ARTICLE IN PRESS

Control Engineering Practice 15 (2007) 187–196

www.elsevier.com/locate/conengprac

Vibration isolation system combining zero-power

magnetic suspension with springs
Takeshi Mizuno, Masaya Takasaki, Daisuke Kishita, Keiichiro Hirakawa
Department of Mechanical Engineering, Saitama University, Shimo-Okubo 255, Sakura-ku, Saitama 338-8570, Japan
Received 21 February 2004; accepted 12 June 2006
Available online 8 August 2006

Abstract

An active vibration isolation system using zero-power magnetic suspension was studied analytically and experimentally. Since a zero-
power magnetic suspension system behaves as if it has negative stiffness, connecting it in series with a normal spring can generate infinite
stiffness against disturbances acting on an isolation table. This property is maintained even when a suspension that supports the weight
of the isolation table is introduced in parallel with the serial combination. Analysis of a simple model involving a single-axis apparatus
clarified the fundamental characteristics of the system. The force–displacement characteristics of the zero-power magnetic suspension
were measured to estimate the negative stiffness properties. It was experimentally confirmed that combining a zero-power magnetic
suspension with a normal spring generates high stiffness against static direct disturbances acting on an isolation table.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: Magnetic suspension; Magnetic bearings; Active control; Vibration damper; Springs; Accelerometers

1. Introduction and higher stiffness is inevitable, so that performance is

Vibration isolation plays an important role in the operation
of many advanced production and measurement systems
such as semiconductor manufacturing and high-precision
measuring (Rivin, 1995, 2003; Yoshioka et al., 2001). Rapid
technological developments in these fields have increased the
need for higher-performance vibration isolation systems.
There are two kinds of vibration that must be reduced by a
vibration isolation system: (1) vibration transmitted from the
ground through the suspension (spring) and (2) vibration
caused by disturbances acting on an isolation table directly
(direct disturbance). A suspension with less stiffness is better
for reducing the former because dynamic coupling between
the vibration source and the isolation table is weakened; thus,
zero stiffness is ideal in this case. However, higher stiffness is
better for suppressing the latter because it reduces displace-
ment of the isolation table from its desired position; thus,
infinite stiffness is ideal in this case. In conventional passive-
type vibration isolation systems, a trade-off between lower
Corresponding author. Tel.: +81 48 858 3455; fax: +81 48 856 2577.
E-mail address: mizar@mech.saitama-u.ac.jp (T. Mizuno).
limited.
In contrast, active-type vibration isolation systems do
not, in principle, suffer from such performance limitations
(Rivin, 2003). One of the fundamental control strategies is
to provide feedback regarding the absolute velocity of the
isolation table—a process often referred to as skyhook
damping (Fuller et al., 1996). The performance limitations
of passive-type vibration isolation systems can be effec-
tively overcome by two-degrees-of-freedom control (Miya-
zaki et al., 1994; Yasuda et al., 1996; Yoshioka et al., 2001).
While various other control methods, such as state
feedback (Miyazaki et al., 1994), H 1 (Watanabe et al.,
1996), feedforward (Mohamed et al., 2005), repetitive
(Daley et al., 2006) and active acceleration (Zhu et al.,
2006) control have also been applied, most active vibration
isolation systems use high-performance sensors, such as
servo-type accelerometers, to detect the vibration of an
isolation table with high sensitivity in a low-frequency
domain. Since these sensors are costly, active systems are
more expensive than passive systems, sometimes by a
factor of 10, which represents a critical obstacle to
expanding their fields of application.

0967-0661/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conengprac.2006.06.001
 ARTICLE IN PRESS
188 T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196

