Modeling and Sensitivity Analysis

Modeling and sensitivity analysis of a pneumatic vibration isolation system
with two air chambers

Jun-Hee Moon
a,
, Bong-Gu Lee
b
a
Department of Mechatronics, Daelim University College, 526-7, Bisan-dong, Dongan-gu, Anyang-si, Gyeonggi-do, 431-715, South Korea
b
Department of Mechanical Engineering, Daelim University College, 526-7, Bisan-dong, Dongan-gu, Anyang-si, Gyeonggi-do, 431-715, South Korea
a r t i c l e i n f o a b s t r a c t
Article history:
Received 17 August 2009
Received in revised form 12 August 2010
Accepted 14 August 2010
Available online 17 September 2010
This paper aims at accurate modeling and sensitivity analysis for a pneumatic vibration isolation
system (PVIS) as a foundation for practical design. Even though the PVIS is widely used for its
effective performance in vibration isolation, its design has depended largely on trial-and-error
methods. In previous studies, nonlinear characteristics of the diaphragm and the air ow
restrictor, which signicantly affect the performance of a PVIS, have been investigated. However,
several hurdles, such as the absence of a mathematical model for the diaphragm, still remain with
regard to the model-based prediction of performance. Therefore, a fractional derivative model for
the diaphragm and a quadratic damping model for the air ow restrictor are newly developed
based on the careful examination of previous studies. Then, sensitivities of vibration isolation
performance indices with regard to major design variables are analyzed and new approximation
formulas are created based on the dynamic characteristics of the PVIS. Our models with a
transmissibility-computing algorithm are veried by comparison with experimental data. The
sensitivity analyses andapproximationformulas are expectedtobe useful for practical PVISdesign
owing to their simplicity and accuracy.
2010 Elsevier Ltd. All rights reserved.
Keywords:
Pneumatic vibration isolation
Diaphragm
Air ow restrictor
Fractional derivative
Equivalent mechanical system
Sensitivity analysis
1. Introduction
As high precision industries such as semiconductor production, precisionmetrology, optics, and microbiology continue to grow,
higher performance vibration isolation systems are needed to meet the corresponding vibration tolerance requirements [13]. To
achieve vibration isolation for local precision equipment, a pneumatic vibration isolation system (PVIS) is widely used because it
needs no energy supply and no control unit, and performs stable and effective vibration attenuation across a wide frequency range.
Even though a PVIS is very useful, its design for better vibration isolation has depended largely on trial and error methods.
Vibration isolation performance enhancement of the PVIS has been attempted by a variety of ways such as reshaping of
elastomeric diaphragm[4], usage of an air owrestrictor of porous media [5], energy dissipation by a gimbal piston in oil chamber
[6], parallelization with a negative-stiffness device [7] and adoption of active control schemes [811]. In all those attempts, the
basic work is the modeling of components of the PVIS since effects of design variables on vibration isolation performance can be
predicted only with accurate mathematical models and corresponding computational techniques.
In this paper, our goal is accurate modeling and vibration isolation performance evaluation of a PVIS such that our results may
be used to predict the performance for practical PVIS design. We examined previous studies and determined that three additional
efforts are required.
The rst is to make a mathematical model of the diaphragm. Some studies pointed out that the diaphragm has an important
role in the elastic and damping characteristic of a PVIS [4,12,13]. However, its nonlinear properties have been thus far neglected in
Mechanism and Machine Theory 45 (2010) 18281850
Corresponding author. Tel.: +82 31 467 4687; fax: +82 31 467 4869.
E-mail address: junheemoon@gmail.com (J.-H. Moon).
0094-114X/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmachtheory.2010.08.006
Contents lists available at ScienceDirect
Mechanism and Machine Theory
j our nal homepage: www. el sevi er. com/ l ocat e/ mechmt
Nomenclature
A area
b
1
, b
2
, b
3
tuned coefcient for approximation formulas
c
E0
, c
E1
, c
E3
material constants of the diaphragm
C damping coefcient
d diameter
E elastic modulus
f force variation
F Fourier transform of f
fr friction coefcient
g acceleration of gravity
G complex nonlinear mapping
h height
H transmissibility
i imaginary unit
I mechanical impedance
K stiffness
L loss coefcient
m air mass in the chamber
M tabletop mass
N volume ratio of damping chamber to spring chamber
P pressure
Re Reynolds number
t time
u uid velocity
V volume
x displacement
X Fourier transform of x
y ordered pair of complex variables
acceleration parameter
E
shift factor
exponent of fractional derivative
the specic heat ratio (=1.4)
variation
strain
loss factor
viscosity
density
stress
time constant
angular velocity
gradient operator
Superscripts
maximum
minimum
time derivative
averaged
Subscripts
0 static or average
a air
at atmosphere
b base
c capillary tube or air ow restrictor
d diaphragm
1829 J.-H. Moon, B.-G. Lee / Mechanism and Machine Theory 45 (2010) 18281850
many cases, since the diaphragm have been treated as a linear element in most analyses [1417] or the properties of the
diaphragm have taken the form of lookup table even in some studies that regarded it as a nonlinear element [12,13].
The mathematical modeling of the diaphragm is essential since it makes transfer functions to be analytic so that the
transfer functions might be used not only for transmissibility calculation but also for sensitivity analysis and performance
optimization.
The second is to consider the air ow restrictor as a nonlinear damper, and the air volume through the restrictor as a
simultaneous independent variable. Some previous studies regarded the ow restrictor as a linear damper, in which case the
transfer function of the PVIS is easily formulated [12,14,18]. However, the nonlinearity of the damping characteristics of the ow
restrictor was demonstrated in many studies [11,1921]. Lee and Kim[13] advanced the analysis of a PVIS by considering the ow
restrictor as a nonlinear damper.
They made the approximation that the air volume passing through the restrictor is proportional to tabletop displacement.
However, since tabletop displacement is not proportional to base displacement of vibration, it is easily inferred that the air volume
coupled with both of them is not proportional to either one of them. The last is to analyze sensitivities of vibration isolation
performance indices to design variables based on nonlinear dynamic characteristics of the PVIS. Many design strategies were
developed using the assumption that nonlinear models can be approximated as linear models [5,6,14,18,22]. However, the design
strategies could not cope with the nonlinear characteristics of a PVIS.
Therefore, the sensitivity analyses using nonlinear models are required for accurate performance prediction and for efcient
design of a PVIS. Consideration of the rst two issues, which is presented in Section 2, results in the transfer functions of a PVIS
consisting of two simultaneous nonlinear complex equations. To derive and calculate the transfer functions, the equal energy
dissipation method is adopted, and an equivalent mechanical system and a recursive numerical method are devised in Section 3.
The models and the PVIS transfer functions are veried by experimental data in Section 4, in which the discrepancies between the
previous air ow restrictor models and experimental data lead to an adjustment of the model. Based on the dynamic
characteristics of the PVIS, sensitivities of PVIS performance indices to design variables are analyzed and approximation formulas
are created in Section 5. Our concluding remarks are given in Section 6.
In exploring the modeling of the PVIS, this research will be limited to considering only vertical rigid-body mode vibration of the
tabletop because horizontal vibration and tabletop exural modes can be controlled by very different mechanisms, which should
be subjects of distinct research [23].
2. Modeling of pneumatic vibration isolation system with two air chambers
A typical pneumatic vibration isolation system (PVIS), whose schematic is shown in Fig. 1, includes a piston, diaphragm, air
chambers, air ow restrictor and wire pendulum. The piston is attached to, and moves along with, the tabletop mass. The
diaphragmseals air in the air chambers and allows smooth movement of the piston. The air chambers are named according to their
functions in the PVIS: the spring chamber acts as a mechanical spring, and the damping chamber acts both as a spring and as a
damper together with the air ow restrictor. The air ow restrictor hampers air ow across the air chambers. The wire pendulum
isolates horizontal vibration and will not be mentioned further because horizontal vibration isolation is beyond the scope of this
study.
2.1. Air chambers
To simplify the analysis, thermodynamic condition of the air chambers is assumed as follows [4,1214]: pressure and
temperature in an air chamber is uniform, and the thermal process in an air chamber is adiabatic.
The pressure variation induced by the volume variation in the damping chamber is derived from the gas law for adiabatic
processes as follows:
P
D
= P
0
1 + V
D
V
D
_ _
1
_ _
1
e experiment
dp experimental data for the diaphragm
D damping chamber
i inertial
mp minimum resonant peak
n iteration number
p piston
S spring chamber
t transmitting
T experimental data for the total PVIS
where P
D
and V
D
are the pressure and volume of the damping chamber, respectively, and P
0
is the average air pressure. The air
volume that has left the damping chamber, V
D
, is equal to the air volume passed through the owrestrictor, and is positive when
air leaves the damping chamber as seen in Fig. 1.
The pressure variation of the spring chamber is derived fromthe gas lawfor adiabatic processes involving volume variations of
the spring and damping chambers as follows:
P
S
= P
0
1 +
V
S
V
D
V
S
_ _
1
_ _
2
where P
S
and V
S
are the pressure and volume of the spring chamber, respectively, and V
S
is the volume variation of the spring
chamber, which results in lifting up the piston, as seen in Fig. 2.
2.2. Diaphragm
The diaphragm in the PVIS is a thin ber-reinforced elastomeric membrane that prevents air leakage in the air chambers and
allows smooth movement of the piston as detailed in Fig. 2. The elasticity and damping properties of the diaphragm cannot be
measured directly because the shape of the diaphragm in the deated condition is quite different from that in the inated
condition, and the deated diaphragm is very difcult to handle due to its exibility. Therefore, the stiffness of the diaphragm is
obtained fromexperimental data, by subtracting the stiffness of an air chamber fromthe total stiffness of a PVIS having a single air
chamber [13].
Fig. 2. Schematic of diaphragm: ; static: , moved.
S
D
D
p
S
b
Fig. 1. Schematic of pneumatic vibration isolation system.
We developed a novel fractional derivative model for the diaphragm, which has not been reported in the literature as far as
we know. Following Eq. (3) is a typical form of the four-parameter fractional derivative Zener model for viscoelastic materials
[2426]:
t +
dt
t = t
dt
t 3
where (t) is the stress, (t) is the strain,

