Design of Gas Bearing Systems For Precision Applications: WWW - Tue.nl/taverne

Design of gas bearing systems for precision applications
Citation for published version (APA):

Wang, J. (1993). Design of gas bearing systems for precision applications. Technische Universiteit Eindhoven.
https://doi.org/10.6100/IR391172
DOI:
10.6100/IR391172
Document status and date:

Published: 01/01/1993
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy

If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
providing details and we will investigate your claim.
Download date: 02. dic. 2021

Design of Gas Bearing Systems
for Precision Applications
Junming Wang
Design of Gas Bearing Systems
for Precision Applications
PROEFSCHRIFf
ter verkrijging van de graad van doctor aan de

Technische Universiteit Eindhoven, op gezag van
de Rector Magnificus, prof. dr. J.H. van Lint,
voor een commissie aangewezen door het College
van Dekanen in het openbaar te verdedigen op
donderdag 7 januari 1993 om 16.00 uur
door
JUNMING WANG
geboren te Acheng, Heilongjiang (China)
dtuk: wibro dissertatiedrukkerij~ helmond.

Dit proefschrift is goedgekeurd
door de promotoren
prof. dr. ir. P. H. J. Schellekeos

prof. dr. ir. M. J. W. Schouten
en de copromotor .
dr. ir. J. W. M. C. Teeuwsen
ISBN 90-386-0112-3
To Xiao-Wei and
our parents
ACKNOWLEDGEMENT
First of all, I would like to express my great thanks to Prof. dr. ir. P.
H. J. Schellekeos and Prof. dr. ir. A. C. H. van der Wolf for the detailed
financial and technical arrangements of the project and to the firm BOTECH,
Helmond for bis continuous funding and Mitutoyo Nederland R & D, for bis
financial contribution to the project. Without their supports, this project
could not have been completed.
Next, I would like to express my gratitude to the staff and students who
have, in one way or another, helped me through the course of this project,
especially to Klaas Struik, Frits Theuws, Harry Sonnenmans, and Adriaan de
Gilde, Metrology Laboratory, for their practical supports, to Jaap Snip and
Eric Driessen, master students, Jinghan Yang, AIO, for their positive
contributions to the work, and to Theo Maas, CTD, for bis great help with the
test set-ups.
I also wish to state my great appreciation to the promotors Prof. dr. ir.
P. H. J. Schellekeus and Prof. dr. ir. M. J. W. Schouten, the co-promotor dr.
ir. J. W. M. C. Teeuwsen, and the other merobers of the reading committee,
Prof. dr. ir. A. C. H. van der Wolf and Prof. dr. ir. A. A. van Steenhoven for
their fruittul discussions and critical readings of this thesis.
I am also obliged to ing. P. L. Holster, Philips Research Laboratories,

for bis helptul discussions, to Mr. Tim Bell and Mr. Nigel Moody for
correcting my English and to P. Mulders (TUE) for his carefully reviewing the
thesis.
Finally, to Xiao-wei, my wife, for her patience, encouragement, and love

tbraughout the endeavor.
CONTENTS
Contents
Nomenclatures iv
Acknowledgements vii
Chapter 1. General Introduetion 1
1.1. Gas hearings and their history 1
1.2. Applications 2
1.3. Advantages and limitations 3
1.4. The scope of the present research 4
PART ONE: DESIGN OF EPG BEARING PADS

Chapter 2. Theory of Extemally Pressurized Gas Bearings 7
2.1. Introduetion 7
2.2. The basic equations 7
2.2.1. The equation of motion 7
2.2.2. The energy equation for gas lubrication 9
2.2.3. The Nusselt number in gas lubrication 11
2.2.4. The continuity equation for EPG hearings 12
2.2.5. The Reynolds equation for EPG hearings 13
2.3. Special aspects 15
2.3.1. Gap shape 15
2.3.2. Bearing body tilt 16
2.3.3. Surface imperfections 17
2.4. Overall characteristics 19
2.4.1. Load capacity 19
2.4.2. Stiffness 20
2.4.3. Possible methods to achleve high stiffness 21
2.4.4. Damping stiffness 22
2.5. Refinement of resttictor flow model 23
2.5.1. Principles of orifice/inherent restrictors 23
2.5.2. Refinement 24
Chapter 3. Numerical Calculations and Experimental Verifications 26
3.1. Introduetion 26
3.2. Numerical methods 26
3.2.1. Linearization 27
3.2.2. FEM formulations 28
ii Design of Gas Bearing Systems for Precision Applications
3.2.3. Bearing design software 31

3.3. Experimental systems 32
3.3.1. The mechanica! structure 32
3.3.2. The instrumentation 37
3.3.3. The test set-up for wall temperature 39
3.4. Tri-conical gap sbaped EPG bearings 39
3.5. The effect of hearing body tilt 44
3.5.1. Higher supply pressure and lower working gap height 45
3.5.2. Higher supply pressure and bigher woricing gap height 47
3.5.3. Lower supply pressure and higher working gap height 48
3.5.4. Lower supply pressure and lower working gap height 48
3.5.5. Rectangular pads 50
3.5.6. Effects of dimensions 50
3.5.7. Experimental verifications 51
3.5.8. Tilt effects on the dynamic performance 51
3.6. Effect of surface imperfection 52
3.6.1. Effect of waviness 53
3.6.2. Effect of rougbness 56
3.7. Wall temperature 57
3.8. Effects of hearing motion velocity 61
3.9. Conclusions 62
PART TWO: DESIGN OF EPG BEARING SYSTEMS

Cbapter 4. Modeling of Linear Guides with EPG hearings 65
4.2. Linear guides 66
4.2.1. Kinematical mounting 66
4.2.2. Loop stiffness 68
4.2.3. Symmetrical, arrangement 69
4.2.4. Kinematical driving 69
4.2.5. Preloading 69
4.3. Model to describe the real hearing gap height 70
4.4. ldealization of gas film stiffness 73
4.4.1. The description of displacement dependenee 74
4.4.2. The description of frequency dependenee 75
4.5. Dynamic model 78
Contents iii
4.5.1. The equation of motion 78
4.5.2. The natural frequency 79
4.5.3. Solution methods 82
4.5.4. Determination of vibration modes 84
4.6. Effects of non-linearity of film stiffness 85
4.7. Conclusions 86
Chapter 5. Experiments on a Linear Guide with EPG Bearing Pads 87
5.2. Design of the linear guide 88
5.2.1. Mechanica! structure 88
5.2.2. Analysis 92
5.3. Ex:periments 96
5.3.1. Ex:perimental set-up 96
5.3.2. The identification of vibration modes 98
5.3.2.1. Fully symmetrical contiguration <n 99
5.3.2.2. Partially symmetrical configuration(II) 100
5.3.2.3. Influence of eccentric mass 103
5.4. Discussions 104
5.4.1. Symmetry 104
5.4.2. Influence of the support of the guide beam 104
5.5. Conclusions 105
Chapter 6. Conclusions and Recommendations 106
6.1. Conclusions 106
6.2. Recommendations 108
Appendix 1. The simplification of the equation of motion 109
Appendix 2. The simplification of the energy equation 112
Appendix 3. The Joule-Thomson coefficient 113
Appendix 4. The FEM formulation of the Reynolds equation 114
Appendix 5. The analysis of ball-cone pair 117
Appendix 6. The analysis of rigid-flexible mounting pair 120
Appendix 7. The equations of the vibration motion of the slide 122
Appendix 8. Test results of vibration of the slide 125
Reierences 128
Summary 134
Samenvatting 136
Curriculum Vitae 138
NOMENCLATURES
English letters
A : hearing area m2
Ad 2
: restcictor hole area m
CJ : the reduced Reynolds numher
CJo : the specific reduced Reynolds numher
Cm : reduced Mach numher
cd : discharge coefficient
cp : heat capacity · at constant pressure J kg-1K1
div : divergency
dd : restcictor diameter m
DD : hearing pad dimeter m
DC : diameter of conical region of tri-/bi-conical pads m
e : eccentricity m
E : elasticity N m-2
f : the vector of body force N
f : frequency -1
s
Fd : dynamic force (moment) N (Nm)
Fm : mass force (inertial moment) N (Nm)
grad : gradient
h : hearing gap height m
h' : dynamic change of hearing gap height m
ho : reference gap height m
he : average gap height in an element m
ht : gap height change due to tilt at edge of a pad m
hw : working gap height m
lm : imaginary
J, j : the moment of inertia Nm
: j = " -1
k : ratio of specific heat, cp/cv
k
i
: permeability of porous materials m2
K : correction factor for turbulent restcictor
K : complex stiffness Nm-1
Kn : the Knudsen numher, lp/h
lp : molecular free path m
L : reference dimension m
Design of Gas Bearlng Systems for Precision Applications V
Lx,Lz : half-wave length of a surface wave m

th : mass flow kg s"1
mb : mass of a pad kg
Mx,Mz: moment capacity Nm
M : the reference Mach number
c
Ma : the Mach number
M : mass or inertial moment kg (kgm)
Nf! : viscosity number
Nu : the Nulsselt number
Pe : environmental pressure Pa
po : the pressure in the upstream of the restcictor Pa
P,p : pressure Pa
Pr : the Prandtl number, Wf!Cp
r : cnordinate m
R : gas constant mz s·2 KI
Re : the Reynolds number, QVh/f!
Re' : the reduced Reynolds number, 1\J Re
s : stiffness N m"1
Sb : average bulk stiffness N m·1
T : temperature in surface 1 K
1
Tz : temperature in surface 2 K
T : temperature K
To : stagnation temperature or reference temperature K
T : static temperature K
s
t : time s.
u : vector form of velocity in the hearing surface m s·1
V : vector form of velocity ms·1
u,v,w : velocity components m s·1
w : load capacity N
w : load capacity per unit area N m·2
-l
U, V, W : reference veloeities ms
U', V', W' : slip veloeities m s·1
x, y, z, X, Z : coordinates m
Greek letters
a, I) : initial phase angles of waviness rad.
ê : amplitude ratio of waviness,
vi NOMENCLATURFS
: b= 1 in resttictor area, b=O others

: wave orientation angle rad.
: exponent concerning heat transfer process,
: the ratio of hearing gap and length or porosity
: coordinate rad.
: tilt angles rad.
: distance to tilt axis m
: ratio of hearing length and width
: the coefficient of heat conductivity W m' 1K' 1
: viscosity of gas Pas
: apparent viscosity Pas
: reference viscosity Pas
: Joule-Thomson coefficient Pa K 1
: density of gas kgm·3
: gas density near the hearing surface kg m- 3
: change
ro : angle frequency rad. s·1
a, (3, y : rotation of slides rad.
superscripts
o : initial
: in-phase part
u : out-of-phase part
footnotes
d : downstteam resttictor I dynamic
e : entrance
: film
h : surface connected to the hearing pads
: index
in : inlet
: Joule-Thomson effect
o : static or reference condition or specific
R : resttictor
: resttictor
m : mass
w :wave
x. y. z. r : coordinates
CHAPTER 1.
GENERALINTRODUCTION
1.1. GAS BEARINGS AND THEIR HlSTORY
An .§xternally ~essurized Qas (EPG) hearing is a kind of gas hearings in

which gas is used as a lubricant Like other fluid lubricated hearings, it
serves two purposes: one is to support an extemal load; the other is to
lubricate a pair of surfaces. In principle, load capacity of gas hearings can
be provided by several physical phenomena, such as by external pressurization,
relative shear motion, squeeze motion, supersonic effect, jet effect,
Bernoulli effect, stretching and thermal effects. The EPG hearings are
commonly used in the field of Precision Engineering so that they are dealt
with in this thesis. In elements with externally pressurized gas lubrication,
any relative motion across the film thickness is compensated by squeeze
effects in the film and by restrietion effects in the hearing gap and inlet
restrictor(s). The first investigation into EPG hearings was the experimental
work on the gas flow between two parallel surfaces [Willis, 1828]. Twenty-six
years later, Hirn [Hirn, 1854] further stated that the gas might be used as a
lubricant, but unfortunately gas hearing technology was limited by the general
technological level of that time. The important theoretica! breakthrough was
the derivation of the pressure equation for thin fluid films [Reynolds, 1886].
Thereafter, with the advancement of the industrial revolution, the fluid film
hearings began to appear in machines, as well as gas hearings. The practical
use of EPG hearings did not emerge until W. H. Wood [Wood, 1890] patented the
counterbalance system using extemally pressurized gas. The early application
to metrology systems was the work at NPL (National Physics Laboratory. UK) in
the 1930s. However, the EPG hearings (generally speaking all the gas hearings)
were not taken seriously until 1959 when the First International Gas Bearing
Symposium was opened [Fuller, 1959]. Thereafter, gas hearings have been
applied in all the fields of engineering. In the field of Precision
Engineering, EPG hearings .are more popular because they provide better
performance than the other types of gas hearings.
2 Design of Gas Hearing Systems for Precision Applications
1.2. APPLICATIONS
Applications of EPG hearings may be found in all the engineering fields,
especially in the field of Precision Engineering. EPG hearings (strictly
speaking, extemally pressurized gas lubricated elements) are mainly used for
four purposes, i.e. motion supports (including joumal and thrust hearings in
spindles, guideways and leadscrews), actuators, fluid sealings and mechanical
couplings, which are illustrated in Fig. 1.1.
Pl
Pl
Hearing Pad
Pl
Linear slide
Pl
P2
Pl
Journal Hearing Air Screw Actuator
Sealing element Motion Coupling
Fig. 1.1. Schematic illustration of EPG hearing family
The practical examples can be found in the following machines :

* metrology systems, e.g. :
-- coordinate measuring machines
CHAPTERl.GENERALINTRODUCTION 3
-- roughness and roundness measuring machines

-- torque measuring systems
-- probe systems
* cutting systems, e.g. :
-- diamond turning and milling machines
-- grinding machines
-- print-board drilling machines
* lithographical systems, e.g. :
-- optical exposure instruments
-- pattem writing and inspeetion instruments
- silicon waver cutting and polishing machines
* astronomical purposes, e.g. :
-- interferometers
-- gyrometers
* turbomachinary, e.g. :
-- turbocompressor
-- steam lubricated blower
1.3. ADVANTAGES AND LIMITATIONS

Because gases are used as lubricant, EPG hearings have the following
distinguishing characteristics:
1). Low and near constant friction: Because gases have very low values of
viscosity and less sensitive to temperature and pressure change, the friction
coefficients of EPG hearings are very low and are nearly constants, e.g. the
reduction of residual friction torques to 10 · 6 of that of the hall hearings
was achieved in the dynamometer used in the cage support of guidance gyros for
roeket vehicles [Rothe, 1959].
2). No metal contact during operation: This is because the gas film
separates the two hearing surfaces, which makes such hearings have a very low
level of wear and mechanical noise.
3). Very high accuracy and very long accuracy life: The distributively
pressurized gas presents averaging effects on hearing surface imperfections,
e.g. roundness of a shaft. Therefore, very high accurate axis definition can
be easily achieved [Li and Wang, 1983].
4). Wide working temperature range: In high temperature lubricated
applications, the working tempersture of gas hearings is limited by the solid
components in a machine. However, at the lower end, the temperature scale

condensation of the gas may become a limitation. Complex gases on the other
hand will have decomposition limitations at the upper end of their usabie
temperature range.
5). Environmental friendliness: The gases used in EPG hearings are the
same as the environmental media, e.g. air in workshops and measuring rooms.
Therefore, no pollution is introduced.
6). Micro-displacement feature: Although it is not really separable, this
is quite useful in precision applications. The difference of supply pressures
in two opposite directions is used to adjust a rotor position. For instance,
it has been used as an actuator for high accuracy positioning mechanism [Kanai
and Miyashita, 1983] and for motion error correcting systems [Shimokohbe et
al, 1986].
There are some necessary conditions in order to make EPG hearings work
properly. The specifications of these conditions depend on the requirements of
a machine. In general, the following conditions are essential:
1). Continuous supply of externally pressurized gas with reasonable
stabilization of pressure and filtration of dust, oil, water and other harmful
matter.
2). Carefully prepared hearing surfaces with acceptable waviness and
roughness.
3). Proper selection of heat expansion coefficients of hearing wall
materials in the cases of high speed motion or of having a big change of
environmental temperature.
4). Correct design of mechanica! connections with sufficient mounting
stiffness and possibility of self-alignment.
5). Taking care of gas temperature and temperature change through
restrietion process.
1.4. TUE SCOPE OF TUE PRESENT RESEARCH

The development of gas hearings follows the general trend of
technological progress. With ever increasing machine accuracy, the EPG
hearings have to provide better performance. In the field of gas hearing
technology, there are still less understanding and poor quantification on
the various subjects, such as:
1). accurate and correct modeling of the restrictor flow;
CHAPTER 1. GENERAL INTRODUCTION 5
2). detailed study into the electronically controlled restrictor;

3). transient behaviours;
4). various subjects related to the application in precision engineering.
The following topics closely related to the applications in Precision

Engineering will be discussed in this thesis:
1). influences of gap shape on the hearing perfonnance;
2). effects of hearing body tilt on the hearing perfonnance;
3). influences of hearing surface imperfections on hearing performance;
4). temperature drops in EPG hearings in quasi-stationary conditions;
5). the optimal use of multiple hearings in a mechanica! system;
6). the use of EPG hearing damping characteristics in the design;
7). effects of motion velocity.
The aim of the research is to provide the up-to-date knowledge and the
efficient tools to design better EPG hearing systems, including the design of
individual hearings and to arrange all the hearings used in an optimal way.
Both circular and rectangular EPG pads with rigid hearing surfaces and rigid
inlet resttictors will be used in this research.
This thesis consists of two parts: Part one presents the design methods
of single pad..'>, as well as the numerical and experimental research on the
above-mentioned special aspects. Part two gives the principles to design more
accurate EPG hearing systems, as well as simulation and experiments on linear
motion systems.
PART ONE
DESIGN OF EXTERNALLY PRESSURIZED GAS BEARINGS
CHAPTER 2.
THEORY OF EXTERNALLY PRESSURIZED GAS HEARINGS
2.1. INTRODUCTION
From a theoretica! viewpoint, gas lubrication is concerned with the study
of gas flow between two surfaces, the distance between these two surfaces
being very small both in absolute value (less than 20 ~m in most applications)
and in comparison with the hearing dimensions (10- 3 -10- 4 ). Nevertheless, under
those conditions, it is reasanabie to neglect the molecular characteristics of
the flow in the lubricant film of Externally Pressurized Gas (EPG) hearings
and to consider the motion of gas as that of continuons fluid. However, if the
permeability of a surface is high, e.g. hearings with porous walls [Majumder,
1988 & Sparrow et al, 1972] or the Knudsen number is greater than unity [Fukui
and Kaneko, 1987], such as self-acting hearings used in the magnetic driver of
information starage systems, the slip velocity at the hearing walls should be
taken into account. In this chapter, we first discuss basic equations for gas
lubrication, particularly in the EPG lubrication. Furthermore, the pressure
equation, known as the Reynolds equation will be discussed in detail with
consideration of hearing gap shape, hearing body tilt and surface
imperfections. Afterwards, the overall characteristics will be described, with
special attention to the relationship between basic concepts in gas dynamics
and fluid mechanics, which can reveal ways to imprave hearing characteristics.
Finally, the models for hearing wall temperature and the refinement of inlet
restcictor flow will be presented.
2.2. THE BASIC EQUATIONS

The equations used in the EPG hearings are adopted from the equations for
viscous compressible fluids which are derived from the physical properties of
gases and from the general equations of mechanics and thermodynamics.
2.2.1 The equation of motion

The equation of motion can be derived by characterizing the dynamic
equilibrium between the inertia farces, the mass farces and the stresses in
the fluid. By camparing the relative order of various terms in the equation of
motion, it can be taken as a simple form in the gas lubrication (the
simplification of the equation of motion will be given in Appendix 1) :
8 Design of Gas Bearing Systems for Precision Applications
(2.1)
Y(h) Y(h)
-z
CIRCULAR PAD RECTANGULAR PAD
Fig. 2.1. Illustration of the coordinate system
From the reference [Wang, 1990], the dimensionless form reads :

2 DÜ Ri> 8 (- 8Ü)
k Ma - =- ~ + Nt-t -::- t-t --
Di ax 8y ay
0 =- ~ (2.2)
By
k O.Ma )2 D~ =-~+Nt-ta_ (JÏ 8~)
Dt az ay ay
where k : ratio of specific heat, k = 1.4 for air,
R : gas constant,
Ma. : the Mach number, Uo/(kRTo) 112 , note that Q/PO "' 1/(RTo)
Uo : reference velocity,
To : reference temperature,
Nf.t : the viscosity number, Nt-t "' f.!O/t-t',
t-t' : the apparent viscosity, t-t'=(ho/L)Poho/U which is a measure
of the combination of the eperation parameters of a bearing.
CHAPTER 2. THEORY OF EPG BEARINGS 9
x = L i:, y = ho y, z = AL z,
u = U Ü, v = q> Uv, w = U A w,
t = (UU) t, p = Po p,
t-t = f.tO f.t, q> = ho/L, A = Lw/L,
L, Lw : reference dimensions in two coordinate directions separately
ho, To, Po, f.tO : reierenee values.
The reierenee parameters should be carefully selected, otherwise the

dimensionless form will lead to the wrong result. The reference velocity for
EPG hearings should take the gas flow velocity not that of the hearing motion
velocity as it is used in the self-acting hearings. This is because the
pressure is generated by hearing motion. From the dimensionless farm of the
equation of motion, it can be seen that the inertia force term in the lefi-
hand side of the equation of motion is govemed by a specific form of the Mach
number, k(Maf The Mach number is the ratio of the gas velocity over the
velocity of sound, its physical meaning can be interprered as a measure of the
weight between the inertia force and the pressure force. In linear motion
systems, the optimal working gap height varles between the 10 ~ 15 t-tm for
parallel gap shaped EPG hearings and around 5 t-tm for conical or tri-conical
gap shaped hearings. From the observation on the flow phenomena of EPG
hearings, given in Fig 2.2, it can be seen that the specific Mach numbers are
negligible in the range of the optimal working gap height. This suggests that
we can even take a simpter form of the equation of motion, i.e. without
concerning the inertia effect :
0 = _ ~ + !!._ (t-t au)

ax ay ay
0 - ~ (2.3)
ay
O= _~ + !!._ (f.t aw)
az ay ay
2.2.2. The energy equation for gas lubrication

Estimation of the relative order in the energy equation yields the
following form for gas lubrication(the derivation will be given in Appendix 2)
DTo = -a (% -aTs) + .:::.t:.

