Keynote Papers

Design for Precision: Current Status and Trends

P. Schellekens (2). N. Rosielle
H.Vermeulen, M.Vermeulen, S. Wetzels, W. Pril
Section Precision Engineering, Eindhoven University of Technology, The Netherlands

Abstract
'Design for Precision' reviews the status quo in Precision Engineering and concludes that today's
precision engineers put repeatability at the top of their list. The design rules, patterns or principles,
quoted here from various authors, are all time-proven insights, to get reproducible results with ultra
precision machines and instruments. Modelling and analysis of different concepts, systems, and
components is required to adapt the progressing design or to confirm its adequacy. Expenditure on such
analysis is worthwhile to avoid realisation of an inadequate design. However, creativity is more important
in keeping the cost down by finding other than locally optimised solutions. World-wide, precision
engineers agree on design principles, the challenge is to apply them creatively to obtain a thought-out
design. In today's most accurate machines, advanced techniques are applied for compensation of e.g.
residual geometric errors, errors caused by machine dynamics, or thermo-mechanically induced errors.
Future developments in Precision Engineering require nanometre- or even subnanometre positioning-
and measuring accuracy, demanding new design concepts with integrated control and error
compensation systems.

Keywords: Design principles and features, predictive design, accuracy, repeatability, reproducibility.

1958, when the integrated circuit was invented. The need

Acknowledgements to have more and more transistors on a chip [Hutcheson,
19961 demanded special machines with an extremely low
The authors would like to thank the following contributors: positioning uncertainty of a few nanometres. An example
K. Blaedel H. van Brussel of such a machine is the waferstepper, used to position
J. Bryan D. DeBra the image of the IC on the surface of a silicon wafer. Such
J. van Eijk C. Evans a machine can be achieved only by highly developed
G. Goch R. Hocken design- and manufacturing technologies. Also the
P. McKeown V. Portman increasing need for high density optical recording systems
S. Sartori H.Spaan (DVD), and as a result for optical disk mastering systems,
C. Teague E. Thwaite asks for advanced machines with nanometre uncertainty.
A. van Tooren D. Trumper The parts used in these machines, like bearings, drives
R. Weill G. X. Zhang and beam shaping optics, have to be manufactured to
submicron accuracy. For the manufacturing process,
machines with submicron even down to nanometre
1. Introduction accuracy have to be developed.

Curren!ly, many activities employed in industry and In Metrology, the technology for high precision
institutes are aimed at creating high precision products measurements is developed e.g. measurement software,
and involve high precision processing. The manufacturing error modelling [Soons, 19931, measurement techniques
of these products demands a highly specialised science and measurement strategies. To measure parts and
called Precision Engineering, which is based on the products with sufficient accuracy, measuring machines for
following disciplines: submicron to nanometre accuracy have to be developed,
0 Design for Precision demanding new design skills since existing design-
0 Optical and Mechanical Metrology methods for high precision hardly allow to reach these
0 Precision Manufacturing levels reague, 1989, 19971. Therefore, Metrology as a
basic discipline has faced an enormous growth resulting
'Design for Precision' here is meant as total design in accurate co-ordinate measuring machines, laser
including materials, mechanics, electronics, control, interferometers and nanosensors such as STM and AFM.
thermo-mechanics, dynamics and software. It may be Additionally much analysis software and software for error
indicated as high precision mechatronic design. Design compensation was developed and implemented.
for higher precision is becoming more and more important
due to the rapidly increasing need for high accuracy Precision Manufacturing concerns the realisation of
machines, instruments and consumer products. products with high shape accuracy and surface quality.
[McKeown, 19871, [Evans, 19891, [Gardner, 19911, The accuracy may be at the nanometre level, so both
[Nakazawa, 19941. Today, this tendency is highly machine design and process behaviour must be well
influenced by developments in computer technology, data understood as well as the interaction between process
processing and data storage. This process started in and machine. i.e. the interaction between tool and

Annals of the ClRP Vol. 47/2/1998 557

workpiece. Several machining techniques can be plays a major role in this field [Nakazawa, 19941. An
mentioned here like diamond turning, grinding, lapping, excellent overview about the 'Japanese State of the Art' is
honing, polishing, ion- and electron-beam machining and presented by Taniguchi in the book 'Nanotechnology',
chemical machining. Excellent overviews about machines raniguchi, 19961. As an example the Canon Super
and machine techniques are given by [Gardner, 19911 Smooth Polishing machine (CSSP) may be mentioned
[Nakazawa, 19941, and raniguchi, 19961. New here [Ando, 19921.
developments in this field are in the nanometre region
[Stix, 19961. So we may conclude that there is an increasing need for
high precision machines and products. The trend towards
Although many interesting examples can be given in the higher precision was shown by [McKeown, 19871, [Evans,
field of precision metrology and manufacturing, this paper 19891 and more specifically for machining accuracy by
focuses on the 'Design for Precision' as a basic activity in raniguchi, 19831 and raniguchi, 19961. His famous
Precision Engineering. This discipline has a significant graph predicting machining accuracy is depicted in fig.
history and has several origins. Nevertheless it is clear 1.2.
that the development of astronomy and metrology had a
major influence on the development of 'Design for
Precision' in the early days. An excellent overview about
historical developments is given in [Evans, 19891. The
19th century brought many inventions, especially in the
field of design. Through the development of linear and
circular dividing machines much knowledge was gained.
Many precision machines were designed and built using
famous design principles like 'kinematic design' and the
'Abbe-principle'. In the 20th century the development of
knowledge of design was further increased, stimulated by
the development of all kinds of measuring instruments
and precision machines. An outstanding example of
precision machine design in the USA is the Large Optics
Diamond Turning Machine (LODTM) [Donaldson, 19831.
Also the excellent design work of Bryan should be
mentioned here. He contributed to several projects like
the 84-inch LDTM [Bryan 1979a1. More recently is the
development of the Molecular Measuring Machine
reague, 1989, 19971, shown in fig. 1.I.

Fig. 1.2: Trends in precision machining by Taniguchi

Current trends are well predicted in this graph. Future

trends in Precision Engineering will mainly be determined
by trends in IC technology, information displays and
storage, biomedical engineering, MEMS and consumer
products needs. Developments will continue at a fast
pace, spurred on by micro-engineering and
nanotechnology developments so the need for high
precision machinery will increase in future.

'Design for Precision' will play a major role in the

development of future high precision products and
machinery. Designs will be realised in multidisciplinary
design groups using total design methodology. Due to the
increasing cost of advanced designs, the design must be
'right the first time'. Therefore 'Predictive Design' is a
Fig. 1.I
: Molecular Measuring Machine necessity. This paper will summarise basic information on
'Design for Precision' and indicate the state of the art and
In Europe, the Cranfield Unit for Precision Engineering future trends in this important area of Precision
and Rank Taylor Hobson developed a wide range of high Engineering.
precision machines including the 'Nanocentre' [McKeown,
19901 while at Philips Research Laboratories
(Netherlands) several high precision machines for internal 2. Elements of precision
purposes have been realised since the early 1950s
[Gijsbers, 19801. In Germany and Switzerland, high In precision instruments and machines many parts
precision measuring and manufacturing machines were interact to achieve a final accuracy. Each part contributes
developed by Zeiss and GSIP, the latter going back as far to the total accuracy due to deviations caused by
as 1875. geometric, kinematic and dynamic effects. Although in
practice the interaction between these effects plays an
Japan also has a long history in the development of high important role in the overall system behaviour, here we
precision machines and instruments. Nowadays, Japan will focus on these effects separately.

558
 Keynote Papers

2.1 Definitions depends on moisture ingress. Other environmental effects

that can affect geometry are vibrations, electric and
Throughout this paper the terms instrument and machine magnetic fields. Some materials ‘age‘, frequently resulting
will both be used to indicate an apparatus. Metrological in dimensional change (secular instability).
terms are defined according to the ‘International
vocabulary of basic and general terms in metrology’. In Deviations of perfect shapes are also introduced due to
‘Design for Precision’ concerning machines and the fact that machines are assembled from many parts.
instruments rather than pure metrology, accurate Here, considerations about form- or force closed
positioning and machine tool operations are key items. constructions, as well as the choice between monolithic
Therefore, definitions given below, are extended from the structures and bolted or glued assemblies are essential.
Internationalvocabulary mentioned above: In the case of assembling, parts can be machined very
accurately on feature specific machines [Knapp, 19981,
0 Accuracy of an operation: closeness of the although hysteresis at the interface might negatively affect
agreement between the actual value resulting from the overall reproducibility. In the classical form closed
an operation and a target value of the quantity. method of assembly each part must have tight tolerances,
Accuracy is a qualitative description. otherwise backlash occurs or, in the case of negative
0 Uncertainty of an operation: parameter, associated over-measure, a high and undefined stress level will be
with the result of an operation that characterises the introduced during assembly. Force closed constructions
dispersion of the values that could reasonably be on the other hand solve this problem by demanding
attributed to the quantity. statically correct connections, i.e. kinematic-, semi-
0 Resolution: smallest difference between indications of kinematic- [McKeown, 19981 or pseudo-kinematic- [Smith,
displaying device that can be meaningfully 19921 design connections, thereby highly reducing
distinguished. tolerances on geometrical shapes. Even in force closed
0 Repeatability (of results of operations): closeness of constructions, some geometrical imperfections, like
the agreement between the results of successive squareness and flatness errors of guiding beams, will.
operations of the same quantity carried out under the affect the overall accuracy. However, these errors are
same conditions. reproducible and might therefore be reduced by software
0 Reproducibility: closeness of the agreement between compensation.
the results of operations of the same quantity carried Due to the fact that the mechanical structure of a machine
out under changed conditions. has a finite stiffness, the geometry will change under
loading. Especially if the load changes, both in place and
Other definitions, for measuring as well as manufacturing size, as is the case with moving carriages, this can
machines are given in [ASME 889.4.1-19971 and [ASME severely affect machine performance. With correct
85.54, 19931 respectively, and quantitative descriptions models, these errors can be predicted and compensated
are given in IS0 guidelines. Bryan in his tutorial on ‘Axis for [Spaan, 19951.
of rotation’ describes practical verification methods for
precision spindles and shafts, starting from the late 1930’s Another problem related to geometry, is the fixation of the
till recently [Bryan, 19961. workpiece itself. Both for manufacturing and measuring
machines, the workpiece must be fixed in such a way that
2.2 Geometry it is not significantly deformed by the fixation itself. At the
same time the piece must be rigidly connected to the
In the initial design of machines and instruments, machine frame or table. Also, especially in the case of
geometry results from the designer’s intentions on the manufacturing, thermal expansion of the workpiece must
functions that the machine has to perform. In that first be possible without intolerable stress levels. Comparable
stage, geometry usually consists of primitive shapes, i.e. problems, related to fixation, apply to the mounting of
cylinders or tubes for shafts, beams or closed box sensors in high precision instruments. This is where
structures for supporting loads and flat or cylindrical kinematic- or semi-kinematic design has an important
surfaces for guiding parts. In practice however, these application.
ideal shapes cannot be reproduced. Lines will never be
perfectly straight and circles never completely round due 2.3 Kinematics
to finite machining accuracy. Here, a careful selection of
the machining operation may significantly improve the Machines are usually not static. Different parts perform
accuracy of the part [Knapp, 19981. In general, more different motions, to be described by kinematic
moving axes during machining give more errors, although relationships. These mathematical descriptions of
small movements from an additional axis might be useful constructions and mechanisms only describe what
for compensation of geometric imperfections. happens in theory, based on theoretical lengths, positions
Accuracy is not only affected by macroscopic shape and set-point curves. In practice however, these factors
errors of a part, but also by microscopic deviations, i.e. are realised with a finite accuracy and therefore, the path
surface finish. For many applications this is an essential followed will be different from the ideal one with respect to
factor in the total performance. In contact bearings, the its form, speed and acceleration.
effect of surface finish on wear and friction is obvious. The
effect of surface finish on connections between clamped In modern machines, displacements are often generated
parts is less obvious but essential when properties like by a combination of mechanical parts, such as actuators
stiffness, damping, hysteresis and thermal conductivity and sensors under servo control. Properties like actuator
and diffusivity are concerned. Geometry not only changes power and speed, sensor resolution, control strategy and
in (mechanical) machining operations. Unless sufficient mechanical reproducibility together determine the
isolation is applied, (e.g. seismic, acoustic and thermal accuracy of the prescribed path. In the case that more
isolation) geometry will be influenced by the environment. than one axis is under control, the synchronisationof axes
Most material components for instance, expand and is another factor that affects accuracy. An example is the
deform under the influence of temperature change, and case of milling a circular contour by simultaneously
for unsealed natural granite, the shape of structures highly controlling two orthogonal linear axes.

559
2.4 Dynamics Vermeulen developed a 3D-CMM, (see fig. 3.1), in which,
by using intermediate bodies (A and B), Abbe errors are
The fact that machines are not static but contain prevented in all three directions in the horizontal mid-
accelerating parts, means that dynamic effects can play a plane [Vermeulen, 19981. This machine satisfies the
major role in their behaviour. An effective way of Bryan principle as well, making the machine non-
minimising the effect of acceleration on relative place susceptible for straightness errors of certain guides.
uncertainty is by choosing appropriate motion profiles,
e.g. set-point curves that contain no jump in the second
derivative, like an inclined sine function instead of a
parabolic one. Preventing 'play' or lost motion is also very
effective in reducing dynamic positioning errors. The parts
themselves can also be designed for minimum forces. In
case of rotating parts this means, that symmetry is
important for reducing unbalance and also the mass
moment of inertia should be minimised. In case of
translations the mass should be kept small and driven as
close through the axes of reaction as possible [McKeown,
1973-19971 (see section 3.6).
Another factor that determines the response of a machine
to dynamic forces is stiffness. For minimising forces and
maximising stiffness in general, it is not only the amount
and sort of material that matters but how it is distributed.
Often dynamic disturbances come from outside the
instrument, like floor and acoustical vibrations. In these
cases, the stiffness to mass ratio again is essential in
minimising the response to these inputs. Isolating the
machine from the disturbance (see section 5.7) can Fig. 3.1: 3D-CMM satisfying Abbe- and Bryan principle to
reduce the input itself. a high degree

3. Desiqn principles 3.2 Kinematic design

Maxwell describes kinematic design as: 'The pieces of our
Design of high precision machines has been analysed by instrument are solid, but not rigid. If a solid piece is
a number of people. Pollard described the mechanical constrained in more than six ways it will be subject to
design of scientific instruments in his Cantor Lectures internal stress, and will become strained or distorted, and
[Pollard, 19221. Loewen has put forth a list of major this in a manner which, without the most exact
principles [Loewen, 19801. McKeown has defined the micrometrical measurements, it would be impossible to
'Eleven Principles and Techniques' in [McKeown, 1986, specify' [Maxwell, 18901. Lord Kelvin established
1987, 19971. Teague and Evans have set out basic Cambridge Scientific Instruments firmly on this design
concepts, published as twelve 'Patterns for Precision philosophy in the interest of achieving highest precision at
Instrument Design' [Teague, 1989-19971. On the basis of lowest cost. GSlP used it extensively, especially in
these surveys, section 3 and 5 have been set up. metrology instruments. Pollard stressed the importance of
kinematic design as opposed to the usual machine tool
3.1 Rules of Abbe and Bryan design of couplings in instrument mechanisms, not only to
the user and the minimising of variance, but also to
The Abbe principle was first published in 1890 in [Abbe, economic manufacture [Pollard, 19221 [Pollard, 1929-
18901: 'The measuring instrument is always to be 19511. The importance of kinematic design for today's
constructed that the distance being measured is a straight precision engineers is clearly described in [Blanding,
line extension of the graduations on the scale that serves 19921. McKeown emphasises its importance in his 'Eleven
as the reference...'. The principle is also known as 'the Principles'. Teague states it gs part of his patterns. Less
principle of alignment' [Rolt, 19291, 'the Abbe comparator known world-wide is Van der Hoek, who was bath an
principle' [Reindl, 1964, and 'the first principle of machine employee of Philips Electronics and professor at the
design and dimensional metrology' [Bryan, 1979~1.As a Eindhoven University of Technology from 1962 till 1985.
restatement to cover those situations where it is not His lecture notebook contained about 200 examples of
possible to design 'in line', Bryan defined a generalised sound versus poor design, in which kinematic design is a
Abbe principle as: 'A displacement measuring system key to solution [Hoek, 1962-19861, [Hoek, 1985-19891.
should be in line with the functional point (i.e. centre of Other books that describe kinematic design are [Slocum,
stylus ball or tool tip) whose displacement is to be 19921, [Smith, 19921, [Nakazawa, 19941 and [Koster,
measured. If this is not possible, either the slide ways that 19961.
transfer the displacement must be free of angular motion,
or angular motion data must be used to calculate the Kinematics evolved from mathematics and is hence
consequences of the Abbe offset'. somewhat 'idealised' in its components. e.g. rigid bodies,
straight lines, perfect circles and 'point contacts' etc.
Another basic principle of measurement is the Bryan Nevertheless it is a good starting point for mechanical
principle [Bryan, 1979a1, formulated as: 'A straightness design that is correct in principle. The basis for kinematic
measuring system should be in line with the functional design is constraining the right number of degrees of
point at which straightness is measured. If this is not freedom (d.o.f.), i.e. translations or rotations. For
possible, either the slide ways that transfer the constraining one d.0.f. for rotation, necessary in e.g.
measurement must be free of angular motion or angular coupling elements in feed drives, bellows are typically
motion data must be used to calculate the consequences used [DeBra, 19981. In fig. 3.2 some examples of fixing
of the offset'. one d.0.f. for translation are given. The classical solution

560
 Keynote Papers

is the use of a slender rod (fig. 3.2a). Due to its finite Fig. 3.6 shows how an xy0-stage can be constructed from
length, sideways displacements result in a parasitic three folded leaf springs.
motion in the constrained direction. The same function is
performed without this disadvantage by a folded leaf
spring (fig. 3.2b). In fig. 3 . 2 ~an alternative consisting of 4
rods is shown.

a. b.-" +4 c.

Fig. 3.2: Single d.0.f. constraints for translation Fig. 3.6: xye-stage with three folded leaf springs

Fig. 3.3 shows some examples for constraining 2 d.0.f. for An illustrative example of the application of kinematic
translation. This can be achieved by 2 rods (fig. 3.3a and design is the kinematic support shown in fig. 3.7. The
b) or a hinged leaf spring (3.3~). classic version with six fixed supporting surfaces is
enhanced by making the surfaces flexible around an
elastic hinge [Schouten, 19971. In this way the friction
force along the surface is greatly reduced while the '

contact stiffness normal to the surface is only slightly

reduced. Due to the improved ratio between stiffness and
friction, the hysteresis (see section 5.l)was reduced from
0.42 for the classic non-hinged version to 0.03 pm for the
new version.

Fig. 3.3: Two d.0.f. constraints for translation

Two d.0.f. for translation and one for rotation are

constrained by for example 3 rods (fig. 3.4a and b) or a
normal leaf spring (fig. 3.4~).