This study presents a novel solution to the problem of Permanent magnet for zero-power control
vibration isolation. In the zero-power magnetic suspension
system described here, the suspension force is generated by
a hybrid magnet consisting of an electromagnet and a
i
permanent magnet. The permanent magnet produces the
bias flux that is necessary for suspending the weight of the
suspended object. The electromagnet is controlled to
stabilize the suspension system and to make the steady
deviation of the coil current converge to zero. As a result of
this type of control, the zero-power magnetic suspension
x
system behaves as if it has negative stiffness. Infinite m
stiffness against direct disturbances acting on the isolation
table can then be achieved by combining the system with a
normal spring (Mizuno, 2002). As a result, direct vibration Fig. 1. Basic model of zero-power magnetic suspension system.
is suppressed while system vibration transmitted from the
ground is reduced. Zero-power magnetic suspension
systems usually need only relative-displacement sensors, The transfer function representation of the dynamics
which cost far less than servo-type accelerometers. More- described by Eq. (1) becomes
over, sensor hardware can be omitted by applying self- 1
sensing techniques (Mizuno and Bleuler, 1995; Mizuno et X ðsÞ ¼ ðb0 IðsÞ þ d 0 F d ðsÞÞ, (2)
t0 ðsÞ
al., 1998; Vischer and Bleuler, 1993).
This paper is organized as follows. First, zero-power in which each Laplace-transformed variable is denoted by
control of magnetic suspension systems is briefly described. its capital, and
Second, the basic configurations of the proposed vibration
t0 ðsÞ ¼ s2
a0 , (3)
isolation system are presented. Third, the fundamental
properties of the system are clarified analytically. Fourth, ks ki 1
experimental results are provided to demonstrate the a0 ¼ ; b0 ¼ ; d0 ¼ .
m m m
effectiveness of the proposed method of generating infinite
stiffness.
2.2. Zero-power control

2. Zero-power magnetic suspension Zero-power control operates to accomplish

lim iðtÞ ¼ 0 for stepwise disturbances. (4)
t!1
2.1. Basic model
In discussing the zero-power control, therefore, distur-
Due to its power-saving properties, zero-power control bance should be considered to be stepwise
has been used in magnetic suspension systems such as F0
momentum wheels for spacecraft stabilization (Sabnis F d ðsÞ ¼ . (5)
s
et al., 1975) and carrier systems in clean rooms (Morishita
et al., 1989). In this form of control, a hybrid magnet Considering linear output feedback control, the control
consisting of an electromagnet and a permanent magnet is input can be represented as
used. Control of the electromagnet makes steady deviation hðsÞ
of the coil current converge to zero. As a result, the air-gap IðsÞ ¼
X ðsÞ, (6)
gðsÞ
length is maintained so that the attractive force generated
by the permanent magnet balances the other, static forces where gðsÞ and hðsÞ are coprime polynomials in s. It is
acting on the suspended object. assumed in the following that they are selected for the
Fig. 1 shows a single-degree-of-freedom-of-motion closed-loop system to be stable.
model for analysis (Mizuno and Takemori, 2002). The Substituting Eq. (6) into Eq. (2) gives
suspended object with a mass of m is assumed to move only gðsÞ
X ðsÞ ¼ d 0 F d ðsÞ, (7)
in the vertical direction translationally. The equation of tðsÞ
motion is given by
where
€ ¼ ks xðtÞ þ ki iðtÞ þ f d ðtÞ,
mxðtÞ (1)
tðsÞ ¼ ðs2
a21 ÞgðsÞ þ a23 hðsÞ. (8)
where x is the displacement of the suspended object, ks and Substituting Eqs. (5) and (7) into Eq. (6) gives
ki are, respectively, the gap-force and current-force
coefficients, i is the control current, and f d is the hðsÞ F0
IðsÞ ¼
d0  . (9)
disturbance force acting on the suspended object. tðsÞ s
 ARTICLE IN PRESS
T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196 189

To achieve the control object described by Eq. (4), hðsÞ The negative sign on the right-hand side indicates that the
must satisfy new equilibrium position is in a direction opposite to the
applied force.
~
hðsÞ ¼ shðsÞ, (10) When a stepwise force acts on the suspended object, it
first moves in the same direction as the force, and then to
~
where hðsÞ is an appropriate polynomial. Therefore, the the new equilibrium position given by Eq. (14) (Mizuno
control input achieving zero-power control is represented and Takemori, 2002). The zero-power magnetic suspension
by system, therefore, has negative stiffness in a static sense; the
~ amplitude of the stiffness is identical to the gap-force
shðsÞ coefficient of the hybrid magnet ks .
IðsÞ ¼
X ðsÞ. (11)
gðsÞ

