Control Engineering Practice 37 (2015) 55–66

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

Active vibration isolation using negative stiffness and displacement

cancellation controls: Comparison based on vibration
isolation performance
Mhia Md. Zaglul Shahadat a,n, Takeshi Mizuno b, Yuji Ishino b, Masaya Takasaki b
a
Graduate school of Science and Engg, Saitama University, Shimo-Okubo 255, Sakuraku, Saitama 338-8570, Japan
b
Department of Mechanical Engineering, Saitama University, Shimo-Okubo 255, Sakuraku, Saitama 338-8570, Japan

art ic l e i nf o a b s t r a c t

Article history: Two vibration control techniques, negative stiffness and displacement cancellation, are characterized
Received 19 February 2014 and employed to achieve active vibration isolation. A horizontal vibration isolation system is developed
Accepted 3 December 2014 and then used to investigate each of these techniques theoretically and experimentally. The respective
responses of the developed system are measured separately while closed-loop poles of the system are
Keywords: kept unchanged; these measured responses are compared. The developed system is based on a series
Zero compliance combination of two isolators and consists of two moving tables; one of the moving tables is mounted on
Infinite stiffness the series-connected isolators and the other is placed between the isolators. Using the negative stiffness
Negative stiffness technique, the isolators are controlled so that one of them has a negative stiffness and the other has
Displacement cancellation
a positive stiffness of equal absolute magnitude; using the displacement cancellation technique, one
isolator is set to cancel displacement while the other behaves as a positive-stiffness isolator. The active
negative, positive, and displacement cancellation isolators are created using voice-coil motors guided by
a negative stiffness controller, proportional derivative controller and integral-proportional derivative
controller, respectively.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction isolation system can be used to cancel out a dislocation in a con-

struction vehicle on an inclined plane.
Precision vibration isolation plays an important role in acquiring There are two main sources for vibration: (i) direct disturbance
the position accuracy in many advanced production and measurement (i.e., on-board disturbance) and (ii) ground vibration (i.e., vibration
industries (Rivin, 1995; Yoshioka, Takahashi, Katayama, Imazawa, & transmitted from the ground to the isolation stage). A vibration
Murai, 2001). During the last three decades, the position and dimen- isolation system should simultaneously be capable of suppressing
sion accuracies in high-tech manufacturing processes have progressed both of these kinds of disturbances. The vibration isolation char-
to the submicron level (Long, Jiang, & Meng, 2013). Such rapid acteristics of passive and active systems have been the focus of
improvements in manufacturing processes have increased the impor- several previous investigations (Trumper & Sato 2002; Rivin, 2003).
tance of research on high-performance vibration isolation systems. Rivin has shown that a vibration isolation system with infinite
Additionally, modern construction vehicles and machineries, such as stiffness and a vibration isolation system with zero stiffness are ideal
bulldozer, excavator, and crane usually, use passive seat-suspension for reducing direct disturbances and ground vibrations, respectively
rather than vehicle-suspension; a horizontal-seat-suspended vehicle (Rivin, 2003). However, a trade-off between high stiffness and low
with a mechanical spring (i.e., passive seat suspension) is subjected to stiffness is inevitable in a conventional passive-type vibration isola-
self-dislocation on an inclined surface. Inherently, the operators of tion system (Yoshioka et al., 2001). Therefore, the applications of a
construction vehicles often confront self-dislocating vibrations that are passive-type vibration isolation system are limited. In contrast, an
responsible of causing premature tiredness and muscular discomfort. active vibration isolation system, at least in theory, is not vulnerable
Because vibration isolation systems typically return to an original or to this problem (Yasuda & Ikeda, 1993). However, an active vibration
set position after being displaced (Tokhi & Hossain, 1996), a vibration isolation system requires high-performance sensors, such as a servo-
accelerometer, which are expensive and make these systems rather
costlier than passive systems. A horizontal vibration isolation system
that utilizes low-cost displacement sensors is thus developed in this
n
Corresponding author. study. Previously, several approaches had been applied to realize a
E-mail address: [email protected] (M.Md.Z. Shahadat). simultaneously infinite stiffness and soft stiffness suspension using

http://dx.doi.org/10.1016/j.conengprac.2014.12.004
0967-0661/& 2014 Elsevier Ltd. All rights reserved.
56 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

an active vibration isolation system; among them, Ho, Matsuhisa, experimentally. The respective responses of the developed experi-
and Honda (2000) used a piezoelectric actuator, Trumper and Sato mental system with both techniques individually are measured
(2002) used an audio speaker, and Jaensch and Lamperth (2007) and the measured responses are compared while the closed-loop
used a magnetic resonance imaging (MRI) scanner. Recently, the poles are kept unchanged in the both cases.
authors have proposed an approach that is based on a series com- This paper is organized as follows: Section 2 illustrates the
bination of two isolators (Mizuno, 2001; Mizuno, Takasaki, & basic principle of vibration isolation using the negative stiffness
Murashita, 2005; Hoque, Mizuno, Takasaki, & Ishino, 2006; Hoque, and displacement cancellation techniques. The designs of the
Mizuno, Takasaki, & Ishino, 2010); this approach is used to develop a controllers associated with the negative stiffness and displace-
horizontal vibration isolation system in this study. Various control ment cancellation techniques are described in Section 3. The
methods, including state feedback control (Miyazaki, Mizuno, theoretical analyses of the developed vibration isolation system
Kawatani, & Hamada, 1994), H1 control (Watanabe et al., 1996), utilizing both control techniques are presented in Section 4. In
feedforward control (Mohamed, Martins, Tokhi, Sada Costa, & Botto, Section 5, the details of the developed experimental system and
2002), repetitive control (Daley, Hatonen, & Owens 2006), active experimental results are presented, followed by the conclusions of
acceleration control (Zhu, Tryggvason, & Piedboeuf, 2006), and neural the study.
network control (Zhang, Mei, & Chen, 2002) have already been
applied to vibration isolation systems. Two different control techni-
ques, “negative stiffness” and “displacement cancellation”, are indi- 2. Principle of vibration isolation
vidually applied to the developed experimental system to investigate
their effectiveness in active vibration isolation. Infinite stiffness and low stiffness suspension are two criteria
In the negative stiffness technique, a positive stiffness isolator required for a vibration isolation system. The series combination
and a negative stiffness isolator of equal absolute stiffness are of two isolators shown in Fig. 1 is used to achieve these two criteria
connected in series. A zero-power controlled magnetic suspension simultaneously in this study. The following paragraphs describe
system was connected with a normal spring in series to isolate how two series-connected isolators using the negative stiffness
vibration because zero-power control magnetic suspension sys- technique and displacement cancelation technique can achieve
tems behave as if they have negative stiffness (Mizuno, Takasaki, infinite stiffness against direct disturbance and low stiffness against
Kishita, & Hirakawa, 2007). Mizuno et al. acquired a vertical ground vibration.
vibration isolation system using a linear actuator connected to a
normal spring in series where the linear actuator was guided by a 2.1. Negative stiffness technique
proper negative stiffness controller (Mizuno, Toumiya, & Takasaki,
2003); the characteristics of that system were further improved In the negative stiffness technique, one of the two series-
using a positive stiffness actuator instead of a normal spring connected isolators is controlled to have a negative stiffness while
(Shahadat, Mizuno, Ishino, & Takasaki, 2012). In this study, a the other is controlled to have a positive stiffness of the same
proper stiffness controller is applied to acquire a negative stiffness amplitude. A negative stiffness system displaces in the opposite
control where the sign of controller parameters are selected so direction of a force applied on it; for example, if the suspended
that a negative displacement takes place against a static direct object shown in Fig. 2 moves upward by δl due to an increase in
disturbance. mass of Δm, then the static negative stiffness of this system can be
In the displacement cancellation technique, a displacement defined as follows:
cancellation isolator and a positive stiffness isolator are connected
in series. Mizuno et al. developed a vertical vibration isolation ðΔmgÞ
 ks ; ð1Þ
system using the displacement cancellation technique (Mizuno,  δl
Furushima, Ishino, & Takasaki, 2010) where a mechanical spring is where l is the initial steady-state distance between base and object.
used as a positive stiffness isolator. In this study, the voice coil Conversely, a system with a positive stiffness moves in the direction
motors (VCMs) guided by a proportional derivative (PD) control of the applied load; for example, a mass-spring system compresses
are used instead of normal springs to create a positive stiffness against a compressive force and expands against a tensile force. The
isolator in the developed system (Shahadat, Mizuno, Ishino, &
Takasaki, 2011).
The series combination of negative and positive stiffnesses of
equal magnitude (i.e., the negative stiffness technique) is in theory
a convenient and easy to understand method to realize a simulta-
neously infinite and low stiffness suspension. However, the main-
taining of same magnitude of stiffness of the isolators in vibration
isolation systems using the negative stiffness technique is difficult
because negative stiffness control is very sensitive and easily
deteriorates into nonlinear behavior that sometimes diverge the
system from producing infinite stiffness. In contrast, the vibration
isolation systems using the displacement cancellation technique
can avoid this difficulty because the displacement cancellation
technique is based on displacement rather than stiffness. Never-
theless, the displacement cancellation control behaves like an
integral control; as a result, the proper ground vibration attenua-
tion may not occur. A soft positive-stiffness middle mass is
connected with the isolation table in the developed system to
create an adjustable stiffness and effectively a mechanical filter
against ground vibration. Considering the points mentioned above,
the influences of the negative stiffness and displacement cancella-
tion techniques on vibration isolation are studied theoretically and Fig. 1. Springs (isolators) in series connection.
 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66 57

