 10911098. Optimum Vibration Absorbers for Linear Damped
Crandall SH and Mark WD (1973), Random Vibration in Systems, Journal of Mechanical Design, 103: 908913.
Mechanical Systems, NY: Academic Press. Ricciardellia F, Occhiuzzib A and Clementec P (2000),
Den Hartog JP (1956), Mechanical Vibrations. 4th ed. Semi-active Tuned Mass Damper Control Strategy for
NY: McGraw Hill. Wind-excited Structures, Journal of Wind Engineering
and Industrial Aerodynamics, 88: 5774.
Frahm H (1911), Device for Damping Vibration of
Bodies. U.S. Patent No. 989958. Ricciardlli F and Vickery BJ (1999), Tuned Vibration
Absorbers with Dry Friction Damping, Earthquake
Fujino Y and Abe M (1993), Design Formulas for
Engineering and Structural Dynamic, 28: 707723.
Mass Dampers Based on a Perturbation Technique,
Earthquake Engineering and Structural Dynamic, 22: Rudinger F (2006), Optimal Vibration Absorber with
83354. Nonlinear Viscous Power Law Damping and White
Noise Excitation, Journal Engineering Mechanics,
Haskett T, Breukelman B, Robinson J and Kottelenberg
132(1): 4653.
J (2003), Tuned Mass Dampers under Excessive
Structural Excitation, Report of the Motioneering Inc. Sadek F, Mohraz B, Taylor AW and Chung RM
Guelph, Ontario, Canada NIK 1B8. (1997), A Method of Estimating the Parameters of
Mass Dampers for Seismic Applications, Earthquake
Hrovat D, Barak P and Rabins M (1982), Semi-active
Engineering and Structural Dynamic, 26: 617635.
Versus Passive or Active Tuned Mass Damper for
Structural Control, Journal of Engineering Mechanics, Soong TT (1990), Active Structural Control: Theory and
109(3): 691705. Practice, New York: Longman Scientific Tech.
Inaudi JA and Kelly JM (1995), Mass Damper Using Soong TT and Dargush GF (1997), Passive Energy
Friction-dissipating Devices, Journal Engineering Dissipation Systems in Structural Engineering, New
Mechanics, 121(1): 142149. York: Wiley.
Johnston B (2006), Project Spotlight: Taipei 101: The Thompson AG (1981), Optimum Tuning and Damping
experience, Elevator World, 54(6): 26132 of a Dynamic Vibration Absorber Applied to a Force
Excited and Damped Primary System, Journal of
Lee CL, Chen YT, Chung LL and Wang YP (2006),
Sound and Vibration, 77: 403415.
Optimal Design Theories and Applications of Tuned
Mass Dampers, Engineering Structures; 28: 4353. Warburton GB (1982), Optimal Absorber Parameters
for Various Combinations of Response and Excitation
Lewis FL (1986), Optimal Control, New York: Wiley.
Parameters, Earthquake Engineering and Structural
Luft RW (1979), Optimal Tuned Mass Dampers for Dynamic, 10: 381401.
Building, Journal of the Structural Division, ASCE,
Warburton GB and Ayorinde EO (1980), Optimum
105: 27662772.
Absorbers for Simple Systems, Earthquake Engineering
McNamara RJ (1977), Tuned Mass Dampers for and Structural Dynamic, 8: 197217.
Building, Journal of the Structural Division, ASCE,
Wu LY, Chung LL, Chen JC and Huang GL (2005),
103: 19851998.
Control Simulation for Taipei 101 with Wind
Nagarajaiah S and Varadarajan N (2005), Short Time Excitation, 7th Conf. on Structural Engineering,
Fourier Transform Algorithm for Wind Response Taiwan. (in Chinese).
Control of Buildings with Variable Stiffness TMD,
Yalla SK, Kareem A and Kantor J C (2001), Semi-active
Engineering Structures, 27: 431441.
Tuned Liquid Column Dampers for Vibration Control of
Randall SE, Halsted DM and Taylor DL (1981), Structures, Engineering Structures, 23:14691479.

Вам также может понравиться