Fig. 2 shows the general form of a zero-power controller 3. Vibration isolation system
(Mizuno and Takemori, 2002).
3.1. Concept
2.3. Negative stiffness
First, it will be shown that infinite stiffness can be
realized by connecting a normal spring with a spring that
A unique characteristic of the zero-power control system
has negative stiffness. When two springs with spring
is that it behaves as if it has negative stiffness. When an
constants k1 and k2 are connected in series, as shown by
external force is applied to the mass in a common
Fig. 4, the total stiffness kc is given by
mass–spring system, the mass moves to the direction of
the applied force, as shown in Fig. 3a. In the system k1 k2
controlled by zero-power, the suspended object moves to a kc ¼ . (15)
k1 þ k2
new equilibrium position located in the direction opposite
to the applied force, as shown in Fig. 3b. This characteristic This equation shows that when normal springs are
is shown analytically in the following. connected the total stiffness becomes lower than the
At steady state, the suspended object is maintained at a stiffness of each spring. However, if one of the springs
position such that has negative stiffness
0 ¼ ks xð1Þ þ ki ið1Þ þ F 0 . (12)
k1 ¼
k2 , (16)
In the zero-power control system, the coil current is
controlled to converge to zero, that is the resultant stiffness becomes infinite:
ið1Þ ¼ 0. (13)
jkc j ¼ 1. (17)
Therefore,
F0
xð1Þ ¼
. (14)
ks

k2
Controller
~ i
x h (s) k1
s
g (s)

Fig. 2. General form of zero-power controller. Fig. 4. Series springs.

∠ mg ∠ mg
(b)
(a)

Fig. 3. Comparison of zero-power magnetic suspension system (b) with a normal spring (a).
 ARTICLE IN PRESS
190 T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196
In the present study, this principle of generating high
stiffness is applied to vibration isolation systems to
counteract direct disturbances acting on the isolation table.
The application of negative stiffness to vibration
isolation was previously reported by Platus (1999) and
Trimboli et al. (1994). In their work, a negative spring was
used in parallel with a positive spring to lower the stiffness
of suspension of the isolation table. In our approach, by
contrast, serial connection of a negative spring and a
positive spring is applied to vibration isolation systems.
3.2. System configuration
Fig. 5 shows the configuration of one of the proposed
vibration isolation systems. A middle mass m1 is connected
to the base through a spring k1 and a damper c1 , which
together work as a conventional vibration isolator. An
electromagnet for zero-power magnetic suspension is fixed
to the middle mass. The part of an isolation table m2 facing
the electromagnet is made of a soft iron material for
confining the magnetic fields produced by the permanent
magnets for zero-power control. This is referred to as the
reaction part.
This system can reduce vibration transmitted from
ground by making k1 small and, at the same time, infinite
stiffness can be produced to counteract direct disturbances
by setting the amplitude of negative stiffness equal to k1 .
To explain the latter more concretely: it is assumed that the
table is subject to a downward force, so that the gap
between the electromagnet and the table becomes smaller
because of the zero-power control; that is, the table would
move upwards if the middle mass were fixed.
Meanwhile, the middle mass moves downwards because
of the increased electromagnetic force. The decrease in the
gap is cancelled by the downward displacement of
the middle mass due to the above-mentioned setting. Thus,
the isolation table is maintained at the same position as
before. The proposed characteristics of the system will first
be shown analytically and then experimentally.
x2
m2
x1
m1
c1
k1
Permanent magnets
x0
Fig. 5. Basic structure of vibration isolation system using zero-power
magnetic suspension.
3.3. Combination with a weight suspension mechanism
In the system shown by Fig.5, the entire weight of the
isolation table is supported by zero-power magnetic
suspension. When the isolation table is large, therefore,
many permanent magnets are needed to suspend its weight,
which will raise the cost of system. Another problem that
can be expected in putting the proposed system to practical
use is that the reaction part must be installed under the
middle table (Fig. 5), because the hybrid magnet can
produce only an attractive force. This makes the structure
of the vibration isolation system rather complex.
These problems can be overcome by introducing an
auxiliary suspension for supporting the weight of the
isolation table, as explained in Fig. 6. A spring kd is added
in parallel with the serial connection of the positive and
negative springs. The total stiffness k~c is given by
k1 k2
k~c ¼ þ kd . (18)
k1 þ k2
When Eq. (16) is satisfied, the resultant stiffness becomes
infinite for any finite value of kd
jk~c j ¼ 1. (19)
Fig. 7 shows the configuration of one of the proposed
vibration isolation systems. A spring kd together with a
damper cd is inserted between the isolation table and the
base. The spring is set to produce upward force in the
k2
kd
k1
Fig. 6. Parallel connection of series springs and a spring.
x2
m2
x1
m1
k1 c1
kd cd
x0
Fig. 7. Modified structure of vibration isolation system using zero-power
magnetic suspension.
 ARTICLE IN PRESS
T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196 191