@:qsF
a.
a.
b.
-.
C.
Fig. 3.7: Symmetrical kinematic support with TC, where
friction is recognised as principal cause of hysteresis

3.3 Thermal loop

A thermal loop is defined as: 'a path across an assembly
Fig. 3.4: Two d.0.f. constraints for translation, one for of mechanical components, which determines the relative
rotation position between specified objects under changing
temperatures'. In principle. a thermal loop should be as
With combinations of these basic elements kinematic small as possible to minimise the influence of spatial
mechanisms or fixtures can be constructed. Fig. 3.5 gives thermal gradients. Thermal expansion in a thermal loop of
an example of a fixation for a table-top. Due to the a machine can be compensated mechanically by adapting
application of three hinged leaf springs, the table-top is the effective length of machine components and by
constrained in 6 d.o.f., with a thermal centre at the choosing materials with the appropriate coefficient of
intersection of the normals to the leaf springs. thermal expansion, e.g. Harrison's 'grid iron' pendulum
[Evans, 19891. The fixed point or axis, from where
components expand with respect to each other, can be
chosen by creating a thermal centre (TC), such as in fig.
3.5. Although coefficients of thermal expansion are not
better known, than within 0.5.106 'C-' [Breyer, 19911, the
influence of thermal expansion can be reduced by
measuring ,expansion of machine components due to
controlled temperature changes (Kunzmann, 19881 and
subsequently creating equal thermal lengths by choosing
the proper fixed point.
To obtain thermal stability of f 0.5 %/day in an air-
conditioned assembly hall and f 0.1 'C/day in a precision-
climatised cabin is quite a problem [Breyer, 19911. Heat
Fig. 3.5: Six d.0.f. constraints with thermal centre (TC). sources localised internally or externally of the machine

561
lead to changes in the temperature profile in the machine.
This may cause unequal thermal expansion rates in a
thermal loop, due to different thermal time constants of -r hydrostatic
workhead spindle
machine components, (see section 5.5). Hence \ Metrology frame
Donaldson strongly recommends, as a principle in his
publication about Machine Tools [Donaldson, 19801 to
take the heat out at the source. Wetzels experienced a
thermo-mechanical stability problem with an integrated
heat source. After having removed this source thermal
drift was reduced with one order [Wetzels, 19981.
Internal heat disturbances have responses that can be
middled and compensated partially (See section 6 ) .
However environmental temperature variations can only
be responsive and not predictive, because the input is
unknown [de Bra, 19981.
Gal
es
3.4 Structural loop
According to [ANSI, 19921 a structural loop is defined as
'an assembly of mechanical components, which maintain attached to z-carriage
relative position between specified objects. A typical pair
of specified objects is a cutting tool and a workpiece: the
structural loop includes the spindle shaft, the bearings and Fig 3.8: Nanocentre, using a Zerodup metrology frame
housing, the slide ways and frame, the drives, and the tool for complying with Abbe in two axes and direct
and work-holding fixtures'. All mechanical components measurement of horizontal straightness error
and joints in the propagation path from the drive to the motion.
point of reaction, e.g. the end-effector (cutting tool or
probe) or the centre of gravity, must have a high stiffness reague, 1989-19971 advises to make the metrology
to avoid deformations under changing load. The design of frame as small as possible, to minimise environmental
a machine or an instrument contains one or more influences. [Bryan, 1979bl suggests either to build the
structural loops. metrology frame in zerodur or to use temperature control
Essential in recognising good structural loop design is the (e.g. an oil shower). Furthermore the plane of supports of
split-up in series and parallel paths. Along a series path the metrology frame should coincide with the bending
stiffness should not change to abruptly. Improvement of neutral axis of the machine base.
the series path is possible by stiffening the most compliant
part preferably by 'transferring' material from more rigid 3.6 Drive offset
parts. Parallel path improvement is in contrary done by
improving the stiffest part further, preferably -for a system By combining sound mechanical design with closed loop
with equal mass- at the cost of more compliant parallel control, enhanced operating speed, accuracy and
paths. flexibility of movement can be achieved. Typical examples
Due to physical limitations, the measuring system of a are the Compact Disc player, the waferstepper, advanced
closed loop system is unavoidably located at a certain CNC milling and turning machines and fast component
distance from the end effector. In addition to good mounting (FCM) machines. In the development of servo-
structural loop design, the path between measuring controlled positioning devices it is essential to consider
system and end-effector has to be as stiff as possible to the place where the actuator force is loading the slide to
minimise deviations, e.g. by minimising the length of this supply the (resulting) force generated by inertia, tool- or
path, called the 'measuring circle' [Kunzmann, 19961. measuring forces, friction, etc. As stated in one of the
'Eleven Principles' [McKeown, 1986, 1987, 19971, drives
3.5 Metrology frame should be placed to operate through the axes of reaction.
If this is not possible, the deviation from the axis of
A metrology frame is a reference frame for displacement- reaction -called drive offset- induces moments on the
measurements, independent of the machine base, i.e. the machine slides. The effect of resulting rotations of a slide
external forces upon the metrology system must be on it's controllability is minimised, if both the drive- and
constant [Bryan, 1979bl. DeBra suggests to see the measurement-axis are at the same side of the centre of
metrology frame as an example of a broader principle, i.e. rotation [Rankers, 19971.
the principle of 'separate functions' [DeBra, 19981. In fact
the routes for force and position information are 3.7 Force compensation
disconnected, an idea which is present as well in the
design of the rotary table of fig. 5.8 [Philips, 19941.
In [Teague, 1989-19971 historical applications of the 3.7.1 Weight compensation
'metrology frame' are discussed, to overcome the problem In many 3D-CMM's vertical rams are used. In order to
of deformation of machine parts. The first example of a avoid loading the vertical slide's driving system with the
metrology frame found so far is on the Rogers-Bond (constant) weight of the ram, force compensation might be
Universal Comparator [Rogers, 18831. More recent applied, thereby eliminating undesired heat dissipation in
examples are found at NlST in measuring machines by the motor. Constant forces can be obtained in many ways,
Hocken et al, at NPL in interferometric time standard e.g. using additional weights, which is unfavourable from
comparators, at LLNL in the 'Ultimat' CMM [Bryan, a dynamic point of view. Alternatives are magnetic fields,
1979b], the 84" SPDTM [Bryan, 1979al and the LODTM cylinders under pressure or vacuum, or 'constant rate
[Donaldson, 19801, and at Cranfield Precision in the springs', e.g. tensators [Tensator, 19971, [Rosielle, 19981.
Nanocentre by McKeown et al (see fig. 3.8) and Wills- Depending on the stroke of the slide and the
Moren [Wills-Moren, 19821and [Wills-Moren, 19891. specifications on admissible force variations, one type of

562
 Keynote Papers

weight compensation is more suitable than another. Hocken mentions some adversaries of symmetry
Weight 'relieving' in which Coulomb friction is reduced on [Hocken, 19951,e.g. vibrational energy is not reduced by
guide ways, was used in many high precision machines symmetric design, in fact it is often enhanced.
designed by GSIP.
3.9 Repeatability
3.7.2 Reaction force compensation
Due to a limited mass and support stiffness of stationary According to the definition, given in section 2,repeatability
machine parts, reaction forces as a result of a driving means that the results of the same machine action carried
forces, cause a motion of these stationary parts [Rankers, out under the same conditions are equal. The machine
19941, [Rankers, 19971. Excitation of machine frames action can either be a measurement on a CMM or
caused by reaction forces becomes increasingly important manufacturing a product.on a machine tool. [Bryan, 19931
in situations of direct drive actuation at high frequency, propounds determinism i.e. the deterministic performance
e.g. fast tool servo cutting [Patterson, 19851 and cutting of of machines under automatic control, as a sound basis for
non-rotational symmetric surfaces [Renkens, 19971. In design, build and performance testing: 'The basic idea is
addition to commonly used approaches to reduce the that automatic machine tools and measuring machines
motion of the frame, such as increasing stiffness and are perfectly repeatable just like the stars*. They obey
mass of the frame, or increasing the (active) damping, cause and effect relationships that are within our ability to
Rankers mentions more fundamental approaches understand and affordably control. There is nothing
[Rankers, 19971.A first example is to accelerate a counter random or probabilistic about their behaviour. Everything
mass between the load and the frame in opposite happens for a reason. The list of reasons is small enough
direction [Weck, 1995bl and [Weck, 19971. Secondly a to manage by common sense, good metrology, and a
simultaneous counteracting force can be applied to the reasonable investment of resources'. Practically, design
frame by a second motor without tight requirements with for repeatability demands:
respect to positioning accuracy.
Application of statically determinate high stiffness
3.7.3 Parasitic stiffness compensation (negative design, minimised hysteresis in the parts-connections
stiffness) (section 5.1)
Instruments and mechanisms based on elastic elements Minimising of friction and maximising stiffness of the
have the advantage that actual backlash and friction, bearing systems, (section 5.2)
causing virtual play [Koster, 19961 are absent. Sizing of Optimisation of Performance of drives (section 5.3)
flexures is based on allowable stress in combination with and control systems (section 5.4)
desired stroke, considering manufacturability and Consideration of sensor qualities including sensor
necessity to suppress the fixed motions rigidly. However, mounting (section 5.5)
the stroke is restricted by the elastic limit of material and Attention to the thermal stability of the design and
the stiffness of the elastic elements results in an opposing adequate isolation against vibrations (section 5.6).
force linear proportional to the displacement. In those
instances where the driving force becomes too high to Repeatability is essential to predictive modelling (section
handle, be it through the sheer size of actuators required 4). The closer the instrument is resembled by the model,
or their heat production as a disturbance to the machine the better the prediction of it's behaviour and the more
functions, either flexures have to be discarded as a design scope for software error compensation (section 6).
option or this undesired effect of force buildup has to be
compensated. This is best done with passive elements for Here Bryan is quoting [Loxham, 19701. He goes on to
obvious reasons. The classical approach is referred to as summarise seven exceptions to the deterministic nature of
creating 'negative stiffness'. With a device also containing classical physical laws, according to Kidder (LLNL) and
flexures and rigidly connecting this with the rotational or Hocken [Bryan, 19931. These exceptions apply to
translational flexure stage so as to obtain a near zero molecular and atomic size masses and have no practical
stiffness design, this problem can be overcome at the cost significance in the field of manufacturing.
of increased complexity. Van Eijk gives several examples
of creating negative stiffness [Eijk, 19851.
4. Modellina for predictive design
3.8 Symmetry
As indicated in section 1, there will be an increasing need
In [Teague, 1989-19971it is recommended to incorporate for ultra precision machinery in the near future. As a
symmetry to the maximum extent possible in properties of result, .design approaches are changing dramatically in
machine elements (e.9. mass- and force distribution or this field. The complete understanding of the behaviour of
stiffness), the entire instrument and properties of the precision machines is necessary to predict the component
environment. In designing, manufacturing, assembling dimensional errors. By adding all the component errors
and operating a precision instrument, any departure from together in an error budget, the machine designer can
symmetry has to be weighed against the resulting make a prediction of the accuracy of the overall machine
compensation to overcome the problems produced by the [Blaedel, 19981,[Thompson, 19891.A very good overview
asymmetry. To avoid thermal asymmetry, inducing of nowadays machinery and processes is given in the
significant distortions of the machine components, a recently published book 'Nanotechnology' raniguchi,
thermal centre as symmetry axis for thermal expansions 19961.As the development of ultra precision machinery
(see section 3.3)can be applied [Vermeulen, 19981.To generally is very expensive, 'right-the-first-time'- design is
overcome the effect of asymmetry about a horizontal getting more and more important. Although extensive
plane, caused by gravitational forces, machines can be analysis of design concepts is expensive, a lot of money
equipped with a vertical axis, e.g. the LODTM, can be saved by systematic analysis early in the design
[Donaldson, 19831. Three-dimensional symmetry is process. Accuracy of machines and instruments is mainly
superbly achieved by a tetrahedral structure, e.g. the determined by five error sources, i.e. deviations
Tetraform by Lindsey of NPL [Lindsey, 19881, [Slocum, concerning kinematic-, thermo-mechanic-, static- and
19921, [Corbett, 19971. In addition to its proponents dynamic- and control system behaviour (see fig. 4.1).

563
 -x- I I system is connected to the sensor which is reading the
Design specifications
scale position of belonging axis i. Vector , R contains a
fixed contribution from squareness errors between
machine axes as well. The total kinematic error at a
Conceptual design specified machine position may be described by an error
vector to be calculated from contributions of all axes
[Spaan, 19951:dp(x,y,z)=T(,Rx ip+ i i ) .
I

Mechanical system Control

Fig. 4.2 contains a schematic representation of this
system modelling approach for one axis i.

analysis

2+ ?I-
Satisfied?

Final design

P / Y
X
Fig. 4.1: Design flow chart for (components of) machines
and instruments Fig. 4.2:Modelling approach for kinematic errors
Starting from the design specifications, one can create a Kinematic modelling must be applied to the subsequent
conceptual design, which must then be analysed by structural loops of a machine, i.e., one can start at the
modelling and simulation and improved. A flowchart can workpiece position passing through all the elements and
be drawn, showing the kind of modelling- and simulation joints of machine carriages, ending at the probe- or tool
activities, necessary to check the conformance to it's position. Rotary tables and spindles are treated in the
functional specifications. In addition to kinematic models same way as linear carriages. In [Spaan, 19951, the
for estimating effects of motion errors, FEM part modelling modelling technique has been applied to a five axis milling
can be used to study (thermo)-mechanic- and static machine. Its relatively simple approach rapidly offers the
aspects including thermal expansion and diffusivity designer information about the effects of a design
respectively stiffness and strength. For both the first proposal for a change in the structural loop.
approximation and necessary validation, calculations Another modelling method based on a description of
based upon simple elastic and thermal theories can be kinematic errors using homogeneous transformation
used. Dynamic aspects of the mechanical system matrices is widely used as well. Here, the error
concerning effects of both inertia and stiffness have to be
analysed as well. Additionally the control system and the parameters ,Rj, iT, are used as matrix elements [Paul,
mechanical system have to be tuned in order to get good 19811, [Soons 19931, [Portman 19971. In the latter, a
dynamic behaviour of the closed loop system. Fig. 4.1 description of second order effects for high precision
shows design to be an iterative process. The further the applications is presented as well.
design process evolves, the more detailed models and
simulations get. Hereafter, the most important analysis 4.2 Thermo-mechanic analysis
activities and approaches are discussed in brief.
As indicated by many authors, thermal effects may have a
significant contribution to the uncertainty of precision
4.1 Kinematic analysis
machines and instruments Voshida, 19671, [Camera,
Having developed a design concept for a precision 19761, [Attia, 19791, [Balsamo, 19901, [Cresto, 19911,
machine or measuring instrument, error modelling [Schellekens, 19921.An excellent overview of work in this
strategies, being developed mainly during the last two field is given by Bryan et a1 in their keynote papers for
decades, can be used to predict kinematic errors [Soons, CIRP - General Assemblies of 1968 and 1990 [Bryan,
19931,[Krulewich, 1995a,b].These modelling techniques 19681,[Bryan, 19901.
are based on estimating the effects of motion errors of Generally, the temperature profile in a machine can be
e.g. linear carriages, rotary tables and spindles. described by an instationary function T = T(x, y, z, t) . As a
Successful use of lumped parameter models to predict result of these spatial and temporal gradients, in
and compensate for bending deflections were used as combination with thermal properties of the material used,
early as 1972 [Wills-Moren, 19821. For each axis i, shape and size of a machine will change in time.
motions errors can be described by means of a Especially the instationary behaviour makes error
translation- ,iand rotation ,R error vector. The effect of estimations very difficult to describe. Therefore, numerous
rotations of carriages at specified positions in the attempts have been made to describe (quasi) stationary
structural loop may be calculated by the outer vector situations T = T(x, y, z) [Soons, 19931, [Trapet, 1997al.
Here, as a first approach, error modelling is mainly based
product of rotation vector iR and local position vector ,P,
on expansion and bending of (large) machine elements in
being defined in the Cartesian co-ordinate system of the thermal loop, considering a stress-free deformation
specific carriage. Here, the origin of the co-ordinate (otherwise modelling becomes very difficult) [Boley, 19601,

564
 Keynote Papers

[Trapet, 1997al. After having calculated expansions and addition to the way the set point is applied (accelerationor
deformation, the kinematic model discussed in section 4.1 ‘jerk’), the vibrational behaviour highly depends on the
can be used to calculate the thermal error vector system’s natural frequencies and the amount of damping
dp( T( x, y, 2)). In [Soons, 19931, this approach was (see section 5.1.2).
successfully applied to a milling machine, giving a thermal In order to foresee deviations as a result of internal
error reduction of about 70%. vibrations, it is of vital importance to model and analyse
As machine tools unavoidably have internal heat sources, machine dynamics, i.e. to find the lowest natural
it may sometimes be necessary to use instationary frequencies and vibration modes of the mechanical
descriptions T = T(x, y,z, t). From this relationship, thermal structure in the relevant frequency range (limited by
effects on the tool position relative to the workpiece can several times the bandwidth of the servo system). Here,
be calculated. Nowadays, finite element (FE) and finite lightweight- and (dynamically) stiff design is particularly
difference (FD) modelling techniques are widely used important for those machine components that determine
the lowest natural frequencies and the static stiffness.
[Soons, 19931. Soons used state-space models as well, to
predict the temperature field in a five axis milling machine Even reaction forces of actuators on machine frames may
due to heat generation in the main spindle drive. cause intolerable deviations, especially for high precision
However, there are still difficulties in specifying thermal machines, such as wafer steppers or SPDT machines
boundary conditions. In contrast to disturbances as a using fast tool servo systems for the manufacturing of
result of internal heat sources, the modelling of non-rotationally symmetric parts. As the mass of the
environmental effects, such as door opens, can only be frame is limited and its connection to the world suffers
responsive, not predictive. finite stiffness, resonance due to reaction forces may
occur (see section 3.7) Weck, 1995bl [Weck, 19971,
[Renkens 1997, [Rankers 19971. A schematic diagram
4.3 Static analysis summarising the indicated effects is given in fig. 4.3.
The structural loop of machines and instruments can be
affected by (quasi) static forces, e.g. changing weight
/I
forces of slowly moving machine parts, (nominal)
machining forces [Spaan, 19951, and forces as a result of
cabling, air- and vacuum hoses. Acceleration forces,
having much higher frequencies, will be considered in the
next paragraph.
Due to finite stiffness of machine components, e.g.
beams, spindles, carriages, including bearing systems
and joints, the forces mentioned above will cause I I
positioning errors at tools or probes. As a first approach,
deviations can be estimated fairly easy by means of
1 (1
......................
stiffness calculations based on simple linear elastic
theories or theories about Hertzian contacts. Nowadays
advanced software packages are available, such as
Unigraphics@, I-DEAS@, Algor@, Pro-Engineer@ [FEM, Fig. 4.3: Simple dynamic model of driving system with
19981to be used for both linear and non-linear analysis at frame (kF), actuator (ka) and bearing stiffness (kb)
e.g. complex plate frame type structures, beams, trusses (no damping considered).
and 3D solids, made of homogeneous isotropic as well as
composite materials (see section 5.1). Interesting For multibody systems like manufacturing machines and
approaches in this field are reported in [Reinhart, 19971, wafer steppers, modelling and evaluating the machine as
which describes 3D-CAD with integrated FEM analysis a whole may be inefficient and excessively time
early in the design stages, called ‘Virtual Prototyping’. consuming. Therefore, [Rankers, 19971 proposes to divide
the total system in substructures and components
successively, and model and analyse them separately.
4.4 Dynamic analysis Afterwards these models are added to an overall system
Since machine structures generally are assembled out of model, to be used for mode shape analysis resulting in
many different components, to be seen as a combination natural frequencies and vibration modes of the overall
of masses and springs, the total structure will behave system.
according to an interaction of these elements Dynamic models, such as lumped mass models, are very
[Timoshenko, 19741. Many good textbooks are published helpful to analyse the dynamic behaviour of a chosen
about this subject, e.g. [RaO, 19901. concept. Here, the model, which consists of a number of
masses connected by a series of mechanical
Since most driving systems do not act in line with the compliances, can be described by a set of ordinary
centre of mass, inertial forces cause rotations of machine differential equations. Model reduction techniques [Hoek,
parts, mainly due to finite stiffness of bearing systems and 1962-19861, [Nijs, 19881, [Koster, 19961, [Rankers, 19971,
joints. In high precision machines, such as 3D CMMs, [Rosielle, 19981 are used to diminish the complexity of the
even low accelerations values may have significant description, advantageous for mathematical treatment and
influence on measuring accuracy [Weekers, 19971. interpretation of results.
For rather complex (sub)-structures ‘mode shape
Machine dynamics may have significant influence on analyses’ offer an excellent tool to analyse the behaviour
system performance. Position- and tracking accuracy as [RaO, 19901, [Ewins, 19841, [Rankers, 19971. For that
well as on smoothness of motion are highly reduced by purpose, the static model of section 4.3 can be used, for
mechanical step response in the structure. In addition to determining natural frequencies, as well as mass, mass
nominal acceleration levels as a result of inertial effects of moment of inertia, and the location of the centre of gravity.
slides and spindles, vibrational accelerations may For a ceramic plate frame type slide construction, to be
continue after an external force or a set point motion applied in an SPDT machine, the result of a modal shape
profile has been applied to a mechanical component. In analysis is depicted in fig. 4.4 [vermeulen, 1996al.