E and are the maximum and minimum of elastic modulus, respectively, is the time
constant, and is the exponent of the fractional derivative. Based on the fact that force is proportional to stress and displacement is
proportional to strain, Eq. (3) can be rewritten as
f
d
+
1
d
_ _
dt
f
d
=
K
d
x

K
d
1
d
_ _
dt
x 4
where f
d
is the force acting on the diaphragm, x is the displacement resulting from the rolling motion of the diaphragm,

K
d
and

K
d
are the maximumand minimum of the stiffness of the diaphragm, respectively, and
d
is the characteristic angular velocity of the
diaphragm. Applying a Fourier transform to Eq. (4) yields
F
d
X
=

K
d
+

K
d
i =
d

1 + i =
d

5
The effect of amplitude can be brought into Eq. (5) using a shift factor
E
as follows [27]:
F
d
=K
d
; jXj X 6
with
K
d
; j Xj =
K
d
+

K
d
i =
d

E
jXj
1 + i =
d

E
jX j
7
log
10

E
x =
c
E1
xc
E0

c
E2
+ xc
E0
8
where c
E0
, c
E1
, and c
E3
are material constants which can be determined experimentally from measured data. The stiffness of the
diaphragmis the real part of K
d
, and its loss factor is the ratio of the imaginary part to the real part of K
d
, which will be calculated for
verication of the diaphragm model in Section 4.1.
2.3. Air ow restrictor
The air owrestrictor of PVIS in this research is a capillary tube placed between the spring chamber and the damping chamber
as shown in Fig. 3.
Fig. 3. Schematic of air ow restrictor.
We propose a simple model for the air owrestrictor described in Eq. (9), which improves upon the previous models as will be
shown by comparing with experimental data (the proof will be provided in Section 4.2 since it uses a newanalysis technique to be
explained in Section 3).
P
c
= L
a
2
ju
a
ju
a
9
where L is the loss coefcient,
a
is the air density and u
a
is the air velocity.
For comparison, a linear model and an nonlinear model, which were developed in previous research that adopted a capillary
tube as an air ow restrictor [12,13], are summarized in Appendix A.
2.4. Constitutive equations describing interactions between components
To describe the complete PVIS system, the constitutive equations expressing the relationships between the models are
required.
The average pressure P
0
can be obtained fromthe static equilibrium condition for the tabletop mass (i.e., the gross mass on the
PVIS including the piston and all mass attached to it such as the platform and payload):
P
0
= P
at
+
Mg
A
p
10
where P
at
is the ambient atmospheric pressure, M is the tabletop mass, g is the acceleration of gravity, and A
p
is the effective
bottom area of the piston. By using the geometric compatibility between the piston movement and the spring chamber volume as
shown in Fig. 1, the volume variation of the spring chamber V
S
is described as
V
S
= A
p
x
p
x
b
_ _
11
where x
p
and x
b
are the displacements of the piston and the base, respectively. The volumetric rate of air ow from the damping
chamber is

V
D
= A
c
u
a
12
where A
c
is the cross-sectional area of capillary tube or d
c
2
/4. Variation of the supporting force to the piston f
S
is caused by
pressure variation in the spring chamber:
f
S
= A
p
P
S
13
External force variations on the tabletop mass include force variation caused by the variation of air pressure in the spring
chamber and force variation caused by diaphragm deformation:
Mx
::
p
= f
d
+ f
S
14
where the dead weight effect is removed naturally since only variations fromequilibriumare considered. Pressure drop due to the
ow restrictor is equal to the pressure difference between the two air chambers:
P
c
= P
S
P
D
15
3. Transfer functions and equivalent mechanical system
In this research, transmissibility from base vibration to tabletop motion is used to evaluate the vibration isolation performance
of the PVIS. To obtain transmissibility analytically, transfer function should be developed in advance.
3.1. Variable conversion and linearization that keeps original properties
For consistent physically-meaningful notation, variables and coefcients in the models, which have been derived from various
mechanics such as dynamics, thermodynamics, uid mechanics, and viscoelasticity, are converted into equivalent forms in
dynamics. The conversion facilitates construction of the equivalent mechanical system that will be introduced in Section 3.3.
Equivalent displacement of volume variation in the damping chamber x
D
is dened as
x
D
= V
D
= A
p
16
As for the air ow restrictor, the equivalent damping forces caused by the capillary tube is dened as
f
c
= A
p
P
c
17
By substituting Eqs. (12) and (16) into Eq. (9) and inserting the result into Eq. (17), we obtain
f
c
=
a
LA
3
p
2A
2
c
x
D
j j x
D
18
Quadratic damping is shown in Eq. (18). In order to insert the model into transfer function, we should apply the Fourier
transform, which processes only a linear function with a single term.
Thus linearization that keeps original nonlinear property is required. Therefore, the equal energy dissipation method is adopted
[29,30]. The basic concept of the method is as follows: the energy dissipated by quadratic damping during a cyclic motion is equal
to that dissipated by the equivalent linear damping coefcient C
c
as described in Eq. (19):
C
c
x
D
dx
D
=
a
LA
3
p
2A
2
c
j x
D
j x
D
dx
D
19
By letting x
D
=

