Optimum Design of a New Tuned

Inerter-Torsional-Mass-Damper
Passive Vibration Control for
Stochastically Motion-Excited

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
Wei-Che Tai
Assistant Professor
Structures
Department of Mechanical Engineering, The inerter is referred to as a two-terminal device that provides inertial forces proportional
Michigan State University, to the relative accelerations between its two terminals. It has been widely applied in vibra-
East Lansing, MI 48824 tion control due to its mass amplification effect. In this paper, a new inerter-based damper is
e-mail: [email protected] proposed to take advantage of the mass amplification effect, which consists of the classic
rack-pinion inerter in conjunction with a torsional tuned mass damper. Unlike any other
topologies of inerter-based dampers, the torsional mass damper is connected to the
pinion of the inerter via a rotational spring and viscous damper. As a result, the weight
of the torsional mass damper can be significantly reduced. The proposed damper is
applied to single-degree-of-freedom primary structures and a two degrees-of-freedom
structure, and the H2 optimization is conducted to obtain the optimum tuning ratio and
damping ratio analytically. When comparing the proposed damper with its counterpart
reported in the literature, the proposed damper achieves 20–70% improvement when
their weights are identical. [DOI: 10.1115/1.4045264]

Keywords: random vibration, smart materials and structures, vibration control

1 Introduction tire normal load, and handling in a quarter car model and demon-
strated the effectiveness of a prototype. Yilmaz and Kikuchi [8] com-
Historically, the idea of using springs to provide restoring forces
pared two lever-type inerter vibration isolators with mass-spring
to dynamically excited primary structures and dampers to dissipate
systems and dynamic vibration absorber-equipped springs and
the vibration energy is among the first passive vibration control strat-
found that the former admit the lowest stop-band frequency. Ikago
egies [1–3]. Recently, a third passive control concept, known as the
et al. [6] developed the viscous mass damper (VMD) which consists
inerter, was introduced [4]. Compared to springs and dampers, the
of a ball-screw inerter and a tubular viscous damper in parallel and
inerter is a two-terminal device that exerts an inertial force propor-
the tuned viscous mass damper (TVMD) which consists of a
tional to the relative acceleration between its terminals. The inerter
tuning spring and the VMD in series. They applied each damper to
can be realized by various methods, e.g., via a rack-pinion [5] or ball-
a single-degree-of-freedom (SDOF) structure and conducted shake
screw mechanism [6]. Either mechanism converts linear motion to
tests using two historical earthquake ground motions. The test
rotational motion of a flywheel of large moment of inertia, which
results showed that the TVMD was found most effective compared
in turn generates large inertial forces. It can also be realized via
to the VMD and the traditional viscous damper. Furthermore, Lazar
mechanical or hydraulic leverage [7,8]. The lever amplifies the
et al. [19] proposed the tuned inerter damper wherein the inerter was
motion of a small mass or fluid, which likewise generates large iner-
used to replace the oscillating mass in the TMD. They compared its
tial forces. Regardless of the method used, the inerter amplifies the
performance with the TMD in multi-degrees-of-freedom structures
inertia effect of a small mass. It was reported in Ref. [9] that a ball-
subjected to seismic excitation and demonstrated similar effective-
screw or rack-pinion inerter with a flywheel of approximately 1 kg
ness between the two. Also, Marian and Giaralis [20] proposed the
can have an apparent mass in a range of 60–200 kg. Furthermore,
tuned mass damper inerter (TMDI) which consists of one TMD
a ball-screw inerter designed by Ikago et al. [6] was capable of gen-
and one inerter in series in an attempt to reduce the physical mass
erating an apparent mass of 350 kg using a flywheel weighing only
of the TMD. In a three degrees-of-freedom (3DOF) structure simula-
2 kg. In other words, the inerter features the “mass amplification
tion, they showed that with similar vibration control performance,
effect” that allows an amplification factor of 10–102.
the weight of the TMDI was four times lighter than the TMD. Fur-
Because of the mass amplification effect, the inerter has been inte-
thermore, De Domenico and Ricciardi [23] incorporated the TMDI
grated with various vibration control strategies in the last decade,
in the base-isolation system and demonstrated that the displacement
including viscous dampers in vehicle and train suspensions [4,9–
demand of base-isolated structures can be significantly reduced.
15], and base-isolation and tuned mass dampers (TMDs) in building
Moreover, Joubaneh and Barry [26] studied the performance of
vibration suppression and seismic control [6,16–26]. Smith and
four models of electromagnetic resonant shunt TMDI on both vibra-
Wang [10] studied various suspension struts that contain at most
tion suppression and energy harvesting and identified the best model.
one inerter and one viscous damper. They showed that the inerter
Their parametric studies showed that increasing the inertance
improved the performance by about 10% or more for ride comfort,
enhances the performance of the best model in terms of both vibra-
tion mitigation and energy harvesting.
In addition to the mass amplification effect, the rack-pinion or
Contributed by the Technical Committee on Vibration and Sound of ASME for ball-screw inerter has another feature that has been less studied:
publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received June
11, 2019; final manuscript received October 11, 2019; published online October 22, the linear-to-rotational motion transmission. Thanks to the rack-
2019. Assoc. Editor: Nick Boechler. pinion or ball-screw mechanism, the inerter converts the linear

Journal of Vibration and Acoustics Copyright © 2019 by ASME FEBRUARY 2020, Vol. 142 / 011015-1
vibration of a primary structure to rotational oscillation of a flywheel.
Because the rotational oscillation is directly related to linear vibra-
tion, an intriguing question then arises: Is suppressing the rotational
oscillation equivalent to suppressing the linear vibration? If it is
true, the weight of such a damper would be much lighter because
the weight of the flywheel is much less than the primary structure.
This motivates the present work. To answer this question, we
propose a new inerter-based damper wherein the flywheel is attached
to the inerter via a rotational spring and a viscous damper. It is worth
noting that in the classic inerter, the flywheel is rigidly attached. In
effect, the flywheel-spring-damper system functions as a “torsional