565
 present, will load the stiffness of components to a
maximum before sliding in any direction. The positional
uncertainty associated is called 'virtual play' [Hoek, 1962-
19861, [Koster, 19961, [Rosielle, 19981, and amounts to
twice the maximum friction force divided by the stiffness.
Uncertainty in position is hence uncertainty in the internal
potential energy content of the construction. Designing for
high stiffness can be conceived as designing for a
minimum potential energy content. 'Bigger is better' might
be true for stiffness reasons, for example when a force
transmitting area of a cross section is concerned. But if
the efficient use of material in a design is also aimed for,
e.g. in high precision slide constructions or aerospace
applications, the mass and thereby the kinetic energy
content should be minimised as well [Riven, 19891,
[Slocum, 19921, [Smith, 19921, [Rivin, 19961. DeBra
emphasises the importance for a design to get it's
functional dependence right, but meanwhile to separate
'material' from 'geometry' [DeBra, 19981. On the basis of
the latter principle, this section about stiffness of
constructions and mechanisms is further divided in
'Geometry at the component level' on the one hand, and
'Material selection' on the other.

Fig. 4.4: Mode shape analysis of plate-frame-typeslide 5.1.IGeometry at the component level
construction Certain movements are desired to obtain functions, others
are preferably avoided. This because the latter result in
4.5 Control system analysis undesired energy storage, e.g. kinetic energy, or potential
energy in elasticity of material components. Vibrations are
In most cases a closed loop control system will be added a typical example of such energies flowing back and forth.
to the mechanical system for positioning and multiaxis If there are many free movements within the design,
contour control. Examples of closed loop control in ultra energy flow becomes chaotic, different vibration modes
precision machines are SPDT machines, ceramic grinding store energy from or release energy to each other.
machines operating in 'ductile mode' as for mag-disc Because this energy flow eventually may affect
substrates, wafei steppers, and optical disk mastering movements, which are necessary for functional reasons, it
machines. In order to get good dynamic behaviour, the is most desirable only to allow functional movements and
control strategy should be matched very carefully to the to suppress deliberately all other movements of
dynamics of the mechanical system. The optimised components in a design. This calls for high stiffness in the
design of the control system highly depends on. the directions in which movements are not functional. Such
behaviour of the mechanical system, e.g. its natural stiffness can best be obtained by correctly placing
frequencies, the presence of friction and other disturbing material in the right shape and direction, and by using as
forces. little material as possible (i.e. stiff-lightweight design),
Generally the control system contains position and especially on functional moving parts. Stationary frames
velocity feedback, and sometimes for the very highest are affected to a much smaller extent by this idea,
performance, velocity and acceleration feedforward loops although here too, saving mass has certain benefits. A
are added as well (see section 5.6). As indicated in design, in which the material involved is used poorly in
[Groenhuis, 19911, the mechanical design mainly terms of deriving stiffness from solids, is often
determines the order of the controlled system. The more characterised by unfavourable loading. If the loading is
degrees of freedom (determined by the number of lumped very localised, large deformations and high local stresses
masses and springs), the higher the order of the model. occur, and unfavourable lever ratios result in high forces
[Rankers, 19971 indicated three main sources of dynamic and undesirably induced bending moments (see fig. 5.2).
influence, i.e. flexibility in the actuator system, guide way In precision equipment, the often-unintended load is
systems flexibility, and mass and stiffness of the machine usually caused by inertial forces, moving-weight forces,
base. All these effects may increase the order of the and internal friction, which should be minimised. Good
controlled systems since they introduce additional design has uniformly distributed loading. Ideally, the
degrees of freedom. As depicted in [Groenhuis, 19911 a stress level under load should be the same value for all
fourth order model (two degrees of freedom) is the material used. However, even if there is a uniform stress
minimum and generally adequate order to describe the level over the construction, it definitively does not mean
dynamic behaviour of a mechanical closed loop system. that the structure is optimised to a point where
improvement is no longer possible. One could have
attained a local design optimum and a different design
5. Desian features for precision may simply be more favourable for the loads and
purposes intended [DeBra, 19981.

5.1 Stiffness of constructions and mechanisms Designs are either assembled from separate parts or
inversely produced by removing material from a volume
In addition to (dynamic) stiffness requirements of high leaving the required components in place. Regardless of
precision components to obtain a desired kinematic the method, the outcome can be regarded to consist of
function, stiffness is necessary to overcome additional truss, beam, plate or solid volume-like parts. Solid
side effects such as positional errors due to friction, volumes should be avoided where possible, i.e. they
essentially Coulomb friction. Friction forces whenever should be hollowed out. Stiff and lightweight structures

566
 Keynote Papers

can then be obtained by using closed boxes of plate applications where high resistance is required. Ceramics
material. These boxes should only to be joined and like B4C, SIC, Si3N4 and A1203 have high refractory values
loaded at the ribs and corners. Large unsupported plate and extremely high resistance against wear, corrosion,
areas will usually resonate due to lack of out-of-plane and erosion. Until now, designing in technical ceramics
bending stiffness in thin plates. Here thicker plates, made meant performing calculations of stress, and stress
of material with a lower density may beneficial. intensity factors [Klostermann, 19871, [With, 19961. The
Alternatively, the plates are made of a lightweight low toughness, in comparison to metals, resulted in a
sandwich design, offering two thin stress skins separated statistical approach of the fracture mechanics (Weibull
by a shear resistant web (honeycomb or foam). statistics). Technical ceramics however, have a high
Alternatively, a statically determined truss structure can specific stiffness. as well (factor of 3 to 7 compared to
be used, which offers high stiffness at the system nodes, steel), and therefore, .application of technical ceramics in
especially for relatively large protruding- or span lengths high precision machines is interesting (Slocum, 1994b].
in a construction. Optimising truss structures involves However, in consequence of transformations at high
preventing sharp angles between the trusses at the joints. temperature sintering, ceramic parts may have internal
Tube-like trusses help both to improve buckling resistance defects, and show considerable non-uniform shrinkage
as well as natural frequency, which is higher than for solid deformation. Therefore, uniform sheet thickness, relatively
rods of equal mass. Tetraform is a prime example. Node simple structures, and the absence of sharp edges and
connections however become more complex. point or line forces are essential.
5.1.2 Material selection Vibrations caused by slide movements, or coming from a
The choice of materials for a machine design is a key cutting process or the environment, will negatively
factor in determining final machine performance. Many influence performance (see section 5.7.2). Whilst the
criteria may be considered, including temporal stability, specific stiffness of massive ceramics is good, vibration
(specific) stiffness, homogeneity, thermal expansion and damping is poor. Therefore, instead of massive ceramic
diffusivity etc. Often, these criteria vary for differing structures, a laminated build-up of thin ceramic tiles is
machine elements and the approaches are numerous highly preferable. Here, glued lap joints are barriers for
[Hocken, 19951. Smith and Chetwynd introduce property vibrations, so that damping is increased (see section
groupings and use charts and profiles in order to compare 5.1.2.2). However, for stiffness reasons, a very small lap
different materials in the performance of precision joint thickness is required, in practice mainly depending
instruments and machines [Smith, 19921. Teague and on the flatness of the tiles used. Semi-manufactured
Evans have listed several criteria to be considered alumina substrate tiles (applied in the manufacture of
[Teague, 1989-1997. Criteria used for thermal loop IC's), having a flatness within a few 0.01 mm, enable the
design are described in section 3.3. For structural loop laminate stiffness to be mainly (>95%) determined by the
design, specific stiffness (the response to static and stiffness of the tiles D/ermeulen, 1996a,b], [Rosielle,
dynamic loading defined as the ratio of the modulus of 19961. Thereby, almost a factor of 3 in mass reduction is
elasticity to the density), internal damping (the specific possible, compared to an aluminium or steel plate frame
damping capacity), and deformation under point loading structure with the same stiffness.
(determined almost exclusively by the modulus of
elasticity) are crucial. The former two are further As a result of the air space program of NASA during the
discussed in this section. 1960's, advanced fibre composites were introduced.
During the last decade, new high-tech fibres like high
5.1.2.1 Specific stiffness modulus (HM) carbon, aramid, and polyethylen, as well as
As a general conclusion from section 5.1.1, it can be advanced thermohardening and thermoplastic matrices
stated that good design of high precision machines and were developed. However the strength of metals can be
instruments, requires design for stiffness instead of improved with additional alloy-elements or heat
admissible stress or lifespan, and as a result, for a readily treatments, its modulus of elasticity is nearly constant,
optimised geometry, it is the modulus which determines except for AlLi alloys [Rivin, 19961. Here, fibre composites
selection of the material. Moreover, in situations where differ from metals, as the former's modulus changes
components are accelerated, e.g., CMMs, next generation dramatically with the change in fibre fraction. In addition to
step-and-scan wafersteppers, and SPDT machines, high the fibre quantity in a matrix, its orientation can be
moving masses cause high inertial forces as well as a changed dependant on the load case of the structure. For
substantial heat dissipation in the slide's driving system, a tensile-compression load, mainly unidirectional (UD)
having a negative influence on the systems accuracy. In fibres can be applied. Especially UD HM carbon fibres
addition, high masses lower the values of the subsequent have a substantially higher specific stiffness than
natural frequencies. Therefore, optimising the stiffness per conventional construction materials. However, for getting
kg for a given geometry means selecting a material with in-plane stiffness, which is required when building plate-
high specific stiffness. frame type structures, only some 40% of the UD modulus
remains as in-plane quasi-isotropic modulus [Lubin,
Metals, like cast iron, steel and aluminium, have in spite of 19821. Even then, for HM carbon fibres a gain factor of 2.5
their moderate specific stiffness, the advantage that in specific stiffness can be obtained.
conventional joining techniques, can be applied, e.g.,
welding and thread tapping. Granite, having a comparable In [Kerstiens,'l990], structural components for high-speed
specific stiffness, is generally used as a structural material grinding and milling machines (slides, spindles), and
in machines, which operate in dry environments, e.g., for bridge-type robot centres (beams), made of carbon fibre
filtered and dried air bearing slide ways. As granite can are described and compared to metal alternatives.
absorb moisture and swell, it may not be appropriate to Dependent on the application and fibre orientation, a
use it in a machine where cutting fluid splashes all over it specific stiffness gain with a factor of 2 to 3 is possible. In
[Slocum, 19921, [Meijer, 19891. For this reason many addition, as a result of internal damping, an increase of
builders seal the granite with very thin epoxy resin. the dynamic stiffness is obtained. Manufacturingcosts are
doubled, but for equal machine performance, costs are
Technical structural ceramics are commonly used in about the same.

567
Reinforced ceramic and metallic composites have By contrast to tuned mass dampers, constrained layer
performance characteristics beyond those of conventional dampers add damping at all frequencies and are
massive ceramics or metals [Lanxide, 19901. Reinforced insensitive to the vibration amplitude. This type of
composites allow for a density and thermal conductivity damping mechanism has been used in many systems,
approximating those of aluminium, combined with steel among others consumer products, helicopter blades, and
like values for modulus and tensile strength. In addition to machine tools. Damping is achieved by imposing shear
a high specific stiffness, reinforced composites are tough strains on the alternate layers of visco-elastic and
and, unlike ceramics, have tolerance for damage. structural materials, which is caused by relative motion.
Reinforced ceramic matrix composites are made by a Motion of a structure is generally greatest far from its
directed metal oxidation process. Another infiltration neutral axis, especially when interface surfaces move in
process (without oxidation) is used to make reinforced opposite directions. The design can be developed using
metal matrix composites. numerical methods [Haranath, 19871, or closed form
solutions that were formulated many years ago [Nashif,
5.1.2.2 Internal damping 19851.
The response of a structure to a time-varying input Viscous shear forces stretch the structural constraining
depends on the stiffness, damping, and mass of the layer, so thin layers are not as effective.
structure. According to [Slocum, 1994a1, high stiffness
and damping are each necessary, but are not individually In the interest of reducing vibration in precision machines
sufficient requirements for a precision machine. and components, e.g. fixtures, [Slocum, 1994al proposes
a concept of a replicated internal viscous damper. The
Although extensively studied, and heavily relied upon in design principle is similar to that of a constrained layer
machine tool design, the mechanism of damping in a damper. The replicated internal viscous damper uses
material is difficult to quantify and one must generally rely internal beams covered with a viscous fluid that are
on empirical results [Lazan, 19681. In fact, damping highly replicated in place inside structural members. In the case
depends on alloy composition, frequency, stress level and where the centre of a structure is cored to make room for
type (i.e. tension or shear), temperature, and joint preload. the damper, although the static stiffness is decreased by
Traditionally, machine tools have been built of cast iron, up to 25% depending on the design, the dynamic stiffness
which has moderately good damping properties, or steel can be increased by an order of magnitude. Design theory
weldments, whose damping properties are not as good. as well as experimental results show that Q factors on the
Polymer concrete was then developed, and typically had order of 20 are obtainable. This is one to two orders of
on the order of 5 to 10 times the material damping of cast magnitude lower than would normally occur in the material
iron and was easier to cast. However, the use of polymer of the structure or in its joints.
concrete led to marginal increase of machine weight and [Marsh, 19961 describes a passive damping method
decrease of thermal diffusivity. However it is quite widely based on both constrained layer and replicated viscous
used for precision grinding machines. damping. Here, visco-elastic layers are embedded within
the structure as opposed to being applied externally. An
Structural joints in machine tools have long been known to analytical solution to the amount of damping in modes is
be a source of damping by the mechanisms of friction and presented using a modal strain energy approach. Several
microslip. Numerous theories are available for predicting full-scale structures have been built and tested, that offer
structural joint damping by these mechanisms [Tsutsumi. modal loss factors (l/Q) between 0.1 and 0.3, depending
19791, [Murty, 19821, [Ferri, 19951. In [Slocum, 19921 and on the constraints of added weight and available space for
[Slocum, 1994a1, the Tetraform structural concept for shear members.
machine tools and instruments, developed by Lin, is
described.,Here, besides viscous shear and squeeze film 5.2 Bearing systems
damping in the legs, damping at the spherical joints is
achieved by application of adjustable sliding bearing
5.2.1 Full film bearings
technology, developed for the Nanosurf 2, in which thin
(<2 pm) PTFE layers are used.
However, the amount of damping obtained in structural 5.2.1 .IStatic bearing stiffness
joints is still an order of magnitude less than what is Full film bearings, mostly aerostatic or hydrostatic
desired for optimal performance. In addition, as far as the bearings, are increasingly used in PE applications to
accuracy of the machine is required, it would be best if the provide a high accuracy long-stroke motion. The latter is
joints behaved as rigid interfaces, instead of slip planes mainly as a result of low limiting friction and a high
[Slocum, 1994a1. averaging effect of surface imperfections. The choice of
lubricant depends upon the type of application.
In machines with rotating components (e.g. grinding Hydrostatic bearings exhibit much greater film stiffness
wheels), there is often enough energy at multiples of the than aerostatic bearings, usually by a factor of about five,
rotational frequency (harmonics) to cause resonant due to the higher supply pressures. Important differences
vibrations. The amplification at a particular frequency can lie in the performance under dynamic loading. Hydrostatic
be reduced using a tuned mass damper, which pulls the bearings have superior damping characteristics to
’

resonant peak down into two smaller peaks. A tuned aerostatic bearings and this may be a critical feature for
mass damper consists of carefully selected mass (at least some application. Instability phenomena in the latter are
3-10% of the structure’s mass), spring, and damper mainly caused by the compressibility of air and molecular
elements attached to a structure. The design of a tuned transit times in the bearing gap, in combination with gap
mass damper is quite straightforward [Meirovitch, 19751. thickness variations and pressure changes [Blondeel,
However, the limitation of these dampers is that it can 19801. In [Plessers, 1988a,b], a stability (Nyquist) criterion
only remove energy at a small frequency range. As the is described to evaluate instability phenomena. Both
machine configuration or boundary conditions change numerical and experimental methods are used to derive
(e.g. the moving parts in a CMM, or the depth of cut taken dynamic stiffness and damping coefficients.
by a boring bar), the damper will lose some of its Compared to hydrostatic bearings, aerostatic bearings
effectiveness. have negligible friction and hysteresis. Viscous friction is

