Received April 21, 2018, accepted May 23, 2018, date of publication June 4, 2018, date of current version

July 6, 2018.
Digital Object Identifier 10.1109/ACCESS.2018.2844086

Analysis and Optimization of the Novel

Inerter-Based Dynamic Vibration Absorbers
XIAORAN WANG1 , XIANDONG LIU1 , YINGCHUN SHAN1 , YONGJUN SHEN2 , AND TIAN HE 1
1 Schoolof Transportation Science and Engineering, Beihang University, Beijing 100191, China
2 Department of Mechanical Engineering, Shijiazhuang Tiedao University, Shijiazhuang 050043, China

Corresponding author: Tian He (hetian@buaa.edu.cn)

This work was supported by the National Natural Science Foundation of China under Grant 51675023 and Grant 51675021.

ABSTRACT Passive dynamic vibration absorber (DVA) is widely used in structural vibration reduction, and
the higher efficient DVA is always required for some special situations. This paper aims to propose the novel
inerter-based dynamic vibration absorbers (IDVAs) to enhance the performance of the passive DVA. First,
several novel IDVAs are presented by matching the inerter with DVA in different places. Then, the closed-
form optimal parameters of six kinds of IDVAs are obtained based on the classical fixed-point theory. The
obtained parameters demonstrates that all the inerters connected between the primary system and absorber
system do not provide improvement for the performance of DVAs, while all the inerters connected to the earth
can improve the performance of DVAs. Moreover, the comparisons among the IDVAs show that the inerter
connected to the earth in the grounded DVA (IR2 in this paper) performs the best performance in vibration
absorption. More than 30 % improvement can be obtained from IR2 as compared with other IDVAs. Finally,
the further comparison among the IDVAs under white noise excitation also shows that IR2 is superior to other
IDVAs. The results may provide theoretical basis for design of the optimal IDVA in engineering practice.

INDEX TERMS Dynamic vibration absorber, fixed-point theory, inerter, optimization, vibration absorption.

I. INTRODUCTION textbooks. However, the conclusions were approximate opti-

The dynamic vibration absorber (DVA) is a passive vibration
control device attached to a vibrating member (called the
primary system) subjected to an exciting force or motion. Due
to simple structure and low cost, it is widely used in engineer-
ing practice such as transportation, industrial machinery and
bridge, and has become one of the most important devices in
vibration control field.
Although the DVA was proposed 100 years ago, the high
performance DVA has always been focused and studied.
In view of different engineering backgrounds, researchers
have proposed various structural forms and working methods
of the DVA. Ormondroyd and Den Hartog [1] found that a
DVA with damping element could suppress the amplitude of
primary system in broader frequency range, which had been
recognized as the typical Voigt type of DVA. Hahnkamm [2]
derived the optimum natural frequency ratio according to
the optimization criterion, and described this method as
the fixed-point theory. Later, the optimum damping ratio
was obtained by Brock [3] in 1946. Currently, the results
based on the fixed-point theory have become the classic
conclusions which were recorded in vibration engineering
mal solutions of Voigt type of DVA relative to the exact solu-
tions. Nishihara and Asami [4], Asami et al. [5] derived the
exact series solutions for the stiffness and damping coefficient
of Voigt type of DVA, and found that the results according
to the fixed-point theory were very close to the exact series
solutions and the approximate forms were much simpler.
Asami and Nishihara [6], [7] presented a three-element type
of DVA, conducted the optimization design, and found that
the three-element type of DVA had better control performance
under the same mass ratio. Ren [8] presented a grounded
type of DVA with the damping element connected to the
earth (a base structure). The results indicated that it could
present better control performance than the Voigt type of DVA
under the same parameter conditions. The above three kinds
of DVAs are denoted as the typical models.
In order to further improve the performance of the DVA,
Shen et al. [9], [10] studied a new type of DVA with a
negative stiffness and finished its parameter optimization.
It was found that the DVA with negative stiffness had good
control performance. In 2017, Wang et al. [11] introduced the
negative stiffness connected to the earth in the three-element

2169-3536
2018 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 6, 2018 Personal use is also permitted, but republication/redistribution requires IEEE permission. 33169
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

type of the DVA and found that the three-element type of
the DVA with negative stiffness also had good control per-
formance. Many researches about the properties and stability
conditions of the negative stiffness system had been reported
in [12]–[18]. However, the negative stiffness structure is
unstable and difficult to achieve in engineering practice.
Recently, the inerter matching with the DVA (called inerter-
based DVA) has attracted several researchers. The inerter has
been successfully used in the suspension systems of Formula
one vehicle and various mechanical systems to attenuate
vibration [19]. Hu et al. [20] investigated the damping per-
formance of inerter-based isolator and compared the optimal
parameters of inerter-based isolator with the traditional DVA.
Shen et al. [21] investigated the inerter-based DVA added
on the body mass and concluded that the new suspen-
sion structure called inerter-based suspension can effectively
improve the damping performance of the suspension system.
Brzeski et al. [22] investigated how additional damping and
inerter introduced by supplementary devices influence the
system dynamics, and compared the effects caused by addi-
tional damper and inerter. These studies show that the intro-
duction of inerter has a potential advantage in improving the
performance of the DVA.
However, IDVA has not yet been systematically studied.
Moreover, Hu and Chen [23] also concluded that adding
single inerter alone to the traditional DVA under certain cir-
cumstances, no matter it is in parallel connection or in series
connection, provides no improvement for the H∞ optimiza-
tion. Therefore it is needed to systematic study the various
IDVAs and propose the novel optimal IDVAs. In this paper,
five novel kinds of IDVAs are presented by introducing the
inerter connected between the primary system and absorber
system and connected to the earth in the Voigt type of
the DVA, the three-element type of the DVA and the grounded
type of the DVA, respectively. And the performances of
six IDVAs are analyzed and optimized.
The organization of this paper is as follows. The models
of the IDVAs are established in section 2. The closed-form
optimal parameters of the IDVAs are derived in section 3. And
section 4 provides the comparison and analysis of the results.
Then, the presented IDVAs are compared with the typical
DVAs under white noise excitation in section 5. Finally,
conclusions are drawn in section 6.
II. THE ESTABLISHMENT OF IDVAS AND THEIR MODELS
Nomenclature:
m1 : mass of the primary system
m2 : mass of the absorber system
k1 : stiffness of the primary system
k2 : stiffness of the absorber system
k3 : stiffness of the viscoelastic
Maxwell model
c: damping coefficient of the
absorber system
b: inertance
F0 : amplitude of the force excitation
33170
X. Wang et al.: Analysis and Optimization of the Novel IDVAs
ω: frequency of the force excitation
A: normalized amplitude magnification
√ factor of primary system
ω1 = √k1 /m1 : natural frequency of the primary system
ω2 = k2 /m2 : natural frequency of the absorber
system
µ = m2 /m1 : mass ratio
δ = b/m2 : inertance-to-mass ratio
υ = ω2 /ω1 : natural frequency ratio
λ = ω/ω1 : forced frequency ratio
ζ = c/(2m2 ω2 ): damping ratio
k = k3 /k2 : ratio of spring constants
δst = F0 /k1 : static displacement
Three typical models of the DVAs by Ormondroyd and Den
Hartog, Asami and Nishihara, and Ren [1], [6], [8] and all
the optimal parameters of the three models are summarized
in Fig.1 and Tab. 1, respectively. Six kinds of IDVAs are
obtained by introducing the inerter connected between the
primary system and absorber system and connected to the
earth in the three typical models, respectively. The six models
of IDVAs are shown in Fig.2. The models are named as ID1,
ID2, IA1, IA2, IR1 and IR2, respectively. In these six models,
ID1 has been studied by Hu and Chen [23], the other five
novel IDVAs have not been studied yet.
FIGURE 1. The three typical models of DVAs: (a) The type by Den Hartog;
(b) The type by Asami; (c) The type by Ren.
III. PARAMETERS OPTIMIZATION OF IDVAS
A. PARAMETERS OPTIMIZATION OF ID1 AND ID2
Usually, the optimum tuning ratio and the optimum damp-
ing ratio of DVA/IDVA are derived by using fixed-points
theory [1], [6], [8], [23]. In the main system amplitude
normalization curves, all curves pass through two points
independent of the damping. The most favorable curve is
the one which passes with a horizontal tangent through the
highest of the two fixed points P or Q. The best obtainable
‘‘resonant amplitude’’ (at optimum damping) is the ordinate
of that point. In this paper, the fixed-points theory is also
used to optimize six IDVAs. For the optimization design of
the DVA, a great deal of in-depth and meticulous research
has been done in the past. In recent years, Japanese scholar
Asami et al. [5] has done a systematic and in-depth theoret-
ical work. According to the different purpose of application,
VOLUME 6, 2018
X. Wang et al.: Analysis and Optimization of the Novel IDVAs
TABLE 1. The formulas of the three typical models of DVAs.
FIGURE 2. The models of IDVAs.
The optimization of DVA design problems can be divided into
the following three categories:
(1) For the case where the main vibration system is sub-
jected to harmonic excitation, the objective is to minimize the
maximum amplitude magnification factor (called H∞ norm)
of the primary system. This is H∞ optimization problem.
VOLUME 6, 2018
(2) For the case where the main vibration system is sub-
jected to random excitation, the objective is to reduce the total
vibration energy of the system over all frequency.
In this optimization criterion, the area (called H2 norm)
under the frequency response curve of the system is mini-
mized. This is H2 optimization problem.
(3) For the case where the main vibration system
is subjected to shock excitation, the objective of the
stability maximization criterion is to improve the tran-
sient vibration of the system. This is maximum stability
problem.
In this paper, the primary system is subjected to harmonic
excitation, and the H∞ optimization is used as the optimiza-
tion criterion in this paper.
In these six models, ID1 has been studied by
Hu and Chen [23], so this paper gives the optimization results
of ID1 firstly.
For the ID1,the optimal parameters can be analytically
obtained as [23]
√ s
1 + (1 + µ)δID1 3µ
υopt = ζopt =
1+µ 8(1 + µ)
33171
 X. Wang et al.: Analysis and Optimization of the Novel IDVAs