Q Cp ~ an + -a [ a~ + -a [ aw) (2.4)
ut ay ay at ay f.tU -a ay f.tW -ay
The dimensionless farm can be written as :

Dto
CJ --- a (-% ---
= ----:- ats) + CJo (an)
.::.:=- + Cm [B
----:- (-- aÜ\
Uf.t --:-I +----:-
BWi] (2.5)
a (-Wf.t- ---_/
Dt ay ay at ay ay ay ay
10 Design of Gas Bearlng Systems for Precision Applications
2.5 ...-------,.---,.-_,"-"--,.....".,
__,-__-_-
__---, 3~------~------.-------.
1.5
1
'iscosity number(N~
0.5
working gap (pm) working gap (pm)

(a) tri-conical gap shaped circular EPG hearing pads
1.2 ~-------.---------,--------, 0.045 r-------...--------.------,

,
/
0.04
1
'
0.035 '
'
0.8 0.03
0.025 ''
'
0.6 ''
'
0.02 ''
''
'
0.4 0.015 '
0.2
0.01
0.005
/
edu, -' Reynolds number(Cj)
o~~ced~_M_a~c_h_n_u_m_be
__ r(~C_m~)----~
5 10 15 20
working gap (pm) working gap (JJm)
(b) reetangu/ar EPG pads with orifice restrietars

Fig. 2.2. The reduced Reynolds number (ÇJ). the reduced Mach number (Cm),
the specific Mach number (kMa) and the viscosity number(Nw
where
Re : the Reynolds number, Re = Q u b/J.t,
Pr : the Prandtl number, Pr=ep IJ/x
Cr : the reduced Reynolds number, Cr = Re Pr q>, q> = h/L
Cro : the specific reduced Reynolds number, Cro = Reo Pr q>
Reo : the specific Reynolds number, Reo = Q0 u0 h/J.t0
Cm : the reduced Mach number, (k-1) M;
'l's = Ts/T, 1'o = To/T, ie = x /1tr, To, 1tr: relerenee parameters
In cases where the inertia effect can be ignored, the gas stagnation
temperature on the left-hand side of the energy equation can be replaced by
the gas static temperature.
From Fig. 2.2b, one can see that both the reduced Mach number and the
reduced Reynolds number are much smaller than the unit for the hearings with
the orifice restrictor. Hence, the energy equation in this situation is
reduced to the conduction term alone. It is also true for the conical gap
shaped pad at the optimal working gap height. However, the convection term
still remains for the conical gap shaped hearings in higher working gap
height :
8 _ at ("-) { CJ D~o for conical/triconical hearings
(x -")+ Cro ~ = Dt (2.6)
ay ay at 0 for hearings with orifice restrictor
2.2.3 The Nusselt number in gas lubrication

The integral of the equation (2.6) with the consideration of the boundary
conditions yields the Nusselt number in two hearing walls. The geometrical
meaning of the Nusselt number is the dimensionless temperature gradient at the
walt. Physically, it is a measure of the weight between the heat conduction
and the convection. Hence, the energy equation can be written as :
h1' h - h 1' h h-
CJ [r 8 _ 0dy +Ju ~y+ 1 J w 8 _0dy +Jv dTo)- CJoJ~dy
Nub + Nuo = o at o ax À o az o o at (Z.?)
h -
{
CJO! ~ dy
o at
In the beat equilibrium condition, it is much simplified
h 1' h 1' h
CJ (J ü a ~dy+ 1 J ~y +I dTo] w v
Nub + Nuo = o 8x À o az o (2.8)
{
0
2.2.4. The eontinuity equation for EPG hearings

By denoting the components of the velocity V of a fluid partiele with u,
v, w, the equation of continuity in a cartesian coordinate system (x,y,z)
reads :
~ + div(Q V) = 0 (2.9)
or the expansion form :
~ + a(QU) + a(QV) + 8(QW) = O (2.10)
at ax ay az
Por the thin film problem, it is convenient to consider the integral form
of the continuity equation :
h ~ dy +Jh a(QU) dy + ~ a(QV) dy +Jh 8(QW) dy 0 (2.11)
J 0 at 0
ax j 0 ay 0
az
Using the mathematica! formula [Constantinuscu, 1967] :

ha a~ ah
J0
ax [f(x,y)1 dy = ax J f(x,y) dy - f(x,h) ax
h (2.12)
I)
the continuity equation becomes :
a~ih) + :,cJ\u dy+ :ztQw dy=-(QVh+ ~ BQYri)+Q(U~+ W~+ :~) (2.13)

0 0
where:
footnotes for U, V, W have the following meanings: h stands for the
velocity of the surface h and o for the velocity of the surface in X-Z
plane shown in Fig. 2.1; r indicates velocity at the restrictor. In fact,
ah/at = Vh, the hearing body motion along the y-axis. Por an EPG hearing
with impermeable walls and one surface slow motion (Uh and Wh => 0), only
one term remains on the right-hand side of the continuity equation, i.e.
~ ö Q. V . , where V . is the function of Q., the l: stands for the sum of
1 r t r 1 1
all the inlet restcictors and ö is an operational factor to indicate

whether the restcictor is included or not.
If the gas film is bounded by permeable materials, the velocity in the

hearing surface h is equal to the slip velocity, so that the terms on the
right-hand side of the continuity equation read :
Q [ Uh:~ +Wh :~ + Vb] + l: öQyri (2.14)
Por the non-permeable hearing walls, the components of slip velocity

disappear and the continuity equation becomes

B(Qh) +
at
!!..___
ax
Jh QU dy + !!.__f'\w dy + :E Q.ÖV . = 0
azj 0 , "
(2.15)
0
The term Q1. V r 1. represents the unit area mass flow from a restcictor in
the perpendicular direction to the gas film. Mathematically, it is a remaining
term from the equation of continuity to describe the mass flow from a
restcictor which cannot be indicated by the two dimensional Reynolds equation.
In order to derive the pressure equation - the Reynolds equation, the
velocity terms in the continuity equation must be replaced by the pressure
terms. The equation of motion gives the relationship between the pressure and
the velocity.
2.2.5 The Reynolds equation for EPG hearings

This type of equation is so called due to the fact that Lord Reynolds
first derived it for a thin oil film [Reynolds, 1886]. The Reynolds equation
is derived by introducing the expression of velocity into the continuity
equation while the velocity expression can be found by solving the equation of
motion. The Reynolds type of equation was also given in general form by
several authors [Cope, 1949] [Dowson, 1962] for lubrication problems. However,
the commonly used Reynolds equation for EPG hearings is adapted from that for
hydrodynamic gas or aerodynamic hearings. The problem is that the entrance
effect is difficult to incorporate because there is no description in the
region between the outlet of the restcictor and the entrance of the hearing
gap. The attempt to solve this problem was to introduce the entrance
correction factor [McCabe et al, 1969] [Vorh, 1969] [Bennett et al. 1976] [Pan
1980] [Wen et al, 1983] [Scharrer and Hibbs 1990] or to modify the "common"
form of the Reynolds equation by solving the equation of motion with inertia
terms [Mori and Mori, 1981]. In fact, the entrance can be incorporated in the
Reynolds equation by introducing the outlet boundary condition of the inlet
restcictor during the derivation of the Reynolds type of equation from the
equation of continuity.
In cases where the inertia effects can be neglected, the velocity
distribution across the film thickness as the tunetion of x or z can be found
by integrating the equations of motion (4). i.e. :
(2.16)
where
pr= Jyrofl dy, (2.17)
By introducing the velocity distributions into the continuity equation,

the corresponding Reynolds equation can be found :
div [-II 1Q gmd(p)+ Q U II 0 ] + BBhtQ + }:ö Q.V . = 0 (2.18)
l rl
where
lil= (Fr. flh (fOrfiOh)], Ifl= (fOrfiOh) (2.19)
Investigations [Kao, 1963][Constantinuscu, 1969] showed that the Reynolds

type equation in cases where two walls have different temperature values can
be replaced by the "normal" Reynolds equation with the average temperature of
two walls. Therefore, only the Reynolds equation under isothermal conditions
is more interesting in detail :
div [- ~;fl p grad(p)+ phU/2] + ~ + f (p) = 0 (2.20)

where
h is the bearing gap height. In general, it is a function of coordinates,
i.e. h = h(x,z), which includes the contribution from bearing float
height {defined as working gap height (hw)}, gap shape (hg), surface
imperfection (hs) and pad body tilt (ht). In some designs, it is also a
function of pressure (hp), where bearing surfaces are flexible. It can be
written as :
h(x,z)= hw + hg(x,z) + ht(x,z) + hs(x,z) + hp(x,z) (2.21)
f(p)= 1; öp .V .(Tr/fri) (2.22)
n n
The relationship between the entrance temperature (T;) and the film
temperature (Tr) can be found through the conservation law of mass flow,
i.e. the mass flow at the exit of the resttictor must be equal to the
mass flow at the entrance of bearing gap :
rhe = rhr (2.23)
or Ve pe Ae/RTe = Vt pt Ar /RTr (2.24)
2
Introducing the expressions for the areas At =x d h and Ae =x d /4
yields :
Ve pe (Ter/fe) = Vr pt 4he/d (2.25)
Eventually, the term f(p) reads:
CHAPTER 2. THEORY OF EPG HEARINGS 15
(2.26)
f(p)= ~ ó pnVn(4he/d)
where d : resttictor diameter; he : entrance gap height
In genera}, for the circular pads with rotation, it is convenient to use

the expanded form of the Reynolds equation in the cylindrical coordinate
system. In this coordinate system, the film thickness direction takes the
height direction (z). The Reynolds equation for joumal hearings takes radial
direction as the film thickness direction in the cylindrical coordinate
system. The Reynolds equation for the rectangular pads and the circular pads
with linear motion is expanded in the Cartesian coordinate system and the
z-direction is the film thickness. For the spherical hearings, the Reynolds
equation is expanded in the spherical coordinate system and the radial
direction is the film thickness direction. For hearings in linear motion
systems, the expanded form of the Reynolds equation in the rectangular
coordinate system is more convenient :
a <trf.1 Bx}
ax
ap.. + a <tff.1 ~)
8Z 8Z
= ~+~hU
8t 8x
+ ~hW
8z
+ f(p) (2.27)
2.3. SPECIAL ASPECTS
2.3.1. Gap shape

Traditionally, the hearing gap shape is a linear line with zero slope,
formed by two parallel surfaces. Research revealed that the modified gap shape
can provide better overall characteristics. such as linear surface with a
non-zero slope (conical gap shape) [Blondeel et al, 1976], multi-linear pieces
with different slopes, two different slopes (bi-conical gap shape) [Blondeel
et al, 19761 and three different slopes (tri-conical gap shape) [Wang, 1990]
and flat surface with grooves (so called surface restriction). The gap shape
of a multi conical circular pad can be described in the following way, the
definitions of the terms are given in Fig. 2. 3 :
n
hg i =a i hv i + ~
j "" i +1
hv j (2.28)
where
a (r. -r )/(r. -r. ) when r > r > r i- l ,
l t t - 1
r 0 : restrictor radius, hv. : ·characteristic gap difference,

I
r. : characteristic radius, i : the index of conical parts.

'
2.3.2. Bearing body tilt
The bearing pad body may be rigidly tilted due to the errors of
manufacturing and assembly, as well as the distartion caused by unsymmetrical
loading and/or thermal loading. As a result, the bearing gap will be changed
by the tilt, the change due to tilting can be expressed as :
ht=xO+yT] (2.29)
where
TJ, 8 : the tilt angles, 9 = htx/(DD/2), TJ = htz/(DD/2) (2.30)
DD : the pad diameter, htx, htz : gap height change due to tilt
y y
without tilting with tilUng

y
gap helght, e.g. hg2

hv3 hg2= a2 • hv2 +hv3
a2 = (r2-r)/(r2-rl)
Fig. 2.3. Illustration of multi-conical circular EPG pads
For circular pads, tilting direction can be easily assigned to one set of
coordinates. Therefore, the change of gap height due to tilt can be simplified
as follows :
ht =x e (2.31)
Because the order of the tilt angle is less than 10-3, the influences of
it on the reference coordinate can be ignored. The theoretica! study on a
particular example was reported [Pande, 1985], later these types of hearing
pads were more extensively studied under one dimensional model [Al-Bender and
Brussels, 1992]. However, this thesis presents the comparative study on the
different types of EPG hearing pads based on two dimensional model.
2.3.3 Surface imperfections

Typical discussions on the roughness effects [Mitsuya et al, 1989]
[Tonder, 1984] [Raad and White, 1989] and waviness effects [Hayage, 1990] in
hydrodynamic lubrication can be found in the literature. However, no related
pubHeation has been found for EPG lubrication. Reviewing the denvation of
the Reynolds equation, it is found that the Reynolds equation is valid only
when the inertia effect is neglected and the dimeosion ratio (the gap height
over a reierenee length) is very small. Engineering surfaces always have
certain imperfections due to errors of manufacturing and assembling, as well
as the distortions of structures. The use of the Reynolds equation in such
surfaces still requires those two conditions. One should note that in such a
condition, the local dimeosion ratio (the roughness/waviness amplitude over
the half wave-length) has to be considered. When this local dimension ratio
meets the requirement, the Reynolds equation can be used to solve rough
surface hearings with real gap shapes. If a surface can be considered as a sum
of sine waves, the gap height contributed from surface imperfection can be
written as :
hs = ho~ ë.
1
sin[3t(x+X)/Lxi+(z+Y)/Lzi+ a.+f3.l
I I
(2.32)
where
h0 : reference gap height
e. : relative amplitude of ith wave
1
a;' 13; : initialphases of ith wave

x,z : local coordinates, fixed on an EPG hearing
X, Y : reference coordinates, fixed on the reference
Lxi, Lzi : half-wavelength in X- and Z-direction
For the clearness and simplicity in the theoretica! investigation into

the effect of wave length and initia! phase, one wave is considered.
Furthermore, the surface waves may have their orientations. In order to
calculate the imperfection effect, the following form can be used :

hs = h [e sin(x (x+X) cos«jl/L)] (2.33)
0
where
Ijl : the orientation angle.
When the direction of motion is different from the wave direction, the
apparent wave length will differ from the actual one. In the special case,
when the moving direction is perpendicular to the wave direction, the position
change will not change the moving body's behaviour. However, as expected, in
the non-perpendicular direction, the performance of a rnaving body will change
with the position change. For circular pads, the orientation of the surface
wave will mean that flow in the gap takes a different pattem in the different
radial direction.
y(h) y(h)
z z
z
y
x
Fig. 2.4. Illustration of the coordinate systems for waviness study
Because the study on roughness effects is closely associated with the

salution methods, it will be treated in Chapter 3. The references [Gans, 1987]
[Patir and Cheng, 1984] [Wang, 1990] present the treatments for the random
pattem of surfaces.
CHAPI'ER 2. THEORY OF EPG BEARINGS 19
2.4. OVERALL CHARACTERISTICS

Because EPG hearings as mechanical elements are used to support rnaving
bodies, the users of EPG hearings are more interested in the overall
characteristics, such as the load capacity, stiffness, mass flow rate and
damping. This section will present the general concepts of the global
characteristics and some common ways to express them.
2.4.1. Load Capacity

The load capacity may he further divided into the hearing carrying force
(W) and moment (M). The hearing carrying force is the integral of pressure
distribution (p) over the whole hearing area (A). The hearing load capacity
can he expressed as follows :
W =JAp dA (2.34)
Using the relationship p Q RT, Q dM/dV and dV h dA, the load

capacity reads :
W = RT M/h (2.35)
where R : gas constant, T : temperature, M : mass content of gas
h : average gap height
The load per unit area can then he written as the following :
W wIA = A 1 1 - (2.36)
JA p dA = A JA KT dA = KT
Which means that the load per unit area equals to the average isotheemal bulk
modulus ( K) of gas in a hearing. KT is the local isotheemal bulk modulus.
The hearing moment relative to the rotation axis of tilt reads :
(2.37)
The net load capacity and moment capacity of circular EPG pads with
atmosphere as environmental pressure have the following forms :
W = fRS: (P-Pe) dz dx (2.38)
Mx = .( .C (P-Pe) z dz dx (2.39)
Mz=.t
-R
rx
-xx
(P-Pe) x dz dx (2.40)
where xx=(R2 - x2)1/2 (2.41)

and rectangular EPG pads have the forms :
W = !~a .r\ (P-Pe) dx dz (2.42)
Mz = räa !bb(P-Pe)
b
x dx dz (2.43)
Mx =!_a !_b(P-Pe) z dx dz (2.44)
z z f..-a-l
f
1-cï
dy l t 1
-1.1'4 .r
x
dx
rx
•
Fig. 2.5. lllustration of pad dimensions
2.4.2. Stiffness
The stiffness is the derivative of the load with respect to the working
gap change which can be expressed as follows :
S = - dW/dhw = ! A d/dhw dA (2.45)
The negative sign indicates that the change of the hearing working gap
bas an opposite direction to the load change. Using the relationship (2.35),
the stiffness can be expressed as follows [Wang et al, 1988] :
(2.46)
s sb + sMi - sMo (2.47)
Where Mout : amount of gas flowing out of a hearing
Min : amount of gas flowing into a hearing

Sb : average bulk stiffness, Sb = ItT A/h = W/h
SMi : inlet stiffness due to mass content change caused by
mass flow into the hearing
SMo : outlet stiffness due to mass content change caused by
mass flow out of the hearing.
This stiffness expression reveals the design principle for high
stiffness, and infinite stiffness hearings. For the hearings with rigid
hearing surfaces and a rigid restrictor, the stiffness takes average bulk
stiffness as the highest limit which can not he achieved in the finite
frequency range hecause the domination of inlet restrietion effect. In
principle, stiffness can he enhanced by designing a hearing with the dominant
effect of outlet restriction. On the other hand, it can he seen that very high
stiffness can he achieved by controlling the restrietion effect in a correct
way. For instance, the hearing with the membrane provides a negative outlet
stiffness. When hearing load increases, the working gap decreases, at the same
time, the membrane is deformed towards more concave. A suitable design of the
membrane can provide the increased net change of mass flow content, i.e.
negative change of mass flow content with respect to the working gap height.
Therefore, the static stiffness can he much enhanced.
2.4.3. Possible methods to achieve high stiffness

The restrietars are key elements in improving the stiffness of EPG
hearings. The pressure distribution in the hearing is determined by the type
of inlet restcictors and the types of hearing gap shapes. A different type of
the inlet restcictor delermines a different boundary value of the inlet
pressure, as well as the pressure gradient with respect to the working gap
height, which delermines inlet stiffness. The hearing gap shapes determine the
pressure gradient with respect to the coordinate and with respect to working
gap height, which means that the hearing gap shape determines the outlet
stiffness.
The types of restcictors are illustrated in Fig. 2.6. The hearing gap
shape controls the gas out-flow which can he seen as an outlet restrictor.
From their functions, the restrietars can he divided into two types: one is
the rigid restcictor and the other is the changeable restcictor which is
applicable for both the inlet and the outlet restrictors. The rigid restcictor
.fl_ n
-----../ '-- b~
7/7777777777
J n~y~ic ~
77777777777
///T//T//7 77?7777777
slot orifice po rous device
l c:=J----@-
n
-./'--
/7//7/7////
~~
/////////7
dL
/////////7
control signal
inherent gaso-elastic piezo-electronic
basic types of inlet restrietars
~
J]_
. Jl . ~~
ooOoo
-J'---
/7//////7 //?'/////// ////////// 7$//////
parallel conical grooved membrane
basic type of outlet restrietars (gap shapes)
Fig. 2. 6. Illustration of restrictor types
is named according to the geometry of resttictor shapes. The rigid restcictor

can be further classified as hole, slot and device resttictors for the inlet
restcictors and as simple (parallel) or modified gap shapes (such as conical
and/or grooves) for outlet restcictor (hearing surfaces). The changeable
resttictors are known by their auxiliary means, such as by the gaso-elastic
properties of matenals or by additional piezo-electric devices. The
changeable restcictor can be used to design extremely high stiffness, as
summarized in the references [Wang, 1988] [Wang, 1990].
2.4.4. Damping stiffness

The shape of the pressure profile is detennined by the gap shape, the
working gap height, and the boundary conditions of the pressure, the flow, the
veloeities and temperature in the ends of the gap outlet and the restriotor
inlet. Therefore, both load capacity and stiffness depend on those parameters.
When those parameters vary with time, the load capacity and stiffness become
complex variables and have two parts: in-phase and out of phase. The stiffness
characteristics of EPG hearings can be expressed by a complex term which may
be called complex stiffness. The imaginary part of the stiffness relates to
the damping properties of the hearing. Therefore, the imaginary part of
stiffness is defined as damping stiffness. The equivalent damping factor can
be defined and will be treated in Chapter 4.
EPG hearings suffer self-excited vibration, termed pneumatic hammer, i.e.
negative damping stiffness occurs. This is due to an unsuitable combination of
operation and design parameters.
2.5. REFINEMENT OF RESTRICTOR FLOW MODEL

2.5.1. Principles of orifice/inherent restrictors
These are two types of restcictors which are commonly used, one is called
an orifice restcictor and the other an inherent restrictor. The flow processes
in those restrictors are traditionally modeled as adiabatic flow because of
the very short flow length. The gas velocity under the conditions of
non-viscous and incompressible flow is equal to the enthalpy change, [Yahya,
1982] i.e. :
u = [2cp (To -T)] 12
1 (2.48)
The mass flow rate can be calculated by :

ril=guA (2.49)
By using the adiabatic relationships :

T!fo = (plp/k-1)/k, Q/Qo = (pip/Ik (2.50)
the commonly used form can be derived :

112
m = cd A P0 [2 k/(k-1)/RTo] x (2.51)
where
X = ([ (p/po)2/k -(P/P o) (k+O/k]1/ 2 when p)Po , [2/(k+ 1)]k/(k-1)
{0.5 (k-1) [2/(k+1)](k+l)/(k- 1)} 112 when pjp .s. [2/(k+1)]k/(k-1)
0
cd is known as the "discharge coefficient" and used to correct the errors due
to the simplification of real flow.
In an inherent restrictor, gas flowing through the restcictor is
proportionally controlled by the hearing working gap height and the restcictor
area (A) which equals to the gap entrance area (A = :n:dhw). In an orifice
restrictor, a recess is introduced between the restcictor and the hearing gap,
its restrietion area is the hole area (A=:n:d 2/4). In applications, a realistic
restrietion area can be obtained by either an equivalent area [Holster and
Jacobs, 1986] or two resttictors connected in series [Gross et al, 1980]. The
consideration on the entrance effect can be found in many references, such as
luminary recovery [McCabe et al, 1969], geometrical effects [Bennett et al,
1976] [Wen et al, 1983] [Scharrer and Hibbs, 1990] and turbulent effect [Vorh,
1969] [Pan, 1980].
2.5.2. Refinement
From the earlier discussion of the derivation of the Reynolds equation
for EPG lubrication, we noted that the resttictor flow was in the y-direction,
which is beyond the limits of the two dimensional equation in the x-z plane
(see Fig. 2.1). Therefore, an extra term was introduced to consicter the
resttictor flow, which directly results from the integral of the continuity
equation. Note that the denvation of the mass flow introduces twice the
uncertainty of a flow process, one by the temperature-pressure transformation
and the other by the density-pressure transformation. The modified Reynolds
equation derived in this chapter shows that the transfarm from density to
pressure is not really necessary. The parameters evolved in the equation are
the local pressure and the local velocity. By doing the density-pressure
transfarm and using mass flow in the salution of the Reynolds equation, the
local temperature appears again. The vague use of the environmental
temperature instead of the local temperature, like that used in many
references, already conflicts with the starting equation (2.48). The second
problem in the traditional treatment of resttictors is to consicter the
turbulent effect. It is physically true that long befare the turbulent occurs,
the inertia effect can no langer be neglected. In practice, the hearing
working gap height is below 20 ~m. even below 10 ~m in many applications so
that the Reynolds number wiJl be below the starting point for turbulent flow.
Those arguments define the present approach to look for a more realistic
way to model the resttictor flow. From the discussion above, it is understood
that only temperature-pressure relationship is important. From the basic
theory of thermodynamics, we understand that devices like the inherent/orifice
resttictor may show the Joule-Thomson effect [Zhang and Shi, 1978]. If the
Joule-Thomson coefficient (~) is added to the equation (2.48), the gas
velocity becomes (the detailed denvation wil! be given in Appendix 3)
112
u = [2cp (To -Td)- ~ (Po - Pd)] (2.53)
J
By comparison with the equation (2.48), the "discharge coefficient" for

local velocity can be detennined by :
112
Cd = [ 1- t-t (Po -Pd)/(To -Td)] (2.54)
1
From experience, it is known that the flow process through the restcictor
is neither adiabatic nor isothermal, but somewhere in between. The
Joule-Thomson coefficient is just a physical quantity to indicate the flow
process. Fig. 2.7 presents the "discharge coefficient" with the pressure
ratio calculated by the equation (2.54).
lr-----~----~-----r----~------~----~-----r-----,
0.9
~ 0.85
1i
t
0.65
0.6~----~----~~--~~----~----~------~----~----~
0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95
pressure ratio (Pd/Po)
Fig. 2. 7. "Discharge coefficient" of the flow through the restrictor

calculated by the Joule- Thomson effect
CHAPTER 3.
NUMERICAL CALCULATIONS AND EXPERIMENTAL VERIFICATIONS
3.1. INTRODUCTION
This chapter deals with the practical aspects of designing an EPG
hearing, including developments of the software and the experimental system.
The Finite Element Methad (FEM) was used to solve the Reynolds equation in
order to calculate the overall characteristics of an EPG hearing. An
experimental system was also developed to perfarm both static and dynamic
tests of EPG hearings. Furthermore, the numerical and experimental results on
both circular and rectangular pads are presented with particular attention to
the effects of gap shape, hearing body tilt, surface imperfections, motion
velocity and wall temperature.
3.2. NUMERICAL METHODS

With the ever increasing development and application of computer
technology, it is convenient to use numerical methods to solve lubrication
problems for achieving more accurate and more sophisticated results. In the
late 60s, the numerical methods for gas hearing film analysis had already been
given more attention [Castelli and Pirvics, 1968]. Recently, several new
methods have also been developed for and applied to the gas lubrication
problem, such as the Boundary Element Methad [Lui et al, 1987], the Finite
Difference Metbod [Tawfik and Stout, 1989], the local partial differential
equation methad [Li and Pan, 1987], and the inner collection methad [Arerkere
and Nelson, 1988]. Nevertheless, the Finite Element Methad (FEM) seems more
popular for EPG lubrications. Two dimensional FEM software was developed for
the calculation of static properties of cylindrical and spherical joumal
hearing [Lui et al, 1983] and of conical and spherical joumal hearings [Lin
et al, 1990]. By using the standard FEM package, one dimensional FEM software
was developed both for static and dynamic properties [Holster and Jacobs,
1989] and was later extended to combine the surface and permeability effects
[Wang, 1990). In this work, two dimensional FEM software was developed to
calculate both static and dynamic characteristics with consideration of body
tilt, surface imperfection and variabie gap shape.
Design of Gas Bearlng Systems for Precision Applications 27
3.2.1. Unearlzation
The Reynolds equation given in Chapter 2. can be further separated into
three equations by a harmonie perturbation around an equilibrium condition, h
= ho + tJt and p = po - t.p. For a small change of periadie working gap height,
say:
tJt = h' sin (rot) (3.1)
if the following change of the dynamic pressure distribution is chosen :
t.p = pi sin(rot) + p 0 cos(rot) (3.2)
the time tenn can be eliminated from the Reynolds equation [Stiffler and
Smith, 1975]. pi is the dynamic pressure change in-phase with working gap
height and p 0 is the out of phase change with working gap height.
The part of the Reynolds equation for static operation reads :
~( pohÖ ~)
éiX ~ éiX
+ ~ro~ho ~)+
élz\~ élz i= I
i öf( ) = élpohoU/2 + élpohoW/2
pi élX élz
(3.3)
where f(p.) p. V. , the expression of V was given in Chapter 2.

J J J
n: the number of restrietars
The part of the equation for dynamic operation with the tenn sin(rot)
reads :
~( pohÖ ~)+ ~(tzhÖ ~) + ~( pi}lÖ ~)+ ~(\'~hÖ ~)
élx ~ ax [jz f.l élz élx ~ élx élz f.l élz
oo az i=l ; élp.
i:
+ 3p h1i' (~ +~) ... ö pi élfo + h' afo
ax ä"fï'
. J .
,. (roh pU)+
0
![(éiJ?~~'U + 8p~~oU ) +(BP~~'W + Bp~~oW)] (3.4)
The part of the equation for dynamic operation with the tenn of cos(rot)
reads :
~( pohd ~)+ ~(pohÖ ~) + ~( p hÖ ~)+ ~(~hÖ ~+i Ö pu_j!f
0 0
ax ~ ax az J2"il az ax 12""il ax az f.l élz i- 1 i ap.

