Вы находитесь на странице: 1из 22

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL.

22, 833-854 (1993)

DESIGN FORMULAS FOR TUNED MASS DAMPERS


BASED ON A PERTURBATION TECHNIQUE

YOZO FUJINO*
Department of Civil Engineering, The Universiiy of Tokyo, Tokyo 113, Japan

AND

MASATO ABE+
Deparimeni of Civil Engineering, Norihwestern Universiiy. Evanston, Illinois 60208, U.S.A.

SUMMARY
Modal properties of tuned mass damper (TMDtstructure two-degree-of-freedom (2DOF) linear systems are studied
employing a perturbation technique. Using the perturbation solutions, formulas relevant to designing the T M D for
various types of loading are obtained; they are expressed as a function of mass ratio, tuning ratio, damping ratio of the
T M D and damping ratio of the structure. Equivalent additional dampings of the structure due to the T M D are derived
for random and harmonic forces. Matched expressions of equivalent damping, which are valid for detuned, i.e.
non-optimal, conditions are also presented. The stability boundary of TMD-structure systems subject to linear
self-excited forces is derived in a closed form. Using the perturbation solutions, procedures for optimizing the T M D
parameters for various types of loading are explained and the optimal values are derived. The formulas obtained in this
study can be used with good accuracy for mass ratios less than 0.02.

INTRODUCTION
Use of the tuned mass damper (TMD) has increased recently in civil engineering structures.
A TMD-structure system generally has non-classical damping, and thus requires a complex eigenvalue
analysis to compute modal properties. In order to avoid this, several studies have aimed at expressing the
eigenvalues in a closed form, utilizing perturbation techniques (for example References 1-3).
In this study, a structure is modelled as a single-degree-of-freedom (SDOF) system, which is justified since
the TMD is designed to suppress a specific vibration mode of the s t r ~ c t u r eThe
. ~ TMD-structure system can
therefore be treated as a 2DOF system, and its eigenvalue problem is equivalent to solving a quartic
characteristic equation, which has an analytic form of solution after Ferarri5 The form of the solution is,
however, too complicated to use in engineering design.
The optimal values of damping and frequency of the T M D for various types of loading have been obtained
when the damping of the structure is negligibly mall.^'^" However, it is rare that they are used in actual
design since there are generally constraints for design of the TMD, for example limitation of permissible
TMD stroke length. It is known that performance of the T M D with the optimal parameters is very sensitive
to the uncertainties of the parameters of the T M D and the structure. With these constraints, one must solve
the eigenvalue problem or equations of motion for each set of parameters, and hence design work for the
TMD becomes a trial-and-error process. In order to overcome this, empirical formulas for design have been
proposed on the basis of numerical results.' Although they are very useful to design, they lack a physical
basis.
In this paper, by modifying a perturbation method proposed by Igusa and Der Kiureghian,'.' we derive
simple and explicit solutions for eigenvalues of a TMD-structure system. Using these, characteristics of

* Professor
Graduate student (formerly graduate student of the University of Tokyo)

0098-8847/93/lOO833-22$16.00 Received 3 August 1992


0 1993 by John Wiley & Sons, Ltd. Revised 2 February 1993
834 Y. FUJINO AND M. ABE

a TMD-structure under free vibration, random excitation, harmonic excitation and self-excitation are
discussed and design formulas are obtained. Because of the closed form of the solutions for the eigenvalues,
the process of optimization of the TMD can easily be understood.

EXPRESSION OF EIGENVALUES
Equations of motion
The TMD-structure system is modelled as a 2DOF system as shown in Figure 1. The equations of motion
are
MX + ell+ KX = F (1)
where

x is the (modal) displacement of the structure to which T M D is attached and y is the displacement of the
TMD relative to the structure. Subscript S denotes structure and T for TMD, while m denotes mass,
5 damping ratio, and w natural frequency. f ( t ) is the external force. The following parameters are introduced
for the purpose of mathematical expression.
w, = (ws + 0T)/2 (average natural frequency) (2)
y = wT/wS (tuning ratio) (3)
5, = (ts + tT)/2 (average damping) (4)
td= tS- tT(damping difference) (5)
p = mT/mS (mass ratio) (6)
Assuming the mode shape as 0 = [&, &IT, then
[A’M + AC + K ] 0 =0 (7)
In order to have a non-trivial solution, the determinant of equation (7), which is called the characteristic
equation, has to be 0
+ 2msws5sA + msw;
det [
msA2
mTA2
- 2mToTtTA - mT&
mTA2 + 2mToT<TA + mT0: ] = o (8)

