You are on page 1of 10

Lei Zuo

e-mail: lei.zuo@abbott.com
Phone: 847-935-0086
Abbott Laboratories,
Bldg. AP52S,
200 Abbott Park Road,
Abbott Park, IL 60064-6212

Samir A. Nayfeh
e-mail: nayfeh@mit.edu
Phone: 617-253-2407
Department of Mechanical Engineering,
Massachusetts Institute of Technology,
Room 3-461A,
77 Massachusetts Avenue,
Cambridge, MA 02139

The Two-Degree-of-Freedom
Tuned-Mass Damper for
Suppression of Single-Mode
Vibration Under Random and
Harmonic Excitation
Whenever a tuned-mass damper is attached to a primary system, motion of the absorber
body in more than one degree of freedom (DOF) relative to the primary system can be
used to attenuate vibration of the primary system. In this paper, we propose that more
than one mode of vibration of an absorber body relative to a primary system be tuned to
suppress single-mode vibration of a primary system. We cast the problem of optimization
of the multi-degree-of-freedom connection between the absorber body and primary structure as a decentralized control problem and develop optimization algorithms based on the
H2 and H-infinity norms to minimize the response to random and harmonic excitations,
respectively. We find that a two-DOF absorber can attain better performance than the
optimal SDOF absorber, even for the case where the rotary inertia of the absorber tends
to zero. With properly chosen connection locations, the two-DOF absorber achieves
better vibration suppression than two separate absorbers of optimized mass distribution.
A two-DOF absorber with a negative damper in one of its two connections to the primary
system yields significantly better performance than absorbers with only positive
dampers. DOI: 10.1115/1.2128639

Introduction

The design of a single-degree-of-freedom SDOF tuned-mass


damper TMD, or dynamic vibration absorber DVA, to attenuate vibration of a single mode of a primary system under various
conditions has been studied extensively e.g., 15. To enhance
the effectiveness and robustness of TMD systems, multiple SDOF
TMDs with frequencies tuned in the neighborhood of a mode of a
primary system have been proposed by Xu and Igusa 6 and
optimized by Zuo and Nayfeh 7. Multiple SDOF TMDs have
been used to damp more than one mode of a primary system by
tuning each TMD to an individual mode of interest in the primary
system 8,9.
Whenever an absorber is attached to a primary system, there is
potential for utilization of motion in more than one degree of
freedom of the absorber body relative to the primary system.
Dahlbe 10 numerically optimized a continuous two-segment
cantilever beam for suppression of SDOF vibration and found the
two-segment beam to be more effective than a SDOF TMD of the
same mass. Recently, Zuo and Nayfeh 1113 and Verdirame and
Nayfeh 14 have optimized the stiffness and damping in multidegree-of-freedom MDOF connections between a rigid body
and primary structure to damp as many as six modes of a structure. In these studies, one mode of vibration of the body relative to
the structure is tuned to each mode of interest in the primary
system.
In this paper, we propose that more than one mode of vibration
of a body relative to a primary structure be tuned to one natural
frequency of a primary system. Such an absorber is often easier to
construct than an SDOF TMD or multiple SDOF TMDs because
of the reduced need for guidance, and can achieve enhanced performance. We cast the design of MDOF TMD systems as decenContributed by the Technical Committee on Vibration and Sound of ASME for
publication in the JOURNAL OF VIBRATIONS AND ACOUSTICS. Manuscript received June 3,
2003; final manuscript received April 1, 2005. Assoc. Editor: D. Dane Quinn.

56 / Vol. 128, FEBRUARY 2006

tralized optimal control problems with static output feedback,


where the feedback gain is a block diagonal matrix composed of
the stiffness and damping parameters to be optimized.
Based on this formulation, we minimize the system response to
random excitation by adapting a gradient-based H2 optimization
technique based on Lyapunov equations. The optimal parameters
are obtained and presented in dimensionless form to be useful for
design. With properly chosen locations for the springs and dampers, the 2DOF TMD achieves better vibration suppression than
two separate TMDs with optimized mass distribution. Next, to
minimize the steady harmonic response under sinusoidal disturbances, we propose an algorithm based on decentralized H optimization, and find that the 2DOF TMD again offers performance
better than the conventional SDOF TMD or two separate TMDs.