0
J
0
ro(h pi+h'p ) + À{8p hoU + élp hoW } (J.S)
o o "' ax az
The part of the equation for the static operation (3.3) can be further
linearized by using the relation, po = p 0 - t.po :
28 CHAPTER 3. CALCULATIONS AND EXPERIMENTS
3.2.2. FEM equations

The basis of FEM is to divide the total salution area into a finite
number of sub-regions (so called sub-mesh). An example of the submesh plot for
circular pads is given in Fig. 3.1. The triangular element is aften used
because it can be easily treated and has sufficient accuracy. The triangular
element defines the interpolation function as a linear function which means
that the pressure distribution over an element is a linear function of
coordinates. At the edge of the hearing pad, the pressure is equal to the
environmental pressure.
Fig. 3.1. Illustration of the FEM mesh of a circular EPG pad
FEM formulation can be performed with a combination of the Galerkin

method, which can lead to a system of linear equations with nice numerical
properties. The FEM formulation of the Reynolds equation can be derived by
following the standard procedures [Cuvelier et al, 1986]. The resulting final
system of linear algebraic equations for noctal pressure can be written in the
matrix form :
KP=R (3.7)
where P [P1, Pz, ... , PNf
R = [Rl, Rz, ... , RNf
From the FEM formulation, the components of the local coefficient matrix
Design of Gas Bearing Systems for Precision Applications 29
Kij and the components of the right-hand side vector, R; in an element can be
written as (the derivation will be given in Appendix 4) :
Kk; = h~ .i!L [(B,; ~ Pk Bk + Cj ~ PI< Ck)/3 + P~ (B,; Bk+ Cj Ck)]

24fl 1::!.2
- ~zl ~I 0
(U & + W"Ck) (3.8a)
3
R; = he .l!L p0 (B,; ~ PI< Bk + Cj ~ Plc Ck) (3.8b)
24 fl 1::!.2 e
where
Pe = (P 1 + P + P )/3 (due to the linear interpolation function)
2 3
!g.dQ = Jb.J/2 ' !g 1jJ; dQ = Jb.J/6
!!I?
a~;
=i Pi as
J=l
a,p;
N = 3 for the triangular element and f; = x, z
1jJ; = (A.j + B,; x + Cj z)/b.
~ -
B"j
,q,(,. Jl1il:
~ -
cJ
ax - 7S" az - 7S"
Ai = x n zn+l - x n+l zn , B,; = z n -z nt-1 Ci = x o+l -x n
with n = mod[Q+l)/3], (mod is the operation to find remainder)
Note that the resttictor area is so small that the restrictor region can
be treated as a point element. The components in the coefficient matrix and
the right-hand side vector read :
0
0 8f
RJ=(f(p)-p-)Ad (3.9)
8p
The total matrix for the whole hearing area is arrived at by assembling
all the local matrices and the right-hand side vectors.
Therefore, the pressure values in all the nodal points can be found by
solving the set of linear equations, because the components in the coefficient
matrix and the right hand side vector are all constants.
If the continuity equation is only used for the area which does not
include the resttictor region, there will be no o terms in the Reynolds
equation. In this case, the restrictor flow is treated as the natural
boundary, i.e. the flow or pressure gradient at the inlet is prescribed.
In the same way, the FEM tormulation for the dynamic parts of the
Reynolds equation can be derived. In an element, the local matrix can be

written as follows :
KP = 3 he PeR (3.10)
where
p = [pi pu pi pu pi pu]T (3.11a)
1' 1' 2' ~ r 3
M 1 1 -2wN 11 M12 -wN 12 M 13 -w N 13

2 wNu M 1 2 wN 12 M 1 2 wN 1 3 M 1 3
M2 1 -wN21 M22 -2wN22 M23 - wN23
(3.11b)
K= wN21 M21 wN22 M22 2wN23 M23
M31 -WN31 M32-WN32 M33-2WN33
wN31 M31 wN32 M32 2wN33 M33
(3.11c)
1~1 {(Bi }.; Pk Bk

3
Mki =
h
~ ~ + Ci }.; Pk Ck)/M + P~ (Bi Bk+ Ci Ck)/2~
- (U"Bk + W"Ck)} (3.12a)
Nki = hdl~l/24 (3.12b)
R; = wp {h~~~ 0 (Bi }.; Pk Bk + Ci }.; Pk Ck)- (UeBk + W"Ck)} (3.12c)
RY = hd P0 (l) 1~1/1/8 (3.12d)
In the point element, the components in the coefficient matrix and the
right hand side vector read :
0
8f
Kiili=i = 8p Ad R (3.13)
j
Therefore, the element subroutines for building the local matrix can be
programmed by using the equations (3.8) - (3.13). The whole matrix is
automatically assembied by the standard routines in the SEPRAN™ package. The
non-linear form of the static part of the Reynolds equation is solved by the
Newton method. The use of the average value of the supply pressure and
environmental pressure for a pad with distributed restrictor out of the centre
(both a rectangular and circular pads) produces a good initia! estimate of the
static pressure Po. For the pad with a centre restrictor, the computation will
converge more quickly to the desired accuracy by taking an initia! estimate as
Design of Gas Bearing Systems for Precision Appücations 31
p(r) = [Po - (Po -Pe) r/R] 112 (3.14)

where
r = (x2 + z2 ) 112 and R is the radius of the hearing pads
3.2.3. Bearing design software

A standard FEM package, SEPRAN™ was installed in the PC computer,
COMPAQ-386/20e, in the Laboratory of Precision Engineering, Eindhoven
University of Technology, for solving problems of fluid flow or heat flow, and
the problems coupled with mechanica! structures. With the help of this
package, a two dimensional, FEM, hearing design software was then developed in
which effects of gap shape, hearing body tilt and surface imperfections were
added. The software was written in FORTRAN supported by the FfN77/386™
compiler which was developed by the University of Salford, England. The
software is capable of designing circular and rectangular EPG pads, as well as
joumal hearings. Both . the static and dynamic properties can he calculated.
For general purposes, the input parameters are divided into four groups:
1). operational parameters: supply and environmental pressures, frequency
range, environmental temperature, woricing gap height and hearing motion
velocity in the specified coordinate direction;
2). geometrical parameters: pad dimension(s) (e.g. diameter of circular
pads), restrictor diameter, characteristic diameters (e.g. conical region
diameter(s) and gap ditterences for conical gap shaped pads and groove
parameters for both rectangular and circular pads;
3). gas properties: gas constant, viscosity, and specific heat ratio;
4). high order parameters: tilting ratios, surface wave lengths and their
amplitudes, initia! phases and orientation angles of waves.
The flow chart of the computer program is given in Fig. 3.2, which
presents a global view of the program structure. The calculation starts with
defining the input parameters and the desired outputs. The pressure
distribution is then calculated under this specified condition. The load
capacity is obtained by the integral of the pressure over the hearing area.
The pressure distribution, load capacity, dynamic load changes, or stiffness
and dynamic stiffness are separately stored in the ASCII files.
Some calculated pressure distributions are shown in Fig. 3.3, which
demonstrate the capability of the software and show the inside view of the
hearing characteristics.
Parameter modification
Calculation of Dynamic
pressure distribution
Fig. 3.2. The flow chart of the software for designing EPG hearings
3.3. EXPERIMENTAL SYSTEMS

An experimental set-up was developed to perform both static and dynamic
tests on hearing pads. Fig. 3.4 schematically illustrates the system. Through
the air conditioning unit, the dust and oil in the inlet air are filtered out
and the inlet air is also dried. The gas parameters, the pressure,- temperature
and flow rate of the inlet air are separately measured by the barometer, the
built-in temperature sensor and the flow meter. The details will be explained
in the following sections: 3.3.1. the mechanica! structure and 3.3.2. the
instrumentation.
3.3.1. The mechanica) structure
The purpose of the mechanica} structure is to support the elements for
delivering the static an1l dynamic forces and to support the test hearings and
piek-up sensors. The basic requirements are that:
1). its natural frequency should be out of the test frequency range;
2). it should be statically and dynamically stabie during the test period;
3). there should be no interference between static and dynamic loading;
Conical gap shaped pads

Fig. 3.3a. Calculated pressure distributions (circular pads)
r +
Rectangular pad
with slot restriator
r
Rectangular pad
Fig. 3.3b. Calculated pressure distributions (rectangular pads)

Design of Gas Bearing Systems for Precision Applications - 35
parallel gap shaped pads
Fig. 3.3c. Calculated pressure distributions (circular pads)

36 CHAPTER 3. CALCULATI ONS AND EXPERIMENTS
Gas supply
r condition uni
Power amplifier
Fig. 3.4. Schematic illustration of experimental systems
4). there should be the following functions:

4a). to deliver the static and dynamic loads and to ind.icate them
4b). to indicate both static and dynamic d.isplacements of the test
bearing
Following these basic considerations, a meebankal structure was designed
[Snip, 1991]. The dynamic load was applied by a magnetic shaker while static
load was provided by a pneumatic cylinder which can also be used to produce
low frequency dynamic load. The dynamic load was measured by the strain gauge.
The displacement of a hearing body was measured by the conductive sensors.
Three displacement sensors were mounted with the same radius. Two of them,
:which were arranged in one line through the geometrical centre, were used as
the main sensors and the third one was put on a line perpendicular to the line
determined by the first two sensors. The average value of two main sensor
· outputs is the true vertical displacement, provided that the influence of the
pad eccentricity is smalt. The tilt of the bearing body can be calculated by
the difference between the outputs of the sensors. The relationships of the
true working gap height, hw and the tilt angles, e, TJ with the sensor outputs,
h1, hz, h3, can be expressed in the following matrix :
Design of Gas Hearing Systems for Precision Applications 37
[el =rr [
1 -1
'l -1 (3.14)
hw r r-ex-ez
where
r : the radius determined by the three sensors
ex,ez : the eccentricities between the circle determined by the sensors
The displacements were measured by the inductive sensors and the force
was measured by the strain gauge. The specifications of the sensors and the
instruments are listed in Table 3.1.
Table 3.1. Specifications of the instruments
sensor/transducer instruments resolution range drift

W1/S inductive KWS S03A 0.8 mVN 0 -1000Hz tem. 0.05%
sensor(HBM) (0.02 !J.m) (-1db)
03-type load een KSO HBM 0-1000 Hz tem. 0.05%
4338 accelerometer 5007 charge 98.7 pc/g

(Brüel & Kjrer) amplifier
5850 flow sensor Model 5878 0.03 nl/min 0 - 30 ni/min.
(Brooks)
Pt-100 (Heraeus) Home-made 0.01 °C
vso mk1 1 KW power 0 -120 N

Vibrator amplifier
LabMaster D/A 8-bits digit I/0 for temperature sensors
interface card, 12 bits AID converters for the others
The dynamic properties of the mechanica! structure were tested. A noise

signal was fed to the shaker. The frequency response in the specified points
was measured. The typical frequency response at measured points is shown in
Fig. 3.5. It can be seen that the natura! frequency of the loading loop is
well above the 1 kHz, which meets the design requirements.
3.3.2. The instrumentation

The standard software, ASYSTANT™ is used to acquire and analyze dynamic
test data. The necessary interfacing programs for temperature and static
property measurement were also developed. The analog outputs from all
instruments were converted into digital sets by the Labmaster interface card.
The specifications of all the instruments are given in Table 3.1.
·---~~------------------·---~--------~ -
300.00
0.0
0.0 HZ 1.6000 K
Fig. 3.5. Vibration mode of loading loop of the experimental system
Fig. 3.6. The test set-up for performing wal/ temperature experiments
3.3.3. The test set-up for wall temperature

Fig. 3.6. shows the test set-up for performing experiments on the hearing
wall temperature. Four flat Pt-100 temperature sensors (No. 4 ~ 7) are
distributed in a radial direction of the standard block, one (No. 3) is put on
the inside wall of the gas supply channel. Two other sensors (No. 1 and No. 2)
are used to determine the local environrnental temperature.
3.4. TRI-CONICAL GAP SHAPED EPG HEARINGS

The EPG pads with a centre inlet restrictor can avoid the problem of dust
or other harmful matter trapping in the hearings. Also, they are easy to make
and relatively cheap. That is why they are favorable in most applications.
Furthermore, the research has shown that the load capacity and the stiffness
can be much improved by designing a suitable gap shape [Wang, 1990]. lt was
found that tri-conical gap shaped pads have better overall characteristics
than the other gap shaped pads. Under the same working conditions (working gap
height, supply pressure etc.), the tri-conical pads provide high load
capacity, high stiffness and high damping stiffness. The comparison of load
capacity of different gap shaped EPG pads is given in Fig. 3.11, from which
P= 0.6 MPa, conical gap P= 0.6 MPa, conical gap

0.38 ,------,----,----,--------. 0.22 ,------,----.----.-------,
hw=S,1.1m hw=5J.Im
0.36 0.2 , ... ,.... . . «" ..... - - - - ......
.... ..,
.,,.#"""'
0.34
0.18
,,,''
~ 0.32
~ 0.16
'
]
j
0.3
~"' 0.14
''
''
j
-~ 0.28 '
-~ 0.12 '
~ ~
~ 0.26
~
hw=IO llm 0.1
0.24
0.22 0.08
0.2 L - - - : - ' - - - - : - ' - - - - ' - - - 0.06 L..__ __,__ __ , __ __.___ __J
0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1

diameter ratio (DC/DD) diameter ratio (DC/DD)
Fig. 3. 7. The load and stiffness of bi-conical gap shaped EPG hearings
in terms of diameter ratio (hv= 5 J.Vll)
the advantage of tri-conical gap shaped pads can be seen easily. The
tri-conical gap shaped pad was developed by the author based on the study of
the relationship between the load/stiffness and the gap difference and between
the load/stiffness and the diameter ratio of conical and bi-conical gap shaped
pads [Wang and Schellekeus 1991]. The dimensionless load is defined as the
load capacity divided by the absolute value of the product of the supply
pressure and the hearing area, i.e. WW = W/(Po A). The dimensionless stiffness
is defined as, SS = S hw/(Po A).
Fig. 3.7 presents the load and stiffness changes with respect to diameter
ratios. It can be seen that when the working gap height is less than 10 f!m,
the maximum load and maximum stiffness correspond to nearly the same diameter
ratio (DC/DD = 0.8 - 0.9). lt is also clear that the optimal diameter ratio is
independent of the pad diameter and the supply pressure. Fig. 3.8 shows how
load capacity and stiffness vary with gap difference. It can be seen that the
load capacity increases with the gap difference [Wang, 1990]. However, the
selection of the gap difference is limited by the occurrence of the "pneumatic
hammer" (Fig. 3.9), a self-excitation due to the unsuitable combination of
hearing parameters. It will be seen in Chapter 4 that the stability of gas
system can be determinèd by the hearing film characteristics.
bi-conical pads, P=0.6 MPa bi-conical pads, P=0.6 MPa
0.32.--------.-------, 0.16,.---------,------,
0.3 hw=5Jtm 0.14
?a
~ 0.12 =10 urn
:! j""
~ 0.26 .... "' 0.1
I
;.,,.,
.9 ".,., .... ;
~
~ 0.24 //_./'hw=lO urn ~ 0.08
/
0.22
~/
0.2 ___J
0.04 ' - - - - - - - - ' - - - - - - - - l
' - i_ _ _ _ _. . . . __ _ _ _
5 10 15 5 10 15
gap difference (Jml) gap difference (/.lm)
Fig. 3.8. The load capacity and the stiffness of conical pads
in terms of gap difference
Fig. 3.9 gives the stability criterion in terrns of diameter ratio, which
reveals the gap shape limits for avoiding the self-excitation. From Fig. 3.8,
one can see that to ensure the same stabie operation there is still room for
gap shape in the inner part to be modified, but not in the outer part. This
implies that an additional conical part will increase load capacity and ensure
the stability at the same time, which leads us to the concept of the
tri-conical gap shaped EPG pads.
ISr----.-----.----~-----.----~----.-----.----.
>:
6
8 Unstable
:a~0.. 5 Stabie
~
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
diameterratio (DC/DD)
Fig. 3.9. The stability criterion of conical pads in tenns of diameter
ratio and gap difference
P = 0.6 MPa, DD= 60 mm P = 0.6 MPa, DD= 60 mm
650r------.------~------~ 60r------.------~------~
\
55 '
50--, \.0.8
45
g 450
l 400
40
1 350
300
250
200
0.8
150 5L-----~10~-----1~5------~20 15 5~----~10~-----1~5~----~20
working gap (j.tm) working gap (pm)

Fig. 3.10. The load and stiffness of tri-conical pads in tenns of
(outer) diameter ratio
Fig. 3.10 shows the effect of the outer diameter ratio of tri-conical gap
shaped pads. lt can be seen that the optimized outer diameter ratio (DC2/DD)
bas the same value as that for conical gap shaped pads.
Fig. 3.11 presents the comparison of the load capacity and stiffness
among parallel shaped, simple, bi- and tri-conical shaped pads, which shows
that tri-conical pads have the highest load capacity and the highest stiffness
900.-----------.---------~-----------.-----------,
FEM calculation:
-- coi}Ïqtl g_ap.sha~ DC/DD=l.O
.. opturu~ bt-corucal gap shape
-· tn-comcal g~p sha~
- optimized fri-conical gap shape
700
Parameters:
diameter: DD=60 mm
g 600 pressure: P=O.S MPa
.€
!.. 500
!3
Working gap height (J.1m)
g
>. 500
l 400
~
6 7 8 9 10 11 12 13 14 15
working gap height (J.Im)
Fig. 3.11. The comparison of different gap shaped EPG pads

over the whole working gap range. The load capacity of the optimized
tri-conical EPG pads at 5 1-1m working gap is 50 % higher than that of the
bi-conical gap shaped EPG pads. The comparison of the experimental results
with the FEM calculation for a tri-conical pad is given in a separate figure
shown in the same figure. The uncertainty of load measurements is ± 1.2 N and
the uncertainty of displacement measurements is ± 0.05 j.lill.
Fig. 3.12 shows the complex stiffness magnitude of those pads. The value
at zero frequency is the static stiffness. In the range of higher frequencies,
the stiffness magnitude approaches to a constant, i.e. the stiffness at
infinitive frequency (K,). This is because of the slow response of the gas
DD= 60 mm, hw=S m. P=0.6 MPa d=O.S mm

150
z~E
"'i!l
~
"'
~
50 100 150 200 250 300 350 400
frequency (Hz)
Fig. 3.12. The comparison of the magnitude of the complex stiffness
25 DD= 60 mm hw=S m P=0.6 MPa d:Q.S mm

,ê. .. para e
z 20 -- bi-conical
~ - tri-conica1
"'"'
~'lii
bi)
c
·a.
E
~
50 100 150 200 250 300 350 400
frequency (Hz)
Fig. 3.13. The comparison of the magnitude of the damping stiffness
supply change, which implies that neither the restcictors nor the film squeeze
effect are functioning in the high frequency range. From this figure, it can
also be seen that K00 increases from the parallel gap shaped pad to. the
tri-conical gap shaped pad. In fact, K00 is the average bulk stiffness of gas
in the bearing. From Fig. 3.11, we understood that the load capacity at the
same working gap height increases from the parallel gap shaped pad to the
tri-conical gap shaped pad. Therefore, the average bulk stiffness should
follow the same pattem
Fig. 3.13 presents the frequency characteristics of the imaginary part of
the stiffness which was defined as the damping stiffness [Wang, 1990]. This
figure shows the frequency characteristics of damping stiffness. If a system
can be so designed that its natural frequency is in the range between 50 to
150 Hz, the hearing damping property is optimally used. Therefore, the dynamic
performance of the system can be enhanced.
3.5. THE EFFECT OF BRARING BODY TILT

In practical applications, the hearing body can be tilted for two
reasons: 1). The presence of an unequal gap height contribution due to errors
of manufacturing and assembly and/or distoetion of the structure; 2). Imposed
momenturn resulting from unsymmetrical loading and/or thermal loading. In
general, the tilt can change overall characteristics of hearings, such as load
and stiffness; and positions of slide or rotor in hearing systems. In the
structures with EPG pads, only the first aspect is important because each pad
functions as a "point" to restriet the degree of freedom of a slide. In these
situations, the design task is to discover an EPG pad with less sensitivity to
tilt and less momenturn to the structure.
Research [Wang and Schellekens, 1992] found that different types of EPG
hearings may have different sensitivities to tilt. Therefore, it is necessary
to consicter the hearing body tilt in the design of EPG hearings in order to
provide more reliable design data or to design a hearing with reduced or
without effects from tilt. Pande [Pande, 1988] conducted a theoretica}
analysis of static performances of EPG pads under the conditions of tilt and
rotation. He only reported the tilting effect of simple conical gap shaped
pads with a small ratio of the pad diameter to the inlet hole diameter.
Because the order of the tilting angles 11 and 8 is less than 10'3 rad.,
the effects on the reierenee coordinate system are of the same order and can
be neglected. Furthermore, the typical parameters used in practice are chosen
in the study, i.e.: the operational parameters were chosen as: 0.3 and 0.6 MPa
supply pressures (absolute), 5 and 10 t-tm working gap heights and 60, 80 mm pad
diameters. The conicities of all the conical gap shaped pads were optimized
for achieving better overall performance.
Fig. 3.14 schematically shows the configuration of a tri-conical gap
shaped pad with a centre resttictor and a parallel gap shaped pad with four
restrictors. In the text, the resttictor information will be omitted and EPG
pads will be referred to tri-conical gap shaped and parallel gap shaped pads.
Tri-coni<~al
gap shaped pad Parallel gap shaped pad
with center restrictor with four restrictors
Fig. 3.14. Schematic illustration of EPG pad configurations
3.5.1 Higher supply pressure and lower working gap height

Fig. 3.15 shows the effects of hearing body tilt on the load capacity and
stiffness of circular EPG pads at 0.6 MPa (absolute) supply pressure and 5 IJ.m
working gap height. The load capacity slightly increases with the tilt for all
the pads. The parallel gap shaped pads have the higher magnitude of change.
The influences on the stiffness are different, the stiffness of tri- and bi-
conical gap shaped pads slightly decreases with tilt ratio while the stiffness
of parallel gap shaped pads increase with tilt.
The impravement of the load capacity under tilt is due to the fact that
the load enhancement due to decreasing the pressure gradient in one side is
more than the load reduction due to increasing the pressure gradient on the
other side. As a result, the hearing load capacity is enlarged. The pressure
disttibution of a tri-conical gap shaped pad is given in Fig. 3.16, from which
one can get an idea what happens under tilt.
P = 0.6 MPa, hw = 5 Jlm P = 0.6 MPa, hw = 5 Jlm

0.38 0.18
tri-conical tri-comca
0.16 --------~-----------~ ---- ... -------
bi-conical
0.36 -----•----w-----------------~----~
bi-conical 0.14
fa
~ 0.34
IS
0.12
] :Ë
·g 0.1
j 0.32 gj
..9:!
5 0.08
-~
.§
§ 0.06
.§ 0.3
'"0 .Ei
'"0
0.04
0.28
......~':lll.l~l.... ·····•·"'·"''"'''''
0.260
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
tilt ratio (ht/hw) tilt ratio (ht/hw)
Fig. 3.15. Tilting' effects on overall characteristics of circular EPG
hearings with a center inlet restrictor (p = 0.6 MPa, hw= 5 !Wf)
Fig. 3.16. Pressure distribution of circular EPG pads under tilt

Design of Gas Hearing Systems for Precision Applieations 47
3.5.2. Higher supply pressure and higher working gap height
Fig 3.17 presents the effects at 0.6 MPa supply pressure and 10 ~m
working gap height. The load capacity of the hearing pads with tri- and
bi-conical gap shapes in these conditions rapidly increases with tilt, while
the load capacity of the pads with parallel gap shape decreases rapidly with
tilt. The stiffness of all the pads increases with tilt, but the stiffness of
the pad with parallel gap shape changes very rapidly.
In comparison with Fig. 3.15, one may note that the perfonnance becomes
more complicated. From these two examples, one can see that there is nothing
in common between the two different working gap heights. From the extensive
study, it was found that the dimensionless overall characteristics of the
supply and hearing dimension pressure show a similar trend.
A condusion can be drawn that the normalized overall characteristics by
the pressure or hearing dimension can be used in qualitative study only.
However, the overall characteristics nonnalized in one working gap teil us
nothing in the other.
P = 0.6MPa, hw = 10J.lm P = 0.6 MPa, hw = 10 pm

0.25 · .. 0.22
"•· ........ bi-corlical
...
• 0.2 -----------------------*--
·· .. _parallel tri-conical
0.245
0.18
tri-conical
~
~ 0.24 ~ 0.16
.",
;tj
.9 t;
"' 0.14
fB ""
l!
] 0.235 6
~ ·~ 0.12
~ ... "" .. ""
,,:-:; .§
-·- .",
".
0.1
0.23 .,. ______ ... __ ...
bi-conical ~
--.... -
parallel
0.08
0.225 0.06
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Fig. 3.17. Tilting effects on overall characteristics of circular EPG

hearings with a center inlet restrictor (p = 0.6 Pa, hw= JO !Jin)
p 0.3 MPa. hw = 10 ,1.1m P = 0.3 MPa, hw = 10 ,1.1m

0.096 0.07
'
0.095 ''
''
' 0.06
'' tri -conical
'
0.094 / ' ''
/ bi-conical
bi-conical / ' 0.05
~ 0.093 '' fa
~
"""
~
0.04
~ 0.092
~
] 0.091 ~
.2
0.03
5
.§
"""
0.09
parallel
j 0.02-
0.089 parallel
0.01
0.088
0.087
0 0.2 0.4 0.6 0.8 0.4 0.6 0.8
tilt ratio (hf/hw) tilt ratio (hf/hw)

hearings with a center inlet restrictor (p = 0.3 MPa, hw= JO j..llll)
3.5.3. Lower supply pressure and higher working gap height

Fig 3.18 gives the results at 0.3 MPa supply pressure and 10 1-lm working
gap height The load capacity of all the EPG hearing pads varies considerably
with the tilt. The load capacity of bi- and tri- conical gap shaped pads
increases with tilt, while that of parallel gap shaped pads decreases with
tilt. The stiffness of all the pads increases with tilt. On the contrary, the
tri- and bi-conical gap shaped pads are much less influenced by tilt. The
direction change of load capacity for different gap shaped pads further
reminds us of the complexity of the performances.
3.5.4. Lower supply pressure and lower working gap height

The tilting effects under these conditions are depicted in Fig. 3.19. The
results are similar to that at 0.6 MPa supply pressure and 5 !-lffi gap height.
From these examples, one can conclude that bi- and tri- conical gap
shaped EPG pads are much less influenced by the hearing body tilt.
P = 0.3 MPa, hw = 5 )Jm P = 0.3 MPa, hw = 5 )Jrn

0.14 0.07
tri -conical
tri -conical
0.135
0.06
0.13 bi-conical
bi-conical
~
0.05
0.125 fa
~
~
0.12
0.115
J..
'&I
0.04
"'
1tl
:§
0.11
j 0.03
5
~ 0.105 J 0.02
0.1
0.01
0.095
parallel
o·
0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
ti1t ratio (ht/hw) tilt ratio (ht/hw}

hearings with a centerinlet restrictor (p = 0.3 MPa, kw= 5 ~
P = 0.6 MPa, dim. 60 x 40 P = 0.6 MPa, dim. 60 x 40

0.375 r------..-----,----..---, 0.12 ,.---~--.....----.....---~
0.1 .................... -
..
-·
~
hw= 10 pm •••• •••
0.08 ,_ ....... ;"""", ....
ra
!§
ra o.355 : 0.06
~
hw=5Jlm
:§
.I
""
0.35 -~
~
0.04
0.345
0.02
0.34
---- hw=
-- ---
....
10
.... _
Jlm
0.335oc__ __._ ___.__;;,;;,~_,_=.::___j OL--~~-~--~---e
0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
tilt ratio (ht/hw} tilt ratio (ht/hw)
Fig. 3.20. Tilting effects on overall characteristics of reetangu/ar