568
 Keynote Papers

proportional to both relative speed and viscosity of the and effective and competitive alternative to both
fluid. The latter is extremely low for air compared to oil magnetic- and ball bearings, for its infinite static stiffness
(factor of 10’ to lo5), and therefore, air spindles can be and active control bandwidth of 700 Hz.
applied in a much wider speed range. When designed for Supply pressure control finally, has in contrary a limited
equal stiffness, the friction in an air spindle might be a bandwidth and relatively low cross over frequency. Until
factor of 3 to 10 lower compared to a hydrostatic system, now, for a system with a great number of aerostatic
but the mass moment of inertia of hydrostatic spindles is bearings, the costs for active stiffness control are
about a factor of 30 less, being advantageous for significant, since at least an equal number of sensors and
application in CNC production machines [Kraakman, actuators is required.
19971, [Schouten, 19971. However, in addition to a
minimum required surface area for a high bearing 5.2.2 Active magnetic bearings
stiffness, the spindle stiffness has to be high enough to In many high precision applications, the limitation of
attain a favourable dynamic behaviour. A drawback of friction losses in bearings is an engineering challenge. As
hydraulic systems is the need for return pipes and much magnetic levitation is contact-free, there is virtually no
more complex filtering and conditioning systems. friction and wear, so magnetic bearings offer long life and
reduced maintenance. The rotational speed is on1 limited
On account of its simplicity, both aerostatic and
Y
by the strength of the rotor material (400 N/mm for soft
hydrostatic thrust as well as journal bearings with parallel magnetic iron), and peripheral speeds of up to 350 m/s
gap shapes (both inherent and orifice types) are often have been reached. Magnetic bearing systems are
used in PE slide constructions and spindles. In addition to particularly suitable for applications in which there should
analytical descriptions characterising fluid film flows, be no contamination from lubricants, as in vacuum. In
design methods and software are available to determine addition, it is fairly easy to adjust the position of the
bearing dimensions, necessary for a certain load capacity bearing forces and suppress the effects of unbalance. A
or stiffness [Holster, 19671, [Kraakman, 19761, [Kraakman, difficulty however, is that in practice, magnetic levitation is
19771, [Blondeel, 19801, [Stout, 19811, [Plessers, always accompanied by an unstable equilibrium of forces,
1988a,b]. In [Wang, 19931, bi- and tri-conical gap shapes so that height control with feedback is necessary. Active
are used, giving a stiffness gain of about 2 to 3, compared magnetic bearing (AMB) systems were built as early as
to parallel gap shape bearings. 1938 by Kemper for experiments and later for momentum
wheels in. space applications. Due to the enormous
As air is highly compressible compared to oil, aerostatic progress in electronics, the number of industrial
bearings need far more surface area and much smaller applications has considerably increased during the last 15
film thicknesses (5 instead of 30 pm) to achieve years. [Schweitzer, 19881, [Higuchi, 19901, [Allaire, 19921,
comparable stiffness values. In consequence of the rrumper, 19931, [Schweitzer, 19941, [Matsumara, 19961.
relatively small modulus of air (about a factor of l o 3 less
than oil), conventional aerostatic bearing systems tend to In general, five control loops are required for an AMB
determine the overall obtainable static machine stiffness. system with one d.b.f. The associated electronics,
However, using a compliant bearing surface (membrane), however, make such bearing rather complicated and
the compressibility of air might be compensated passively expensive, and for this reason, AMBs have been
[Al-Bender, 19921. Thereby, even infinite static stiffness designed that have one or more non-controlled d.0.f.
can be obtained over a large range of the load capacity, [Kamerbeek, 1983, 1984). The drawback here is that the
provided that a stiffness of the membrane and the air gap load-carrying capacity and stiffness corresponding to the
are equal. This principle of stiffness compensation can be non-controlled d.0.f. are reduced [Hendrikson, 19741,
used in axial as well as journal bearing systems [Blondeel, [Sabnis, 19741, [Kamerbeek. 19971. Electromagnets are
19761, [Snoeys, 19771, [Bryant, 19861, [Snoeys, 19871, usually arranged in two radially opposed coils for each
[Holster, 19871. Additionally, the load-displacement d.o.f., giving an eight-pole radial bearing system.
characteristic of the bearing may even compensate for the However, a minimum of six poles with six windings would
(static) elastic deformation of connected machine parts, suffice, but, contrary to systems using passive elements,
e.g. slides and spindles [Kazimierski, 19921. At Philips such as flexures and fluid film bearings, control becomes
Research Laboratories, double hydrostatic bearings are more difficult.in AMB systems, since force generation in
used with pressure feedback to a double-restriction the x and y directions is then coupled.
membrane, to create infinite stiffness for Single Point
Diamond Turning (SPDT) operation [Kraakman, 19691. For small rotor displacements and small control currents,
(compared to the nominal air gap and the bias current
5.2.1.2 Dynamic bearing stiffness respectively), the AMB force is quite linear. However, due
To improve the dynamic aerostatic bearing stiffness as to a negative coefficient of proportionality of the rotor
well, active stiffness compensation can be applied. In [AI- displacement (the open-loop stiffness), AMB systems are
Bender, 19941, three methods of active control are characterised by open-loop instability. Besides position
described to enhance dynamic stiffness, viz. support feedback, Kamerbeek applied velocity feedback, having
control, conicity control, and supply pressure control. the advantage of less heat dissipation in the control coils
Support control is most simple and applicable to all [Kamerbeek, 19831. Although the losses in AMBs are
bearing types. However, it is not compact and quite usually less then 20% of that in fluid film bearings, and
expensive. In conicity control the deflection of a compliant less then 5Y0 of the loss in mechanical bearing systems,
convergent surface is additionally controlled by a piezo- they may be examined in some applications. Especially
electric actuator. As described in [Al-Bender, 19971, this for larger rotor displacements, non-linear and non-ideal
method of active compensation is found to be most effects appear, such as copper losses, hysteresis effects,
effective, owing to the low actuated mass and the large stray fields, saturation of the iron, and eddy current
bearing force gain and bandwidth. It achieves infinite losses. Especially for high performance AMBs, these non-
static stiffness and still, at 100 Hz, a tenfold increase in ideal properties, which mainly depend on the material
stiffness is possible compared to the passive value. [AI- used in the flux-loop and the bearing arrangement
Bender, 19981 describes the development and application (hetropolar or homopolar configuration), have to be
of active front radial bearings in HF electro-spindles, as considered. For conventional soft magnetic iron, such as

569
silicon iron, the maximum flux density is limited to about some execution examples (a,b,c) as well as the schematic
1.5 T, thereby limiting the maximum static bearing force representation (d).
density to about 20-40 N/cm2 (comparable to air
bearings). -6

For the general class of linear permanent magnet electric

machines, rrumper, 19961 presents a design and
analysis framework, with which complex magnet arrays
and winding patterns can be addressed. The models
resulting from the analysis, serve as design tools for the a. b.
development of high-resolution, power efficient, linear
magnetic levitators, such as used in wafer steppers (fig.
5.1). To improve power efficiency, the levitator uses a
permanent magnet Halbach array, primary developed for
use as an optical element in particle accelerators. In
rrumper, 19931, useful geometries of such magnet arrays
for synchronous machines are presented in different co-
ordinate systems.

9 f l
levitated platen
&inch wafer C. d
Fig. 5.2: Execution examples of one rotation

The alternative 5.2b is preferable to 5.2a from the point

that a symmetrical loading results in four times higher
stiffness. This becomes clear from fig. 5.3, where the
dimensionless stiffness reduction c(a)/c(O) is set against
the dimensionless distance of the force’s line of action to
lihearmotordator the cross section’s neutral axis a/h . Alternative 5 . 2 ~is
‘
(thermally) symmetrical about the rotational axis (TC).
Fig. 5.1: Six d.0.f. magnetically levitated stage The schematic representation is shown in fig. 5.2d.
rrumper, 19961.

Another six d.0.f. magnetically suspended stage is

described in [Holmes, 19961. Here, additional neutrally
buoyant oil floatation is used in order to achieve extremely
high disturbance rejection and position resolution,
competitive with piezoelectric actuators. This hybrid stage,
which was designed for STM applications, attains atomic-
scale motion control inside 100 pm cube of travel.
-I

5.3 Flexures
- 2.
The advantages of flexures over classical hinges are the
absence of play, and sliding- or rolling friction as a source
of structural hysteresis [Koster, 19961, [Rosielle, 19981:
hence, there is no need for lubrication systems and
maintenance. Even for ultra precision mechanisms, the Fig. 5.3: Dimensionless stiffness dependent on the force’s
level of material hysteresis in flexures, e.g. as a result of line of action.
non-elasticity, or dislocation mechanisms in metals, is
insignificantly low, and extremely precise measurements Fig. 5.4a contains an example of an elastic universal joint,
are necessary to detect this phenomenon [Lazan, 19681. (with it’s schematic representation shown in fig. 5.4b),
The linear operating force of flexures presents no big leaving two rotational d.0.f. free, whilst constraining the
challenge in motion control (see section 5.6).Care must remaining rotation and three translations.
be taken to avoid mechanical flip flops due to internal over
constraints as a result of for example design or
manufacturing defects, undesired damage such as pre-
buckling, external assembly- or thermal miss-alignment.
Provided that such problems are avoided, overload
protection is envisaged, and buckling-, as well as modal
analysis is required. Although the length of stroke is
limited, flexures are often advantageous in ‘Design for
Precision’, featuring predictability and repeatability
[Eastman, 19351, [Eastman, 19371, [Jones, 19511, [Jones,
19561, [Geary, 19541, [Hoek, 1962-19861, [Hoek, 1985-
19891, [Paros, 19651, [Eijk, 19851, [Smith, 19921, [Koster,
19961, [Rosielle, 19981. a. b.
Single motion flexures can be divided in two main groups,
i.e. rotation and translation. For rotation, fig. 5.2 contains Fig. 5.4: Elastic universal joint: two rotational d.0.f.

570
 Keynote Papers

An elastic spherical pivot is shown in fig. 5.5a. Different The elastic guide in fig. 5.7 on the other hand, has a true
from the schematic representation (fig. 5.5b), where the axis of symmetric translation without lowering effects. An
three translational d.0.f. are fixed with three rods, the additional 'folded' leaf spring is applied for optimal
execution example contains three 'folded' leaf springs (cut symmetry. Internal stress is avoided by the monolithic
by wire EDM), as explained in section 3.2. As the fold- fabrication process, using wire EDM [Rosielle, 1994a,b],
lines are orthogonal, translation stiffness at the (fictitious) [Haitjema, 19971.
intersection point P is equal in all directions (580 NIpm)
[Rosielle, 19961. 5.4 Drives: properties and accuracy
Machine tool feed drives control the position and velocity
of slides or axes in accordance with commands generated
Intersection by CNC interpolators. In [Srinivasan, 19971, a list of
requirements on feed drive performance is given: 'control
; point
over a wide range of speeds, precise control of position,
ability to withstand machining loads while maintaining
accuracy, rapid response of drive system to command
inputs from the machine tool CNC system, and precise
co-ordination of the control of multiple axes in contouring
operations'. These requirements are best met by closed
loop control of feed drives based upon real-time sensing
of position.
Demands on feed drive performance have become more
stringent as machining technology has evolved to meet
a. b. the requirements of a broader range of applications, e.g.
high speed machining. Precision machining applications,
Fig. 5.5: Elastic spherical pivot: three rotational d.0.f. such as SPDT have similarly tightened positioning
accuracy requirements on feed drives. In [Smith, 19921, a
For translation, fig. 5.6 contains examples of one d.0.f. variety of actuators is described, that are capable of
When applying the driving force F halfway length I of the operating at the nm level. Each actuator is discussed in
leaf springs, as in alternative 5.6b, these springs are free terms of its relative merits, such as dynamic response,
of (opposed) normal forces. The disadvantage of the repeatability, size, linearity and cost.
lowering effect (0.6x2/1)of the parallel motion alternative of
fig. 5.6 might be compensated by a double parallel 5.4.1 Short stroke actuators
mechanism, which was among others implemented in the Whereas recent low voltage type piezos based on thin,
roughness measurement instrument by Rank Taylor layers are giving more extension at typically 100 V, many
Hobson. precision applications still apply high voltage piezos,
which, at present, tend to be stiffer under static conditions.
Compact actuators can generate large forces. Piezo
elements typically have 0.1% strain under 1000 V/mm.
Hence, single piezo elements are limited in maximum
stroke to the pm .range, which often necessitates two
stage actuation. The total length of stroke obtainable
depends on the number of discs in a piezo stack, and
amounts typically to a maximum of about 100 to 200 pm.
Here, typical material expansion rates are 1 mm/s.
Piezo material itself may have an internal hysteresis,
b. C.
which can be in the order of 20% of the stroke length.
Therefore, accurate positioning over large strokes is best
done with closed loop control. In [Ge, 19961, an approach
is described for modelling the hysteresis non-linearity of
Fig. 5.6: Five d.0.f. constraint parallel motion without piezo-ceramic actuators under non-cyclic input variations,
symmetry. By implementing the model in a feedforward control loop,
tracking control accuracy was improved significantly (see
section 5.6.1).

Manufacturers tend to clamp piezo stacks with a

preloaded spring to avoid mechanical hysteresis in the
many contact surfaces. A preload itself, e.g. by a weight,
shifts the range of piezo movement without reducing it,
thin-walled whereas stiffness, e.g. from an external elastic
triangular tube mechanism will reduce the stroke that can be obtained.

Nowadays, linear piezo actuators are used in numerous

monolyth applications like positioning systems [Wulp, 19961, shape
control of optical components [Schothorst, 19971, vibration
control, smart structures [Koster, 19971, laser cavities
[Schellekens, 19861, measuring lasers [Wetzels, 19981,
ed leafspring (6x) optical delay lines [Mierlo, 19931, AFMs and STMs
[Vorburger, 19971, piezo driven Stewart Platform
Fig. 5.7: Five d.0.f. axis-symmetric linear motion [Beltman, 19971, etc.
produced by wire EDM.

57 1
5.4.2 Long stroke actuators line power, but rely upon sophisticated electronic control
of frequency and amplitude. Smoothness of generated
5.4.2.1 Piezosteppers force, i.e. low force- and velocity ripple, is an important
performance measure for all motor types and requires
The above limitations tend to shadow the otherwise clean specific design features and electrical power control to
electrical to mechanical transfer. Designers have thought minimise force ripple [Bollinger, 19821, [Slocum, 19921.
around the obstacles and created actuators for longer
strokes like the classical inchworm@,a stepping device Linear power transmission elements are used in feed
which can be used for long stroke actuation by repeated drives both to convert rotary actuator motion and to
stepping motion, as well as for single movement with achieve speed reduction. Lead screw drives, using nut
nanometre resolution. Besides linear motion types, transmission elements, are most commonly used at this
designers have come up with numerous rotational drive stage of development [Gross, 19831, v e c k , 19841. Many
systems, e.g. for rotary tables and mirrors. Large of the problems with lead screw drives are the result of an
rotations, when required are mostly step type or inertial improper kinematic mounting of the nut. When properly
drive systems, providing tangential motion at large radii. designed however, they are capable of submicron-level
Though all bearing types mentioned in section 5.2 are motion repeatability. The advantage of lead screw drives
potential candidates for high precision piezo driven rotary in some applications is that stiffness is present above the
tables, it can be useful to consider driving more d.0.f. than bandwidth of control, as opposed to linear motors, which
just the rotation. [Philips, 19951 describes the design of a therefore require excessive bandwidth. Backlash in lead
rotary table, depicted in fig. 5.8, that can be moved within screw drives is reduced by preloading the nut, at the
0.1 p a d and some nm position accuracy of the rotation expense of increasing Coulomb friction, and thereby,
axis, due to closed loop control and calibration facilities. increasing virtual play and lowering drive efficiency.
Three piezo stepper drive-units are equally spread around Smoothness of motion in ball screw drives is limited due
a rotary table with an axial air bearing. All mechanisms to entry and exit of the balls in the ball-nut, causing
are cut into a single plate by wire EDM, which is freely velocity ripple.
programmable, and thereby other than traditionally The behaviour of friction wheel drives - also known as
circular holes can be applied in order to make a more traction drives, or capstan.drives - on the other hand, is
compact design [Rosielle, 19911. Due to the application of more predictable and reproducible due to a prescribed
high accuracy sensors, many disturbances including level of preload at the statically determinate wheel
temperature change, can be compensated for in closed contacts, thereby superior for the machining of optically
loop control. The relatively low rotational speed of 1 rpm is smooth surfaces [Takahashi, 19901, [Rosielle, 19941. At
a challenge to increase further. CUPE, LLNL [McKeown, 19981 and Philips Research
Laboratories, friction wheel drives have been used in
SPOT machines since the 1980s, in combination with
laser interferometer measuring systems [Bryan, 1979a1,
-Radial piezo [Gijsbers, 19801, [Donaldson, 19841, [Donaldson, 19861. In
actuatw (3x) [Bispink, 19921, ball screw as well as friction wheel drives

8 actuator (3x.X
[Tangential piezo are investigated in both constant velocity and reverse
tests. Bispink found that friction wheel drives have a factor
of 10 smaller position error signal. However, they have
lower stiffness and a lower maximum force level. In
[Taniguchi, 19961, experiments with nano-positioning
systems are described. Here, an aerostatic nut-and-screw
transmission is used in combination with a nano-
servomotor with air bearings as well. The nut surface is
made slightly concave, giving a 50 Nlpm thrust stiffness.
Applying motor encoder feedback as well as scale
feedback, result in a deviation of the feedback signal from
sine wave commands of some nm. CUPE/CPE have done
the same with oil hydrostatic lead screws to 2 nm
resolution and control [McKeown, 19981.

According to [Klafter, 19891, one of the primary causes of

poor tracking in indirect driven machines, especially at low
speed, is the mechanical link between the actuator and
the output. Hence a design without such transmission, i.e.
a direct drive system, is desirable in some cases. For
Fig. 5.8: rotary table,with nm range positioning system for these systems, e.g. linear motors, voice coil actuators,
the axis of rotation. and torque motors, it is essential to employ very high
5.4.2.2 Electromagnetic actuators resolution encoders, being available nowadays for linear
as well as rotational measurements. Direct drive systems,
Currently, electrical actuation is predominantly used in are advantageous for their ease of maintenance, little
machine tool feed drives. Rotary motors are used in friction and higher stiffness due to fewer compliant parts.
conjunction with rotary-to-linear power transmission Direct drives include lower drive inertia, higher
elements. The need for the latter is eliminated using linear acceleration capability, elimination of transmission
DC or AC motors. While brushed DC motors retain a resonance, and transmission related frictional effects and
purchase price advantage over the other motors because (virtual) play [Pritschow, 19901. Pritschow indicated higher
of simpler associated electronics, they are subject to positioning performance achievable using direct drive
brush wear and require higher maintenance. Furthermore, motors.
brush friction in such motors is non-linear and can limit An advantage of indirect drive systems is the accurate
control loop performance. AC motors are powered by AC tachometer velocity information, even at a very low slide

572
 Keynote Papers

travelling speed due to the application of a transmission Parameters characterising the behaviour of a sensor are
element. Moreover, the inertia of the load as well as the uncertainty, repeatability, resolution, bandwidth and
external- and internal forces converted to the motor shaft measuring range. Below, a short list of most'important
are reduced by the transmission ratio, thereby reducing sensors is dealt with [Doebelin, 19901. [Smith, 19921,
the size of the actuator. Depending on motor constant and [Slocum, 19921.
electrical resistance, heat dissipation of the actuator,
might be smaller, especially when hold currents are 5.5.1 Digital position sensors
supplied. In addition, the source of heat might be placed
at a greater distance and thereby isolated from the 5.5.1.1 Laser interferometers
workpiece. Laser interferometersare based on the interference of two
Transmission elements affect the nature of the feed drive beams, i.e. a reference and a measuring beam, emitted
control problem significantly. Feed drives including gears by a coherent laser source. These beams are split in the
or ball screws for speed reduction generally have high interferometer either by a polarising beam splitter or by an
transmission ratio values. Consequently, the feed drive amplitude beam splitter. Both possibilities are used by
control loop is effectively uncoupled from the machining different manufacturers. To increase stability often slightly
process and the structural dynamics loop. In the case of different frequencies (heterodyne interferometry) are
direct linear actuation however, where the ratio equals used. Whether this is done and which frequency
unity, the structural loop and the feed drive control loop difference is used also varies for different manufacturers.
interact significantly. Robust controllers are needed here, One or more photodiodes detect changes in the optical
such as H" synthesis and sliding mode control [Brussel, path difference of the beams by counting interference
1 9981. fringes, generated in the measurement after the
In [Renkens, 19971, a spindle unit is described, using a measuring- and reference beam are made to interfere.
direct drive voice coil actuator. With respect to current Interpolation between the fringes is used to increase the
spindles the objective was to reduce the axial error motion resolution to the nm level By using different optics, in
down to i0.01 pm, increase the axial stiffness up to 1000 addition to position, also angle, flatness, and straightness
Nlpm with a bandwidth of 700 Hz, and create the ability to can be measured [VDI, 19851, [Kunzmann, 19931,
produce non-rotationally symmetric parts with optical [Flugge, 19961.
surface quality (Ra c 10nm). The accuracy of laser interferometers is limited by the
stability of the wavelength of the laser light, which
Direct drive systems have the important drawback that changes due to variation in the refractive index of air. This
load variations are directly felt by the motor, and in index varies with temperature, pressure, humidity, COZ
reverse, variations in motor force, such as cogging and content and contamination by other gasses, and can be
force- and torque ripple, directly influence the accuracy at computed from Edlen's formula [Edlen, 19661, [Birch,
the load. Cogging results from end effects of steel 19941. An other source of error is non-linearity in the
laminations, crossing the alternate magnetic poles. Force- interpolation process, due to misalignment of and
and torque ripples arise from higher-order field harmonics imperfections in optics (beam splitters) optics [Quenelle,
due to both the magnet array geometry and the stator 19831, [Hou, 19921, and pNu, 19961. When laser
current distribution. In addition to the application of non- interferometers are used as a measuring tool for
ferrous core motors, state feedback, feedforward, and measurement of length, rotation, straightness or flatness,
motor ripple compensation can be applied to reduce the the equipment used, e.g. laser source, optics, mechanics,
variation in actuator force. [Braembussche, 19961 electronics and software, have to be calibrated
describes that only a simple ripple compensation method [Schellekens, 19863, [Kunzmann, 19921. Although nm
is necessary and sufficient in order to achieve acceptable resolution is easily achievable nowadays the relative
tracking errors. In [Kim, 19961,an analysis of the force uncertaint will -for measurements in air- not come far
ripple in a surface-wound permanent magnet linear motor Y
below 1 0 due to the error sources indicated above.
with Halbach array is presented, as being highly
dependant on the number of phases. The force ripple is 5.5.1.2 Scales
found to be less than 0.1% for motors using four or more Scales are based on a repetitive pattern of a reflective-
phases. Disc armature motors, normally used in high transmissive- conductive- or magnetic material embedded
acceleration applications for their small mass moment of in a substrate, giving a periodic signal. The accuracy is
inertia, have many self supporting copper coils, running mostly limited due to deviations in the embedded pattern
through an axial ma netic field. For the very high number and lies in the pm to sub-pm range. Conventional optical
9
of coils (order of 10 ), compared to traditional radial field encoders typically operate on the principle of counting
permanent magnet DC motors, the torque ripple is Moire lines by means of a light source and a photodiode,
negligibly small and in consequence, these motors might they can be linear or rotary.
be applied as direct drive in applications where high
smoothness of motion is required, e.g. SPDT operations Diffractional optical scales used as grating, increase the
[Vermeulen, 1996al. resolution to the nm level, whilst an uncertainty of about
10 nm. Here, two beams split from a laser of LED fall on
5.5 Sensors the ruler and diffract with different orders. After that, the
beams are brought to interfere. The interference signal is
Sensors can be incorporated in precision machines for essentially counting the lines of the scale passing by. To
two reasons: either in measuring instruments for which improve the interpolation, other signals with different
the sensor is a prerequisite, e.g. CCD cameras in vision phases are generated, e.g. by means of quarter
system, or to obtain higher system accuracy, for example wavelength plates. Main advantages to conventional
using a position sensor in a control loop or a temperature optical scales are the higher line density, up to 2000
sensor for thermal error compensation. The sensors, instead of 125 lines per mm, and the improved
which are dealt with in this section are limited to the latter possibilities for interpolation [Kunzmann, 19931, [Flugge,
group. These are position and velocity sensors for control, 19961, [Spies, 19971. Travelling speed at maximum
acceleration sensors for active vibration isolation, and resolution however, is presently limited by electronics in
temperature sensors for thermal error compensation. commercially available systems.