x
D
sint, we obtain
f
c
=C
c
;

x
D
_ _
x
D
20
with
C
c
;

x
D
_ _
=
4
a
LA
3
p
3A
2
c
x
D
21
Even though Eqs. (20) and (21) constitute a linear form appropriate to be Fourier-transformed, they virtually conceive
quadratic damping since the equivalent damping coefcient has the product of frequency and amplitude.
As for the air chambers, the models described by Eqs. (1) and (2) are linearized by Taylor series expansion only up to the rst
order term as follows:
P
D
=P
0
V
D
V
D
22
P
S
=P
0
V
S
V
D
V
S
23
Since the ratio of volume variation to original volume or V = V is less than 0.01 for a typical PVIS, the difference between
nonlinear Eq. (1) and corresponding linearized Eq. (22) is less than 1.22%. Thus, the linearized Eqs. (22) and (23) can be
substituted for Eqs. (1) and (2) with negligible error.
By multiplying both sides of Eqs. (22) and (23) by A
p
and substituting Eqs. (11) and (16) into the result, the equivalent elastic
forces caused by the damping and spring chambers can be expressed as
f
D
=K
D
x
D
24
f
S
=K
S
x
p
x
b
x
D
_ _
25
where the equivalent stiffnesses are given by
K
D
=
P
0
A
2
p
V
D
26
K
S
=
P
0
A
2
p
V
S
27
3.2. Transfer functions and computation
The following are the transfer functions derived in Appendix B:
X
p
X
b
=
1
1 +

2
M
K
d
;X
b
X
p
= X
b
1 j j +
1
1
K
S
+
1
K
D
+ iC
c
; X
D
j j
28
X
D
=
K
S
X
b
X
p
= X
b
1
_ _
K
S
+ K
D
+ iC
c
; X
D
j j
29
We see that Eqs. (28) and (29) with the coefcient equations of Eqs. (7), (8), (21), (26) and (27) constitute the PVIS transfer
functions (Eqs. (28) and (29) are the same as Eqs. (B.13) and (B.14), respectively. Eqs. (B.10), (B.3) and (B.13) are the exact forms
of Eqs. (7), (8) and (21) for calculation, respectively). The term the PVIS transfer functions will be used to reference these
equations in subsequent analyses.
In previous studies, the computations of transfer functions of the PVIS needed no iteration because the transfer functions were
linear [1214]. Since our transfer functions are a complex nonlinear system that cannot be solved with one-step calculation, a
computational algorithm, which is composed of the xed-point iteration and under-relaxation, is newly devised and applied. The
details are described in Appendix C.
3.3. Equivalent mechanical system
Among the PVIS transfer functions, Eq. (28) is the major equation that renders transmissibility. Terms in Eq. (28) can be
classied by their physical meaning as follows:
H =
1
1 +
I
i
I
t
=
I
t
I
i
+ I
t
30
where
H = X
p
= X
b
31
I
i
=
2
M 32
33
Here, H, I
i
and I
t
denote the transmissibility, inertial mechanical impedance and transmitting mechanical impedance, respectively
(mechanical impedance, also referred to as dynamic modulus in Ref. [28], represents force per unit displacement [29]).
The corresponding equivalent mechanical systemis depicted in Fig. 4, where mechanical impedances are connected in serial or
in parallel according to positions of corresponding terms in Eqs. (30)(33). This equivalent mechanical system claries the
physical meanings of components and their effects on the vibration isolation performance of the PVIS, and will be used to verify
the model of the air ow restrictor in Section 4.2.
4. Experimental verication of the models and PVIS transfer functions
4.1. Experimental verication of the diaphragm model by stiffness comparison
The curve evaluated using the diaphragmmodel of Eqs. (7) and (8) is tted for the experimental data of Ref. [13] by tuning the
coefcients in the model. The experimental data and tted curves for displacements of 0.05 mm, 0.15 mm and 0.5 mm are shown
in Fig. 5. The tuned coefcients are as follows:

K
d
= 1:0 10
3
N m
1
,

K
d
= 2:5 10
4
N m
1
,
d
=6.310
4
rad s
1
, =0.15,
c
E0
=5.010
3
m, c
E1
=4.0, and c
E0
=1.010
3
m.
The closeness of the curve and the experimental data indicates the validity of the model. Since the curve tting is performed
mainly with the real part and complementarily with the loss factor, the loss factor differs from the experimental data. Because the
difference between the loss factor from the experimental data and that of the model is about 50% as seen in Fig. 5, and the ratio of
the damping force from the diaphragm to that of the entire PVIS in experimental data, i.e. (K
dp
dp
)/ (K
T
T
) is about 10% near the
natural frequency, the overall damping force of the PVIS differs from the experimental results by about 5% (=50%10%).
Accordingly, the inuence of this difference on the overall damping performance of the PVIS is so small.
4.2. Experimental verication of the air ow restrictor model by mechanical impedance comparison
The schematic of the experimental setup of Ref. [13] is the same as Fig. 4 except that their system contains no payload, and the
displacement of the base is xed or X
b
=0. Thus, the total stiffness of the dual-chamber pneumatic spring in Ref. [13] is the same as
10
1
10
0
10
1
0
2000
4000
6000
8000
10000
12000
14000
16000
Frequency [Hz]
10
1
10
0
10
1
Frequency [Hz]
R
e
a
l

p
a
r
t

o
f

K
d

[
N
/
m
]

0.05 mm (experiment)
0.5 mm (experiment)
0.05 mm (model)
0.15 mm (model)
0.5 mm (model)
(a)
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
L
o
s
s