EPG hearings with jour in/et restrictors
3.5.5. Reetangwar pads
Fig. 3.20 shows the tilting effects on rectangular pads, from which it
can be seen that the effects on the load capacity and stiffness at higher
working gap height are similar to those of the circular pads with parallel gap
shape. However, at lower working gap height, the behaviour is different, both
load capacity and stiffness increase.
3.5.6. Effects of dimensions
The dimensionless load and stiffness of two circular EPG pads with
different diameters are shown in Fig. 3.21. The diameter ratio and the gap
difference ratio (the gap difference over working gap height) are exactly the
same in the two pads. From the graphs, it can be seen that there is not much
difference between them. It suggests that the supply pressure and the pad area
can be easily normalized because they have the simple relationships with the
Ioad capacity and stiffness.
-DD=60mm, DD=80mm - DD-60 mm, -- DD•80 mm

0.4 0.18
hw=5,10~m
hw• 5JJm
0.38
- 0.16
0.14
0.36 Û)
~
!:!3- 0.12
"0
0.34
j
~ !El 0.1
"'~ "'
"'"'
<l)
]
<I)
0.32 '"2
.Q
0.08
c:
<l) "'c:
<l)
.5
"0 .§ 0.06
0.3 "0
0.04
0.28
0.02
hw= l01.1m
0.26 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Fig. 3.21. The tilt effects on EPG pads with different dimensions
186
184 - experiments with uncertainty
o calculations
~ 182
"0
.§ 180
0
178
176
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75
tilt ratio (ht/hw)
Fig. 3.22a. Expertmental verifwation of tilt effects on parallel pads
164 circular d with tri-conical a

- experiments with uncertainty
162 o calculations
~
"0 160
.§
158
156
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
tilt ratio (ht/hw)
Fig. 3.22b. Experimental verification of tilt effects on tri-conical pads
3.5.7. Experimental verifications

Fig. 3.22 presents the comparison of load capacity between the
calculation and the experiment on a tri-conical gap shaped pad and a parallel
gap shaped pad at the supply pressure 0.3 MPa. lt can be seen that the
numerical and experimental results are agreed fairly well. This proves that
the results of the numerical analysis are correct.
3.5.8. Tilt effects on the dynamic performance

Fig. 3.23 shows the tilt effects on the damping stiffness. The damping
stiffness of bi-/tri-conical gap shaped pads increases under tilt. However,
the damping stiffness of the parallel gap shaped pad decreases after about 40
Hz exciting frequency. The increasing rate of the damping stiffness of
bi-conical gap shaped pad is from 37 % in high frequency to 62 % in lower
frequency. The example was calculated at the tilt ratio 0.8, ht/hw. This
infonnation further suggests that the bi-/tri-conical gap shaped pads are not
sensitive to the tilt. Their overall characteristics under tilt are even
improved.
20r-----.-D==D-=~6~0rmm~~h~w_=~5~pm~P~=~0~.6~M~Pa~d~=0~.~5~mm~.~h~Vh~w~=~0~.8~----~
15
------ ....... ___ _

bi-conical
-- .. __ .... ___ ..
tri-conical
-5~--~--~=---~--~~--~----~--~--~
0 50 100 150 200 250 300 350 400
frequncy (Hz)
Fig. 3.23. Tilt effects on dynamic performance of EPG pads
3.6. EFFECTS OF SURFACE IMPERFECTIONS

Engineering surfaces are generally wavy and rough. In addition, a
mechanical body may be distorted under load or tempersture change. As a
result, the surfaces fonn a non-ideal shape for the hearing gap. Fig. 3.24a
shows waviness of a large granite beam, typically used in CMMs. It can be seen
that waviness can have a big influence on the EPG pad behaviour. The most
important point is that overall characteristics of the hearing on such a
surface will change from place to place.
MtHtMUM ZONF. F'LATMF.S$ <ttF.IGMT.,UM)
. LOCATN.. • . S"-P tUZF. Ut X -tee.. I MM. CRAOE <BS 817• IOIMD .. 8.

. EffVtJaO., , . 5TW S!lF: IN Y •1811. 8 ft4M. ~·---- t NOMINAL CRtO.
. DATE • e-fto-80
Fig. 3.24a. A measurement example of granite surface used in CMMs

Design of Gas Hearing Systems for Precision App6cations 53
Fig. 3.24b presents the micro view of typical granite surfaces. At first
glance, it may be doubted that the permeability would play a significant role.
lndeed, the experiments showed that the permeability of this kind of granite
can be neglected, although the porosity may cause problems due to the local
gas trapping.
Fig. 3.24b. Micro view of typical granite surfaces (15 .:1;)
3.6.1. Effects of waviness

As already noted [Wang, 1990], the hearing perfonnance is more sensitive
to the phase angle shift of the surface wave fonn. For practical purposes,
this implies that the EPG hearing behaves differently in different positions.
It is necessary to specify the surface wave shape when designing a gas hearing
system with high accuracy. Fig. 3.25 shows the changes of load capacity and
stiffness with the phase of a surface wave. lt can be seen that in the worst
case (amplitude is equal to 3 ~J.m), the change of load capacity is up 12
percent and stiffness up to 250 percent for the parallel gap shaped pad, while
10 percent in the load capacity and 13 percent in the stiffness for
tri-conical gap shaped pad. The reason is due to the gap shape change with the
pad position. When the phase in the negative 90 degrees, the hearing gap shape
contributed from guide beam is concave, while when the phase in the positive
90 degrees, it is convex. Therefore, the load capacity and stiffness are much
reduced in the later case.
oorallel
0 •3 r----,,.---a-"'m=p"i)li;:;.tu""de-:--:--,-----, 0.05
0.28
--lum
,·····-., - 3 um
Ql
~ 0.26 ~--·--~~~~~--- !a
1 0.24
~------------. ~
'........ ,. ''
J
'i!
0
0.22 ..................... "'

02~--~----~----~--~ -0.05
-200 -100 0 100 200 -200 -100 0 100 200
phase (deg.) phase (deg.)
0.4 bi-conical 0.18 bi-conical

,....................
~ ~
0.38
0.16
~
~ 0.340.36 .._
ij 0.14 ' '' ''
''
:§
amplitude:
--lum
\
\ !S'liS .' ' ''
',
-3pm \'" ,/ 0.12 ............______ ,,'
0.32 ... Opm .. \.,..,
......... ../
0.3 0.1
-200 -100 0 100 200 -200 -100 0 100 200
tri -conical
0.42 0.18 . ' ' tri-conical
........ ,'
0.4 . -- . ''
···--···••t•••
~ ~
..
'
~
0.17
0.38
~
'
..
"0
<U
.9
0.36
0.34
amplitude:
--lum
-3um
.. Oum
''
-........ - ""
.. -
.s
·.:: 0.16
"' ''
''
.. '
'
.
.. '
'
'' ... ..... ' .........

0.32 0.15 ''
-200 -100 0 100 200 -200 -100 0 100 200
Fig. 3.25. The change of load capacity and stiffness of EPG hearings
with phase of surfaces wave (DD=60 mm, hw= 5 f1111, P= 0.6MPa)
Fig. 3.26 shows the effects of the wavelength of the wave form. It can be
seen that if the ratio of the half wave length over the pad diameter is
greater than 1.5, the influence on the bi-/tri-conical pads is small,
especially on the stiffness. On the other hand, the influences on the load and
stiffness of parallel gap shaped pads increase with decreasing of wave length.
Design of Gas Bearing Systems for Precision Appllcations 55
0.3 rallel rallel

wavelength ratio: 0.1
0.28 0.5
-1.0
~
~
•• 1.5 0.05
0.26
~
] 0.24 ~
'iJ 0
0.22
0.2 -0.05
-200 -100 0 100 200 -200 -100 0 100 200
0.45 .------,..--...!b::.!.i-con~~i~ca:!!l_-,.---____, 0.2 bi-conical
-,
___wavelength ratio: 0.18
~ 0.4 _,- --, -- 0.5
~ '
~
~
0.16
] 0.35 j 0.14
.......... ,//'
"' 0.12 ...................
0.3 0.1
-200 -100 0 100 200 -200 -100 0 100 200
tri-conical 0.22 tri-conical

0.42
_...............
m 0.2 ' '
'
~"g
fa
fa 0.18
''
''
''
.
~
wavelength
0.36 ratio: ~
.9
--0.5
0.34 -1.0
·..:::
"' 0.16 . ''
' ,. ____ .. ,
.. 1.5 0.14
0.32
-200 -100 0 100 200 -200 -100 0 100 200
Fig. 3.26. The change of load capacity and stijfness of EPG hearings
with wave-length ofsurfaces(DD=60 mm, hw= 5 !JIIt, P= 0.6MPa)
Fig. 3.27 shows the change of the damping stiffness in two extreme phase
conditions (+90~ top figure, and -90~ bottorn figure) when the pads operate on
the wavy surface. The figure only shows the change of the damping stiffness
relative to the ideal condition. In the example, the single sine wave was
used. The half wave length is taken the same as the pad diameter (60 mm) with
1 fJ-ffi amplitude. It can be seen that the change is insignificant. The values of
the damping stiffness in this situation can be determined by actding the

damping stiffness operating on smooth surfaces, given in Fig. 3.13.
0.4
- tri-conical
-- bi-conical
:@ 0.3 -.parallel
~ 0.2
~
~ 0.1 -........... ---- ---- ....... -.. -- .. --- .... -------------------------------
50 100 150 200 250 300 350 400

frequency (Hz)
] ----------------------------------------------------
z
6
~
~ -15
--- - tri-conical
-- bi-conical
-.parallel
-20
0 50 100 150 200 250 300 350 400
frequency (Hz)
Fig. 3.27. The change of damping stiffness of EPG hearings

with surface waves (DD=60 mm, hw= 5 ~ P= 0.6MPa)
3.6.2. Effects of roughness

The importance of the study of the effects of roughness is to select
suitable roughness parameters. In lubrication problems, the exact meaning of
roughness and waviness may be different from the definitions in the standard
of the surface characterization. The flow average model was successfully used
in the oil film lubrication [Patir and Cheng, 1984]. but there is still
disagreement in the application of gas lubrication [Raad and White, 1989]
[Tooer, 1984]. In this study, we limit ourselves in the roughness (local
waviness) in the condition that the Reynolds equation is still valid, i.e. so
called "Reynolds roughness". This means that the local dimensional ratio, an
amplitude of a wave over its half wave length, should have the same order as
that we used in the denvation of the Reynolds equation. In this treatment, we
Design of Gas Rearing Systems for Precision Applications 57
can simply use the geometrical average of gap height. The derivation of local
gap height from the ideal surface is averaged in an element. An alternative
flow averaged model was also proposed [Wang, 1990] to deal with more
complicated surface profile.
fh=0,8mm
~20000
Fig. 3.28a. A surface profile of EPG pads
Fig. 3.28b. FEM averaged surface pattem of a typical surface
A grinding surface of an EPG pad is given in Fig. 3.28a. When the surface
pattem is introduced to the FEM formulation, the surface profile is locally
averaged in an element, shown in Fig. 3.28b. The derivations of load capacity
and stiffness from the ideal conditions are shown in Fig. 3.29. It can be seen
that both load capacity and stiffness show a significant change.
3.7. WALL TEMPBRATURE

In genera!, the hearing wall temperature depends on the heat equilibrium
in the hearing walls, or the heat exchange between the gas film and the walls
and the heat exchange between the walls and the environment. The gas film in
most applications acts as heat sink sourees due to the cooling effects or a
cooler gas supply, or both.
P = 0.6 MPa, DD= 60 mm P = 0.6 MPa, DD• 60 mm

120,-----,-----~----~ 0.-----~------~----~
Lw=l mm Lw=l mm
Aw=lJJm Aw=lpm
100 - tri-oonical -2
•• bi·conical
-.parallel
80
i
60
40
20 -JO
0 -12
-20 -14
-.parallel
-40 L - - - - ' - - - - - - ' - - - - - -16 L..__ _ _ __,__ _ __,__ ___.J
5 10 15 20 5 10 15 20
working gap (pm) working gap (J.Im)
Fig. 3.29. Load capacity and stiffness change on a rough surface
21.8
to tum on gas
21.l
21:---~--~----~--~---L----L---~---L----~~
0 lO 20 30 40 50 60 70 80 90 100
relative time
Fig. 3.30. The temperafure of supply gas in the Metrology Laboratory,

Eindhoven University of Technology
TR-60. P:Sbar
21.6 ,..~-... --..-.• ~-----.-------
~,.._..,~w;'.Jh~:o"!--"'--~~-~'llo.II'--,"Y,...,.f"t:.~l
-
[
· sensor 2
7
W . _ Ut S
sensor4 -
.,u
~
[
= 21
~e.nsor 5 &7
~ 20.8l
20.4o-----=-so:-:o----looo...._____ls. ._o_o____2.oo_o
. . ____2_jsoo
relative time 1 sec.
C0-60, P=S bar, Q=3.39 nl/min.
,,V
!:!:::1
sensor6
ë 19.9
8.
E
g
19.8
19.7L
19.6
0 200 1600
rciative time
Fig. 3.31. The measured temperature drop in the steel plate

in the opposite hearing pad
Fig. 3.30 shows the temperature measurement of the inlet air in the
Metrology Laboratory, TUE. The experimental set-up was given in Fig. 3.6. It
can be seen that the temperature of the supply gas (sensor 3) is lower than
the room temperature. As a result, the cooler gas is introduced into the
bearings, and also machine surface.
Additionally, the Joule-Thomson effect will occur in devices such as
restrictors, which means that the gas will be further cooled down through a
restrictor. Some experimental data on the three types of circular pads with
two different type restrietars are given in Fig. 3.31 and 3.32. From these
results, we can conclude that the gas temperature is indeed lowered through
restrictor and a gas film acts as a heat sink source. The recessed pads (the
bearing with orifice restrictor) cause a larger temperature drop. The bi- and
tri-conical gap shaped pads cause a less temperature drop in the outlet
restrictor because of the relatively fast gas speed in the region. Different
After turning on supply air, 0.5 MPa
0
u
_..... -.. ""
50 100 150 200 250 300

Time min.
Fig. 3.32. The measured temperafure drop in the aluminum plate

opposite a recessed hearing pad
Design of Gas Hearing Systems for Precision AppHcations 61
types of hearings (especially, restrictor types) will cause a different

tempersture drop.
3.8. EFFEC'I'S OF HEARING MOTION VELOCITY

From the Reynolds equation, it can be seen that the motion velocity will
influence the pressure distribution. The significanee of this influence
depends on the magnitude of the velocity. With increasing machine velocity, it
is necessary to evaluate the effect of motion velocity. Until now, there is no
such information available for EPG pads, particularly circular EPG pads, in
the linear motion.
From the numerical analysis, it is understood that the motion velocity
will not cause the problems for the machine with velocity less than one meter
per second which is considered as the highest velocity in the current
commercially available machines. Therefore, only a numerical study is
presented. In principle, the motion velocity influences both static and
P = 0.6 MPa, hw =5pm P = 0.6 MPa, hw =5 pm

0.2 r - - - - - - , , . - - - - - - - ,
tri -coni cal
~~-----... ··--...
0.15
0.34
~ __
è
~ 0.32
'•,,',,,
~
~ ',,
.§ 0.3 raUel paraHel
"'
ij
E
:.s 0.28
0.26 -0.05
0.24 -0.1 ' - - - - - - - " - - - - - _ _ )

0 5 10 0 5 10
velocity (m/s) velocity (m/s)
Fig. 3.33. Velocity effects on load capacity and stijfness of EPG pads
dynamic performance of an EPG hearing.

The velocity effects on load capacity and stiffness are shown in Fig. 3.
33, from which one can find that the bi- and tri-conical EPG pads are
relatively sensitive to velocity. The change is up to 15 % in load capacity
and 63 % in stiffness at a velocity of 10 meter per second. Fortunately, the
velocity range of currently used linear machines is less than 1 meter per
second. From the graphs, it can he seen that in this velocity range, the
influences are very small.
Note that the calculation is under the assumption that the hearing gap
height does not change when the hearing moves. If a hearing pad is so designed
that it can he fully self-aligned, i.e. gap height can freely change with
velocity, the influences will he the same as the effect of tilt. In this
situation, the bi- and tri-conical EPG pads will he less sensitive to the
velocity than the parallel gap shaped pads.
The damping stiffness at 1 meter per second is shown in Fig. 3.34. The
damping stiffness of bi-/tri-conical gap shaped pads greatly increases with
velocity, up to 43 %.
S.-----.-~D~D~=-6~0~m~m~·~hw~=5T=um~,P~=~0~.6~M~P~a~,d~=~Or.5~m~m~,~v,=~1~.0~m/~s~-----.
.. parallel
-- bi-conical
6 - tri -conical
50 100 150 200 250 300 350 400

frequncy (Hz)
Fig. 3.34. Velocity effects on damping stiffness of EPG pads
3.9. CONCLUSIONS
The developed software and experimental system can he effectively used in

the design, study and testing of the EPG hearings.
The numerical and experimental research further reveals that the EPG
hearings with concave gap shapes provide higher load capacity and higher
stiffness. The tri-conical gap shaped pads are superior to the parallel and
bi-conical gap shaped pads.
The hearing body tilt will greatly reduce the load capacity and stiffness
of the parallel gap shaped hearing, maximum change of the stiffness is up 60 %
at the optimal working gap height. On the other hand, the tri-conical gap
shaped pads are much less sensitive to tilt, e.g. the stiffness change is less
than 1 % at the optimal working gap height.
The tri-/bi-canical gap shaped pads are also less influenced by surface
imperfection, especially if the half wavelength is greater than 1.5 times pad
reference dimension (e.g. diameter for circular pads). The worst case is that
the half wave length is equal to the pad reference dimension.
In the velocity range of less than one meter per second, the load
capacity and stiffness rarely change. However, the relatively high velocity
will greatly reduce the hearing load capacity and stiffness.
The temperature of the gas supply is generally lower than the standard
room temperature, 0.5 - 1 °G. The gas through a restcictor can be further
caoled down by the Joule-Thomson effect. In very accurate applications, it may
cause significant distartion of the structure.
From the study in this chapter, a general guideline to designing
mechanica} systems with EPG pads beoomes clear. The bi/tri-conical EPG pads
are preferabie because they are less sensitive to tilt and surface
imperfection.
PART TWO
DESIGN OF MECHANICAL SYSTEMS WITH EPG HEARING PADS

CHAPTER 4.
MODELLING OF LINEAR GUlDES WITH EPG BEARINGS
4.1. INTRODUCTION
Linear guides with EPG hearings are basic elements in CMMs (Çoordinate
Measuring Machines) and some other machine tools. The geometrical errors due
to the finite stiffness of materials and thermal-elastic effects have been
extensively studied, e.g. [Soons and Schellekens, 1991]. In this work, we
concentrate on the role of EPG hearings in a linear guide. For cleamess, both
the guide beam and the slide are taken as the rigid body. In the previous
chapters, we have already stuclied the overall characteristics of EPG pads, but
the task still remains to investigate the influences of arrangements of EPG
pads on system hehaviour. Based on some typical linear guides, some important
design concepts will he discussed, which include the loop stiffness,
kinematica! mounting and driving, symmetrical arrangement and preloading
methods.
Furthermore, the appropriate models both for studying static and dynamic
hehavour of linear guides are developed. The homogeneous transformation, the
matrix form of coordinate transform, is used to descrihe the real gap height
of an EPG hearing in a structure. Through this model, the slide hehaviour, due
to the change of hearing working gap height, can he determined, which can he
used, for instance, to specify a guide surface quality for a desired accuracy.
In genera!, the stiffness of the gas film is frequency and displacement
dependent For vibrations with a small amplitude (e.g. 0.5 f.Lm), constant
stiffness at a specific frequency can he assumed. Therefore, the response of a
gas hearing system can he obtained by the loci of the response of a continuous
series of linear systems. At each frequency, the corresponding system is
represented by a linear differential equation with constant coefficients. For
a relatively high vibration amplitude, the response at a specific frequency
can he determined by characterizing the film stiffness. It was found that the
displacement dependenee of hearing film stiffness can he characterized by a
cubic function. The morlal distribution at a specific space coordinate can he
found by a transformation of the morlal from the general coordinate system into
the spatial coordinate system. The response in the time domain under the given
excitation can he determined by using the same transformation.
66 Design of Gas Bearing Systems for Preeision Applications
4.2. LINEAR GUlDES

Some possible forms of linear guides are summarized in Fig. 4.1, some of
which are the typical configurations in practice. Based on those we can reveal
some general principles to achleve better mechanica! systems. In geneml, the
basic principles used to design precision structures were summarized in the
ASPE tutorlal [C. Teague and C. Evans, 1990]. Some of those principles can be
directly applied in the design of structures with EPG hearings, while others
require more attention, because the EPG hearings have lower stiffness than the
solid cantacts and a small working gap height. Those principles include
kinematic mounting, loop stiffness, symmetry and kinematic driving. For EPG
hearing systems, there is another important aspect preloading.
General form Commonly used Vert.ical Side vacuum

Z guide in CMMs Weighl prelaad prelaad
Pad prelead Commonly used Alternalive Top surface

in two direction Y guide o! CMMs form for light as main guide
weight surface
Fig. 4.1. Illustration of some typkal configurations of linear guides

with EPG hearings (section view)
4.2.1. Kinematica! mounting
If a pad is rigidly mounted (no self-alignment), the maximum rotation
angle (until touch down to the guide heam) is limited by the pad dimension (1;)
and its working gap height (hw), i.e. :
CHAPTER 4. MODELLING AND DESIGNING 67
41 .s. hw/S (4.1)

The ratio 41 is approximately in the order 10- 4 (111 20 are. sec). In
practice, the allowable rotation angle will he much smaller than this, which
is actually defined by the allowable change of the hearing gap height, i.e. :
(4.2)
Under normal working conditions, it is within a few are seconds. If we

consider this value is within the error range, one can say that one pad will
he sufficient to restmin two rotations, particularly in the case of small
structures. In most applications, an EPG pad has a self-alignment mechanism.
If the self alignment mechanism does not work, the guide will he over
JX>Sitioned. Strictly speaking, if any error in JX>Sition of a pad is introduced
in a hearing system, the system will he oveC}X)Sitioned. Furthermore, it should
he noted that the rotation about the motion axis can he completely controlled
by two pads mounted on two perpendicular surfaces. This implies that if one
more pad in any surface is used, the slide will he OVerJX>Sitioned. In the
suppression of a translational degree of freedom, a pad limits only one
direction translation. In practice, pretoading is necessary to cestrain the
degree of freedom in another direction.
Summarizing the above discussion, we can conetude that a purely kinematic
mounting of a slide needs only two preloaded EPG pads (one with and one
without self-alignment) mounted on the other side of two perpendicular
surfaces. In practice, four preloaded EPG pads with effective self-alignments
are preferable, especially in a large linear guide. Two pads mounted on the
same side should he well-aligned to avoid ovef}X)sitioning in the rotation
about the motion axis.
Another imJX>rtant }X)int in design of EPG pad systems is that with a small
change of displacement it provides one elas6c restrain in the direction of
working gap height. This particular feature provides that there are less
critica! requirements on the kinematic mounting in the small change of slide
JX>Sition. 'If the number of pads used in a slide is more than the numhers for
purely kinematic mounting and the toleranee is within 10 f.im, all pads can
still work properly. On the other hand, the finite stiffness of EPG pads
provides an imperfect restrain, which determines the existence of error
motions of a slide. The hearing stiffness will he further reduced if the loop
stiffness is not carefully designed.
4.2.2. Loop stiffness

This is one of the important conceptual pattems for precision structure
design. It becomes more important in gas hearing systems because high film
stiffness is easily lost by poor design of joints, which unfortunately occurs
in applications, even in commercial products. In a linear guide system, all
the elements involved in a chain from the gas film to the moving slide form a
group of springs in series. In a linear guide, the loop elements include the
gas film, the pad body, the pad mounting, the guide beam, the slide frame and
the guide mounting. The stiffness loop is illustrated in Fig. 4.2. It is known
that the overall stiffness of a linear guide equals the reciprocal of the
reciprocal sum of all the element stiffness. In a mathematical form, it reads:
(4.3)
1. foundation
2. guide
3.4. conneetion
5. slide
6-9. springs
Fig. 4.2. Illustration of loop stiffness in a mechanica/ system with

EPG hearing pads
In general, the gas film stiffness is more difficult to achieve than the
stiffness of mechanical elements. Therefore, in the design, sufficient
stiffness of all the mechanical elements should be ensured, a study on this
subject was conducted to design a better mounting mechanism [Driessen, 1992]
and some good examples will be illustrated in the next chapter.
-------·-
CHAPTER 4. MODELLINO AND DFSIGNING 69
4.2.3. Symmetrical arrangement

Symmetrical arrangements are very important for systems with gas hearings
because gas hearings provide active farces to the support, for instance, the
guide heam and the slide frame in Fig. 4.1. A fully symmetrical arrangement
can provide the cancellation of distortions introduced by hearing farces.
However, it will become more difficult to utilize a pair of opposite pads in
large structures, like those in 3D CMMs. An ideal salution can be the
pretoading in the same side which can be realized by vacuum or magnet
preloading. In order to eliminate the troubles introduced by EPG hearings, it
is better to arrange all the hearings symmetrically whenever it is possible.
4.2.4. Kinematical driving