Problem Formulation

Figure 1 shows the configuration of a MDOF TMD attached to


a SDOF primary system. We take the 2DOF TMD as an example;
the general multi-DOF TMD can be handled similarly. The primary system has a natural frequency s = ks / ms and damping
ratio s = cs / 2ksms and is subject to a base excitation x0, an external disturbance force f, or both. The absorber body has two
planar degrees of freedom, translation xd and rotation d. Its mass
is md and the rotational inertia about its center of mass is Id
= md2, where is the radius of gyration. The absorber is connected to the primary system at distances d1 and d2 from its center
of mass via springs and dashpots. Our goal is to design the parameters k1, c1, k2, and c2 as well as the locations of the connections d1 and d2 in order to minimize the response of the primary
system. We consider first the case where d1 = d2 = d, then in Sec.
4.5 discuss the case where d1 d2. The effect of absorber mass md
and rotary inertia Id will also be explored in Sec. 4.
By taking the springs and dashpots as elements which feed back
locally the relative displacements and velocities, Zuo and Nayfeh
15 have cast the optimization of a wide class of mechanical

Copyright 2006 by ASME

Transactions of the ASME

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm




md/2

x1

0 0 0

x1

0 0 0

x1

md/2 md/2

x2

x2

xs

+ 0 0 0
0 0 ks

x2

ms

+ 0 0 0
0 0 cs

md/2
0

xs

0
0
1
1
0
= 0 x0 + 0 x0 + 0 f + d2/2 d2/2
1
1
1
cs
ks

u1

u2

xs

or
M p p + C p p + K p p = B px0 + Bvx0 + Bd f + Buu

where p = x1 , x2 , xs. From Eq. 6, we see that the performance


of the TMD system does not depend on the rotational inertia Id.
Rather, it depends on the ratio of the radius of gyration to the
distance d from the mount points to the center of mass of the
absorber.
Defining the state variables of the system as

Fig. 1 2DOF TMD for one mode of a primary system

x=
systems as zeroth-order decentralized control problems. In this
way, the role of the spring and dashpot pairs are replaced by the
control-force vector u1 , u2, where the prime denotes the matrix
transpose, as shown in Fig. 2. The control forces in this case are
given by
u1 = k1x1 xs + c1x1 xs
u2 = k2x2 xs + c2x2 xs

p
p M 1
p B vx 0

we can write the governing equations in first-order form as


9

x = Ax + B1w + B2u
where w = f , x0 and
A=

where x1 and x2 are the displacements of the absorber in the direction of xs at the connection locations.
The equations governing the vibration of the coupled system
with small absorber rotation take the form
mdxd = u1 u2

B1 =

1
M 1
p Kp M p Cp

M 1
p Bv

1
1
M 1
p Bd M p B p C p M p Bv

B2 =

10

12

M 1
p Bu

The cost output can be taken as the absolute or relative displacement, velocity, or acceleration of the primary system, which
can be expressed in the form
z = C1x + D11w + D12u

Id d = u1d u2d

msxs + csxs + ksxs = csx0 + ksx0 + f + u1 + u2

Noting that xd = x1 + x2 / 2, d = x2 x1 / 2d, and Id = md2, we


write the above governing equations in matrix form as

11

13

For example, if the cost output is the displacement response of the


primary system, we write the cost as z = xs = C1x, where C1
= 0 , 0 , 1 , 0 , 0 , 0.
To complete the state-space description, we rewrite the control
force given by Eqs. 1 and 2 as a static feedback gain F multiplied by the measurement output y. That is,
u=

k1 c1 0
0

0 k2 c2

def

y = Fy

14

where y is given by
y = x1 xs,x1 xs,x2 xs,x2 xs = C2x + D21w + D22u
15
The form of C2 follows from the definition of the state given by
Eq. 8 and the matrices D22 = 0 and
D21 = C2

Fig. 2 Control formulation of a passive 2DOF TMD

Journal of Vibration and Acoustics

M 1
p Bv

16

Equations 9, 13, and 15 cast the design of the MDOF


TMD system as a decentralized control problem, as shown by the
block diagram of Fig. 3.
Based on this formulation, we use decentralized control techniques to directly optimize the stiffness and damping coefficients
of the springs and dampers to achieve performance measured by
z under the disturbance w. Decentralized H2 optimization minimizes the output variance under random excitation, and decentralized H optimization minimizes the worst-case response magniFEBRUARY 2006, Vol. 128 / 57

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

minH22 = traceB1 + B2FD21KB1 + B2FD21


F

22

subject to
KA + B2FC2 + A + B2FC2K + C1 + D12FC2C1 + D12FC2
23

=0

Fig. 3 Block diagram of the primary system and MDOF TMD


viewed as a system with decentralized control

tude under harmonic excitation 15. In the following section, we


will briefly review our method for decentralized H2 optimization,
and then present the results of optimization of a 2DOF TMD
under random excitation in Sec. 4. In Sec. 5, an algorithm for
decentralized H optimization is proposed along with some results for the 2DOF TMD under harmonic excitation.