573
Other scales, based on a magnetic, inductive or large ranges, differing from 0.1 mm to about 5 m, and
capacitive measuring principle have resolutions and linearities between 0.05 and 1%. The resolution can be
uncertainties in the same range as conventional optical very low down to one tenth of a nm and the accuracy can
scales. be as good as a few tenth on a nm (Wetzels, 19981.The
bandwidth is mostly limited to several hundreds of Hz.
5.5.2 Analogue position sensors
In general, analogue position sensors like LDVTs, eddy 5.5.2.3 Eddy current sensors
current- and photonic sensors, have dynamic ranges In varying magnetic fields, currents are induced at the
(defined as the ratio of position measuring ran e and surface of a conductive material (for aluminium: typically
resolution) in the order of about lo3 to lo'? Only 0.1 mm skin depth). Measuring the change of self-
capacitive position sensors can have dynamic ranges induction of the coils generating the alternating magnetic
comparable with optical scales. In fig. 5.9,the dynamic field can sense these so-called eddy currents. Although
range of most common position sensors is depicted these probes have a significant non-linearity of 0.1 to
against the bandwidth. All sensors regarded here are non- lo%, resolutions of a few nm can be obtained, and the
contact sensors avoiding friction and hysteresis. maximum stoke amounts to about 10 mm [Doebelin,
19901.The best target material for eddy current sensors is
capacitive optical several other photonic aluminium; copper and brass are good alternatives. The
sensors scales sensors sensors use of steel however, is better avoided because of the
\ 1 / / varying permeability due to the microstructure of steel or
surface layers [Weekers, 19971.The disadvantage of all
high-magnification, eddy current sensors is their
2106
sensitivity to micro cracks, sub-surface stresses, and lack
of homogeneity in the target material. If the target moves
in and out with respect to the head, there is no problem,
but transverse movements can cause errors as high as
2.5 pm. [Bryan, 19671. The advantage of eddy current
sensors is their relative immunity to oil and dirt on the
surface.
5.5.2.4 Optical sensors
, , A wide range of optical sensors is available including laser
1oc
I [ I I I I 1 ,I I I I I 1 , I I I

triangulation sensors [Mohrke, 19911,focal error detection

1oo 10' 1o2 1o3
[Bouwhuis, 19851, [Renkens, 19941, and fibre optic
Bandwidth FHZ] sensors, CCDs, PSDs and array detectors [Doebelin,
19901.The range and resolutions depend on the applied
Fig. 5.9:Dynamic range against sensor bandwidth hardware and software: Focal error detection sensors
[Nijsse, 1997 have nm resolution and pm range, whereas triangulation
sensors, CCDs and PSDs have sub pm resolution and
5.5.2.1 Capacitive sensors mrn range. A clean target surface is a prerequisite for
Two electrically conducting parallel plates or concentric these types of sensors.
cylindrical parts, having a certain voltage difference and
overlapping surface may form a capacitor. By changing 5.5.3 Velocity sensors
the distance or the overlapping surface of the two parts, Velocity sensors are often used in servo systems to
the capacitance value will change, and thereby capacitors implement damping as will be explained in section 5.6.
can measure distance or displacement. By adding a third Linear velocities can be measured with voice coils
parallel plate in between the two others, being attached [Renkens, 19971,[Goossens, 19971. For angular velocity
elastically to the sensor housing, capacitive sensors can information in both direct and indirect drive systems,
be used as well for accurate angle measurement mostly tachometers are used. In voice coils as well as
(differential method). Practical values vary from 0.01 pF tachometers, a back-EMF signal is induced, linear
for small probes, to 100 pF for the larger ones. Capacitive proportional to the speed. Tachometers are available in
sensors can have very high resolutions down to the versions with or without brushes. Increasing the number
picometre level at about 10 kHz bandwidth [Bonse, 1995). of coils can reduce voltage ripple in tachometers, which
The range however, is limited to at most 1 mm. Capacitive increases costs as well, or by filtering which causes time
sensors are highly temperature sensitive. They are used delay. Resolution and accuracy depend on the
for velocity and acceleration measurement as well. application: in measuring instruments or to obtain higher
[Doebelin, 19901. system accuracy. Linearity is typically in the order of
5.5.2.2 Inductive sensors
0.1%.
Inductive sensors measure the variation of inductance 5.5.4 Accelerometers
between two or more coils, which depends on the position Measuring and differentiating the displacement of an
of a ferromagnetic core. There are two coil configurations. accelerating body relative to a compliantly suspended
The first type uses two coils in a Wheatstone measuring mass can derive acceleration information. Here, most
bridge. The second configuration type uses three coils: a commonly strain gauges and piezo electric transducers
primary coil, inductively coupled to two secondary coils. are used. During the last decade however, silicon
The use of two measuring coils allows for differential acceleration sensors have become commercially available
measurements; these sensors are called LVDT or RVDT as well, being produced in large quantities at small costs,
for linear- and rotary- variable displacement measurement (using etching technology) for the application in air bags.
respectively. From the two measuring coils additional For high accuracy measurements however, force
information about the moving direction is obtained, and in balanced accelerometers are applied. Here, a mass-
the mid-position of the core, the sensors are independent spring combination is used in a feedback loop with a voice
of temperature changes. Inductive sensors may have very

574
 Keynote Papers

coil actuator keeping the mass in zero position. The In a review about the last three decades, Koren describes
current (force) needed is a linear proportional measure for three types of motion control, depicted in fig. 5.10, i.e.
acceleration. servo control of individual machine-axes, interpolators to
co-ordinate the motion of several axes, and adaptive
5.5.5 Temperature sensors control to adjust real time the process variables in order to
Temperature detectors are usually based on either a rnaximise system performance.
temperature dependent resistance (Resistance
Temperature Detector, RTD) or on the Seebeck effect, 5.6.1 Servo control
which generates a thermo-voltage proportional to the In the past few decades, many types of feedback
measured temperature change (thermo-couple). controllers have been developed, both analog and digital,
Two types of RTDs are mainly used for practical as well as hybrid implementations [Srinivasan, 19971. In
temperature measurement i.e. thermistors and platinum digital controllers the effects of limited cycle frequency of
resistance thermometers (Pt-RTD). Thermistors have a sampling might reduce system performance [Renkens,
high negative temperature coefficient dR/dT and are non- 19971, [Goossens, 19971. In addition, controllers applying
linear but very stable. They show high sensitivity and state space estimation are used. The majority of industrial
resolution and uncertainty down to the mK range at feedback control systems contains a PID-controller,
normal workshop conditions. No complicated electronics having proportional gain (P-action) for servo stiffness,
are necessary. Pt-RTDs show better linearity but integral gain (I-action) to eliminate steady state errors and
sensitivity is much lower and asks for complicated to reject external disturbances, and differential gain (D-
electronics means to reach mK resolution and uncertainty. action) to provide damping, which gives the system
Thermocouples may be used too but they ask for a very stability. For the latter either velocity feedback, using
stable reference temperature and require very sensitive tachometer information, can be applied, or the
and stable voltage measurement equipment to get mK differentiated position feedback signal. In the case that
accuracy. transmissions are used in combination with feedback of
the actuator position, a resonancelanti-resonance pair is
5.5.6 Sensor mounting induced inside the control loop near to the transmission
As sensors are often part of a structural loop, kinematic, resonance frequency [Srinivasan, 19971. Here,
stiff-, and thermo-mechanically stable sensor mounting is transmission related deviations are not dealt with inside
essential, as close as possible to the place of which the control loop, unlike feedback of the position of the
information is required. The location of sensors is of vital load.
importance for the accuracy of high precision machines, For effective rejection of disturbances, such as motor
and sometimes, even a metrology frame is essential. The torque- and tachometer voltage ripple, as well as
Abbe and Bryan principles have already been mentioned disturbances from contact bearings, a sufficiently high
in section 3.1. Besides, sensors should be isolated from control loop bandwidth is required [Donaldson, 19841. The
heat sources and vibrations (see section 5.7). Sometimes effect of disturbances on the output position is expressed
it is advantageous to measure displacements before the by the sensitivity [Bosch, 19981. H"-controllers
transmission for the higher resolution at lower costs. All simultaneously filter disturbances and handle model
the errors in the transmission however, will affect the uncertainties, for example at the subsequent resonance
output. A final rule for sensor mounting here, is that a frequencies [Damen, 19971, [Braembussche, 19981. In the
slide's measurement and driving system should be at the latter case, a notch filter, used in classical controllers to
same side of the centre of rotation. Othewise, an filter a small frequency band, will not avail.
additional phase shift of 180" will occur, which is The required servo stiffness of the control loop depends
disadvantageous for stability [Slocum, 19921. on the application. For diamond turning, a static stiffness
of 10' to lo9 Nlm is acceptable [Kouno, 19841, but in
5.6 Control precision grinding of brittle materials, more than 10" Nlm
is required [Kanai, 19901. Classical PID-controllers are
As disturbances acting upon high precision machines Single-Input-Single-Output (SISO) systems. More
introduce inadmissible deviations from prescribed advanced controllers have Multiple Inputs and Multiple
functions, machine control is essential. Control of machine Outputs (MIMO), e.g. observer based controllers, which
tools commenced in the early 1950's with the invention of apply a state space approach [Eijk, 19981. MIMO systems
Numerically Controlled (NC) machines by Parsons [Koren, can be simplified to multiple SISO systems by conversion
19971. A major progress, especially on flexibility, occurred of dependent to independent variables.
with the introduction of the Computer Numerical Control In the design and analysis of feedback systems, various
(CNC) and adaptive control in the early 1970s. optimisation methods can be used. A rule of thumb to
guarantee sufficient stability tells the ratio of the first and
programmed required programmed second undamped resonance frequency to be in between
speed and feed 'Ore, 4
positions machine
pperature 3 and 10. A very powerful tool is the frequency response
1 machine
geometry
method, using Bode- and Nyquist diagrams, and having a
adaptive close link to experimental information [Rankers, 19971. As
an example, in order to investigate the frequencies of
or power- workpiecefiool
deflections interest, e.g. bandwidth, notch filter frequency, and tamed
workpiece action frequency, the Bode diagram of the open loop
2
- interpolator temoerature
response can be used. Two main problems with PID
axial commands controllers in contouring applications are the relatively
servo control poor tracking of corners and non-linear contours, and
1
significant overshoot [Koren, 19971. To reduce these
drive drives position effects, a smooth motion profile can be chosen, e.g. an
speed feedback
feedback inclined sine, instead of a trapezoidal function [Rosielle,
19981. Another effective method is the use of different
Fig. 5.10: Hierarchical levels in CNC controllers, from controller strategies, for example, feedforward controllers,
[Koren, 1997) sophisticated axial controllers, like f u u y logic and

575
repetitive controllers, or cross coupling controllers [Koren, environmental disturbances: 1) uncouple the machine
19971. from the environment, 2) design the machine so that
environmental disturbances are minimised, and finally, 3)
5.6.2 Feedforward control control the environment [ANSI, 19731. Other references
In a feedback system, controller action is generated only regarding this matter are: [Pohl, 19861, p’oung, 19721,
after an error signal has occurred. Application of reague, 19861and [Binnig, 19821.
feedforward control however, can lead to significant error
reduction up to a factor of 20 [Bosch, 19981, [Eijk, 19981. 5.7.1 Thermal isolation
Feedforward control -this can be velocity or acceleration Machine drives are one of the major sources of internal
feedforward- uses a priori knowledge of the mechanical heat dissipation. To minimise this effect, the drive systems
system, e.g. inertial effects in a machine, or cogging have to be adjusted to the load (see section 5.4). Small
effects in a motor. Application of intelligence however, is cycle times can reduce power dissipation, but at the same
at the expense of calculation time, resulting in lower time limit productivity. For external heat sources, other
sampling frequency and lower control power. Feedforward techniques can be applied, e.g. encapsulation of heat
can be implemented as the inverse of the feedback loop source, and insulation of temperature sensitive materials
[Koren, 1997). However, if the inverse function includes (see section 3.3) from its surroundings rrapet, 1997bI
unstable poles, alternative forms of feedforward
controllers have to be used. Tomizuka and others propose In [Smith, 19921, the physical meaning of thermal material
the use of frequency domain criteria romizuka, 19891, properties is dealt with. The smaller a material’s
e.g. zero phase error tracking controllers (ZPETCs). coefficient of thermal expansion a,the smaller the effect
These controllers however, act very well in tracking of differential temperatures within a component.
complex trajectories but have poor performance in Temperature differences depend reciprocally proportional
machine systems with large unmodelled disturbances to the thermal conductivity 1, and as a result, it is the
romizuka, 19871. quotient a/1 that expresses the material sensitivity for
spatial gradients. Mainly therefore, and for its good
5.6.3 Interpolators availability and machinability as well, aluminium is a
Interpolators are used successfully, in contouring favourable construction. material. Therefore, CMM
systems, where a co-ordinated movement of separately manufacturer Zeiss developed a special ceramic-surface-
driven axes is necessary [Koren, 19751, [Koren, 19811. coated ageing resistant aluminium (CARAT@).In scales,
The interpolators, located at level 2 of fig. 5.10, generate exotic materials with negligible a are used, such as Invar@
command signals for each segment of the produced part, and Zerodur@glass ceramics. The latter is applied for
based upon an initial and final point, and some type of structural components as well, e.g. in the Nanosurf by
curvature, which is typically linear, circular and Lindsey, Smith and Robbie at NPL [Lindsey, 19881.
occasionally parabolic. Precision and productivity can be In the case of an time dependent temperature distribution
improved using real-time interpolators, instead of in a machine, different time constants of components
conventional CAD/CAM off-line decomposition [Shpitalni, become predominant. The exponential path temperature
19941. change with time is governed by the volumetric thermal
diffusivity Upq, (p: density, cp: specific heat). For rapid
5.6.4 Adaptive control and compensation
response, this coefficient should be large. A mixture of
Adaptive control and compensation, located at level 3 of different materials however, might cause potential
the control architecture of fig. 5.10, improves problems [Lingard, 19911. From temperature experiments,
performance, either by enlarging productivity, e.g. in performed at PTB by rampwise increasing the
rough cutting, or by improving accuracy, e.g. in optical environmental temperature with 2OoC, the time constants
quality finishing operations. Ro and Hubbel applied two of different CMM components appear to vary from
different types of adaptive control in ball screw driven typically 4 to more than 30 hours [Kunzmann, 19881.
positioning systems, for both micro and macro When a body is over-constrained, internal stresses arise;
movements, having very different underlying friction For a small stress level the material’s modulus should be
mechanisms [Ro, 19931. For high precision purposes, small, which is in conflict with the demand for high
real-time compensation techniques are used for errors stiffness (see section 5.1). Therefore, kinematic design,
caused by varying machine temperature, geometry errors when possible, is highly preferable (see section 3.2).
of the machine, tool wear, etc. [Peklinik, 19701, [Soons,
19931, [Spaan, 19951, (see also section 6). Other frequently applied manners to minimise the effect of
environmental temperature changes are temperature
5.7 Isolation controlled rooms [Neumann, 19881, [Kunzmann, 19881,
and oil - and air-showers, with which a thermal stability
For many different reasons, precision machines have to within a few mK is possible [Bryan, 1979b],[Bryan, 19821.
be isolated against external disturbances. In reague, This approach to thermal stability is widely used today on
1989-19971, Hocken mentions a quotation of Maxwell ultra precision machines, e.g. the LDTM [Bryan, 1979al
[Hocken, 19951: ‘In designing an experiment the agents and the LODTM [Donaldson, 19841.
and phenomena to be studied are marked off from all
others and regarded as the field of investigation. All others 5.7.2 Vibration isolation
must be so arranged that the effects of disturbing agents In [DeBra, 19921, successful precision engineering is
on the phenomena to be investigated are as small as described as ‘The balance of robustness of the machine
possible’ [Maxwell, 18901. The major disturbing agents are and how benign the environment can be made through
vibration, temperature changes, humidity, acoustic noise, isolation to minimise the strains caused by vibration that
and electric and magnetic fields. The former two are compromise a machine’s accuracy’. Seismic disturbances
discussed below in more detail. In specific situations, as a result of natural or cultural ground motions can
other effects, like variations of gravitational load, compromise measuring accuracy and affect surface finish
atmospheric pressure and EMC might have significant in diamond turning operations. In machine shop floors,
influence. Teague and Hocken mention three design typical ground vibrations have frequencies of 10 to 20 Hz
strategies for machines to become resistant to