A kinematica! driving system requires a purely driven force (or a pure
moment for rotation systems) in the motion direction if possible. Driving
stiffness should be as high as possible, because weak driving stiffness will
shift the lowest vibration mode to the lower side of the frequency range. This
lowest natura! frequency will influence the vibration motion in the other
coordinate directions, nat only in the driving direction. Perforated hinge is
one of suitable elements in such applications: it provides high stiffness in
transverse direction and flexibility in other directions.
4.2.5. Pretoading
Pretoading can be used not only to restrain unwanted degrees of freedom
completely but also to set the optima! working gap height of EPG pads. In
genera!, pretoading can be provided by an opposite EPG pad, a vacuum pad or a
magnetic strip (force preloading) and structure weight (mass preloading).
Mass pretoading is a simpte way to provide preloading, but it can be only
used in a vertical direction (or vertical direction and one rotation in the
guide with eccentric mass). but the optimization of hearing characteristics is
limited by mass value. This methad is useful for relatively heavy structures.
In a light structure extra mass may be necessary, which will lower the lowest
natura! frequency of the system.
The opposite EPG pad pretoading requires the high accuracy, both in
flatness and in parallelism of the two opposite surfaces of the guide beam, in
the process of manufacturing and assembling of a structure. Due to technical
limits, it may become impossible to build a highly accurate and large
structure by using thls methocl. The benefit by doing so is the enhancement of

the total stiffness of the system. In thls situation, the total stiffness will
be the sum of the stiffness in the two sides. lt can also be used as an
actuator for realizing sub-micron displacement in the stationary situation. It
is very useful to realize small feeding in processing machines or small
displacement in measuring machines. With one side elastic mounting, the
critica! requirements on assembling can be relaxed. But the total stiffness
can only be about one side film stiffness.
Both magnet and vacuum preloading can be arranged in the same side as EPG
pads. In thls way, only one precise surface needs to be made, which leads to a
cheaper manufacturing and assembling process. There are plenty of examples of
light structures in which the magnet or vacuum is used as means of preloading.
For large structures, they can also be the better means of setting optima!
working gap height. The disadvantage of vacuum pretoading is the additional
vacuum device (or pump). The use of the Bernoulli effect [Wang, 1990] or
supersonic effect in the form of EPG pads could be a best solution. The
drawback of magnet pretoading is the special requirement on materials.
The calculation of the attraction force due to vacuum can be performed by
the similar procedure as that of EPG hearing force. The difference is that the
flow direction in the vacuum pad flow goes inwards whlle in EPG pads it is
outwards. The inward gas can be the outlet gas of EPG pads or using a
separation grooves to avoid unfiltered gas going into the hearing working
area. The calculation of the attraction force in a magnet can be performed by
using the material specification of magnets.
4.3. MODEL TO DESCRIBE THE REAL BEARING GAP HEIGHT

The real gap height for each hearing mounted on a slide can be derived by
using the homogeneaus transformation. The homogeneaus transformation has been
used by many researchers to establish a generalized model to describe the
geometrical errors of machines, e.g. [Teeuwsen et al 1990]. In order to derive
the model to describe the real gap height of an EPG pad in a mechanica!
system, two coordinate systems can be defined: a reference coordinate system,
let (X,Y,Z), connects to an unmovable guide beam and its centre is assigned in
the mass centre of the guide beam; the other coordinate system, 6cs
(X' ,Y' ,Z'), is fixed on the slide. The coordinate systems are shown in Fig.
4.3. Because the hearing gap height is the distance between the hearing pad
CHAPTER 4. MODELLINO AND DESIGNING 71
and the guide beam, the change of working gap height will be the relative
change of the coordinates 6cs in the pad mounting position with respect to the
reference coordinates les.
In this study, a rigid slide was used, which allows us to concentrata on
studying the coordinates in the mass centre and the attitude angles of the
slide. In a linear guide, the slide position is defined by all the EPG pads
mounted on it. Therefore, the mass centre coordinates and the attitude angles
of the slide can be described by all the hearing working gap heights. In
practice, the working gap height will change due to the change of the
effective lood capacity, e.g. lood capacity varlation on a rough surface. By
using the relationship between the working gap height and the slide
bebaviours, the slide motion errors can be determined. Moreover, for a desired
level of accuracy, the manufacturing and assembling errors of a slide for a
fixed type hearings can be defined.
In an arbitrary condition, the hearing position will be modified by the
following matrix :
Fig. 4.3. An example of a linear guide with EPG hearings with the
definition of coordinates and hearing gap height
72 Design of Gas Bearlng Systems for Precision Applieations
thw = [ ~ -r -~1 dzc

-~ a0
~yc] (4.4)
0 1
where dyc, dzc : mass centre changes of the slides
a, f3, 'Y : the attitude angles of the slides
lt should be noted that the translations parallel to a hearing surface

and the rotations about the axis perpendicular to this hearing surface have no
significanee in descrihing this hearing working gap height. Furthermore,
because there is no pad parallel to the Y-Z plane, the translational
derivations in the X- direction do not contribute to the hearing working gap
height. lt should a1so be noted that the attitude angles of slides are very
small, the trigonometrical function in the transformation matrix can be
replaced by their radiant angles directly. The products of angles (af3, ay, f3y,
af3y) are far smaller than the angles themselves so that the terros with the
angle products in the transformation matrix can also be ignored.
Using the transformation matrix, any coordinate on the slide coordinate
system 6cs can be transformed to the reference coordinate system ~es.
Therefore, the real working gap height can be determined by the difference
between two veetors in the hearing gap direction: the vector defining a pad
position, [ap;,bp;,cp;,1], and the reference vector on the guide
[ag;,bgi,Cgi,l] defining the initial working gap height. Therefore, the
working gap height can be calculated by the corresponding row in the
transformation matrix (a transformation vector), for instance, the second row
for the hearings mounted on the surface(s) perpendicular to the Y-axis and the
third row for the hearings perpendicular to the Z-axis, i.e. :
hwyi = [y, 1, -a, dyc] [ap;, bp;, cp;. 1]T- bg; (4.5)
hwyi = api y - Cpi a + dyc h~yi (4.5a)
hwzi = [-f3, a, 1, dzc] [api, bp;, Cp~ 1f- Cgi (4.6)
hwzi == -api f3 + bp; a + dzc + h~zi (4.6a)
where
a;, bi, c; the mounting position coordinates of pad in X,Y,Z
coordinate directions.
footnote
p. g : pad, guide, i : ith pad, x,y : coordinates
superscript
0 : initia! gap heights, h~zi == Cpi - Cgi and h~yi == bp; bgi
CHAPTER 4. MODELLING AND DFSIGNING 73
If all the pads in a linear guide are arranged fully symmetrically with
respect to the Y-Z plane, the change of the centre coordinates and the
attitude angles of the slide can be expressed by the working gap height :
dyc = (DJ.{\ry; - :Dlwy;)/n (4.7)
dzc = (.th{\rzi
- .thwzï)/n (4.8)
y = (.t.Mtw;x - .t.Mtwf,t)/2ay (4.9)
13 = (.t.Mtw!x - .t.Mtw~)/2az (4.10)
a = (.t.Mtw;z - .t.Mtwf.;:)/2cy = (.thw!y - .thw;y)/2bz (4.11)
or a = [(.t.Mtw;z + .t.Mtwf.;:)+(.t.Mtw~y + .t.Mtw;y)]/(2cz+2cy) (4.11a)
where
.ó.hwy; = h{\ry; - hWyi, and .ó.hwzi = b{\rzi - hWzi
superscript:
+ referring to the gap height corresponding to the pad in
the positive coordinate direction
referring to the gap height corresponding to the pad in
the negative coordinate direction
footnote: x,y,z, coordinates, e.g.
.thwY,. stands for the sum of all the gap heights in the Y
direction with the positive z coordinate
ay implies the coordinate X for the bearing mounted in the
surface perpendicular to Y axis
n number of pads
In Chapter 3, it was established that the overall characteristics of EPO

hearings vary with the phase shift of a surface wave. A gas hearing system has
a constant weight, or constant force in general, so that the real woricing gap
height will change to balance the slide weight (or forces acting on the
slide). As a result, the changes of the pad woricing gap height directly
contribute to the slide motion errors. It follows that the rotational motion
error will be reduced if the distance between two pads is not equal to the
half wave length of the surface wave. The use of those formulas to describe
the motion errors of a linear guide will be demonstrated in the next chapter
with specific examples.
4.4. IDEALIZATION OF GAS FILM STIFFNESS

Before going any further in modeHing the dynamic behaviour of a slide,
we need to idealize the gas film stiffness in order to couple EPG pads into
74 Design of Gas Bearlng Systems for Precision Applieations
structures in the usual manner.
4.4.1. The descrlption of displacement dependenee

The stiffness of EPG hearing films, in general, is not constant. However,
the vibration always starts from a small amplitude, this allows us to limit
our study to a small change of the hearing displacement, e.g. ± 0.5 f.tm, in
which the hearing stiffness ean be estimated to be a constant at an
equilibrium gap height. For the pad studied in this thesis, the stiffness
error due to non-linearity is less than 5 % and the error of the natural
frequency for the single degree of freedom will be 2.2 %. If necessary, more
accurate results ean be obtained by introducing the stiffness function from
the curve fitting into the equation of the motion of the bearing system. It
was found that film stiffness can be accurately described by a cubic function
of the hearing displacement, as illustrated in Fig. 4.4.
tri-conical pad paratlel pad

29
odata 28
- cubic fitting
DD=60mm 27
50
dd= O.Smm
P= 0.6MPa 26
:§: 45 :§:
:z: :z: 25
6 40
6
~
~- 35 ~
"'
24
odata
"' 23
- cubic fitting
30 DD= 60mm
22
dd= 0.5 mm
25 P= 0.6MPa
21
20 20
5 10 15 5 10 15
gap height (pm) gap height (pm)
Fig. 4.4. Illustration of the hearing stijfness characteristics

CHAPI'ER 4. MODELLING AND DESIGNING 75
4.4.2. The description of frequency dependenee

From the solution of the Reynolds equation under dynamic operations, we
often find the dynamic stiffness in the frequency domain. Therefore, it is
convenient to use the concept of the damping (loss) factor (the structural
damping factor) which was introduced to study the material damping, e.g.
[Newland, 1989] rather than the damping coefficient used to investigate the
viscous damping. The damping factor for the general resilient element is
defined as the ratio of the imaginary modulus over the real modulus. Following
this definition, we can defme the (equivalent) damping factor for gas film.
It is a ratio of the damping stiffness over the total stiffness which is the
sum of the real part of the dynamic stiffness and the static stiffness
(expression in page 86) :
TJ(ro) = lm[K(ro)]/Re[K(ro)] (4.12)
By using the damping factor, the total stiffness of an EPG pad reads
Tri-eonical gap shaped pad.

0.4 .------,----.-----.----.---...,.-.:__-.,----.----,
0.35
--realstiffness, tilt
.. damping factor, tilt
- real stiffness, normal
-. damping factor, normal
0.2
..... "" ........ --- ........
~ ...........
--- -----
.. /' ... ............................ ..
'
---- ..
50 100 150 200 250 300 350 400
frequency (Hz)
Fig. 4.5. The real stiffness and damping toss factor of tri-conical
gap shaped EPG pads (p = 0.6 MPa, kw = 5 ~)
KQro)= Re[K(ro)][l + j rJ(ro)] (4.13)
With the comparison of the damping coefficient, it can be seen that the
damping factor at resonance is . twice the damping coefficient in the viseaus
vibration model and the inverse of the quality factor.
Fig. 4.5 shows the distribution of the Re[K(ro)] and iJ(w) for lin EPG pad
witb tri-conical gap shape from which one can see that they are nonlinear
functions of, the frequency. Therefore, the response of a gas hearing system
can only he obta1ned by the loci of the response of a continuous series of
linear systems. We take this bearing as an example to illustrate the idea by
assuming this hearing support a 60 kg mass. At a specific frequency, the
damping factor and the total stiffness are constants. At this frequency, we
solve the linear differential equationsconstant coefficients. In this
with
example, we solve 20 equations and the loci of frequency response is given in
Fig. 4.6, from which it can he seen that the characteristics of tri-conical
gap shaped pads are improved under tilt of pad body and the vibration
amplitude decreases. This is due to the increase of the damping factor. The
role of the damping factor in the control of the vibration amplitude will he
discussed later by using the equation of the vibration motion.
5 xl0-5 tri-conical
.-. 4
~
.a 3
l
~
2
0
0 50 100 350 400
frequency (Hz)
Fig 4.6. The frequency response of one degree of freedom system

with an EPG hearing pad with tri-conical gap shape
In order to avoid the frequency-dependent term, the film stiffness can he

approximated by a proper combination of certain numbers of constant springs
and dasbpots [Wang, 1990]. In general, five elements, three spring elements
and two dashpots, are needed to model the overall characteristic of a gas
film. In the special case, e.g. for the absolutely stabie or absolutely
unstable hearings, three elements, two spring elements and one dashpot, are
sufficient to describe the characteristic of the gas film. The physical
meaning of the model can be described by the overall compressibility of gas
film, the reciprocal · of the mass content bulk modulus (K.) and the modulus
associated with the change of mass content. lt can be directly seen in the
three-element model, the representations of the element models are given in
Fig. 4.7.
Fig. 4. 7. Representation of hearing films stiffness in the combination

of constant spring elements and constant damping elements
The constants of equivalent hearing stiffness and damping dashpots in an

element model can be determined from the hearing characteristics, e.g. the
constants in the three element model are :
Kt = Re[K(w)] I00 => 00
= Koo (4.14)
Re[K(w)] !00 => 0
= Kt/(l+Kt/K2) =Ko
K2 = Ko/(1-Ko/K00) (4.15)
D=K2't 1 (4.16)
where 't : time constant
1
From the structural analysis point of view, a three-element model can be

78 Design of Gas Bearing Systems for Precision App1ications
seen as a mechanical system with the zero mass and one spring in the top and
one spring and one dashpot in the bottom. If this model is used in the
equation of the vibration motion, one more degree of treedom for a hearing pad
will be introduced. The benefit is that all stiffness in the system is
constant.
4.5. DYNAMIC MODEL

4.5.1. The equation of motion
In genera!, the equation of motion for the mass centre of a hearing
system reads :
11
M {X} ; {W} - {Fm} - {F} (4.17)
where W : vector of load capacity, both static and dynamic parts

W = Wo + (aW/ax) dx + (aW/at) (at/dx) dx (4.18)
in the frequency domain, it reads :
W = Wo + (ko + Re[Kdfjw)] + j Im[Kdfjw)J) dx (4.19)
W = Wo + (Re[Kfjw)] + j Im[Kyw)]) dx
Wo: vector of static load capacity
x : static coordinate vector, X : dynamic change coordinate vector,
dx: in the equation of motion for dynamics, only the relative
change with respect to the equilibrium point is involved,
so that dx = X
M : mass matrix; if we takes the mass centre as the reference,
the mass matrix is always a diagonal matrix
Kd: dynamic stiffness, K : total stiffness, ko: static stiffness
11
X : Acceleration vector, Fm: mass force vector

F : force vector, e.g. pretoading force, excitation force etc.
By consictering the initia! balance between the static forces and the
static moment, such as mass and/or other pretoading forces and the hearing
static load capacity, the equation of motion for an EPG hearing system
beoomes:
- M w2 {X} = (Re[K(w)] + j Im[K(w)]) {X} + {Fd} (4.20)
where Fd: dynamic force vector
By introducing the concept of the damping factor, the equation of motion

CHAPI'ER 4. MODELLINO AND DFSIGNING 79
becomes:
- M ro2 {X} = - Re[K(ro)](1 + j {TJ(ro)}) {X} + {Fd} (4.21)
If we use the element approach to model the film stiffness, the equation
of the motion for a gas hearing system can he written with constant
coefficients and with increased degrees of freedom. In the three-element
model, one additional degree of freedom will he introduced for a pad.
From the definition of the damping factor [Newland, 1979], it is
understood that the natural frequency is not influenced by the damping factor.
If we are interested in the (dynamic) natural frequency only, the equation of
the motion can he further reduced to the following form :
- M ro 2 {X} =- (Re[K(ro)]) {X} (4.22)
The components in each matrix or vector depend on the specific design.

The abstract forms of the equations of the vibmtion motion from (4.17) to
(4.22) are valid for all kinds of pretoading structures and they are also the
same whether or not the structures are symmetrical. The detailed treatment
will he illustrated with a specific example given in the next chapter.
4.5.2. The natural frequency

Before discussing the solution, it is necessary to devote some attention
to the natuml frequency. Due to the frequency dependenee of hearing film
stiffness, it is puzzling to select the stiffness value to delermine the
natuml frequency, especially for the more complicated hearing systems. In
order to use a single pammeter to indicate the stability criteria, the
dynamic natuml frequency [Wang and Schellekens, 1990a] was introduced.
The dynamic natuml frequency can he derived by considering an EPG
hearing as a servomechanism. The servo model of EPG hearings was put forward
by Wilcock [Wilcock, 1967] and used by many other researchers [Stowell, 1971]
[Haycock, 1976] [Plessers, 1985]. In this way, the system chamcteristics can
be expressed by means of the transfer function of load change to working gap
change. The block diagmm of tthe servo system for an EPG hearing system is
shown in Fig. 4.8. The frequency response of the hearing supported mass is the
2
main loop with tmnsfer function of Mljro) =- 1/Mro and the response of the
gas film is a feedback loop with a tmnsfer function Kljro) - the stiffness
chamcteristic of the EPG hearings. The gain factor of the tmnsfer function
80 Design of Gas Bearing Systems for Precision AppJications
for the feedback loop is the static stiffness, the real part of the transfer
function is the dynamic stiffness and the imaginary part is the damping
stiffness. The product of these two transfer functions reads :
/:,W 1 /:,h
+;=== ---
Mcu 2
K(jcu) r--
r--'
Fig. 4.8. The block diagram of an EPG hearing system
1
SQw) =-- 2
{Re[KQw)] + j Im[KQw)]} (4.23)
Mw
According to the Nyquist stability criteria, the stability of a system

with single degree of freedom requires that :
Im[SQw)] < 0, if Re[SQw)] = -1 (4.24)
In terms of hearing film stiffness and mass, it implies :

Im[KQWn)] > 0, if IRe[KQWn)]I/M~ = 1 (4.25)
This implies that a stabie hearing system should have positive damping
stiffness, which further illustrates the stability criteria used in Chapter 3.
From this expression, it can also he seen that the natura! frequency of a
hearing system is determined by a specific value of the total stiffness which
results from the frequency dependenee of stiffness characteristics. For
convenience, this value was defined as the dynamic natura! stiffness. It is
readily to see that this value is dependent of the mass. For a selected system
with a fixed mass value, the dynamic natura! stiffness can he detennined by
the trial-and-error metbod through the equation IRe[Kfjro)JI = Mro 2 • This
equation can only he satisfied at a specific frequency value, which is the
natural frequency.
If the transfer function of a hearing stiffness characteristics has the
cross over frequency (the plot goes through the real axis), the stability can
he expressed in an alternative way :
IRe[Kfjro)J IIMro! IIm•=> Im- > 1 if Im[Kfjro) = 0 (4.26)
or IRe[Kfjro)JI/Mro~ lrm-=> Im• < 1 if Im[Kfjroc] =0
For convenience, the term, (Re[K(ro )]!M)! 12 is defined as the dynamic

c
natural frequency. Then, the stability criteria will he stated as follows: If
the product plot goes through the real axis from the positive (negative) to
the negative (positive) half plane of the imaginary part and the dynamic
natura! frequency is greater (less) than the resonant frequency, the system is
stable. For a more complicated system, e.g. having more cross over
frequencies, the dynamic natura! frequency is that cross over frequency near
the resonant frequency.
For a system with a single degree of freedom, it is easy to see that Mro2
is an increasing function of ro so that the cross-over frequency can only he
produced by the transfer function of the hearing stiffness. In genera!, the
transfer function of an EPG hearing can he expressed by a function with two
poles and two zeros :
(4.27)
It can he verified that there exists one cross over frequency for the
transfer function with two poles and two zeros.
The time constants, \, 't 2 , 't and 't 4 can he determined from the hearing
3
stiffness characteristics (as illustrated in Fig. 4.9). However, it should he
noted that they are all working gap-dependent for a selected hearing under the
same supply pressure. If all the hearings in a structure do not operate in the
same working gap height, the total stiffness characteristics will beoome more
complicated, more cross over frequencies may he introduced, particularly in
the system with mixed hearing characteristics.
In special cases, i.e. absolutely stabie (1; > 1: ) or absolutely unstable,
1 2
h 1 < 't2 ), the transfer function is simplified as the function with one pole
and one zero, i.e. :
Kfjro) = K0 (1 +jm't )/(1 +jrot
1 2
) (4.28)
or Kfjro) = K0 [(l+ro 2 t t )+jro(1; -t )]/[(l+(t ro) 2 ] (4.29)
1 2 1 2 2
In this case, no cross over frequency is involved. The stabie hearing

system requires positive damping stiffness, i.e. 'tl > 't2 (see equation 4.29).
K(w
1/T1 1/'Te
Frequency (f)
Fig. 4.9. Illustration of log-log plot of an EPG hearing stiffness
4.5.3. Solution methods

In a system with large degrees of freedom, it is very difficult to get a
direct correlation between system parameters and natura! frequencies of the
system by using analytical formulas [Yang, 1989]. Since a lot of standard
software is available · to solve matrix problems, it is convenient to solve a
system with the large degrees of freedom numerically.
If we are only interested in the natural frequency, we can disregard the
damping stiffness of the EPG hearing. Therefore, the solution of the hearing
system at a specific frequency beoomes a normal eigenvalue problem, i.e. to
solve the following equation :
detHMr1[Re[K(ro)]-À[I]} =o (4.30)
In the three-element model, the mass matrix is singular due to the zero
mass elements are introduced, e.g. for M X M mass matrix with N zero mass.
The first thing to do is to reduce the mass matrix and reconstruct the damping
and stiffness matrices in order to get the eigenvalnes of the matrix, in other
words, the natural frequency of the system. The new matrix Mr, (M-N) x (M-N),
can be obtained by cutting off the rows and columns corresponding to the zeros
mass. New stiffness matrix Sr, (M-N) x M, can be built up simply by canceling
the rows with respect to the zero mass. However, the rows and columns in the
damping matrix corresponding to the zero mass in the damping matrix must be
made to zero if they are not zero and the new damping matrix Dr becomes (M-N)
x (M-N) matrix.
Thereafter, the problem becomes the normal vibration problem with
damping, which can be solved by the reduced order metbod [Newland, 1989]. A
new square matrix with dimension, (2M-N)x(2M-N) is then composed. In the OS
matrix, the rows (iis) corresponding to the zero mass diagonals, m(ii,ii),
have the new coefficients which are the ratio of the elements in the original
stiffness matrix, S(m, n), over the element, D(ii,ii), in the original damping
I
OS(19x19) liD (19X9)
I
A= (4.31)
I
M~ 1 Kr I M~ Dr
1
(9x19) I ( 9x9)
matrix for all the columns. In the ID matrix, there is the similar treatment,
except only the elements corresponding to the non-zero mass diagonal (ii) are
taken into account :
OS(m,n) Im(ii,ii)=O = S(m,n)/D(ii,ii)
ID(mJ) Im(ii,ii)=O = D(m,j)/D(ii,ii) (4.32)
where
ii = 1,2, ... , number of zero mass, j 1,2, ... , number of non-zero mass
The eigenvalnes of the matrix A will contain the natural frequencies of

the systems. Physically, only the eigenvalues related to the non-zero mass are
significant, which represent the natural frequencies of the slide.
Solving this matrix is the normal eigenvalue problem which can be easily
solved by using MATLAB™ The computer program based on the MATLAB was
developed to solve this kind of problem.