F
a
c
t
o
r

o
f

K
d
0.5 mm (experiment)
0.05 mm (model)
0.15 mm (model)
0.5 mm (model)
(b)
Fig. 5. Experimental data and tted curve for mechanical impedance of diaphragm K
d
: (a) real part of K
d
, (b) loss factor of K
d
.
Fig. 4. Equivalent mechanical system of PVIS with two air chambers.
the transmitting mechanical impedance of our PVIS model described in Eq. (33) in terms of physical meaning, because both imply
the ratio of supporting force variation to base displacement variation.
By subtracting K
d
from both sides of Eq. (33) and replacing the transmitting mechanical impedance I
t
and the complex
diaphragm stiffness K
d
with K
T
(1+i
T
) and K
dp
(1+i
dp
), respectively, we obtain
K
T
1 + i
T
K
dp
1 + i
dp
_ _
=
1
1
K
S
+
1
K
D
+ iC
c
; X
D
j j
34
where K
T
and K
dp
are real parts of mechanical impedances of the total PVIS and of the diaphragmstiffness, respectively, and
T
and
dp
are loss factors of the total PVIS and of the diaphragm, respectively.
For comparison between experiment and model, we divided both sides of Eq. (34) into Eqs. (35) and (36):
I
a
=
1
1
K
S
+
1
K
D
+ iC
c
; X
D
j j
35
I
a;e
= K
T
1 + i
T
K
dp
1 + i
dp
_ _
36
where I
a
denotes mechanical impedance caused by two air chambers and an air owrestrictor, and I
a, e
denotes the corresponding
experimental mechanical impedance. By setting X
b
=0 in Eq. (29) for the calculation of C
c
,
X
D
=
K
S
X
p
K
S
+ K
D
+ iC
c
; X
D
j j
37
Since Eq. (37) is a recursive nonlinear equation, X
D
is evaluated by the xed-point iteration method with under-relaxation,
mentioned in Section 3.
Among the coefcients of Eq. (35), K
S
and K
D
, which are determined by Eqs. (26) and (27), respectively, are well-dened and
have been accepted in numerous prior studies of air chambers [5,6,12,13,18]. Thus, Eq. (35) together with Eq. (37) can be used for
the verication of only one coefcient, C
c
. Therefore, the models are accessed by graphical comparison between Eqs. (35) and (37)
(determined by the air ow restrictor model) and Eq. (36) (determined by experimental data).
For comparison between models, two air ow restrictor models developed in previous studies is applied to the equivalent
linear damping coefcient of air ow restrictor C
c
to be inserted into the PVIS transfer functions. By the equal energy dissipation
method and variable conversion explained in Section 3, Eqs. (A.1) and (A.2) become corresponding equivalent linear damping
coefcients, respectively, as follows:
C
c;1
=
128
a
h
c
A
2
p
d
4
c
38
C
c;2
=
h
c
d
c
fr + L
_ _
4
a
A
3
p
3A
2
c
X
D
j j 39
where the friction coefcient fr is calculated using Eqs. (A.3)(A.5) with the constitutive Eq. (12) and the dimension conversion
formula of Eq. (16).
In Figs. 68, the shape of the mechanical impedance of Erin and Wilson's model of Eq. (38) is very different from experimental
results and is independent of the base vibration amplitude because the model is linear.
In Fig. 6, Lee and Kim's model of Eq. (39) with the loss coefcient L=0.7, and our model of Eq. (B.3), which is the exact formof
Eq. (21) for calculation, with L=1.2, coincide very well with experimental results. As the base vibration amplitude decreases, the
plot based on Lee and Kim's model shows discrepancy with experimental results as seen in Fig. 8, which is the case when air ow
rate through the capillary tube is slow (Reb2300). The discrepancy is thought to originate from following two deciencies.
The rst deciency is that they assumed that the amplitude of air ow passing through the capillary tube

u
a
is proportional to
the amplitude of tabletop displacement X
p
. Although this assumption made the transfer function of the PVIS simple enough to
need only a single step calculation,

u
a
is found not to be proportional to X
p
either in magnitude and phase when

u
a
is taken as a
variable independent of X
p
, as in our analysis.
The second deciency is that they applied uid mechanics for internal owin a fully developed region to a short capillary tube.
Entrance length, where the fully developed region starts, ranges from18d
c
(for turbulent ow) to 138d
c
(for laminar ow) [31,32].
However, the length of capillary tube h
c
is about 13.3d
c
in Lee and Kim's specication, so fully developed ow does not occur for
almost all ow rates. Therefore, we improved the capillary tube model by removing the pressure drop by internal ow term from
Lee and Kim's model of Eq. (A.2). The criterion between laminar owand turbulent owof Eq. (A.3) is not necessary for our model
due to the removal of the term.
The fact that our model shows good agreement with experimental results validates our inferences. The loss coefcient for the
air owrestrictor L is tuned to 1.2 for the best t. The value is quite reasonable because the value is in the interval of the sumof loss
coefcients of entrance and exit: the capillary tube has both an entrance and exit, and the loss coefcient ranges from 0 to 1 for
various shapes of entrance, and is about 1 for exit regardless of shape [31].
Dynamic characteristics of the mechanical impedance of the air chambers and the capillary tube in Figs. 68 can be explained
using Eq. (35). The mechanical impedance of air chambers and capillary tube I
a
changes from 1/ (1/ K
S
+1/ K
D
) to K
S
as C
c
increases, and C
c
increases quadratically as frequency or amplitude of vibration increase according to Eq. (21). This corresponds to
the fact that the mechanical impedance changes approximately from 1= 1= K
S
+ 1= K
D
= 8:39 10
3
N m
1
to
K
S
=2.3910
4
N m
1
as frequency or amplitude of the base vibration increases, both in the analytic results and in the
experimental results. This implies that the effect of the spring chamber is dominant due to high damping by the capillary tube at
high frequency or large amplitude, and that the two air chambers act as one due to low damping at low frequency or small
amplitude.
4.3. Experimental verication of the PVIS transfer functions by transmissibility comparison
Experimental transmissibility data that were obtained from the research of Ref. [12], are used to verify the PVIS transfer
functions by transmissibility comparison. The following are the specications and conditions of Ref. [12]: V
S
=7.3210
5
m
3
,
V
D
=4.1810
4
m
3
, A
p
=1.8510
3
m
2
, M=110 kg,
a
=1.82410
5
Pa s, h
c
=7.2710
3
m, and d
c
=6.1010
4
m
2
.
There are seven coefcients to be determined in the diaphragm model of Eqs. (7) and (8). Since
d
, c
E0
, c
E1
, c
E2
, and are
material constants, and is the angular velocity of movement or

K
S
+ K
d
= M
_
, only

K and

K
d
are tuned to t the model using
the experimental data. Thus, we obtain

K = 4:4 10
3
N m
1
and

K
d
= 12:5 10
3
N m
1
.
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
x 10
4
R
e
a
l

P
a
r
t

o
f

I
a

[
N
/
m
]
(a)
0
0.1
0.2
0.3
0.4
0.5
L
o
s
s

F
a
c
t
o
r

o
f

I
a
(b)
10
1
10
0
10
1
Frequency [Hz]
10
1
10
0
10
1
Frequency [Hz]
Fig. 6. Comparison of mechanical impedances caused by air chambers and capillary tube I
a
at X
b
=0.5 mm: (a) real part of I
a
, (b) loss factor of I
a
: +, experiment;
, Erin and Wilson; , Lee and Kim (L=0.7); , Moon and Lee (L=1.2).
With all the model coefcients identied, transmissibility is calculated by the PVIS transfer functions at X
b
=210
6
m and is
compared graphically in Fig. 9 with the experimental results. The good agreement validates the PVIS transfer functions and
calculation algorithms. The second resonance in Fig. 9 can be ignored in our analysis since it is from a rocking mode caused by
multiple isolator legs [12].
5. Sensitivity analysis of vibration isolation performance
Our aim in using sensitivity analysis is to know how the major design variables affect the major performance indices of the
PVIS, and to extract simple approximation formulas for design. The fundamental tasks required in a sensitivity analysis are to
identify a model of the object, and to screen out major inputs (or design variables) and outputs (or performance indices) of the
model [33,34].
The selected design variables are the spring chamber volume V
S
, tabletop mass M, effective piston area A
p
, volume ratio of air
chambers N, and capillary tube cross-sectional area A
c
because these values have a strong effect on the vibration isolation
performance of a PVIS as will be shown in this section. The selected performance indices are the frequency and magnitude of the
resonance peak since a PVIS uses the attenuation of displacement which occurs at a frequency range higher than the resonant
frequency, and resonant peaks should be suppressed if the frequency of base vibration is close to resonant frequency.
The specications of PVIS for the transmissibility calculation are the same as in the experiment of Ref. [13] as follows:
V
S
=8.110
4
m
3
, V
D
=1.510
3
m
3
, A
p
=5.310
3
m
2
, M=212 kg (calculated by Eq. (10) with P
0
=4.9310
5
Pa),
P
0
=4.9310
5
Pa,
a
=5.97 kg m
3
,
a
=1.7910
5
Pa s, h
c
=1.210
2
m
2
and d
c
=9.010
4
m. The tuned coefcients
0
0.1
0.2
0.3
0.4
0.5
L
o
s
s