where F is in the given block-diagonal form. In practice, we require that the parameters be nonnegative. So we replace F by
FdFd, where the symbol denotes multiplication entry by
entry.
Employing the Lagrange multiplier method 17 and matrix calculus, we obtain the gradient of the square of the H2 norm with
respect to Fd where FdFd is the controller gain in the form 11

H22
D12FdFdC2 + D12
C1 + B2KLC2
= 4D12
Fd

24

Fd
+ B2KB1 + B2FdFdD21D21

25

where the observability Grammian K and Lagrange multiplier matrix L are obtained by solving the two decoupled Lyapunov matrix
equations with a given matrix Fd:
KA + B2FdFdC2 + A + B2FdFdC2K + C1
+ D12FdFdC2C1 + D12FdFdC2 = 0

Review of Decentralized H2 Optimization

LA + B2FdFdC2 + A + B2FdFdC2L + B1

The system H2 norm is defined as the energy of the system


impulse response:
H22 =

thzwtdt
tracehzw

1
2

jHzwjd
traceHzw

17

2
zz
= lim E
T

1
T

ztztdt = H22

x = Acx + Bcw
19

z = C cx + D cw
is infinite if Ac is unstable or Dc is nonzero. Otherwise
H22 = traceBc KBc

20

where K is a symmetric matrix known as the observability Grammian which satisfies the Lyapunov equation
Ac K + KAc + Cc Cc = 0
Under the decentralized feedback u = Fy, the closed-loop system w z shown in Fig. 3 is given by
Ac Bc
Cc Dc

A + B2FC2 B1 + B2FD21
C1 + D12FC2 D12FD21

21

To obtain a finite H2 norm, D11 + D12FD21 must be zero. Then the


decentralized H2 optimal control problem becomes
58 / Vol. 128, FEBRUARY 2006

27

Fd =

k/2 c/2
0

k/2 c/2

28

where k and c are the optimal stiffness and damping of an SDOF


TMD of the same mass ratio md / ms obtained from the analytical
tuning formulas developed by Den Hartog 1 or Asani et al. 5.

18

where zz is the RMS value of z. Therefore, minimization of the


H2 norm minimizes the RMS response of the output zt under
wide-band random excitation wt.
Based on the definition given by Eq. 17, it can be shown that
the H2 norm of an LTI system can be evaluated by solving a
Lyapunov equation 16, as summarized in the following: The H2
norm of the LTI system from w z given by

+ B2FdFdD21B1 + B2FdFdD21 = 0

We can therefore adapt a gradient-based optimization method,


such as the FBGS quasi-Newton method 18, to solve for the
matrix Fd efficiently and then obtain the stiffness and damping
parameters from controller gain FdFd. A good initial guess for
Fd is

where hzw is the system impulse-response matrix and Hzw is the


system transfer matrix. Another interpretation of system H2 norm
is the asymptotic value of output variance under unit white noise
input. That is, if Ewt = 0 and Ewtwt + = I, then

26

The H2 Optimal 2DOF TMD

Using gradient-based decentralized H2 optimization, we obtain


the optimal parameters of k1, k2, c1, and c2 as a function of the
ratio / d. We consider initially the case where d1 = d2 = d, and
later the case where d1 d2. As for the SDOF absorber 5, the
optimal H2 norm is proportional to the square-root of the natural
frequency s of the primary system, and we therefore normalize

the H2 norm by s in the following. In this section, we give a


comprehensive study of the 2DOF TMD, taking the excitation as
the base motion x0w = x0 and the performance index as the RMS
value of the displacement xs.
4.1 Effect of / d. As pointed out following Eq. 6, the performance of the TMD system depends on the ratio of the absorbers radius of gyration to the distance d from the mount points to
the center of mass of the absorber. The solid line in Fig. 4 shows
the normalized optimal H2 norm as a function of / d for
= md / ms = 5% and s = 0.
On this plot, / d = or d = 0 corresponds to the optimal

SDOF TMD, which attains H2 = 2.108s. If / d = 1, we obtain


a system equivalent to two separate SDOF TMDs with m1 = m2
= md / 2, as shown in Fig. 5. And / d = 1 / 3 is the case of a
uniform-density bar supported at its two ends. But the optimal / d
for = 5% and s = 0 is 0.780, which yields H2 = 2.0189s. The
trend is similar if the primary system is lightly damped, as indiTransactions of the ASME