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
tuned mass damper” that aims to suppress the rotational oscillation
of the inerter. In this regard, the proposed damper is referred to as
tuned inerter-torsional-mass-damper (TITMD) for the rest of the
paper. Fundamentally different from the classic two-terminal
inerter, the TITMD utilizes a third terminal to connect the torsional
tuned mass damper. Thus, it is deemed a “three-terminal” device.
The rest of the paper is organized as follows: In Sec. 2, the con-
figuration and working principle of the proposed TITMD are intro-
duced and explained. Next, the equation of motion of the TITMD is
derived. Furthermore, the TITMD is applied to SDOF primary
structures subjected to random ground motion excitation. In Sec. 3,
the H2 optimization is conducted to obtain the optimum parameters
of the damper. The optimum parameters are in closed-form when
the primary damping is absent and in series approximation when
the primary damping is nonzero and small. Next, the optimum
parameters are validated via numerical simulations. In Sec. 4, the
TITMD are assessed compared with its counterpart: the TMDI is
proposed in the literature [20] because the TMDI also comprises
one inerter and one tuned mass damper. In Sec. 5, the TITMD is
applied to a two degrees-of-freedom (2DOF) structure and the H2
optimization is conducted to obtain the optimum parameters of
Fig. 1 SDOF primary structure incorporating the proposed
the damper in closed-form when the primary damping is absent. TITMD: (a) simplified mathematical model and (b) design
The optimum parameters are again compared with numerical simu- concept of TITMD
lations. Furthermore, the effect of the inerter’s apparent mass (iner-
tance) is studied. Finally, this work is concluded in Sec. 6.
defined as the angular displacement of the pinion and the flywheel,
respectively. Then, the linear-to-rotational motion transmission
2 Proposed Tuned Inerter-Torsional-Mass-Damper for factor R, which converts the linear relative motion x1 − x0 of the
a Single-Degree-of-Freedom Motion Excited Primary primary structure to rotational motion of the pinion, is written as
Structure x1 − x0
R= (1)
Figure 1(a) shows a linear damped SDOF primary structure θ1
modeled by a linear spring of stiffness k, a mass M, and a viscous Conveniently, one can define xd as the equivalent linear displace-
damper of damping coefficient c. The displacement of the ment of the flywheel by
primary structure is denoted by x1 and the ground motion by x0.
To control the vibration of the primary structure, it is herein pro- xd = Rθ2 (2)
posed to consider the classic rack-pinion inerter of moment of
inertia I1, in conjunction with a torsional tuned mass damper. The By Newton’s second law, the equations of motion governing θ1
tuned mass damper consists of a flywheel of moment of inertia I2, and θ2 are derived as
a torsional spring of stiffness kθ, and a torsional damper of

damping coefficient cθ. This new type of inerter-based damper is I1 θ̈1 = kθ (θ2 − θ1 ) + cθ θ̇2 − θ̇1 + RF
referred to as TITMD. It is noteworthy that although the present
 (3)
I2 θ̈2 = −kθ (θ2 − θ1 ) − cθ θ̇2 − θ̇1
TITMD considers a rack-pinion mechanism, it can also be readily
realized by a ball-screw mechanism. where F is the force exerted on the rack.
In practice, the TITMD can be achieved by a design inspired by Let us define mp = I1/R 2 and md = I2/R 2 as the equivalent mass of
the dual mass flywheel (DMF), as shown in Fig. 1(b). The DMF is a the pinion and flywheel, respectively, and kd = kθ/R 2 and cd = cθ/R 2
type of torsional tuned mass damper that is widely applied to isolate as the equivalent damper tuning stiffness and damping coefficient,
the automobile transmission from the torsional vibration of the respectively. Then, substitution of (1) and (2) into (3) leads to
engine [27]. As shown in Fig. 1(b), the arc spring provides the tor-
sional stiffness kθ. When the relative rotational velocity is small, the F = mp (ẍ1 − ẍ0 ) + kd [(x1 − x0 ) − xd ] + cd [(ẋ1 − ẋ0 ) − ẋd ]
contact friction of the arc spring can be approximated by viscous (4)
0 = md ẍd + kd [xd − (x1 − x0 )] + cd [ẋd − (ẋ1 − ẋ0 )]
damping, and thus provides the torsional damping coefficient cθ.
The damping coefficient can be adjusted by including a viscous Several things in (4) are worth noting. First, mp (ẍ1 − ẍ0 ) is pro-
liquid damper inside the DMF, e.g., see Ref. [28]. Adjustment of portional to the relative acceleration between the two terminals
the damping coefficient can be achieved by changing the viscosity and thus denotes the inerter force of the pinion. The proportionality
of the viscous liquid. mp = I1/R 2 is known as the inertance. The inertance can be readily
amplified by increasing the moment of inertia, e.g., attaching a
2.1 Equations of Motion of the Tuned Inerter-Torsional- flywheel to the pinion. Second, kd[(x1 − x0) − xd] and
Mass-Damper. Referring to Fig. 1(a), suppose θ1 and θ2 are cd [(ẋ1 − ẋ0 ) − ẋd ] are the forces attributed to the torsional tuned

011015-2 / Vol. 142, FEBRUARY 2020 Transactions of the ASME

mass damper. Unlike traditional two-terminal devices, such as The analytical expression of (11) can be obtained by solving the
springs, viscous dampers, and the inerters, these forces depend on corresponding Lyapunov equation as described in Ref. [13]
the relative value of three displacements x1, x0, and xd, each repre-
senting one terminal of the TITMD. In other words, the TITMD is r3 ξ3d + r2 ξ2d + r1 ξd + r0
deemed a “three-terminal” device. Third, the second equation in (4) I= (12)
represents the equation of motion of the torsional tuned mass q3 ξ3d + q2 ξ2d + q1 ξd + q0
damper.
It is noted that the coefficients in (12) are provided in (A1) in the
2.2 Equations of Motion of the SDOF Primary Structure. Appendix. To find the optimum parameters, Eq. (12) is differenti-
Referring to Fig. 1(a), the equation of motion of the primary ated with respect to ν and ξd, separately. This gives the following
mass M can be derived as simultaneous equations:

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
M ẍ1 + k(x1 − x0 ) + c(ẋ1 − ẋ0 ) = −F (5) a4 ξ4d + a3 ξ3d + a2 ξ2d + a1 ξd + a0 = 0
(13)
Substitution of (4) into (5) yields b4 ν4 + b3 ν3 + b2 ν2 + b1 ν + b0 = 0
M ẍ1 + c(ẋ1 − ẋ0 ) + k(x1 − x0 ) + mp (ẍ1 − ẍ0 ) It is noted that the coefficients in (13) are provided in (A2) in the
+ cd [(ẋ1 − ẋ0 ) − ẋd ] + kd [(x1 − x0 ) − xd ] = 0 (6) Appendix.
md ẍd + cd [ẋd − (ẋ1 − ẋ0 )] + kd [xd − (x1 − x0 )] = 0 Solutions of (13) correspond to the optimum tuning ratio νopt and
the damping ratio ξopt
d that minimize the mean square value of the
Define the relative displacement x = x1 − x0 and rearrange terms. relative displacement E[x 2] of the primary mass. When the

 primary mass’s damping ratio ξ ≠ 0, closed-form solutions of νopt
M + mp ẍ + cẋ + kx + cd (ẋ − ẋd ) + kd ( x − xd ) = −M ẍ0 and ξopt
(7) d are difficult to obtain. Instead, a series solution using the
md ẍd + cd (ẋd − ẋ) + kd ( xd − x) = 0 perturbation method [29] will be obtained. First, the primary
damping ratio ξ is assumed to be a small perturbation, i.e.,
The following dimensionless parameters are introduced to
rewrite the equations of motion in a dimensionless form: ξ=ϵ (14)
 Second, ξd and ν are expanded into a power series in ϵ, i.e.,
mp md k c
μ= , μd = , ωn = , ξ=
M M M 2Mωn ξd = ξd0 + ϵξd1 + · · ·
 (15)
kd cd ωd ν = ν0 + ϵν1 + · · ·
ωd = , ξd = , ν= , τ = ωn t (8)
md 2md ωd ωn Finally, Eqs. (15) are substituted into (13), and terms of equal
d powers of ϵ are collected and equated to zero.
(·)′ = ω−1
n (· ) By solving these equations in each order of ϵ successively, one
dt
can obtain solutions with different orders of accuracy. The
With (8), the equations of motion are rewritten as zeroth-order solutions and the first-order approximation will be

 shown in Secs. 3.1 and 3.2.
1 + μ x′′ + 2ξx′ + x + μd x′′d = −x′′0

 (9)
x′′d + 2νξd x′d − x′ + ν2 ( xd − x) = 0 3.1 Zeroth-Order Approximation. Equating the zeroth-order
terms in (13) to zero, one can obtain the following equations:
There are two things worth noting in (9). First, the torsional mass
damper creates a resisting inertia force μd x′′d to counter the primary

ν4 3μ2 + 6μμd + 6μ + 3μ2d + 6μd + 3
structure’s motion x1. This working principle is very similar to the

classic TMD; thus, it is expected that the proposed TITMD will + ν2 4μξ2d −2μ + 4μd ξ2d + μd + 4ξ2d −2 −1 = 0
share some similar features with the TMD. Second, the mass ratio
 (16)
μ of the inerter only appears in (1 + μ)x′′ . It is clear that the only ν4 −μ2 −2μμd −2μ − μ2d −2μd −1
effect of μ is to increase the equivalent mass of the primary structure.

+ ν2 4μξ2d + 2μ + 4μd ξ2d + μd + 4ξ2d + 2 −1 = 0
Equation (16) has only one set of positive solutions:
3 H2 Optimization of the Tuned Inerter-Torsional-


Mass-Damper μd (4μ + 3μd + 4)
ξopt =


Assuming that the motion excitation x¨0 is a stationary random d0
24μμd + 32μ + 24μd + 8μ2d + 16μ2 + 16
process with a uniform power spectrum density S0 (the unit of S0 (17)
is m2/(s3 rad)), the variance of the relative displacement process x 2 + 2μ + μd
νopt
0 =
of the primary structure is written as 2(1 + μ + μd )2

∞ 2 ∞ 2
  x x 3.2 First-Order Approximation. Equating the first-order
E x2 = S0 dω = S0 dλ (10)
x¨ ω3 x′′ terms to zero and substituting ξopt opt
d0 and ν0 from (17), one can
−∞ 0 n −∞ 0
obtain the first-order equations
where |x/x′′0 | is the magnitude of the transfer function from x′′0 to x.
Given a mass ratio μ of the inerter and μd of the torsional mass c 1 ν1 + c 0 = 0
(18)
damper, it is sought to determine optimum values for the tuning d 1 ν1 + d 0 = 0
stiffness ratio ν and damping ratio ξd, which minimizes the variance
E[x 2]. To simplify the computation, a dimensionless performance It is noted that the coefficients in and the solutions to (18) are pro-
index is defined by vided in (A3) and (A4)–(A5) in the Appendix, respectively, due to
their lengthy expressions.
  
E x2 1 ∞ x 2
I= = dλ (11) 3.3 Numerical Validation. To validate the approximation
2πS0 /ω3n 2π −∞ x′′0 solutions in (17) and (A4)–(A5), a standard Matlab built-in

Journal of Vibration and Acoustics FEBRUARY 2020, Vol. 142 / 011015-3

minimization function fmincon is used to solve the H2 optimization excitation has a constant power spectrum density, shifting the
numerically. First of all, the zeroth-order optimum parameters ξopt d
natural frequency has little effect on reducing the vibration. This
and νopt in (17) are compared with the numerical solutions. The feature, however, will be found useful when the primary structure
comparison is presented graphically in Fig. 2. The optimum param- is subjected to excitation of colored noise or in multi-degrees-
eters ξopt
d and ν
opt
with several mass ratios μd of the torsional mass of-freedom structures (see Sec. 5).
damper are plotted against the mass ratio μ of the inerter. To demonstrate the effect of μd and μ, the frequency response of
There are several things worth noting in Fig. 2. First of all, as |x/x′′0 |2 is plotted with various μd in Fig. 2(d) and with various μ in
shown in Figs. 2(a) and 2(b), the difference between the analytical Fig. 2(e). There are two things worth noting. First, Fig. 2(d) shows a
solutions and numerical solutions is indiscernible. Second, ξopt decreasing trend of the frequency response as μd increases, which
d
increases and νopt decreases as μd increases. This trend is the same confirms the observation in Fig. 2(c). Second, Fig. 2(e) shows