x2 kd Dx2
m2 g þ f̄ e ¼ 0. (24)
m2
According to Eq. (24),

f̄ e ¼ m2 g
kd Dx2 , (25)
x1
which indicates that the steady-state force to be generated
m1
kd cd by the hybrid magnet can be reduced by using the upward
force of the spring kd . In addition, when f̄ e o0, the
configuration shown by Fig. 8 can be applied.
k1
c1
According to the discussion in Section 2.2, the control
x0 current achieving the zero-power control is generally
represented by
IðsÞ ¼
c2 ðsÞsðX 2 ðsÞ
X 1 ðsÞÞ, (26)
Fig. 8. Another modified configuration of vibration isolation system.
where
~
hðsÞ
equilibrium state. It reduces the static load force that the c2 ðsÞ ¼ . (27)
gðsÞ
zero-power magnetic suspension has to support. Moreover,
when the upward force is greater than the gravitational
force, the zero-power magnetic suspension must produce 4.2. Response to direct disturbance
downward force so that the configuration is modified, as
shown in Fig. 8. Since the reaction part is installed above For simplicity it is assumed that the initial values are
the middle table, the structure is simpler than the original zero. From Eqs. (20) to (24) and (26), we get
one shown in Fig. 5. It should be noted that isolation from k~1 ðsÞðt2 ðsÞ þ k~2 ðsÞÞ k~2 ðsÞ
ground vibration can be maintained by using a soft spring X 1 ðsÞ ¼ X 0 ðsÞ þ F d ðsÞ, (28)
tc ðsÞ tc ðsÞ
as kd .
k~1 ðsÞk~2 ðsÞ t1 ðsÞ þ k~2 ðsÞ
X 2 ðsÞ ¼ X 0 ðsÞ þ F d ðsÞ, (29)
4. Analysis tc ðsÞ tc ðsÞ
where
4.1. Basic equations
t1 ðsÞ ¼ m1 s2 þ c1 s þ k1 , (30)
The model shown in Fig. 7 is treated because it includes
the model shown in Fig. 5 as a special case, in which kd ¼ 0 t2 ðsÞ ¼ m2 s2 þ cd s þ kd , (31)
and cd ¼ 0. The equations of motion for the translation
motion in the vertical direction are k~1 ðsÞ ¼ c1 s þ k1 , (32)
m1 x€ 1 ¼
m1 g þ k1 Dx1
k1 ðx1
x0 Þ
c1 ðx_ 1
x_ 0 Þ
f e , k~2 ðsÞ ¼ ki c2 ðsÞs
ks , (33)
(20)
tc ðsÞ ¼ t1 ðsÞt2 ðsÞ þ k~2 ðsÞðt1 ðsÞ þ t2 ðsÞÞ. (34)
m2 x€ 2 ¼
m2 g þ kd Dx2
kd ðx2
x0 Þ
cd ðx_ 2
x_ 0 Þ
To estimate the stiffness for the direct disturbance f d , it is
þ f e þ f d, ð21Þ assumed to be stepwise
where x0 , x1 , and x2 are, respectively, displacements of F0
Fd ¼ ðF 0 : constÞ. (35)
the floor, the middle mass, and the isolation table, f e is s
the control force produced by the electromagnet, f d is the When the vibration of the floor is neglected (x0 ¼ 0Þ, the
direct disturbance acting on the isolation table, Dx1 is steady-state displacement of the table is obtained as
the initial compressed length of spring k1 , and Dx3 is the
initial compressed length of spring kd . x2 ð1Þ t1 ðsÞ þ k~2 ðsÞ k1
ks
¼ lim ¼ . (36)
The attractive force of the hybrid is approximately F0 s!0 tc ðsÞ ðk1
ks Þkd
ks k1
represented by
Therefore, when
f e ¼ f̄ e þ ks ðx2
x1 Þ þ ki i, (22) k1 ¼ ks , (37)
where f̄ e is the attractive force in the equilibrium state, in for any finite value of kd , we get
which the following equations are satisfied
x2 ð1Þ
¼ 0. (38)
k1 Dx1
m1 g
f̄ e ¼ 0, (23) F0
 ARTICLE IN PRESS
192 T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196