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 4 The normalized system H2 norm as a function of / d


with = 5% for s = 0 solid and for s = 1% dashed

cated by the dashed line in Fig. 4, where the optimal / d for


= 5% and s = 1% is 0.777. The frequency responses are compared
in Fig. 6.
Figure 4 also shows that the optimal H2 norm at / d = 0 is
smaller than that of a SDOF TMD / d = , indicating that a
2DOF TMD without rotary inertia can achieve better vibration
suppression than the SDOF TMD. This can be explained by comparison of the impedance of the 2DOF TMD with no rotary inertia
as sketched in Fig. 7a to that of the SDOF TMD sketched in Fig.
7b. The latter is given by a second-order transfer function,
whereas the former is given by a third-order transfer function,
which when optimized accounts for the improvement in
performance.
4.2 Comparison to Two TMDs With Optimal Mass
Distribution. As mentioned previously, the performance of the
optimal 2DOF TMD exceeds that of two separate TMDs with
equal masses and optimal springs and dampers. The authors 7
have also obtained the optimal H2 norm of two SDOF TMDs with
unequal mass distribution as plotted in Fig. 8 for = 5% and s
= 0.
In this curve, m1 / m1 + m2 = 0.5 corresponds to the case of
equal masses, and m1 / m1 + m2 = 1 corresponds to the case of a
single SDOF TMD. The optimal mass distribution is m1 / m1

Fig. 5 Two separate SDOF TMDs for one mode of a primary


system

Journal of Vibration and Acoustics

Fig. 6 Frequency responses of the H2 optimal TMD system for


= 5% and s = 0: / d = 0.780 solid, two separate TMDs dash,
uniform bar supported at two ends dash-dot, and SDOF TMD
dot

+ m2 = 53.35% with the corresponding H2 = 2.0426s, which is


only 0.02% less than that of two SDOF TMDs with equal masses,
and is larger than the H2 norm of the 2DOF TMD with optimal
/ d. Thus the performance of the 2DOF TMD generally exceeds
that of two separate SDOF TMDs, even if the mass is distributed
optimally among them.
4.3 Negative Damping. Figure 4 shows two sharp corners in
the dependence of the optimal H2 norm on / d. This is because at
these values of / d the constraint ci 0 become active. That is,

Fig. 7

a Two-DOF TMD with / d = 0 and b SDOF TMD

Fig. 8 Normalized system H2 norm versus mass distribution


among two SDOF TMDs for = 5% and s = 0

FEBRUARY 2006, Vol. 128 / 59

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Comparison of designs obtained with and without the constraint that the parameters
remain nonnegative for / d = 0.2, = 5%, cs = 0, ms = 1, and ks = 1

the constraint is active for / d greater than about 1.25 or less than
0.751. An interesting set of designs are found beyond these two
values if we do not constrain the parameters to be positive.
Take as an example the case of / d = 0.2, = 5%, and s = 0
with the stiffness ks and mass ms of the primary system normalized to one. We optimize the parameters of stiffness and damping
using the algorithm of Sec. 3 which guarantees that ki 0 and
ci 0 and list as the first entry in Table 1 the optimal parameters,
the poles of the coupled absorber and primary system, and the
poles of the absorber mounted to ground. With the parameters
constrained to be nonnegative, we see that for this value of / d,
one of the modes of the absorber is tuned close to s, but the other
is over damped. The Bode plot of xss / x0s is shown as a dashed
line in Fig. 9.
If we modify the algorithm to allow negative stiffness and
damping, we obtain the results shown as the second entry of Table
1 and indicated by the solid line in Fig. 9. From Table 1, we see
that although there is a negative damper in the system, the coupled
absorber and primary system are stable. Moreover, the performance of this system is significantly better than that of the system
with the dampers constrained to be non-negative. Such a design
could readily be implemented in an active absorber, though there
would be potential for instabilities in the presence of modeling
uncertainties.
It may be suggested that a different configuration of springs and
dampers could attain the same performance without resort to
negative dampers. For this particular example, it can be shown
algebraically that it is not possible to find positive values of k1, k2,
c1, and c2 that yield the same dynamics, even if their locations are
allowed to vary independently. However, it may be possible to
obtain improved performance in passive systems by considering a
somewhat more general configuration e.g., including torsional
springs and dampers.