The natural frequency in the complex stiffness model is not influenced by
damping factor. Therefore, the natura! frequencies of the slide can be solved
by only using the real part of the complex stiffness of the hearings. However,
due to the frequency dependence, the solution of the system natural frequency
can not be obtained in one time. In this treatment, the mass matrix and
stiffness are all the M x M matrices. At one frequency, we have a constant
stiffness matrix, K, and we can have the difference between the given
frequencies and the eigenvalues of the matrix M- 1K(w), i.e.
DD = ldiag(À(w)I) - diag(eig[M- 1K(w)]) I (4.33)
where
eig : operation to find eigenvalues of the matrix
diag: diagonal vector of a matrix
At a given frequency, DD will be a vector with length of M. Using the

formula (4.33) in a frequency range, DD becomes a matrix with the dimension M
by the number of frequencies. It is obvious that the minimum value of the
matrix DD will be the natura! frequencies of the system. This model gives the
direct forward solution and smaller dimension of matrices, the mass is also
regular. However, it needs more calculations to find the natural frequencies
of the system.
4.5.4. Determination of vibration modes

In order to find eigenvalues of a matrix, the spatial variables {X} are
transformed into the general coordinate system, {y}, through the similarity
transformation of the matrix :
{y} = u- 1 {X} (4.34)
The transformation matrix U can be automatically derived through the matrix

diagonalization which is composed by the eigenvectors, i.e.
U = [{.ui}, {w}, ... ,{&}] (4.35)
Because the length of the eigenveetors is arbitrary, the real use of the
transformation matrix needs to normalize the eigenvectors. The vibration modes
which we have found are uncoupled modes in the generalized coordinates {y}. In
practice, we like to know the relationship between a mode and the system
parameters. This suggests that we need to analyze the composition of each mode
in tenns of the real spatial coordinates. From the relation (4.34), it is

easily to find that this can be done by using the transformation matrix, i.e.:
{X} =U {y} (4.36)
By doing so, we can determine the relationship between the vibration

modes and the system parameters. By using the eigenvector matrix, the
observable modes in a spatial coordinate can be determined. Using this
information, the system can be improved by adjusting the suitable parameters.
The transformation matrix is also useful to determine the vibration
displacement with time in a real coordinate system :
{x(t)} = U {diag eÀi 1}[!{diag eÀi 1}U- 1 B {Fd(t)}dt+C) (4.37)
where : eigenvalue
In our problems, the mass matrix is always diagonal so that the matrix B
equals the inverse of the mass matrix, i.e. B = M 1•
4.6. EFFECTS OF NON-LINEARITY OF FILM STIFFNESS

The displacement dependenee of gas bearing film stiffness can also be
studied by the loci of the response. The difference from the previous case is
that the hearing stiffness is no longer a constant, but a fourth order
polynomial function of the displacement. At a specific frequency, the equation
of the vibration reads :
(4.38)
If a simpte single harmonie input is used at a specific frequency, it was

proved [Storer and Tomlinson, 1991] that the higher order transfer function
can be constructed by Volterra series, which presents the diagonals of the
higher order frequeney response functions. Therefore, the first order transfer
tunetion under this single harmonie excitation will be described as following:
(4.39)
By this technique, the non-linearity of hearing film stiffness can be

studied. However, one should note that the dynamic displacement is no longer
one harmonie wave. Therefore, the dynamic characteristics given in Chapter 3
are no longer true. This can be a significant subject in the further study.
4.7. CONCLUSIONS
This chapter presented the models to describe the real gap height, which
are the inter:..connection tools between EPG pads and machine structures. lt can
be used to define a requirement on the guide beam toleranee in terms of
surface waviness and the parallelism in the opposite arrangement of EPG pads.
It may be also used in software correction by coupling the overall
characteristics of EPG hearing pads.
The stiffness of the gas film is generally a non-linear function of
frequency and displacement Nevertheless, in the study of the natural
frequencies of a gas hearing system, the hearing stiffness can be taken as a
specific constant value, i.e. the dynamic natural stiffness, to adapt the
general methods to solve the time-invariant, linear systems. Under a specified
excitation with a constant frequency, it is also reasonable to treat the gas
film with constant characteristics: constant stiffness and constant damping.
Thereafter, the metbod was developed to determine natural frequencies of a
system, the observable modes in a spatial coordinate and the vibration
displacement with time.
In the design of a gas hearing system, there are some more things to be
carefully considered, such as loop stiffness, kinematically positioning and
driving, symmetry and preloading.
CHAPTER 5.
EXPERIMENTS ON A LINEAR GUlDE WITH EPG HEARING PADS
5.1. INTRODUCTION
This chapter will present the detailed analysis and experiments for a
large linear guide with EPG hearings, which can act as an important component
in CMMs (,Çoordinate Measuring Machines) and some other machines with linear
motion components. The linear guide used in this research was designed in a
way simHar to one slide in a CMM. The purpose of the study is to find out the
role of the EPG hearings in the slide, especially referring to the natural
frequencies of gas bearing systems and the design principles such as loop
stiffness, symmetry, preloading, which were discussed in Chapter 4. In order
to reach this goal, the slide frame was designed with high transmissive
stiffness so that the slide can be considered as a rigid body. Therefore, the
effect of the gas hearing on the slide behaviour can be easily studied. Such a
slide can only be constructed by consirlering constructional details very
carefully, not only the slide frame but also the conneetion of the EPG pads to
the structure, the detailed design considerations were reported earlier
[Driessen, 1992]. The transmissive stiffness of 10 3 N/!lm was realized on the
frame at bearing mounting positions and on the bearing conneetion in which
elastic binges were used. The soft plate spring was selected as a pretoading
mechanism due to its compact size and other special features: to guarantee the
parallel movement of a pad and to realize the self-alignment in the same
mechanism. The binge-soft spring pair reduces the influence of non-parallelism
of the guide beam on the bearing behaviour.
Furthermore, this chapter will present detailed treatments of the models
to determine the natura! frequencies of the linear guide and their
experimental verifications. The slide was studied in three different
configurations bath theoretically and experimentally. The purposes are :
1). to verify the models;
2). to emphasize some essential concepts in the design of a gas hearing
system. These concepts include:
1). fully symmetrical arrangement will decouple vibration modes;
2). eccentric mass will lower the lowest natural frequency;

3). any additional weak element will introduce lower natural
frequencies than the lowest natural frequency determined by
the gas hearings alone.
5.2. DFSIGN OF THE LINEAR GUlDE

5.2.1. Mechanical structure
Fig. 5.1 shows the schematic diagram of the slide structure, Fig. 5.1a
shows the frame with the pad mounting positions and Fig. 5.1b presents the
whole slide : the frame with pads is supported by the granite beam. The
overview of the realized guide is shown in Fig. 5.5. The granite guide beam
was chosen because it is often used as the guide of X-slide in CMMs. The
tri-conical gap shaped pads were used in the guide due to the fact that they
provide the best overall characteristics as pointed out in chapter 3. Two pads
of 80 mm diameter were mounted on the top of the slide in order to test
possible situations with a heavier load. Pads of 60 mm diameter were used on
the two sides of the slide. The frame structure of the slide is optimized in
the sense that the slide can provide very high transmissive stiffness in the
hearing mounting points and the weight of the moving part of the guide is
camparabie with that of CMMs. In order to study the effect of the hearing
arrangement, pads can he mounted in 14 positions. Three basic arrangements
were used in the experiments in order to compare the slide hehaviour, i.e.
I). tully symmetrical arrangement of all the pads, ten pads were used
(positions 1 to 10 in Fig. S.la);
II). not-tully-symmetrical arrangement, eight pads were used (there are
no pads in the positions 3 and 7);
III). eccentric loading, the arrangement is similar to the contiguration
II but without balancing load.
In the contiguration I, four of the eight hearing pads were rigidly
mounted on one side of the slide through elastic binges (Fig. 5.2c) and the
other four were mounted on the other side of the slide through plate springs
(Fig. 5.3). The four hearings connected through plate springs tunetion as the
pretoading mechanism. The plate spring mounting enables the slide to reduce
the requirement for manufacturing and assembling accuracy. In contiguration 11
three pads were mounted on each side and all the connections were the same as
in contiguration I. In these two configurations, an additional mass was used
CHAPrER 5. EXPERIMEN1S ON TUE GUlDE 89
to adjust the mass center in the vertical plane whlch goes through the line
determined by two top hearings. In configuration lil, the balancing mass was
taken away and the weight of the slide structure acts as an eccentric toading
as it is the case in many CMMs.
F frame
z
y: 17
(b) {J~X
Fig. 5.1. Schematic illustration of the linear guide
Fig. 5.2 shows the configuration of the elastic hinge and two other
possible elements to mount an EPG pad on the slide. The elastic hinge is a
kind of ideal element to construct a precision mechanism, particularly for
thls kind of application. The hinge conneetion provides high transmissive
stiffness to ensure loop stiffness whlch is a weak point in traditional
designs. At the same time, it functions as a self-alignment mechanism which is
extremely important to ensure the hearings work properly with less strict
requirements on manufacturing and assembly accuracy. In the hinge, the

transmissive stiffness is guaranteed by the proper selection of the hinge
parameters and in the same element self-alignment is ensured by the material
elasticity.
In the pair of ball and conical surface (or other surfaces, e.g. a
spherical surface) (Fig. 5.2a), the contact stiffness is often too low in
normal working conditions for typical dimensions used in practice. The
stiffness can be estimated by the formula used for line contacts in roller
hearings. For high transmissive stiffness, e.g. for experiments with our
slide, it is not practical to select such elements. The reason is that we do
not want a large space between the slide and the guide beam, which will
obviously weaken the structure. (The detailed analysis will be given in
Appendix 5)
a) ball conical pair b) straight bar
c) elastic hinge
Fig. 5.2. Mechanisms to mount EPG pads on the slide
An additional important problem in the ball-cone pair is the the friction

behaviour in the contact. This friction behaviour is unpredictable due to a
lot of causes, e.g. the friction coefficient is not a constant and the line
CHAPTER 5. EXPERIMENTS ON THE GUlDE 91
contact situation may vary with the appearance of corrosion, dust and
lubricants, etc. The self-alignment results from the balance between the
friction moment and hearing moment. If we think: one step further, we would
doubt what really happens if one wants such a pair making a very small
rotation (10- 4 rad.)! The straight bar conneetion (Fig. 5.2b) really gets rid
of the line contact but the conneetion with high transmissive stiffness and
the efficient self-alignment can not be realized through the same bar.
;c>----- 11 _ _ _ _...,
plate spring
,-12~
hearing pad
Fig. 5.3. Platespring mechanisms to mount EPG pads on the slide
Furthermore, in the opposite EPG pad configuration, it is neeessary to

introduce a kind of soft conneetion (its stiffness is much lower than the
hearing film stiffness) in the preloading side. The soft conneetion functions
as the absorber of the displacement in the hearing gap height direetion. If
there is any change, e.g. due to non-parallelism of the two opposite surfaces
of the granite peam, this change will dominantly go to the soft connections
because its stiffness is much lower than that of the gas film (the detailed
analysis is given in Appendix 6). By analysis and comparison, a plate spring
shown in Fig. 5.3 was chosen in the slide. The plate spring is favorable in
mounting a preloading pad because it can be designed with the desired
stiffness in less space. At the same time, it can also be designed to
guarantee the parallel movement of a hearing pad by proper selection of the

length of two ann lengths, i.e. L1 = 2 L2, shown in Fig. 5.3. In addition, due
to lower bending stiffness and torsion stiffness, it can also act as a
self-alignment mechanism. The experience in the developed slide shows that the
plate spring works properly and is indeed a Sllitable element in Sllch an
application. Fig 5.4 shows the detailed view of a pair of pads : one is
mounted through the elastic hinge and the other through the plate spring.
Frame
Plate
spring
Hinge
Fig. 5.4. /llustration of the hinge-plate spring mounting pair
The stiffness of the finally realized guide in the hearing mounting

position is higher than 109 N/m, about 10 times the hearing film stiffness.
The stiffness in the transmissive direction of the hinge is also higher than
109 N/m. These stiffness values ensure an acceptable loop stiffness.
Therefore, the slide can be considered as a rigid body. Furthermore, the
lowest natural frequency of the slide frame is about 900 Hz which is far out
of the frequency range for testing the slide. Fig. 5.5. illustrates the
realized linear guide.
5.2.2. Analysis
Before discussing the experimental verification, we first repeat the aim
Fig. 5.5. lllustration of the realized slide
of the research on the slide and explain the techniques used in the solution.
Until now, the dynamics of complex, large machines, like CMMs, are rarely
investigated, particularly the machines with EPG hearings. With further
increasing requirements on machine accuracy and reliability, the dynamic
behaviour of the maclllne becomes the most important factor for designers. The
EPG hearing is a basic element in those machines and a key factor influencing
the dynamic behaviour of these machines. Therefore, there is an urgent need to

develop appropriate roodels to study the machine's dynamic behaviour in
relation to EPG hearings. In Chapter 4, two roodels were developed: 1). the
three element model in which the gas film characteristics are represented by
three constants: two springs and a dashpot; 2). the damping factor model (the
complex stiffness model) in which the complex stiffness of a gas film is used.
For clarity, the coordinates used in the roodels are shown in Fig. 5.6.
(b) coordinates for

the complex
stiffness model
(a) coordinates for the three

element model
Fig. 5.6. /llustration of the coordinates used in the models
If we use the three element model, the generalized coordinate vector

reads:
S=[Y, Z, a, !3, y, Çl,Ç2,Ç3,Ç4, t;l,t;2,1;,3,t;4,t;5,1;,6,p);8,1;,9,t;10] (S.la)
for the contiguration I (as described in 5.2.1.) and :
S =[Y, Z, a, (3, y, l;l, l;2, l;3, ~1, ~. t;3, ~4, t;S, f;6, l;,7, l;,8] (5.1b)
for the contiguration II, (also described in 5.2.1.).
The first five coordinates descrihe the slide motions, i.e. Y, Z, a, (3,
y, l;'s are the motions of the hearing body attached to the plate spring, and
t's are the motions of hearing zero mass elements.
The main diagonal element veetors of the mass matrices read as :
{M} =[ms, ms, jx, jy, jz, mb,mb,mb,mb, 0,0,0,0,0,0,0,0,0,0] (5.2a)
{M} =[ms, ms, jx, jy, jz, mb,mb,mb, 0,0,0,0,0,0,0,0] (5.2b)
where ms : the mass of the slide, mb : the mass of a pad,
jx,jy,jz : moments of inertia of the slide about X, Y, and Z.
In the damping factor model, the coordinate veetors become :
S=[Y, Z, a, (3, y, l;l,l;2,l;3,l;4] (5.3a)
for the contiguration I and :

S =[Y, Z, a, (3, y, l;l, l;2, l;3] (5.3b)
for the contiguration IL Due to the constructional feature of the slide and
the selection of the coordinate system, the mass matrix is the main diagonal
matrix. The element veetors in the diagonal read respectively :
{M} =[ms, ms, jx, jy, jz, mb, mb, mb, mb] (5.4a)
{M} =[ms, ms, jx, jy, jz, mb, mb, mb] (5.4b)
It is clear that one more degree of treedom for one hearing is introduced
in the three element model. In addition, the mass matrix is singular.
Nevertheless, there exists a special technique to deal with it, as introduced
in Chapter 4. The damping matrix and stiffness matrix can he filled according
to the equations of vibration motion of the slide, which are the symmetrical
matrices. All the elements in the stiffness matrix are constants so that the
problem becomes an eigenvalue problem. In the complex stiffness model, the
total degrees of freedom are less and the mass matrix is regular. However, the
elements in the stiffness matrix are the dynamic stiffness of gas film which
is frequency-dependent. The final solution can only he determined by the loci
of a series of solutions, as explained in Chapter 4. For a given hearing,
either the constants in the three element model or the dynamic stiffness
distribution in the damping factor model can he obtained by using the hearing
design software mentioned in Chapter 3.
By using the techniques described in Chapter 4, the spectrum of natural

frequencies of the slide can be calculated. Fig. 5.7 shows nine of them
2500
2023
~ 2000
N
::c platesp ngs
I
-- 1500
~
~ 1000 ,... ~ a 'Y y
I 500
z
0286 0385 0424
0126 0145
I I
00 1 2 3 4 5 6 7 8 9 10
modenumber
Fig. 5. 7. The calculated spectrum of the natura/ frequencies
450
400
'N' 350
e;
~s 300
"'
!a
:Ë
·r;s 250
!
5 200
150
100
100 150 200 250 300 350 400 450
Three element model (Hz)
Fig. 5.8. The comparison of the calculated natura[ frequencies between

the three element model and the complex stiffness model
(configuration I) (the test results will be given in Appendix 8)
CHAPTER 5. EXPERIMEN'IS ON THE GUlDE 97
corresponding to the non-zero masses, calculated by the three element model

for the contiguration I. Note that the eigenvalnes are obtained by
diagonalization of the matrix A wbich implies the coordinate transformation
from the spatial coordinate system into a generalized coordinate system. The
observable natural frequencies in a spatial coordinate can be determined by
the eigenvector wbich can be used to transform the vibration modes from the
generalized coordinate system back to the spatial coordinate system. In other
words, the eigenvector indicates the movement of all the masses in the system
along the corresponding spatial coordinate direction at the natural
frequencies. With the help of the eigenvector, one can identify the mode shape
of a vibration mode in the spatial coordinate(s), as indicated on the top of
the frequency bars. Fig. 5.8 presents the comparison of the natural
frequencies of the system between the complex stiffness model and the three
element model. The straight line in the figure represents the perfect
agreement between these two results and the symbols 'o' repcesent the natura I
frequencies calculated by the models. It can be seen that there is no conflict
between these two models. Due to the fuJJy symmetrical arrangement of all the
pads, the vibration modes are not coupled.
5.3. EXPERIMEN'IS
5.3.1. Experimental set-up
Experiments were performed on the guide to verify the theory derived in
this thesis. The experimental set-up for testing single pads (Fig. 2.5) was
modified to perform the experiment on the guide, schematically shown in Fig.
5.9. The overview of the whole test system is given in Fig. 5.10. In order to
eliminate the influence of the test set-up in the dynamic test of the slide,
measurements on the structure were also performed, including the structure for
supporting the shaker and the structure for supporting the guide beam. From
the model, we deal with the slide frame vibration relative to the guide
granite beam. Therefore, it is necessary to measure the vibration at reference
points on the slide. The most accurate reierenee will be the point nearest to
the test point on the frame.
The dynamic behaviour of the frame itself was also checked by a modal
analyzer. The excitation was delivered by a hammer and the acceleration was
measured. It confirms that the slide can be considered as a rigid body in the
range for testing the slide, the test results are given in Appendix 8.
Gaa aupply
Fig. 5.9. Illustration of the test set-up for the dynamic test of a slide
5.3.2. The identification of vibration modes

The electro-magnetic ·exciter was used to generate the dynamic force and
the force transducer was mounted between the exciter and the slide to measure
the dynamic force in the random and step sine testing. The accelerometer
or/and displacement sensor can be mounted on the slide to measure the response
of the slide. In the detailed experiment, the stepwise sine testing signa! was
chosen in order to concentrate the energy on the measuring frequency. The
dynamic force was delivered in the vertical direction at the center top of the
slide, as indicated in Fig. 5.1. In this loading situation, it provides two
excitation forces : the force in Z-direction and the moment about X-axis. This
is because the top pads were designed at the off-center along the negative
direction of the Y-axis, see Fig. 5.1. The accelerometer was mounted on the
slide through a permanent magnet and on the granite through wax, which can
provide sufficient accuracy in the specified frequency range. In the
identification of the vibration in Y-direction and the rotations about the Z
and about the Y axes, the hammer excitation was used. The hammer was also used
to check the natura! frequencies of the frame and the support. In this type of
test, the sensitive excitation point was chosen for determining a desired
natura! frequency. Appendix 8 presents a typical test result.
Fig. 5.10. Overview of the test set-up jor the dynamic test of a slide
5.3.2.1. Fully symmetrlcal oonf.guration (I)

The arrangement with 10 pads, two on the top and eight on two sides, was
the first contiguration in which the pads were fully symmetrically positioned
on the slide referring to the mass center. Therefore, if each pair of
symmetrically arranged pads bas the same stiffness there will not be any
cross-coupling between the vibration modes. After experimentally determining
the reai working parameters of the top hearings, the experimental results
should correlate well with the theoretica! predictions.
The analysis by using the three-element model indicated that the natural
frequencies of the slide which can be observed in Z-direction are 145 Hz and
the natural frequencies which can be observed in the rotation about the X-axis
and the Y-axis are 286 Hz and 126 Hz (Fig. 5.7). By placing the accelerometer
100 Design of Gas Hearing Systems tor Precision Applications
in different positions, the natura} frequencies corresponding to different

spatial coordinates can be distinguished through oomparing the results. As
mentioned earlier, the dynamic toading on the slide from the shaker was
delivered in the Z-direction and the rotation about X-axis, the moment
generation was due to the eccentric toading (Fig. 5.1). Therefore, the really
ohservable natura! frequencies are only those in the Z-direction and the
rotation about X-axis, i.e. 145 Hz and 286 Hz. Fig. 5.11 shows the comparison
between the experimental results and the theoretica! predictions. The straight
line indicates the ideal conditions under which the experiments coincide with
the theoretical predictions. If the line determined by the comparison points,
'o's, perfectly matches the straight line, it shows good agreement between the
experiments and the predictions. If there is a systematic dispersion, it means
that there is an unsuitable modeling of film stiffness or a wrong estimation
of rnass/moment of inertia. If a random dispersion appears, it may indicate
deficiency in the experiments. At a natural frequency, the upward (downward)
deviation of a comparative point from the straight line implies the under-
(over-) estimation of that natural frequency by the model.
From Fig. 5.11, it can be seen that the experiments have very close
correlations to the theoretica! predictions. The two 'o's in one natural
frequency are the calculated results from two modets. Th:is is a very
encouraging result. Thereafter, more experiments were performed and compared
with both the three-element model and the damping factor model. They all show
very good agreement with the tl}eoreticat predictions.
500
N' 400
6
"'5 300
:~
l 200
100
100 150 200 250 300 350 400 450
Experiments (Hz)
Fig. 5.11. The comparison of the natura/ frequencies of the slide between
the experiments and theoretica/ predictions (configuration . /)
CHAPI'ER 5. EXPERIMENTS ON THE GUlDE 101
2500 22!r.5
2192
~ 2000 2011
t=
N
1500 r
plate spring
~ 1000
l 500
0240 0284 0336 I
0591
0148
0 I I I
0 1 2 3 4 5 6 7 8 9
moelenurnher
Fig. 5.12. The spectrum of natura! frequencies (conftguration II)
600
550
500
450
'N
e; 400
"'
.ii6 350
l 300
250
150 200 250 300 350 400 450 500 550 600
Experiments (Hz)
Fig. 5.1 3. Comparison between the theoretica/ predictions and experimental

verifications in the configuration 1I
102 Design of Gas Bearing Systems tor Precision Applieations
5.3.2.2. Partially symmetrieal contiguration (11)

The three pads on one side are rigidly mounted and the other three are
mounted on the slide through plate springs. In this configuration, the working
parameters of all the pads on the sides can be determined. Therefore, a more
detailed and more accurate comparison can be performed. Fig. 5.12 shows the
spectrum of the natural frequencies of the slide under the contiguration IT.
Fig. 5.13 presents the comparison between the experiments and theoretica}
predictions. This result once more shows a good correlation between the model
predictions and the experiments.
Because the pads in the contiguration II are not fully symmetrically
arranged in the side, the responses in Z, Y, a, and 'Y show strong cross
coupling between the vibration modes. As one example, Fig. 5.14 presents the
predictions on the mode shapes in Y-direction and a-rotation. The vertical
axis represents the relative amplitude of each mode appeared in this
coordinate. One can expect that if a weak driving mechanism is introduced in
the system one lower natural frequency than those given in the figure will
Observable modes in Y direction
1
240Hz 591Hz
·~
0.5
.8
·~ 0
:a
336Hz
~ -o.s
-1
0 1 2 3 4 5 6 7 8 9
modenumber
Observable modes in a - direction
1
0.8
s
] 0.6
240Hz 336Hz
·s
VJ
:a 0.4
~ 0.2
0
0 1 2 3 4 5 6 7 8 9
modenumber
Fig. 5.14. Mode shapes in Z-direction and a-rotation (conjiguration Il)
appear in these two coordinates. This is because the driving mechanism can not
be mounted through the mass center.
5.3.2.3. lnfluences of eccentric mass

In practice, most machines work under eccentric loading, which is a
typical situation in CMMs. The model to determine system natura! frequencies
is based on the mass center coordinates. A slide with eccentric mass,
generally speaking, refers to the condition in which the mass center of the
slide is different from the center determined by all the pads, which implies
the change of mass and the moments of inertia and the change of the hearing
position related to the mass center used in the model. If the model is used to
study the machine with eccentric mass, we only need to change the constants
related to the hearing positions and the moments of inertia of the slide.
Obviously, one will expect the lowest natural frequency of the system will be
decreased due to the increase of mass and moment of inertia. In most CMMs, the
moment of inertia is much higher than that of our test slide, especially that
about the X-axis and the stiffness is lower than that we used. Therefore, the
CMMs will have lower lowest natura! frequencies than the slide which we used
in this study. Fig. 5.15 presents a test result on the slide with eccentric
mass, contiguration III (the contiguration II without the balancing mass, see
Fig. 5.1). The mass center bas a distance 30 mm from the line determined by
the two top pads. The calculated natura! frequencies in the Z- and a-
directions are 165 Hz and 256 Hz. Comparing these results with those shown in
Fig. 5.15, one can see that it once more shows a very good agreement between
the experiments and the model predictions.
2.------r------.------.------~----~------.------.
1.5
0.5
250 300 350

frequency (Hz)
Fig. 5.15. A test result on the slide under eccentric toading

5.4. DISCUSSION
5.4.1. Symmetry
A fully symmetrical arrangement of all the pads will eliminate the
cross-coupling among vibration modes, which is illustrated in Fig. 5.14, in
which the calculation was based on the assumption that all the pads have the
same characteristics. However, in practice it is almost impossible to ensure
fully symmetrical contiguration due to the difficulties in the control of the
woricing parameters. As a result, weak cross-coupling, at least, will he
observed in most machines. If we calculate the slide with real hearing
parameters, the cross coupling can he predicted. For the slide with variabie
toading like those used in the Y-slide of CMMs, the pads work under different
gap heights, which means that in general the pads operate with different
stiffness and damping.
There is another point which is interesting to he mentioned, i.e. the
driving mechanism. In most machines, the acting point of driving force is not
through the mass center, which means that the vibration mode related to the
driving system will he cross-coupled to all the coordinate directions. If
driving stiffness is very low, it will introduce a much lower natural
frequency than those determined by the hearings.
In practice, a machine slide may he heavier than the slide used in the
study and the machine size rnay also he larger than that of the slide.
Therefore, one will expect the appearance of lower natural frequencies than
those determined in the study.
5.4.2. lnfluence of the support of the guide beam