Eq. (38) shows that the suspension system between the

isolation table and the floor has infinite stiffness statically
because there is no steady-state deflection even in the
presence of the stepwise disturbance acting on the table.
The steady-state displacement of the middle mass and
the variation of the gap are obtained as
x1 ð1Þ
ks 1
¼ ¼ , (39)
F0
ks k1 k1

x1 ð1Þ
x2 ð1Þ
k1 1

¼ ¼ . (40)
F0
ks k1 ks

Eqs. (39) and (40) show that the middle mass moves
downward and the gap decreases when a downward force
acts on the isolation table (F 0 o0Þ. This well supports
predictions regarding the behavior of the proposed
vibration isolation system, as described in Section 3.2.

5. Experiment
Fig. 10. Photo of experimental apparatus.
5.1. Experimental apparatus and methods

Figs. 9 and 10 are a schematic diagram and a
photograph of a single-axis apparatus that was built for
the experimental study of the original system (Fig. 5). It is
not equipped with any of the weight suspension mechan-
isms shown in Figs. 7 and 8. The height, diameter and mass
of the apparatus are 200, 226 mm and 18 kg, respectively.
The isolation table and the middle mass weigh 3.5 and 5 kg,
respectively, and are guided to move translationally in the
vertical direction by linear air bearings. A ring-shape
electromagnet with a 448-turn coil is fixed to the middle
mass corresponding to m1 in Fig. 5; its inner and outer
diameters are 68 and 138 mm, respectively. Ten
10  10  5-mm permanent magnets made of NdFeB
provide bias flux. These magnets, rather than the electro-
magnet, are built in the reaction part of the isolation table.
This configuration widens the range of operation because
Isolation table
Sensor
Auxiliary
electromagnet
Electromagnet
for suspension
Permanent Base
magnet
Reaction part
Base
Middle mass
Sensor
Fig. 9. Schematic diagram of experimental apparatus.
repulsive force can be generated (Mizuno and Takemori,
2002). The nominal gap between the electromagnet and the
permanent magnets is about 3 mm. The middle mass is
suspended by four mechanical springs. An electromagnet
for adjusting the positive stiffness k1 and the damping c1 is
installed on the base, and its reaction part is built in the
middle mass. The electromagnet is referred to as an
auxiliary electromagnet and is used to equalize the positive
stiffness and the amplitude of the negative stiffness, as
shown by Eq. (37) in the following experiment. The relative
displacement of the middle mass to the base and that of the
isolation table to the middle mass are detected by eddy-
current gap sensors.
In the experiments, we use a zero-power controller in the
form of

 Z
d
i ¼
pd þ pv ðx2
x1 Þ þ pi i dt. (41)
dt
Air bearing This is a combination of PD (proportional-derivative)
control and a local integral feedback of current (Mizuno
Middle mass and Takemori, 2002). Eq. (41) states that
Base sðpd þ pv sÞ
IðsÞ ¼
ðX 2 ðsÞ
X 1 ðsÞÞ ¼
c2 ðsÞsðX 2 ðsÞ
s
pi