and the resonant amplitudes at the two fixed points are

s
2 + µ + 2δID1 (1 + µ)
 
 X1 
  ∼
=
δ  µ
st max

For the case of ID2 in Fig.2, the differential motion

equation can be established as
m1 ẍ1 + k1 x1 + k2 (x1 − x2 ) + c(ẋ1 − ẋ2 ) = F0 sin(ωt)
m2 ẍ2 + k2 (x2 − x1 ) + c(ẋ2 − ẋ1 ) + bID2 ẍ2 = 0 (1)
where m1 , m2 , k1 , k2 are the mass, linear stiffness coefficient
of the primary system and the DVA respectively. c is the
damping of the absorber, and bID2 is the inertance. x1 and x2
is the displacement of the primary system and the absorber
respectively. F0 and ω is the amplitude and frequency of the
excitation. FIGURE 3. The amplitude-frequency curves under µ = 0.1, υ = 0.9,
Using the following parametric transformation δ = 0.1.
s s
k1 k2 c
ω1 = , ω2 = , ζ =
m1 m2 2m2 ω2 There exist two equal values if ζ = 0 and ζ → ∞
m2 bID2 F0 in (4).
µ= , δID2 = , f =
m1 m2 m1 (1 + δID2 )λ2 − υ 2
A=
Equation (1) becomes (−1−δID2 )λ4 −υ 2 +[1+δID2 +(1+µ + µδID2 )υ 2 ]λ2
( 1
ẍ1 +ω12 x1 +µω22 (x1 −x2 )+2µζ ω2 (ẋ1 − ẋ2 ) = f sin(ω t) =±
(1 + µ + µδID2 )λ2 − 1
(6)
(2)
(1 + δID2 )ẍ2 + ω22 (x2 − x1 ) + 2ζ ω2 (ẋ2 − ẋ1 ) = 0
It could be found that there is no meaning when the right
Letting f sin(ωt) in (2) be represented by fejωt , the steady- part in (6) is positive. Accordingly, taking the negative one
state solutions take the forms as and simplifying the equation, we can get

x1 = X1 ejωt , x2 = X2 ejωt (3) (1 + δID2 )(2 + µ + µδID2 )λ4

Substituting (3) into (2), and introducing the parameters − 2[1 + δID2 +(1+µ + µδID2 )υ 2 ]λ2 + 2υ 2 = 0 (7)

ω ω2 F0 Supposing that the roots of (7) are λP and λQ , we can get

λ= , υ= , δst =
ω1 ω1 k1 the following equation
the steady-state response can be established as (λ2 − λ2P )(λ2 − λ2Q ) = λ4 − (λ2P + λ2Q )λ2 + λ2P λ2Q = 0 (8)
ζ 2 A2ID2 + B2ID2
 2
 X1 
2
A =   = (4) Comparing (7) and (8), we can conclude that
δst ζ 2 CID2
2 + D2
ID2
2[1 + δID2 + (1 + µ + µδID2 )υ 2 ]
where λ2P + λ2Q = (9)
(1 + δID2 )(2 + µ + µδID2 )
AID2 = 2λυ The values at P and Q should be the same in order to get
BID2 = −(1 + δID2 )λ2 + υ 2 the optimum natural frequency ratio
CID2 = 2λυ[1 − (1 + µ + µδID2 )λ2 ] 1 1
DID2 = (1 + δID2 )λ4 + υ 2 =− (10)
(1 + µ + µδID2 )λP − 1
2 (1 + µ + µδID2 )λ2Q − 1
− [1 + δID2 + (1 + µ + µδID2 )υ 2 ]λ2 (5)
Simplifying (10) we can get
By simple deduction of (4), it can be found that the nor-
2
malized amplitude-frequency curve will pass through two λ2P + λ2Q = (11)
points which are independent of the damping ratio. The 1 + µ + µδID2
normalized amplitude-frequency curves under the different Combining (9) and (11), and solving the equation we can
damping ratios as 0.1, 0.2 and 0.3 are shown in Fig. 3. It could obtain the optimum natural frequency ratio
be clearly seen that there exist two common fixed points P √
and Q on all the curves, which are independent of the damping 1 + δID2
υopt = (12)
ratio. 1 + µ + µδID2
33172 VOLUME 6, 2018
X. Wang et al.: Analysis and Optimization of the Novel IDVAs

The two fixed points can be obtained under this condition

as
s !
1 µ(1 + δID2 )
λ2P = 1− (13a)
1 + µ + µδID2 2 + µ + µδID2
s !
1 µ(1 + δID2 )
λ2Q = 1+ (13b)
1 + µ + µδID2 2 + µ + µδID2
Based on the optimum natural frequency ratio, the response
at P and Q can be obtained
s
2 + µ(1 + δID2 )
 