Fig. 2. Concept of negative stiffness.

realization of the negative stiffness technique utilizing negative- Fig. 3. Concept of displacement cancellation.
stiffness and positive-stiffness isolators is described in below.
Two series-connected isolators with stiffness coefficients k1
and k2, provide a combined stiffness “kc” (see Fig. 1) that can be Eq. (6) indicates that the static zero-compliance to a direct
expressed as follows: disturbance using the displacement cancellation technique occurs
k1 k2 when the compression in
one
isolator


is equal to the extension in
kc ¼ : ð2Þ the other isolator (i.e.,
Δy1
¼
 Δy2
), as shown in Fig. 3. For a
k1 þ k2
ground vibration, the combination of a middle mass and a soft
Eq. (2) shows that the combined stiffness kc is lower than the isolator works as a mechanical filter that attenuates the transmit-
individual stiffness of each isolator when k1 and k2 are both ting of ground vibrations to the upper table. In addition, a low-
positive (i.e., conventional springs). However, if one of the two pass filter in the feedback loop between the displacement of
isolators has


a negative


stiffness and is equal in absolute magni- the middle mass and the actuator of the upper table can further
tude (i.e.,
k1
¼
 k2
), then the combined stiffness kc becomes improve the performance of ground vibration transmissibility
infinite as follows: (Mizuno et al., 2010). In this study, an electronic low-pass filter
ð  k2 Þk2 is inserted between the two tables of the developed experimental
kc ¼ ¼ 1: ð3Þ system using the displacement cancellation technique. The
 k2 þ k2
displacement-cancellation isolator and positive-stiffness isolator
Consequently, the relative displacement of the upper table with are created using VCMs guided by I-PD and PD control, respec-
respect to the base shown in Fig. 1 becomes zero, which is tively. A simple application of PID control hardly can achieve the
expressed as follows: criteria associated to a vibration isolation system (infinite stiffness













and low stiffness). Therefore, an integral control with regard to a

k1
¼
 k2
)
Force
¼
 Force
) xu  xb ¼ 0; ð4Þ

xm  x

xu  xm
command signal and a PD control are applied in the same time to
b
attain the infinite stiffness in the experimental system; this is
where xm, xu and xb denote the displacements of the middle mass,
known as I-PD control.
the upper table and the base, respectively. Eq. (4) indicates that
In fact, Eqs. (4) and (6) represent the basics of vibration isolation,
the static zero-compliance to a direct disturbance can be achieved
maintaining a position against a direct static disturbance using the
using the negative stiffness technique, and this zero-compliance is
negative stiffness and displacement cancellation techniques, respec-
independent of the magnitude of the stiffness of each isolator.
tively. The detailed dynamic behaviors of the system, using the both
However, for a simultaneously soft suspension, series-connected
techniques individually, are presented mathematically in Section 4.
isolators with low stiffness are required. In this study, negative and
positive stiffness isolators are created using VCMs guided by a
proper negative stiffness controller and PD controller, respectively.
3. Design of controllers

2.2. Displacement cancellation technique In Section 2, it is demonstrated that the vibration isolation
system using the negative stiffness technique involves negative
In the displacement cancellation technique, one of the two stiffness control and PD control, and the vibration isolation system
series connected isolators has a soft positive stiffness (e.g., a soft using the displacement cancellation technique involves displace-
coil spring) and the other is controlled to cancel displacement ment cancellation (I-PD) and PD control. This section explains the
against a force, which is shown in Fig. 3. Because the stiffness k1 is designs of the negative stiffness, I-PD and PD controllers indivi-
positive, the displacement of the upper table “y” takes place along dually with respect to the same model (i.e., a horizontal vibration
the direction of disturbance. This displacement is canceled by the isolation system). In each case, the parameters of the controllers
upper isolator (Fig. 3), which is controlled with integral- are determined based on the pole assignment method.
proportional derivative (I-PD) control. Inherently, the upper iso- Although the developed experimental system is composed of
lator behaves as if it has a negative stiffness. The displacement y two moving masses, the isolation table and the middle mass, the
can be defined as follows: controller parameters are derived from a single-mass system because
y ¼ ðy1 þ y2 Þ ðy1  Δy1 þ y2  Δy2 Þ; ð5Þ the structure of the developed system is restricted to have a moving
middle mass (the middle table) between the isolation table and the
where Δy1 and Δy2 are the displacement of the lower and upper
base. As a consequence, there is an influence of middle mass and its
isolators, respectively. The displacement y would be zero if the
motion on the feedback parameters to control the isolation table.
following condition is satisfied:
To avoid the complexities due to this influence in determining the
0 ¼ ðy1 þ y2 Þ  ðy1  Δy1 þ y2  Δy2 Þ ) Δy1 ¼  Δy2 : ð6Þ controller parameters by the pole assignment method, two individual
58 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

single-mass systems are assumed independently where their control

stabilities are acquired individually; the systems are then connected
in series. By using the similar approach mentioned above, the two-
stage vibration isolation systems were confirmed to be stable in some
previous researches (Zhang, et al., 2002; Hoque, et al., 2006; Mizuno,
et al., 2007). Additionally, because a system of positive stiffness is
needed in both techniques, PD control is applied to the middle mass;
hence, by selecting the same PD parameters in both techniques, the
same control condition (i.e., the same closed-loop poles) can be used
without repeating the complex calculation; this is another reason why
the controller parameters are derived from a single-mass system. Fig. 5. Block diagram of negative stiffness controller.