Figure 1. TMD-structure system


TUNED MASS DAMPERS 835

that is,
A4 + 2CwS5.5 + (1 + p)wT5T113 + + ( l + p)w: + 4wSwT5S5TlAr
+ 20swT [ wT 5s +0 s (T] 1 + 0;w: =0 (8‘)
-_
Solving this equation, one obtains two pairs of complex conjugate roots A1, A 2 , A1,Az ( - stands for complex
conjugate) in general. Here, the jth modal frequency and modal damping are defined as follows:

wj = Jm (9)
t j= -a j 1 J m ( j= 1,2) (10)
where the j-th eigenvalue is expressed as
l j = aj + bji ( a j , bj:real) (11)

Perturbation point
The purpose of this section is to express the complex eigenvalues as
A = io*{i<* + Jm+ (i = J-1)
where E is the parameter of perturbation. The perturbation point is defined as
A* = - w*<* + iw*J1 - 5*2 (12’)
Here we choose the perturbation point to be the duplicate roots of the characteristic equation (8).
Eigenvalues are then expressed as
1 = a k bi (a, b: real) (13)
Then the characteristic equation is
A4 -4 d3 + ( 6 ~ +’ 2b2)A2- 4a(a2 + bZ)A+ ( a 2 + b2)’ = 0 (14)
Comparing the coefficients in equation (14) with those in equation (8), one obtains

under the condition 4; < p. This is a sufficient condition to obtain positive and real values of wT, tT,w*
and (*. When the damping of the structure, ts, is negligible,

w* = w,/Jl +p
t* = &I2
These values are indeed equal to the optimal parameters of the TMD for free vibration.’. lo
836 Y. FUJINO A N D M. ABE

For simplifying equations (15)--(18), the first-order terms with respect to rsand p are maintained:
OT - p)wS (1Sb)
5r = JTr + 5s U6b)

r-

Considering

and comparing equation (19) with equation (17b),


w* = 0, (20)
Similarly, the average damping ratio, ta,is

Then

When equations (15b) and (16b) are satisfied, two eigenvalues are equal, and are approximated from
equations (20) and (22) as

In this paper, I* defined in equation (23) is used as the perturbation point. If equations (15b) and (16b) are
satisfied in a TMD-structure system, E in equation (12) becomes 0.
A parameter to express the closeness of the natural frequencies of the structure and T M D is introduced
and is defined on the basis of equation (I5a) as

This fl will be used later in analysis.


Perturbation solution of eigenvalues
To apply a perturbation method, the following order relations are assumed, which are generally satisfied in
civil engineering structures:

O(B) = a t , ) = O(&) 6 O(1) (25)


Note that the mass ratio, p, is around 0.01 and at most 0.02 in civil engineering applications. Substituting
equations (12), (1 3) and (24) into equation (8), and maintaining the first-order terms, the characteristic
equation is expressed as
4E2 - ( B + itd)’ - =0
Solving this with respect to E,

& = ktJp + (a + icd)’ (27)


Then, the two eigenvalues are

l j = iaa(l + it, Isgn(B)Jp + ( B + itd)’), ( j = 192) (28)


where sgn(8) denotes the sign of p. This sign is introduced so that mode 1 under detuned conditions
corresponds to the structure’s dominant mode.’
TUNED MASS DAMPERS 837

According to equations (9H1l), the modal frequency and modal damping are expressed as

The upper sign is for j = 1 and the lower for j = 2, and Re and Im denote the real and imaginary part,
respectively.
The following parameters regarding the modal frequency and damping are introduced:
-
p=- 0 1 - 0 2 -
- ReJp + (fi + itd)' (31)
ma

fd = 51 - (2 ?Z ImJP + ( b + itd)2 (32)


These will also be used later.
The difference between equations (29) and (30) and the results by Igusa and Der Kiureghian'.' lies in the
definition of B; they used p = (ws - wT)/wainstead of equation (24).In this paper, recognizing that efficiency
of the T M D is strongly influenced by the tuning condition, b has been carefully defined.