F
a
c
t
o
r

o
f

I
a
(b)
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
x 10
4
R
e
a
l

P
a
r
t

o
f

I
a

[
N
/
m
]
(a)
10
1
10
0
10
1
Frequency [Hz]
10
1
10
0
10
1
Frequency [Hz]
a
at X
b
a
a
: +, experiment;
, Erin and Wilson; , Lee and Kim (L=0.7);, Moon and Lee (L=1.2).
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
x 10
4
R
e
a
l

P
a
r
t

o
f

I
a

[
N
/
m
]
(a)
0
0.1
0.2
0.3
0.4
0.5
L
o
s
s

F
a
c
t
o
r

o
f

I
a
(b)
10
1
10
0
10
1
Frequency [Hz]
10
1
10
0
10
1
Frequency [Hz]
a
at X
b
a
a
: +, experiment;
, Erin and Wilson; , Lee and Kim (L=0.7);, Moon and Lee (L=1.2).
Frequency [Hz]
T
r
a
n
s
m
i
s
s
i
b
i
l
i
t
y
10
1
10
0
10
0
10
1
10
1
Fig. 9. Comparison between transmissibility of experimental data and transmissibility evaluated by the PVIS transfer functions: +, experiment; , model.
were determined in Section 4 as follows: L=1.2,

K
d
= 1:0 10
3
N m
1
,

K
d
= 2:5 10
4
N m
1
,
d
=6.310
4
rad s
1
,
=0.15, c
E0
=5.010
3
m, c
E1
=4.0, and c
E0
=1.010
3
m. If specications and tuned coefcients are not mentioned in the
remainder of this paper except for a certain value of design variable, the remaining design variable values are the same as those
just cited. The frequency range is selected as 0.1 Hz to 100 Hz, and the amplitude range is selected from 0.1 m to 1 mm, because
these values represent the general condition of base vibration [4,12].
X
b
=1 mm
X
b
=0.1 mm
X
b
=0.01 mm
(a)
Amplitude [m]
Frequency [Hz]
T
r
a
n
s
m
i
s
s
i
b
i
l
i
t
y
(b)
T
r
a
n
s
m
i
s
s
i
b
i
l
i
t
y
10
1
10
1
10
0
10
0
10
1
10
2
10
-10
10
-5
10
0
10
1
10
2
10
0
10
-2
10
-4
Frequency [Hz]
10
0
10
1
10
-7
10
-1
10
0
10
1
10
2
10
-6
10
-5
10
-4
10
-3
Frequency [Hz]
A
m
p
l
i
t
u
d
e

[
m
]
(c)
A
B
C
Fig. 10. Transmissibility of the PVIS: (a) transmissibility curve, (b) transmissibility surface, (c) contour view: A, one-chamber region; B, two-chamber region;
C, minimum resonant peak.
5.1. Dynamic characteristics of pneumatic vibration isolation system
Transmissibility of the PVIS is shown in Fig. 10(b) in the form of a surface in 3-dimensional space. Fig. 10(c) is a contour view,
which is viewed from the transmissibility axis. As shown in Fig. 10, the transmissibility depends on amplitude of base vibration as
well as frequency, whereas the transmissibility of a linear second-order systemdepends only on frequency, and not on amplitude.
The nonlinear dynamic characteristic originates mainly from the damping force or restriction force by the air ow restrictor as
veried in Section 4.2.
Consequently, at high frequencies or large amplitudes of vibration in the region labeled A in Fig. 10(c), the spring chamber
works alone as a mechanical spring, and the damping chamber nearly does not work, because the ow restriction force is strong
enough to signicantly block air ow between air chambers. Thus, for the remainder of this paper, we call this the one-chamber
region. The corresponding equivalent mechanical system is illustrated in Fig. 11.
At low frequencies or small amplitudes of vibration in the region labeled B in Fig. 10(c), the two air chambers act as a single
chamber, since the owrestriction force is weak enough to freely allowthe air owacross air chambers. Thus, for the remainder of
this paper, we call this the two-chamber region. The corresponding system is illustrated in Fig. 12.
5.2. Sensitivity of resonant frequency to air chamber volume, tabletop mass, and piston area
Eqs. (26) and (27) can be consolidated to the effective stiffness of air chambers K
a
as follows:
K
a
=
P
0
A
2
p
V
a
40
where the effective volume of air chambers V
a
is assigned to V
S
for the one-chamber region and to V
S
+V
D
for the two-chamber
region. By inspection of the equivalent mechanical systems shown in Figs. 11 and 12, the natural frequency of the PVIS with the
effective air volume becomes
f
n
=
1
2
K
d
+ K
a
M
_
41
M
Payload
Base Excitation
Tabletop Motion
Spring
Chamber
Damping
Chamber
Diaphragm
K
d
K
S
K
D
X
p
X
b
Fig. 12. Equivalent mechanical system for the two-chamber region.
Fig. 11. Equivalent mechanical system for the one-chamber region.
Thus the resonant frequency of the PVIS can be derived by substituting Eqs. (10) and (40) into Eq. (41) as follows:
f
n
=
1
2
K
d
M
+
A
2
p
P
at
MV
a
+
A
p
g
V
a
42
The approximation formula of Eq. (42) cannot be used for the transition region between the one-chamber region and the two-
chamber region because the effective air chamber volume V
a
cannot be dened in the transition region. As seen in this
approximation formula, it was found that the PVIS resonant frequency depends on effective air chamber volume V
a
, tabletop mass
M, and piston area A
p
. Among these, M and A
p
are dependent on each other as described by Eq. (10). Considering the maximum
gauge pressure is about 5.5 bar (80 psi), which is general practice in industry, we conne the ranges of V
a
, M and A
p
in this
subsection so that P
0
in Eq. (10) may be 2 through 6.5 bar in absolute pressure.
The solid line in Fig. 13 represents results calculated using the PVIS transfer functions by following procedure: 1) search a
resonant peak of PVIS transmissibility curve for a given amplitude of base vibration by the golden section search method [35]; 2)
store the frequency at the found peak; 3) repeat the procedure by varying chamber volume; and 4) draw the graph of resonant
frequency with respect to chamber volume. The solid lines in Figs. 14 and 15 can be obtained by applying the same procedure to
the tabletop mass and piston area, respectively.
The dashed lines in Figs. 1315 are evaluated by the approximate formula of Eq. (42). Since the resonant frequency for a base
vibration amplitude of X
b
=1 m is in the two-chamber region, as seen in Fig. 10, the approximate resonant frequency should be
calculated using Eq. (42) with V
a
=V
S
+V
D
and K
d
=

K
d
+

K
d
_ _
= 2. Similarly, since the resonant frequency at X
b
=1 mmis in the
1 2 3 4 5 6 7 8 9
x 10
3
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
Air chamber volume, V
S
+V
D
[m
3
]
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(a)
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 10
3
1
1.5
2
2.5
3
3.5
4
Air chamber volume, V
S
[m
3
]
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(b)
Fig. 13. Resonant frequency with respect to air chamber volume: (a) X
b
=1 m, (b) X
b
=1 mm: , calculated by the PVIS transfer functions; , calculated by
approximation formula of Eq. (42).
one-chamber region, the approximate resonant frequency should be calculated with V
a
=V
S
and K
d
=