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
with the optimum parameters in the classic TMD, e.g., see that the frequency response slightly increases and shifts toward a
Ref. [29]. Third, to examine the effectiveness of the TITMD, the lower frequency range as μ increases, which also complies with
minimum performance index Imin with various μd is plotted as a func- Fig. 2(c).
tion of μ in Fig. 2(c). As shown, the performance index Imin decreases Next, the first-order approximation in (A4) and (A5) in the
as μd increases, i.e., the larger the torsional mass damper’s inertia, the Appendix is compared with the numerical solutions for the case
better vibration suppression can be achieved, which is the same with the primary damping ξ ≠ 0. Because μ has little effect, μ = 0.06 is
the classic TMD. It should be noted that μd is proportional to the assumed in this comparison. The approximated and numerical
equivalent mass of the torsional mass damper not its physical optimum ξopt d and ν
opt
with various μd are plotted against ξ in Fig. 3.
mass. Due to the mass amplification effect of the inerter, the physical There are several things worth noting in Fig. 3. First, the first-
mass can be ten or even hundred times smaller than the classic TMD order approximation is very close to the numerical solutions, espe-
and this makes the proposed TITMD very attractive. Fourth, the cially when ξ < 0.3 which is a practically large value for common
inerter mass ratio μ seems to have little effect on Imin, especially viscous damping. In other words, the first-order approximation is
when μd < 0.3. This finding can be explained as follows. As sufficient for practical applications. Second, the optimum ξopt d is
explained after (9), the only effect of μ is to change the equivalent little affected by ξ as shown in Fig. 3(a). This trend is very different
mass of the primary structure. Equivalently, it decreases the from the classic TMD, where the optimum damping ratio is signifi-
natural frequency of the primary structure. Because the motion cantly affected by ξ, especially after ξ > 0.1, e.g., see Ref. [29].

(a) (b) (c) (d)

(e)
Fig. 2 Optimum parameters for H2 optimization when primary damping ξ = 0: (a) optimum damping ξopt d , (b) optimum tuning ν
opt
,
(c) minimized performance index Imin, (d) frequency response curve with different μd and μ = 0.1, and (e) frequency response curve
with different μ and μd = 0.1

011015-4 / Vol. 142, FEBRUARY 2020 Transactions of the ASME

 Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
(a) (b) (c) (d)
Fig. 3 Optimum parameters for H2 optimization (μ = 0.06) with varying primary damping ξ: (a) optimum damping ξopt
d ,
(b) optimum tuning νopt, (c) minimized performance index Imin, and (d) frequency response curve with different ξ

This is an advantageous feature because it means that the proposed The minimum performance index of the TMDI is provided in the
TITMD is less susceptible to the change in the primary damping. literature [20]
Third, the minimum performance index Imin with several different 

μd is plotted in Fig. 3(c) to show the influence of ξ. Finally, the fre- 1 (1 + μ̂ )3 (u(3 − μ̂ ) + (4 − μ̂ )(1 + μ̂ ))

d

min =
d d d
quency response with increasing ξ is plotted in Fig. 3(d ). As I TMDI (19)
2 μ + μ̂d 1 + μ + μ̂d
expected, the response decreases with an increase of ξ.
For the sake of comparison, the minimum performance index of the
proposed TITMD is provided as follows:
4 Assessment of the Tuned Inerter-Torsional- 

Mass-Damper When Compared With the Tuned Mass 1 4 + 4μ + 3μ
Damper Inerter I min = 
TITMD
d
 (20)
2 μd 1 + μ + μd
To further assess the proposed TITMD, we compare its perfor-
mance with the TMDI proposed by Marian and Giaralis [20]. To assess the performance, a ratio J is defined by
Because the TMDI also consists of one inerter and one TMD, it
is a good counterpart to compare with. I TITMD
J= min
(21)
Figure 4 shows the configuration of the TMDI. As shown, the I TMDI
min
TMDI consists of a linear tuned mass damper m̂d in conjunction
with the inerter m in series. In this regard, the inerter is used as In this way, the proposed TITMD is more effective when J < 1 and
the conventional two-terminal device, distinctly different from the less otherwise.
proposed three-terminal TITMD (see Sec. 2). The criterion for J < 1 can be obtained by solving the inequality
Because the TMDI uses a linear tuned mass damper, we define I TITMD
min < I TMDI
min . The solution is written as
μ̂d = m̂d /M to be the physical mass ratio of the TMDI to be distin-
2
3

guished from the equivalent mass ratio μd = md/M in (8). Further- u < 3 Aμ̂d +2 Aμ̂d +O μ4d (22)
more, because the proposed TITMD’s equivalent mass can be where O(·) is the order operator. For small μd < 1, the fourth order
amplified through the rack-pinion inerter, let us define A as the and higher terms can be neglected.
mass amplification factor. That is, if the physical mass ratio of the For practical values 0 < μ < 1, Eq. (22) is satisfied when
TITMD’s torsional mass damper is μ̃d , then μd = Aμ̃d . For a fair
comparison, we assume that the proposed TITMD and its counter- 1
part TMDI have an identical physical mass ratio, i.e., A> (23)
2μ̂d
μ̂d = μ̃d = μd /A.
Provided that a typical range of the mass ratio for the classic TMD is
0.01 < μ̂d < 0.1 [30], (23) is equivalent to a mass amplification
factor A > 50, which can be readily achieved knowing that a
typical mass amplification factor of the inerter is in a range of
10–102 [6,9]. To examine the performance for practical applica-
tions, we plot J with A = 50 in Fig. 5.
As shown in Fig. 5, given an identical mass ratio μ of the inerter
and physical mass ratio μˆd of the torsional damper, the proposed
TITMD is more effective than its counterpart TMDI. The improve-
ment is in a range of 20–70% for 0 < μ < 1 and 0 < μ̂d < 1.

5 H2 Optimization for Undamped Two Degrees-of-

Freedom Primary Structures
In this section, the effectiveness of the proposed TITMD on a
Fig. 4 SDOF primary structure incorporating the TMDI configu- 2DOF structure is studied. The purpose is to examine how the iner-
ration [20] ter’s mass ratio μ plays a role in reducing the vibration. To this end,

Journal of Vibration and Acoustics FEBRUARY 2020, Vol. 142 / 011015-5

 Given the modal matrix, let us define the following dimensionless
parameters:
α = 1 − ϕ1
β = 1 − ϕ2
M̂1 + ϕ1 M̂2
γ=
M1
M̂1 + ϕ2 M̂2
δ=
M1 (26)