3.1. Basic equation Substituting the Laplace transform of Eq. (11) into the Laplace
transform of Eq. (10) leads to
A basic model of a horizontal vibration isolation system with a !
ðnÞ pðnÞ s þ P ðnÞ
v s
2
ðnÞ
single degree-of-freedom is shown in Fig. 4. Here, it is assumed I ðsÞ ¼  d
þ P p XðsÞ; ð12Þ
that the table (mass: m) is driven by a VCM and can move along s  P ðnÞ
i
the x-axis (i.e., horizontal translation motion) without any internal
where P(n) (n)
d , Pv and Pi
(n)
are the proportional, derivative and integral
interference from the other axes. The x-axis motion equation of
parameters of the negative stiffness controller, respectively. The
the table is written as follows:
transfer function representation of the displacement from a direct
mx€ ¼ F a þ F d ; ð7Þ disturbance in the negative-stiffness-controlled system is obtained
by substituting Eq. (12) into Eq. (9) and is given as follows:
where x is the relative displacement of the table to base, Fa is the
actuator's thrust force, and Fd is the direct disturbance acting on ðnÞ
XðsÞ ðs  P i Þ
the table. ¼ ðnÞ
; ð13Þ
F d ðsÞ mt c ðsÞ
The thrust force generated by a VCM is proportional to its coil
current i. Thus, the thrust force Fa can be defined as follows: where t(n)c (s) defines the characteristic equation of the controlled

F a ¼ ki i; ð8Þ system, which is expressed as follows:

 
where ki is actuator's thrust force coefficient. The transfer function ki ðnÞ k ki ðnÞ ðnÞ
t ðnÞ 3
c ðsÞ ¼ s þ P v  P ðnÞ
i s2 þ i ðP ðnÞ þ P ðnÞ
p Þs  P p P i : ð14Þ
representation of the actuator's dynamics is obtained by substitut- m m d m
ing Eq. (8) into Eq. (7) and is written as a Laplace transform as
Eq. (14) indicates that the system with negative stiffness control is a
follows:
3rd-order system. To find the steady-state displacement xð1Þ, the
 
1 k 1 disturbance Fd is assumed to be stepwise and is modeled as follows:
XðsÞ ¼ 2 i IðsÞ þ F d ðsÞ ; ð9Þ
s m m
F0
F d ðsÞ ¼ ðF 0 : step size ðconstantÞÞ: ð15Þ
where each Laplace transform variable is denoted by its capital. s
Therefore, the steady-state displacement xð1Þ against a unit step-
3.2. Design of negative stiffness controller wise disturbance can be specified as follows:

A typical negative stiffness controller is presented in the block xð1Þ ðs  P ðnÞ Þ  P iðnÞ 1
¼ lim ðnÞi ¼ ðnÞ ðnÞ
¼ ðnÞ
: ð16Þ
diagram shown in Fig. 5, which contains a normal PD control F0 s-0 mt ðsÞ
c  k P
i p P k i P
i p
followed by a local integral feedback control where a reference
input to the controller is assumed to be zero by referring to a Because the steady-state displacement of a system against a unit
vibration isolation system (Mizuno et al., 2007). Hence, the control load (i.e., a disturbance) is the inverse of the static stiffness of that
current i(n) for negative stiffness control can be expressed as system, Eq. (16) can be rewritten as follows:
follows:
xð1Þ 1 1
¼ ¼ ; ð17Þ
iðnÞ ¼  P pðnÞ x þi; ð10Þ F0 ki P pðnÞ ks

where P(n) p is stiffness adjustment parameter and i is defined as where ks is the magnitude of the negative stiffness.
follows: The characteristic equation of a 3rd-order ideal system is
Z
represented as follows:
i ¼  P ðnÞ
d
x  P vðnÞ x_ þ P ðnÞ
i idt: ð11Þ
t d ðsÞ ¼ ðs2 þ 2ζ r ω1 s þ ω1 2 Þðs þ ω2 Þ ¼ s3 þ α2 s2 þα1 s þ α0 : ð18Þ

where α2 ¼ 2ζω1 þ ω2 ; α1 ¼ 2ζω1 ω2 þ ω1 2 ; and α0 ¼ ω1 2 ω2 . To cal-

culate the controller parameters, it is assumed that ω1 ¼ ω2 ¼ ωr :
The symbols ωr (2πfr) and ζr denote the angular frequency and
damping ratio, respectively. Usually, these parameters are used
to describe the dynamics of a 2nd-order system. In this paper,
however, they are used to specify the closed-loop poles of a
controlled system because they are more comprehensible than
the poles themselves. According to the pole assignment method,
the controller parameters are determined uniquely by comparing
the characteristic Eqs. (14) and (18) while considering Eq. (17).
Fig. 4. Basic model of horizontal vibration isolation system with VCM. Finally, the parameters of the negative stiffness controller are
 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66 59