X 1 ðsÞÞ, ð42Þ
[See Eq. (26)]. The designed control algorithm is imple-
mented with a digital controller. The control period is
100 ms. In the actual operation, stable magnetic suspension
is achieved by initially activating only the PD control.
Subsequently, the integral feedback for zero-power control
is activated. This procedure is effective in stably achieving
 ARTICLE IN PRESS
T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196 193

zero-power suspension. The feedback gains pd , pv and pi
are tuned by trial and error.
5.2. Experimental results
5.2.1. Estimation of negative stiffness
To estimate the negative stiffness of the zero-power
magnetic suspension, its force–displacement characteristics
are measured when the middle mass is fixed; downward
force is produced by placing weights on the isolation table.
Fig. 11 presents the measurement results. The upward
displacement of the isolation table is plotted against the
downward force produced by the weights. As shown in the
figure, the direction of the displacement is opposite to that
of the applied force so that the stiffness is negative. Fig. 12
shows the magnitude of negative stiffness versus the
applied force, which is calculated based on the measure-
ment results shown in Fig. 11. As the downward force
increases, the gap between the electromagnet and the
reaction part decreases so that the gap-displacement
2 estimated stiffness between the isolation table and the base
1.5
Displacement [mm]
coefficient ks becomes larger. As a result, the amplitude
of negative stiffness also becomes larger. However, it will
be assumed to be constant in the following experiments, in
which the zero-power magnetic suspension system is
combined with a suspension mechanism with positive
stiffness. The average value of ks is 14:3 kN=m in the
range of force 0–9 N, which is treated as a nominal value.
5.2.2. Responses to direct disturbance
In the second experiment, the middle mass is released so
that it is supported by the springs, and weights are again
added onto the isolation table as a direct disturbance. Since
the positive stiffness by the springs is 12:5 kN=m, it is
adjusted to equal the nominal value 14:3 kN=m by the
auxiliary electromagnet. It should be noted that this type of
adjustment can be achieved by changing the springs.
Fig. 13 shows the displacement of the isolation table to
the base, that of the isolation table to the middle mass, and
that of the middle mass to the base. The figure shows that
the position of the table is maintained at the same position
while the position of the middle mass changes proportion-
ally to the force applied to the isolation table. The
is 892 kN/m in this region, which is about 63 times k1 and
ks ð14:3 kN=mÞ. This result demonstrates well that combin-
ing a zero-power magnetic suspension with a normal spring

can generate high stiffness against a static direct dis-

1
turbance acting on the isolation table.
Since the magnitude of negative stiffness is a function of
the gap between the electromagnet and the reaction part,
0.5 the stiffness against the direct disturbance will decrease
when the amplitude of the disturbance exceeds a certain
level. Two approaches are proposed for resolving this
0 problem. One is to use a linear actuator instead of the
0 10 20 30 40 50 hybrid magnet to produce negative stiffness (Mizuno et al.,
Force [N] 2003). The other is to use a nonlinear spring to produce
Fig. 11. Force–displacement characteristics of the zero-power magnetic positive stiffness (Mizuno et al., 2003).
suspension.

0.8

0.6
80
0.4
Displacement [mm]

60 0.2
Stiffness [kN/m]