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
 
Fig. 5 Comparison of the minimum performance index of the ω2 k1 k2
proposed TITMD and TMDI in Ref. [20]. J = Imin /Imin is the
TITMD TMDI w = , ω1 = , ω2 =
ω1 M1 M2
ratio between the performance index of TITMD and TMDI. ⎡ m m ⎤
  −
1 a ⎢ M1 M1 ⎥
μ = ΦT ⎣ m m ⎦Φ
a b −
let us add one more oscillating mass to the SDOF primary structure M1 M1
shown in Fig. 1. The resultant 2DOF structure is shown in Fig. 6,
where M1 is the modal mass corresponding to the fundamental
where M̂1 and M̂2 are the mass of the primary and secondary struc-
mode of the uncontrolled primary structure and μ is the correspond-
tures, respectively, and k̂1 and k̂2 are the stiffness, respectively. For
ing modal mass ratio of the inerter.
simplicity, both viscous damping coefficients c1 and c2 are
The modal equations of motion governing the modal coordinates
assumed to be zero. Then, the equations of motion of the system
q1 and q2 are written as
are written as
(1 + μ)q′′1 + aμq′′2 + q1 = −αμd x′′d − γx′′0
M K

   
  
M̂1 + m −m ẍ1 k̂1 −k̂1 x1 (1 + b)μq′′2 + aμq′′1 + μw2 q2 = −βμd x′′d − δx′′0 (27)
+
−m M̂2 + m ẍ2 −k̂1 k̂2 + k̂1 x2 (24) x′′d + 2νξd (x′d − (αq′1 + βq′2 )) + ν2 (xd − (αq1 + βq2 )) = 0
 
−md ẍd − M̂1 ẍ0 Note that μ, μd, ξd, and ν are normalized with respect to the modal
= mass M1 in (25).
−md ẍd − M̂2 ẍ0 Following the procedure in Sec. 3, the performance index is
derived and written as
 
To facilitate deriving the optimum parameters, we transform the E x21 c2 ξ2 + c1 ξd + c0
I= = d (28)
equations of motion of the system into modal domain using the 2πS0 /ω1
3 d1 ξd + d0
mode shapes of the uncontrolled 2DOF primary structure.
Suppose there exists a modal matrix Φ consisting of the mode It is noted that the coefficients in (28) are provided in (A6) in the
shapes of the uncontrolled primary structure such that Appendix.
To find the optimum ξopt d , ν , and μ , (28) is differentiated with
opt opt

      respect to ξd, ν, and μ, respectively. The solutions to the simulta-

M1 0 k1 0 1 1 neous equations correspond to the optimum parameters. It is
ΦT MΦ = , ΦT KΦ = , Φ=
0 M2 0 k2 ϕ1 ϕ2 worth noting that unlike the case for the SDOF primary structure,
there exist optimum μ opt for the 2DOF structure.
(25)
The optimum νopt and ξopt d are
 
νn opt ξnd
ν =
opt
, ξ = (29)
νd d
ξdd
where νn, νd, ξnd , and ξdd in (29) are provided in (A7) in the Appendix
due to their lengthy expressions. Furthermore, the optimum μ opt sat-
isfies the following implicit equation:
e5 (μopt )5 + e4 (μopt )4 + e3 (μopt )3 + e2 (μopt )2
(30)
+ e1 μopt + e0 = 0
where the coefficients in (30) are provided in (A8) in the Appendix.
Next, a numerical simulation is conducted to validate the
optimum parameters. The simulation parameters are listed in
Table 1. These parameters give two natural frequencies ω1 =
2.40 rad/s and ω2 = 4.66 rad/s. The comparison of ξopt d and νopt
and the corresponding minimized Imin are plotted in Fig. 7.

Table 1 Simulation parameters

j Mass, M̂j (kg) Stiffness, k̂j (N m−1)

1 1 10
2 2 25
Fig. 6 2DOF structure incorporating the proposed TITMD

011015-6 / Vol. 142, FEBRUARY 2020 Transactions of the ASME

 Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
(a) (b) (c) (d)
Fig. 7 Optimum parameters for H2 optimization for a 2DOF primary structure when primary damping ξ = 0: (a) optimum
damping ξopt
d , (b) optimum tuning ν
opt
, (c) minimized performance index Imin, and (d) frequency response curve with
optimal and off-tuned μd

There are three things worth noting. First, as seen in Figs. 7(a) oscillating mass and is connected with the inerter in series. The
and 7(b), the difference between the closed-form and numerical comparison showed that given a mass amplification factor of 50,
solutions is indiscernible. Therefore, the closed-from solutions are which is commonly seen in the literature, the proposed TITMD
accurate. Second, as seen in Fig. 7(c), there exists a curve indicating achieves better performance by 20–70% when the physical
the optimum μoptd or equivalently μ . For parameters that are away
opt
weight is identical to its counterpart. This study confirmed the
from the curve, the performance index increases. As a result, advantage of the proposed device.
increasing the mass ratio μd of the inertia damper does not necessar- Furthermore, the TITMD is applied to a 2DOF structure to
ily improve the performance. In fact, in this particular numerical examine the effect of the inerter. Optimum tuning ratio and
example, for μ < 0.1, increasing μd actually deteriorates the perfor- damping ratio are obtained in closed-from when the structure is
mance. To demonstrate this, let us set μ = 0.08 and plot the corre- undamped. Again, the closed-form solutions are validated by
sponding frequency response with the optimum μopt d and two numerical simulation. Based on the study, given a mass ratio of
other, larger μd in Fig. 7(d ). As shown, the first resonance peak is the torsional mass damper, there exists an optimum mass ratio of
significantly increased when μd is larger than μopt d . This finding is the inerter. Increasing the equivalent mass of the torsional mass
different from the observation in Fig. 2(d) and worth further inves- damper without considering the inertance no longer guarantees
tigation in the future. better vibration suppression. This is particularly true when the
inerter has small mass ratios.