K. The volume ratio was
kept constant both in analytic and approximate solutions since the ratio between air chambers affects the shape of the
transmissibility surface.
By inspection of Figs. 13 and 14, the resonant frequency decreases and approaches a horizontal asymptote as air chamber
volume or payload mass increases. Such phenomena are also found in commercial air springs [36]. FromFig. 15, it is found that the
resonant frequency increases as the piston area increases. The approximation Eq. (42) explains the reasons as follows: resonant
frequency f
n
approaches

K
d
= M
_
= 2 as the air chamber volume V
a
goes to innity; resonant frequency f
n
approaches
A
p
g = V
a
_
= 2 as the tabletop mass M goes to innity; resonant frequency f
n
increases as the effective piston area A
p
increases.
Since the supplied pneumatic pressure generally has an upper bound in factories, there is a limitation on the enhancement of
PVIS attenuation performance resulting from the lowering of resonant frequency by increasing tabletop mass or reducing piston
area, as described by Eq. (10). Moreover, the diaphragm stiffness determines the lower bound of the PVIS resonant frequency
K
d
= M
_
= 2 that can be decreased by no more than the increase in air chamber volume. Therefore, making the diaphragm as
exible as possible is an important factor with which to lower PVIS resonant frequency.
5.3. Sensitivity of resonant peak magnitude to volume ratio of air chambers
In Fig. 16, the magnitude of the resonant peak with respect to the volume ratio N is depicted at X
b
=1 m, which causes the
transmissibility curve to lie in the two-chamber region. Since the resonant peak in the one-chamber region is maintained, only the
magnitude in the two-chamber region is shown.
As shown in Fig. 16 by the solid line calculated using the PVIS transfer functions, the resonant peak magnitude decreased as the
volume ratio increased. However, when N is large, the curve levels off.
50 100 150 200 250 300
1
1.5
2
2.5
3
3.5
Tabletop mass, M [kg]
50 100 150 200 250 300
Tabletop mass, M [kg]
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(a)
1.6
1.8
2
2.2
2.4
2.6
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(b)
Fig. 14. Resonant frequency with respect to tabletop mass: (a) X
b
=1 m, (b) X
b
1 2 3 4 5 6 7 8 9 10
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
9
Volume ratio, V
D
/V
S
R
e
s
o
n
a
n
t

p
e
a
k

m
a
g
n
i
t
u
d
e
,

|
X
p
/
X
b
|
Fig. 16. Resonant peak magnitude with respect to volume ratio of air chambers at X
b
=1 m: , calculated by the PVIS transfer functions; , calculated by
approximation formulas of Eqs. (43) and (48).
0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
Piston area, A
p
[m
2
]
0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Piston area, A
p
[m
2
]
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(a)
1.5
2
2.5
3
3.5
4
4.5
R
e
s
o
n
a
n
t

f
r
e
q
u
e
n
c
y
,

f
n

[
H
z
]
(b)
Fig. 15. Resonant frequency with respect to piston area: (a) X
b
=1 m, (b) X
b
To develop an approximation formula, we exploit damping characteristic of the second order mechanical system. The resonant
peak magnitude can be expressed with the loss factor as follows [37]:
H
peak
=
1 +
2
_

43
Meanwhile, the magnitude of damping force from the air ow restrictor is proportional to the square of the equivalent
displacement of the volume of air oscillating through air ow restrictor as described in Eqs. (B.2) and B.3). Thus,
F
D
j j X
D
j j
2
44
Because, at resonant frequency, air ows through the ow restrictor at a high rate, it is assumed that the air in the spring
chamber and the air in the damping chamber is compressed by the same ratio at resonant peak. Then, the volume variations in
both air chambers can be expressed as
1 + N : NV
S
: V
D
45
By substituting Eqs. (11) and (16) into Eq. (45), we obtain
X
D
N
N + 1
X
p
X
b
1
_ _
X
b
46
By using Eq. (46) and the fact that transmissibility X
p
/ X
b
is proportional to X
b
at resonant peak and that X
b
is an input variable,
Eq. (44) becomes
F
D
j j
N
N + 1
_ _
2
47
Since the loss factor is related to systemdamping by the air owrestrictor and the diaphragm, the following can be extracted
from Eq. (47) and from the fact that the damping force of the diaphragm F
d
is independent of hN.
= b
1
N
N + 1
_ _
2
+ b
2
48
where the rst termis related to the damping force caused by the air owrestrictor, and the second termis related to the damping
force caused by the diaphragm. By curve-tting Eqs. (43) and (48) to the resonant magnitude calculated by the PVIS transfer
functions, b
1
=0.115 and b
2
=0.03 are obtained, and the tted curve is displayed in Fig. 16 by a dashed line. Since coefcients b
1
and b
2
are also dependent on the other design variables such as air chamber volume, piston area, tabletop mass, and capillary tube
area, the usage of approximation formulas of Eqs. (43) and (48) is limited only to predict the tendency of peak magnitude variation
with respect to volume ratio variation.
5.4. Sensitivity of vibration amplitude at minimum resonant peak to air ow restrictor area
As seen in the transmissibility surface of the PVIS, it is found that the resonant peak reached its minimum, whose position is
labeled C in Fig. 10(c), at a certain value of the base vibration amplitude. In Fig. 17, the solid line in the graph represents the results
calculated by the PVIS transfer functions using the following procedure: 1) search for the minimumpoint among resonant peaks of
the transmissibility surface of the PVIS for a given capillary tube area by the golden section search method; 2) store the base
vibration amplitude at the found minimum peak; 3) repeat the procedure by varying the area; and 4) draw the graph of base
vibration amplitude of minimum resonant peak with respect to the area.
The plot of the base vibration amplitude minimizing the resonant peak magnitude X
b, mp
is approximately linear with respect to
the capillary tube area A
c
in the loglog graph shown in Fig. 17, so it can be expressed by
X
b;mp
=
A
c
b
3
_ _
2:3
49
where coefcient b
3
remains constant as A
c
changes while the other design specications are held constant. The curve evaluated by
Eq. (49), which is tted to the exact value evaluated by the PVIS transfer functions, is shown by the dashed line in Fig. 17, and the
tuned coefcient value is b
3
=2.610
5
m
2
.
A useful approximation formula can be derived from Eq. (49) as follows:
A
c;2
= A
c;1
X
b;mp;2
X
b;mp;1
_ _
0:435
50
This approximation formula is supposed to be very helpful in suppressing the amplication of vibration in following case: the base
vibration amplitude X
b, mp, 1
and the capillary tube area A
c, 1
are known only for a small range of the frequency and amplitude of base
vibration because of the limitation of actuation and/or measurement instruments. The vibration amplitude of the base X
b, mp, 2
where
the object of vibration isolation stands is known. In addition, the frequency of the base vibration is close to the resonant frequency.
With this known information, the capillary tube area A
c, 2
can be determined in order to minimize resonant peak magnitude by using
Eq. (50). This is equivalent to placing the resonant peak on point C in Fig. 10(c).
6. Conclusions
The nonlinear properties of the diaphragm of a PVIS have been described using look-up tables in previous studies since it is
very difcult to model the stiffness and damping, which depend on the amplitude and frequency of deformation. In this paper,
a fractional derivative model for the diaphragm was newly developed and was shown to agree very well with experimental
data.
The capillary tube in the PVIS, referred to as an air ow restrictor, has previously been regarded as a linear damper or as a
nonlinear damper with its air ow volume being proportional to the tabletop displacement. We have described the limitations of
these assumptions of air ow restrictors and improved the model for the air ow restrictor. In our model, the damping force is
proportional to the square of the ow rate and the air volume through the capillary tube is dened as an independent variable.
Based on comparisons with experimental data, the new air ow restrictor model has proved to be more accurate than previous
models.
The PVIS transfer functions were derived from our new models by the equal energy dissipation method and the conversion to
equivalent dynamic variables. To calculate the obtained transfer functions that cannot be calculated in one step due to the
nonlinearity, xed-point iteration with under-relaxation was applied. The computational algorithm was proven to be valid by
comparison of its results with experimental ndings. The algorithm enabled sensitivity analysis as well as transmissibility
calculations.
The equivalent mechanical system, which was newly proposed in the formulation of the transfer functions, claried the
physical meanings of components of the PVIS and was useful in adjusting the air owrestrictor model. The dynamic characteristic
of the PVIS, namely, that the transmissibility surface should be divided into two characteristic regions according to the ranges of
the frequency and amplitude, could be explained using the equivalent mechanical system.
For the two characteristic regions, the sensitivities of vibration isolation performance indices to design variables were analyzed
using the PVIS transfer functions and the golden-section search method. In order to facilitate efcient design, approximation
formulas were created based on the dynamic characteristics of the PVIS. The results agreed very well with the results of sensitivity
analyses through the tuned coefcients.
10
-6
10
-7
10
-6
10
-5
10
-4
10
-3
Capillary tube area, A
c
[m
2
]
B
a
s
e