4.4 Optimal Parameters. The optimal normalized modal frequencies undamped 1 / s and 2 / s, and modal damping ratios 1 and 2 of the TMD subsystem attached to ground for
= 5% and s = 0 are shown in Fig. 10 as functions of / d. As we
have observed in the foregoing, for small / d, we can only tune
one of the natural frequencies of the TMD close to s and the
other mode is over damped.
To provide results more readily used for design, in Fig. 11 we
give the optimal parameters in the form of dimensionless fre 1 / s and
2 / s and damping factors 1 and 2
quencies
1 = k1/md,

= c /2m k ,
1
1
d 1

2k2/md

= c /2m k
2
1
d 1

from which the optimal stiffness and damping can be constructed.


These are convenient dimensionless parameters, but they do not
correspond to the resonant frequencies or damping ratios of the
modes of the TMD subsystem.
Next, we minimize the H2 norm of xs / x0 as the mass ratio is
varied. Figure 12 shows the optimal / d for different values of .
The achieved minimal H2 norms of 2DOF TMD system with
optimal / d are shown in Fig. 13 and compared with those of the
SDOF TMD and two separate TMDs. The corresponding optimal
tuning of stiffness and damping for the optimal / d are given in
1 / s,
2 / s,
Fig. 14 in terms of the dimensionless parameters
, and . Figures 13 and 14 can be used to design the optimal
1
2
2DOF TMD for a given mass ratio. Comparing the curves in Fig.
13, we see that the optimal 2DOF TMD whose mass is 5% of that
of the primary system provides roughly the same performance as
the optimal SDOF absorber whose mass is 6% of that of the
primary system.

Fig. 9 Bode plots of xss / x0s for / d = 0.2, = 5%, and s = 0: original system
dot, optimized with nonnegative constraint dashed, optimized without nonnegative constraint solid

60 / Vol. 128, FEBRUARY 2006

Transactions of the ASME

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Optimal modal frequencies and damping ratios of the 2DOF TMD subsystem as a function for / d for = 5% and s = 0

4.5 Asymmetric Connection Locations. Thus far, we have


maintained the symmetry of the locations of the connections between the TMD and primary system by holding d1 = d2 = d. In this
subsection, we relax this constraint and examine the H2 optimal
2DOF TMD allowing d1 and d2 to vary independently. The design
method is the same as before: For a given d1 and d2, we use
decentralized control techniques to optimize the stiffness and
damping values.
The achievable performance is a function of / d1 and / d2.
Figure 15 shows the minimal normalized H2 norm for = 5% and
s = 0 attained by optimizing k1, k2, c1, and c2 for a range of / d1
and / d2. From this figure, we see that the optimal / d1 = / d2
= 0.780 for symmetric connections is a saddle point and that we
can attain better performance by allowing d1 and d2 vary independently. The contour plot indicates that the global minimum for a
mass ratio of 5% is attained when / d1 tends to zero and / d2
is approximately 3. To avoid a singular computation, we take
/ d1 = 0.01 and find the corresponding best / d2 to be 3.055. The
corresponding optimal parameters and closed-loop poles are
shown in Table 2 and compared with those of the optimal symmetric case. The corresponding frequency responses are compared
in Fig. 16.

Decentralized H Optimization

We now turn to the problem of minimizing the worst-case response to sinusoidal inputs of the primary system with a 2DOF
TMD. For single-input-single-output LTI systems, the H norm is

the peak of the magnitude of the frequency-response. For multiinput-multi-output LTI systems, it is the supremum of the largest
singular value over all frequencies:
2
jHzwj 29
H2 = sup max
Hzwj = sup maxHzw

where max and max respectively, denote the largest singular


value and eigenvalue of their arguments. We see that the H norm
is the steady magnitude response under worst-case sinusoidal excitation. Note that Den Hartogs design formulas for an SDOF
TMD mounted to a SDOF primary system yield a close approximation to the H optimal design.
For an SDOF TMD attached to multi-DOF primary system, the
efficient algorithms for optimization of centralized static output
feedback proposed by El Ghaoui et al. 19 or by Geromel et al.
20 can be used. However, as we have seen in Sec. 2, the optimization of an MDOF TMD is equivalent to decentralized optimal
control with static output feedback.
Decentralized H optimization has been investigated by many
researchers in the controls community, and various techniques
such as alternative linear matrix inequalities LMIs 21, homotopy 22,15, and LMI iteration 23 have been proposed. But
none of these algorithms can be guaranteed to converge to a local
optimum or a stationary point. Our experience is that these methods generate sequences that decrease quickly when the controller
gain is far from the optimum, but become very inefficient when
the sequences come close the optimum. In the engineering application of structural optimization, a commonly used frequency-