6 Conclusions
Acknowledgment
In this paper, a new inerter-based damper, TITMD, is proposed.
The damper consists of the classic rack-pinion inerter in conjunc- This material is based upon work supported by the start-up
tion with a tuned torsional mass damper. Unlike any other inerter- funding in Michigan State University (Funder ID: 10.13039/
based dampers, the torsional mass damper is connected with the 100007709). Any opinions, findings, and conclusions or recom-
pinion of the inerter via a rotational spring and a viscous damper. mendations expressed in this material are those of the author
The equivalent mass of the torsional mass damper can be signifi- and do not necessarily reflect the views of Michigan State
cantly amplified by a factor of 10 − 102 due to the mass amplifica- University.
tion effect of the inerter. Thus, a torsional mass damper with light
physical weight can be used to achieve high level of vibration
suppression. Appendix
The proposed TITMD is applied to SDOF primary structures sub-
jected to random excitation to examine its effectiveness. The H2 The coefficients in (12) are listed as follows:
optimization is conducted to minimize the relative displacement
r0 = μd ν3 ξ
of the primary structure. Optimal tuning ratio and damping ratio


of the torsional mass damper are obtained in closed-form when r1 = ν4 (μ + μd + 1)2 − ν2 2μ + μd −4ξ2 + 2 + 1
the primary structure damping is absent and in series approximation

when the structure damping is small. The closed-form and approx- r2 = 4νξ ν2 (μ + μd + 1) + 1
imated solutions are validated with numerical simulations. The
study showed that the proposed TITMD effectively suppresses r3 = 4ν2 (μ + μd + 1)
the vibration, and the effectiveness is increased as the torsional (A1)
q0 = 4μd ν3 ξ2
mass damper’s equivalent mass increases. On the other hand, the


 
mass of the inerter has little effect on the performance for SDOF q1 = 4ξ ν4 (μ + μd + 1)2 −2ν2 μ−2ξ2 + 1 + 1
primary structures.

 
Next, the TITMD is compared with its counterpart reported in q2 = 4ν 4ξ2 ν2 (μ + μd + 1) + 1 + μd
the literature [20] which also consists of one inerter and one
tuned mass damper. Only the tuned mass damper uses a linear q3 = 16ν2 ξ(μ + μd + 1)

Journal of Vibration and Acoustics FEBRUARY 2020, Vol. 142 / 011015-7

The coefficients in (13) are listed as follows: c = 2νopt 4 opt 8
0 (−1 + 5(1 + μ + μd ) (ν0 )

a0 = μd ν3 ξ + 2(νopt opt 2
0 ) (3 + 3μ + 2μd + 8(1 + μ + μd )(ξd0 ) )
2

 + 2(1 + μ + μd )2 (νopt 6
a1 = ν4 (μ + μd + 1)2 − ν2 2μ + μd −4ξ2 + 2 + 1 0 ) (−3−3μ−5μd

 + 28(1 + μ + μd )(ξopt 2
a2 = 4νξ ν2 (μ + μd + 1) + 1 d0 ) )

a3 = 4ν2 (μ + μd + 1) + (νopt
0 ) (−4−8μ−4μ −10μd −10μμd −5μd
4 2 2

a4 = 4ν2 (μ + μd + 1) −40(1 + μ + μd )2 (ξopt 2 opt 4

d0 ) −16(1 + μ + μd ) (ξd0 ) ))
2

(A2)

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
b0 = μd ν3 ξ d = −3 + 21(1 + μ + μd )4 (νopt 8 opt 2
0 ) + 10(ν0 ) (2 + 2μ + μd (A5)
b1 = 0 + 4(1 + μ + μd )(ξopt 2
d0 ) )


b2 = 4νξ ν2 (μ + μd + 1) + 1 + 14(1 + μ + μd )2 (νopt
0 ) (−2−2μ − μd
6

b3 = 4ν2 (μ + μd + 1) + 12(1 + μ + μd )(ξopt 2

d0 ) )

b4 = 4ν2 (μ + μd + 1) + (νopt
0 ) (−5(2(1 + μ) + 8(1 + μ)μd + 5μd )
4 2 2

The coefficients in (18) are listed as follows: −40(1 + μ + μd )(2 + 2μ + μd )(ξopt

d0 )
2

c1 = μd (ξopt opt opt −112(1 + μ + μd )2 (ξopt 4

d0 ) )
d0 ) (−2(2 + 2μ + μd )ν0 ξd0
2

+ 12(1 + μ + μd )2 (νopt 3 opt opt opt 3

0 ) ξd0 + 8(1 + μ + μd )ν0 ξd0 )

c0 = μd (ξopt opt opt 3 2 opt 5 The coefficients in (28) are listed as follows:
d0 ) (−2ν0 + 2μd (ν0 ) + 2(1 + μ + μd ) (ν0 )
2

+ 8(1 + μ + μd )(νopt 3 opt 2 2 opt 4

0 ) (ξd0 ) −3ξd1 + 9(1 + μ + μd ) (ν0 ) ξd1


+ (νopt opt 2
d0 ) (−3(2 + 2μ + μd ) + 20(1 + μ + μd )(ξd0 ) )ξd1 )
2 c2 = w4 8aδγμ3 ν2 + 8αβδγμ2 μd ν2


d1 = −2μd νopt opt 2 opt 4
0 ξd0 (1 + (1 + μ + μd ) (ν0 )
+ w6 4α2 γ2 μ3 μd ν2 + 4γ2 (μ + 1)μ3 ν2


+ (νopt opt 2
0 ) (−2−2μ − μd −8(1 + μ + μd )(ξd0 ) ))
2 + w2 4(b + 1)δ2 μ2 ν2 + 4β2 δ2 μμd ν2
c1 = 0
d0 = μd ξopt opt 4 opt opt 2
d0 (−2μd (ν0 ) − ν1 ξd0 + (ν0 ) (3(2 + 2μ + μd )

c0 = 2δγμν2 w2 (μw2 (ν2 α2 μd + μ + 1 (aμ + αβμd )
−5(1 + μ + μd )2 (νopt opt opt 4 opt 2
0 ) )ν1 ξd0 + 8(1 + μ + μd )(ν0 ) (ξd0 )
2


−2aμ − αβμd ) + ν2 bμ + β2 μd + μ (aμ + αβμd ))
+ 12(1 + μ + μd )(νopt opt 3
0 ) ν1 (ξd0 ) )
2

(A3) + γ2 μ2 w4 ν4 (aμ + αβμd )2

"
2
 #
The first-order approximation of the solutions to (18) is provided as + γ2 μ3 w6 ν2 ν2 α2 μd + μ + 1 − α2 μd + 2μ + 2 + 1
follows:

+ δ2 μν2 w2 ν2 (aμ + αβμd )2 −2(b + 1)μ + β2 ( − μd )
a
νopt
1 =
2
b
(A4) + δ2 ν4 bμ + β2 μd + μ +μ2 w4
c
ξopt
d1 = d1 = 4μ3 μd νw6
d
d0 = 0 (A6)
where

a = 2(νopt 4 opt 8
0 ) (−2 + 2(1 + μ + μd ) (ν0 )
2
The numerators and denominators in (29) are listed as
+ (νopt 2
+ 4μ + 7μd + 12(1 + μ + μd )(ξopt 2 follows:
0 ) (4 d0 ) )