RESPONSE TO FREE VIBRATION


It is important to optimize the T M D in order to reduce residual (i.e. free) vibration of the structure caused
from a shock. The perturbation solutions obtained are used to derive the optimal parameters of the T M D for
free vibration.
Damping of the whole system is governed by the smaller modal damping because both modes are, in
general, excited by an impulsive force. Hence, under optimal conditions the two modal damping ratios are
equal, i.e. = t2.Since the TMD is tuned to the structure, i.e. wi x w, the modal damping is expressed as
t j x 5, f f&?from equations (30) and (32). It is clear from this that 5 d = 0 is the necessary condition for
= t2,which means that either t d = 0 or = 0 (under p - 5; 2 0).
If <d ( = 5, - tT)= 0, both modal dampings and t2 are indeed the same and equal to Ca( = 5, = tT)
since ws ?Z wT.However, consider an extreme case when 5, = 0. In this case tTshould be equal to zero. It is
clear that t d = 0 does not lead to the optimal parameter.
The other condition, B = 0, eventually gives the optimal condition; the optimal tuning ratio is given from
equation (24) as
Y ( w T / O S ) = l/(l + p ) (33)
When the damping of the structure is negligible, relations among the damping parameters become simple
such as td= tT,5, = 5T/2 and t j = t,. In this case the modal damping is equivalent to maximizing tTunder
the condition that p - t: 2 0. The optimal damping ratio of the TMD is therefore expressed as
tT =JP (34)
and the modal damping of the TMD-structure system under the optimal parameters is

tj = &/2 O' = 192) (35)


Equations (33), (34)and (35) are good approximations to the exact solutions shown in equations (15), (16)and
(1 8), respectively.
In the presence of damping of the structure, the optimal parameters of the T M D can similarly be
obtained as
Y = 1/(1 + P I (33')
(34')
The errors in y and tTcompared with the exact solutions (equations (15) and (16))are 0.04 and 0.3 per cent,
respectively, when p = 0.02 and ts= 0.02; these errors would be acceptable.
838 Y.FUJINO AND M. ABE

The modal properties were obtained numerically as well as by perturbation solutions (equations (29) and
(30))and are compared in Figures 2 and 3 for the case when ts = 0 and p = 0.01. In Figure 3, the solutions by
Igusa and Der Kiureghian'V2 are also shown.
It is seen from these figures that the perturbation solutions agree well with the exact onesin the case when
the TMD and the structure are tuned (Figure 2) but not well when they are detuned (Figure 3). Their
accuracy can be improved by means of the matching method'." using solutions under the detuned
condition.
When the T M D and the structure are not tuned, interaction between them is weak. The eigenvalues of the
system converge to the following:
= - costs +i w S , / ' m

1.0

0.9 \
0 0.05 0.10 0.15 0.20
(a) 5T
Sj
TUNED MASS DAMPERS 839

while in equation (28), if /? is large,

Comparing equation (36) with equation (37), it is to be expected that replacing tS and tTby ( s w s / o ,and
tTuT/w,,respectively, should increase the accuracy. With these corrections, both ws/w,and wT/w, approach
unity when the system is tuned. Hence, it retains its accuracy even when they are tuned. Solutions shown in

PERTURBATTON

0.8 0.9 1.0 1.1 1.2


Y

0.12 I t I
AFI'ER IGUSA
0.10

0.08

0.06

0.04

0.02

"
0.8 0.9 1.o 1.1 1.2
(b) Y
Figure 3. Modal frequency and damping as a function of y(p = 0.01; tS= 0;
damping
tT = Jm):
(a) modal frequency; (b) modal
840 Y. FUJINO AND M. ABE

equations (28H30)are termed tuned expressions and the solutions after correction are termed matched
expressions.' Matched expressions of the modal frequency and the damping are shown as

The matched expressions are shown in Figure 4 for the same case as Figure 3 and it is found that they are
more accurate over a wide range of the tuning ratio, y.