40 -0.2

-0.4

20 -0.6

-0.8
0 2 4 6 8 10
0 Force [N]
0 10 20 30 40 50
Force [N] Fig. 13. Displacements of the isolation table and the middle mass when
downward force acts on the table: K displacement of the isolation table to
Fig. 12. Amplitude of the negative stiffness of the zero-power magnetic the base; m: displacement of the isolation table to the middle mass; ’:
suspension system. displacement of the middle mass to the base.
 ARTICLE IN PRESS
194 T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196
-20
-40
Gain [dB]
-60
Isolation table
Middle mass
-80
0.1 1 10
(a) Frequency [Hz]
0 -0.1
Isolation table
Middle mass
-90
Phase [deg]
-180
-270
-360
0.1 1 10
(b) Frequency [Hz]
Fig. 14. Frequency response of the vibration isolation system to direct
disturbance (a) Gain; (b) Phase.
Fig. 14 shows a frequency response of the system to
direct disturbance. A sinusoidal disturbance was produced
by an electromagnet, which was installed over the isolation
table for measurement. Since the electromagnet was driven
by a current-output power amplifier, the command signal
inputted to the amplifier was treated as an input signal. The
displacement of the isolation table to the base and that of
the middle mass to the base were treated as output signals.
As can be seen in the figure, the displacement of the
isolation table is reduced at a frequency range lower than
1 Hz. This result also supports the conclusion that the
proposed system can generate high stiffness against a static
direct disturbance acting on the isolation table.
Figs. 15 and 16 show the movements of the isolation
table, the middle mass and the control current of the
electromagnet when the frequency of the sinusoidal direct
disturbance is 0.25 and 6 Hz, respectively. When the
frequency is 0.25 Hz, the movement of the middle mass
and the relative displacement of the isolation table to the
middle mass are almost in opposite phase. They cancel
each other so that the movement of the isolation table
to the base is attenuated. When the frequency is 6 Hz,
the movement of the middle mass and the relative
displacement of the isolation table to the middle mass
are almost in phase so that the movement of the isolation
table to the base becomes larger than the movement of
the middle mass. These results indicate that cancellation
Middle mass to the base
isolation table to the middle mass
Isolation table to thebase
0.15
0.1
Displacement [mm]
0.05
0
-0.05
-0.15
(a) Time [1s/div]
0.1
0.05
Current [A]
0
-0.05
-0.1
(b) Time [1s/div]
Fig. 15. Response to a sinusoidal direct disturbance with a frequency of
0.25 Hz. (a) Displacements; (b) Control current.
of the displacement can be achieved in the low-frequency
range by connecting the positive and negative springs in
series.
Fig. 17 shows a response to a stepwise direct disturbance
that was produced by the electromagnet over the isolation
table. An upward force applied to the isolation table
initially was quickly removed by making the coil current
zero. The middle mass begins to move downward and stay
at a position that is lower than the initial position. The
relative displacement of the isolation table to the middle
mass is negative just after the applied force is removed and
then positive at steady state. The former displacement is
cancelled by the latter one so that the position of the
isolation table returns to its initial position at steady state.
The dynamic performances of the system, i.e., its
responses to sinusoidal and stepwise direct disturbances,
depend on the control performance. In this work, the
controller was tuned by trial and error, as mentioned
above. To improve more effectively the dynamic perfor-
mance of the system, further intensive study on the
applications of advanced-control design methods will be
necessary. Since the performances of the system also
depends on physical parameters such as k1 , c1 and m2 ,
 ARTICLE IN PRESS
T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196 195

Middle table to the base Isolation table to the base

Isolation table to the middle mass Isolation table to the middle mass
Isolation table to the base Middle mass to the base
0.2
0.15

0
0.1
-0.2
Displacement [mm]

0.05
-0.4
0
0.2

Displacements [mm]
-0.05
0

-0.1 -0.2

-0.15 -0.4
(a) Time [0.05s/div]

0.1 0.4

0.2

0.05 0
Current [A]

-0.2
0 Time [0.2s/div]

Fig. 17. Response to a stepwise direct disturbance.

-0.05
Acknowledgments
-0.1
(b) Time [0.05s/div] This work is financially supported in part by a Grant-in-
Aid for the Creation of Innovations through Business-
Fig. 16. Response to a sinusoidal direct disturbance with a frequency of Academic-Public Sector Cooperation, a Grant-in-Aid for
6 Hz. (a) Displacements; (b) Control current.
Scientific Research (B) from the Ministry of Education,
Culture, Sports, Science and Technology of Japan. The
the integrated design of mechanism and control using authors thank Mr. T. Oya, an undergraduate student at
optimization techniques offers a promising approach to Saitama University, for help with the experiments.
optimizing performance.