obtained as follows: into Eq. (23) leads to the closed-looped transfer function represen-
ks tation of the I-PD controlled system, shown as follows:
P ðnÞ
p ¼  ; ð19Þ
ki XðsÞ ki P ðdÞ
¼ i
; ð26Þ
mα0 RðsÞ mt ðdÞ
c ðsÞ
P ðnÞ
i ¼  ðnÞ
; ð20Þ
ki P p where t(d) c (s) represents the characteristics equation of the con-
trolled system and is expressed as follows:
mα1    
P ðnÞ
d
¼  P pðnÞ ; ð21Þ ki ðdÞ 2 ki ðdÞ k
ki t ðdÞ ðsÞ ¼ s 3
þ P s þ P s þ i P ðdÞ : ð27Þ
c
m v m d m i
mðα2 þ P ðnÞ
i Þ
P ðnÞ
v ¼ : ð22Þ Eq. (27) indicates that the system with I-PD control is a 3rd-order
ki system. According to the pole assignment method, the parameters
of the I-PD controller are determined uniquely by comparing the
3.3. Design of displacement cancellation controller characteristic Eq. (27) with the characteristic equation of an ideal
3rd-order system, shown in Eq. (18). The parameters of the I-PD
The basic concept of displacement cancellation technique is to controller are obtained as follows:
cancel the disturbances on the system by inserting a command mα0
into the controller. However, a system with the displacement P ðdÞ
i ¼ ; ð28Þ
ki
cancellation control in principle is unable to measure the distur-
bances on it. Hence, to design the controller parameters of the mα1
P ðdÞ
d
¼ ; ð29Þ
displacement cancellation control (i.e., I-PD controller), the dis- ki
turbance shown in Eq. (9) is assumed to be compensated by an
mα2
integral compensator (i.e., a command signal); a reference input to P ðdÞ
v ¼ : ð30Þ
ki
the controller is assumed as the command signal (r) in this study.
Moreover, if a disturbance cancellation by a command signal (i.e., Because PD control is applied to achieve a soft positive-stiffness
Fd ¼0, r a0) or no disturbance cancellation using no command system in both techniques, the derivation of the parameters of a
signal (i.e., Fd a0, r ¼0) occurs, the characteristic equation of the PD controller is illustrated in this section as well. The parameters
system would remain the same. Because the design of a controller of the PD controller are determined by comparing the character-
depends on the characteristic equation governing the system, the istic equations of an ideal 2nd-order system with the characteristic
magnitudes of controller parameters will be the same values for equation of a PD controlled system. In general, the characteristic
the both cases. Therefore, the actuator dynamic described by Eq. equation of an ideal 2nd-order system is represented as follows:
(9) are revised with respect to I-PD control and is given as follows:
  t^d ðsÞ ¼ ðs2 þ 2ζ^ r ω ^ r 2 Þ ¼ s2 þ α^ 1 s þ α^ 0 ;
^ rsþω ð31Þ
1 k
XðsÞ ¼ 2 i I ðdÞ ðsÞ : ð23Þ where ω^ r and ζ^ r denote the angular frequency and damping ratio,
s m
respectively, and are used to specify the closed-loop poles of the
where I(d)(s) denotes the Laplace transform of the I-PD control PD controlled system. Thus the parameters of the PD controller are
current. The block diagram of an I-PD controller is shown in Fig. 6 obtained as follows:
where a reference input to the controller is assumed to be zero by
mα^ 0
referring to a vibration isolation system (Mizuno et al., 2010). The P ðpÞ
d
¼ ; ð32Þ
ki
control current i(d) can be defined as follows:
Z
mα^ 1
iðdÞ ¼  P ðdÞ ðx  rÞdt  P ðdÞ x  P ðdÞ
v x:
_ ð24Þ P ðpÞ
v ¼ ; ð33Þ
i d ki
The Laplace transform of Eq. (24) is written as follows: where P(p) (p)
d and Pv are the proportional and derivative parameters
! of the PD controller, respectively. Because the proportional para-
P ðdÞ P ðdÞ
I ðdÞ ðsÞ ¼ i rðsÞ  i
þ P ðdÞ
d
þ P ðdÞ
v s XðsÞ; ð25Þ meter of a PD controller defines the static positive stiffness of that
s s PD controlled system, the parameter P(p) is adjusted so that the
d

where P(d) (d) (d) acquired positive stiffness kp is equal to the absolute magnitude of
d , Pv , and Pi are the proportional, derivative, and integral
parameters of the I-PD controller, respectively. Substituting Eq. (25) the negative stiffness ks in the negative stiffness technique. This
value is given as follows:
kp
P ðpÞ
d
¼ : ð34Þ
ki

4. Characteristic analysis of the system

The developed system is a double-stage vibration isolation

system with the stages connected in series as shown in Fig. 1.
The controllers for these series connected tables (i.e., stages) are
designed individually, considering each table as a single-degree-
of-freedom system, as shown in the basic model in Fig. 4. The basic
motion equations of the middle mass and the isolation table
associated with the entire system (i.e., two individual systems in
series) are written, respectively, as follows:

Fig. 6. Block diagram of I-PD controller. m1 x€ 1 ¼ ki1 i1  ki2 i2 ; ð35Þ

60 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

m2 x€ 2 ¼ ki2 i2 þ F d ; ð36Þ r 2d ðsÞ ¼ m1 s2 þ t 1 ðsÞ þt 2 ðsÞ;

where x1, x2 are the displacements of the middle mass and the
isolation table, respectively. In these basic motion equations, the t 1 ðsÞ ¼ ki1 ðP ðpÞ
d
þP ðpÞ
v sÞ;
subscripts 1 and 2 with the corresponding symbols represent
the middle mass and the isolation table, respectively. The Laplace !
P ðnÞ
d
s þP ðnÞ 2 ðnÞ ðnÞ ðnÞ
v s þ Pp s  Pp Pi
transforms of Eqs. (35) and (36) are written, respectively, as follows: t 2 ðsÞ ¼ ki2 ;
s  P ðnÞ
i
2
m1 s X 1 ðsÞ ¼ ki1 I 1 ðsÞ  ki2 I 2 ðsÞ; ð37Þ

m2 s2 X 2 ðsÞ ¼ ki2 I 2 ðsÞ þ F d ðsÞ: ð38Þ t c ðsÞ ¼ ðm1 s2 þ t 1 ðsÞ þ t 2 ðsÞÞðm2 s2 þ t 2 ðsÞÞ  ðt 2 ðsÞÞ2 :

The dynamics of the middle mass and the isolation table shown in
Eqs. (41) and (42) involve the responses of the tables to both direct
4.1. Negative stiffness control system disturbance and ground vibration where X0 and Fd denote the base
displacement (i.e., floor vibration) and the direct disturbance,
A schematic diagram of the developed vibration isolation respectively. In this paper, the theoretical responses to the direct
system using the negative stiffness technique is shown in Fig. 7, disturbance and ground vibration are obtained while X0 and Fd are
where the middle mass is controlled with PD control, and the respectively assumed to be zero in these equations. Moreover, the
isolation table mounted on the middle mass is controlled with static responses of the tables against the direct disturbance and
proper negative stiffness (ks) control. The control currents for the ground vibration are obtained by substituting the condition “s ¼0”
PD (i.e., positive stiffness kp) control i(p) and the negative stiffness into Eqs. (41) and (42).
control i(n) are written, as the Laplace transforms respectively, as To estimate the steady-state displacements of the tables to a
follows: direct disturbance, it is assumed that the disturbance Fd is stepwise,
and there is no ground vibration (X0 ¼ 0). Here, the steady-state
I ðpÞ ðsÞ ¼  ðP ðpÞ
d
þ P ðpÞ
v sÞðX 1 ðsÞ X 0 ðsÞÞ; ð39Þ
displacement of the isolation table is obtained by Eq. (42) and is
! given as follows:
pðnÞ s þ P ðnÞ
v s
2
I ðnÞ ðsÞ ¼  d
þ P ðnÞ ðX 2 ðsÞ X 1 ðsÞÞ: ð40Þ
s  P ðnÞ
p
X 2 ð1Þ r ðsÞ m1 s2 þ t 1 ðsÞ þ t 2 ðsÞ
i ¼ lim 2d ¼ lim
F0 s-0 t c ðsÞ s-0 t c ðsÞ
where x0(s) denotes the base displacement of the developed
system. Substituting these control currents (I1(s)¼I(p)(s), I2(s)¼ 1 1 1 1
¼ þ ¼ þ ðby Eqs: ð17Þ and ð34ÞÞ: ð43Þ
I(n)(s)) into the motion Eqs. (37) and (38) yields the dynamics of ki2 P pðnÞ ki1 P ðpÞ ks kp
d
the moving tables of the system using the negative stiffness
technique, which are given as follows: Eq. (43) indicates that there is no steady-state displacement of the
r 1g ðsÞ r ðsÞ isolation table when the positive