 X1 
A|λp ,λQ =   ∼
= (14)
δst max µ(1 + δID2 )
Next step one could adjust the damping ratio so as to make
the maximum amplitude of the primary system at the fixed
points. The condition can be achieved when the derivatives of FIGURE 4. The amplitude-frequency curve under the optimum
parameters µ = 0.1, υ = 0.9449, δ = 0.1, ζ = 0.202.
the amplitude amplification factor are zero at the two fixed-
points
∂A2  where m1 , m2 , k1 and k2 are the masses and linear stiffness
λ ,λ = 0 (15)
∂λ2 p Q coefficients of the primary system and the DVA respec-
Solving (15) and substituting the optimum natural fre- tively. bIA1 is the inertance. k3 and c are linear stiffness
quency ratio and the optimum negative stiffness ratio into the and damping coefficient of the viscoelastic Maxwell model
results, we can get which is origination of three-element. x1 , x2 and x3 are the
q displacement of the primary system, the absorber and the
µ+µδID2
µ(1 + δID2 )2 (3 − 2+µ+µδ ID2
) split point of spring and damping in Maxwell model respec-
ζp =
2
(16a) tively. F0 and ω is the amplitude and frequency of the force
8(1 + µ + µδID2 )
q
µ+µδID2
excitation.
µ(1 + δID2 )2 (3 + 2+µ+µδ ID2
) Using the following parametric transformation
ζQ2 = (16b)
8(1 + µ + µδID2 ) s s
k1 k2 c m2
Taking an average of ζp2 and ζQ2 , we can get the optimum ω1 = , ω2 = , ζ = , µ=
m1 m2 2m2 ω2 m1
damping ratio
k3 bIA1 F0
s k = , δIA1 = , f =
3µ(1 + δID2 )2 k2 m2 m1
ζopt = (17)
8(1 + µ + µδID2 ) Equation (18) becomes
At this time, the optimum parameters of ID2 are shown as
ẍ1 + ω1 x1 + µω2 (x1 − x2 ) + µkω2 (x1 − x3 )
2 2 2

follows:


 + µδIA1 (ẍ1 − ẍ2 ) = f sin(ωt)

√ s 
1 + δID2 3µ(1 + δID2 )2 (19)
υopt = ζopt = ẍ2 +ω22 (x2 − x1 )+2ζ ω2 (ẋ2 − ẋ3 ) + δIA1 (ẍ2 − ẍ1 ) = 0
1 + µ + µδID2 8(1 + µ + µδID2 )





2ζ ω2 (ẋ3 − ẋ2 ) + kω22 (x3 − x1 ) = 0
s 
2 + µ(1 + δID2 )
 
 X1 
  ∼
=
δ  µ(1 + δID2 )
st max Letting f sin(ωt) in (19) be represented by fejωt , the steady-
Fig. 4 shows the amplitude-frequency curve under the state solutions take the forms as
above optimum parameters. It can be found that the optimiza-
x1 = X1 ejωt , x2 = X2 ejωt , x3 = X3 ejωt (20)
tion purposes are basically achieved.
Substituting (20) into (19), and introducing the
B. PARAMETERS OPTIMIZATION OF MODELS IA1 AND IA2 parameters
For the case of IA1 in Fig.2, the differential motion equation
ω ω2 F0
can be established as λ= , υ= , δst =
 ω1 ω1 k1
 m1 ẍ1 + k1 x1 + k2 (x1 − x2 ) + k3 (x1 − x3 )
the steady-state response can be established as


 + b (ẍ − ẍ ) = F sin(ωt)
IA1 1 2 0
(18)
ζ 2 A2IA1 + B2IA1
 2
m ẍ
2 2 + k (x
2 2 − x 1 ) + c(ẋ2 − ẋ3 ) + bIA1 (ẍ2 − ẍ1 ) = 0  X1 
A2 =   =

(21)


δst ζ 2 CIA1
2 + D2

c(ẋ3 − ẋ2 ) + k3 (x3 − x1 ) = 0 IA1

VOLUME 6, 2018 33173

 X. Wang et al.: Analysis and Optimization of the Novel IDVAs

where Because the three fixed points have nothing to do with the
damping ratio, we could obtain
AIA1 = 2λ[(−1 − δIA1 )λ2 + (1 + k)υ 2 ]  
 X1 
BIA1 = kυ[(1 + δIA1 )λ2 − υ 2 ] |A| =  
δst
CIA1 = 2λ{(1 + δIA1 + µδIA1 )λ4 − (1 + k)υ 2 (1 + δIA1 )λ2 − υ 2
 
 
=
4 2 2 2

(1+δIA1 +µδIA1 )λ +υ −[1+δIA1 +(1+µ)υ ]λ 
− [1 + δIA1 + (1 + k)(1 + µ)υ 2 ]λ2 }
(26)
DIA1 = kυ{(1 + δIA1 + µδIA1 )λ4 + υ 2
when ξ = 0, and, (27), as shown at the top of the next page,
− [1 + δIA1 + (1 + µ)υ 2 ]λ2 } (22) when ξ = ∞.
Accordingly we can get
By simple deduction of (21), it can be found that the nor-
 X1   (2 + k)υ 2 − 2(1 + δIA1 )λ2 
   
malized amplitude-frequency curve will pass through three |A| =   =   (28)
points which are independent of the damping ratio. The δst kυ 2 [1 − (1 + µ)λ2 ] 
normalized amplitude-frequency curves under the different
For simplicity, the three real roots of (24) are supposed
damping ratios as 0.1, 0.2 and 0.3 are shown in Fig. 5. It could
as λ2P , λ2Q and λ2R . Once the values of λP , λQ and λR are
be clearly seen that there exist three common fixed points P, Q
determined, the ordinates of the three fixed points could be
and R on all the curves, which are independent of the damping
obtained
ratio.
  = (2 + k)υ − 2(1 + δIA1 )λP
 
 X1  2 2
δ  kυ 2 [1 − (1 + µ)λ2P ]
st P
(2 + k)υ 2 − 2(1 + δIA1 )λ2Q
 
 X1 
  =−
δ  kυ 2 [1 − (1 + µ)λ2Q ]
st Q

  = (2 + k)υ − 2(1 + δIA1 )λR

 
 X1  2 2
δ  (29)
st R kυ 2 [1 − (1 + µ)λ2R ]
Optimum tuning of IA1 will be accomplished with the
condition that the three fixed points are adjusted to equal
height. At first, by adjusting the fixed points P and R to equal
height, that is to say |X1 /δst |P = |X1 /δst |R , we could obtain
2(1 + δIA1 )
k= −2 (30)
(1 + µ)υ 2
Substituting (30) into (24), we can get
FIGURE 5. The amplitude-frequency curves under µ = 0.1, υ = 0.9,
( 1 + δIA1 )(1 + δIA1 + µδIA1 )λ4
 
κ = 0.7, δ = 0.1.
[(1 + µ)λ2 − 1]  − 2(1 + δIA1 )2 λ2 =0
+ 2(1 + δIA1 )υ 2 − (1 + µ)υ 4
In order to find the values of the fixed points, the following
(31)
equation should hold
AIA1 BIA1 Solving (31) for λ2 yields, (32), as shown at the top of the
= (23) next page.
CIA1 DIA1
Then (29) can be written as
Substituting (22) into (23), we can get
1 + δIA1
 