576
 Keynote Papers

and amplitudes of a few pm [Riven, 1979al. real time in order to get an accurate result. The effect of
In addition to ground motions, vibrations might result from error compensation strongly depends on the design of the
machine utilities, such as electricity, signal wires, air, machine and the ability to measure the errors of the
vacuum, hydraulics and cooling fluid. Even acoustic machine which are in principle deterministic in a major
coupling from neighbouring machine tools and fans can sense as Bryan indicated in his papers [Bryan, 19841,
be significant, since acoustic energy is broadband in most [Bryan, 19931. Therefore high repeatability is a need,
cases. asking for an excellent design.
Passive vibration isolation implies the application of a high As the first method of error minimisation has been
mass on low stiffness springs, i.e. a second order system. described in previous sections, this section focuses on the
Therefore, the attenuation of the amplitudes of vibration of latter method of error compensation.
the isolation on the one hand, and, for example, the tool
or measuring system on the other, is given by the 6.1 Error compensation of measuring machines
reciprocal squared ratio of their first natural frequencies.
Multiple isolation stages can have higher natural Error compensation in measuring machines is generally
frequency and give more attenuation above the critical based on the addition of a compensation vector
frequency due to a steeper slope. dp(x, y, z,. .) to scale- or interferometer readings
In [DeBra, 19921 and [Riven, 1979a,b], different kinds of p(x, y, z,. .). The vector dp may be a vector function of co-
isolators for precision machines are explained, some of
ordinates dp = f(x, y, 2 , ..) . The functions F(x, y, z, ..)
them commercially available. For many applications
polymer isolators may provide adequate isolation at low depend on the mechanical machine structure and the
cost. Riven, for example, describes rubber-metal accuracy of the scales and the probe and geometry errors
equifrequential isolators, based on the volumetric like straightness, squareness and waviness. These
incompressibility of rubber and a proportionality between properties however may be influenced by the temperature
load and stiffness. However, when isolation requirements distribution within the machine.
are tighter, pneumatic isolators are better. Air springs do Also dynamic errors due to the inertial and modal
not have the distributed property of slinky modes, present behaviour of measuring machines can be corrected
in metal helical springs. Hysteresis can be minimised [Weekers, 19971. It may be clear that compensation of this
using rubber roll seals between the piston and the type of errors is much more difficult than the (quasi)-static
isolation cylinder. errors discussed before. [Weekers, 19971 and [Knapp,
In isolation systems, additional damping is required to 19881 showed that machine dynamics i.e. inertial effects
improve transient settings. Pneumatic isolators have the and vibrations may cause considerable errors during
advantage that damping can be applied easily, by probing.
connecting each air spring through a laminar restriction to To apply error compensation, different error models and
a different chamber, which has, in the ideal case, eight measurement techniques were developed by many
times the volume of the compliance.chamber. In addition researchers [Hocken, 19771, [Zhang, 19851, [Jouy, 19861,
to damping, automatic height levelling is easy by [Sartori, 19881, peeuwsen 19891, [Balsamo, 19901,
individually changing the pressure in the legs [DeBra, [Kunzmann, 19901, [Soons, 19931, [Kruth, 19941, [Sartori,
19921. However, as air is filled through the damping tanks 19951, [Lingard, 19911. They showed that it was possible
to the air spring chamber, the levelling system might to develop very good error compensation for measuring
become too slow. machines. The compensation techniques are nowadays
For faster system response, active suspension and commercially available and used by many manufacturers
isolation can be applied [DeBra, 19921, [DeBra, 19941. of CMMs.
Here, a feedback loop is used to influence the height and
orientation of the machine by means of a position or The techniques to analyse the error structure of CMM
acceleration sensor and a magnetic or electrostatic may be divided in parametric and volumetric techniques.
actuator. Permanent magnet DC actuators based on In the parametric description the error parameters ,R,, ,T,
Lorentr forces are easily controllable due to the linear (as indicated in chapter 4) are measured directly at
reaction between current and the supplied force discrete positions and from these results, the error vector
[Vermeulen, 19981. In fully active isolation systems -which can be calculated, using adequate models. For these
will be used much more in future- the natural frequency calibration techniques, laser interferometers, levelling
can be adjusted electronically, so that the frequency instruments, autocollimatorsand hole- and ball plates may
response of the vibration transmission can be shaDed be used [Kunzmann, 19901, [Soons, 19931.
arbitrarily.
The volumetric approach to determine the error structure
of a CMM is based on volumetric length measurements
6. Error reduction and comDensation with artefacts like a double ball bar (DBB), a checking
gauge of Renishaw or with laser interferometer, slip- or
step gauge. It is important to measure sufficient lengths
In Precision Engineering nowadays two methods of error throughout the measuring volume [Krulewich, 1995a,b],
minimisation are applied, as mentioned in [Sartori, 19951, [Jouy, 19861, [Soons, 19931, [Florussen, 19981,
[Weck, 1995al and [Teague, 1989-19971. In the first [Krulewich, 19981. Using sophisticated modelling
method, called ‘error reduction’, positioning uncertainty is
techniques, parametric errors ,T,(x,y,z). ,R,(x,y,z), may
minimised by proper design, based on design principles
(section 3), modelling (section 4) and design features be estimated from the volumetric errors. In return, these
(section 5). Since a design will always be a compromise parametric errors are used to calculate the error vector
of many design philosophies, certain errors remain. A dp by an appropriate model [Florussen, 19971.
second method, called ‘error compensation’, can be Impressive results are reported by some authors [Soons,
applied to compensate for residual errors, for instance 19931, [Spaan, 19951, [Krulewich, 1995a,b] and
using geometric calibration- temperature measurement [Krulewich, 19981, but is has been proven too that it is not
data. This type of error compensation may be applied in easy to estimate all the error parameters correctly from
measuring machines. However, errors in positioning- volumetric measurements using the DBB or the checking
machines and machine-tools have to be compensated gauge as manufactured by Renishaw.

577
Recently activities are reported dealing with the so-called Compared to measuring machines, error compensation of
‘Virtual CMM’ (VCMM) [Trapet, 1997al. In this approach machine tools has been developed not so well due to both
extensive error compensation of geometric-, thermo- measurement- as well as implementation problems of the
mechanic- finite stiffness- and probing errors is applied. algorithms and while additional measurement equipment
Residual errors are estimated from measurements of well- is needed. For instance, due to the internal heat sources
known artefacts and careful analysis. As a result each in a milling machine, ten to fifteen temperature sensors
(corrected) measurement point p, has a residual error have to be connected to the machine body to estimate the
temperature distribution of the machine efficiently. Sensor
field dPi where IdPil<A,. where A, is the upper error placement and effect analysis is of major importance
limit. From pi the measurement result M is calculated in [Spaan, 19951, [Ni, 19971.
Now first error compensation of geometrical and some
the usual way for CMMs, but now many different values of
thermal effects are commercially available for machine
M may be calculated using dP, and simulate the tools. Also some finite stiffness effects (straightness) may
measurement many times. From these results a mean be compensated. Recently, first results were published
value and an statistical uncertainty value SG can be about compensation of cutting force induced errors [Yang,
calculated. In this way an approach for traceable 19961. These errors may become significant in hard
measurements of products on CMMs may be realised. machining.
Generally the real time software error compensation is
embedded in the controller of the machine tool. An
6.2 Real time error compensation of machine interesting development is the so-called NC-code
tools compensation [Spaan, 19951. This method adapts the NC-
Real time error compensation of CNC machine tools and code to compensate the programmed trajectory for the
positioning machines (e.g. wafersteppers) by programmed modelled errors.
positions or trajectories has been applied since the early
1970’s [Hocken, 19861, [Soons, 19931, [Krulewich,
1995a,b], [Spaan, 19951, [Ni, 19971, The main problem is, 7. Future trends in precision desiqn
that the compensation values for each position have to be
known beforehand. Particularly error compensation of As indicated before, Precision Design is one of the basic
thermo-mechanic and finite stiffness effects of disciplines in Precision Engineering. Precision Design
manufacturing machines is difficult since these errors are offers the tools to improve the accuracy of precision
time- and task dependent. Many authors have reported machinery.
about different approaches to overcome these problems During the last decades Precision Design has been
[Hocken, 19861, [Donmez, 19911, [Schellekens, 19921, developed very rapidly which resulted in numerous high
[Spaan, 19951, [Weck. 1995a1, and [Krulewich, 1995a,b]. precision machines and instruments, such as CMMs and
Thermo-mechanic errors are the worst to compensate for SPDT machines. Knowledge on Precision Design further
due to their complicated structure [Spaan, 19951. Most developed in order to thrust back the frontiers of accuracy
approaches use many temperature sensors to measure and resolution. The development of supporting software
the outer machine temperature distribution and as a result for design activities was of great importance, such as
to calculate the resulting tool drift. Here statistical tools to predict the behaviour of machines, and foresee
techniques as well as neural network approaches are problems, without hardware realisation. Moreover, the
used to get robust relations between measured integration of high accuracy sensors, closed loop control
temperatures and tool drift [Spaan, 19951, [Veldhuis, systems and advanced control software led to more
19941, [Ziegert, 19941, [Srinivasan, 19921, [Moriwaki, accurate and flexible machinery.
19961. Impressive improvements of machine accuracy
were reported, showing that both approaches are effective Nowadays, needs for precision are mainly driven by
in thermal error compensation [Spaan, 19951, [Krulewich, developments in information technology. Research on IC-
1995a,b]. Fig. 6.1 shows the effect of geometrical and technology and technology for data handling and storage
thermal error compensation on a milled workpiece with on tapes and discs, but also liquid crystal and plasma
and without online error compensation. Here the extracted display technology are driving forces for Precision
zero drift from the measured workpieces as a function of Engineering in general and Precision Design in particular.
manufacturing time, is depicted.
The growth of the IC-technology is well indicated by the
expansion of DRAM capacity from 245 kb in 1982 to 256
bml ,
Mb in 1997. Line width and distance of traces have been
0
reduced from 2 to 0.2 pm. Very large- and even ultra large
-20 with compensation scale ICs (VLSls and ULSls respectively) are common. All
the many processing and illumination steps of the photo
lithography process on silicon wafers of up to 12”
diameter, ask for extreme requirements on velocity,
acceleration, and especially reproducibility of the scanning
and positioning stages, down to some tens of
nanometres.

Next step will be x-ray lithography to overcome the

restraints of the illumination wavelength, limiting the line
width of traces to about 0.15 pm. Due to the short
wavelength of about 13 nm for soft x-ray lithography, only
mirrors can be used to deflect the x-ray beam from the
Fig. 6.1 : Zero-point drift (z-direction) extracted from source, through the mask at the wafer surface. Here, the
two workpieces milled with and without uncertainty of the mirror shape has to be below 20 nm
compensation [Spaan, 19951. and a surface roughness less than 0.5 nm Ra is required

578
 Keynote Papers

at 200 mm diameter. In order to get enough x-ray [Allaire, 19921

reflection, high precision multi-layer coatings are Allaire, P. (ed.), 1992, Proc. of the 3rd International
necessary raniguchi, 19961. As a result, new Symposium on Magnetic Bearings, Washington, DC.
technologies have to be developed, not only for the [Ando, 19921
stepper itself, but also for machines for mask making and Ando, M.,1992, Super smooth surface polishing on
inspection, as well as the manufacturing of high precision aspherical optics, Proc. of the SPIE, 1720, pp. 22-33.
optical components. Here, also 'Design for Low [ANSI, 19731
Contamination' is getting part of concern of 'Design for ANSVASME B89.6.2-1973(R1979), Temperature and
Precision'. Humidity Environment for Dimensional Measurement.
[ASME B89.4.1-1997l
Research and development of high density optical disk Methods for performance evaluation of coordinate
systems further shift the limits for data handling and measuring machines, revision of ASME 689.1.12 M-
storage. For systems like CD-Audio, CD-ROM, CD-Video 1990
and recently DVD-systems as well, maximum tolerances [ASME 65.54, 19931
of track pitches and pit dimensions further decrease. Both Methods for performance evaluation of computer
recordable and rewritable versions are available. Here numerically controlled machining centres
again, decreasing wavelengths, in combination with [Attia, 19791
increasing numerical apertures (NA) are required. Recent Attia, M.H., Kops, L., 1979, Calculation of thermal
developments like dual layer (DVD9) and even double- deformations of machine tool, in transient state, with
sided dual layer systems (DVDl8) further increase the the effect of structural joints taken into account, Annals
information density. The writing accuracy of laser beam of the CIRP, Vol. 2811.
recorders asks for nanometre accuracy rotary tables and [Balsamo, 19901
mastering beam systems. Consumer DVD rewritable Balsamo, A,, Margues, D., Sartori, S., 1990, A Method
systems (DVD-RW) will soon reach the tens of nanometre for thermal deformation correction of CMMs, Annals of
accuracy level for writing reproducibility. the CIRP, Vol. 3911.
[Beltman, 1994
For PC terminals and HDTV, new flat screen systems are Beltman, A.J., 1997, A Piezo driven Stewart Platform,
recently developed. For the former, thin-film-transistor Proc. of NVPT Jubileumcongres Precisie Technologie,
colour LCDs are available, and for HDTV, plasma display Leusden, The Netherlands.
panels (HD-PDP) were developed recently with new [Binnig, 19821
fabrication techniques. Both LCD and PDP systems need Binnig, G., Rohrer, H., Gerber, C.H., and Weibel, E.,
very accurate alignment techniques of opposite electrodes 1982, Surface studies by scanning tunnelling
and well-controlled etching and film deposition processes microscopy, Physical Review Letters, Vol. 49, p 57.
[Taniguchi, 19961. [Birch, 19941
Birch, K.P., Downs, M.J., 1994, correction to the
Finally, developments in astronomy have to be mentioned updated Edlen equation for the refractive index of air,
here. In the next decade, astronomic measurements will Metrologia, Vol. 31, pp. 315-316.
require stability in length of nanometres and nanoradians [Bispink, 19921
in angular position [Sartori, 19981. Bispink, T., 1992, Neue Wege zur Beurteilung
Precision Engineering and therefore Precision Design as hochgenauer Vorschubantriebe fur Ultraprazisions-
well, faces a decade of fast growth and striking new maschinen, Ph.D. Thesis, Technische Hochschule
challenges. Aachen, Germany, ISBN 3-86111-275-2.
[Blaedel, 19981
Blaedel, K.L., 1998, Lawrence Livermore National
References Laboratory, CA, Private communication
[Blanding, 19921
[Abbe, 18901 Blanding, D.L. 1992, Principles of Exact Constraint
Abbe, E., Messapparate fur Physiker, Zeits, Fur Mechanical Design, Eastman Kodak Company,
Instrumenten-Kunde,Vol. 10, pp. 446-448. Rochester, NY, 14650-1822 716-722-6382
[Al-Bender, 19923 [Blondeel, 19761
Al-Bender, F., 1992, Contributions to the design theory Blondeel, E., Snoeys. R., Devrieze, L., 1976,
of circular centrally fed aerostatic bearings, Ph.D. Aerostatic Bearings with Infinite Stiffness, Annals of
Thesis, Katholic University of Leuven, Belgium, ISBN the CIRP, Vol. 25/1, pp. 291-296.
90-73802-22-9. [Blondeel, 19801
[Al-Bender, 19941 Blondeel, E.,Snoeys, R.. Devrieze, L., 1980, Dynamic
Al-Bender, F., Brussel, H. van, 1994, Active Dynamic Stability of Externally Pressurised Gas Bearings,
Compensation of Aerostatic .Bearings, Proc. of the Transactions of the ASME, Journal of Lubrication
ISMA 19, Tools for Noise and Vibration Analysis, Technology, Vol. 102, pp. 511-518.
Leuven, Belgium, pp. 187-197. [Boley, 19601
[Al-Bender, 19971 Boley, B., Weiner, J., 1960, Theory of thermal
Al-Bender, F., Brussel, H. van, 1997, Active Aerostatic stresses, Krieger Publishing Company, Malibar, FL.
Bearing through Control of Film Geometry, Proc. of the [Bollinger, 19821 '

9th IPES-4UME, Braunschweig, Germany, Vol. 2, pp. Bollinger, ' J.G., Stute, G., Van Brussel, H., 1982,
389-392. Digital Controls and Feed Drives: State of the Art and
[Al-Bender, 19981 New Developments,Annals of the CIRP, Vol. 29/2, pp.
Al-Bender, F., Brussel, H. van, 1998, Development of . 497-506.
high frequency (HF) electrospindles with passive and [Bonse, 19951
active air bearings, Proc. of the International Seminar Bonse, M.H.W., 1995, Capacitive position transducers,
on Improving machine tool performance, San theoretical aspects and practical applications, PhD-
Sebastian, Spain, Vol. 1, pp. 175-184. thesis, Delft University of Technology, The
Netherlands, ISBN 90-370-0135-1

579
[Bosch, 19981 Bearing, Transactions of the ASME, Journal of
Bosch, P.P.J. van den, 1998, Eindhoven University of Mechanisms, Transmissions, and Automation in
Technology, The Netherlands, Private communication. Design, Proc. of Design Engineering Technical
[Bouwhuis, 19851 Conference, Columbus, Ohio, Paper No. 86-DET-32.
Bouwhuis, G., Braat, J, Huijser, A., Pasman, J., [Camera, 19761
Rosmalen, G. van, Schouhamer Immink, K., 1985, Camera, A., Favareto, M., D'Aprile, F., 1976, Analysis
Principles of optical disc systems, Adam Hiler LTD, of thermal behaviour of a machine tool table using
ISBN 0-85274-785-3. finite element method, Annals of the CIRP, Vol. 2511.
[Braembussche, 19961 [Cresto, 19911
Braembussche, P. van den, Swevers, J., Brussel, H. Cresto, P.C., Di Ciommo, M., Kancheva, T., Margues,
van, Vanherck, P., 1996, Accurate Tracking Control of D., Mudronja, V., Sartori, S., 1991, A method for the
Linear Synchronous Motor Machine Tool Axes, identification and correction of thermal deformations in
Mechatronics, Vol. 6, No. 5. Pp. 507-521. ~D-CMMS, Measurement,9(1).
[Braembussche, 19981 [Corbett, 19971
Braembussche, P. van den, Swevers, J., Brussel, H. Corbett, J., McKeown, Course notes in Precision
van, 1998, Control of the structural flexibility of a Engineering, School of Industrial and Manufacturing
machine tool, Proc. of the International Seminar on Science, Cranfield University, United Kingdom.
Improving machine tool performance, San Sebastian, [Damen, 19971
Spain, Vol. 1, pp. 339-350. Damen, A., 1997, Complete systemlcontrol strategies,
[Breyer, 19911 PAT0 Curcus Atieve' Magnetische Lagering,
Breyer, K.H., Pressel, H.G., 1991, Paving the way to Eindhoven University of Technology, The Netherlands.
thermally stable Coordinate Measuring Machines, [DeBra, 19911
Progress in Precision Engineering, pp. 56-75. DeBra, D.B., 1991, Vibration isolation, Stanford
[Brussel, 19981 University, CA, Tutorial Notes, ASPE Annual Meeting.
Brussel, H. van, Braembussche, P. van den, 1998, [DeBra, 19921
Robust control of feed drives with linear motors, DeBra, D.B. 1992, Vibration Isolation of Precision
Annals of the CIRP. Machine Tools and Instruments, Annals of the CIRP,
[Bryan, 19671 Vol. 4112, Keynote Paper.
Bryan, J.B. et al, 1967, American Machinist, Special [DeBra, 19941
Report. DeBra, D.B., Rivin, G., 1994, Active vibration isolation,
[Bryan, 19681 Stanford University, CA, Wayne State University,
Bryan, J.B:, International status of thermal error Detroit, Michigan, Tutorial Notes, ASPE Spring Topical
research, Annals of the CIRP, Vol. 1611, 1968. Meeting, Cincinnati, OH.
[Bryan, 1979al [DeBra, 19981
Bryan, J.B, 1979, Design and Construction of an Ultra- DeBra, D.B., 1998, Stanford University, CA, Private
Precision 84 inch Diamond Turning Machine, communication.
Precision Engineering, Vol. 1, No. 1, pp. 13-18. [Doebelin, 19901
[Bryan, 1979bl Doebelin, E.O., 1990, Measurement systems,
Bryan, J.B., Carter, D.L.. 1979, Design of a new error- Application and Design, McGraw-Hill International
corrected co-ordinate measuring machine, Precision Editions, ISBN 0-07-100697-4
Engineering, Vol. 1, No. 3, pp. 125-128. [Donaldson, 19801
[Bryan, 1979~1 Donaldson, R.R., 1980, Error Budgeting in Technology
Bryan, J.B., 1979, The Abbe Principle revisited: an of Machine Tools, Vol. 5, Machine Tool Accuracy (ed.
updated interpretation, Precision Engineering, Vol. 1, R.J. Hocken, 1980).
NO.3, pp. 129-132. [Donaldson, 19831
[Bryan, 19821 Donaldson, R.R., Patterson, S.R., 1983, Design and
Bryan, J. et al, 1982, An order of magnitude Construction of a Large Vertical Axis Diamond Turning
improvement in thermal stability with use of liquid Machine, (LODTM), Proc. SPlE Tech. Symposium, pp.
shower on a general purpose measuring machine, 433-438, UCRL. 89738.
ASME Technical Paper MR82-936. [Donaldson, 19841
[Bryan, 19841 Donaldson, R.R., Maddux, A.S., 1984, Design of a
Bryan, J.B., 1984, The Power of Deterministic High-Performance Slide and Drive System for a Small
Thinking in Machine Tool Accuracy, The first Precision Machining Research Lathe, Annals of the
International Machine Tool Engineers Conference, CIRP, VOI.3311, pp. 243-248.
Tokyo, Japan, Lawrence Livermore National [Donaldson, 19861
Laboratory Report UCRL-91531. Donaldson, R.R., Thompson, D.C., 1986, Design and
[Bryan, 19901 Performance of a Small Precision CNC Turning
Bryan, J.B., 1990, International status of thermal error Machine, Annals of the CIRP, Vol. 3511, pp. 373-376.
research, Annals of the CIRP, Vol. 3912. [Donmez, 19911
[Bryan, 19931 Donmez, M.A., 1991, Progress Report of the Quality in
Bryan, J.B., 1993, The deterministic approach in Automation Project for F490 NIST, Gaithersburg, MD,
metrology and manufacturing, Proc. of the ASME NlSTlR 4536.
International Forum on Dimensional Tolerancing and [Eastman, 19351
Metrology, Dearborn, Michigan. Easman, F.S., 1935, Flexure pivots to replace knife
[Bryan, 19961 edges and ball bearings, Univ. Wash. Engng Exp. Sta.
Bryan, J.B., 1996, Tutorial on Axis of Rotation, ASPE Bull. No. 86.
Annual Meeting, Monterey, CA. [Eastman, 19371
[Bryant, 19861 Easman, F.S., 1937, The design of flexure pivots, J.
Bryant, M.R., Velinsky, S.A., Beachley, N.H., Aero. Sci., 5(1),pp. 16-21.
Fronczak, F.J., 1986, A Design Methodology for [Edlen, 19661
Obtaining Infinite Stiffness in an Aerostatic Thrust Edlen, B., 1966, The refractive index of air, Metrologia,