and

negative stiffnesses are equal
X 1 ðsÞ ¼ X 0 ðsÞ þ 1d F d ðsÞ; ð41Þ in absolute magnitude (i.e.,
kn
¼
kp
). Similarly, the steady-state
t c ðsÞ t c ðsÞ
displacement of the middle mass is obtained as follows:
r 2g ðsÞ r ðsÞ
X 2 ðsÞ ¼ X 0 ðsÞ þ 2d F d ðsÞ; ð42Þ X 1 ð1Þ r ðsÞ k2 P pðnÞ 1
t c ðsÞ t c ðsÞ ¼ lim 1d ¼ ¼ ðby Eq: ð34ÞÞ: ð44Þ
F0 s-0 t c ðsÞ k1 k2 P ðpÞ P ðnÞ
p
kp
where d

r 1g ðsÞ ¼ ðm2 s2 þ t 2 ðsÞÞt 1 ðsÞ; The gap between the isolation table and the middle mass is
determined as follows:
r 1d ðsÞ ¼ t 2 ðsÞ;
X 2 ð1Þ  X 1 ð1Þ 1
¼ : ð45Þ
r 2g ðsÞ ¼ t 2 ðsÞt 1 ðsÞ; F0 ks

Eq. (44) demonstrates that the middle mass moves in the direction
of the disturbance applied to the isolation table, while the gap
between the two tables increases for F0 40. It is thus recognized
theoretically that the developed system using the negative stiffness
technique can realize the characteristics of vibration isolation
associated with direct disturbance.
To estimate the displacements of the middle mass and the
isolation table utilizing the negative stiffness technique against
ground vibrations of low frequency, it is assumed that the floor
moves stepwise ðs-0Þ, and there is no direct disturbance on the
isolation table; this allows the calculation of the displacements as
follows:

X 1 ðsÞ r 1g ðsÞ ðm2 s2 þ t 2 ðsÞÞt 1 ðsÞ

¼ lim ¼ lim ¼ 1; ð46Þ
X 0 ðsÞ s-0 t c ðsÞ s-0 t c ðsÞ

x2 ðsÞ r 2g ðsÞ t 2 ðsÞt 1 ðsÞ

¼ lim ¼ lim ¼ 1: ð47Þ
X 0 ðsÞ s-0 t c ðsÞ s-0 t c ðsÞ

Eqs. (46) and (47) indicate that, at very slow floor movements,
Fig. 7. Schematic diagram of the developed system using the negative stiffness the displacements of the tables are similar to the displacement of
technique. the base.
 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66 61

4.2. Displacement cancellation control system decreases of the input magnitude at frequencies above the cutoff
frequency. The magnitudes of the input signals are reduced by
A schematic diagram of the developed vibration isolation system 20 dB and 40 dB per decade above the cutoff frequency when first
using the displacement cancellation technique is shown in Fig. 8. and second order filters are used, respectively. First and second
In the displacement cancellation technique, the middle mass is order electronic low-pass filters are used in this study and they are
controlled with PD (i.e., positive stiffness) control while the isola- expressed, respectively, as follows:
ω ω 2
tion table is controlled with I-PD control. Because the middle mass GF ðsÞ ¼ s þ fωf and GF ðsÞ ¼ s2 þ 2ωf s þ ω 2 , where ωf is the cut-off
f f
is guided by a VCM with PD control in both techniques, the control frequency of the filters. Under the condition of (51), the transfer
current needed for the middle mass in the displacement cancella- function (1  GF ðsÞ) can be represented as follows:
tion technique is the same as that in the negative stiffness tech- 1  GF ðsÞ ¼ cF ðsÞs; ð52Þ
nique shown in Eq. (39). Since, the displacement cancellation


technique focuses on the cancellation of displacement with respect where lim
cF ðsÞ
o 1.
s-0
to the base, the command signal r for the developed system (i.e., Substituting of control currents shown in Eq. (39) and (50)
two-mass system) is selected carefully so that the relative displace- [I1(s)¼I(p)(s), I2(s) ¼I(d)(s)] into the motion Eqs. (37) and (38) yields
ment of the isolation table with respect to the base (x2  x0) that the dynamics of the moving tables of the system using the
needs to be canceled becomes feedback parameter of the integral displacement cancellation technique, which are given as follows:
control. The eddy current gap sensors used in the developed system
can detect the relative displacements of an object with respect to a r^ 1g ðsÞ r^ ðsÞ
X 1 ðsÞ ¼ X ðsÞ þ 1d F d ðsÞ; ð53Þ
reference point (i.e., the base); i.e., the sensor outputs yield the ^t c ðsÞ 0 t^c ðsÞ
calculation of a displacement x as “x x0”. Inherently, for the
developed system, the command signal r and the variable x shown r^ 2g ðsÞ r^ ðsÞ
X 2 ðsÞ ¼ X 0 ðsÞ þ 2d F d ðsÞ; ð54Þ
in Eq. (24) can be defined as follows: t^ c ðsÞ t^c ðsÞ
   
rðtÞ ¼ x2 ðtÞ  x0 ðtÞ  x2 ðtÞ  x1 ðtÞ ¼ x1 ðtÞ  x0 ðtÞ; ð48Þ where
r^ 1g ðsÞ ¼ ðt^ 1 ðsÞ  t^ 3 ðsÞð1  GF ðsÞÞÞðm2 s2 þ t^ 2 ðsÞ þ t^ 3 ðsÞÞ
xðtÞ ¼ x2 ðtÞ  x1 ðtÞ: ð49Þ
þ t^ 3 ðsÞð1  GF ðsÞÞðt^ 2 ðsÞ þ t^3 ðsÞÞ;
An electronic low-pass filter is inserted into the feedback loop
between the displacement of the middle mass and the actuator
r^ 1d ðsÞ ¼ ðt^ 2 ðsÞ þ t^ 3 ðsÞÞ;
of the isolation table to attenuate the magnitude of the ground
vibration at frequencies above the cutoff frequency of the filter.
r^ 2g ðsÞ ¼ t^3 ðsÞGF ðsÞðm1 s2 þ t^1 ðsÞ þ t^ 2 ðsÞ þ t^ 3 ðsÞð1  GF ðsÞÞÞ
Therefore, the Laplace transform of the control current of the I-PD
controller can be expressed as follows: þ ðt^1 ðsÞ  t^ 3 ðsÞGF ðsÞÞðt^ 2 ðsÞ þ t^3 ðsÞð1 GF ðsÞÞÞ;

P ðdÞ   r^ 2d ðsÞ ¼ ðm1 s2 þ t^1 ðsÞ þ t^ 2 ðsÞ þ t^ 3 ðsÞð1  GF ðsÞÞÞ;