Fig. 11 Optimal parameters of the 2DOF TMD versus / d for = 5% and s = 0

Journal of Vibration and Acoustics

FEBRUARY 2006, Vol. 128 / 61

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 15 Contour map of normalized minimal H2 norm for various values of / d1 and / d2 for = 5% and s = 0

Fig. 12 Optimal ratio / d of the radius of gyration to the connection spacing d versus mass ratio for s = 0

domain optimization method is based on the evaluation of a transfer function at discrete frequencies e.g., 24,10. To capture the
maximal magnitude response, the frequencies have to be closely
spaced and hence the method is also computationally inefficient.
We therefore develop a method with better efficiency for minimization of the peak of the frequency response. It is based on
these two points: 1 The peak of the frequency response, or the
H norm of a LTI system, can be computed very efficiently using
-iteration. 2 Finite differences can be used to approximate the
gradient, and the computational efficiency should be much better
than direct search if the objective function is easy to evaluate. The
-iteration algorithm is based on this fact: For the stable LTI
system of Eq. 19 assuming that Dc = 0, H if and only if
the Hamiltonian matrix

Fig. 13 Optimal H2 norm of xs / x0 versus the mass ratio for


s = 0: 2DOF TMD with optimal / d solid, two separate TMDs
dashed, and SDOF TMD dotted

Ac

1
B B
c c

1
Cc Cc

Ac

has no eigenvalue on the imaginary axis. So starting with an upper


and a lower bound we can use a bisection algorithm to calculate
the system H norm i.e., the value of that makes the above
Hamiltonian matrix have an imaginary eigenvalue. A similar
Hamiltonian matrix exists for the case Dc 0; details can be
found in the text 16. Standard routines for -iteration, such as
the Matlab function normhinf in the Robust Control Toolbox, are
available.
The algorithm for minimization of the worst-case response to
harmonic excitation is summarized as follows:

Fig. 14 Parameters of the optimal 2DOF TMD as a function of the mass ratio
for s = 0

62 / Vol. 128, FEBRUARY 2006

Transactions of the ASME

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 2 Comparison of symmetric and asymmetric connection locations for = 5%, cs = 0,


ms = 1, and ks = 1

Step 1: Choose an initial value of Fd composed of k1, k2,


c1, and c2. To ensure that we obtain nonnegative parameters, we take controller gain as FdFd.
Step 2: Evaluate the H norm J using the Matlab function
normhinf or any other -iteration routine 16 for a given
matrix Fd and Fd + Fd. Then approximate the gradient
J / Fd using the finite difference J / Fd. If the decrease of
J is small enough, stop; otherwise go to Step 3.
Step 3: Based on the gradient J / Fd, calculate a search
direction DF e.g., the FBGS quasi-Newton direction 18.
Choose a proper step size using the Armijo rule, or any other
rule 18. Update Fd with Fd + DF. Go to Step 2.
Note that the system H norm is a nonsmooth function of the
matrix Fd and therefore, at some points in the design space,
J / Fd may approximate a mixture of subgradients. Such a
mixture might not yield a true descent direction, in which case, we
can either provide a slight perturbation to the current iterate, or
choose a descent direction directly from the finite difference.
Using this algorithm, we minimize the H norm of the 2DOF
TMD system to obtain the optimal stiffness and damping parameters for harmonic excitation. Figure 17 shows the minimal H
norms achieved using a 2DOF absorber of mass ratio = 5% attached to a primary system with initial damping s = 0 and s
= 1% as a function of / d. The trend is similar to that obtained for
H2 optimal design as shown in Fig. 4. The optimal values of / d
for = 5% are, respectively, 0.751 and 0.747 with s = 0 and s
= 1%, which are close to the optimal values of 0.780 and 0.777 for
the H2 optimal designs with = 5%. As in the case of H2 optimal
design, the frequency peak of the optimal 2DOF TMD without
rotary inertia / d = 0 is smaller than that of the optimal SDOF

Fig. 16 Frequency responses of the H2 optimal 2DOF TMD


with symmetric and asymmetric connection locations for
= 5% and s = 0: / d1 = / d2 = 0.780 solid, / d1 = 0.010 and / d2
= 3.055 dashed, SDOF TMD dotted