580
 Keynote Papers

VOI.2, pp. 71-80. [Haitjema, 19971

[Eijk, 19851 Haitjema, H., Rosielle, P.C.N.J., Steijaert, H.A.M.,
Eijk, J. van, 1985, On the design of plate-spring 1997, A Calibration Platform for Generating Parallel
mechanisms, Ph.D. Thesis, Delft University of Displacements with Nanometre Accuracy, Proc. of the
Technology, The Netherlands. 9th IPES-IUME, Braunschweig, Germany, Vol. 2, pp.
[Eijk, 19981 456-459.
Eijk, J. van, 1998, Department Mechatronics, Centre [Haranat, 19871
For Manufacturing Technology, Philips Electronics Haranat, S., Ganesan, N., Rao, B., 1987, Dynamic
N.V., The Netherlands, Private communication. analysis of machine tool structures with applied
[Evans, 19891 damping treatment, International Journal of Machine
Evans, C., 1989, Precision Engineering, An Tools and Manufacturing, Vol. 27, pp. 43-55.
Evolutionary View, Cranfield Press, Bedford, [Henrikson, 19741
ISBN 1 87131501 8 Henrikson, C.L., Lyman, J., Studer, P.A., 1979,
[Ewins, 19841 Magnetically suspended momentum wheels for
Ewins, D.J., 1984, Modal Testing: Theory and spacecraft stabilization, AlAA 12th Aerospace
Practice, Research Studies Press Ltd, United Sciences Meeting, Washington, DC.
Kingdom. [Higuchi, 19901
[FEM, 19981 Higuchi, T. (ed.), 1990, Proc. of the 2nd International
Unigraphics@EDS Plano Texas; I-DEAS@SDRC Milfid Symposium on Magnetic Bearings, Tokyo, Japan.
Ohio; Algor@Pitsburg PA; Pro-Engineer@ [Hocken, 19771
[Ferri, 19951 Hocken D.J. et al, 1977, Three dimensional metrology,
Ferri, A.A., 1995, Friction Damping and Isolation Annals of the CIRP, Vol. 2612. Pp. 403-408.
Systems, Transactions of the ASME, Special 50th [Hocken, 19861
Anniversary Design Issue, Vol. 117, pp. 196-206. Hocken, R., et al. NIST, 1986, A general methodology
[Florussen, 19971 for machine tool accuracy enhancement by error
Florussen, G., 1997, Accuracy Analysis of Multi-Axis compensation, Precision Engineering, Vol. 814, pp.
Machines by Double Ball-Bar method, Masters Thesis, 187-196.
Eindhoven University of Technology, The Netherlands. [Hocken, 19951
[Florussen, 19981 Hocken, R., 1995, Course Notes, Principles Active in
Florussen, G., Delbressine, F., Schellekens, P., 1998, the Design of Precision Machines.
Accuracy analysis of multi-axis machines using a [Hoek, 1962-19861
double ball bar, Proc. of the lntzrnational seminar on Hoek, W. van der, 1962-1986. Het voorspellen van het
improving machine tool performance, San Sebastian, Dynamisch Gedrag en Positioneringsnauwkeurigheid
Spain. van constructies en mechanismen, Lecture note-book,
[Flugge, 19961 Eindhoven University of Technology, The Netherlands,
Flugge, J., 1996, Vergleichende Untersuchungen zur in Dutch.
mentechnischen Leistungsfahigkeit von [Hoek. 1985-19891
Laserinterferometern und inkrementellen Hoek, W. van der, Koster, M.P., 1985-1989, Des
Manstabmel3systemen. PTB-Bericht F-23, Thesis Duivels Prentenboek, Lecture note-book, Eindhoven
Technischen Hochschule Aachen, Germany, in University of Technology, The Netherlands, in Dutch.
Geman. [Holmes, 19961
[Gardner, 19911 Holmes, M., Trumper, D., 1996, Magnetidfluid-bearing
Gardner, J.W., Hingle, H.T., 1991, From stage for atomic-scale motion control (the angstrom
instrumentation to nanotechnology, Gordon and stage), Precision Engineering, Vo1.18, No.1, pp. 38-49.
Breach Science publishers USA, ISBN 2-88124-794-6. [Holster, 19671
[Ge, 19961 Holster, P.L., 1967, Gaslagers met uitwendige
Ge, P., Jouaneh, M., 1996, Modelling and Control of drukbron, P.T. Werktuigbouw, Vol. 22, No. 9, pp. 363-
Hysteresis in a Piezoceramic Actuator, Proc. of the 370, No. 10, pp. 415-425, No. 26, p. 1105, in Dutch.
ASPE Annual Meeting, Monterey, CA, pp. 396-399. [Holster, 19871
[Geary, 19541 Holster, P.L., Jacobs, J.A.H., 1987, Theoretical
Geary, P.J., 1954, Flexure devices, Published by the analysis and experimental verification on the static
British Scientific Instruments Research Association, properties of externally pressurized air-bearing pads
Research Report M.18. with load compensation, Tribology International, Vol.
[Gijsbers, 19801 20, NO.5, pp. 276-289.
Gijsbers, T.G., 1980, Colath, a numerically controlled [Hou, 19921
lathe for very high precision, Philips Technical Review, Hou, W., Wilkening, G., 1992, Investigation and
Vol. 39, NO.9, pp. 229-244. Compensation of the nonlinearity of heterodyne
[Goossens, 19971 interferometers, Precision Engineering, Vol. 14, p.p.
Goossens, H.J., 1997, Design of a long range tool 9 1-98
servo, Mechatronics post-masters design Thesis, [Hutcheson, 19961
Eindhoven University of Technology, The Netherlands, Hutcheson, G., Hutcheson, J., 1996, Technology and
ISBN 90-5282-785-0. Economics in the Semiconductor Industry, Scientific
[Groenhuis, 19911 American.
Groenhuis, H., 1991, A design tool for [Jones, 19511
electromechanical servo systems, Ph.D. Thesis, Jones, R.V., 1951, Parallel and rectilinear spring
Twente University, The Netherlands, movements, J. Sci. Instrum., Vol. 28, No. 38.
ISBN 90-9004089-1. [Jones, 19561
[Gross, 19831 Jones, R.V., Young, I.R., 1956, Some parasitic
Gross, H., 1983, Electrical Feed Drives for Machine deflections in parallel spring movements, J. Sci.
Tools, John Wiley & Sons Ltd, NY. Instrum., Vol. 33, No. 11.

581
[Jouy, 19861 Netherlands, Private communication.
Jouy, F., 1986, Theoretical modelisation and [Kouno, 19841
experimental identification of the geometrical Kouno, E., 1984, A Fast Piezoelectric Actuator for
parameters of coordinate measuring machines by Servo Correction of Systematic Errors in Precision
measuring a multidirected bar, Annals of the CIRP, Machining, Annals of the CIRP, Vol. 33,pp. 369-372.
Vol. 3511. [Kraakman, 19691
[Kamerbeek, 19831841 Kraakman, H.J.J., Gast, d., J.G.C., 1969, Een
Kamerbeek, E.M.H., 1983184, Magnetic bearings, precisiedraaibank met hydrostatische lagering en
Philips Technical Review, Vol. 41,No. 11112,pp. 348- aandrijving, Philips Technisch Tijdschrift, Jaargang 30,
361. No. 5, pp. 121-137,in Dutch.
[Kamerbeek, 19931 [Kraakman, 19761
Kamerbeek, E.M.H., 1997, Magnetische velden, Kraakman, H.J.J., 1976, Hydrostatische lagers met
krachten en actuatoren, PAT0 cursus Actieve verhoogde stijfheid, De Constructeur, Nov. 1976,No.
Magnetische Lagering, Eindhoven University of 1 1, pp. 51-57,in Dutch.
Technology, The Netherlands. [Kraakman, 19771
[Kanai, 19901 Kraakman, H.J.J., 1977, Hydrostatische lagers met
Kanai, A.. Sano, H., Yoshioka, J., Miyashita, M., 1990, verhoogde stijfheid 1-111, P.T. Werktuigbouw, Vol. 32,
Positioning of a 200 kg Carriage on Pain Bearing NO. 10,pp. 552-557,NO. 1 1 , pp. 611-616,NO. 12,pp.
Guideways to Nanometre Accuracy with a Force- 690-698,in Dutch.
Operated Linear Actuator, Nanotechnology, pp. 43-51. [Kraakman, 19971
[Kazimierski, 19921 Kraakman, H.J.J., 1997, Ontwerp van een precisie
Kazimierski, Z., Brzeski, L., Horodko, L., 1992, Gas bewerkingsmodule, kracht, precisie en doelmatigheid
Bearing of Infinite Stiffness, Transactions of ASME, door hydrauliek, Mikroniek, No. 6, pp. 182-187,in
Journal of Tribology, Vol. 114,No. 1 , pp. 270-273. Dutch.
[Kerstiens, 19901 [Krulewich, 1995al
Kerstiens, P., 1990,Strukturelemente fur Produktions- Krulewich, D., 1995, Bibliography on Error
maschinen aus carbonfaserverstarkten Kunststoffen, Compensation, LLNL, 94551,Livermore, CA.
Ph.D. Thesis, Technische Hochschule Aachen, [Krulewich, 199513)
Germany, in Geman. Krulewich, D., Hale, L., Yordy, D., 1995, Rapid
[Kim, 19961 Mapping of Volumetric Errors, Proc. of ASPE Annual
Kim, W-J., Trumper, D.L., 1996, Force Rippte in Meeting, Austin, USA, pp. 412-415.
Surface-Wound Permanent-Magnet Linear Motors, [Krulewich, 19981
Proc. of the lntermag, Seattle, WA. Krulewich, D.A., 1998, Rapid Mapping of Volumetric
[Klafter, 19891 Errors Using Distance Measurements, Proc. of
Klafter, R.D., et al., 1989, Robotic Engineering, an International Seminar on Improving Machine Tool
integral approach, Englewood Cliffs, Prentice-Hall Int. Performance, San Sebastian, Spain, Vol. 2,pp. 487-
ISBN. 0-13-782053-4 496.
[Klostermann, 19871 [Kruth, 19941
Klostermann, J.A., 1987,Een metaalkundige visie op Kruth, J.P., Vanherck. P., Jonge, d. L., 1994, Self-
de mechnanische eigenschappen van keramische calibration method and software error correction for
materialen, De Constructeur, Okt. 1987, No. 10, pp. three-dimensional coordinate, measuring machines
42-51,in Dutch. using artefact measurements, Measurement 14.
[Knapp, 19881 [Kunzmann, 19881
Knapp, W., 1988,Accuracy of Length Measurement Kunzmann, H., Waldele, F., 1988, Performance of
and Positioning: Statical Measurement and Contouring CMMs, Annals of the CIRP, Vol. 3712,pp. 633-640.
Mode, Annals of the CIRP, Vol. 3711. [Kunzmann, 19901
[Knapp, 19981 Kunzmann, H., Trapet, E., Waldele, F., 1990, A
Knapp, W., 1998, Ingenieurburo Knapp, Private uniform concept for calibration, acceptance test and
communication. periodic inspection of coordinate measuring machines
[Koren, 19751 using reference object, Annals of the CIRP, Vol. 3911.
Koren, Y., 1975,Interpolator for a CNC System, IEEE [Kunzmann, 19921
Trans. Comp. Vol. C-25 No. 1, pp. 32-37. Kunzmann H., Herrmann K., 1992, Calibration and
[Koren, 1981) testing of laser interferometers, Collection of papers
Koren, Y., Massory. O., '1981, Reference-Pulses and printed information of participants in the
Circular Interpolators for CNC Systems, ASME Journal EUROMET-Workshop held on 11th and 12th March
of Engineering for Industry, Vol. 103,No. 1, pp. 131- 1992,PTB-Bericht F-I 5,Berlin, Germany.
136. [Kunzmann, 19931
[Koren, 19971 Kunzmann, H., Pfeifer, T., Flugge, J., 1993,Scales vs.
Koren, Y., 1997,Control of Machine Tools, Journal of laser interferometers performance and comparison of
Manufacturing Science and Engineering, Vol. .I 19.pp. two measuring systems, Annals of the CIRP, Vol.
749-755. 4212,pp 753-767.
[Koster, 19961 [Kunzmann, 19961
Koster, M.P., Constructieprincipes, Uitgeverij Kunzmann, H., 1996, PTB, Braunschweig, Germany,
Universiteit Twente, Enschede, The Netherlands, Private communication.
ISBN 9036508320,in Dutch. [Lanxide, 19901
[Koster, 19971 LanxideTM,Reinforced ceramics and metals, Product
Koster, M.P., 1997, Smart Disc, Proc. of NVPT information, Lanxide, Newark, Delaware.
Jubileumcongres Precisie Technologie, Leusden, The [Lazan, 19681
Netherlands, in Dutch. Lazan, B.J., 1968,Damping of Materials and Members
[Koster, 19981 in Structural Mechanics, Pergamon, London, United
Koster, M.P., 1998, Twente University, The Kingdom.

582
 Keynote Papers

[Lindsey, 19881 [Mierlo, 19931

Lindsey, K., Smith, S.T., Robbie, C.J., 1988, Sub- Mierlo, H.A., van, 1993, Ontwerpen en testen van
Nanometre Surface Texture and Profile Measurement delay line systemen voor optische apertuur synthese,
with Nanosurf 2, Annals of the CIRP, Vol. 3711, pp. Masters Thesis, Eindhoven University of Technology,
519-522 The Netherlands, in Dutch.
[Lingard, 19911 [Mohrke, 19911
Lingard, P.S. Purss, M.E., Sona, C.M., Thwaite, E.G., Mohrke, G., 1991, Mehrdimensionale Geometrie-
Mariasson, G.H., Temperature perturbationeffects in a erfassung mit optoelektronischen Triangulations-
high precision CMM, Precision Engineering, Vol. 13. sensoren, Ph.D. Thesis Technischen Hochschule
NO.1, pp. 41-51. Aachen, Germany, in German.
[Loewen, 19801 [Moriwaki, 19951
Loewen, E.G., 1980, Metrology problems in general Moriwaki, T., Schamamoto, E., Kawano, M., 1995,
engineering, a comparison with precision engineering, Estimation of Thermal Deformation of Machine Tool by
Annals of the CIRP, Vol. 2912, pp. 451-453. Applying Neural network, Transaction of JSME, Part
[Loxham, 19701 C, Vol. 61, No. 584.
Loxham, J., 1970, The Commercial Value of [Murty, 19821
Investigations into Repeatability, Cranfield Unit for Murty, A.S.R., Padmanabhan, K.K., 1982, Effect of
Precision Engineering, Bedford, United Kingdom. surface topography on damping in machine joints,
[Lubin, 19821 Precision Engineering, Vol. 4, pp. 185-190.
Lubin, G., 1982, Handbook of composites, Van [Nakazawa, 19941
Nostrand Reinhold, NY. Nakazawa, H., 1994, Principles of Precision
[Marsh, 19961 Engineering, Oxford University press UK, ISBN 0-19-
Marsh, E.R.. Slocum, A.H., 1996, An integrated 856266-7.
approach to structural damping, Precision [Nashif, 19851
Engineering, Vol. 18, No. 213, pp. 103-109. Nashif, A.H., Jones, D.I.G., Herderson, J.P., 1985,
[Matsumura, 19961 Vibration Damping, John Wiley & Sons, NY.
Matsumura, F., Okada, Y., Futjita, M., Namerikawa, T., [Neumann, 19881
(ed.), 1996, Proc. of the 5th International Symposium Neumann, H.J., et al, 1988, CNC-Koordinaten-
on Magnetic Bearings, Kanazawa, Japan. rnentechnik, Expert Verlag, Ehningen, ISBN 3-8169-
[Maxwell, 18901 0220-0.
Maxwell, J.C., 1890, General considerations [Ni, 19971
concerning scientific apparatus, The Scientific Papers Ni. J., 1997, CNC Machine Accuracy Enhancement
of James Clerk Maxwell, W. D., Niven, Ed., Dover through Real Time Error Compensation, Journal of
Press, Reprint from the Handbook to the Special Loan Manufacturing Science and Engineering, Vol. 119.
Collection of Scientific Apparatus, 1876. [Nijs, 19881
[McKeown, 19731 Nijs, J. de, Schellekens, P., 1988, Modelling of
McKeown, P.A., Loxham, J., 1973, Some Aspects of Coordinate Measuring Machine for its Dynamic
the Design of high precision machines, Annals of the behaviour, Guyan Reduction Technique, Annals of the
CIRP, Vol. 22/1 pp. 139-140. CIRP, Vol. 3711.
[McKeown, 1973-19971 [Nijsse, 19971
McKeown, P.A., Dunsdale, J., and Wills-Moren, W.I., Nijsse, G., 1997, Verplaatsingssensoren, PAT0
1973-1997, The design of high precision machines cursus Actieve Magnetische Lagering, Eindhoven
and systems, Course Notes, College of Manufacturing, University of Technology, The Netherlands, in Dutch.
Cranfield, United Kingdom. [Paros, 19651
[McKeown, 19861 Paros, J.M., Weisbord, L., 1965, How to design flexure
McKeown, P.A., 1986, High Precision Manufacturing hinges, Machine Design, Nov. 1965.
and The British Economy, James Clayton Lecture, [Patterson, 19851
Inst. Mech. E. Roc., Vol. 200, No. 76. Patterson, S.. Magrab, E., 1985, Design and Testing
[McKeown, 19871 of a Fast Tool Servo for Diamond Turning, Precision
McKeown, P.A., 1987, The role of Precision Engineering, Vol. 7, No. 3, pp. 123-128.
Engineering in Manufacturing of the Future, Annals of [Paul 19811
the CIRP, Keynote, Vol. 3612 Paul, R.P., 1981, Robot Manipulators: Mathematics,
[McKeown, 19901 programming and Control, MIT Press, Cambridge, MA.
McKeowm, P.A., Carlisle, K., Shore, P., Read, R.F.J., [Peklinik, 19701
1990, Ultra-precision, high stiffness grinding CNC Peklinik, J., 1970, Adaptive Control of Manufacturing
grinding machines for ductile mode grinding of brittle Systems, Annals of the CIRP, Vol. 18/1, pp. 265-272.
materials, SPIE, Vol. 1320, Infrared Technology and [Plessers, 1988aj
Application, pp. 301-313 Plessers, P., Snoeys, R., 1988, Dynamic Identification
[McKeown, 19971 of Convergent Externally Pressurized Gas-Bearing
McKeown, P.A., 1997, New and Advanced Materials, Gaps, Transactions of the ASME, Journal of Tribology,
Special Article, Nanotechnology, Emerging Vol. 110, pp. 263-270.
Technology Series. [Plessers, 1988bl
[McKeown, 19981 Plessers, P., Snoeys, R., 1988, Dynamic Stability of
McKeown, P.A., 1998, Cranfield University, United Mechanical ’ Structures Containing Externally
Kingdom, Private communication, Pressurized Gas-Lubricated Thrust Bearings,
[Meijer, 19891 Transactions of the ASME, Journal of Tribology, Vol.
Meijer, J., 1989, From straightness to flatness, Ph.D. 110, pp. 271-278.
Thesis, Twente University,The Netherlands. [Philips, 19951
[Meirovitch, 19751 Philips, D.M.H., Rosielle, P.C.J.N., Schellekens,
Meirovitch, L., 1975, Elements of Vibration Analysis, P.H.J., 1995, Design of a High Precision Rotary Table,
McGraw-Hill, NY, pp. 93-100, 115-117. Proc. of 8th IPES, Compiegne, France, pp.221-224.