The influence of the support can he studied by expanding the model to
include the degree of treedom of the support. With similar extension, one can
study the whole machine. With the extended model, it is possible to optimize
the structural rnass and hearing dynamic stiffness of X-slide and Y-slide in a
machine, e.g. a CMM, to obtain better overall dynamic hehaviour. In principle,
the interaction of vibration modes in two bodies can he used to obtain minimal
amplitude in the desired position, e.g. the tooi holding position. The
generalized theory of vibration isolation and vibration absorption can he the
guide to perform the research. This strategy refers to the "hardware"
correction of machine errors. With the help of the model, the effective
algorithm for software correction of the machine error due to vibrations can
be developed. These topics are beyond the current scope of research and will
not be treated in detail.
5.5. CONCLUSIONS
This chapter explained the theory and the solution techniques, as well as
the experimental system for designing and analyzing a gas hearing slide. The
experimental identification of the natural frequencies of the slide proves the
correctness of the developed models. Therefore, the model developed in this
work can be used in design and analysis of gas hearing slides, even for more
complicated structures.
In the design and test of the slide, constructional details were
emphasized, particularly the conneetion mechanisms were discussed in detail.
It shows the importance of taking care with these details.
The metbod and model developed in this thesis were proved by experiments
on the slide in three configurations. The model is not limited by the machine
dimension and hearing working condition. Under a given condition, the
parameters in the models remian constants. Therefore, the model can be
directly used to design new slides and to analyze existing slides. It can also
be extended to analyze complete multi-axis machines.
CHAPTER 6.
CONCLUSIONS AND RECOMMENDATION
6.1. CONCLUSIONS
The research in this project shows that the developed software and the
experimental system can he effectively used in the design, study and testing
of both single EPG hearings and EPG hearing systems.
From the research on the characteristics of hearings and hearing systems,
one can draw the following conclusions:
1). The numerical and experimental research reveal that the EPG hearings
with concave gap shapes provides higher load capacity and higher stiffness.
The tri-conical gap shaped pads are superior to the parallel and bi-conical
gap shaped pads. Under the same stabie conditions, the tri-conical gap shaped
pads provide higher load capacity and higher stiffness than those provided by
the other two.
2). The hearing body tilt will greatly reduce the load capacity and
stiffness of the parallel gap shaped hearing: maximum change of the stiffness
is up to 60 % at the optima! working gap height. On the contrary, the
tri-conical gap shaped pads are much less sensitive to tilt, the stiffness
change is less than 1 % at the optima! working gap height.
3). The tri-/bi-canical gap shaped pads are also less influenced by
surface imperfection, especially when half a wavelength is greater than 1.5
times the pad reference dimeosion (e.g. diameter for circular pads). The worst
case is when half a wavelength is equal to the pad reference dimension. The
overall characteristics follow the change of the surface profile, particularly
load capacity. Under constant load, this means that the change of the working
gap height follows the surface pattem.
4). In the velocity range of less than one meter per second, which is the
current status of CMMs, the load capacity and stiffness are rarely changed.
However, relatively high velocity will greatly reduce the hearing load
capacity and stiffness.
5). The temperature of the gas supply is generally lower than the
standard room temperature, 0.5 ~ 1 °C. The gas through a restcictor can be
Design of Gas Bearing Systems for Precision AppUcations 107
further cooled down. In very accurate applications, it may cause significant

distoetion of the structure. From the study, a general guideline was
established for designing mechanica! systems with EPG pads. The bi/tri-conical
EPG pads are preferabie because they are less sensitive to hearing body tilt
and surface imperfections.
6). In dealing with gas hearing systems, this thesis presents the model
descrihing the real worldng gap height in a system, two models descrihing the
hearing film stiffness and the metbod for applying the models to the gas
hearing systems for the purposes of design and analysis.
7). The models descrihing the real gap height can he used to define a
requirement on the guide beam toleranee in terms of surface waviness and the
parallelism in the opposite arrangement of EPG pads.
8). The stiffness of the gas film is generally a non-linear function of
frequency and displacement Nevertheless, in the study of the natural
frequencies of a gas hearing system, the hearing stiffness can he represented
by springs and dashpots, as in the three-element model: one dashpot and two
springs. The purpose of the treatment is to adapt the general methods to solve
the time-invariant, linear systems. Under a specified excitation at a constant
frequency, it is also reasanabie to take the dynamic stiffness of gas film at
that frequency into the vibration equation. The final salution can he derived
from loci of a series of linear systems with constant stiffness and constant
damping. Based on these, two models were developed to predict the natural
frequencies of a gas hearing system. These two models all work well and can he
conveniently used in designing and analyzing the gas hearing systems.
9). Theoretica! analysis and the experiments on the three different
configurations of a large linear guide showed the validity of these models.
The experimental results correlate well with the predictions of the models.
Furthermore, from the study, some essential concepts were revealed. It can he
concluded that the models can he used to analyze the machines with EPG
hearings, e.g. CMMs. In the design of a gas hearing system, there are some
more constructional details to he carefully considered, such as loop
stiffness, kinematica! positioning and driving, symmetry and preloading. This
thesis also provides some discussions and solutions for those aspects. In the
system design, weak elements (or components) should he paid more attention
because of the cross-coupling of vibration modes.
108 CHAPTER 6. CONCLUSIONS AND RECOMMENDATIONS
6.2. RECOMMENDATIONS
As a result of this research, we may also derive some areas for future
studies. First of all, it may be concluded that the various investigations in
this thesis could be extended to study the joumal gas hearings, including
modification of gap shape, effect of rotor tilt and surface imperfections. The
metbod of determining the system's natural frequencies may also be used in the
spindie system.
Secondly, the non-linear model for gas film stiffness is important for
the large dynamic change of working gap height, as the case of Y-slide in a
CMM.
Thirdly, on the question of developing the EPG hearings, it is worthwhile
spending time and effort developing cheap, reliable, integrated, pretoading
mechanisms, such as the supersonic-pressurization combination or the Bemoulli
-pressurization combination, for pad type hearings used in both linear motion
and rotation systems.
Fourthly, in relation to the development of close-loop controlled EPG
hearings, we would suggest stopping any attempt to control the supply
pressure, including the control of the restrictor.
Finally, in the design of gas hearing systems, if one would like to
design a close-loop controlled system, the ideal position to mount the
actuator is at the conneetion between the hearing pad and the structure, which
is also true for the tilting pad hearings used in spindles. For high speed
spindles, this kind of controlled system could best be left to the active
magnetic hearings.
APPENDIX 1.
THE SIMPLIFICATION OF THE EQUATION OF MOTION
The equation of motion can be derived by characterizing the dynamic

equilibrium between the inertia forces, the mass forces and the stresses in
the fluid and can be expressed in a compact form [Constantinuscu, 1969] :
pre ssure force

I viscos i ty force
Q g~ = f - grad p + 2 div(~ D) - ~ grad(~ div V) (Al.l)
-,..--- T s t resses
mass farces
inertia forces
where V : velocity vector, t : time, Q : density, ~ : viscosity

D : strain-rate tensor
Bu/ax (au/ay+av/ax)/2 (au/az+aw/ax)/211

D = (Bu/ay+av/ax)/2 av/ay (av/az+aw/ay)/2 (Al.2)
11 (8U/8z+aw/ax) /2 (av/ az+aw/ 8y) /2 aw/az
The expansions of the equation of motion in the rectangular coordinates

read :
Q gut -~X
..,
+ ~8X(~ éJ8UX) +!!._~(8U
8y ay ax
+8V )]+ !!._ [~( 8U + 8~~- ~(~div V)
az az ax 38x
Q gv = _ !!e.. + ~(~ a v) +!!_[~(au +av )]+ !!._ [~( av + a~] _ ~(~div v)
1 ay ay ay ax 8y ax az az ay JBy
Dw _ !!e_ + ~(IJ BW) +!!._(IJ(8U +aw ))+!!._(IJ( 8V + BW\) _~(IJ div V)
Q ITT = az Bz r- 8z ax ... Bz ax By ... Bz ay' 38Z r
(Al.3)
whe re
(Al.4)
The dimensionless form of the equation of motion can be written in

the following form if the dimensionless parameters are introduced :
x= lo X, y =ho Y =lo (ho/lo) y, z =bo Z = lo À i, À= bo/lo
u= Uo ü, v=Uo (ho/lo) ü, w=Uo'J..w, p=pop, Q=QoQ,
~ = ~o i!, u = Uo Ü, t (lo/Uo) Ï,
Ma= UÖ/(kRTo), QO =po/(RTo)
110 Design of Gas Hearing System For Precision Applications
kMi(N)2 g n~ = _ ~ +N~{z(N)2 8 _(JÏ ä~ )+(~)2 8 JJÏ(~ +(~2 ä~J

Dt 8y ay ay 8x 8y 8x
+ (~)2 8_ [JÏ(ä~ +(N)2 8~] _ ~ (~2 8 _(JÏ div V ))

8z 8z 8z 8y
2 - Dw = - .::.:----
k (ÀMa ) Q - -
8n +N~ X
{2 (h0)2 8
w --:- (-
~
8w ) + --:-
-- 8 [-(8w (bo)2 8-Y\J
~ -- + 10 --:-1
Dt 8z 8z 8z 8y 8y 8z
+ (N)2 8_ [JÏ(8~ + 8~)] _ ~ (~)2 8 _(JÏ div V )}

8z 8z 8x 8z
(ALS)
Therefore, the relative order of the velocity derivatives can he listed

in Table Al :
TABLE Al The Relative Order of Magnitude of the

Derivatives !!! the Equation of Motion
Relative
D e r i va t i ves
order
~ ~~
8x' 8z' 8y 1
2 2
N~, kMa,ÀkMa
2
kMi (h/L)
N~(h/L)
2
Note that the working gap height (h) of EPG hearings is very small in
comparison to hearing dimensions (L), h/L a! 10- 3- 10-4, in the restrictor
region 10- 1- 10- 2 • Therefore, the terms of the derivative of velocity
helonging to the third and fourth rows can he omitted. Hence, the equation of
APPENDIX 1. THE SIMPLIFICATION 111
motion reduces to the following form :
=- ~ + Nf.l a- (ii a~)

ax ay ay
0 =- ~ (Al.6)
By
k ( ÀM a )
2 Dw
--
Af\ a (- Bw)
= - .::.:- +Nf! --:- f.l --
Dt az ay ay
or the dimensional form :
(A1.7)
APPENDIX 2.
THE SIMPLIFICATION OF THE ENERGY EQUATION
The (thennal) energy equation for a perfect gas can be expressed in the
following form [Kakaç, et al, 1987] :
QCp~T· = V(xVTs) + ~ + ~-tel> (A2.1)
cl> = [(8U)2+(8!i2+(8~.)2)+(8U+8~.)2+(8V+8~2+(8W+8V)2_ ~ (V•V)2 (A2.2)

ax 8y' az az ax ax 8y' ay aZ! 3
By virtue of the equation of motion and the use of the concept of

stagnation tempemture (To), the energy equation reads
= V•(x VTs) + ~ + 8 (A2.3)

I at I
convect10n conducttOn :rdt'-s-!::-,st"="pa::-:;t".,to=n
where
2 2 2
To=Ts + (u + v + w )/2cp (A2.4)
(A2.5)
x : heat conduction coefficient

To : gas stagnation tempemture
Ts : gas static temperature
Cp : heat capacity at constant pressure
Noting the relative order, the dissipation term 8 can be simplified as :
~ 8 [ 8u] + -8 [ 8w] (A2.6)

ay !J.U -ay
~ !I! -
ay !J.W -ay
lt is also true for the heat conduction terms in the hearing film, i.e.
V (x VTs) = ~ (x BTs) (A2.7)

ay ay
APPENDIX 3.
THE JOULE-THOMSON COEFFICIENT
When the Joule-Thomson coefficient is considered in the flow process, the

energy equation reads :
udu = ep dT - I"}P (A3.1)
where ll : the Joule-thomson coefficient in the condition of zero-velocity,

for air under 150 bar pressure, it can be expressed as follows
[Zhang and Shi, 1978]:
!!J ~ (2.73 x 10"3 - 0.0895 x 10-6 p)(273/f) 2 (A3.2)
To integrate the equation (A3.1) and take the square root of it yields :
u "' [2ep (To -Td)- !! (Po - Pd)] 112 (A3.3)

Jm
where (A3.4)
APPENDIX 4.
THE FEM PORMULATION OF THE REYNOLDS EQUATION
The procedures of the derivation of the FEM formula are illustrated by

taking the static part of the Reynolds equation as an example. In order to
clarify the derivatión, the Reynolds equation is written as a compact form :
div{ 1 ~! [ 0
p grad po + po grad p
0
] - po hoV/2}+ IJ<?
div(~ p grad p~
0
0
"'
Uó~
0
p )- .i Ö
I= 1
(f 0 (p.)I - 1
8pBf ) (A4.1)
Note that the operations div and grad are only performed for x and z.
In the present formulation of the Reynolds equation, only the Dirichlet

boundary condition, ro, appears, i.e. the peripheries of a hearing pad.
The FEM formulation of the Reynolds equation can he derived by following

the standard procedures [Cuvelier et al, 1986], i.e.
1). to multiply the equation by a test function 'ljJ which vanishes
on the hearing peripheries;
2). to integrate the modified form of the Reynolds equation derived
from step (1) over hearing area, Q;
3). to apply Green's formula to the integral form.
Perlorming the steps (1) and (2) gives :
(div{ 1 ~~...
0
.f [ p 0 grad po + po grad p 0 ] pohoV/2}+ pq af )1jJk dQ
' 8p
.f(div(~ p grad p~
0
"' 0
p 0 )- .i Ö (f 0 (p .)- 8pBf )} 'ljJk dQ (A4.2)
~'-~ '• I ' '
After the step (3), the equation becomes :
.r({l~! :~ '\jJk dQ
0
0
[ pograd po +po grad p ] -po hoV/2}•grad $) dQ - .fJX?
(~ p p~
0
0
"' .f grad p 0 ) grad 'ljJk dQ +.i Ö (f 0 (p.)- a8pf )) lJ>k dQ (A4.3)
1.:0~ I • 1 I I
Note that J' (div 1jJk {~t [ 0

p grad po + po grad p
0
] - po hoV/2}) dQ and
J' div( 1 ~! 0
1jJk p grad p 0 ) dQ can be converted into linear integral over the
boundaries, whlch are equal to zero, i.e. J' dr = 0.

ro
The unknown p in the domain Q can be written as :
(A4.6)
from whlch we deduce the following system of linear algebraic equations for
pressures in all nodal points, Pt, P2, ... , Pn :
As mentioned earlier, the hearing area, Q, in the finite element

approximation is subdivided into a finite number of triangles. Assume that
there are N vertices of triangles: al, a2, ... , on. For each nodal point, a;,
a basic tunetion 'ljli is defined with the following properties :
1 if i='
1). 'Ijl; (m) = 09 = { 0 if i*~ iJ, = 1, 2, ... ,N
2). 'Ijl; is linear in each triangle
3). 'Ijl; is continuous on Q = Q n ro,
If the element does not include a part of a restrictor, a component of

the coefficient matrix and a component of the right hand side vector are
separately expressed as follows :
Kij = h~ ~2 [ (B.i l: Pi Bi + Cï l: Pi Cï)/3 + ~ (B.i Bi+ Ci Ci))

24!! ll
- ~i I~ I (U eBi + wee;) (A4.8a)
3
Rï • he ~ p (B.i l: P; Bi + Ci l: Pi
0
Ci) (A4.8b)
24!J. fl2 e
116 APPENDIX 4. TUE FEM FORMULATION
If the element does enclose a part of a restrictor, a component of the

coefficient matrix and a component of the right hand side vector become :
3
Kii = [!!..!...._ ill2 [(B.i l: P; B; + Ci l: P; C;)/3 + P~ (B.i B;+ Cï C;)]
24~ 11
- ~zl~l (U"B; + W"Ci))(l- f*l) (A4.9a)
3
R.iï = [!!..!...._ ill p0 (B.i l: P; B; + Ci l: P; C;))(l- ~ (A4.9b)
24~ 112 e TKI
where
Pe = (P 1 + P 2 + P 3)/3 (Because of the linear interpolation function)
Ig.dQ = 1111/2 • Ig,dQ = 111rl/2 Ig 1jJj dQ = 1111/6
N
~ =.l: p 8'1jlj
81; j•l ja;
N =3 for the triangle element, s= x,z
1jJj = (Ai + Bi x + Ci z)/11
~-Bi ~-Ci
ax - ?S: az - ?S:
Ai = x k z kd - x kd z k , Bi = zk -zk+l Ci = x k+l -x k
with k = mod[ G+1)/3], (mod is operation to find remaindering)
Note that if the resttictor area is small, the equation (A4.9a) can take
the same form as that in the non-resttictor region, i.e. the equation (A4.8a).
If a point in the element is inside of the inlet resttictor region, the
components read :
(A4.10)
where
2
Ad : resttictor area, Ad = xd /4
Therefore, the unknowns can be found by solving the set of linear

equations, because the matrix and the right-hand side vector are already
expressed by the constants.
The FEM tormulation for the dynamic equations can be derived by following
the same procedures.
APPENDIX 5.
ANALYSIS OF BALL-CONICAL SURFACE CONTACT
The most relevant characteristics of ball-conical surfaces (or other

surfaces, e.g. spherical surface) are the stiffness in the contact region in
form of line contact and the rotation about a very small angle for
self-alignment.
rotation axis
a) ball-conical pair b) mustration of the arm

of friction moment
Fig. A5.1. Ball-conical surface contact and its line contact
As shown in Fig. AS.la, in the ideal condition, the contact between these
two surfaces can be considered as a line contact, as illustrated in batehing
lines.
The stiffness (deformation) can be estimated by the formula used in the
analysis of rolling hearings. One of them is that given by Palrogren [Palmgren,
1964], which reads :
9
s. __ 2.45 Fg.
u (AS.l)
E~·9L2ts
where Fn : force perpendicular to contact surface, Fn "' F cos a

Let : effective length, 2 xR sin a
R : hall radius, a : conical angle
Er : the reduced modulus of elasticity of materials

2
1 - 1 [ 1 - 1J1 + _1---r..-:---1 (A5.2)
ffi"- -2 Et
Er = 2.29 x 105 N/mm2 for steel-steel contact
Er = 1.36 x 105 N/mm2 for steel-brass contact
u : the Paission ratio of material i

E; : the elasticity modulus of material i
The deformation in the transverse direction bas the following relation

with the normal deformation, i.e. Öt = ó cos a. Therefore, the deformation in
the transverse direction can be written as follows :
09
& = 0 .62 F ' (cos a) L 9 (A5.3)
E~ 9 08
· R ' (sin a) 0·8
The contact stiffness in the transverse direction, Kc, can be expressed

as :
(A5.4)
From this relation, it is easy to find out that the optimal conical angle
a is 30°. Mter introducing this optimal angle and the reduced elasticity, the
contact stiffness can be written :
Kc = Cc R0 ' 8 Fl· 1 (A5.5)
where Cc = 0.56 (0.33) for steel-steel (steel-brass)

For the general purpose, a plot of the contact stiffness change with the
steel-steel contact
3000
ê
g 2000 F=800(N)
lt::.
-
~
~ 1000
F =lOO(N)
e
0
2 3 4 5 6 7 8 9 10
ball radius (mm)
Fig. A5.2. The contact stiffness of ball-conical surface pair
APPENDIX 5. ANALYSIS OF BALL-CONICAL SURFACE CONTACT 119
hall radius and toading force is given in Fig. A5.2.

In conceming the self-alignment in ball-cone pairs, it is difficult to
say what really happens in the contact. In an ideal condition of contact, one
may say the self-alignment depends on the friction behavior at contact.
Due to close contact, the component of loading force in tangential
direction is halanced by itself. The rotation axis of the friction moment is a
line through the hall centre and parallel to the bearing surface, which
implies that the local force on the circle will have different contribution to
the total friction moment due to the difference in the arm length. At one
point, the local friction force, if we do not consider the material
hysteresis, can be expressed as :
dFt= f F cos a /(2:n:R sin a) (R d6) (A5.6)
and the distance of the local friction force to the rotadon axis reads :
La=Rsin6 (A5.7)
Hence, the local friction moment reads :

dMr= if/2:n:) F R cos a sin 6 d6 (A5.8)
and total friction moment becomes :

Mt = (2J/:n:) F R cos a (A5.9)
For the general purpose, a plot of the ratio of friction moment over
friction coefficient with the hall radius and toading force is given in Fig.
A5.3.
3 4 5 6 7 8 9 10
ball radius (mm)
Fig. AS. 3. The friction moment of ball-conical sw:face pair

APPENDIX 6.
ANALYSIS OF A PAIR OF RIGID-FLEXIBLE MOUNTING
We concentrate on a pair of the pad connections: one side is rigidly

mounted on the slide through the hinge and the other side through a kind of
soft spring, which Can be schematically shown in Fig. A6.1.
ffis
Fig. A6.1. Schematically illustration of a mounting pair
On the side with rigid mounting, the stiffness is equal to the gas film
stiffness and on the side with the soft conneetion the stiffness is the result
of two springs in series: gas film stiffness and soft spring stiffness.
(A6.1)
If the stiffness of the soft conneetion (ks) is much lower than the
stiffness (kg') of gas film of the pad mounting in the same side, the result
stiffness will be approximately equal to the soft spring stiffness, kt !!! ks.
In the situation from one equilibrium condition to another, the changes of
forces in two sides must be equal. Hence, we have the following relation :
Design of Gas Hearing Systems for Precision Applleations 121
(A6.2)
or
Llliw = (ks/kg) tJ (A6.2a)
where kg the stiffness of the pad mounted in the opposite side of the
soft conneetion
h : the stiffness of soft conneetion
Llliw : the change of bearing gap height in the rigid conneetion
Lll : the displacement of the soft spring
From the equation (A6.2a), it can be easily seen that if the stiffness
ratio (ks/kg) is very small the pad gap change will be very small, which can
be illustrated by Fig. A6.2.
F F
6F
l hw
61 6hw
h = LlF/Lll kg = ÁF/Llliw
a) soft spring b) gas film
Fig. A6.2. Illustration of the stiffness characteristics of the soft

spring and gas film pair
APPENDIX 7.
THE EQUATIONS OF VlDRATION MOTION OF THE SLIDE
In order to present the general forms of the equation of vibration

motion, we define the stiffness vector, k, damping vector, D and the hearing
position vectors, ax, ay, az. In the damping factor model, the stiffness
vector bas the length of the numbers of hearing used in a gas hearing system.
In the three element model, the length of the stiffness vector is doubled. If
soft springs are introduced, the length of the stiffness vector is extended to
the sum of the original length plus the numher of the soft springs. However,
the length of the damping vector is equal to the numher of hearings in most
cases in which no additional damping is introduced. Por convenience, the
orders of the stiffness and damping veetors follow the rule: the closer the
stiffness component to the guide, the earlier the number. Therefore, the plate
springs were numbered as 7 (6) to 10 (8) in the contiguration I (II & III).
With those notations and referring to the coordinates defined in Fig.
5.6, the equations of vibration motion for the three element model read :
The motion in Y -direction :
ms y" + Sk(3,10) y -Skz(3,10) a+ Skx(3,10) 'Y + SD(3,6) y + SDz(3,6) à

4 .
- SDx(3;6) y -;; [k(i+6) l;i + k(i+2) ~i+ d( i +2) ~;] = 0
1
The motion in Z-direction :
" Sk(l ,2) y + Skz(l ,2) a - Skx(1,2)

ms z+ ~ + SD(1,2) z+ SDz(l,2) a
, 2 •
Skx(l ,2) ~ - }; [k(i) ~; + d(i) ~;] =0
The rotation about X-axis :
11
Jxa-Sky(l,2) z + Skz(3,10) y + [Skyy(1,2) + Skzz(3,10)] a

-Skxy(1,2) á.- Skxz(3,10) i'+ SDy(1,2) z- SDz(3,6) y +
- [SDyy(1,2) + SDzz(3,10)] á. SDxy(l,2) ~- SDxz(3,10) i' -

4
-.}; [k(i+4) az(i+4).1;i + k(i+4) az(i+4) l;i + d(i+2) az(i+2)l;i]
•• 1
2
\~
1 [k(i) ay(i) l;i + d(i) ay(i) l;i] =0
Hearing Systems for Precision App6cations 123
The rotation about Y-axis :
Jy ~+ Skx(1,2) z + Skxy (1,2) a +Skxx(1,2) 13- SDx(1,2) z- SDxy(1,2) a-

• 2
- SDxx(3,10) 13 - .:El [k(i+2)
I=
~; + d(i+2) ~l] = 0
The rotation about Z-axis :
Jz y- Skx(3,10) y + Skxz(3,10) a - Skxx(3,10) y - SDx(3,6) y- SDxz(3,10) a

4
- SDxx(3,10) 1 -.:E [k(i+6) ax(i+6) l;i + k(i+2) ax(i+2) ~i+ d(i+2)
I • 1
ax( i +2) ti] =0
The motion of plate springs, j = 1, 4(3), i = 7(6), 10(8), for configuration I

(11 & lil) :
mb fij - k(i) y + k(i) ~(i) a + ~(i) ax(i) y + [k(i) + k(i+4)] ~

- k(i) ti + d(i) ~ - d(i) ti =0
The motion of two top hearings, i = 1,2 :
- k(i) z + k(i) ay(i) a - k(i) ax(i) 13 + [~(i) + k(!+ 14)] ti

- d(i) z + d(i) ay(i) a - d(i) ax(i) j3 + d(i) ti = 0
The motion of side bearings, i = 3, 6 (5), for configuration I (11 & lil) :
- k(i) y - k(i) az(i) a + k(i) ax(i) y + [k(i) + k(~+ 14)] ~i

- d(i) y - d(i) az(i) a + d(i) ax(i) 1 + d(i) ti = 0
The motion of hearings connected to plate springs, i = 7(6), 10(8), j 1, 4

(3) for configuration I (11 & ITI) :
. .
-k(i) ~ + [k(i+4) + k(i+ 14)] ti - d(i +4) ti + d(i +4) l;i =0
The equations of vibration motion for the complex stiffness model read :
The motion in Y-direction :
ms y" + Sk(3,10) y + Skz(3,10) a - Skx(3,10) y + j [SD(3,6) y + SDz(3,6) a

4
- SDx(3,6) y] 1
-;~ k(i+6) l;i =0
124 APPENDIX 7. THE EQUATIONS OF VIBRATION MOTION OF TUE SLIDE
The motion in Z-direction :

11
ms z + Sk(1,2) y + Skz(1,2) a - Skx(1,2) f3 + j [SD(l,2) z + SDz(l,2) a

- Skx(1,2) f3] = 0
The rotation about X-axis :
Jx a" - Sky(l,2) z + Skz(3,10) y + [Skyy(1,2) + Skzz(3,10)] a - Skxy(1,2) f3

- Skxz(3,10) y + j {SDy(1,2) z - SDz(3,6) y + [SDyy(1,2)+ SDzz(3,10)] a
4
- SDxy(1,2) f3 - SDxz(3,10) 1l - .l: k(i+4) az(i+4) 1;i = 0
1 =1
The rotation about Y- axis :

Jy ~ + Skx(1,2) z + Skxy (1,2) a + Skxx(1,2) f3 - j[SDx(1,2) z + SDxy(l,2) a
+ SDxx(3,10) f3 ] = 0
The rotation about Z-axis :

Jz y" - Skx(3,10) y + Skxz(3,10) a - Skxx (3,10) 1 -j[SDx(3,6) y + SDxz(3,10) a
4
- SDxx(3,10) y] - .l: kl(i+6) ax(i+6) 1;i
I= 1
=0
The motion of plate springs, i = 1, 4 (3) for contiguration I (ii & III) :
" - k(i+6) y + k(î+6) az(i) a + k(i+6) ax(i) y + [k(î+6) + k(i+ 10)] 1;i
mb 1;i
+ j d(i+lO) l;i =0
where
j j
S m (i,j) : l:.m(k), Smn(i,j) : l:.m(k) an(k),
k•t k=t
j
Smnl(i,j) :k~im(k) an(k),at(k)
n=x, y, z; m = k, d
k, d : stiffness and damping veetors
x, y, z, a, f3, 1 : coordinates for movable frame
1;; : coordinates for plate springs
l;; : coordinates for hearing zero mass elements
ms : mass of moving part of the slide
mb : mass of hearing attatching to the soft spring
According to these equations, one can easily build the detailed matrices
for a specific contiguration and the selected model.
APPENDIX 8.
TFST RESULTS OF VmRATION OF THE SLIDE
Fig. A8.1 shows the test results at the reference point to determine the
vibration modes in the Z-axis and in the rotations about X- and Y-axes. The
vertical axis in the figures is the response of magnitude. The slide was
excited with the exciter and the response was measured with an accelerometer.
Two natura! frequencies with relative high value can be clearly identified,
i.e. 50 Hz and 350 Hz. The natural frequencies of slide are located between
these two frequencies. Therefore, they present no trouble on the
identification of the slide's natural frequencies.
450.00
-.
.....
0
c
0.0
o.o 800.00
Fig. A8.1. The test result at the rejerence point on the granite
Fig. A8.2 shows a test result for determining the vibration modes in the
126 Design of EPG Hearings for Precision Applications
Y-direction. The slide was excited with the hammer and the response was
measured with an accelerometer. From the figure, it can be seen that the
natural frequency can be readily determined. Two relative low peaks are the
viberation of basis and the third one is the rotation about X-axis.
TRANS fA: 6
2.5000
.,r
I
0.0
80.000 HZ 170.00
Fig. A8.2. A test result to determine the natura/ frequencies in Y-direction
Fig. A8.3 shows the test results on checking the rigid body assumption of
the frame. The slide was excited with the hammer and the response was measured
with an accelerometer. The top figure is the response when the slide was
excited on the right-hand side of the slide and the bottorn figure is the
result excited on the left-hand side of the slide. From Fig. A8.3, it can be
seen that two figures have the exect same number of paeks and first one with
APPENDIX 8. TFST RFSULTS 127
the same magnitude. It can be concluded that the slide can be considered as a
rigid body. For the additional information, the coherence tunetion of the test
results is shown in Fig. A8.4.
TRANS IA:. 6
BOO.OO
""0
0
e~
0.0
140.00 HZ 320.00
TRANS
1.4000
i
0.0
140.00 HZ 320.00
Fig. A8.3. Test results on slide in the different position