 X1 
  =
δ  1 + δIA1 − (1 + µ)υ 2
λ6 + qIA1 λ4 + qIA1 λ2 + qIA1 = 0 (24) st P,R

  = (1 + µ)[1 + δIA1 − (1 + µ)υ ]

2
 
 X1 
where δ  (33)
st Q µ
−2(1+δIA1 )2 +(−2−k)(2+2δIA1 +µ + 2µδIA1 )υ 2
qIA1 = Secondly, adjusting the points P (or R) and Q to equal
2(1 + δIA1 )(1 + δIA1 + µδIA1 ) height, that is to say |X1 /δst |P,R = |X1 /δst |Q in (33), we can
2(1 + k)(1 + µ)υ 4 + 2(2 + k)(1 + δIA2 )υ 2 obtain the optimum tuning condition with respect to υ
qIA2 =
2(1 + δIA2 )(1 + δIA2 + µδIA2 ) s
1 + δIA1 µ
r
−(1 + k)υ 4 υopt = (1 − ) (34)
qIA3 = (25) 1+µ (1 + µ)(1 + δIA1 )
(1 + δIA3 )(1 + δIA3 + µδIA3 )
33174 VOLUME 6, 2018
X. Wang et al.: Analysis and Optimization of the Novel IDVAs

(−1 − δIA1 )λ2 + (1 + k)υ 2

   
 X1   
|A| =   =   (27)
δst (1 + δIA1 + µδIA1 )λ − (1 + k)υ − [1 + δIA1 + (1 + k)(1 + µ)υ ]λ 
4 2 2 2
p
(1 + δIA1 )2 − (1 + δIA1 ){(1 + δIA1 )3 + (1 + δIA1 + µδIA1 )υ 2 [(1 + µ)υ 2 − 2(1 + δIA1 )]}
λP =
2
(1 + δIA1 )(1 + δIA1 + µδIA1 )
1
λ2Q =
1+µ
p
(1 + δIA1 )2 + (1 + δIA1 ){(1 + δIA1 )3 + (1 + δIA1 + µδIA1 )υ 2 [(1 + µ)υ 2 − 2(1 + δIA1 )]}
λR =
2
(32)
(1 + δIA1 )(1 + δIA1 + µδIA1 )

Substituting (34) into (30), the optimum value of k could Using the approximation (39) and substituting (34), (35)
be established and (38) into (36), we can get
√
2(µ + µ(1 + µ)(1 + δIA1 )) y4IA1 +dIA11 y3IA1 +dIA12 y2IA1 +dIA13 yIA1 + dIA14 = 0 (40)
kopt = (35)
1 + δIA1 + µδIA1
where
For the condition of the three fixed points with equal
height, optimum damping ratio ξopt is achieved when 1
dIA11 = 2r 7 µ4
the two resonance peaks are adjusted to equal height. 1IA1 (1 + µ) IA1
Rearranging (21), we can get × {1 + µ + rIA1 [−2 + rIA1 (−1 + 2rIA1 )](1 + µ)
 2 + zIA1 + (−2 + rIA1 )rIA1 (1 + 2rIA1 )zIA1 }
  = (1 + yIA1 zIA1 )αIA11 + (1 − yIA1 zIA1 )αIA12 (36)
 X1 
1
δ  (1 + yIA1 zIA1 )αIA13 + (1 − yIA1 zIA1 )αIA14 dIA12 = r 7 µ4
st 1IA1 (1 + µ)2 IA1
where × h−2 (1 + µ) + rIA1 {1 + µ + zIA1
yIA1 = λ2 + rIA1 [−2tIA1 zIA1 + 6(1 + µ + zIA1 )
 2
2ζ + (−6 + tIA1 )rIA1 (1 + µ + zIA1 )] }i
zIA1 = 1
kυ dIA13 = 2r 8 µ4
1IA1 (1 + µ)3 IA1
αIA11 = y2IA1 (1 + δIA1 )2 − (2 + k)yIA1 (1 + δIA1 )υ 2 × {1 + µIA1 + zIA1 + rIA1 [(−3 + tIA1 )(1 + µ)
1 + (−1+tIA1 )zIA1 −(−2+tIA1 )rIA1 (1+µ+zIA1 )]}
+ [2 + k(2 + k)]υ 4
2 1
1 dIA14 = 8
(−1 + tIA1 )rIA1 µ4
αIA12 = kyIA1 (1 + δIA1 )υ 2 − k(2 + k)υ 4 1IA1 (1 + µ)4
2 × [(−1 + rIA1 )2 (1 + µ) + (1 + rIA1
2
)zIA1 ]
αIA13 = y2IA1 [−1 + yIA1 + (−1 + yIA1 + µyIA1 )δIA1 ]2
1IA1 = rIA1
6 2
(1 − rIA1 )µ4 [(rIA1
2
−1)(1+µ+zIA1 )−2rIA1 zIA1 ]
1
+ [2 + k(2 + k)](−1 + yIA1 + µyIA1 )2 υ 4
s
2 (1 + µ)(1 + δIA1 )
rIA1 = (41)
− (2 + k)yIA1 (−1 + yIA1 + µyIA1 )[yIA1 µ
+ (−1 + yIA1 + µyIA1 )δIA1 − 1]υ 2 The (40) is a monic quartic polynomial. The existence of
1
αIA14 = k(−1 + yIA1 + µyIA1 )υ 2 {2yIA1 two double roots leads to the factorized form
2
× [−1 + yIA1 + (−1 + yIA1 + µyIA1 )δIA1 ] F() = ( − A )2 ( − B )2 (42)
− (2 + k)(−1 + yIA1 + µyIA1 )υ } 2
(37) Vieta’s theorem gives the alternative expressions for
In order to find the values of the optimum damping ratio di (i = 1, . . . , 4)
and the resonance amplitude magnification factor, the follow-
dIA11 = −2(A + B ) (43)
ing equation should hold
 2 dIA12 = 2A + 4A B + 2B (44)
 X1  dIA13 = −2A B (A + B ) (45)
  = h2 ≡ 1/tIA1 (38)
δ  IA1
st dIA14 = 2A 2B (46)
where hIA1 denotes the resonance amplitude magnification
factor. By eliminating A and B , these four equations are
The trial simplification is to use the approximation which reduced to two equations
p
Asami presented [11] f1 = dIA11 dIA14 − dIA13 = 0 (47)
1 − yIA1 zIA1 ∼ 1 − yIA1 zIA1 d 2 p
= (39) f2 = IA11 + 2 dIA14 − dIA12 = 0 (48)
1 + yIA1 zIA1 1 + zIA1 /(1 + µ) 4
VOLUME 6, 2018 33175
 X. Wang et al.: Analysis and Optimization of the Novel IDVAs

Substitution of definitions (41) into (47) and (48) yields the where
simultaneous equations {f1 = 0, f2 = 0} with respect to zIA1  2 3 5
and tIA1 . The first equation is rearranged as aIA1 = rIA1 − 2rIA1 + rIA1

4 6 7 8

IA1 = rIA1 − 2rIA1 − 2rIA1 + 3rIA1
b
2
f1 ≡ h( − 1 + rIA1 )[1 + rIA1 (−2 + tIA1 rIA1 )](1 + µ) 2 − 2r
cIA1 = rIA1
 IA1 − 1
2 2
× {−1+rIA1 [1+2rIA1 +tIA1 (−1−2rIA1 +rIA1 )]} zIA1 i