583
[Pohl, 19861 of tool, International Journal of Machine Tools and
Pohl, D.W., 1986, Some design criteria in scanning Manufacturing,Vol. 29, pp 361-376.
tunnelling microscopy, IBM Journal of Research and [Riven, 19961
Development, Vol. 30, p 428. Riven, E.I., 1996, Stiffness in Design, Wayne State
[Pollard, 19221 University, Detroit, Michigan, Tutorial Notes, ASPE
Pollard, A.F.C., 1922, The Mechanical Design of Annual Meeting, Monterey, CA.
Scientific Instruments, Cantor Lectures, Imperial [Ro, 19931
College of Science and Technology, Parsons Ltd. Ro, P.I., Hubbel, P.I., 1993, Model Reference Adaptive
Hastings. Control of Dual-Mode MicroIMacro Dynamics of Ball
[Pollard, 1929-19511 Screws for Nanometre Motion, ASME, Transactions
Pollard, A.F.C., 1929-1951, The Kinematic Design of Journal of Dynamic Systems, Measurement, and
couplings in Instrument Mechanisms, Imperial College Control, Vol. 115, pp. 103-108.
of Science and Technology, Hilger and Watts Ltd., [Rogers, 18831
London, United Kingdom. Rogers, W.A., 1883, Studies in metrology, Proc. of the
[Portman, 19971 American Academy of Arts and Sciences, Vol. 18, pp
Portman, V.T., Weill. R.D., Shuster, V.G., 1997, High 287-398.
order Approximation in Accuracy Computations of [Rolt, 19291
Machine Tools and Robots, Journal of Manufacturing Rolt, F.H., 1929, Gauges and fine measurements,
Science and Engineering, Vol. 119. Macmillan, London, United Kingdom.
[Pritschow, 19901 [Rosielle, 19911
Pritschow, G., Philipp, W., 1990, Direct Drives for Rosielle. P.C.J.N., 1991, Een meer efficiente vorm van
High-Dynamic Machine Tool Axes, Annals of the Gatscharnieren, Mikroniek, No. 4, pp. 104-108, in
CIRP, Vol. 39, pp. 413-416. Dutch.
[Quenelle, 19831 [Rosielle, 19941
Quenelle, R.C., Wuerz, L.J., 1983, A New Rosielle, P.C.J.N., 1994, Development and design
Micracumputer-Controlled Laser Dimensional rules for friction wheel drives, Proc. of ASPE Spring
Measurement and Analysis System, Hewlett-Packard Topical Meeting, Mechanisms and Control for Ultra-
Journal, 2-22. precision Motion, Cincinnati, OH.
[Rankers, 19941 [Rosielle, 19961
Rankers, A.M., Eijk, J. van, 1994, The Influence of Rosielle, P.C.J.N., Vermeulen, J.P.M.B., Reker,
Reaction Forces on the Behaviour of High E.A.G., 1996, Laminaten, vervaardigd uit epoxy-
Performance Motion Systems, Proc. of the Second verlijmde, dunne AlzO3-keramische tegels, geschikt
International Conference on Motion and Vibration voor sledeconstructies in precisietoepassingen,
Control, Yokohama, pp. 71 1-716. Licence no. 1002369,in Dutch.
[Rankers, 19971 [Rosielle, 19981
Rankers, A.M., 1997. Machine dynamics in ‘ Rosielle, P.C.J.N., Reker, E.A.G., 1998,
mechatronic systems, An engineering approach, Ph.D. Constructiepricipes 1, Bedoeld voor het nauwkeurig
Thesis, Twente University, The Netherlands, ISBN 90- bewegen en positioneren, Lecture note-book,
365-0957-2 Eindhoven University of Technology, The Netherlands,
[RaO, 19901 in Dutch.
RaO, S.S., 1990, Mechanical Vibrations, Addison [Sabnis, 19741
Wesley Publishing Company, USA, ISBN 0-201- Sabnis, A.V., Dendy, J.B., Schrnitt, F.M., 1974,
50156-2. Magnetically suspended large momentum wheels,
[Reindl, 19671 Proc. of AlAA Mechanics and Control Flight Conf.,
Reindl, R., 1967, Dr Otto Eppenstein - A Scientist of Anaheim, CA.
the Jena Zeiss works, Jena Review, No. 5, pp 263- [Sartori, 19881
267. Sartori, S., Cresto, P.C., Dicionimo, M., Kancheva, T.,
[Reinhart, 19971 1988, A way to improve the accuracy of a coordinate
Reinhart, G., Weiszenberger, M., 1997, measuring machine, Measurement E(2).
Eigenschaftsoptimierung am virtuellen Prototyp, ZWF [Sartori, 19951
92, 1997,11,Germany. Sartori, S, Zang, G.X., 1995, Geometric Error
[Renkens, 19941 Measurement and Compensation of Machines, Annals
Renkens, M.J.M., 1994, Ontwerp van een geregeld of the CIRP, Vol. 4412
axiaal lager, Mechatronics post-masters design [Sartori, 19981 *

Thesis, Eindhoven University of Technology, The Sartori, A., 1998, lstituto di Metrologia “G. Colonnetti”,
Netherlands, ISBN 90-5282-329-4,in Dutch. Italy, Private communication
[Renkens, 19971 [Schellekens, 19861
Renkens, M.J.M., 1997, Design of an Axially Schellekens, P.H.J., 1986, Absolute meetnauwkeurig-
Controlled Spindle Unit for High Precision Diamond heid van technische laserinterferometers, Ph.D.
Turning, Ph.D. Thesis, Eindhoven University of Thesis, Eindhoven University of Technology, The
Technology, The Netherlands, ISBN 90-74445-35-7. Netherlands, in Dutch.
[Riven, 1979al [Schellekens, 19921
Riven, E., 1979, Vibration Isolation of Precision Schellekens, P.H. et al, 1992, Modelling of thermal
Machinery, Sond and Vibration behaviour of machine tools, Development of Methods
[Riven, 1979bl for numerical error correction of Machine,tools, Bureau
Riven, E., 1979, Principles and Criteria of Vibration Communantaire de References, 33201110116018918 -
Isolation of Machinery, ASME, Journal of Mech. BCR - NL, Brussels, Belgium.
Design, Vol. 101, p 682. [Schothorst, 19971
[Riven, 19891 Schothorst, G., 1997, Piezo aangedreven vervormbare
Riven, E., Kang, H., 1989, Improvement of machining spiegel, Proc. of NVPT Jubileumcongres Precisie
conditions for slender parts by tuned dynamic stiffness Technologie. Leusden, The Netherlands, in Dutch.

584
 Keynote Papers

[Schouten, 19971 [Takahashi, 19901

Schouten, C.H., 1997,Het ontwerp van een precisie Takahashi, M., Otsuka, J., Ono, K., 1990, Study of
spindel voor hoge snelheden, Mechatronics post- Precision Positioning using a Friction Drive (I),
masters design Thesis, Eindhoven University of Structure and Traction Characteristics, Bull. JSPE,
Technology, The Netherlands, ISBN 90-5282-754-0, in VOI.24,NO.1, pp. 21-26.
Dutch. [Taniguchi, 19831
[Schweitzer, 19881 Taniguchi, N., 1983, Current status in and future
Schweitzer, G. (ed.), 1988, Proc. of the 1st trends of ultra-precision machining and ultra fine
International Symposium on Magnetic Bearings, materials processing, Annals of the CIRP, Vol. 3U2
Springer-Verlag, Berlin, Germany. [Taniguchi, 19961
[Schweitzer, 19941 Taniguchi, N., 1996, Nanotechnology: Integrated
Schweitzer, G., Siegwart, R., Herzog, R. (ed.), 1994, processing systems for ultra precision and ultrafine
Proc. of the 4th International Symposium on Magnetic products, Oxford University Press, UK, ISBN 0-19-
Bearings, VdF-Verlag, Zurich, Switzerland. 8562837.
[Shpitalni, 19941 [Teague, 19861
Shpitalni, M. Koren, Y., Lo, C., 1994,Real-Time Curve Teague, E.C., 1986,Room temperature gold-vacuum-
Interpolators, Computer Aided Design, Vol. 26, pp. gold-gold tunnelling experiments, NBS Journal of
832-838. Research, Vol. 91,p 171.
[Slocum, 19921 [Teague, 1989-19971
Slocum, A,, 1992. Precision Machine Design, Teague, E.C., Evans, C., 1989,Patterns for precision
Englewood Cliffs, NJ, Prentice Hall. instrument design (mechanical aspects), National
[Slocum, 1994al Institute of Standards and Technology, Tutorial Notes,
Slocum, A.H., Marsh, E.R., Smith, D.H., 1994,A new ASPE Annual Meeting, Norfolk, Virginia.
damper design for machine tool structures: the Teague, E.C., 1997, Basic Concepts for Precision
replicated internal viscous damper, Precision Instrument Design: Designing Instruments and
Engineering, Vol. 16,No. 3,pp. 174-183. Machines to Have a High Degree of Repeatability,
[Slocum, 1994bJ National institute of Sfandards and Technology,
Slocum, A., Scagnetti, P., Kane, N., 1994, Ceramic Tutorial Notes, ASPE Annual Meeting, Norfolk,
Machine Tool Axis with Self-compensated Water- Virginia.
Hydrostatic Linear Bearings, Proc. of the ASPE Teague, E.C., 1997, Basic Concepts for Precision
Annual meeting, Cincinnati, OH, pp. 57-60. Instrument Design: Some Draft Notes and Papers,
[Smith, 19921 National Institute of Standards and Technology,
Smith, S.T., Chetwynd, D.G., Foundations of Ultra- Tutorial Notes, ASP€ Annual Meeting, Norfolk,
precision Mechanism Design, Gordan and Breach Virginia.
Science Publishers. [Teeuwsen, 19891
[Snoeys, 19771 Teeuwsen, J., Soons, J., Schellekens, P., 1989, A
Snoeys, R., Devrieze, L., Vanherck. P., 1977, Self General Method for error description of CMM's using
aligning aerostatic shoe bearing, Annals of the CIRP, polynomial fitting procedures, Annals of the CIRP, Vol.
Vol. 2511,pp. 205-210. 3811.
[Snoeys, 19871 [Tensator, 19971
Snoeys, R., Al-Bender, F., 1987, Development of Tensator, Constant Force Springs, Catalogue 2nd
improved externally pressurized gas bearings, KSME Edition, Tensator Limited, United Kingdom.
Journal, Vol. 1, No. 1, pp. 81-88. [Thompson, 19891
[Soons, 19931 Thompson, D.C., 1989, The design of an ultra-
Soons, H., 1993, Accuracy Analysis of Multi-Axis precision CNC measuring machine, Annals of the
Machines, Ph.D. Thesis, Eindhoven University of CIRP, Vol. 3911
Technology, The Netherlands. rimoshenko, 19741
[Spaan, 19951 Timoshenko, S., Young, D.H., Weaver, W., 1974,
Spaan, H.A.M., 1995,Software Error Compensation of Vibration Problems in Engineering, John Wiley, NY.
Machine Tools, Ph.D. Thesis, Eindhoven University of [Tomizuka, 19871
Technology, The Netherlands. Tomizuka, M., Zero Phase Error Tracking Algorithm
[Spies, 19971 for Digital Control, ASME Journal of Dynamic
Spies, A., 1997,Linear and Angular Encoders for the Systems, Measurement and Control, Vol. 109,pp. 65-
High-Resolution range, Proceedings of the gth 68.
International Precision Engineering Seminar. [Tomizuka, 19891
[Srinivasan, 19921 Tomizuka, M., 1989, Design off Digital Tracking
Srinivasan, N., Ziegert, J., Smith, S., 1992,Prediction Controllers for Manufacturing Applications, ASME,
of Positional Errors of a Three Axis Machine Tool Manufacturing Review, Vol. 2,No. 2,pp. 134-141.
Using Neural Network, JapanlUSA, Symposium on [Trapet, 1997al
Flexible Automation. Trapet, E., 1997, Physikaiisch Technische
[Srinivasan, 19971 Bundesanstalt, Braunschweig, Germany, Workshop
Srinivasan, K., Tsu Chin Tsao, 1997, Machine Tool on Traceability of Coordinate Measuring Machines,
Feed Drives and Their Control - A Survey of the State Oct. 9-10,1997.
of the Art, Journal of Manufacturing Science and rrapet, 1997b1
Engineering, Vol. 119,pp. 743-748. Trapet, E., 1997, PTB, Braunschweig, Germany,
[Stix, 19961 Private communication.
Stix, G., 1996,Waiting for Breakthroughs, Scientific [Trumper, 19931
American. Trumper, D.L., Williams, M.E., Nguyen. T.H., 1993,
[Stout, 19811 Magnet Arrays for Synchronous Machines, Proc. of
Stout, K.J., 1991,The design of externally pressurized the IEEE Ind. Appl. SOC. Annual Meeting, Toronto,
gas bearings. Canada.

585
rrumper, 19961 [Wetzels, 19981
Trumper, D.L., Kim, W.J., Williams, M.E., 1996, Wetzels, S.F.C.L., 1998, Laser Based Displacement
Design and Analysis Framework for Linear Calibration with Nanometre Accuracy, Ph.D. Thesis,
Permanent-Magnet Machines, IEEE Transactions on Eindhoven University of Technology, The Netherlands,
Industry Applications, Vol. 32, No. 2, pp. 371-379. ISBN 90-386-0730-X
[Tsutsumi, 19791 [Wills-Moren, 19821
Tsutsumi, M., Ito, Y., 1979, Damping mechanism of a Wills-Moren, W.J., Modjarrad, A. Read, R.F.J., and
bolted joint in machine tools, Proc. of the International McKeown, P.A., 1982, Some Aspects of the Design
Machine Tool Design Research Conference, 1979, pp. and Development of a large, High Precision CNC
443-448. Diamond Turning Machine, Annals of the CIRP, Vol.
[VDI, 19851 31, pp. 409-414.
VDI Berichte 548, 1985, Dokumentation [Wills-Moren, 19891
Laserinterferometrie in der Langenmesstechnik, Wills-Moren, W.J., Wilson, T., 1989, The Design and
Verlag des Vereins Deutcher lngenieure Dusseldorf, Manufacture of a large CNC Grinding Machine for Off-
Germany, in German. axis Mirror Segments, Annals of CIRP, Vol. 3811
[Veldhuis, 19941 [With, 19961
Veldhuis, S., Elbestawi, M., 1994, Modelling and With, d., G., Dortmans, L.J.M.G., 1996, Constructief
compensation for five-axis machine tool errors, gebruik van brosse materialen, De Constructeur, Okt.
International Mechanical Engineering Congress and 1996, No. 10, pp. 30-37, in Dutch.
Exposition Conference, ASME, Chicago, Vol. 6812. [Wu. 19961
[Vermeulen, 1996al Wu, C., Su, C., 1996, Nonlinearity in measurements of
Vermeulen, J.P.M.B., 1996, Development of a length by optical interferometry, Meas. Sci. Technol.
lightweight and ultrastiff single point diamond turning Vol. &, pp. 62-68
machine with submicron machining accuracy, [wulp, 19961
Mechatronics postmasters design Thesis, Eindhoven Wulp, H. van der, Beek, H.F. van, Mertens, B.M.,
University of Technology, The Netherlands, ISBN 90- 1996, A piezodriven TEM stage with subnanometre
5282-702-8. position stability, Proc. of ASPE Annual Meeting, pp.
[Vermeulen, 1996bl 26-31.
Vermeulen, J.P.M.B., Rosielle, P.C.J.N., Schellekens, (Yang, 19961
P.H.J., 1996, Innovations in the design and Yang, S., Yuan, J., Ni, J., 1996, Real Time Cutting
development of ultraprecision machines, Lightweight Force Induced Error Compensation on a Turning
and ultrastiff single point diamond turning machine Center, Proceedings of the Second S.M. Wu
with submicron accuracy, Proc. of ASPE Annual Symposium, Ann Arbor.
Meeting, Monterey, CA, pp. 631-636. [Yoshida, 19671
[Vermeulen, 19981 Yoshida, Y., Honda, F.. 1967, Thermal deformations of
Vermeulen, M.M.P.A., Rosielle, P.C.J.N., Schellekens, machine tool structure-the bed of a lathe, Annals of
P.H.J., 1998, Design of a High-precision 3D- the CIRP, Vol. 15/1.
Coordinate Measuring Machine, Annals of the CIRP, [Young, 19721
Vol. 4711 Young, R.D., Ward, J.. Scire, F., 1972, The
[Vorburger, 19971 Topografiner: an instrument for measuring surface
Vorburger, T.V., Dagata, J.A., Wilkening, G., lizuka, topography, Review of Scientific Instruments, Vol. 43,
K., 1997, Industrial Uses of STM and AFM, Annals of p 999.
the CIRP, Vol. 4612, Keynote Paper [Zhang, 19851
[Wang, 19931 Zhang, G.J., et al, 1985, Error compensation of
Wang, J., 1993, Design of Gas Bearing Systems for coordinate measuring machines, Annals of the CIRP,
Precision Applications, Ph.D. Thesis, Eindhoven Vol. 3411.
University of Technology, The Netherlands, ISBN 90- [Ziegert, 19941
386-0112-3. Ziegert, J.C., Mize, C.D., 1994, The laser Ball Bar: A
[Weck, 19841 new Instrument for Machine Tool Metrology, Journal of
Weck, M., 1984, Handbook of Machine Tools, Vol. 3, the American Society for Precision Engineering, Vol.
Automation and Controls, Wiley; NY. 16, No. 4.
[Weck, 1995a)
Weck, M., McKeown, P., Bonse, R., Herbst, U., 1995,
Reduction and Compensation of Thermal Errors in
Machine Tools, Annals of the CIRP, Vol. 4412.
[Weck, 1995bl
Weck, M., dezmeral, H., Mehlkopp, K., Terwei, T.,
1995, New Hybrid Concept for long Stroke Fast-Tool-
Servo-System, American Society For Precision
Engineering, Proc. Vol. 12, pp. 211-214.
[Weck, 19971
Weck, M. Oezmeral, H., Mehlkopp, K., Terwei, T., A
Long Stroke Fast-Tool-Servo with Air Bearings,
Progress in Precision Engineering and
Nanotechnology, Proc. of 9th IPES4UME, Vol. 2, pp.
393-394
weekers, 1997)
Weekeis, W.G., 1997, Compensation for Dynamic
Errors of Coordinate Measuring Machines, Ph.D.
Thesis, Eindhoven University of Technology, The
Netherlands. ISBN 90-386-0178-6.

586