COHER IA:. B
900.00
""0
'"'"
........
le
0.0
140.00 HZ 320.00
Fig. A8.4. The measured coherence of the test results shown in Fig. A8.3
REFERENCES
1828
* Willis, R. B. A. "On the pressure produced on a flat plate when opposed
to a stream of air issuing from an orifice in a plane surface",
Trans. of Camb. Phi. Soc., Vol. 3, no. 1, pp 129 - 141
1854
* Hirn, G. A. "Sur les pricipaux phenomenes qui presentent les frottement
mediats", Bull. Soc. lnd. Mulhouse, Vol. 26, no. 129
1886
* Reynolds, 0. "On the theory of lubrication and its application to Mr.
Beauchamp Tower's experiments, including an experimental determination
of the viscosity of olive oil", Phil. Trans. R. Soc. A., Vol. 177
1892
* Wood, W. H. "Counter balance for journals", U. S. A. Patent 46664531
1949
* Cope, F. "The hydrodynamic theory of film lubrication",
Pro. Roy. Soc. A, Vol. 197, pp 201 - 217
1959
* Fuller, D. D. editor, 1st International Symposium on Gas-lubricated
Bearings, Washington D.C., USA
* Rothe, H. C. "Air hearings for guidance components of ballistic
missiles and their production aspects", Paper in 1st International
Symposium on Gas-lubricated Bearings, Washington D.C., USA
1962
* Dowson, D. "A generalized Reynolds equation for fluid film lubrication",
Int. J. Mech. Sci., Vol. 4, pp 159 - 170
1963
* Kao, H. C. "A theory of self-acting gas lubricated hearings with heat
transfer through surface", J. of Basic Eng., pp 324 - 326
1967
* Wilcock, D. F. "Extemally pressurized hearings as servomechanism
I - The simple thrust hearing", J. of Lub. Tech., pp 418 - 424
1968
* Castelli, V., and J. Pirvics "Review of numerical method in gas hearing
film analysis", J. Lub. Tech., pp 777 - 792
Design of Gas Bearing Systems for Precision AppUcations 129
1969
* Constantinuscu, V. N. "Gas Lubrication", Publisbed by ASME
* McCahe, J. T., Elord, H. G., Carfago, S. and R. Colsher "Summary of
investigations of entrance effects of circular thrust hearings",
Paper 17, 5th Gas Bearing Symposium, Southampton, UK
* Jona, M. G. "Non-linear effects and pneumatic hammer in a recessed
thrust hearing", Sth Gas Bearing Symposium, Southampton, UK
* Vorh, J. H. "A study of inherent restrictor characteristics for
hydrastatic gas hearings", Paper 30, 5th Gas Bearing Symposium
1971
* Stowell, T. B. "Application of control theory to pneumatic hammer in
externally pressurized gas-lubricated thrust collar hearings",
Paper 2, 6th Gas Bearing Symposium, Southampton, UK
1972
* Sparrow, E. M., Beavers, G. S. and I. T. Hwang "Effect of velocity
slip on porous-walled squeeze films", J. of Lubr. Tech.
1975
* Stiffler, A. K. and D. M. Smith "Dynamic characteristic of an inherently
compensated, square, gas film hearing", J. of Lub. Tech., pp 52 - 62
1976
* Bennett, J., Hudson, B. G. and H. Marsh "The flow characteristics of
small orifices used in externally pressurized gas hearings", Paper E3,
7th Gas Bearing Symposium, Southampton, UK
* Blondeel, E., Snoeys, R. and L. Devrieze "Externally pressurized
hearings with variabie gap geometries", Paper E2, the same as above
* Haycock, R. "Statie and dynamic characteristics of a precision gas-
lubricated interferometer slide", Paper Cl, the same as above
1978
* Zhang, Z. Y. and B. S. Shi "Refrigeration and low temperature
technology",(text book in Chinese), Publisbed by the Machinery Press
* Kanai, A. and M. Miyashita "Nanometer positioning characteristics of
closed looped differential hydro- or aerostatic actuator",
Annual of CIRP
1980
* Blondeel, E., Snoeys, R. and L. Devrieze "Dynamic stability of extemally
pressurized gas hearings", J. Lub. Tech., Vol. 102, pp 511 - 519
130 REPERENCES
* Gross, W.A., Matsch, V., Castelli, V., Eshel, A., Vohr, J. H. and
M. Wildmann "Fluid Film Lubrication", A Wiley International PubHeation
* Pan, C. H. T. "Gas Bearings", Chapter 4. of < <Tribology, Friction,
Lubrication and Wear», edited by Szeri, A. Z.
1981
* Mori, A. and H. Mori "Modified Reynolds equation and boundary conditions
for analysis of dynamic properties of externally pressurized gas
hearings", Paper 16, Gas Bearing Symposium, Lanchiester, U.K.
1982
* Yahya, S. M. "Fundamentals of compressible flow", Wiley Basten Limited
1983
* Li, Y. D. and J. M. Wang "Experimental studies of double semi-spherical
aerostatic spindles", (in Chinese), Paper for Second National Conference
on Gas Bearing Technology, Changchun, China
* Liu, D., Liu, Y. H. and X. W. Shi "The finite element methods for
extemally pressurized gas hearings", (in Chinese),
lntemal Report, no. 155, Harbin University of Technology, China
* Wen, Z. Z., Lin, J. and X. C. Qiao "An improvement of the Powell method
for design of aerostatic hearings with orifice restrictors", (in Chinese)
lntemal Report, Xi'an JiaoTong University, China
1984
* Patir, N. and H. S. Cheng "An average flow flow model for determining
the effect of three dimensional isotropie on partial hydrodynamic
lubricating films", J. Lubr. Tech., Vol. 100
* Tender, K. "A numerical assessment of the effect of striated roughness
on gas lubrication", J. of Tribology, Vol. 106, pp 315 - 321
1985
* Pande, S. S. "Analysis of tapered land aerostatic thrust hearings under
conditions of tilt and rotation", Wear, Vol. 104, pp 297 - 308
* Plessers, P. "Dynamic instability of aerostatic hearings in mechanical
systems", Ph. D. Thesis, Catholic University, Leuven, Belgium
* Roblee, J. W. "Design of externally pressurized gas hearings for dynamic
applications", Ph. D. Thesis, California University, USA
1986
* Cuvlier, C., Segal, A. and A. A. van Steenhoven "Finite Element Method
and Nervier-Stocks Equation"
* Holster, P. L. and J. A. H. Jacobs "Theoretica! analysis and experimental

verification of the static properties of externally pressurized air-
hearing pads with load compensation", 9th Gas Hearing Symposium, USA
* Roblee, J. W. and C. D. Mote jr. "Vibration damping in externally
pressurized gas hearings"Paper for International Conference on Vibration
Problem in Engineering, June, 17-20, Xi'an China
* Shimokohhe, A., Aoyama, H. and I. Watanahe "A high precision
straight-motion system", Precision Engineering, Vol. 8, no. 3
1987
* Fukui, S. and R. Kaneko "Experimental lnvestigation of Externally
Pressurized Gas Hearings Under High Knudsen Number Conditions",
ASME Paper 87-trib-15
* Gans, R. F. "On random Reynolds roughness", J. of Tribology, Vol. 109
* Kakaç, S., Shah, R. K. and W. Aung "Handbook of single-phase convective
heat transfer", John Wiley and Sons, lnc.
* Li, W. and C. H. T. Pan "A new numerical technique for the analysis of
lubrication films", IMechE Paper C225/87
* Liu, D., Liu, Y.H., Shi, X. W. and J. Y. Li "The finite element and
boundary element methods in gas lubrication", IMechE Paper, C240/87
* Wang, J. M. "On the principle and possible methods to improve aerostatic
hearing stiffness", Master Thesis, Eindhoven Univarsity of Technology
(TUE), The Netherlands
1988
* Arakere, N. and H. D. Nelson "An interlor collection method for static
and dynamic analysis of finite length gas joumal hearings",
J. Tribology, Vol. 110, pp 456 - 461
* Kato T. and Y. Hori "A fast method for calculating dynamic coefficient of
fini te width joumal hearings with quasi Reynolds boundary condition",
J. Tribology, Vol. 110, pp 387 - 393
* Majumder, M. C. "Study of the pneumatic instability of externally
pressurized gas thrust hearings with slip velocity", Wear, Vol. 124
* Plessers, P. and R. Snoeys "Dynamic stability of mechanica! structure
containing externally pressurized gas-Lubricated thrust hearings",
J. of Trib., Vol. 110, pp271 - 278
* Wang, J. M., Kodde, L. and M. J. W. Schouten "Approach to the high stiff
and stabie aerostatic hearings", Proceedings of Conference on High Speed
132 REPERENCES
Technology, Lappeeranda, Finland

1989
* Mitsuya, Y., Ohkubo, T. and H. Ota "Averaged Reynolds Equation extended
to gas lubrication possessing surface roughness in the slip flow regime:
approximation method and confirmation experiments", J. of Trib., Vol. 111
* Newland, D. E. "Mechanica! vibration Analysis and computation",
Longman Scientific & Technical
* Raad, P. E. and J. E. White "Entrance and stationary roughness effect on
the load carrying capacity of a wide wedge gas hearing",
J. of Trib., Vol. 111, pp 41 - 48
* Teeuwsen, J. W. M. C. "Performance evaluation and a quality control
system for three coordinate measuring machines", Ph. D. Thesis, TUE
1990
* Faires, V. M. and C. M. Simmang "Thermodynamics", Sixth Edition,
Collier Macmillan International Editions
* Lin, J. F., Jou, J. C. and G. S. Lu "Two-dimensional flows in gas
hydrostatic hearings in thin films of variabie thickness analyzed by
finite element method", Tribology Int., Vol. 23, no. 5, pp 351 - 360
* Scharrer, J. K. and I. Hibbs Jr. "Flow coefficient for the orifice of a
hydrastatic hearing", Trib. Trans., Vol. 33, pp 543 - 550
* Teague, C. and C. Evens "Patterns for precision instrument/machine
design (mechanica! aspects) ", Tutorlal in ASPE Annual Meeting
* Wang, J. M. "Design of externally pressurized gas hearings: Theory and
Practice", Intemal Report, no. WPA 0921, TUE
* Wang, J. M. and P. H. J. Schellekens, (a) "Dynamic natural frequency and
stability of hearibg systems", Presented in the 45th STLE Annual
Meeting, Dever, Cororado, USA
* Wang, J. M. and P. H. J. Schellekens, (b) "Stiffness characteristics of
externally pressurized gas hearings", Proceeding of 6th ASPE Annual
Meeting, Rochester, New York, USA
1991
* Holster, P. L., Jacobs J. A. H. and J. Roblee "The messurement and finite
element analysis of the dynamic stiffness of nonuniform clearance, gas
thrust hearings", J. of Trib., Vol. 113, pp 768-776
* Hagreaves, D. J. "Surface waviness effects on the load-carrying capacity
of rectangular hydrastatic pads", Wear, Vol. 145, pp 137 - 151
* Snip, J. G. "Development of experimental set-up for testing static and

dynamic properties of aerostatic hearing pads", (in Dutch),
Internal Report, no. WPA 1090, TUE
* Wang,J. M. and P. H. J. Schellekens "Design of high perfonnance
externally pressurized gas hearings", Progress in Precision Engineering,
Proreeding of 1PES6/UME2, Braunschweig, Gennany
* Yang,J. H. "Modeling of dynamic perfonnance of linear guide with
externally pressurized gas hearings", Internat Report, no. WPA 1152, TUE
1992
* Dricssen, E. G. M. "The modeling and design of air hearing guide for
precision applications", (in Dutch), Master Thesis, TUE
* Soons, J. A, Theuws, F. C. and P. H. J. Schellekens "Modeling the
errors of multi-axis machines: a general methodology",
Precision Engineering, Vol. 14, no.l
* Wang, J. M. and P. H. J. "Tri-conical gap shaped externally
pressurized gas hearing", to be publisbed in Trans. STLE
SUMMARY
Extemally pressurized gas (EPG) hearings are widely used in precision

machines. However, in practice, a lot of problems regarding both industrial
applications and sdentific research have not been fully solved, some of them
are even little understood. The motivation and aim of the research are to try
to solve some of the problems. Based on providing up-to-date knowledge and
efficient tools, especially for precision applications. The research shows
that the developed software and experimental system can be effectively used in
the design, study and testing of both individual EPG hearings and EPG
hearing systems.
In the beginning of this thesis, the basic concepts of gas hearings are
presented as the general introduction. These concepts include the working
principle, short history, properties and the definition of the current
research. The main body of the thesis consists of two parts: Part one presents
the design methods of single pads, as well as the numerical and experimental
research on some special aspects, including the effects of pad body tilt,
surface imperfections, motion velocity and temperature drop. Part two
describes the principles in the design of more accurate systems with EPG pads,
as well as simulation and experiments on linear motion systems.
Part one presents the theoretica} basis and the results of numerical
simulation and experimental verification of single EPG hearings, including
descriptions of effects of hearing body tilt, surface imperfection,
temperature drop in the downstream restrictor and the refinement of the
restrictor region in hearings. Based on the theoretica! basis the design
software was developed with solving two dimensional solutions of the
time-dependent Reynolds equation. The detailed studies include optimizations
of hearing gap shape, effects of hearing body tilt, surface imperfections,
motion velocity and temperature drop. The tri-conical gap shaped pads are
superior to the parallel and bi-conical gap shaped pads in all aspects: high
load capacity, high stiffness, wide stability region, less sensitive to tilt,
surface imperfection and motion velocity, and low temperature drop.
Part two discusses the conceptual arguments and the roodels to design EPG
hearing systems. At first, the design principles such as loop stiffness,
symmetry, kinematic mounting and driving and preloading are discussed in
detail. Two roodels were developed to descrihe the gas film stiffness for
modeling the dynamic hehavior of a gas hearing system. The solution methods
and techniques to use the roodels were also discussed in this chapter.
Thereafter, it presents the detailed study on a large linear guide with a
comparable size to a guide used in CMMs, particularly dealing with the
constructional details in the design of linear guides and the determinations
of the natural frequencies of the gas hearing systems. In this chapter, the
comparisons of the experiments and the model predictions were presented in
three different configurations of the same linear guide. All the experimental
results are wen correlated to the model predictions.
SAMENVATriNG
Uitwendig van druk voorziene gaslagers, doorgaans aangeduid als luchtlagers, worden
momenteel veelvuldig toegepast in precisiemachines. In de praktijk echter zijn vele
problemen met betrekking tot de industriële toepassing van en het wetenschappelijk
onderzoek naar luchtlagers nog niet volledig opgelost. Over enkele aspecten is er zelfs
nauwelijks kennis aanwezig. De motivatie en het doel van dit onderzoek zijn daarom
gericht op het oplossen van deze problemen door het verkrijgen van meer inzicht in en
het ontwikkelen van efficiëntere gereedschappen voor luchtlagers met name voor
precisie toepassingen. Het onderzoek heeft o.a. geleid tot de ontwikkeling van een
softtwarepakket en een experimentele opstelling waarmee op effectieve wijze zowel
individuele luchtlagers als luchtlagersystemen bestudeerd, ontworpen en getest
kunnen worden.
In de algemene inleiding van dit proefschrift worden allereerst de basisaspecten van

luchtlagers gepresenteerd. Deze aspecten hebben betrekking op het werkingsprincipe,
een kort historisch overzicht, luchtlagereigenschappen en de definiëring van het
huidige onderzoek. Het hoofdgedeelte van dit proefschrift bestaat uit twee gedeelten:
het eerste gedeelte behandelt ontwerp methodieken voor individuele luchtlagers
alsmede het numeriek en experimenteel onderzoek naar enkele specifieke aspecten,
zoals de invloed van lagerkantelingen, oppervlakte-onnauwkeurigheden, snelheid en
temperatuurdalingen. Deel twee beschrijft de ontwerpprincipes voor meer
nauwkeurige systemen met luchtlagers en de simulatie en experimenten m.b.t. lineair
bewegende systemen.
Deel een geeft de theoretische basis en de resultaten van numerieke simulatie en

experimentele verificatie van individuele luchtlagers, inclusief de beschrijving van de
invloed van lagerkanteling, oppervlakte-onnauwkeurigheid, temperatuursdaling in de
"downstream" restrictie en de verbetering in het restrictie gebied van luchtlagers. Op
basis van de theoretische kennis is een softwarepakket ontwikkeld voor het geven van
een 2-dimensionale oplossing voor de tijdsafhankelijke Reynolds-vergelijking.
De gedetailleerde studies bevatten o.a. de optimalisatie van de luchtlagerspleet, de
invloed van lagerkanteling, oppervlakte onnauwkeurigheid, temperatuurdaling en
snelheid. Luchtlagers met een "triconical" lagerspleet blijken in vergelijking tot
luchtlagers met parallelle en "bi-conical" lagerspleten over veel betere eigenschappen

te beschikken en wel wat betreft alle aspecten: hoge belastbaarheid, hoge stijfheid,
stabiliteit over een groot gebied, minder gevoelig voor kanteling,
oppervlakte-onnauwkeurigheden, snelheid en temperatuursdalingen.
Deel twee behandelt de ontwerpregels en de modellen voor het ontwetp van

luchtlagersystemen. In de eerste plaats komen ontwetpprincipes met betrekking tot de
stijfheid, symetrie, kinematische ophanging, de aandrijving en de voorspanning in
detail aan de orde. Twee modellen zijn ontwikkeld voor de beschrijving van de
stijfheid van het luchtlager ten behoeve van de modellering van het dynamisch gedrag
van luchtlagersystemen. De oplossingsmethodieken en technieken voor deze modellen
komen eveneens aan de orde in dit hoofdstuk. Vervolgens wordt een gedetailleerde
studie van een grote lineaire geleiding gepresenteerd; welke met name gericht is op de
constructieve details in het ontwetp van lineaire geleidingen en op de bepaling van de
eigen frequenties van luchtlagersystemen. De betreffende geleiding is wat afmeting
betreft, vergelijkbaar met geleidingen welke in coördinaten meetmachines gebruikt
worden. In dit hoofdstuk is hierbij een vergelijking gemaakt tussen het experimenteel
bepaald gedrag en de door het model voorspelde gedrag van de geleiding en wel voor
drie verschillende configuraties van dezelfde geleiding. Alle experimenteel bepaalde
gegevens komen goed overeen met de voorspellingen aan de hand van het model.
CURRICULUM VITAE
PERSONAL
Name: Junming, WANG
Bom: April 21, 1960 in Acheng, Heilongjiang, China
EMPWYMENT
4/1990 - 11/1992 BOTECH, Boogers Technology in Precision Instrument
and Metrology Systems, Helmond, The Netherlands
3/1988 - 3/1990 IVO, Eindhoven University of Technology, Eindhoven
The Netherlands
10/1988- 12/1988 Eindhoven University of Technology, Eindhoven
8/1982 - 10/1986 Research Institute of Opties and Electronics
Chinese Academy of Science, Chengdu, China
EDUCATION
3/1988 - 11/1992 Ph. D. in the Section for Precision Engineering
Department of Production Technology & Automation
Faculty of Mechanica! Engineering
3/1988 - 3/1990 Further Education under Subject of the Computer Aided
(Simultaneously) Design and Manufacturing for Discrete Products
IVO, Eindhoven University of Technology
6/1987 - 1/1988 ir. (M.Sc.) in the Section of Power Transmission and
Tribology, Department of Design Engineering
Faculty of Mechanica! Engineering
10/1978- 7/1982 B.S. in Section of Precision Production Technology
and Manufacturing Proces..<>, Mechanica! Department I
Harbin University of Science and Technology
Harbin, China
ACTIVITY
Session Chainnan in the Session of Gas and Magnetic Bearings
International Conference in High Speed Technology
Finland (1988)
Member of American Society for Precision Engineering (ASPE)
Member of Society of Tribolagist and Lubrication Engineers (STLE)
Member of Klvl (Koninklijk Instituut van Ingenieurs)
STATEMENTS
DFSIGN OF GAS HEARING SYSTEMS FOR PRECISION APPLICATIONS

J. M. WANG
With the ever increasing demands on machine accuracy, also systems using
gas hearings have to provide better perfonnance. This can only he achieved by
paying more attention to the practical details and the development of
efficient tools to design gas hearing systems.
- this thesis
11
In the design of a closed-loop controlled gas hearing system, the ideal

position for mounting the actuator is at the joint between the hearing pad and
the structure.
- this thesis
III
High film stiffness of a gas hearing is easily lost by a poor design of

the joint, which unfortunately occurs in many applications, even in commercial
products.
- this thesis
IV
The use of the concept of error correction in machine tools is a

fundamental step towards higher precision in production.
- J. A. Soons. F. C. Theuws and P. H. J. Schellekeus
«Precision Engineering» Vol. 14. No.l 1992
V
A building bas to be as comfortable as our skin; it lives, it breathes,

it surrounds us without oppressing us. A building bas to be as healthy as we
want our body to be.
- Ton Albects <<Healthy Buildings > >
© Longman Group UK Limited 1992
VI
A better strategy to solve the deadlock problem in computer operating

systems can only be derived trom a fuller understanding of the deadlock
problem itself. This also holds true for the problem solving in human
activities, as stated in the book <<The Ideal Problem Solver> >, by J.D.
Bransford and B. S. Stein.
VII
As the new millennium approaches, the crucial need of the human race is
to find a unifying vision of the nature of man and society.
< < Bahá'u'lláh> > Bahá'i International Community,
Office of Public lnfonnation, New York, 1991
VIII
What exists serves for possession. What does not exist serves for
effectiveness.
- < <Tao De Jing> >
IX
Symmetry is not only a form of natural beauty, but also a fundamental

rule to avoid mistakes, e.g. in the design of precision machines.
Eindhoven, 10th October 1992

Design of Gas Bearing Systems For Precision Applications: WWW - Tue.nl/taverne

Uploaded by

Copyright:

Available Formats

Design of Gas Bearing Systems For Precision Applications: WWW - Tue.nl/taverne

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Design of Gas Bearing Systems For Precision Applications: WWW - Tue.nl/taverne

Uploaded by

Copyright:

Available Formats

Design of gas bearing systems for precision applications

Citation for published version (APA):

Document status and date:

Please check the document version of this publication:

Take down policy

Download date: 02. dic. 2021

ter verkrijging van de graad van doctor aan de

dtuk: wibro dissertatiedrukkerij~ helmond.

prof. dr. ir. P. H. J. Schellekeos

dr. ir. J. W. M. C. Teeuwsen

I am also obliged to ing. P. L. Holster, Philips Research Laboratories,

Finally, to Xiao-wei, my wife, for her patience, encouragement, and love

PART ONE: DESIGN OF EPG BEARING PADS

3.2.3. Bearing design software 31

PART TWO: DESIGN OF EPG BEARING SYSTEMS

Lx,Lz : half-wave length of a surface wave m

: b= 1 in resttictor area, b=O others

1.1. GAS BEARINGS AND THEIR HlSTORY

An .§xternally ~essurized Qas (EPG) hearing is a kind of gas hearings in

Sealing element Motion Coupling

Fig. 1.1. Schematic illustration of EPG hearing family

The practical examples can be found in the following machines :

-- roughness and roundness measuring machines

1.3. ADVANTAGES AND LIMITATIONS

components in a machine. However, at the lower end, the temperature scale

1.4. TUE SCOPE OF TUE PRESENT RESEARCH

2). detailed study into the electronically controlled restrictor;

The following topics closely related to the applications in Precision

2.2. THE BASIC EQUATIONS

2.2.1 The equation of motion

CIRCULAR PAD RECTANGULAR PAD

Fig. 2.1. Illustration of the coordinate system

From the reference [Wang, 1990], the dimensionless form reads :

The reierenee parameters should be carefully selected, otherwise the

0 = _ ~ + !!._ (t-t au)

2.2.2. The energy equation for gas lubrication

DTo = -a (% -aTs) + .:::.t:.

The dimensionless farm can be written as :

working gap (pm) working gap (pm)

1.2 ~-------.---------,--------, 0.045 r-------...--------.------,

(b) reetangu/ar EPG pads with orifice restrietars

2.2.3 The Nusselt number in gas lubrication

2.2.4. The eontinuity equation for EPG hearings

Using the mathematica! formula [Constantinuscu, 1967] :

the continuity equation becomes :

a~ih) + :,cJ\u dy+ :ztQw dy=-(QVh+ ~ BQYri)+Q(U~+ W~+ :~) (2.13)

all the inlet restcictors and ö is an operational factor to indicate

If the gas film is bounded by permeable materials, the velocity in the

Q [ Uh:~ +Wh :~ + Vb] + l: öQyri (2.14)

Por the non-permeable hearing walls, the components of slip velocity

disappear and the continuity equation becomes

2.2.5 The Reynolds equation for EPG hearings

By introducing the velocity distributions into the continuity equation,

Investigations [Kao, 1963][Constantinuscu, 1969] showed that the Reynolds

div [- ~;fl p grad(p)+ phU/2] + ~ + f (p) = 0 (2.20)

In genera}, for the circular pads with rotation, it is convenient to use

2.3. SPECIAL ASPECTS

2.3.1. Gap shape

r 0 : restrictor radius, hv. : ·characteristic gap difference,

0.2 L - - - : - ' - - - - : - ' - - - - ' - - - 0.06 L.. , , ._ J