 2 − r 4 + r 5 − 2r 6 + r 7
dIA1 = rIA1 IA1 IA1 IA1 IA1
× {4rIA1 [(−1 + rIA1 )(1 + µ) − zIA1 ]zIA1
2 s
+ tIA1 (−1 + rIA1 )3 (1 + rIA1 )(1 + µ + zIA1 )2 } = 0 (1 + µ)(1 + δIA1 )
rIA1 =
(49) µ

Equation (49) is solved with respect to zIA1 Fig. 6 shows the amplitude-frequency curve under the
above optimum parameters. It can be found that the optimiza-
IA1 (−2 + tIA1 rIA1 )](1 + µ)
2 )[1 + r
(−1 + rIA1 tion purposes are basically achieved.
zIA1 = 2 {1 + 2r
1 − rIA1 IA1 + tIA1 [−1 + (−2 + rIA1 )rIA1 ]}
(50)
Substitution of (50) into f2 = 0 eliminates zIA1 in f2
f2 ≡ 2 + rIA1 h− 1 + rIA1 {−3 + tIA1 rIA1
2
× [−2+rIA1 (6 − 2tIA1 −tIA1 rIA1 +tIA1 rIA1 )] }i = 0
(51)
Equation (51) is solved with respect to tIA1
q
3 − 3r 4 + 2 r 4 − 2r 6 − 2r 7 + 3r 8
rIA1 IA1 IA1 IA1 IA1 IA1
tIA1opt = 6 − r 5 − 2r 4
rIA1 IA1 IA1
(52)
Substituting (52) into (50), we can get
2 )(1 + µ)(a
√
(−1 + rIA1 IA1 − bIA1 )
zIA1opt = √ 2
(53)
cIA1 bIA1 + dIA1 − rIA1 FIGURE 6. The amplitude-frequency curve under the optimum
parameters µ = 0.1, υ = 0.844, k = 0.8069, ζ = 0.2803, δ = 0.1.
where
aIA1 = 2
rIA1 3
− 2rIA1 5
+ rIA1 For the case of IA2 in Fig.2, the differential motion
bIA1 = 4 6 7
rIA1 − 2rIA1 − 2rIA1 8
+ 3rIA1 equation can be established as
2

cIA1 = rIA1 − 2rIA1 − 1 m1 ẍ1 + k1 x1 +k2 (x1 −x2 )+k3 (x1 − x3 ) = F0 sin(ωt)

2 4 5 6 7
dIA1 = rIA1 − rIA1 + rIA1 − 2rIA1 + rIA1 (54) m2 ẍ2 + k2 (x2 − x1 ) + c(ẋ2 − ẋ3 ) + bIA2 ẍ2 = 0 (57)

c(ẋ3 − ẋ2 ) + k3 (x3 − x1 ) = 0

Then, the optimum damping and the minimized resonance
amplitude are obtained the steady-state response can be established as
1p  2
ζ 2 A2IA2 + B2IA2
ζopt = kopt υopt zIA1opt 2
 X1 
A =   = (58)
2 δst
s √ ζ 2 CIA2
2 + D2
IA2
rIA1 µ(1 + rIA1 ) aIA1 − bIA1
= · √ (55) where
1+µ cIA1 bIA1 + dIA1
AIA2 = 2λ[(1 + k)υ 2 − (1 + δIA2 )λ2 ]
s s
  6 − r 5 − 2r 4
rIA1
 X1  1 IA1
 
δ  = = 3 4 +2 b
√IA1 (56) BIA2 = kυ[λ2 (1 + δIA2 ) − υ 2 ]
st max tIA1opt rIA1 − 3rIA1 IA1
CIA2 = 2λ{(1 + δIA2 )λ2 (−1 + λ2 )
At this time, the optimum parameters of IA1 are shown as
− (1 + k)[−1 + (1 + µ + µδIA2 )λ2 ]υ 2 }
follows:
√ DIA2 = kυ{(1 + δIA2 )λ4 − (1 + δIA2 )λ2
2(µ + µ(1 + µ)(1 + δIA1 ))
kopt =
1 + δIA1 + µδIA1 + [1 − (1 + µ + µδIA2 )λ2 ]υ 2 } (59)
s
1 + δIA1
r
µ According to the parameter optimization procedures
υopt = (1 − ) of IA1, we can get the optimum parameters of IA2:
1+µ (1 + µ)(1 + δIA1 ) v
√
s u s
u 1+δ µ(1 + δIA2 )
rIA1 µ(1 + rIA1 ) aIA1 − bIA1 υopt = t
IA2
ζopt = · √ 1 + µ + µδIA2
(1 −
1 + µ + µδIA2
) (60)
1+µ cIA1 bIA1 + dIA1

33176 VOLUME 6, 2018

X. Wang et al.: Analysis and Optimization of the Novel IDVAs

kopt = 2 µ(1 + δIA2 )(1 + µ + µδIA2 ) + 2µ(1 + δIA2 )

p
where
(61)
s √ AIR1 = 2λυ
1+µ aIA2 − bIA2 BIR1 = −(1 + δIR1 )λ2 + υ 2
ζopt ∼
= · (62)
µ(rIA2 − 1) cIA2 CIR1 = 2λυ[(1 + µδIR1 )λ2 − µυ 2 − 1]
s
DIR1 = [1 + δIR1 + µδIR1 ]λ4 + υ 2
 
 X1  dIA2
  ∼
= √ (63)
δ 
st max eIA2 − 2 bIA2 − [1 + δIR1 + (1 + µ)υ 2 ]λ2 (68)
where According to the parameter optimization procedures
aIA2 = rIA2 + 3rIA2 − 1 + µ(rIA2 + rIA2 + rIA2 − 1)
3 3 2

 of ID2, we can get the optimum parameters of IR1:
4 3 2


 bIA2 = 12rIA2 − 8rIA2 + 4rIA2 − 4rIA2 + 3 s
1 + δIR1 + µδIR1


+ µ(12rIA24 − 12r 3 + 6r 2 − 8r
υopt =

IA2 + 6) (69)

IA2 IA2

1 − µ − µ2 δIR1


+ µ (3rIA2 − 4rIA2 + 2rIA2 − 4rIA2 + 3)
2 4 3 2



 s
6 − 2r 5 + 2r 3 + r 2
cIA2 = rIA2 (64) µ(2rIR1 + µ)
IA2 IA2 IA2 ζopt = (70)
+ µ(2rIA2 − 2rIA2 + 4rIA2 + 2rIA2 + 2rIA2 )
6 5 4 3 2 4(rIR1 + µ − rIR1 µ)




 √
+ µ2 (rIA2
6 + 2r 4 + r 2 )

(1 − µ − µ2 δIR1 ) 3
 

 IA2 IA2  X1 


2 +2r 3 −2r 4 +µ(3r 2 + 2r 3 − r 4 )
  ∼
= (71)

dIA2 = 3rIA2
 δ  µ(1 + µδIR1 )2

 IA2 IA2 IA2 IA2 IA2 st max

eIA2 = 3 − 2rIA2 + 6rIA2 2 + µ(3 − 2r 2
IA2 − 3rIA2 ) where
s
1 + µ(1 + δIA2 )
rIA2 = (65) 3 = µ{2 + δIR1 [2 + µ(2 + δIR1 )]} (72)
µ(1 + δIA2 )
rIR1 = µ(1 + µδIR1 ) (73)
Fig. 7 shows the amplitude-frequency curve under the
above optimum parameters. It can be found that the optimiza- Fig. 8 shows the amplitude-frequency curve under the
tion purposes are basically achieved. above optimum parameters. It can be found that the optimiza-
tion purposes are basically achieved.