6. Conclusions
An active vibration isolation system combining zero-
power magnetic suspension with a conventional suspension
was studied analytically and experimentally. Since a zero-
power system behaves as if it has negative stiffness,
connecting it with a normal spring in series can generate
infinite stiffness against disturbances acting on the isolation
table. This property is maintained even if a suspension to
support the weight of the isolation table is introduced in
parallel with the serial combination. The principles and
fundamental characteristics of the system were studied
using a single-axis apparatus developed for the experi-
ments. The force–displacement characteristics of the zero-
power magnetic suspension were measured to estimate the
negative stiffness properties. It was experimentally con-
firmed that combining a zero-power magnetic suspension
with a normal spring in series generates high stiffness
against static direct disturbance acting on an isolation
table.
References
Daley, S., Hätönen, J., & Owens, D. H. (2006). Active vibration isolation
in a ‘‘smart spring’’ mount using a repetitive control approach. Control
Engineering Practice, 14(9), 991–997.
Fuller, C. R., Elliott, S. J., & Nelson, P. A. (1996). Active control of
vibration (pp. 213–220). New York: Academic Press.
Miyazaki, T., Mizuno, K., Kawatani, R., & Hamada, H. (1994).
Consideration about feedback feedforward hybrid control for active
control of micro-vibration control. Proceedings of the second interna-
tional conference on motion and vibration control (pp. 29–34).
Mizuno, T. (2002). Vibration isolation system using zero-power magnetic
suspension. Preprints of 15th World Congress IFAC (955pp.).
Mizuno, T., & Bleuler, H. (1995). Self-sensing magnetic bearing control
system design using the geometric approach. Control Engineering
Practice, 3(7), 925–932.
Mizuno, T., Ishii, T., & Araki, K. (1998). Self-sensing magnetic sus-
pension using hysteresis amplifiers. Control Engineering Practice, 6,
1133–1140.
Mizuno, T., Iwashita, S., Takasaki, M., & Ishino, Y. (2003). Vibration
isolation system combining zero-power magnetic suspension with a
magnetic spring. Proceedings of the Asia-Pacific Vibration Conference
2003 (pp. 469–474).
 ARTICLE IN PRESS
196 T. Mizuno et al. / Control Engineering Practice 15 (2007) 187–196
Mizuno, T., & Takemori, Y. (2002). A transfer-function approach to the
analysis and design of zero-power controllers for magnetic suspension
system. Electrical Engineering in Japan, 141(2), 67–75.
Mizuno, T., Toumiya, T., & Takasaki, M. (2003). Vibration isolation
system using negative stiffness. JSME International Journal, Series C,
46(3), 807–812.
Mohamed, Z., Martins, J. M., Tokhi, M. O., Sáda Costa, J., &
Botto, M. A. (2005). Vibration control of a very flexible manipulator
system. Control Engineering Practice, 13(3) Aerospace IFAC 2002
(pp. 267–277).
Morishita, M., Azukizawa, T., Kanda, S., Tamura, N., & Yokoyama, T.
(1989). A New Maglev system for magnetically levitated carrier
system. IEEE Transactions on Vehicular Technology, 38(4), 230–236.
Platus, D. L. (1999). Negative-stiffness-mechanism vibration isolation
systems. Proceedings of the SPIE—the international society for optical
engineering (Vol. 3786, pp. 98–105).
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.
Sabnis, A. V., Dendy, J. B., & Schmitt, F. M. (1975). A magnetically
suspended large momentum wheel. Journal of Spacecraft, 12, 420–427.
Trimboli, M. S., Wimmel, R., Breitbach, E. (1994). A quasi-active
approach to vibration isolation using magnetic springs. Proceedings of
the SPIE—the international society for optical engineering (Vol. 2193,
pp. 73–83).
Vischer, D., & Bleuler, H. (1993). Self-sensing active magnetic levitation.
IEEE Transactions on Magnetics, 29(2), 1276–1281.
Watanabe, K., Hara, S., Kanemitsu, Y., Haga, T., Yano, K.,
Mizuno, T., et al. (1996). Combination of H1 and PI control
for an electromagnetically levitated vibration isolation system.
Proceedings of the 35th conference on decision and control
(pp. 1223–1228).
Yasuda, M., Osaka, T., & Ikeda, M. (1996). Feedfoward control
of a vibration isolation system for disturbance suppression. Pro-
ceedings of the 35th conference on decision and control
(pp. 1229–1233).
Yoshioka, H., Takahashi, Y., Katayama, K., Imazawa, T., & Murai, N.
(2001). An active microvibration isolation system for hi-tech manu-
facturing facilities. ASME Journal of Vibration and Acoustics, 123,
269–275.
Zhu, W. H., Tryggvason, B., & Piedboeuf, J. C. (2006). On active
acceleration control of vibration isolation systems. Control Engineering
Practice, 14(8), 863–873.