I ðdÞ ðsÞ ¼  i
ðX 2 ðsÞ  X 1 ðsÞÞ þ GF ðsÞðX 1 ðsÞ  X 0 ðsÞÞ
s
 ðP ðdÞ
d
þ P ðdÞ
v sÞðX 2 ðsÞ  X 1 ðsÞÞ; ð50Þ t^1 ðsÞ ¼ ki1 ðP ðpÞ
d
þ P ðpÞ
v sÞ;

where GF(s) is the transfer function representation of the inserted

t^2 ðsÞ ¼ ki2 ðP ðdÞ þ P ðdÞ
v sÞ;
electronic low-pass filter that satisfies the steady-state criteria d

given as follows: !
P ðdÞ
GF ð0Þ ¼ 1: ð51Þ t^3 ðsÞ ¼ ki2 i
;
s
In this paper, the effect of an electric low-pass filter of different
orders on ground vibration transmissibility is investigated as well. t^c ðsÞ ¼ ðm1 s2 þ t^1 ðsÞ þ t^2 ðsÞ þ t^ 3 ðsÞð1  GF ðsÞÞÞðm2 s2 þ t^2 ðsÞ þ t^3 ðsÞÞ
The order of a low-pass filter determines the gradient descent of
 ðt^2 ðsÞ þ t^3 ðsÞÞðt^ 2 ðsÞ þ t^ 3 ðsÞð1  GF ðsÞÞÞ:

The dynamics of the middle mass and the isolation table shown in
Eqs. (53) and (54) involve the responses of the tables to both direct
disturbance and ground vibration where X0 and Fd denote the base
displacement (i.e., floor vibration) and the direct disturbance,
respectively. In this paper, the theoretical responses to the direct
disturbance and ground vibration are obtained while X0 and Fd are
respectively assumed to be zero in these equations. Moreover, the
static responses of the tables against the direct disturbance and
ground vibration are obtained by substituting the condition “s ¼0”
into Eqs. (53) and (54).
To estimate the steady-state displacements of the tables to a
direct disturbance, it is assumed that the disturbance Fd is
stepwise, and there is no ground vibration (X0 ¼ 0). Here, the
steady-state displacement of the isolation table is obtained by
Eq. (54) and is given as follows:
x2 ð1Þ r^ ðsÞ
¼ lim 2d ¼ 0: ð55Þ
F0 s-0 t^ c ðsÞ

Eq. (55) indicates that there is an infinite stiffness between the

Fig. 8. Schematic diagram of the developed system using the displacement isolation table and the base as well as no steady-state displacement
cancellation technique. of the isolation table. Similarly, the steady-state displacement of the
62 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

middle mass is determined as follows:

x1 ð1Þ r^ ðsÞ 1 1
¼ lim 1d ¼ ¼ ðby Eq: ð34ÞÞ: ð56Þ
F0 s-0 t^ c ðsÞ ki1 P ðpÞ kp
d

The gap between the two tables is given by the following equation:
x2 ð1Þ  x1 ð1Þ 1
¼ : ð57Þ
F0 kp
Eqs. (56) and (57) indicate that the middle mass moves in the same
direction as the disturbance applied on the isolation table and the
gap between the two tables increases for F0 40. It is thus confirmed
theoretically that the developed system using the displacement
cancellation technique can effectively isolate the vibrations asso-
ciated with direct disturbance.
To estimate the steady-state displacements of the middle mass
and the isolation table utilizing the displacement cancellation Fig. 9. Photograph of the experimental system.
technique against ground vibrations of low frequency, it is
assumed that the floor moves stepwise ðs-0Þ, and there is no
direct disturbance on the isolation table; the displacements are
thus obtained as follows:
X 1 ðsÞ r^ 1g ðsÞ t^ 1 ðsÞðm2 s2 þ t^ 2 ðsÞ þ t^3 ðsÞ
¼ lim ¼ lim ¼ 1; ð1 GF ð0Þ ¼ 0Þ
X 0 ðsÞ s-0 t^c ðsÞ s-0 t^ c ðsÞ
ð58Þ

x2 ðsÞ r^ 2g ðsÞ
¼ lim
X 0 ðsÞ s-0 t^c ðsÞ
t^ 3 ðsÞGF ðsÞðm1 s2 þ t^1 ðsÞ þ t^2 ðsÞ þ ðt^1 ðsÞ  t^ 3 ðsÞGF ðsÞÞt^2 ðsÞ
¼ lim ¼ 1:
s-0 t^c ðsÞ
ð59Þ
Eqs. (58) and (59) indicate that, at very slow floor movements,
the displacements of the tables are similar to the displacement of
the base. Fig. 10. Arrangements of actuators and sensors for each table.

5. Experiment dimensions including permanent magnets. The average thrust

coefficient of the VCMs was 22 N/A. The designed control algo-
5.1. Experimental setup rithms were implemented with a digital signal processor from
dSPACE™ at a 1 kHz sampling frequency.
The developed vibration isolation system shown in Fig. 9 The isolation table had dimensions of 620  620  10 mm3
consisted of two moving tables: the isolation table and the middle and a mass of 20.5 kg. The dimensions of the isolation table were
mass (table). Both tables were supported vertically by free bear- selected based on a standard driver seat of a heavy vehicle. In
ings at the four corner points of each table, making them free to addition, the dimensions of the middle table were selected to be
move in two dimensions horizontally and rotate about the vertical 530  530  10 mm3 with a mass of 15.5 kg.
axis. Although the three-degree-of freedoms (3-DOFs) of motion of
each table are controlled, only the responses along a single axis 5.2. Experimental results
(i.e., translation along the x-axis) were analyzed in this paper.
To control these three axes of motion of each table, four VCMs In the experimental analysis, the static and dynamic responses
were used; they were attached with the respective table like that of the developed vibration isolation system using the negative
as shown in Fig. 10. The motion of each table was detected by six stiffness and displacement cancellation techniques individually
eddy current gap sensors positioned around the table as shown in were measured against both direct disturbance and ground vibra-
Fig. 10. With the frequency range of the signals to be measured set tion. The direct disturbance applied on the isolation table was
to 0–100 Hz, the maximum length to be measured was 75 mm. generated by the two additional VCMs that were mounted on the
Additionally, to achieve the associated precision accuracy for an isolation table and were fixed respect to the base. The same
effective vibration isolation system, eddy current gap sensors were disturbance signal was fed to these two VCMs to produce a pure
selected; these Baumer Electric sensors had a sensing coefficient translational disturbance (i.e., no rotational disturbance) on the
and resolution of 1550 V/m and 0.005 mm, respectively. The tables. Similarly, the ground vibrational effect on the developed
contact areas of each table with vertical supports were lubricated system was generated by shaking the system with respect to the
by grease to make frictional effects negligible. Every individual- floor; this shaking was created by four VCMs fixed between the
axis motion control of the tables was based on feedback para- base of the system and the floor. The VCMs used to generate
meters; consequently, the VCMs used in the developed system disturbances (i.e., direct disturbance and ground vibration) were
generated resisting torques and thrust forces at the same time operated independently from those used to control the motion of
without interfering with each other. The motion of the middle the tables.
table is guided by the VCMs placed between the middle table and To compare the negative stiffness and displacement cancella-
the base whereas the VCMs used to guide the isolation table are tion techniques, experiments were conducted using unchanged
placed between the two tables. The VCMs had nearly the same closed-loop poles in both cases. In this paper, the closed-loop
 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66 63

poles are defined as follows:

s1 ¼  ωr ;
 qffiffiffiffiffiffiffiffiffiffiffiffi
s2 ; s3 ¼  ζ r 7j 1  ζ 2r ωr ;

 qffiffiffiffiffiffiffiffiffiffiffiffi
2
s4 ; s5 ¼  ζ^ r 7j 1  ζ^ ω r ^ r:

The closed-loop poles of a stable control system take place at

the left side of a complex plane (i.e., imaginary-real), where the
real parts of these poles are always negative. Indeed, the pole
assignment method determines the controller parameters by
making a comparison between the characteristic equations of
the system to be controlled and an ideal system of the same order.
It shows that the controller parameters are mainly functions of
the angular frequency and damping ratio (see Eqs. (18) and (31)).
In this study, the magnitudes of the angular frequency and the
damping ratio are selected (ωr ¼ ω ^ r ¼ 2π  4 rad=s; ζ r ¼ ζ^ r ¼ 0:5)
based on a practically feasible stable condition so that all the
closed-loop poles remain in the left side of the complex plane. To
maintain the same closed-loop poles throughout the experiments,
the magnitudes of the parameters ωr ; ω ^ r ; ζ r and ζ^ r are kept
unchanged in both techniques.
The static characteristics of the moving tables against a direct
disturbance applied on the isolation table are shown in Fig. 11. In
these experiments, it is observed that the isolation table behaves
with an infinite stiffness against a static load (i.e., 0 Hz) for the both
techniques; this result has also been confirmed theoretically in
Section 4. It also appears that the displacement cancellation techni-
que can maintain this zero-compliance characteristic for a wider
range of static loads compared to the negative stiffness technique.
The reasons behind this circumstance are stated in the interpreta-
tion of Fig. 12. The displacement of the middle table and the relative
displacement between the two tables are measured and shown
together in Fig. 11. The displacement of the middle table with
respect to the base is positive whereas the relative displacement
between the two tables is negative; the absolute value of these
displacements are shown to be equal in magnitude for a specific
load, which is also confirmed theoretically in Section 4. Fig. 12. Step response to stepwise direct disturbance: (a) step disturbance,
Fig. 12 illustrates the step responses of the tables to a stepwise (b) negative stiffness technique, and (c) displacement cancellation technique.
direct disturbance on the isolation table. It is found that the isolation
table returns to its original position exactly and can keep this direct disturbance when the absolute magnitudes of both stiffnesses
position in steady states when using the displacement cancellation of the series-connected isolators are equal (described in Section 2);
technique; however, a small static error occurs under the same however, in actual practice, it is very difficult to maintain the
conditions when the negative stiffness technique is used. The equality of the positive and negative stiffnesses in the system. As a
unequal stiffness in the negative stiffness technique is the primary result, the system using the negative stiffness technique is very
reason for this static error shown in Figs. 11 and 12(b). The negative sensitive and thus easily deteriorates to nonlinearity (Shahadat,
stiffness technique can solely possess an infinite stiffness against a Mizuno, Ishino, & Takasaki, 2014). The displacement cancellation
technique does not suffer from such static error because it includes
integral control with regard to the displacement feedback. However,
in transient states, the displacement peak of the isolation table with
the displacement cancellation technique is longer than that of the
negative stiffness technique. The control currents against a step
disturbance are measured using the both techniques individually
and are shown in Fig. 13. Considering the peak and fluctuation of the
control current in transient states, it can be stated that the vibration
isolation with displacement cancellation technique is slightly worse
compared to that with the negative stiffness technique.
In this paper, the dynamic responses of the developed experimen-
tal system against both the direct disturbance and ground vibration are
presented by bode plot which is also known as a graph of transfer
function T(s) in the Laplace domain versus frequency. The frequency
responses of tables of the experimental system, using both techniques
under the same closed-loop poles (i.e., selected for “ωr ¼ ω ^r ¼
Fig. 11. Static response to direct disturbance. 2π  4 rad=s; ζ r ¼ ζ^ r ¼ 0:5”) over the experiments, are measured
64 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

individually and shown by the bode plots given in Figs. 14–18. The
frequency responses of the isolation table to direct disturbance are
shown in Fig. 14 that includes the magnitude (in decibels) and phase
angle (degree) of transfer function of the isolation table. Here, the
transfer function denotes the ratio of displacement of the isolation
table (m) to the direct disturbance (N) applied on it, which is defined
as FXðsÞ
d ðsÞ
in the text. The Laplace variable s means jω. In this figure, the
magnitudes (dB)
and phase

angles

 (degree) of the transfer function are
measured as 20log 10
TðjωÞ
and ∠TðjωÞ, respectively, where TðjωÞ
denotes FXðjωÞ d ðjωÞ
. It is recognized that the isolation table can have a high
stiffness at low frequency and even infinite stiffness at zero frequency
when using either the negative stiffness or displacement cancellation
techniques (i.e., the dB magnitude is very small and the system thus
has a large stiffness) at 0 Hz. In these experiments, a frequency range
Fig. 15. Frequency response to direct disturbance of the middle table.
from 1 to 30 Hz is considered because the developed system is heavy
(i.e., the isolation table has a mass of 20.5 kg), and the zero-compliance
controls are applied. Nevertheless, the nonlinearity due to the induced
friction in the system behaves differently for different controllers
(Shahadat, et al., 2014); this may be one of the reasons for the varying
dynamic responses shown in Fig. 14, specifically at low frequencies (i.e.,
significant friction is experienced under slow movement). Although

Fig. 16. Frequency response to direct disturbance of the control current.

Fig. 13. Control currents of actuators against a stepwise direct disturbance.

Fig. 17. Frequency response to ground vibration of the isolation table: (a) negative
stiffness technique and (b) displacement cancellation technique.

the closed-loop poles for both techniques are the same, they may still
have different zeros; and besides the poles also the zeros heavily
determine the frequency response of the system, which may be
Fig. 14. Frequency response to direct disturbance of the isolation table. another reason for the variations shown in Fig. 14.
 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66 65

technique. The experimental behaviors of the developed vibration

isolation system utilizing both techniques are measured, and a
comparison between them is carried out using the same closed-
loop poles. With regard to static responses, the isolation table
using the displacement cancellation technique can maintain zero-
compliance for a wider range of static loads applied on the isolation
table than that using the negative stiffness technique. In the case of
a step response, a shorter transient displacement and a longer
settling time are observed with the negative stiffness technique.
The frequency responses of the isolation table with both techniques
are almost identical at higher frequencies but vary significantly at
lower frequencies. The resonance peak of ground vibration trans-
missibility is shown to be shorter with the negative stiffness
technique. The ground vibration transmissibility with the displace-
ment cancellation technique could be improved by inserting an
electric low-pass filter into the feedback-loop between the two
Fig. 18. Effect of low-pass filter to suppress ground vibration with the displacement tables of the developed system.
cancellation technique.