+ (1 + μ + μd )2 (νopt 6
0 ) (−4−4μ−9μd

+ 28(1 + μ + μd )(ξopt opt 4 νn = μopt w2 2δγμopt w2 (2aμopt + αβμd )

d0 ) ) + (ν0 ) (μd (2 + 2μ + μd )
2


−8(1 + μ + μd )(5 + 5μ + 7μd )(ξopt 2 + μopt w2 γ2 (μopt )2 w4 α2 μd + 2μopt + 2
d0 )

+ 16(1 + μ + μd )2 (ξopt 4 + μopt w2 δ2 (2(b + 1)μopt + β2 μd )

d0 ) ))


b = ξopt 4 opt 8 νd = 4δγμopt w2 (aμopt + αβμd )(μopt w2 α2 μd + μopt + 1
d0 (3−21(1 + μ + μd ) (ν0 )

+ 14(1 + μ + μd )2 (νopt opt 2 + (b + 1)μopt + β2 μd )

0 ) (2 + 2μ + μd −12(1 + μ + μd )(ξd0 ) )
6

+ 10(νopt opt 2 + 2γ2 (μopt )2 w4 ((aμopt + αβμd )2

0 ) (−2−2μ − μd −4(1 + μ + μd )(ξd0 ) )
2

2
+ (νopt
0 ) (5(2(1 + μ) + 8(1 + μ)μd + 5μd )
4 2 2 + μopt w2 α2 μd + μopt + 1 )

+ 40(1 + μ + μd )(2 + 2μ + μd )(ξopt

d0 )
2 + 2δ2 (μopt w2 (aμopt + αβμd )2

2
+ 112(1 + μ + μd )2 (ξopt 4
d0 ) )) + bμopt + β2 μd + μopt )

011015-8 / Vol. 142, FEBRUARY 2020 Transactions of the ASME

 [4] Smith, M. C., 2002, “Synthesis of Mechanical Networks: The Inerter,” IEEE
ξnd = 4αβ3 δγμopt μ2d w2 (δ2 + 2γ2 μopt w2 ) Trans. Autom. Control, 47(10), pp. 1648–1662.
[5] Wang, F.-C., Liao, M.-K., Liao, B.-H., Su, W.-J., and Chan, H.-A., 2009, “The
+ β4 δ2 μ2d (3δ2 + 4γ2 μopt w2 ) + 4αβ(μopt )2 μd w2 (2aδ4 Performance Improvements of Train Suspension Systems With Mechanical
Networks Employing Inerters,” Veh. Syst. Dyn., 47(7), pp. 805–830.
+ 2δγ((1 + b)γ2 μopt + δ2 (1 + μopt + α2 μd ))w2 [6] Ikago, K., Saito, K., and Inoue, N., 2012, “Seismic Control of
Single-Degree-of-Freedom Structure Using Tuned Viscous Mass Damper,”
+ γ3 μopt (2aγμopt + α2 δμd )w4 ) Earthquake Eng. Struct. Dyn., 41(3), pp. 453–474.
[7] Flannelly, W. G., 1967, “Dynamic Antiresonant Vibration Isolator,” U.S. Patent
+ (μopt )3 w2 (4δ2 (−aδ + γ + bγ)2 3,322,379, May 30.
[8] Yilmaz, C., and Kikuchi, N., 2006, “Analysis and Design of Passive Low-Pass
−4δγ(2(aδ − (1 + b)γ)(aγμopt − δ(1 + μopt )) Filter-Type Vibration Isolators Considering Stiffness and Mass Limitations,”
J. Sound Vib., 293(1–2), pp. 171–195.