FIGURE 7. The amplitude-frequency curve under the optimum

parameters µ = 0.1, υ = 0.8240, k = 0.9189, ζ = 0.3047, δ = 0.1.
FIGURE 8. The amplitude-frequency curve under the optimum
parameters µ = 0.1, υ = 1.111, ζ = 0.199, δ = 0.1.

C. PARAMETERS OPTIMIZATION OF MODELS IR1 AND IR2

For the case of IR2 in Fig.2, the differential motion equa-
For the case of IR1 in Fig.2, the differential motion equation tion can be established as
can be established as (
( m1 ẍ1 + k1 x1 + k2 (x1 − x2 ) = F0 sin(ωt)
m1 ẍ1 +k1 x1 +k2 (x1 −x2 )+bIR1 (ẍ1 − ẍ2 ) = F0 sin(ωt) (74)
(66) m2 ẍ2 + k2 (x2 − x1 ) + cẋ2 + bIR2 ẍ2 = 0
m2 ẍ2 + k2 (x2 − x1 ) + cẋ2 + bIR1 (ẍ2 − ẍ1 ) = 0
the steady-state response can be established as the steady-state response can be established as

ζ 2 A2IR1 + B2IR1 ζ 2 A2IR2 + B2IR2

 2  2
 X1   X1 
A2 =   = (67) A2 =   = (75)
δst ζ 2 CIR1
2 + D2
IR1
δst ζ 2 CIR2
2 + D2
IR2

VOLUME 6, 2018 33177

 X. Wang et al.: Analysis and Optimization of the Novel IDVAs

where
AIR2 = 2λυ
BIR2 = −(1 + δIR2 )λ2 + υ 2
CIR2 = 2λυ[λ2 − µυ 2 − 1]
DIR2 = (1 + δIR2 )λ4 + υ 2
− [1 + δIR2 + (1 + µ + µδIR2 )υ 2 ]λ2 (76)
According to the parameter optimization procedures
of ID2, we can get the optimum parameters of IR2:
s
1 + δIR2
υopt = (77)
1 − µ − µδIR2
s
(1 − rIR2 )2 √ p p
ζopt = √ ·( 2+ 31 + 32 ) (78)
4 2µ(1 + rIR2 ) FIGURE 10. Comparison between the Voigt type of DVA by Den Hartog
√
(1 − µ − µδIR2 ) 2(µ + µδIR2 ) and ID1 with µ = 0.1 but different δ.
 
 X1 
  ∼
= (79)
δ  µ + µδIR2
st max
where
( √
31 = 3 − rIR2 − 2 2 − 2rIR2
√ (80)
32 = 3 − rIR2 + 2 2 − 2rIR2
rIR2 = 1 − µ − µδIR2
Fig. 9 shows the amplitude-frequency curve under the
above optimum parameters. It can be found that the optimiza-
tion purposes are basically achieved.
FIGURE 9. The amplitude-frequency curve under the optimum
parameters µ = 0.1, υ = 1.112, ζ = 0.219, δ = 0.1.
IV. ANALYSIS OF THE RESULTS
Hu and Chen [23] found that ID1 provide no improvement
as compared with the Voigt type of the DVA by Den Hartog.
With the increase of inerter-to-mass ratio, the performance
worsens. Such an observation is shown in Fig. 10.
The comparison between the Voigt type of the DVA by Den
Hartog and ID2 with different δ is shown in Fig. 11. With
the increase of the inerter-to-mass ratio, the performance
becomes better. So it is concluded that the case of ID2 provide
33178
(81)
FIGURE 11. Comparison between the Voigt type of DVA by Den Hartog
and ID2 with µ = 0.1 but different δ.
improvement as compared with the Voigt type of the DVA by
Den Hartog.
Comparison between Fig. 10 and 11 clearly shows that the
amplitude magnification factor of primary system increases
with increase in inerter-to-mass ratio in the case of ID1,
and the amplitude magnification factor of primary system
decreases with increase in inerter-to-mass ratio in the case
of ID2. So the performace of these models may be ranked as:
ID2, the Voigt type of the DVA by Den Hartog, ID1 under the
same mass ratio and inertance-to-mass ratio.
Fig. 12 and 13 show the comparison between the
three-element type of the DVA by Asami and IA1 with
different δ and the comparison between the three-element
type of the DVA by Asami and IA2 with different δ, respec-
tively. Further, comparing Fig. 12 with Fig. 13 may reach
that the amplitude magnification factor of primary system
increases with the inerter-to-mass ratio in IA1, and the ampli-
tude magnification factor of primary system decreases with
inerter-to-mass ratio in IA2. So the performace of these
models may be ranked as: IA2, the three-element type of DVA
VOLUME 6, 2018
X. Wang et al.: Analysis and Optimization of the Novel IDVAs
FIGURE 12. Comparison between the three-element type of DVA by
Asami and IA1 with µ = 0.1 but different δ.
FIGURE 13. Comparison between the three-element type of DVA by
Asami and IA2 with µ = 0.1 but different δ.
by Asami, IA1 under the same mass ratio and inertance-to-
mass ratio.
Fig. 14 and 15 show the comparison between the grounded
type of the DVA by Ren and IR1 with different δ and the
comparison between the grounded type of the DVA by Ren
and IR2 with different δ, respectively. Further, the Compar-
ison between Fig. 14 and Fig. 15 is performed, and it can
be found that the amplitude magnification factor of primary
system increases with inerter-to-mass ratio in IR1, and the
amplitude magnification factor of primary system decreases
with inerter-to-mass ratio in IR2. So the performace of these
models may be ranked as: IR2, the grounded type of the DVA
by Ren, IR1 under the same mass ratio and inertance-to-mass
ratio.
As mentioned above, comparison between the IDVAs and
typical DVAs under the same mass ratio and inertance-to-
mass ratio shows that the inerter connected between the pri-
mary system and absorber system provides no improvement
for the performance of the DVAs. On the contrary, the inerter
VOLUME 6, 2018
FIGURE 14. Comparison between the grounded type of DVA by Ren and
IR1 with µ = 0.1 but different δ.
FIGURE 15. Comparison between the grounded type of DVA by Ren and
IR2 with µ = 0.1 but different δ.
connected to the earth can improve the performance of the
DVAs, and IR2 shows best effect. The performance is ranked
as IR2, IA2, then ID2. Such a conclusion is also confirmed
by Fig. 16 under µ = 0.1, δ = 0.1.
However, since the mass ratio is normally quite small and
practically less than 0.25, the optimal maximum amplitude
of primary system are plotted in Fig. 17 over the range of
0 < µ ≤ 0.25. As shown in Fig. 18, if using IR2, 32 %
improvement can be attained as compared with ID2 when
µ = 0.25, δ = 0.1, and 30 % improvement can also
be obtained as compared with IA2. So IR2 is superior to
ID2 and IA2.
V. THE RESPONSE OF THE PRIMARY SYSTEM UNDER
WHITE NOISE EXCITATION
The primary system is usually subject to random excitation in
practical engineering, so it is very important and meaningful
to investigate the system response to random excitation. For
convenience, here we just investigate the primary system
33179
 X. Wang et al.: Analysis and Optimization of the Novel IDVAs