a
m
p
l
i
t
u
d
e

a
t

m
i
n
i
m
u
m

p
e
a
k
,

X
b
,
m
p

[
m
]
Fig. 17. Base vibration amplitude at minimum peak magnitude with respect to capillary tube area: , calculated by the PVIS transfer functions; , calculated
by approximation formula of Eq. (49).
Acknowledgement
We wish to express our gratitude for the support of the Ministry of Education, Science and Technology (BK21) and the Korea
Science and Engineering Foundation (ERC: Micro Thermal SystemResearch Center). We are also grateful to Prof. Heui Jae Pahk, Dr.
Byung-Hoon Lee and Dr. Jeung-Hoon Lee for their helpful comments.
Appendix A. Previous models of air ow restrictors
The Erin and Wilson's model [12] of the air ow restrictor is
P
c
=
32
a
h
c
d
2
c
u
a
A:1
where P
c
is the pressure drop across the capillary tube,
a
is the viscosity of the air, h
c
is the height of the capillary tube, d
c
is the
diameter of the capillary tube, and u
a
is the velocity of air.Subsequently, Lee and Kim [13,22] adopted a nonlinear model for the
capillary tube as follows:
P
c
=
h
c
d
c
fr + L
_ _
a
2
u
a
j ju
a
A:2
where L is the loss coefcient and
a
is the air density. The friction coefcient fr is a function of Reynolds number Re for laminar
ow and turbulent ow as follows:
fr =
64
Re
for Re 2300 laminar flow
0:3164
Re
1=4
for 4000bRe10
5
turbulent flow
_
_
A:3
with
Re =

a
u
a
d
c
a
A:4
where the cycle-averaged air velocity u
a
is given by
u
a
=

2
2 =
0
u
a
j j t
=
2
u
a
A:5
with u
a
=

u
a
sint.
Appendix B. Derivation of PVIS transfer functions
The pressure relation in Eq. (15) can be converted into the force relation in Eq. (B.1) by multiplying both sides by A
p
:
f
c
= f
S
f
D
B:1
To obtain the transfer function of the PVIS, Fourier transforms are performed on the models and constitutive equations. Fourier
transforms of variables and functions are signied by capital letters as follows: xX,

xj Xj , xiX, x
::
2
X, and f F.
Accordingly, Eqs. (20), (21), (24), (25), (14), and (B.1) are Fourier-transformed into Eqs. (B.6) through (B.7), respectively.
F
c
=iC
c
; X
D
j j X
D
B:2
C
c
; X
D
j j =
4
a
LA
3
p
3A
2
c
X
D
j j B:3
F
D
= K
D
X
D
B:4
F
S
=K
S
X
p
X
b
X
D
_ _
B:5
2
MX
p
= F
d
+ F
S
B:6
F
S
= F
D
+ F
c
B:7
The diaphragmmodel described in Eqs. (6), (7) and (8) should be adjusted into Eqs. (B.8), (B.9) and (B.10), because the relative
displacement of the piston to the base X
p
X
b
is the effective rolling displacement, as shown in Fig. 1:
F
d
=K
d
; X
p
X
b
_ _
X
p
X
b
_ _
B:8
K
d
; X
p
X
b
_ _
=
K
d
+

K
d
i =
d

E
X
p
X
b
_ _ _ _
1 + i =
d

E
X
p
X
b
_ _ _ _
B:9
log
10
E
X
p
X
b
_ _
=
c
E1
X
p
X
b
c
E0
_ _
c
E2
+ X
p
X
b
c
E0
B:10
By substituting Eqs. (B.2) and (B.4) into Eq. (B.7) and rearranging for X
D
, we obtain
X
D
=
F
S
K
D
+ iC
c
; X
D
j j
B:11
By substituting Eqs. (B.11) into Eq. (B.5) and rearranging for F
S
,
F
S
=
1
1
K
S
+
1
K
D
+ iC
c
; X
D
j j
X
p
X
b
_ _
B:12
Finally, by inserting Eqs. (B.8) and (B.12) into Eq. (B.6) and rearranging, we obtain the transmissibility from base vibration to
tabletop motion or X
p
/ X
b
as follows:
X
p
X
b
=
1
1 +