Downloaded from https://asmedigitalcollection.asme.org/vibrationacoustics/article-pdf/142/1/011015/6448062/vib_142_1_011015.pdf by Indian Institute of Technology Kanpur user on 06 May 2020
+ α2 δ(−2aδ + γ + bγ)μd )w2 + 4γ2 (a2 γ2 (μopt )2 [9] Papageorgiou, C., and Smith, M. C., 2005, “Laboratory Experimental Testing of
Inerters,” Proceedings of the 44th IEEE Conference on Decision and Control,
−2aδγμopt (1 + μopt ) + δ2 (1 + μopt + α2 μd )2 )w4 IEEE, Seville, Spain, Dec. 12–15, IEEE, New York, pp. 3351–3356.
[10] Smith, M. C., and Wang, F.-C., 2004, “Performance Benefits in Passive Vehicle
(A7)
+ α2 γ4 μopt μd (4 + 4μopt + 3α2 μd )w6 ) Suspensions Employing Inerters,” Veh. Syst. Dyn., 42(4), pp. 235–257.
[11] Papageorgiou, C., and Smith, M. C., 2006, “Positive Real Synthesis Using Matrix
Inequalities for Mechanical Networks: Application to Vehicle Suspension,” IEEE
+ 2β2 μopt μd (4aδγ3 (μopt )2 w4 Trans. Control Syst. Technol., 14(3), pp. 423–435.
[12] Wang, F.-C., and Su, W.-J., 2008, “Impact of Inerter Nonlinearities on Vehicle
+ 2α2 γ4 (2 μd w6 + 2δ4 (1 + b + α2 μd w2 ) Suspension Control,” Veh. Syst. Dyn., 46(7), pp. 575–595.
[13] Scheibe, F., and Smith, M. C., 2009, “Analytical Solutions for Optimal Ride
+ δ2 γ2 μopt w2 (4 + 4b + (−2−2μopt + α2 μd )w2 )) Comfort and Tyre Grip for Passive Vehicle Suspensions,” Veh. Syst. Dyn.,
47(10), pp. 1229–1252.
ξdd = 8(δ2 (μopt + bμopt + β2 μd ) + 2δγμopt (aμopt + αβμd )w2 [14] Wang, F.-C., and Chan, H.-A., 2011, “Vehicle Suspensions With a Mechatronic
Network Strut,” Veh. Syst. Dyn., 49(5), pp. 811–830.
+ γ2 (μopt )2 (1 + μopt + α2 μd )w4 )(δ2 (2(1 + b)μopt [15] Hu, Y., Chen, M. Z. Q., and Shu, Z., 2014, “Passive Vehicle Suspensions
Employing Inerters With Multiple Performance Requirements,” J. Sound Vib.,
+ β2 μd ) + 2δγμopt (2aμopt + αβμd )w2 333(8), pp. 2212–2225.
[16] Ikago, K., Sugimura, Y., Saito, K., and Inoue, N., 2012, “Modal Response
Characteristics of a Multiple-Degree-of-Freedom Structure Incorporated
+ γ2 (μopt )2 (2 + 2μopt + α2 μd )w4 ) With Tuned Viscous Mass Dampers,” J. Asian Arch. Build. Eng., 11(2),
pp. 375–382.
The coefficients in (30) are listed as follows: [17] Garrido, H., Curadelli, O., and Ambrosini, D., 2013, “Improvement of Tuned
Mass Damper by Using Rotational Inertia Through Tuned Viscous Mass
e5 = 2γ2 (νopt )4 w6 Damper,” Eng. Struct., 56, pp. 2149–2153.
[18] Nakamura, Y., Fukukita, A., Tamura, K., Yamazaki, I., Matsuoka, T., Hiramoto,
K., and Sunakoda, K., 2014, “Seismic Response Control Using Electromagnetic
e4 = 2aδγ(νopt )4 w4 + a2 γ2 (νopt )4 w4 −2γ2 (νopt )2 w6 Inertial Mass Dampers,” Earthquake Eng. Struct. Dyn., 43(4), pp. 507–527.
[19] Lazar, I. F., Neild, S. A., and Wagg, D. J., 2014, “Using an Inerter-Based Device
+ 2γ2 (νopt )4 w6 + 2α2 γ2 μδ (νopt )4 w6 for Structural Vibration Suppression,” Earthquake Eng. Struct. Dyn., 43(8),
pp. 1129–1147.
+ 4γ2 (νopt )2 w6 (ξopt
d )
2
[20] Marian, L., and Giaralis, A., 2014, “Optimal Design of a Novel Tuned
Mass-Damper-Inerter (TMDI) Passive Vibration Control Configuration for
e3 = 0 Stochastically Support-Excited Structural Systems,” Probab. Eng. Mech., 38,
pp. 156–164.
e2 = (−2aβ2 δγμδ (νopt )4 w2 −2α3 βδγμ2δ (νopt )4 w4 [21] Javidialesaadi, A., and Wierschem, N. E., 2018, “Optimal Design of Rotational
Inertial Double Tuned Mass Dampers Under Random Excitation,” Eng. Struct.,
− α2 β2 γ2 μ2δ (νopt )4 w4 −2αβδγμδ (νopt )4 w2 (1 + b + w2 ) 165, pp. 412–421.
(A8) [22] Joubaneh, E. F., Barry, O. R., and Zuo, L., 2018, “On the Vibration Suppression
−2αβδγμδ (νopt )2 w4 (−1 + 4(ξopt 2
d ) ) and Energy Harvesting of Building Structures Using an Electromagnetic-
Inerter-Absorber,” ASME 2018 Dynamic Systems and Control Conference,
− δ2 (w4 + (νopt )4 (1 + 2b + b2 + 2aαβμδ w2 ) Atlanta, GA, Sept. 30–Oct. 3, American Society of Mechanical Engineers, New
York, p. V002T18A005.
+ 2(1 + b)(νopt )2 w2 (−1 + 2(ξopt 2 [23] De Domenico, D., and Ricciardi, G., 2018, “An Enhanced Base Isolation System
d ) ))) Equipped With Optimal Tuned Mass Damper Inerter (TMDI),” Earthquake Eng.
Struct. Dyn., 47(5), pp. 1169–1192.
e1 = (−4αβ3 δγμ2δ (νopt )4 w2 [24] Pan, C., Zhang, R., Luo, H., Li, C., and Shen, H., 2018, “Demand-Based Optimal
Design of Oscillator With Parallel-Layout Viscous Inerter Damper,” Struct.
−2β2 δ2 μδ (νopt )2 ((νopt )2 (2 + 2b + α2 μδ w2 ) Control Health Monit., 25(1), p. e2051.
[25] Sun, H., Zuo, L., Wang, X., Peng, J., and Wang, W., 2019, “Exact H2 Optimal
+ w2 (−1 + 4(ξopt 2
d ) ))) Solutions to Inerter-Based Isolation Systems for Building Structures,” Struct.
Control Health Monit., 26(6), p. e2357.
e0 = −3β4 δ2 μ2δ (νopt )4 [26] Joubaneh, E. F., and Barry, O. R., 2019, “On the Improvement of Vibration
Mitigation and Energy Harvesting Using Electromagnetic Vibration
Absorber-Inerter: Exact H2 Optimization,” ASME J. Vib. Acoust., 141(6), p.
061007.
References [27] Albers, A., 1994, “Advanced Development of Dual Mass Flywheel (DMFW)
[1] Constantinou, M. C., Soong, T. T., and Dargush, G. F., 1998, MCEER Design-Noise Control for Today’s Automobiles,” Proceedings of the 5th LuK
Monograph Series, Vol. 1, Multidisciplinary Center for Earthquake Symposium, Buhl, Germany, May 27, pp. 5–41.
Engineering Research, Buffalo, NY. [28] Bopp, W. G., 1988, “Two Mass Flywheel Assembly With Torsional Damping
[2] Kelly, J. M., 1999, “The Role of Damping in Seismic Isolation,” Earthquake Eng. Means,” U.S. Patent 4,782,936, Nov. 8.
Struct. Dyn., 28(1), pp. 3–20. [29] Asami, T., Nishihara, O., and Baz, A. M., 2002, “Analytical Solutions to H∞ and
[3] Symans, M. D., Charney, F. A., Whittaker, A. S., Constantinou, M. C., Kircher, H2 Optimization of Dynamic Vibration Absorbers Attached to Damped Linear
C. A., Johnson, M. W., and McNamara, R. J., 2008, “Energy Dissipation Systems Systems,” ASME J. Vib. Acoust., 124(2), pp. 284–295.
for Seismic Applications: Current Practice and Recent Developments,” J. Struct. [30] Connor, J. J., 2003, Structural Motion Control, Pearson Education, Inc., New
Eng., 134(1), pp. 3–21. York.

Journal of Vibration and Acoustics FEBRUARY 2020, Vol. 142 / 011015-9