DVAs and the IDVAs are, respectively

SD (ω) = |X1D |2 S0 , SID1 (ω) = |X1ID1 |2 S0 ,
SID2 (ω) = |X1ID2 |2 S0
SA (ω) = |X1A |2 S0 , SIA1 (ω) = |X1IA1 |2 S0 ,
SIA2 (ω) = |X1IA2 |2 S0
SR (ω) = |X1R2 |2 S0 , SIR1 (ω) = |X1IR1 |2 S0 ,
SIR2 (ω) = |X1IR2 |2 S0 (82)
Further, comparing Fig. 12 with Fig. 13 may reach that the
where the subscripts D, A and R represent the DVA model
by Den Hartog, the model by Asami and Ren respectively.
The subscripts ID1,2, IA1,2 and IR1,2 represent the IDVAs
proposed in this paper. The mean square response of the
primary system of these DVAs and IDVAs can be deduced
FIGURE 16. The comparison among ID2, IA2 and IR2 under µ = 0.1, as
δ = 0.1. Z ∞ Z ∞
π S0 YD
σD =
2
SD (ω)dω = S0 |X1D |2 dω =
−∞ −∞ 2ω13 µζ υ
π S0 YID1
Z ∞ Z ∞
σID1
2
= SID1 (ω)dω = S0 |X1ID1 |2 dω =
1 µζ υ
2ω 3
−∞ −∞
Z ∞ Z ∞
σID2
2
= SID2 (ω)dω = S0 |X1ID2 |2 dω
−∞ −∞
π S0 YID2
=
2ω3 µζ υ(1 + δID2 )2
Z ∞1 ∞ πS0 YA
Z
σA2 = SA (ω)dω = S0 |X1A |2 dω =
−∞ −∞ 2ω13 µζ k 2 υ 3
Z ∞ Z ∞
σIA1
2
= SIA1 (ω)dω = S0 |X1IA1 |2 dω
−∞ −∞
π S0 YIA1
=
2ω3 µζ υ 3 k 2 (1 + δIA1 + µδIA1 )
Z ∞1 Z ∞
FIGURE 17. Optimal maximum amplitude of primary system with
different mass ratio under δ = 0.1. σIA2
2
= SIA2 (ω)dω = S0 |X1IA2 |2 dω
−∞ −∞
π S0 YIA2
=
2ω3 µξ υ 3 k 2 (1 + δIA2 )2
Z ∞1 Z ∞
σR2 = SR (ω)dω = S0 |X1R |2 dω
−∞ −∞
π S0 YR
=
2ω3 µζ υ 5
Z ∞1 Z ∞
σIR1
2
= SIR1 (ω)dω = S0 |X1IR1 |2 dω
−∞ −∞
π S0 YIR1
=
2ω3 µζ υ(υ 2 − δIR1 )2
Z ∞1 ∞ π S0 YIR2
Z
σIR2
2
= SIR2 (ω)dω = S0 |X1IR2 |2 dω =
−∞ −∞ 2ω13 µζ υ 5
(83)
FIGURE 18. Percentage improvement of IDVAs under δ = 0.1. where
YD = 1 + υ 4 (1 + µ)2 + υ 2 (4µζ 2 + 4ζ 2 − µ − 2)
subject to white noise excitation with zero mean value and
constant power spectral density (S(ω) = S0 ). Then, PSD YID1 = (1 + µ)2 υ 4 + (1 + δID1 )2
function of the displacement response of the three typical + υ 2 [−2 − µ + 4ζ 2 (1 + µ) − 2(1 + µ)δID1 ]