1.2

1.1

1.0

0.9

0.8 1 1
0.8 0.9 1.0 1.1 1.2
(a)

0.12 1 1 1
--.-
-MATCHED EXPRESSION
0.10

0.08

0.06

0.04

0.02

(b) OB 0.9 1.0 1.1 y 1.2

Figure 4. Modal frequency and damping matched expression ( p = 0.01; tS= 0; tT = d m ) :(a) modal frequency; (b) modal
damping
TUNED MASS DAMPERS 84 1

RESPONSE WHEN SUBJECTED T O RANDOM FORCES


In this section, the external force f ( x ) is assumed to be a white-noise random process for the purpose of
simplicity. Excitations due to turbulent winds or earthquakes are not white noise, but are often modelled as
a white-noise process for T M D design. The efficiency of the T M D is evaluated in terms of the mean-square
responses of the structure and TMD.

Exact solution
The mean-square response of a SDOF system subject to a white-noise input with spectral density Socan be
obtained from the following integral:I2
co
E [ z 2 ] = So IH(w)I2d o
-a,

where z is the displacement and H ( w ) is the frequency response function of the system.
For the 2DOF system shown in Figure 1, the frequency response functions are
H,(o) = ( - W2 + 2 i < ~ o +~ w+)/(rnsA)
o
Hy(W) = 02/(msA)
where
A = o4 - 2i[Os<s + (1 + p)WT(T]W3 - [Oi + (1 + p)’$ + 4wSwT<S<T]w2
f 2iWsoT[oT<s + os<T]w + Wio:

Performing the integral of equation (40) using equations (41aH41c), one obtains
r .

E [ x 2 ] = nSo
c J

where

a
-E[x’] =0
X T

In the case of rs= 0, it is found as

tT = zId-=- (1 + p)(1 + p/2)


842 Y.FUJINO AND M.ABE

From the analogy of response of a SDOF system, which is

the equivalent damping teq


of the TMD-structure system is defined as

Using this definition, the equivalent damping under the optimal condition is expressed as

The non-dimensionalized amplitude of the T M D is defined as

Under the optimal conditions, amplitude AT of the TMD is

Perturbation solution
As is shown in equation (42), the mean-square responses are obtained analytically. Here the perturbation
solution is applied to obtain simpler formulas. Using these an optimization procedure is explained in the
following.
Taking the lowest-order terms in equation (42), one obtains the mean-square response of the structure and
TMD:

The equivalent damping and non-dimensionalized amplitude of the T M D are, respectively,


< 2w3 t S t T ( P 2 + 4r:) + Pt:
=a
cq
pla + 2tT(P2 + 45:) -"

As shown in the previous section, the optimal parameters of the T M D are determined by taking the
derivatives of the mean-square response of the structure. When the damping of the structure, ts,is zero, the
optimal parameters are obtained as
optimal tuning ratio:
(54)
optimal damping of TMD:

(55)
TUNED MASS DAMPERS 843

and
equivalent damping under optimal condition:

non-dimensionalized amplitude of the TMD:


1
AT = __ (571
4%
These are good approximations of equations (44), (47) and (49) which are exact solutions: the errors of the
perturbation solutions (equations (54H57)) are less than 1 per cent for the mass ratio p = 0.02.
In Figures 5 and 6, the perturbation solutions of equivalent damping teqand non-dimensionalized
amplitude AT are compared with exact values computed from numerical analysis. It can be seen that the
perturbation solution for the equivalent damping loses its accuracy under detuned conditions; the matching
technique is again considered. When the system is detuned, which means that jl is large, the perturbation
solution of equivalent damping approaches

Physically, when the system is detuned, the TMD is ineffective, i.e. the equivalent damping should approach
zero. Replacing w,/ws in equation (52)with 1 gives a matched expression. When the system is tuned, o,/os is
almost equal to unity and this matching expression is applicable to the tuned case. The matched expression
for 5, is shown in Figure 6(a). The non-dimensionalized amplitude of the TMD is expected to approach zero
because structural vibration is dominant when the system is detuned. From equation (53), when fl is large, it
also approaches zero; equation (53) is a matched expression as well.
Note that equations (44a) and (44b) as well as equations (54) and (55) are obtained for tS= 0. The
numerical study by Warburton' showed that in the presence of the structural damping, tS,the optimal
tuning ratio of the T M D is very slightly affected, while the optimal damping is unchanged. When tsis low
( < 0.02) the optimal parameters (equations (54) and ( 5 5 ) ) may be used.

RESPONSE WHEN SUBJECTED TO HARMONIC FORCES


The efficiency of the T M D is usually evaluated at the peak of the frequency response of a structure when it is
designed for periodic loading.
It is assumed herein that the frequency response function (equation (41a)) takes a peak value when the
excitation frequency is equal to one of the modal frequencies. Substituting the perturbation solutions of the
natural frequency (equation (29)) into equation (41a), and maintaining the first-order terms, peak responses
of the structure and the T M D are

where
844 Y. FUJINO AND M. ABE

0 0.05 0.10 0.15 0.20


(a) ST

AT

0 0.05 0.10 0.20


(b) 0”5 5T

Figure 5. Efficiency of TMD subjected to random input as a function of tT (p = 0.01; ls = 0.01; y = ,/-2/(1 + p), 095):
(a) equivalent damping req;
(b) non-dimensionalizedamplitude of TMD AT

In general, the damping of the structure is much smaller than the damping ratio of the TMD. Then, after
some manipulation, it can be shown that the peak value of mode 1 is larger than that of mode 2.
Equivalent damping from the analogy of a SDOF is defined as
1
res = 2msoiH; - rs (60)

Substituting equation (59a) into equation (60),the equivalent additional damping due to the T M D is
TUNED MASS DAMPERS 845

0.04 I I r
-.I.-I-. MATCHED EXPRESSION
_ - - _" E D EXPRESSION

0.03

0.02

0.01

0
0.8 0.9 1.0 1.1 I .2
(a) Y

Ar
12 r I I I

10 - /~=0.005

I I I
0
0.8 0.9 1.0 1.1 1.2
(b) Y
Figure 6. Efficiency of TMD subjected to random input as a function of y (p = 0@05, 0.01, 0.02; ts= 0.01;
tT = Jp(l + 3p/4)/((1 + p)(1 + p / 2 ) } ) :(a) equivalent damping teq;
(b) non-dimensionalized amplitude of TMD, A,

obtained as

The equivalent damping, teq, approaches zero when the system is detuned. As for the case of random
excitation, replacing w,/os (in equation (61)) with 1, one obtains a matched expression as
846 Y.FUJINO AND M. ABE

The non-dimensionalized amplitude is defined as AT = H:/Hf..


Then,

It is shown in equation (63) that the non-dimensionalized amplitude approaches zero when the system is
detuned (fl is large); hence, equation (63) does not require a matched expression.
Now, we consider the optimization of the TMD. It is assumed that the two peaks of frequency response of
the structure are equally small when TMD is operating optimally.6 From equations (59a) and (59c), it is
shown that both peaks have the same value when fl is equal to zero. Therefore, the optimal tuning ratio is
Y = 1/(1 + P) (64)
This agrees with the exact solution6 when the damping of the structure is negligibly small.
The effects of TMD damping on peak response of the structure are examined under optimal tuning
conditions (equation (64)). The damping of the structure is assumed to be negligibly small. It can be shown
after some manipulation that the structural response has two equal peaks when <: < p:

In this case, the equivalent damping, Ceq, becomes


4
eq
=-/-tT
2
4p- 3t?
p + 35:
When tJ.f 2 p, both modal frequencies coincide, as is shown in Figure 2(a) and

The optimal damping, tT,is obtained by taking the derivative of equation (66) with respect to tT:
i,=+5 - 1

The exact solutions corresponding to these are6

In the case of p = 0.01, an exact value of 0.0609 compares with an approximate value of 0.0642 for optimal
damping of the TMD, cT, while an exact value of 0.0354 compares with 00357 for equivalent damping, teq.
For p = 0.02, the errors in tTand tcqare 9.4 and 1.4 per cent, respectively. The use of equations (68) and (69)
may be limited to p less than 0.02.
These perturbation solutions are compared with the exact solutions in Figures 7 and 8. Both figures show
that agreement is good.
However, there is some discrepancy under optimally tuned and optimally damped conditions in
Figure 7(a). This is because the peak value of the frequency response occurs not at one of the modal
frequencies but somewhere between these when the two modal frequencies are very close, and furthermore
TUNED MASS DAMPERS 847

scq
0.06 I I

0.04

0.02

0
0 0.05 0.10 0.15 0.u)
(a) ST
AT
12 1 I I 1

Figure 7. Efficiency of TMD subjected to harmonic input as a function of tT ( p = 0.01;ts= 0.01; y = 1/(1 + p), 0.95):(a) equivalent
damping Ceq; (b) non-dimensionalized amplitude of TMD, AT

the corresponding peak values are almost equal. In this case, equivalent damping may be obtained by
substituting ma( = (wl + m2)/2) into equation (41a) and taking the first-order terms:

Equation (72) for y = 1/(1 + p) is also shown in Figure 7(a) and gives a better approximation in a range
around the optimal damping. However, the error in equation (66) is not as disturbing when the TMD is
slightly detuned as shown in Figure 7(a) for the case where y = 0.95. If the tuning ratio and damping are very
close to the respective optimal values, one may take smaller values of equations (67) and (72) to give a more
accurate approximation.
848 Y. FUJINO AND M. ABE

0.06 r I I 1

I- PERTURBATION
EXACT

0.04

0.02

0
0.8 0.9 1.o 1.1 1.2
(a) Y

10 -
8 -

n
"
I
I
1
I
0.8 0.9 1.0 1.1 1.2
(b) Y
Figure 8. Efficiency of TMD subjected to harmonic input as a function of y ( p = 0.005, 0.01, 0.02; tS = 0.01; tT =
(a) equivalent damping tea;
(b) non-dimensionalized amplitude of TMD, AT
Jm):
Warburton and Ayorinde4 showed numerically that the presence of the damping of the structure, &,
affects the optimal tuning ratio and the optimal damping of the T M D slightly. The formulas (equations (44),
(68) and (69))can be used only for a small value of ts, say < 0.02.

STABILITY UNDER SELF-EXCITED FORCES


TMD-structure under negative damping
The TMD has often been used to suppress self-excited vibration such as wind-induced galloping and
flutter. In this case, when the amplitude is not large, the self-excited effect can be modelled as negative and
TUNED MASS DAMPERS 849

linear damping. Stability of the system under a self-excited force is examined by adjusting the sign of modal
damping or the real parts of the complex eigenvalues of the TMD-structure system, and the stability
boundary is given when modal damping becomes zero. The T M D has to be designed theoretically such that
the real parts of both eigenvalues are nearly equal to zero. This indicates that a perturbation point different
from the previous case should be determined.
First, the condition that the real parts of both eigenvalues are zero is obtained. In that case, the eigenvalues
are expressed as
ill = ia (734
i12 = ib (73W
where a and b are real numbers. Then, the characteristic equation is
A4 + (a2 + b2)A2+ a 2 b 2= 0 (74)
Comparing the coefficients in equation (74) with those in equation (8'), and from the condition permissible
that all coefficients are real, we obtain

t* = o (79)
where ts is the maximum permissible negative damping (including the inherent damping of the structure) for
stability.
The results in equations (75H79) are the same as the relations in the optimal T M D for negative
In this case, a = b = w* and the characteristic equation has duplicate roots.
Taking the first-order terms with respect to ts and m, the frequency ( w * ) of the centre of perturbation is
approximated as w* x w,. For the damping coefficients (t*), similarly, this is expressed as t* GZ 5, since t, is
0 ( ~and very~small.
~ / ) Therefore, equation (23) with equations (78) and (79) can be used as the perturbation
point in this case also. The tuning parameter is defined on the basis of equation (75) as

instead of equation (24).

Stability boundary
The stability of a TMD-structure system under a self-excited force is determined by the sign of its two
modal damping ratios. If both the modal dampings are positive, the system is stable; otherwise, the system is
unstable.
The modal damping can be evaluated from the perturbation solution given by equation (30), where p is
replaced with p'. The necessary and sufficient condition for stability is that both modal damping ratios are
positive, which is equivalent to the smaller modal damping being greater than zero. Considering wi x warthis
yields
ta-+ILI 20 (8 1a)
850 Y.FUJINO A N D M. ABE

Taking the square of equation (81a) and using equation (32), one obtains

This is the stability criterion. The boundaries of tscomputed by equation (81) and by complex eigenvalue
analysis are compared in Figure 9 where the ordinate is the negative value of the maximum permissible
negative damping for stability. This figure shows that this simple first-order equation (81a and 81b) is very
accurate.
Under the optimal TMD damping ratio given in equation (76), the negative damping ratio corresponding
to the stability boundary is computed by equation (81b) and is shown in Figure 10 in which the exact values
are also plotted. Very good agreement can be seen even for the detuned condition, say y x 0.8 or 1.2.
When the system is completely detuned, the T M D is not effective in suppression at all. In this case the
stability condition is obviously ts > 0. When P’ is large, equation (32) can be approximated by

Hence, recognizing that tT> 0 and ts 6 0, one can obtain the stability condition

This indicates that equation (81 b) is a matched expression also.


In order to examine the stability of a TMD-structure system subject to self-excited force, complex
eigenvalue analysis or Routh’s criterion has been commonly used. Using the perturbation solutions, a very
simple and explicit formula is derived.

Optimal parameters
With a given value of ta,the optimal condition is fd = 0 from equation (81a). From equation (32), this
indicates that either (d = 0 or P’ = 0 under the condition p - > 0. As explained before, fl‘ = 0 is the

0.06 1 I I

0.04

0.02

a
0.05 0.10 0.15 0.20
5T
Figure 9. Efficiency of TMD under negative damping of structure as a function of <T ( p = 001;y = l/m
095) ,
TUNED MASS DAMPERS 85 1

-5s
0.06 I I I

0.04 - -

0.02 -

-0.8 0.9 1.0 1.1 1.2

Figure 10. Efficiency of T M D under negative damping of structure as a function of y ( p = 0.01; TT = J{1 - ( l / m ) ) / 2 )

necessary condition for optimality, i.e.

This is the same as the corresponding exact solution shown in equation (75).
In the following, the optimal damping of tSis obtained.
First consider the case when the tuning ratio is optimal (p’ = 0). Under the condition p - 5: 2 0, rd =0
when B’ = 0. Hence, 5, 2 0 is the stability condition. It is expressed as

- 5 s G 5T (5-r G Jim (85)

If p - (2 < 0, fd is not equal to zero (see equation (32)); the two modal frequencies are therefore different.
The stability condition that the smaller damping is greater than zero leads to 5, - , . / G I 2 2 0, i.e.

From equations (85) and (86), it can be shown that the optimal damping of the TMD is

and that the maximum allowable negative damping in this case becomes
7

J P
5s= -2
Both equations (87) and (88) are good approximations of the exact expressions shown in equations (76) and
rs
(77), respectively; the errors in tTand are 0.7 and 2.5 per cent for p = 0.02.
852 Y. FUJINO AND M. ABE

CONCLUSIONS
This study can be summarized as follows:
(1) The dynamic characteristics of TMD-structure systems, i.e. modal frequencies and damping, are
expressed in simple and explicit forms by modifying the perturbation technique proposed by Igusa and Der
Kiureghian.'.

Table I. Formulas on T M D

Free vibration (modal damping), =

For jth mode

j = 1,2; upper sign for j = 1 and lower for j = 2

Random excitation, /3 = P1

Non-dimensionalized T M D amplitude: AT = ta

Harmonic excitation, P = P1

1
Non-dimensionalized T M D amplitude: AT =

Self-excitation, B = P2
For stability of the structure: ta - ~IL 2 0I or
t: -i{Jh + P2 - + 4P2tdZ - (P + P2 - t:)} 2 0
Note: tS ( < 0) means negative self-excited damping of the structure including the inherent structural
damping

where

p: mass ratio; 5: damping ratio; o:natural frequency; subscripts S: structure; T: TMD; i = ,/ - 1


Table 11. Optimal parameters of T M D
Perturbation
Exact solutions
Free vibration Y 1 1
1+P
5T

J& J7r
Jsr2 Jsr2
Random excitation* Y ~
1
1+P
5T
4
2

A?

Harmonic excitation* Y -
.
1
1+P
5T

-
1 JL
2 2+p

Self-excitation+ Y 1 1
fi J1+1
5T
J;I 2

5s

Note: tS is assumed to be zero.


*Displacement is the quantity to be controlled.
'For self-excitation, ts means maximum allowable negative damping.
y : optimal tuning ratio (wT/ws); tT:optimal damping ratio of TMD, te,,;equivalent damping: AT: non-dimen-

sionalized amplitude of TMD; p: mass ratio (mT/ms); ct =


854 Y. FUJINO AND M. ABE

(2) The effectiveness of the T M D under the non-optimal as well as the optimal T M D parameters for free,
harmonic, random and self-excited vibrations of the structure is expressed as closed-form formulas using the
perturbation solutions.
(3) The formulas derived in this study (summarized in Table I) predict well the TMD’s efficiency and its
amplitude.
(4) Using the perturbation solutions, procedures of optimizing the T M D parameters for various types of
vibrations are explained in an explicit manner and their optimal values are obtained as seen in Table 11.
Finally, it should be noted that the formulas are derived under the assumption that the mass of the T M D is
small. Although the limitation of the formulas is not extensively investigated, they can be used for the mass
ratio p < 0.02. Although the formulas for expressing the T M D effectiveness include the damping of the
structure tS,most of the optimal TMD parameters for various types of input are derived for 5, = 0. These
should be used with care when 5, is not small, say > 0.02.

REFERENCES
I , T. Igusa and A. Der Kiureghian, ‘Dynamic characterization of two-degree-of-freedom equipment-structure systems’, J. eng. mech.
ASCE 111, 1-19 (1985).
2. T. Igusa and A. Der Kiureghian, ‘Dynamic analysis of multiple tuned and arbitrarily supported secondary systems’, Report No.
UCB,/EERC-83/073,Earthquake Engineering Research Center, University of California, Berkeley, 1983.
3. B. M. Pacheco and Y. Fujino, ‘Approximate explicit formulas for complex modes of a two-degree-of-freedom (2DOF) system’,
Struct. englearthquake eng. JSCE ql), 191-200 (1989).
4. G. B. Warburton and E. 0. Ayorinde, ‘Optimum absorbers for simple systems’, Earthquake eng. struct. dyn. 8, 197-217 (1980).
5. A. D. Aleksandrov et al. Mathematics-its Contents, Methods, and Meanings; Vol. 1, 266-270, MIT Press, Cambridge, MA, 1963.
6. J. P.Den Hartog, Mechanical Vibrations, 4th edn, McGraw-Hill, New York, 1956.
7. G. B. Warburton, ‘Optimum absorber parameters for various combinations of response and excitation parameters’, Earthquake eng.
struct. dyn. 10, 381-401 (1982).
8. H. Horiuchi, N. Fujisawa and N. Tsumura, ‘Practical formulas for estimating design parameters of TMD’, J. struct. eng. J S C E
(in Japanese) 37A, 781-788 (1991).
9. Y. Kamiya and M. Takano, ‘A study on a high-speed intermittent motion mechanism’, Precision machinery (in Japanese) 439).
8 1-89 ( 1 977).
10. M. D. Rowbottom, ‘The optimization of mechanical dampers to control self-excited galloping oscillations’, J . sound uib. 75, 559-576
( 1 98 1).
1 1 . A. H. Nayfeh. Perturbation Methods, Wiley, New York, 1973, pp. 110-114.
12. S. H. Crandall and W. D. Mark, Random Vibration QfMechanical Systems, Academic Press, New York, 1963.
13. T. lkeda and T. loi, ‘On suppression of self-excited vibration by means of the T M D , Proc. J S M E (in Japanese) 43,2551-2556 (1973).

Вам также может понравиться