Similarly, the frequency responses to direct disturbance of the

Acknowledgments
middle table are measured and shown in Fig. 15. It is shown that
the middle table has a lower stiffness compared to the isolation
table when the disturbance is below the resonance frequency. The authors gratefully acknowledge the financial support
Inherently, the developed system can be used to suppress ground received from the Ministry of Education, Culture, Sports, Science
vibration as well. and Technology of Japan, and a Grant-in-Aid for Scientific Research (B).
Furthermore, the characteristics of the current required to The authors thank Mr. Takehiko Furushima for his contribution with
control the isolation table with respect to the frequency of the this study.
applied direct disturbance are measured and shown in Fig. 16. It is
observed that the system using the displacement cancellation
technique produces higher current consumption and abrupt fluc- References
tuations near the resonance frequency compared to the system
using the negative stiffness technique. Daley, S., Hatonen, J., & Owens, D. H. (2006). Active vibration isolation in a smart
spring mount using a repetitive control approach. Control Engineering Practice,
The frequency responses of the isolation table to ground vibra- 14, 991–997.
tion are shown in Fig. 17 that includes the magnitude (dB) and phase Ho, S. T., Matsuhisa, H., & Honda, Y. (2000). Passive vibration suppression of beam
angle (degree) of transfer function of the isolation table. Here, the with piezoelectric elements. JSME International Journal, Series C, 43(3), 740–747.
Hoque, M. E., Mizuno, T., Takasaki, M., & Ishino, Y. (2006). Development of a three-
transfer function denotes the ratio of displacement of the isolation axis active vibration isolation isolator using zero-power control. IEEE/ASME
table (m) to the ground vibration (m) applied to the base, which is Transactions on Mechatronics, 11(4), 462–470.
defined as XXðsÞ0 ðsÞ
in the text. In this figure, the magnitudes (dB) and Hoque, M. E., Mizuno, T., Takasaki, M., & Ishino, Y. (2010). Development of an active
vibration isolation system using linearized zero-power control with weight
phase

angles


(degree)
 of the transfer function are measured as support springs. Journal of Vibration and Acoustics, 132, 041006/1–041006/9.
20log 10
TðjωÞ
and ∠TðjωÞ, respectively, where TðjωÞ denotes XXðjωÞ
0 ðjωÞ
. Jaensch, M., & Lamperth, M. U. (2007). Development of a multi-degree-of-freedom
It is observed that the dB magnitude decreases beyond the reso- micro positioning, vibration isolation and vibration suppression system. Smart
Materials and Structures, 16(2), 409–417.
nance frequency in both techniques; inherently, the system is Long, X., Jiang, H., & Meng, G. (2013). Active vibration control for peripheral milling
effective at attenuating ground vibrations of high frequency. In these processes. Journal of Materials Processing Technology, 213, 660–670.
experiments, it is observed that the resonance peak is longer but Miyazaki, T., Mizuno, K., Kawatani, R., & Hamada, H. (1994). Consideration about
feedback feedforward hybrid control for active control of micro-vibration
quicker to decline when using the displacement cancellation tech-
control. In Proceedings of the Second International Conference on Motion and
nique. The effect of actively damping ground vibration transmissi- Vibration Control (pp. 29–34).
bility is illustrated in Fig. 17 as well. Solely for recognizing the Mizuno, T. (2001). Proposal of a vibration isolation system using zero-power
differences between the two techniques with respect to ground magnetic suspension. In Proceedings of Asia Pacific Vibration Conference: Vol. 2
(pp. 423–427).
vibration transmissibility, the experimental results shown in Fig. 17 Mizuno, T., Toumiya, T., & Takasaki, M. (2003). Vibration isolation using negative
(b) were measured under the condition of “GF(s)¼1” GF(s) is transfer stiffness. JSME International Journal, Series C, 46(3), 807–812.
function of low-pass filter]. The ground vibration transmissibility is Mizuno, T., Takasaki, M., & Murashita, M. (2005). Pneummatic zero-compliance
mechanism using negative stiffness. In IUTAM Symposium. Munich.
investigated with regard to the order of an electronic low-pass filter Mizuno, T., Takasaki, M., Kishita, D., & Hirakawa, K. (2007). Vibration isolation
attached between the two tables of the developed system utilizing system combining zero-power magnetic suspension with springs. Control
the displacement cancellation technique these results are shown in Engineering Practice, 15, 187–196.
Mizuno, T., Furushima, T., Ishino, Y., & Takasaki, M. (2010). Realization of zero-
Fig. 18. It is shown that the ground vibration transmissibility using compliance system by using displacement cancellation control. Journal of
the displacement cancellation technique could be improved by Vibration and Control, 16(4), 585–599.
inserting an electric low-pass filter (see Fig. 8) into the feedback- Mohamed, Z., Martins, J. M., Tokhi, M. O., Sada Costa, J., & Botto, M. A. (2002).
Vibration control of a very flexible manipulator system. Control Engineering
loop from the displacement of middle table to the actuator of the Practice, 13(3), 267–277 (Aerospace IFAC).
isolation table. Rivin, E. I. (1995). Vibration isolation of precision equipment. Precision Engineering,
17, 41–56.
Rivin, E. I. (2003). Passive vibration isolation (pp. ix–xv)New York: ASME Press.
Shahadat, M. M. Z., Mizuno, T., Ishino, Y., & Takasaki, M. (2011). Development of
6. Conclusions three-axis active vibration isolator using displacement cancellation technique.
Journal of System Design and Dynamics, 5(5), 1077–1093.
The zero-compliance against direct disturbance and the soft Shahadat, M. M. Z., Mizuno, T., Ishino, Y., & Takasaki, M. (2012). Improvement of
transient characteristics of negative stiffness control system by feedforward
suspension against ground vibration can be acquired by using either control. Journal of the Japan Society of Applied Electromagnetics and Mechanics,
the negative stiffness technique or the displacement cancellation 20(3), 619–625.
66 M.Md.Z. Shahadat et al. / Control Engineering Practice 37 (2015) 55–66

Shahadat, M. M. Z., Mizuno, T., Ishino, Y., & Takasaki, M. (2014). Effect of control). Transaction of Japan Society of Mechanical Engineers, C, 59(562),
nonlinearity caused by friction on a negative stiffness control system. IEEE 1694–1701.
Transactions on Control Systems Technology, 22(4), 1385–1395. Yoshioka, H., Takahashi, Y., Katayama, K., Imazawa, T., & Murai, N. (2001). An active
Tokhi, M. O., & Hossain, M. A. (1996). A unified adaptive active control mechanism micro vibration isolation system for hi-tech manufacturing facilities. ASME
for noise cancellation and vibration suppression. Mechanical System and Signal Journal of Vibration and Acoustics, 123, 269–275.
Processing, 10(6), 667–682. Zhang, C. L., Mei, D. Q., & Chen, Z. C. (2002). Active vibration isolation of a micro-
Trumper, D. L., & Sato, T. (2002). A vibration isolation platform. Mechatronics, 12, manufacturing platform based on a neural network. Material Processing
281–294. Technology, 129, 634–639.
Watanabe, K., Hara, S., Kanemitsu, Y., Haga, T., Yano, K., & Mizuno, T. (1996). Zhu, W. H., Tryggvason, B., & Piedboeuf, J. C. (2006). On active acceleration control
Combination of h1 and pi control for an electromagnetically levitated vibra- of vibration isolation systems. Control Engineering Practice, 14(8), 863–873.
tion isolation system. In Proceedings of 35th Conference on Decision and Control
(pp. 1223–1228).
Yasuda, M., & Ikeda, M. (1993). Double-active control of micro vibration isolation
systems to improve performances (application of two-degree-of-freedom