33180 VOLUME 6, 2018

X. Wang et al.: Analysis and Optimization of the Novel IDVAs

YID2 = (1 + δID2 )2 + (1 + µ + µδID2 )2 υ 4 + (1 + δID1 )2 decreases with inerter-to-mass ratio in ID2, IA2 and IR2.
+ υ 2 [4ζ 2 (1 + µ + µδID2 ) Then the comparisons among ID2, IA2 and IR2 show that the
IR2 performs the best performance in vibration absorption.
− (1 + δID2 )(2 + µ + µδID2 )]
h i More than 30 % improvement can be obtained in IR2 as
YA = k 2 υ 2 1 − (2 + µ)υ 2 + (1 + µ)2 υ 4 compared with ID2 and IA2. These results may provide
n io foundation for the design of IDVA in engineering practice.
+ 4ζ 2 1 + (1 + k)υ 2 [−2+ (1 + k)(1 + µ)υ 2
YIA1 = 4ζ 2 [(1+δIA1 )3 +(1+k)2 (1+µ)υ 4 (1+δIA1 +µδIA1 ) ACKNOWLEDGMENT
The authors would like to thank the associate editor and the
− 2(1 + k)υ (1 + δIA1 )(1 + δIA1 + µδIA1 )]
2
anonymous reviewers for their careful reading and helpful
+k υ (1 + δIA1 + µδIA1 )
2 2
comments.
× [(1 + µ)2 υ 4 + (1 + δIA1 )2
− υ 2 (2 + µ + 2δIA1 + 2µδIA1 )] REFERENCES
[1] J. Ormondroyd and J. P. Den Hartog, ‘‘The theory of the dynamic vibration
YIA2 = 4ζ [−2(1 + k)υ (1 + δIA2 ) + (1 + δIA2 )
2 2 2
absorber,’’ J. Appl. Mech., vol. 50, no. 7, pp. 9–22, Jan. 1928.
+ (1 + k)2 υ 4 (1 + µ + µδIA2 )] [2] E. Hahnkamm, ‘‘The damping of the foundation vibrations at varying
excitation frequency,’’ Master Archit., vol. 4, pp. 192–201, 1932.
+ k 2 υ 2 [(1 + δIA2 )2 + υ 4 (1 + µ + µδIA2 )2 [3] J. E. Brock, ‘‘A note on the damped vibration absorber,’’ J. Appl. Mech.,
vol. 13, no. 4, p. A284, 1946.
− υ 2 (1 + δIA2 )(2 + µ + µδIA2 )] [4] O. Nishihara and T. Asami, ‘‘Closed-form solutions to the exact opti-
YR = 1 + υ 4 + (µ + 4ζ 2 − 2) mizations of dynamic vibration absorbers (minimizations of the maxi-
mum amplitude magnification factors),’’ J. Vib. Acoust., vol. 124, no. 4,
YIR1 = υ 4 + (1 + δIR1 )2 + υ 2 pp. 576–582, Apr. 2002.
[5] T. Asami, O. Nishihara, and A. M. Baz, ‘‘Analytical solutions to H∞ and
× [−2 + µ − 2δIR1 + 4ζ 2 (1 + µδIR1 )] H2 optimization of dynamic vibration absorbers attached to damped linear
YIR2 = υ 4 + (1 + δIR2 )2 + υ 2 systems,’’ J. Vib. Acoust., vol. 124, no. 2, pp. 284–295, Apr. 2002.
[6] T. Asami and O. Nishihara, ‘‘Analytical and experimental evaluation of
× [4ζ 2 + (1 + δIR2 )(−2 + µ + µδIR2 )] (84) an air damped dynamic vibration absorber: Design optimizations of the
three-element type model,’’ J. Vib. Acoust., vol. 121, no. 3, pp. 334–342,
The mean square response of the primary system under the Apr. Jul. 1999.
optimal parameters can be respectively obtained for µ = 0.1, [7] T. Asami and O. Nishihara, ‘‘H2 optimization of the three-element type
dynamic vibration absorbers,’’ J. Vib. Acoust., vol. 124, no. 4, pp. 583–592,
δ = 0.1, as follows: Apr. Oct. 2002.
[8] M. Z. Ren, ‘‘A variant design of the dynamic vibration absorber,’’ J. Sound.
6.401π S0 6.718π S0 6.099π S0
σD2 = , σID1
2
= , σID2
2
= Vib., vol. 245, no. 4, pp. 762–770, Apr. 2001.
ω13 ω13 ω13 [9] Y. Shen, X. Wang, S. Yang, and H. Xing, ‘‘Parameters optimization for
a kind of dynamic vibration absorber with negative stiffness,’’ Math.
6.030π S0 6.341π S0 5.755π S0
σA2 = , σIA1
2
= , σIA2
2
= Problem Eng., vol. 2016, Jul. 2016, Art. no. 9624325.
ω13 ω13 ω13 [10] Y. Shen, H. Peng, X. Li, and S. Yang, ‘‘Analytically optimal parameters
of dynamic vibration absorber with negative stiffness,’’ Mech. Syst. Signal
5.780π S0 5.958π S0 5.445π S0 Process., vol. 85, no. 15, pp. 193–203, Feb. 2017.
σR2 = , σIR1
2
= , σIR2
2
=
ω13 ω13 ω13 [11] X. R. Wang, Y. J. Shen, S. P. Yang, and H. J. Xing, ‘‘Parameters opti-
mization of three-element type dynamic vibration absorber with negative
(85) stiffness,’’ J. Vib. Eng., vol. 30, no. 2, pp. 177–184, Apr. 2017.
[12] R. S. Lakes, ‘‘Extreme damping in composite materials with a nega-
Above equations show that IR2 is with smallest coefficient tive stiffness phase,’’ Phys. Rev. Lett., vol. 86, no. 13, pp. 2897–2900,
as 5.445, so it also shows excellent performance under the Mar. 2001.
[13] R. S. Lakes, ‘‘Extreme damping in compliant composites with a negative-
white noise excitation. stiffness phase,’’ Phil. Mag. Lett., vol. 81, no. 2, pp. 95–100, Nov. 2001.
[14] R. S. Lakes, T. Lee, A. Bersie, and Y. C. Wang, ‘‘Extreme damping in
VI. CONCLUSION composite materials with negative-stiffness inclusions,’’ Nature, vol. 410,
no. 6828, pp. 565–567, Mar. 2001.
In this paper, IDVAs are proposed by introducing the inerter [15] R. S. Lakes and W. J. Drugan, ‘‘Dramatically stiffer elastic composite
connected between the primary system and absorber system materials due to a negative stiffness phase,’’ J. Mech. Phys. Solids., vol. 50,
and connected to the earth in the three typical DVAs, respec- no. 5, pp. 979–1009, May 2002.
[16] Y.-C. Wang and R. Lakes, ‘‘Extreme stiffness systems due to negative
tively. The performance of six kinds of IDVAs is investi- stiffness elements,’’ Amer. J. Phys., vol. 72, no. 1, pp. 40–50, Jan. 2004.
gated by using the fixed-points theory, and the closed-form [17] Y.-C. Wang and R. Lakes, ‘‘Negative stiffness-induced extreme viscoelas-
optimal parameters of six kinds of IDVAs are obtained. The tic mechanical properties: Stability and dynamics,’’ Philos. Mag., vol. 84,
no. 35, pp. 3785–3801, Jan. 2004.
comparisons between the typical DVAs and the IDVAs under [18] Y.-C. Wang and R. Lakes, ‘‘Stability of negative stiffness viscoelastic
the harmonic and white noise excitation show that the inerter systems,’’ Quart. Appl. Math., vol. 63, no. 1, pp. 34–55, Mar. 2005.
connected between the primary system and absorber system [19] M. C. Smith, ‘‘Synthesis of mechanical networks: The inerter,’’ IEEE
Trans. Autom. Control, vol. 47, no. 10, pp. 1648–1662, Oct. 2002.
(ID1, IA1 and IR1 in this paper) provides no improvement [20] Y. Hu, M. Z. Q. Chen, Z. Shu, and L. Huang, ‘‘Analysis and optimisation
for the performance of DVAs. On the contrary, the inerter for inerter-based isolators via fixed-point theory and algebraic solution,’’
connected to the earth (ID2, IA2 and IR2 in this paper) can J. Sound Vib., vol. 346, pp. 17–36, Jun. 2015.
[21] Y. Shen, L. Chen, X. Yang, D. Shi, and J. Yang, ‘‘Improved design of
improve the performance of DVAs. It also been observed dynamic vibration absorber by using the inerter and its application in vehi-
that the amplitude magnification factor of primary system cle suspension,’’ J. Sound. Vib., vol. 361, no. 20, pp. 148–158, Jan. 2016.

VOLUME 6, 2018 33181

 X. Wang et al.: Analysis and Optimization of the Novel IDVAs
[22] P. Brzeski, T. Kapitaniak, and P. Perlikowski, ‘‘Novel type of tuned mass
damper with inerter which enables changes of inertance,’’ J. Sound. Vib.,
vol. 349, no. 4, pp. 56–66, Aug. 2015.
[23] Y. Hu and M. Z. Q. Chen, ‘‘Performance evaluation for inerter-based
dynamic vibration absorbers,’’ Int. J. Mech. Sci., vol. 99, pp. 297–307,
Aug. 2015.
XIAORAN WANG received the degree in engi-
neering from Inner Mongolia University for the
Nationalities, Inner Mongolia, China, in 2014,
and the master’s degree in mechanical engi-
neering from Shijiazhuang Tiedao University,
Shijiazhuang, China, in 2017. He is currently pur-
suing the Ph.D. degree with the Department of
Automotive Engineering, Beihang University. His
current research interests include noise and vibra-
tion control and fault diagnosis.
XIANDONG LIU received the B.S. degree in
automobile engineering and the M.S. degree in
computational mechanics from the Jilin Uni-
versity of Technology (now, Jinlin University),
in 1986 and 1989, respectively, and the Ph.D.
degree in aerospace propulsion theory and engi-
neering from Beihang University, China, in 1999.
He is currently a Professor with the School of
Transportation Science and Engineering, Beihang
University. His research interests include vehicle
system dynamics, noise and vibration control, fault diagnosis, acoustic emis-
sion, and vibration signal processing.
33182
YINGCHUN SHAN received the Ph.D. degree
in aerospace propulsion theory and engineering
from Beihang University, Beijing, China, in 2002.
She is currently with Beihang University. Her cur-
rent research interests include control of structure
vibration and noise.
YONGJUN SHEN received the degree in engi-
neering from the Tianjin University of China,
Tianjin, China, in 1995, the master’s degree in
mechanical engineering from Shijiazhuang Tiedao
University, Shijiazhuang, China, in 2002, and the
Ph.D. degree in the mechanical engineering from
Beijing Jiaotong University, Beijing, China,
in 2006. His research interests include dynamic
analysis and vibration control of mechanical sys-
tems, fault diagnosis, and signal processing of
rotating machinery.
TIAN HE received the B.S. and M.S. degrees in
mechanical engineering from Shijiazhuang Tiedao
University, China, in 2001 and 2004, respectively,
and the Ph.D. degree in aerospace propulsion
theory and engineering from Beihang University,
China, in 2008. He is currently an Associate Pro-
fessor with the School of Transportation Science
and Engineering, Beihang University. His research
interests include fault diagnosis and vibration con-
trol of mechanical systems.
VOLUME 6, 2018