2
M
K
d
;X
b
X
p
= X
b
1 j j +
1
1
K
S
+
1
K
D
+ iC
c
; X
D
j j
B:13
The equivalent displacement of the air volume passing through the air ow restrictor X
D
is obtained by substituting Eqs. (B.2),
(B.4) and (B.5) into Eq. (B.7) as follows:
X
D
=
K
S
X
b
X
p
= X
b
1
_ _
K
S
+ K
D
+ iC
c
; X
D
j j
B:14
Appendix C. Computational algorithm for the PVIS transfer functions
The PVIS transfer functions can be abstracted as follows.
y = Gy; y = X
p
= X
b
; X
D
_ _
C:1
where Gis the complex nonlinear mapping, and the ordered pair of y are complex numbers. We applied the xed-point iteration
for calculation of the PVIS transfer functions. When the xed-point iteration method was directly applied to the PVIS transfer
functions, the speed of contraction was very slow. Thus, the under-relaxation method is applied to hasten convergence by
damping out oscillations of y as follows [38]:
y
n + 1
= 1 y
n
+ y
n + 1
; y = X
p
= X
b
; X
D
_ _
C:2
where is the acceleration parameter and subscript n is the number of iterations. Conditions for numerical calculation of
transmissibility of the PVIS are as follows: the initial values are X
p
=0 and X
D
=0; input values are the base vibration amplitude X
b
and the base vibration frequency ; and iteration stops when the product of relative errors of variables between two adjacent
iteration steps is less than relative error criterion of 10
6
. The acceleration parameter is set to 0.75, since iteration convergence is
the fastest at this value.
References
[1] E.E. Ungar, D.H. Sturz, H. Amick, Vibration control design of high technology facilities, Sound and Vibration July (1990) 2027.
[2] C.G. Gordon, Generic criteria for vibration-sensitive equipment, Proceedings of SPIE, vol.1619, Nov 1991, pp. 7185, San Hose, CA.
[3] H. Amick, M. Gendreau, T. Busch, C. Gordon, Evolving criteria for research facilities: I vibration, Proceedings of SPIE Conference 5933: Buildings for
Nanoscale Research and Beyond, San Diego, CA, 2005, pp. 113.
[4] K.P. Heiland, Recent advancements in passive and active vibration control systems, Proceedings of SPIE, Vibration Control in Microelectronics, Optics, and
Metrology, vol.1619, Feb 1992, pp. 2233.
[5] Newport corporation homepage, http://www.newport.com (Accessed 29 July 2008).
[6] Technical manufacturing corporation homepage, http://www.techmfg.com (Accessed 29 July 2008).
[7] C.-M. Lee, V.N. Goverdovskiy, A.I. Temnikovb, Design of springs with negative stiffness to improve vehicle driver vibration isolation, Journal of Sound and
Vibration 302 (2007) 865874.
[8] K.G. Ahn, H.J. Pahk, M.Y. Jung, D.W. Cho, A hybrid-type active vibration isolation systemusing neural networks, Journal of Sound and Vibration 192 (4) (1996)
793805.
[9] D.K. Han, P.H. Chang, A robust two-time-scale control design for a pneumatic vibration isolator, Proceedings of the 46th IEEE Conference on Decision and
Control, New Orleans, LA, USA, Dec 2007, pp. 16661672.
[10] T. Kato, K. Kawashima, K. Sawamoto, T. Kagawa, Active control of a pneumatic isolation table using model following control and a pressure differentiator,
Precision Engineering 31 (2007) 269275.
[11] H. Porumamilla, A.G. Kelkar, J.M. Vogel, Modeling and verication of an innovative active pneumatic vibration isolation system, Transactions of the ASME:
Journal of Dynamic Systems, Measurement, and Control 130 (3) (2008) 112 031001.
[12] C. Erin, B. Wilson, J. Zapfe, An improved model of a pneumatic vibration isolator: theory and experiment, Journal of Sound and Vibration 218 (1) (1998)
81101.
[13] J.H. Lee, K.J. Kim, Modeling of nonlinear complex stiffness of dual-chamber pneumatic spring for precision vibration isolations, Journal of Sound and Vibration
301 (35) (2007) 909926.
[14] D.B. DeBra, Design of laminar ow restrictor for damping pneumatic vibration isolators, CIRP Annals 33 (1) (1984) 351356.
[15] G. Popov, S. Sankar, Modelling and analysis of non-linear orice type damping in vibration isolators, Journal of Sound and Vibration 183 (5) (1995) 751764.
[16] M. Heertjes, N. Wouw, Nonlinear dynamics and control of a pneumatic vibration isolator, Transactions of the ASME: Journal of Vibration and Acoustics 128
(2006) 439448.
[17] A.J. Nieto, A.L. Morales, A. Gonzalez, J.M. Chicharro, P. Pintado, An analytical model of pneumatic suspensions based on an experimental characterization,
Journal of Sound and Vibration 313 (2008) 290307.
[18] C.M. Harris, C.E. Crede, Shock and Vibration Handbook, McGraw-Hill, New York, 1961.
[19] M. Hino, M. Sawamoto, S. Takasu, Experiments on the transition to turbulence in an oscillating ow, Journal of Fluid Mechanics 75 (1976) 193207.
[20] S. Washio, T. Konishi, K. Nishii, A. Tanaka, Research on wave phenomena in hydralic lines (9th Report, Experimental Investigation of Oscillatory Orice
Flows), Bulletin of the JSME 25 (210) (1982) 19061913.
[21] T. Fujita, H. Okumura, T. Yamada, N. Inoue, S. Endoh, T. Kagawa, Affection of connecting conduit to characteristics of air spring with subtank, Journal of JSME
(part C) 63 (610) (1997) 19201926.
[22] J.H. Lee, K.J. Kim, A method of transmissibility design for dual-chamber pneumatic vibration isolator, Journal of Sound and Vibration 323 (2009) 6792.
[23] R.A. Ibrahim, Recent advances in nonlinear passive vibration isolators, Journal of Sound and Vibration 314 (2008) 371452.
[24] M. Caputo, Linear models of dissipation whose Q is almost frequency independent II, Geophysical Journal of the Royal Astronomical Society 13 (1967)
529539.
[25] T. Pritz, Analysis of four-parameter fractional derivative model of real solid materials, Journal of Sound and Vibration 195 (1996) 103115.
[26] D.I.G. Jones, Handbook of Viscoelastic Vibration Damping, John Wiley & Sons, 2001.
[27] A.D. Nashif, D.I.G. Jones, J.P. Henderson, Vibration Damping, John Wiley & Sons, 1985.
[28] J.T. Xing, Y.P. Xiong, W.G. Price, Passive-active isolation systems to produce or innite dynamic modulus: theoretical and conceptual design strategies, Journal
of Sound and Vibration 286 (2005) 615636.
[29] F.S. Tse, I.E. Morse, R.T. Hinkle, Mechanical Vibrations: Theory and Applications, second ed.Allyn and Bacon, 1978.
[30] J.H. Moon, Analysis and optimal design of a pneumatic vibration isolation system, PhD Thesis, Seoul National University, 2002.
[31] F. White, Fluid Mechanics, third ed.McGraw-Hill, 1994.
[32] T.S. Zhao, P. Cheng, Experimental studies on the onset of turbulence and frictional losses in an oscillatory turbulent pipe ow, International Journal of Heat
and Fluid Flow 17 (4) (1996) 356362.
[33] A. Saltelli, K. Chan, E.M. Scott, Sensitivity Analysis, John Wiley & Sons Ltd., England, 2000.
[34] J.S. Arora, Introduction to Optimum Design, second ed.Elsevier, 2004.
[35] D.G. Luenberger, Linear and Nonlinear Programming, second ed.Addison-Wesley Publishing Company, 1984.
[36] Firestone Airstroke Actuator/Airmount Isolator Catalog, Firestone company, USA, 1996.
[37] J.C. Snowdon, Vibration and Shock in Damped Mechanical Systems, Wiley, New York, 1968.
[38] R.L. Burden, J.D. Faires, Numerical Analysis, Sixth ed.International Thompson Publishing, 1997.
Jun-Hee Moon received the B.S., M.S. and Ph.D. degrees in the Department of Mechanical Design and Production Engineering from
Seoul National University, South Korea, in 1992, 1994, and 2002, respectively. From 2000 to 2004 he was a Research Engineerat SNU
Precision Co., Ltd. From 2005 to 2008 he was a Senior Researcher at the Micro Thermal System Research Center. Since 2009 he has
taught in the Department of Mechatronics at Daelim University College, South Korea. His research interests include the analysis and
control of precision systems using air pressure, piezoelectric materials, and mechanical linkages.
Bong-Gu Lee received the B.S., M.S. and Ph.D degrees in the Department of Mechanical Engineering from Yonsei University, South
Korea, in 2000, 2003, and 2009, respectively. He has been working as a Senior Researcher at the Korea Institute of Industrial
Technology (KITECH) from 2000 to 2002. He is currently a Professor in the Department of Mechanical Engineering at Dealim
University College. His main research interests are micro machining and material processing using lasers and ultrasonic vibration.

Modeling and Sensitivity Analysis

Încărcat de

Informații document

Descriere originală:

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Modeling and Sensitivity Analysis

Încărcat de

Drepturi de autor:

Formate disponibile

Modeling and sensitivity analysis of a pneumatic vibration isolation system

with two air chambers

S-ar putea să vă placă și