Journal of Vibration and Acoustics

TMD corresponding to / d = . The frequency responses of the


optimal 2DOF TMD, SDOF TMD, and two SDOF TMDs are
compared in Fig. 18. It is seen that the 2DOF TMD attains much
better performance than either the SDOF TMD or two SDOF
TMDs.
In Fig. 17 the two sharp corners of the curve are again the result
of the constraint that the parameters of stiffness and damping
remain non-negative. Without taking the feedback gain F to be
FdFd, the forgoing algorithm can also be used to minimize the

Fig. 17 The effect of / d for H optimal design with = 5% for


s = 0 solid and for s = 1% dashed

Fig. 18 Frequency responses of the H optimal 2DOF TMD


design for = 5% and s = 0: optimal / d = 0.751 solid, two
separate TMDs dashed, uniform bar supported at two ends
dashed-dotted, and SDOF TMD dotted

FEBRUARY 2006, Vol. 128 / 63

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 19 Bode plots of xss / x0s for / d = 0.2, = 5%, and s = 0 obtained by H
optimization: original system dotted, optimized with non-negative constraint
dashed line, peak magnitude of 6.071, optimized without nonnegative constraint solid line, peak magnitude of 3.384

H norm while allowing the parameters to be negative. Special


attention must be paid to the step size to avoid a destabilizing
design. The frequency responses obtained with and without the
constraint that the parameters remain nonnegative are compared in
Fig. 19. As in H2 optimization, decentralized H optimization
yields significant improvement with one damper allowed to be
negative, and the total system is still stable.
To better understand the physics of the 2DOF TMD, it is useful
to examine the apparent mass of the TMD, i.e., the transfer func-

tion from the acceleration s2xs of the primary system to the force
F exerted upon it by the TMD. Figure 20 gives the normalized
apparent mass F / s2mdxs for three designs of the 2DOF TMD.
The total apparent mass denoted by the solid lines is the sum of
contributions from the two modes of the TMD denoted by dashed
lines. The total normalized apparent mass at zero frequency must
be unity and is equal to the sum of the residues of the two modes.
From Fig. 20a, we see that the two modes of the optimum

Fig. 20 Normalized apparent mass F / s2mdxs of various H optimal 2DOF absorbers obtained without parameters constraints: a / d = 1, b / d = 0.751, and c / d = 0.2. The solid
lines denote the total apparent mass and the dashed lines denote the contribution of each
absorber mode

64 / Vol. 128, FEBRUARY 2006

Transactions of the ASME

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

design with / d = 1 have equal residues and are tuned to frequencies just below and just above the resonance frequency of the
primary system. This is in contrast to the absorber obtained with
/ d = 0.751, shown in Fig. 20b, in which the residue of the first
mode is slightly larger than unity and that of the second mode is
negative. In this case, most of the vibration suppression arises
from the first absorber mode. The difference in the contribution of
the absorber modes is even more pronounced when / d = 0.2 and
a parameter is allowed to be negative, as shown in Fig. 20c.
Here the residue of the first mode is approximately three and that
of the second mode is approximately negative two. The first mode
acts to suppress vibration and is tuned close to the resonance of
the primary system, whereas the second mode magnifies vibration
and is tuned to a significantly higher frequency.

Conclusions

In this paper, we propose the use of a MDOF TMD for one


mode of primary system and show that, for a given mass, an
optimal 2DOF TMD performs better than a traditional SDOF
TMD or two separate TMDs with optimal mass distribution. We
cast the parameter optimization of MDOF TMD systems as a
decentralized control problem, where the block-diagonal controller gain is directly composed of the stiffness and damping parameters of the connections between the absorber and primary system.
Based on this formulation, we adapt decentralized H2 and H
optimization techniques to optimize the system response under
random and harmonic excitation, respectively.
First, we employ gradient-based decentralized H2 optimization
to minimize the RMS response under random excitation and provide a comprehensive study of the performance of a 2DOF TMD
attached to a SDOF primary system. Design charts for passive
TMD implementation in which all of the springs and dampers are
required to be positive are provided. We then discuss the case
where the dampers are allowed to be negative, and find that the
performance is considerably improved. This suggests that an effective reaction-mass actuator can be constructed with a 2DOF
reaction mass.
We propose an algorithm for decentralized H optimization to
minimize the peak of the frequency response under harmonic excitation. The maximal response is obtained efficiently using
-iteration and finite differences are used to approximate its gradient with respect to the design parameters. We then optimize the
2DOF absorber and find that its frequency-domain performance
is again better than that of the SDOF absorber or two SDOF
absorbers.

References
1 Den Hartog, J. P., 1947, Mechanical Vibration, McGraw-Hill, New York.
2 Brock, J., 1946, A Note on the Damped Vibration Absorber, ASME J. Appl.

Journal of Vibration and Acoustics

Mech., 684, pp. A284.


3 Snowdon, J. C., 1968, Vibration and Shock in Damped Mechanical Systems,
Wiley, New York.
4 Warburton, G., 1982, Optimum Absorber Parameters for Various Combinations of Response and Excitation Parameters, Earthquake Eng. Struct. Dyn.,
10, pp. 381401.
5 Asani, T., Nishihara, O., and Baz, A. M., 2002, Analytical Solutions to H
and H2 Optimization of Dynamic Vibration Absorber Attached to Damped
Linear Systems, J. Vibr. Acoust., 124, pp. 6778.
6 Xu, K., and Igusa, T., 1992, Dynamic Characteristics of Multiple Substructures With Closely Spaced Frequencies, Earthquake Eng. Struct. Dyn., 21,
pp. 10591070.
7 Zuo, L., and Nayfeh, S., 2005, Optimization of the Individual Stiffness and
Damping Parameters in Multiple Tuned Mass Dampers, ASME J. Vibr.
Acoust., 127, pp. 7783.
8 Kitis, L., Wang, B. P., and Pilkey, W. D., 1983, Vibration Reduction Over a
Frequency Range, J. Sound Vib., 89, pp. 559569.
9 Rice, H. J., 1993, Design of Multiple Vibration Absorber Systems Using
Modal Data, J. Sound Vib., 160, pp. 378385.
10 Dahlbe, T., 1989, On Optimal Use of the Mass of a Dynamic Vibration
Absorber, J. Sound Vib., 132, pp. 518522.
11 Zuo, L., 2002, Optimal Control With Structure Constraints and its Application
to the Design of Passive Mechanical Systems, Masters thesis, Massachusetts
Institute of Technology.
12 Zuo, L., and Nayfeh, S., 2002, Design of Multi-Degree-of-Freedom TunedMass Dampers: A Minimax Approach, in 43rd AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, AIAA 2002
1283.
13 Zuo, L., and Nayfeh, S., 2004, Minimax Optimization of Multi-Degree-ofFreedom Tuned-Mass Dampers, J. Sound Vib., 272, pp. 893908.
14 Verdirame, J., and Nayfeh, S., 2003, Design of Multi-Degree-of-Freedom
Tuned Mass Dampers Based on Eigenvalue Perturbation, in 44th AIAA/
ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, AIAA 20031686.
15 Zuo, L., and Nayfeh, S., 2002, Design of Passive Mechanical Systems via
Decentralized Control Techniques, in 43rd AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, AIAA 2002
1282.
16 Zhou, K., Doyle, J. C., and Glover, K., 1995, Robust and Optimal Control,
Prentice-Hall, New Jersey.
17 Mendel, J. M., 1974, A Concise Derivation of Optimal Constant Limited
State Feedback Gains Systems, IEEE Trans. Autom. Control, AC-19, pp.
447448.
18 Bertsekas, D. P., 1995, Nonlinear Programming, Athena Scientific.
19 El Ghaoui, L., Oustry, F., and AitRami, M., 1997, A Cone Complementarity
Linearization Algorithm for Static Output-Feedback and Related Problems,
IEEE Trans. Autom. Control, 428, pp. 11711176.
20 Geromel, J., de Souza, C., and Skelton, R., 1998, Static Output-Feedback
Controllers: Stability and Convexity, IEEE Trans. Autom. Control, 431, pp.
120125.
21 Goh, K. C., Turan, L., and Safonov, M. G., 1994, Bilinear Matrix Inequality
Properties and Computational Methods, in Proceedings of the American Control Conference, pp. 850855.
22 Zhai, G., Ikeda, M., and Fujisaki, Y., 2001, Decentralized H Controller
Design: A Matrix Inequality Approach Using a Homotopy Method, Automatica, 37, pp. 565572.
23 Cao, Y.-Y., Sun, Y.-X., and Mao, W.-J., 1998, Output Feedback Decentralized
Stabilization: ILMI Approach, Syst. Control Lett., 35, pp. 183194.
24 Kitis, L., Pilkey, W., and Wang, B., 1984, Optimal Frequency Response
Shaping by Appendant Structures, J. Sound Vib., 95, pp. 161175.

FEBRUARY 2006, Vol. 128 / 65

Downloaded 25 Jun 2009 to 129.49.32.179. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm