Professional Documents
Culture Documents
Machine Tool Feed Drives
Machine Tool Feed Drives
Y. Altintas (1)
a,
*, A. Verl (2)
b
, C. Brecher (2)
c
, L. Uriarte (3)
d
, G. Pritschow (1)
b
a
Manufacturing Automation Laboratory, University of British Columbia, Vancouver, Canada
b
Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW), Stuttgart, Germany
c
Machine Tool Laboratory (WZL), University of Aachen, Germany
d
Mechatronics Dept. Tekniker-IK4, Eibar, Basque Country, Spain
1. Introduction
Feed drives are used to position the machine tool components
carrying the cutting tool and workpiece to the desired location;
hence their positioning accuracy and speed determine the quality
and productivity of machine tools. A general architecture of feed
drive hardware and its computer control structure are shown in
Fig. 1. Feed drives are either powered by linear motors directly, or
by rotary motors via ball screw and nut assembly as shown in
Fig. 2. The drive trainconsists of a machine tool table resting on the
guide, and moved linearly either by a ball screwdrive-nut or by a
linear motor system. The ball screw may be connected to the
rotary servo motor directly or via gear reduction for large
machines. The motor is powered by amplier electronics
connected to a Computer Numerical Control (CNC) system as
shown in Fig. 1a. The table is positioned by the servo drives by
following a trajectory generation and control algorithm as shown
in Fig. 1b. An NC program generated in CAD/CAM systemis loaded
to the CNC unit of the machine tool. CNC parses the NC program
into tool path segments which may consist of linear, circular,
spline or other geometric motions. The feedrate entered in the NC
program is combined with the acceleration and jerk limits of the
feed drives, and time stamped discrete position commands are
sent to each drive servo by the real time trajectory generation
algorithm. The trajectory generation algorithm considers the
kinematics of the machine in decoupling the spatial tool motion
into each feed drive. Modern CNC units use jerk continuous, i.e.
fth order polynomials, to generate position commands at each
discrete time interval [3]. The discrete position commands are
processed by real time control laws of each servo drive, and the
corresponding digital velocity commands are converted into
electrical signals which are fed to the amplier and motor of the
drive. The speed and accuracy of positioning the machine tool are
affected by the trajectory generation and control algorithms,
mechanical drives and guides, ampliers, motors and sensors
used in each feed drive.
By considering the importance of the topic, CIRP published two
key note articles on feed drive systems. Koren et al. presented a
survey of feedback, feedforward, and cross-coupled controllers
applied on feed drives [62,63]. Pritschow et al. compared the
performance of linear drives against the conventional electro-
mechanical ball screw drives Electromechanical Drives [83]. The
latest CNC design architecture has also been surveyed by
Pritschow et al. [84]. This keynote paper presents a review of
recent technological developments and academic advances
achieved in feed drive systems. Ongoing research challenges
are also discussed in order to push the feed drive accuracy and
performance to higher levels.
The paper is organized as follows. The machine tool guides
based on friction, roller bearing, hydrostatic and levitation
principles are presented in Section 2. The rack-pinion, ball
screw and linear drive structures are given in Section 3, followed
by their structural dynamic models in Section 4. Electric
motors and sensors used in feed drives are discussed in Sections
5 and 6, respectively. The control of rigid and exible feed
drives is presented in Section 7. The paper is concluded by
highlighting the current research challenges in feed drive design
and control.
2. Machine tool guides
The roller-based guides have gained popularity due to their
high performance and modular integration to machine tools. On
the other hand, hydrostatic guides are preferred in the applications
where higher accuracy, stiffness and damping are required.
Aerostatic, magnetic or vacuum guides are used in the precision
positioning applications where the external load is small.
CIRP Annals - Manufacturing Technology 60 (2011) 779796
A R T I C L E I N F O
Keywords:
Feed
Drive
Machine tool
A B S T R A C T
This paper reviews the design and control of feed drive systems used in machine tools. Machine tool
guides designed using friction, rolling element, hydrostatic and magnetic levitation principles are
reviewed. Mechanical drives based on ball-screw and linear motors are presented along with their
compliance models. The electrical motors and sensors used in powering and measuring the motion are
discussed. The control of both rigid and exible drive systems is presented along with active damping
strategies. Virtual modeling of feed drives is discussed. The paper presents the engineering principles and
current challenges in the design, analysis and control of feed drives.
2011 CIRP.
* Corresponding author.
Contents lists available at ScienceDirect
CIRP Annals - Manufacturing Technology
j ournal homepage: ht t p: / / ees. el sevi er. com/ ci rp/ def aul t . asp
0007-8506/$ see front matter 2011 CIRP.
doi:10.1016/j.cirp.2011.05.010
The required functional r for the longitudinal guides are the
following:
Geometric accuracy since it is translated to the part directly.
Stiffness to withstand the machining process and the inertial
forces with minimum deformation.
Wear resistance and low friction to avoid gripping, stickslip
phenomena and aging of the surfaces.
Toughness to withstand impacts from the machining process.
2.1. Friction guides
Friction guides have good damping, strength against impact
loads and high load capacity with up to 140 MPa. They are
primarily used in speeds under 0.5 m/s. Uniform contact with
minimum adhesion between the bed and slide is obtained by
scraping and leaving uniform marks on the contact surfaces. The
slide-ways are lubricated by 1 mm deep lubrication slots opened
on the moving part of the guides. The guides can also be coated
with few mm thick polymers in order to reduce the friction.
Various congurations of friction guides are shown in Fig. 3.
Different pairs of materials are used to manufacture friction
guides. Casting, steel, bronze and certain polymers are used as
sliding materials. A key factor to ensure the controllability and
smooth operation of the guide is to avoid the stickslip
phenomenon which appears when the static friction coefcient
is higher than the dynamic friction coefcient. The development
of polymer based materials with additives which favour
lubrication, for example, Turcite
TM
, SKC
1
or Moglice
1
, has
allowed to a great extent the reduction of the stickslip
phenomenon (see Figs. 4 and 5).
2.2. Rolling guides
Recirculating and stationary roller-based guides are most
widely used in present machine tool applications (see Fig. 6).
The stationary rolling bearings tend to be used when the stroke of
the slide is relatively short. The rolling elements can be steel balls,
rollers or needles, which are preloaded between two cages
attached to stationary and moving parts of the guides. They have
Fig. 1. Architecture of a feed drive control system. (a) Physical components of a feed
drive (WZL). (b) Feed drive control algorithms (UBC).
Fig. 2. Linear and ball-screw drive mechanisms.
Fig. 3. Congurations of friction guides (by Busak Shamban
1
).
Fig. 5. Coated slideways in a table of a gantry milling machine (by SKC
1
).
Fig. 4. Comparative diagram of frictional behaviour (SKC
1
).
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 780
low friction, high load capacity and stiffness, but with low
structural damping. The recirculating rolling guides are manu-
factured with different size and load capacity by the manufacturers
as listed in Table 1. They can be supplied with integrated position
sensors or racks to expedite the design and assembly of machine
tool drives.
Fig. 7 shows the clamping systems for linear guide technologies.
To release the clamping action by safety clamps (see Fig. 7 top), the
chamber between the two spring diaphragms is lled with
compressed air. The diaphragms deect outwards, allowing the
clamping body to return to its original relaxed position. The brake
blocks lift off the rail. When the chamber is not inated, the
diaphragms move back pushing apart the upper clamp body. With
the horizontal strut acting as fulcrum point, the brake blocks are
forced against the linear rail, thus clamping the carriage. The
operating pressure amounts to 56 bar. To activate the clamping
action by clamps with air (see Fig. 7 bottom), the chamber under
the spring diaphragm is lled with compressed air. The spring
sheet is pushed upwards and stretched. With the horizontal strut
acting as fulcrum point, the brake blocks are forced against the
linear rail, thus clamping the carriage. When the chamber is not
inated, the diaphragm moves back into the bended position and
the upper clamp body goes back in its relaxed position. The brake
blocks lift off the rail. By clamps with air the operating pressure
amounts to 46 bar.
2.3. Hydrostatic guides
The sliding surfaces of the components are separated by cells
pressurized with thin oil, hence stick slip and static friction are
avoided in hydrostatic guides [55,81,96]. Oil is released to the
surrounding land as it loses pressure (see Fig. 8). The distance
between the land and the surface which slides over it is the oil
gap, h, which is about 1040 mm. The oil gap generates resistance,
R
c
against the ow of the uid, Q, from the oil cell to the outside.
The pressure difference between the cavity and the atmospheric
pressure which acts on the outside is known as cell pressure, p
c
.
The oil ow resistance along the gap can be estimated as:
R
c
Dp
Q
12hL
bh
3
(1)
where h is the dynamic viscosity coefcient of the oil, L is the land
length and b is the total width of the land along its perimeter. The
pressure gradient along the length of the land can be assumed to be
linear, and the pressure acts on up to one half of the length of the
land for initial force prediction. The complete pressure acts on the
effective area (A
eff
). The hydraulic suspension force, F, is calculated
as follows:
F P
c
A
e f f
(2)
Table 1
Load capacity of guiding systems.
Size no. Outer dimensions Basic load rating, Radial, reverse radial and side
Height, M (mm) Width, W (mm) Length, L (mm) Dynamic rating, C (kN) Static rating, C0 (kN)
15 24 34 64.4 14.2 24.2
20 30 44 79 22.3 38.4
25 36 48 92 31.7 52.4
30 42 60 106 44.8 66.6
35 48 70 122 62.3 96.6
45 60 86 140 82.8 126
55 70 100 171 128 197
65 90 126 221 205 320
Fig. 6. Linear guides with rolling bearings. (a) Guides with linear, stationary rolling
bearings (INA
1
). (b) Recirculating rolling guide model (THK
1
). (c) Linear rolling
guides by Schneeberger
1
.
Fig. 7. Clamping systems for linear guides (HEMA
1
).
Fig. 8. Hydrostatic V-at and wrap-around guides (by Hydrostatic
1
).
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 781
The hydrostatic guides are designed with several cells, such that
they can support off-centre forces and moments. Each cell is
supplied at a different pressure, in order to withstand forces at
varying operating conditions. Usually a single pump is used with
restrictors to supply each cell at the appropriate pressure. The
pressure in a cell is estimated as:
p
c
p
p
R
c
R
k
R
c
(3)
where p
p
is the supply pressure and R
k
is the restrictor ow
resistance. The restrictors are built in the form of capillaries where
the resistance depends on the viscosity of the oil in the cells. The
resistance to the oil ow of a capillary is estimated as:
R
k
8hL
k
pr
4
k
(4)
where L
k
and r
k
are the length and radius of the capillary,
respectively. The short restrictors should have a very small
diameter to provide the necessary resistance. This diameter is
limited by the size of suspended particles in the oil which can block
the capillary. The use of short capillaries is also limited because the
design is very sensitive to their diameter; the resistance depends
on the fourth power of the diameter. Capillaries with larger
diameters and a longer length are also used, which tend to be in
spiral form.
The restrictor automatically increases the cell pressure (Eq. (3))
and tends to decrease the oil gap and increase the ow resistance
(Eq. (1)) when external forces are applied. The oil gap is kept
almost stable near the equilibrium gap height, h
0
. The stiffness of
the system is formed from the pump, restrictor and cell at the
equilibrium state of the gap. In case of a cell supplied through a
capillary by means of a pump which operates at a constant
pressure, the resulting stiffness is,
K 3
F
0
h
0
R
k
R
k
R
c0
(5)
where F
0
is the suspension force at equilibrium. As the stiffness
depends on the load supported by the cell, hydrostatic guides with
retaining plate are used in order to apply high preloads. In
addition to gaining stiffness, the guide can absorb loads in both
directions.
The hydrostatic guides provide much higher damping values
than the roller-based guides against motions perpendicular to the
cell. The source of the damping is the friction force that the oil
lamina presents on sliding. In case of motions parallel to the cell,
the damping is reduced because there is no displacement of uid.
The temperature changes in the oil affect the viscosity, hence
the damping and lifting force of the system. The temperature can
be estimated from the total work done by the pump and friction
forces. The work carried out by the pump is:
W
p
Q p
p
e
(6)
where e is the pump efciency. The source of the friction work (W
r
)
is the translational motion of the table on the guides. The shear
stress of the oil is:
t
r
h
v
h
(7)
where v is the translational velocity. The work carried out to
overcome the hydrodynamic friction can be determined as
W
r
t
r
A
r
v A
r
h
v
2
h
(8)
where A
r
is the land area.
The dimensioning of a hydrostatic guide is relatively complex
due to the strong dependency on the temperature rise of the oil
throughout the hydraulic circuit.
There are several challenges in optimal design of hydrostatic
guides. It is desired to have as small oil gaps as possible. However,
the manufacturing accuracy, elastic deformations of the slides, the
loss of oil viscosity due to temperature rise, and poor regulation of
oil ow and pressure from a central pump may diminish the gap
between the sliding surfaces. Especially, the oil viscosity and
temperature rise are highly coupled; hence the mathematical
modeling of the interaction between the two needs to be iterative
in order to predict the equilibrium of their states. A comprehensive
mathematical model of the hydrostatic guide design, which
predicts oil circuit dynamics (i.e. pressure, ow velocity, normal
and friction force distribution), ow energy, temperature, oil
viscosity, elastic deformation of sliding surfaces, and prediction of
dynamic oil cell gap and its stiffness could be highly useful tool for
machine tool designer.
Most machine builders opt for proled linear rail guides, which
provide a good combination of performance and ease of installa-
tion. Hydrostatic guides are relatively expensive, time-consuming
to mount and require a larger design envelope compared to
proled rail guides. Paepenmu ller have combined the positive
properties of both rolling element and hydrostatic guideways to
form a new hydrostatic compact linear guide [127]. This new
guideway system is based on the standardized dimensions of
rolling element guideways. The new hydrostatic linear guides are
now a cost-effective option, as they offer both high precision and
better damping characteristics for high dynamic rigidity and much
greater design freedom (see Fig. 9).
2.4. Guides for ultraprecision applications
The use of cylindrical non-circulating roller bearings is
commonly used in the linear carriages of ultra precision machines.
The conguration should be kinematic [104] with the rollers
contacting the minimum degrees of freedom needed (see Fig. 10).
The preload force is estimated to t the application and the related
loads that the carriage is expected to experience during operation.
When high precision is the primary objective, a constant preload
Fig. 9. Hydrostatic compact linear guide (Schaefer
1
).
Fig. 10. Ultraprecision rolling guides in a semi-kinematic conguration.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 782
method is advantageous over constant displacement solutions,
due to its ability to hold the imperfections of the rolling elements.
Constant force preload can be achieved in different ways, e.g.
created by the weight of the system itself in the case of horizontal
carriages, by means of counterbalancing masses in vertical
carriages or by spring-based locking systems.
Hydrostatic and aerostatic guides are also widely used in
precision positioning of tables due to reduced friction and wear
[69,70]. They are insensitive to small random irregularities in the
ways and pads, producing the so-called averaging effect. In the
case of aerostatic guides, where load is supported over a thin lm
of air, the low viscosity and compressibility of air require high
manufacturing and assembly tolerances. The aerostatic guides
present low losses due to low friction [35,98,109] and the low
viscosity of air, even at high speeds. The air viscosity is
approximately three orders of magnitude less than those of the
oils used in hydrostatic guides at the ambient temperature. In
addition, it is not necessary to return the uid. The viscosity of air is
very low and does not vary in a wide range of temperature. The
aerostatic guides are the best option for applications with a wide
range of temperature variation. The compressibility and poor
damping of the air make the dimensioning of aerostatic guides
more difcult than hydrostatic guides. They may exhibit self-
excitation, known as air hammering, which may degrade the
operation of the guides. This vibration can be mitigated through
the use of a high number of cells, each with its own restrictor.
Recent developments in modeling have allowed the improvement
of their stability and their active compensation [1,2].
Presently, the guides are manufactured from porous materials
[80,99], and the porosity emulates the operation of multiple
restrictors. Aerostatic guides are seldom used in machine tool
applications, mainly due to the difculty in attaining sufcient
load capacities at a reasonable cost [105]. It is necessary to
machine the sliding surfaces to a very high degree of accuracy, such
that the gap can be extremely small. Typical gap values can be in
the order of 510 mm for a standard air pressure supply of 6 bar
owing in a continuous way to the atmosphere. However
aerostatic guides are widely used in high speed, precision
positioning steppers for electronics manufacturing and micro/
nano-machining machine tools. In particular, the vacuum pre-
loading of air bearings offers an excellent method of achieving a
compact aerostatic guideway designed, for planar motion, which
provides both vertical and angular pitching stiffness [40,98]. They
can meet precision positioning demands [8,9]. The lack of damping
of aerostatic guides is overcome by additional damping elements,
as electro-rheological uid dampers [100]. A new principle of uid
bearing, which utilizes travelling waves produced by piezoelectric
actuators, is proposed in [97]. Denkena et al. presented an
aerostatic linear guide for microsystems with dimensions of
8.4 mm 1 mm and the air is supplied from capillaries with
0.15 mm diameter [29]. The application of various aerostatic and
electro-magnetic guide design concepts can be found in [24].
2.5. Magnetic levitation guides
Magnetic levitation guides are mainly used in precision
positioning applications. They have the advantage of zero friction
with its associated benecial effects of wear free operation without
any lubrication. In combination with linear motors, they avoid
mechanical contact between the moving slide and the xed part.
There are few cases of application in machine tools [28], and
limited to one linear axis [129]. The disturbances caused by
periodic milling forces are attenuated by linear guides which are
used as both sensors and actuators in [25]. Their main drawbacks
are the lack of damping and the complexity of the control system
which tries to control ve degrees of freedom motion. Fig. 11
presents a two dimensional magnetic drive [41], where the moving
part of the system is free of any contact and wires. When
mechanically well designed, the dynamic behaviour and overall
precision of such magnetic drives can be increased as they allow
the integration of the measuring system in the centre of movement
location.
There are other ultraprecision guiding systems for low range of
movement based on compliant mechanisms. Compliant mechan-
isms base their performance in the elastic properties of materials.
Their main advantages as guiding solution are freedom from
friction, backlash and stickslip effects. Their main limitations are
the short stroke (micrometers or few millimeters) and low load
capacity. Although the basic principles of compliant mechanisms
are well known, design methods and applications have remained
fragmented without a detailed methodology. The advance in
precision machining by wire-EDM allows currently the production
of complex monolithic structures with good accuracy and surface
roughness. Recently, the growing number of applications by
compliant mechanisms (space, scientic instruments, and ultra-
precision machines) has produced a systematic approach to their
design. Fig. 12 shows a 3D positioning device for a micro-EDM
machine with a travel range of 6 6 6 mm, and a positioning
accuracy of 0.1 mm by means of a parallel kinematic mechanism.
3. Mechanical drives
The ball-screw and linear drives are most commonly used as
machine tool feed drives, and their fundamental principles are
Fig. 11. Planar motor magnetically levitated (by Tekniker
1
). Stroke (X and Y):
100 mm, max velocity: 60 m/min, max acceleration: 20 m/s
2
, jerk: 1000 m/s
3
, K
v
:
10 m/mmmin, linear resolution: 60 nm, angular resolution: 0.046 arcsec, linear
accuracy: 0.2 mm, workpiece weight: 120 kg, linear force: 2000 N.
Fig. 12. 3D compliant mechanism for a Micro-EDM Machine (by of Agie
1
).
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 783
highlighted rst followed by a review of recent advances. Machine
tools with long working paths use rack-pinion-drive.
3.1. Rack-pinion-drive
Rack-pinion-drives are recommended for long travel distances.
By adding several racks together, very long feed travel can be
realized. The resulting total stiffness of rack-pinion-drive is always
independent of the length of travel distance. The total stiffness is
dominated by the torsional stiffness of gear and pinion shaft as
well as contact stiffness of rack-pinion-combination. The power
transfer on the pinion is characterized by low revolutions and high
torque. It needs additional gear steps. The whole drive line should
be designed with high torsional stiffness and free of clearance. A
feed transfer with clearance freedom in both movement directions
could be achieved through the separation of the pinion. Fig. 13
shows a feed drive with clearance elimination through pinions
with helical gearing that combine with a rack. The lower pinion is
moved axially through spring force on a spline-shaft shoulder,
which allows both pinions bearing against opposite ank of the
rack and compensating the gearing error.
Another possibility to realize clearance freedom in the rack-
pinion-drive is the application of tension by driving the pinion with
two motors in opposite directions. While the main motor applies
high torque to deliver the motion, the second delivers less torque
to remove the clearance. Fig. 14 shows the torque lines.
3.2. Ball screw drives
The ball-screw is currently the most frequently used in feed
drives of the machine tools. Ball screw drives are characterized by
high efciency (h = 9598%) and thus by low heating, low wear and
high service life without a stickslip effect [73]. The ball-screw
drive consists of a screw supported by thrust bearings at the two
ends, and a nut with recirculating balls (Fig. 15) [57]. The nut is
connected to the table. One end of the ball-screw is either attached
to a rotary motor directly or through gear/belt speed reduction
mechanisms. The nuts are preloaded [42] to avoid backlash by
adjusting the spacer, creating offset between the leads or using
oversized balls as shown in Fig. 16. It is rather difcult to grind the
pitch at uniform intervals, and the pitch errors are transmitted as
position errors unless they are compensated [43,57,78].
The design, calculation and acceptance terms of ball screwdrives
are described in DIN standards [31,32] and [33]. Ball screw drives
with a length of approximately 12 m may be used in machine tools
with long travel strokes. Depending on the application, the screw
diameter and pitch may vary between 16 and 160 mm, and 5 and
40 mm, respectively. The current ball-screw drives can deliver up to
100 m/min travel speed with 2 g acceleration. Optimizations of the
ball screwdrive design [47], the coating of balls to reduce the friction
and wear, deection and nut preload control [107], led to signicant
increases in the speed and accuracy performance of ball screw
drives. The balls roll between the guide slots of the screw and nut
based on either the internal or external recirculation design
principle shown in Fig. 17. External ball recirculation is achieved
by a recirculation tube or channel. An adapted design of the tube
Fig. 14. Clearance-free rack-pinion-system with electric generated preload.
Fig. 15. Structure of ball screw system.
HIWIN.
Fig. 13. Clearance-free rack-pinion-system with split pinion.
Fig. 17. Ball recirculation systems [Hiwin, WZL].
Fig. 16. Ball-screw and nut mechanism.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 784
allows the balls to exit the bearing nut area and to enter it more
tangentially, allowing a more even and smooth ow as well as higher
speeds. A signicant drawback of this design is the slight damage
that may occur to the recirculation tube, which hinders the ball
transport and leads to damage of the nut system. Recirculation
channel based systems may be subdivided into various types, such
as end cap recirculation or front end recirculation. The internal
recirculation system guides the balls via channels at the end of each
thread. While this design has the advantage of requiring less space,
the unfavorable ball entry and exit angles have adverse effects on
even rolling and noise development.
The design of feed motors and the mechanical components is
initially carried out by considering only the rigid body dynamics
and static stiffness of the system. From the application require-
ments, a suitable combination of the design parameters of ball
screw, nut, bearings, motor torque, nominal motor speed, spindle
pitch and transmission ratio are determined.
The total feed force (F
feed
) acting on the drive consists of cutting
and friction loads
F
feed
F
cut
F
friction
(9)
The force is transmitted to the motor as an external disturbance
torque t
feed
t
feed
h
s
2p
1
h
e
r
t
F
feed
(10)
where h
s
, h
e
, r
t
are the pitch length of the screw, mechanical
efciency of the overall system and transmission ratio if the motor
is connected to the screw via pinion-gear system (r
t
> 1),
respectively. The mass of the table/work is transmitted to the
motor as an equivalent rotary inertia (J
e
)
J
e
1
r
2
t
m
l
2
r
2
l
h
s
2p
_ _
2
m
t
_ _
m
m
2
r
2
m
(11)
where m
l
, m
t
, m
m
are the mass of the screw, table and motor,
respectively. r
l
and r
m
are the pitch radius of the screw and motor,
respectively. The motor must deliver torque larger than the total
load torque at speeds and loads to be experienced by the drive,
t > J
e
d
2
u
dt
2
B
du
dt
t
feed
(12)
where u [rad] and B [N m/(rad/s)] are the angular rotation of the
motor shaft and equivalent viscous damping in the drive train,
respectively. The feed force and torque cause axial and torsional
deformation of the ball screw drives. The deformations are
coupled, and distort the positioning accuracy of the table. While
static deformations occur during constant travel speeds under
constant loads, the drive may experience vibrations during
dynamic positioning and under interrupted cutting operations.
The static stiffness is determined primarily by the equivalent
axial stiffness of the ball screwnut contact as outlined in [DIN
69051-6]. The ball screw drive system with bearings and the
intermediate transmission or clutch, has a nite stiffness that
assists in determining the static displacement of the table under
load during high speed positioning of the table. The ball screw is
supported by thrust bearings at two ends. Bearings provide radial
guidance to the screw and absorb the feed forces in the axial
direction. If the bearings at both ends are xed, the equivalent axial
stiffness of the ball screw system is given by (see Fig. 18),
k
eq
1
k
i
k
ii
1
k
M
_ _
1
(13)
where the stiffness terms contributed by the left (k
i
) and right (k
ii
)
bearings are dened as,
k
i
1
k
1
1
k
_ _
1
; k
ii
1
k
2
1
k
_ _
1
(14)
The axial stiffness is reduced when the right bearing is preload
free (k
ii
= 0). As the table position changes, the axial stiffness of the
drive varies which leads to time varying dynamics of ball screw
drives. It must also be noted that the motion delivery between the
ball screw and the nut exhibits a hysteresis type of nonlinear
behavior [22]. In feed drives of large milling machines with high
machining forces, highly preloaded stiff bearings are used and
resulting friction loads are disregarded. However, when high
positioning accuracy is required such as in grinding machines,
low friction is required. The ball screw system also expands due to
thermal loads produced by friction on the guides, bearings and nut.
Thermal simulation of the ball screw system may assist the designer
to predict the changes in the stiffness of the overall system [45].
3.3. Linear direct drives
The table is moved by the magnetic force between the primary
and secondary parts of the linear motor, hence there is no exible
motion transmission element in linear drives. The guide system is
the same as ball screw drives. While the exibilities of screw, nut,
coupling, motor shaft and thrust bearings are avoided in direct
drives, the cutting load and mass are directly transmitted to motor.
Linear drives allow higher acceleration, feed speed and rapid
positioning with high servo bandwidths over ball screw drives
[86]. The acceleration capacity of linear motor is inversely
proportional to total moving mass (m
total
).
Acceleration
F
shear
m
total
(15)
where F
shear
is the magnetic force capacity of the linear motor. A
comparison of accelerations between the ball screw system with
two different pitches (h
1
= 40 mm, h
2
= 20 mm) and the linear
direct drive with two different maximum forces (8000 N, 2000 N)
is shown in Fig. 19. While the linear motor can reach to high
accelerations for light payloads, the ball screw drive can maintain
its acceleration capacity for a larger variation of payload mass due
to reduction of inertia reected to the rotary motor. The heat
generated by the motor needs to be dissipated through cooling
strategies in order to minimize thermal distortion of the drive
system and machine tool. Heavy machine tool table and column
masses carried by high speed, high acceleration linear drives may
excite low frequency modes of machine tool structures [136,137].
The resulting inertial vibrations are picked up by linear encoders
placed on the table, which may destabilize the controller and
produce poor surface nish.
Fig. 18. Axial stiffness of the ball screw with single and double sided thrust bearings.
WZL.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 785
4. Structural dynamic model of drives
Ball screw drives exhibit torsional exibilities at the motor
shaftscrew coupling, screw itself, and nut as shown in Fig. 20. The
axial displacement of the screw is coupled with its torsional
exibility, and the screw may experience lateral exibilities which
apply tension and compression loads on the tableguideway
interface. The structural vibrations caused by the ball screw
assembly occur typically above the bandwidth frequency of the
servo drive, i.e. above 100 Hz. However they affect the surface
nish quality and precision positioning accuracy during machin-
ing, hence they need to be avoided. The linear cutting force (F
1
) and
table mass (m
t
) are transmitted to motor as a reduced torque (t),
see Fig. 20. The bandwidth and speed are increased by using two
parallel ball screw drives in most recent, high speed machine tools.
The mechanical drive system G
m
(s) is represented by its rigid body
motion when the structural dynamic exibilities are neglected.
Ball screw G
m
s
u
2
t
2
1
J
e
s b
K
e
s
Linear drive G
m
s
x
2
F
2
1
m
e
s b
K
e
s
(16)
where K is the encoder gain, b is the viscous damping, and J
e
, m
e
are
the equivalent inertia and mass as seen by the motor, respectively.
However, it may be important to damp the structural vibrations
of the machine which are excited by the cutting and inertial forces
during high speed motions. The general mechanical transfer
function between the loads and table/motor position can be
represented as
x
2
s
x
1
s
_ _
G
11
s G
12
s
G
21
s G
22
s
_ _
F
2
s
F
1
s
_ _
(17)
where x
1
, x
2
are the linear positions of the table and motor; F
1
, F
2
are the forces acting on the table and motor, respectively. In the
case of ball screw drives, the angular position (u) and torque (t) are
given by,
u
x
r
g
; t r
g
F (18)
Transfer functions G
pq
(s) determine the relationship between
the forces and positions of the table and motor due to the exibility
of the mechanical drives, and they replace the rigid body based
transfer functions G
m
(s). A simple structural dynamic model of the
ball screw drive system can be approximated by the reected
inertias at the motor (J
m
) and lead screw J
l
connected by a torsional
spring (k
t
) and damping c
t
elements as shown in Fig. 20. The
connection can be considered at the motorscrew coupling or
screwnut coupling junctions for simplicity. By neglecting viscous
friction, the structural dynamics of the ball-screw system can be
expressed by:
s
2
J
m
0
0 J
l
_ _
s
c
t
c
t
c
t
c
t
_ _
k
t
k
t
k
t
k
t
_ _ _ _
u
m
u
l
_ _
t
m
t
l
_ _
(19)
which leads to the following transfer function of the system:
Gs
J
l
s
2
c
t
s k
t
c
t
s k
t
c
t
s k
t
J
m
s
2
c
t
s k
t
_ _
s
2
J
l
J
m
s
2
c
t
J
l
J
m
s k
t
J
l
J
m
(20)
The ball screw system has the torsional natural frequency:
v
0
k
t
J
l
J
m
=J
l
J
m
(21)
The inertias (J
m
, J
l
) can be replaced by equivalent masses (m
m
,
m
t
) at the motor and table for the linear drives, respectively. The
natural frequency will remain the same. Methods of estimating
model parameters can be found in [23].
The system may have more natural modes depending on the
structural conguration of the drive train. The parameters are
experimentally identied by exerting impulse load on the table
(F
1
) and measuring the displacement x
1
with vibration sensor or
linear encoder, and u
2
with the angular encoder of the motor.
Motor torque (t
2
) can be used as an excitation load by injecting a
white noise or harmonic signal to the current amplier of the
motor. The modal parameters are estimated through least squares
based identication algorithms [77].
Research efforts have been primarily concentrated on the
dynamic modeling of the screw, nut, bearings, coupling, motor
shaft, table and guide [67,128]. Generally, computer models of ball
screw drives are obtained using hybrid nite element methods
(FEM). Relatively rigid components of the drive are modeled as
lumped masses connected by springs while components with
distributed mass and stiffness, like the ball screw, are modeled
using nite element structures. The method is able to capture the
essential dynamics of the drive while maintaining computationally
advantageous low-order models, which is an important require-
ment for virtual prototyping. Varanasi et al. [121] and Whalley
et al. [130] modeled the ball screw using EulerBernoulli beam
formulations, which capture the axial and torsional dynamics of
the ball screw drive. However, neither model considers the lateral
Fig. 20. Ball-screw drive mechanism.
UBC.
Fig. 19. Acceleration capacity comparison between linear and ball screw drives.
WZL.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 786
deformations of the ball screw, which could affect the positioning
accuracy and performance of the machine tool. Similarly, the
researchers in [57] modeled the ball screw using nite-element
beam formulations. With the exception of Zaeh et al.s model [131],
the past models include only the axial and torsional dynamics of
the ball screw while neglecting its lateral dynamics. Zaeh
considered the lateral deformations of the ball screw but the
model used for the screwnut interface stiffness matrix fails to
capture the coupling between the lateral, axial and torsional
dynamics of ball screw drives. Okwudire et al. [77] modeled the
ball screw using Timoshenko beam elements, while representing
relatively more rigid components of the drive (nut, table and
guideway) as lumped masses/inertias connected by multi-direc-
tional springs at exible joint interfaces. The ball-nut stiffness is
modeled by considering the Hertzian contact of each re-circulating
ball and projecting their stiffness at the nut node of the FE model.
Stiffness values of the guideway and thrust bearings are obtained
from the manufacturers catalogs. Screwnut interface model is
developed to capture the dynamic interaction between the axial,
torsional and lateral dynamics of ball screw drives. The simulated
and experimentally measured natural frequencies and mode
shapes are shown in Fig. 21. The most dominant mode (218 Hz)
that affects the axial positioning accuracy of the table is due to
coupled torsional, axial and lateral deformation of the ball screw
which varies as the table moves from one extreme towards the
motor side. The second and third modes are due to twisting of the
screw and yaw motion of the table, and they do not affect the table
position signicantly. The discrepancy between the simulation and
measurements are mainly due to analytically estimated contact
stiffness of nut and assumed bearing stiffness values. However,
virtual design and analysis model of the ball-screw drive is useful
during prototyping of the machine tool and controller [106].
5. Electrical drives
5.1. Introduction
A variety of electrical motors are used in machine tools as
summarized in Fig. 22. Usually synchronous permanent magnet ac
servo drives are used in feed drives, while asynchronous motors
are typically employed in spindles. At rst, asynchronous drives
became widely accepted for spindle drives because of their
overload capacity and synchronous drives for feed drives because
of their degree of efciency and the related lower heating. The
synchronous motors became dominant for feed drives in 1970s.
Ball-nut screw drives for large traveling length and linear direct
drives were developed in 1980s. The latter made the mechanical
transformation from rotary to translatory motion redundant. New
electrical drive concepts like transversal or axial ux [49] drives
have not been commonly used yet. Typical feed drives have a rated
power up to 20 kW and a speed range up to 8000 rpm, whereas
spindle drives can go up to 100 kW power and 2060,000 rpm
angular speed.
5.2. Permanent magnet AC synchronous and AC induction motors
Permanent magnet synchronous motors (PMSM) are most
widely used in machine tool drives. Similar to brushless DC motors,
the PMSM has a permanent magnet rotor and windings on the
stator. The PMSM is driven with sinusoidal current generated by
Field Oriented Control (FOC). High-torque motors can deliver more
30,000 N m torque [122]. Some manufacturers offer motors with
higher inertia than standard designs in order to have more
favorable motor/load inertia ratios for achieving a better dynamic
performance in feed drives with varying inertia. The further
developments of rotary synchronous motors in recent years have
Fig. 21. Mode shapes and frequency variation of ball screw drive as table travels
360 mm. M: Measured, S: Simulated with FE model.
UBC [77].
Fig. 22. Common feed drive motors with applications and characteristics.
ISW.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 787
lead to power improvement and reduction of the cogging effects by
rare earth magnets embedded in the rotor. In well-designed
motors the cogging effect is less than 1% of the operational torque.
The AC induction motor consists of a cage-like rotor and a stator
with three brushless windings. Control methods of AC Induction
motor can be divided into four major groups, namely the constant
Voltage/frequency, sensor-less vector, eld-oriented and the direct
control method [10,113].
5.3. Linear drives
A linear motor [83] can be viewed as a rotary motor that has been
cut to its axis, rolled out and stretched in length. The direct drives
powered by linear motors have high mechanical stiffness, low
inertia and zero backlash. The primary part of the linear motor is
similar to the stator of a PMSM. The secondary part consists of
permanent magnets. There is a high attraction force (about three
times more than the driving force) between primary (with iron) and
secondary part (with magnets) which has to be supported by the
guide ways. This leads to the position dependent disturbance or
cogging forces. Force ripple occurs due to an electromagnetic
disturbance resulting in periodic variations of the motors constant,
which depend on the instantaneous motor force. The iron cores and
thus the attraction forces can be omitted when the secondary part is
designed as a comb with magnets on both sides. The ironless linear
direct drives have much lower electrical time constants and thus
high bandwidth. For applications where permanent magnets in the
secondary part are prohibited or too costly for large travel lengths,
there are new linear direct drives with permanent magnet-less
secondary part (e.g. Siemens 1FN6 drive series). The basic principle is
a synchronous linear motor where the magnets are integrated into
the primary part with its windings for each phase.
5.4. Hybrid feed drive concepts
A combination of two drives is used in some applications, i.e. to
prevent the tilting of gantry bridges. The resulting force vector can
be placed at the center of gravity with double drives. Feeding the
same reference position commands simultaneously may distort
the mechanically coupled drives. It is more common to use a
master-slave setup, which compares the torque values and feed
back to each drive. Since linear motors need permanent magnets
for the entire traverse, their application to machines with long
travel lengths can be complex and costly. Instead, the concept of
redundant axes may be used for long strokes as shown Fig. 23. A
standard ball-screw drive is used for the entire traverse, and a short
linear direct drive is used to enhance the dynamic performance.
The motions of both drives are kinematically combined and
controlled to position the tool at the required location.
Since linear drives are capable of delivering more than 10 g
acceleration, the excitation of the machine bed caused by reaction
forces of the table motions can be quite signicant. One possibility
of compensating these forces is to accelerate a second slide on the
same guide but in the opposite direction [46,73]. Another solution
would be to damp the peak forces. Instead of stiff mounting of the
feed drive to the frame, it can be stacked on a second guide to avoid
the transmission of acceleration forces to the machine frame. The
second guide moves in the opposite direction to the feed drive, and
absorbs the acceleration load [13,79]. The dynamic stroke of the
second table is limited and damped by a spring-damper-system,
and its counter-motion is measured and included in the control
loop. This concept also works for ball screw and rack and pinion
feed drives [61].
6. Sensors
Feed drives use position, velocity, acceleration and load sensors
to improve their positioning accuracy and response bandwidth.
The most widely used sensors are briey reviewed here along with
their associated performances.
6.1. Position measurement
The absolute position of the feed drive needs to be measured for
precision positioning of the tool on the workpiece. The table
position is directly measured using optical encoders in machine
tools. Laser interferometer methods are used in precision machines
designed for manufacturing of optics, electronic circuits or in
machines. Rotary encoders or syncro-resolvers are measuring to
the rotary servo motor shafts for indirect measurement of the table
position from the angular position of the motor shafts.
Optical encoders are based on the principle of transmitting or
reecting light by a glass or metal grating. A light source and photo
detector array are used to sense the position of moving encoder disk
or scale containing equally spaced reective grating. The absolute
encoders have binary (gray or pseudo-random) coded absolute
positions. Incremental encoders have an additional reference mark
which allows the controller to track the absolute position by
counting the equally spaced marks relatively to the reference mark.
In case of a rotary or linear encoders with equivalent pitch spaces (i.e.
around 10 mm), the scanning unit moving in the direction of motion
illuminates the scale at the measurement point. Motion between the
scale and the scanning unit is evaluated through a scanning grid by
means of photo elements. The received modulated signal has a
nearly sinusoidal shape per scanned increment. The alignment of
two receiving elements is used to detect the direction of the motion.
The resolution of the encoders is improved by a factor of 1000 or
more by interpolating the sinusoidal pattern of the encoder readings.
Interferometric measurement principle is used for encoders with
higher resolution.
Recent developments improve the non-ideal sinusoidal signal
output of incremental sensors by reducing the harmonics of
error frequencies, hence a position-signal is taken with at very
high sampling rates, which allows to process harmonics of the
error-frequencies.
6.2. Speed measurement
The velocity of the feed drive is needed by the servo controller
for tracking and damping of the table motion. Rotary tacho-
generators, which provide DC voltage proportional to the angular
speed, are integrated to the motor shaft. Its operation principle is
equivalent to that of a DC machine with a stationary exciting eld
and rotary sensor winding unit. The brush contacts of DC
tachogenerators limit their service life. AC tachogenerators have
a permanent-eld rotor with a stationary winding system, which
gives longer service life. AC tachogenerators produce trapezoidal
AC voltage with an amplitude and frequency proportional to the
speed. However, both tachogenerators have low signal to noise
ratio at low speeds, which negatively affect the velocity control of
feed drives. Recently, speed sensors based on eddy-current
principle have been investigated by researchers [59,115].
It is most common to estimate the velocity by digitally
differentiating the position measurements obtained from the Fig. 23. A concept for redundant axes test bench.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 788
encoders [16]. The accuracy of the velocity estimation depends on
the resolution of encoders, quantization [12,15], speeds and
harmonic error-frequencies [59]. When the change in the velocity
is too small in a very short digital integration interval, i.e. control
interval of the servo, the velocity prediction becomes highly
inaccurate and noisy. Low pass and FIR lters are employed to
smooth the velocity estimation, which adds undesired phase delay
to the drive controller. There are research attempts in treating the
system at low and high frequency zones separately in order to
improve estimation of the velocity from discrete position samples.
6.3. Acceleration measurement
Acceleration feedback is used in control laws for damping the
structural dynamics and inspecting the actual trajectory of the feed
drive. The acceleration can be measured directly, or from the
second digital derivative of the position measurements
[44,66,135]. Standard, piezoceramic-based accelerometers provide
absolute acceleration of the drive; hence both rigid body and
structural vibrations are mixed in the signal. Pritschow et al. [85]
introduced electro-magnetic based Ferraris sensor which mea-
sures the relative acceleration between the moving drive (i.e. table)
and stationary base (i.e. guide), which is advantageous for active
damping of structural vibrations. The bandwidth of the system
with a stationary exciting eld is up to 1500 Hz.
6.4. Current measurement
The current of a servo motor is used in the rst loop of a cascade
controller, where the given value for the force or moment in form
of the current is controlled via a PI controller. The current is also
used to compensate friction and cutting force disturbances. There
have been several research attempts to in predict the cutting forces
from the motor current as well [4,26]. The current is measured
from the shunt resistors, inductive transformers or utilizing
magnetic effects. Shunt resistors are placed in series to the load.
A voltage drop across the shunt-resistor is proportional to the
current through the resistor. This method has its key benet in a
wide frequency-range. Inductive transformers operate by trans-
forming the current of a motor phase through magnetic coupling,
leading to AC signals proportional to the current. Hall-sensors are
most commonly used in measuring the current-induced magnetic
eld, which has a drawback of limited bandwidth. Recent
developments replace the Hall-sensor by GMR (giant magnetore-
sistance) semiconductor-elements, which have magneto-resistive
detecting elements with about 1 MHz bandwidth.
7. Control of feed drives
The general architecture of most feed drive controllers is
illustrated in Fig. 24. The mechanical structure of the drive is
represented by G
m
which is either considered to be rigid or exible.
Amplier (G
A
) and electrical winding of the motor (G
E
) have fast
dynamics, and are usually modeled as gains. However, they may be
considered as rst order lags or with higher order dynamics in
some advanced controllers. The digital control law tries to
minimize the position error e(k) at each control interval (k):
ek x
r
k xk (22)
where x
r
(k) and x(k) are the reference and actual position of the
table, respectively.
There have been a signicant number of advanced controllers
reported in the literature. In addition to position, the control laws
may use velocity ( x
r
k; xk), acceleration ( x
r
k; xk) and jerk
(x
_
r
k). Feedforward (G
FF
) and feedback (G
FB
) compensators are
used to minimize the effects of friction, cutting force disturbance,
and unmodeled or varying dynamics of the drive.
The cascaded control structure shown in Fig. 25 is most
commonly used in industrial feed drives; hence it is used as a
reference controller against new algorithms published in the
literature. The position commands are transmitted from the CNC to
the drive via eld bus. The block diagram (Fig. 25) represents a
direct drive system with rigid mass (m
g
), and the controller has a
current loop inside, surrounded by velocity and position control
loops. Cascaded controllers use a proportional gain (K
v
) on position
error (e), proportional gain (K
p
) on the velocity error ( x
r
x), and
integral action with time constant (T
n
) to minimize the steady state
error caused by the disturbance (F
d
) and lag caused by the transfer
function of the system. The inertia and viscous damping forces are
compensated by the feedforward and feedback terms, respectively.
The current controller is in general designed as a PI controller in
series with the power converter. It usually has a bandwidth of
1 kHz with a PWM (Pulse Width Modulation) converter of 4
20 kHz. The maximum bandwidth of the velocity loop is less than
10% of the current loop (i.e. 100 Hz) when the velocity gain is
tuned around K
p
= 600 s
1
for linear drives [89,93]. The mass is
compensated by a gain factor of m
g
because it affects the
proportional velocity control gain K
p
. The position control loop
usually has 30% less bandwidth than the velocity loop, i.e. with
K
v
= 200 s
1
the bandwidth is about 30 Hz. However, the
bandwidth can be increased by three fold with modern control
and compensation methods [87]. If the external disturbance
sources are well known, such as friction and pitch error induced
ball screw loads, they can be compensated using feedforward and
set point variations in the controller [57,76,82]. To nshoff et al.
presented the challenges and possible solutions in compensating
friction and cogging forces in linear drives [112].
7.1. Rigid body controllers
The structural exibility of the machine tool reected to the
drive is not explicitly considered in rigid body based controllers.
The objective is to widen the positioning bandwidth of the drive as
much as possible. However, to avoid interfering with the rst
vibration mode, these types of designs are frequently implemented
alongside notch or low-pass lters that avoid the control
(actuation) signal from exciting the machine structure. A low-
pass lter provides good high frequency gain attenuation, and
therefore a robust solution against unknown or varying high
frequency structural dynamics. If the frequency and damping of
the mode(s) are well-known and do not change over time (or can
be accurately predicted), notch lters provide a more favorable
means of attenuating the amplitude due to mechanical resonances,
without inducing as much phase lag as low-pass lters. This
provides better stability margins and the ability to achieve higher
control bandwidths. When implementing the control laws, the Fig. 24. General structure of a feed drive control system.
Fig. 25. Equivalent block diagram of the cascade controller for linear drives.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 789
saturation limits of the feed drive motor and amplier [125] must
be avoided as shown in Fig. 25.
Although cascaded control system parameters are tuned to
accomplish rigid drive control, several modern algorithms are also
proposed as alternative methods. Tomizuka et al. [111] developed
a zero phase error-tracking controller (ZPETC) by canceling the
stable dynamics of the servo drive in a feedforward fashion. The
bandwidth of the overall system, hence the tracking accuracy of
the drives increases dramatically with ZPETC provided that the
drive model is accurate and does not vary with time. Weck and Ye
[124] noted that ZPETC generates feedforward commands with
rich frequency content, leading to distorted motions at high speed
tool paths with sharp corners. They removed the high frequency
content by pre-ltering the position commands. Dumur et al.
proposed several predictive control methods [34] which generate
control commands by predicting the behavior of the drive
performance few sampling periods ahead. The contouring accuracy
in high speed machining depends on the tracking bandwidth and
the disturbance rejection of the axis control system, as well as the
smoothness of the reference trajectory. Van Brussel et al. [117]
addressed the disturbance compensation problem by adding an
observer to the control law.
The main drawback of the classical controller design techniques
is the sensitivity to modeling errors, as addressed by Van den
Braembussche [120]. Especially when feedforward control is used,
it is vital to establish a robust feedback loop that will mitigate the
inuence of changes in the inertia, friction as well as other
dynamics which may be dependent on the drives position [132].
Otherwise, as the open or closed loop dynamics change, the
feedforward controller would try to cancel the incorrect model
[87]. In addition, process forces, which in most cases are difcult to
predict and compensate in real-time, have to be rejected by the
feedback loop. Hence, recent research articles focused on the
design of controllers which are capable of coping with changing
friction and external disturbances, and uncertainties in the drive
inertia. Jamaluddin et al. compared the performances of cascaded
controllers against sliding mode controller with disturbance
compensation [52,53]. They showed improved dynamic stiffness
and better tracking accuracy in comparison to cascaded controller.
Choi et al. considered the effect of cutting force disturbances on the
linear drive stiffness [20]. Altintas et al. [6] presented a sliding
mode controller with disturbance compensation (Fig. 26). The
controller was implemented on ball-screw and linear drives in
current control mode, and demonstrated to have as high
bandwidth as ZPETC in command following while being robust
against inertia changes up to 30%. The control command needs to
use reference position, velocity, acceleration commands, and
measured position and velocity of the drive.
7.2. Control algorithms with exible drive structure
The relative structural vibrations between the tool and work-
piece degrade the surface quality and tolerance of the parts.
Vibrations accelerate the wear of the power train components, and
overload the drives and machine tool structure. Furthermore, when
the vibrations are felt by the servo system, the controller can become
unstable which leads to unsafe operation of the machine tool.
The source of vibrations reected on the feed drives includes
cutting forces, unbalanced spindle loads, friction and backlash in
the drives as listed in Fig. 27. The discontinuities in the motion
trajectory algorithms, which generate position, velocity, accelera-
tion and jerk command proles for each drive, also excite the
natural frequencies of the machine tool structure. More impor-
tantly, interaction between the drives mechanical response, the
current loop, and the digital servo controller can be one of the main
causes of vibrations. If the controller is not designed properly, this
can lead to stability and robustness problems.
The vibrations of mechanical structures can be rst reduced at
the design stage by changing the topological structure, using stiffer
components connected with materials having higher damping
ratios. However, stiffer designs usually lead to larger moving
masses, which reduce the high speed positioning performance of
the drives. If the dominant natural modes of the structure cannot
be avoided during the design stage, damping elements, which are
tuned to damp specic natural frequencies, can be used between
external force and vibrating structure [18]. Alternatively, the
effects of oscillating forces can be countered with active dampers
actuated by electromagnetic, piezoelectric or hydraulic actuators
[11,50,72,74,133]. While these measures typically lead to good
results, it can be difcult and more costly to incorporate them into
the machine tool design. Fundamental real time algorithms, which
are used to avoid and damp vibrations, are discussed in the
following sections.
7.2.1. Trajectory generation
The motion commands to feed drives constitute a major source
for excitation of the machine tool vibrations. If the reference
trajectory motion, e.g. position, velocity, acceleration and jerk
commands at discrete control intervals, have discontinuities, they
will have wide frequency content. Drives with high speeds and
acceleration create high reactive forces between the moving mass
and the stationary bodies of the machine. Inertial forces for high
frequency content excite the machine tool structure causing
undesired transient vibrations, which are suppressed by smooth
trajectory motion proles, passive dampers, or active control
methods are used.
The present CNC systems use jerk limited-trapezoidal accel-
eration motion proles (i.e displacement commands are at least a
cubic function of time) while avoiding the saturation of drive
motors [3]. When three trajectory generation algorithms are
compared in Fig. 28, cubic acceleration with a second order
continuous jerk has the least amount of acceleration amplitude at
high frequencies. That is translated as having smaller inertial
forces at the high frequencies, hence the excitation of structural Fig. 26. Sliding mode controller for a rigid feed drive [6].
Fig. 27. The source of feed drive vibrations and their compensation method.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 790
modes at those frequencies are reduced. In addition, the
frequencies which excite the natural modes can be removed from
the motion commands by pre-ltering before injecting them to the
servo system. Such pre-ltering techniques are covered broadly in
the literature, including low pass and notch lters [31,32,33],
moving average lters [48,54,95,114], and input-shaping using
impulse sequences and form functions [36,56,101]. The notch lter
prevents the excitation of the problematic modes by the trajectory
command signals, but cannot stop the table disturbance forces
from exciting them. Furthermore, the bandwidth of the drive
becomes less than the structural resonance of the drive. This
conservative approach reduces the productivity of the machines,
and they are especially not suitable for high speed machines with
light structures. The lters have to be applied with care as they
may alter the path geometry or cause overshoot. As these
trajectory proles are smooth, they introduce relatively low
excitation compared to acceleration limited motion proles.
Nevertheless, low acceleration and jerk settings may be required
for machines with prominent resonances. More advanced trajec-
tory generation techniques include spline based path and velocity
proles, and smoothing of jerk proles [19,51,65,71]. One
interesting idea that has been proposed by ISW researchers is
the optimization of axis jerk durations to avoid excitation of
frequency content around certain resonances [30]. A sample
application of smooth trajectory generation avoids excitation of
rams natural frequency as shown in Fig. 29. In addition to smooth
trajectory generation at CNC, passive damping systems can be used
to absorb the energy created by the inertial vibrations. Denkena
et al. presented a novel, energy efcient guide system which
decouples the jerk from the stationary base through a tuned mass,
spring and damper system [27].
7.2.2. Active damping of drive vibrations
Independent of the individual techniques, only vibration modes
that are controllable by the actuators and that are observable
through the sensors available on the machine can be damped.
Various closed loop control techniques have been demonstrated
to damp the vibrations with the feed drive motor by using
position, velocity, acceleration, current and force feedback
[14,17,21,75,102]. The algorithms are designed assuming constant
or time varying and position dependent structural dynamics of the
drive.
An easy to apply damping method is illustrated here using
cascade control structure applied to a ball-screw drive. The
velocity loop is modied by considering the structural vibrations of
the drives and adding an acceleration feedback at the dominant
natural frequency as shown in Fig. 30. The current control loop is
neglected, and the velocity is controlled by a PI controller. The
transfer function between the motor torque and velocity measured
at the motor shaft can be obtained from Eqs. (17)(20):
G
mot
s
v
mot
t
2
sG
11
s
r
g
J
l
s
2
c
t
s k
t
sJ
l
J
m
s
2
c
t
J
l
J
m
s k
t
J
l
J
m
(23)
with natural frequency v
0
as given in Eq. (19). The less dominant
mode v
01
and the bandpass lter is neglected for a single mode
damping. The transfer function between the velocities at the table
and motor shaft is given as:
G
mec
s
v
v
mot
sG
12
s
sG
11
s
c
t
s k
t
r
g
J
l
s
2
c
t
s k
t
(24)
Fig. 29. Reduction of transient vibrations through smooth trajectory generation
(ISW).
Fig. 30. Velocity loop of ball screw drives with multi-resonance modes and active
damping [64,90].
Fig. 28. Frequency content of trajectory generation algorithms with innite,
constant and continuous jerk proles.
UBC MAL.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 791
The indirect velocity loop uses the motors angular velocity
measured from a tachogenerator or encoder mounted on the motor
shaft (v
mot
), whereas the tables velocity (v
ma
) or acceleration (sv
ma
)
is measured in the direct velocity loop. When the tables position is
measured from linear encoders, its second derivative (accelera-
tion) is used in the feedback but normalized at the vibration mode
to be damped (v
0
). Alternatively a Ferraris acceleration sensor can
be used directly [85]. A sample application to a milling machine
with a dominant vibration mode at v
0
and a less exible mode at
v
01
is illustrated (see Fig. 30). First, only the mode (v
0
) is damped.
The feedback s/v
0
raises the phase by 908. The closed loop
bandwidth is practically not affected due to the open feedback
attenuation below resonance frequency (v
0
), see Fig. 31, where
two coupled rotational and translational modes of a turning
machine is shown. The magnitude peak at the resonance (v
0
) is
attenuated, and the effective phase of the open loop becomes less
than 1808. The position control loop gain (K
v
) can be increased from
15 s
1
to 150 s
1
with active damping, hence improving the
bandwidth close to the resonance frequency v
0
[90].
When additional natural frequency (i.e. v
01
) exists ahead of
dominant mode (v
0
), the acceleration feedback (s/v
0
) may not be
sufcient to widen the bandwidth. In such cases, a bandpass lter
tuned to (v
01
) is used in the direct velocity loop. While (s/v
0
)
damps mode (v
0
), the bandpass lter attenuates all amplitudes
around v
01
which is also damped [64], see Fig. 32. The effect of this
active damping strategy can be seen in the closed loop response of
the position loop (Fig. 33), where the position gain K
v
can be
doubled from 50 to 100 [s
1
]. The effect of active damping strategy
on the transient vibrations measured at the table is shown in
Fig. 34.
There are various versions of active damping with acceleration
feedback. When the active damping strategy is incorporated to the
virtual simulation of entire machine tool, it is possible to optimize
the masses and design conguration of feed drives [107].
7.2.3. Advances in active damping of feed drive vibrations
Considering the drawbacks of the classical approaches, recent
efforts are directed towards improving the bandwidth of the
drives using robust controllers with disturbance compensation.
Zatarainet al. attachedan accelerometer at the tool center point to
estimate the relative position between the tool and slide to
compensate the deformations [134]. Dumur et al. used an
adaptive generalized predictive control to compensate the
structural modes in the presence of variations of drive inertia
[34] but it does not take into account the effects of external
disturbance forces. The general sliding mode controller (SMC) was
rst introduced by Utkin [116] which required switching around
the sliding surface resulting in a discontinuous control law, and
considered to be more robust than linear controllers. To alleviate
this problem, Slotine and Li [103] proposed an adaptive sliding
mode controller (ASMC) to estimate and cancel various uncer-
tainties (including disturbance forces) which do not necessarily
Fig. 31. Frequency response of direct velocity loop with active damping of mode
v
0
= 65 Hz (ISW).
Fig. 32. Velocity control loops with a multi-resonances at v
01
and v
0
with a
bandpass lter at v
01
and phase lead at v
0
.
Fig. 33. Closed loop response of the position control with active damping. Bandpass
lter is set at v
01
.
Fig. 34. Experimentally measured effect of active damping network.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 792
vanish at the equilibrium point. A mode-compensating version of
ASMC was proposed by Kamalzadeh and Erkorkmaz [39,57] for
active control of ball-screw drives. Okwudire et al. [78] applied
discrete time sliding mode controller (DADSC) to compensate
vibration modes of a ball screwdriven table and linear drives [7].
Symens et al. proposed a gain scheduling and robust algorithms to
control feed drives which have position dependent dynamics
[108,119].
The approaches based on closed loop control presented so far
have the advantage that they can react to any vibration
independent of its cause. At the same time, they have the
disadvantage that they only react once a vibration is present,
which will be too late for some cases. It is desirable to use the
combination of feedforward blocks to compensate friction, back-
lash and measurable disturbance forces, active damping of
vibration modes using various control laws, and notch lters to
eliminate frequencies from the control signal [68,76,91,110]. An
adaptive sliding mode controller being used in conjunction with
ripple and friction compensation is shown in Fig. 35 [58].
7.2.4. Coupled exible multi-body and feed drive simulation of
machine tools
Machine tool feed drives are currently designed in virtual
environment using Computer Aided Control Engineering Program
[92]. MATLAB/Simulink
TM
and ADAMS
TM
software systems allow
integration of control, kinematics and structural models of the
machine tool and feed drives [60,118,126,127]. Virtual analysis of
interaction between the feed drive control and machine tool is
illustrated in Fig. 36. To simulate the machine tool a exible multi-
body simulation model is used that describes its structural and
kinematic behavior in one position. Large displacements in the
machine axes can be described by a movable exible multi-body
simulation model [5].
The control models of all drives are created either in MATLAB/
Simulink environment or the real CNC is connected to the
structural dynamic model of the drives as in Fig. 36 [88]. The
torque/force created by inertial motion as well as external friction/
cutting loads are simulated and steady state results are applied at
control loop intervals onto the machine tool structure at feed drive
and cutting points [37,38]. The resulting structural vibrations are
projected to the drive positions and fed back to the controller. The
time varying and machine tool position dependent dynamics can
be simulated along the tool path, and the inuence of trajectory
generation and controller parameters on the overall structural
dynamics of the machine tool can be analyzed [92]. The system
also allows tuning of control parameters for damping the
structural dynamic modes.
7.2.5. Hardware in the loop simulation
Instead of simulating the entire system in virtual environment
(Fig. 36), the virtual dynamic model of the machine is connected to
the physical, actual CNC for real time simulation as shown in
Fig. 37. The real time simulation of virtual machine tools is called
Hardware in the Loop Simulation. The mathematical model of
the machine tool, which consists of structural model of bed,
column, spindle, feed drives as well as the servo motors and
sensors are assembled to achieve virtual model of the machine
tool. The structure can either be represented by rigid body or Finite
Element models [88,94]. The virtual model of the machine tool is
connected to the real CNC, hence the control commands are sent to
the mathematical model in real time. The mathematical model
predicts the vibrations and rigid body motion of the machine
reected at each drive. The machine positions are passed through
the models of the sensors which are sampled by the real CNC at
control intervals as outlined in Fig. 38. The system also allows
visualisation of the machine tool motion and material removal in
real time [88]. The method allows realistic testing of feed drives
during the design stage, well ahead of costly physical trial of
prototypes. A more detailed description of hardware in the loop
system used in practice under the name virtuous can be found in
[123].
Fig. 35. Adaptive sliding mode control with notch ltering and ripple compensation
[58].
Fig. 36. Coupled exible multi-body simulation of machine tools.
WZL.
Fig. 37. Design of a hardware in the loop simulator.
ISW.
Fig. 38. Hardware in the Loop architecture.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 793
8. Conclusion
Recent advances in electrical motors, power electronics,
computers, actuators and sensors technology have been pushing
the speed limits of spindle and feed drives upward continuously.
There are already spindles which reach over 10
5
rpm which require
feed drives travelling over 5 10
4
mm/min with 10 g acceleration
rates. Micro-machining applications use air spindles up to
5 10
5
rpm, which require fast and precision feed drives. There
are research challenges in meeting such high speed accurate
positioning demands from industry.
The reduction of friction in the drives and their real time
compensation by the CNC systems, avoidance of vibrations
through adaptronic devices, advanced trajectory generation and
control algorithms; energy efcient design of linear drives,
compensation of backlash, thermal and geometric deformation
of drives require more research in order to push the speed and
accuracy boundaries of present systems further.
The physical prototypes and trials are costly and slow down the
design and development of new machine tool concepts. Compu-
tationally efcient, accurate and reliable mathematical models of
machine tools with feed drives need to be developed in order to
test design concepts in the virtual environment realistically. The
mathematical assembly of models representing each machine tool
component and drive in a modular but computationally affordable
way still requires major research efforts, especially in multi-body
dynamic modelling of machine tool and feed drive structures.
Acknowledgements
The authors thank all CIRP members who sent their contribu-
tions to this article. Special thanks are extended to Dipl. Ing.
Alexander Haer of ISW Stuttgart; Dipl. Ing. Turker Yagmur of
WZL Aachen; Prof. K. Erkorkmaz of University of Waterloo for
their contributions to various sections in the paper. Prof. H. Van
Brussel of KLU Leuven and Dr. H.C. Mo hring of IFW Hannover
kindly reviewed the draft paper and suggested improvements.
References
[1] Aguirre G, Al-Bender F, Van Brussel H (2010) A Multiphysics Model for
Optimizing the Design of Active Aerostatic Thrust Bearings. Precision Engi-
neering 34(3):507515.
[2] Al-Bender F (2009) On the Modelling of the Dynamic Characteristics of
Aerostatic Bearings Films: From Stability Analysis to Active Compensation.
Precision Engineering 33(2):117126.
[3] Altintas Y (2000) Manufacturing Automation. Cambridge University Press,
Cambridge, UK.
[4] Altintas Y (1992) Prediction of Cutting Forces and Tool Breakage in Milling
from Feed Drive Current Measurements. Transactions of ASME Journal of
Engineering for Industry 114:386392.
[5] Altintas Y, Brecher C, Weck M, Witt S (2005) Virtual Machine Tool. Annals of
the CIRP 54(2):651674.
[6] Altintas Y, Erkorkmaz K, Zhu W-H (2000) Sliding Mode Controller Design for
High Speed Feed Drives. Annals of CIRP 49(1):265270.
[7] Altintas Y, Okwudire C (2009) Dynamic Stiffness Enhancement of Direct-
Driven Machine Tools using Sliding Mode Control with True Disturbance
Recovery. Annals of CIRP 58(1):335338.
[8] Aoyama T, Koizumi K, Kakinuma Y, Kobayashi Y (2009) Numerical and
Experimental Analysis of Transient State Micro-bounce of Aerostatic Guide-
ways Caused by Small Pores. Annals of the CIRP 58(1):367370.
[9] Aoyama T, Kakinuma Y, Kobayashi Y (2006) Numerical and Experimental
Analysis for Small Vibration of Aerostatic Guideways. Annals of the CIRP
55(1):419422.
[10] Arias A, Romeral L, Aldabas E, Jayne M (2005) Stator Flux Optimised Direct
Torque Control System for Induction Motors. Electric Power Systems Research
73(3):257265.
[11] Ast A, Braun S, Eberhard P, Heisel U (2007) Adaptronic Vibration Damping for
Machine Tools. Annals of the CIRP 56(1):379382.
[12] Baehr A, Mutschler P (2006) Comparisonof Speed AcquisitionMethods based on
Sinusoidal Encoder Signals. Journal of Electrical Engineering Edition 4 15824594.
[13] Berkemer J, Altenburger R, Koch T, Lehner W-D (2004) Effektive Nutzung des
Leistungspotenzials von Direktantrieben. wt Werkstattstechnik, Springer,
Berlin. online Jahrgang 94 H. 5.
[14] Brecher C, Schulz A, Weck M (2005) Electrohydraulic Active Damping System.
Annals of the CIRP 54(1):389392.
[15] Bu nte A, Beineke S (2004) High-Performance Speed Measurement by Sup-
pression of Systematic Resolver and Encoder Errors. IEEE Transactions on
Industrial Electronics 5(1):4953.
[16] Brown R, Schneider S, und Mulligan M (1992) Analysis of Algorithms for
Velocity Estimation from Discrete Position Versus Time Data. IEEE Transac-
tions on Industrial Electronics 39(1):1119.
[17] Chen Y, Tlusty J (1995) Effect of Low-Friction Guideways and Lead-Screw
Flexibility on Dynamics of High-Speed Machines. Annals of the CIRP
44(1):353356.
[18] Chen Y-D, Fuh C-C, Tung P-C (2005) Application of Voice Coil Motors in
Active Dynamic Vibration Absorbers. IEEE Transactions on Magnetics 41:
11491154.
[19] Cheng C-W, Tsai M-C (2004) Real-time Variable Feed Rate NURBS Curve
Interpolator for CNC Machining. International Journal of Advanced Manufac-
turing Technology 23(1112):865873.
[20] Choi C, Tsao T-C (2005) Control of Linear Motor Machine Tool Feed Drives for
End Milling: Robust MIMO Approach. Mechatronics 15:12071224.
[21] Chung B, Smith S, Tlusty J (1997) Active Damping of Structural Modes in
High-Speed Machine Tools. Journal of Vibration and Control 3:279295.
[22] Cuttino JF, Dow TA, Knight BF (1997) Analytical and Experimental Identica-
tion of Nonlinearities in a Single-Nut, Preloaded Ball Screw. ASME Journal of
Mechanical Design 119(1):1519.
[23] Dadalau A, Mottahedi M, Groh G, Verl A (2010) Parametric Modeling of Ball
Screw Spindles. Production Engineering 4(6):625631.
[24] Denkena B, Brehmeier S, Mo hring H-C (2008) Active Linear Guidances for
Micro Actuators: Alternative Concepts and First Prototypes. Journal of Micro-
system Technologies 14(12):19611973.
[25] Denkena B, Hackeloeer F, Neuber C-C (2009) 5-DOF Harmonic Frequency
Control Using Contactless Magnetic Guides. CIRP Journal of Manufacturing
Science and Technology 2(1):2128.
[26] Denkena B, To nshoff HK, Li X, Imiela J, Lapp C (2004) Analysis and Control/
Monitoring of the Direct Linear Drive in End Milling. International Journal of
Production Research 42(24):51495166.
[27] Denkena B, Hesse P, Gu mmer O (2009) Energy Optimized Jerk-decoupling
Technology for Translatory Feed Axes. Annals of CIRP 58(1):339342.
[28] Denkena B, Kallage F, Ruskowski M, Popp K, To nshoffr HK (2004) Machine
Tool with Active Magnetic Guides. Annals of the CIRP 53(1):333336.
[29] Denkena B, Li J, Kopp D (2004) An Aerostatic Linear Guidance for Micro-
systems. Annals of the German Academic Society for Production Engineering
XI(2):203208.
[30] Dietmair A, Verl A (2009) Drive Based Vibration Reduction for Production
Machines. Modern Machinery Science Journal 3(4). Prague, Czech Republic.
[31] N.N., DIN 69051-3: Werkzeugmaschinen; Kugelgewindetriebe; Abnahmebe-
dingungen und Abnahmepru fungen, Mai 1998.
[32] N.N., DIN69051-4: KugelgewindetriebeBerechnung der statischen und
dynamischen Tragzahl sowie der Lebensdauer, Entwurf April 1989.
[33] N.N., DIN69051-5: Werkzeugmaschinen; Kugelgewindetriebe; Anschlussmae
fu r Kugelgewindemuttern, September 1992.
[34] Dumur D, Boucher P, Ramond G (2000) Direct Adaptive Generalized Pre-
dictive Control. Application to Motor Drives with Flexible Modes. Annals of
the CIRP 49(1):271274.
[35] Ekinci TO, Mayer JRR, Cloutier GM (2009) Investigation of Accuracy of
Aerostatic Guideways. International Journal of Machine Tools & Manufacturing
49(6):478487.
[36] Eloundou RF, Singhose WE (2003) Command Generation Combining Input
Shaping and Smooth Base-line Functions, US Patent Application Publication
US0218440A1, US Patent Ofce.
[37] Erkorkmaz K, Altintas Y (2001) High Speed CNC System Design. Part II.
Modeling and Identication of Feed Drives. International Journal of Machine
Tools and Manufacture 41(10):14871509.
[38] Erkorkmaz K, Altintas Y, Yeung CH (2006) Virtual Computer Numerical
Control System. Annals of CIRP 55(1):399402.
[39] Erkorkmaz K, Kamalzadeh A (2006) High Bandwidth Control of Ball Screw
Drives. Annals of the CIRP 55(1):393398.
[40] Erkorkmaz K, Gorniak JM, Gordon DJ (2010) Precision Machine Tool XY
Stage Utilizing a Planar Air Bearing Arrangement. Annals of CIRP 59(1):
425428.
[41] Etxaniz I, Izpizua A, San Martin M, Arana J (2006) Magnetic Levitated 2D Fast
Drive. IEEJ Transactions on Industry Applications 126(12):16781681.
[42] Verl A, Frey S (2010) Correlation between Feed Velocity and Preloading in Ball
Screw Drives. CIRP Annals 59(1):429432.
[43] Frey S, Walther M, Verl A (2010) Periodic Variation of Preloading in Ball
Screws. Production Engineering 4(2):261267.
[44] Gianchandani YB, Tabata O, Zappe H (2008) Comprehensive Microsystems, vol.
2. Elsevier, Amsterdam/Heidelberg. ISBN 978-0-444-52192-7.
[45] Gleich S (2007) Approach for Simulating Ball Screws in Thermal Finite
Element Simulation, Workshop on Supervising and Diagnostics of Machining
Systems. Journal of Machine Engineering 7(1):101107.
[46] Gromann K, Mu ller J (2009) Untersuchungsergebnisse zur Wirksamkeit der
Impulskompensation von Lineardirektantrieben. ZWF Zeitschrift fur
wirtschaftlichen Fabrikbetrieb 104(9):S761S767.
[47] Golz HU, Weule H (1990) Vorspannungsregelung im Kugelgewindetrieb ein
neuer Ansatz fu r den Werkzeugmaschinenbau? Wt Werkstattstechnik 80,
Springer. 366368.
[48] Hara R (1989) Apparatus for Controlling Acceleration and Deceleration for
Servo Control, European Patent EP 0378708 A1, Paris.
[49] Huang S, Luo S, Leonardi F, Lipo TA (1999) A Comparison of Power Density for
Axial Flux Machines Based on the General Purpose Sizing Equation. IEEE
Transactions on Energy Conversion 14(2):185192.
[50] Hwang CL, Chen YM (2004) Discrete Sliding-mode Tracking Control of High-
displacement Piezoelectric Actuator Systems. ASME Journal of Dynamic Sys-
tems Measurement and Control 126:721725.
[51] Heng M, Erkorkmaz K (2010) Design of a NURBS Interpolator with Minimal
Feed Fluctuation and Continuous Feed Modulation Capability. International
Journal of Machine Tools & Manufacturing 50:281293.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 794
[52] Jamaludin Z, Brussel HV, Swevers J (2007) Classical Cascade and Sliding Mode
Control Tracking Performances for a XY Feed Table of a High-speed Machine
Tool. International Journal of Precision Technology 1(1):6574.
[53] Jamaludin Z, Van Brussel H, Pipeleers G, Swevers J (2008) Accurate Motion
Control of XY High-Speed Linear Drives using Friction Model Feed-Forward
and Cutting Forces Estimation. Annals of the CIRP 57(1):403406.
[54] Jeon JW, Ha YY (2000) A Generalized Approach for the Acceleration and
Deceleration of Industrial Robots and CNC Machine Tools. IEEE Transactions
on Industrial Electronics 47(1):133139.
[55] Jeon SY, Kim KH (2004) A Fluid Film Model for Finite Element Analysis of
Structures with Linear Hydrostatic Bearings. Proceedings of the Institution of
Mechanical Engineering Part C Journal of Mechanical Engineering Science
218(3):309316.
[56] Jones SD, Ulsoy AG (1999) An Approach to Control Input Shaping with
Application to Coordinate Measuring Machines. ASME Journal of Dynamic
Systems Measurement Control 21(2):242247.
[57] Kamalzadeh A, Erkorkmaz K (2007) Compensation of Axial Vibrations in Ball
Screw Drives. Annals of the CIRP 56(1):373378.
[58] Kamalzadeh A, Erkorkmaz K (2007) Accurate Tracking Controller Design for
High Speed Drives. International Journal of Machine Tools & Manufacturing
47:13931400.
[59] Kavanagh RC (2000) Improved Digital Tachometer With Reduced Sensitiv-
ity to Sensor Nonideality. IEEE Transactions on Industrial Electronics
47(4):890897.
[60] Kim M-S, Chung S-C (2006) Integrated Design Methodology of Ball-Screw
Driven Servomechanisms with Discrete Controllers. Part I. Modelling and
Performance Analysis. Mechatronics 16:491502.
[61] Knorr M (2001) Impulsgekoppelter TransmissionsantriebEuropean Patent
and Trademark Ofce, Patent Specication EP 1 247 613 B1 (date of ling
06.04.2001).
[62] Koren Y, Lo CC (1992) Advanced Controllers for Feed Drives. Annals of the CIRP
41(2):689698.
[63] Koren Y (1997) Control of Machine Tools. Transactions of ASME Journal of
Manufacturing Science and Engineering 110:749755.
[64] Ku nzel S, et al. (2002) Verfahren zur Da mpfung mechanischer Schwingungen
von Achsen von Werkzeugmaschine. German Patent and Trademark Ofce,
Patent specication DE 102 46 093C 1, Siemens AG, Mu nchen (date of ling
02.10.2002).
[65] Ladra U, Scha fers E, Schu r T (2007) Operating Method for an Assembly
Having a Mechanically Movable Element, Patent Application WO 2007/
115874 A1.
[66] Lee SH, Song JB (2001) Acceleration Estimator for Low-velocity and Low-
acceleration Regions Based on Encoder Position Data. IEEE/ASME Transaction
on Mechatronics 6(1):5864.
[67] Lin MC, Ravani B, Velinsky SA (1994) Kinematics of the Ball Screw Mechan-
ism. ASME Journal of Mechanical Design 116:849855.
[68] Ma rton L, Lantos B (2009) Control of Mechanical Systems with Stribeck
Friction and Backlash. Systems & Control Letters 58(2):141147.
[69] Mekid S (2000) High Precision Linear Slide. Part I. Design and Construction.
International Journal of Machine Tools & Manufacturing 40(7):10391050.
[70] Mekid S, Olejniczak O (2000) High Precision Linear Slide. Part II. Control and
Measurements. International Journal of Machine Tools & Manufacturing
40(7):10511064.
[71] Nam S-H, Yang M-Y (2004) A Study on a Generalized Parametric Inter-
polator with Real-time Jerk-limited Acceleration. Computer-Aided Design
36(1):2736.
[72] Neugebauer R, Denkena B, Wegener K (2007) Mechatronic Systems for
Machine Tools. Annals of the CIRP 56(2):657686.
[73] Neugebauer R, Drossel W-G, Ihlenfeldt S, Harzbecker C (2009) Design
Method for Machine Tools with Bionic Inspired Kinematics. CIRP Annals
59(1):371374.
[74] Neugebauer R, Wittstock V, Illgen A, Naumann G (2006) Piezo-based
Actuator-Sensor-Units for Uni-axial Vibration Reduction in Machine Tools.
Annals of the German Academic Society for Production Engineering WGP
XIII(2):211216.
[75] Novotny L, Strakos P, Vesely J, Dietmair A (2009) Potential of State-feedback
Control for Machine Tool Drives, Modern Machinery Science Journal
2009(2):98112, Paper No. 4, Prague.
[76] Nordin M, Gutman P-O (2002) Controlling Mechanical Systems with Backlash
A Survey. Automatica 38(10):16331649.
[77] Okwudire C, Altintas Y (2009) Hybrid Modeling of Ball Screw Drives with
Coupled Axial, Torsional and Lateral Dynamics. Transactions of ASME Journal
of Machine Design 131(071002):19. 10.1115/1.3125887.
[78] Okwudire C, Altintas Y (2009) Minimum Tracking Error Control of Flexible
Ball Screw Drives using a Discrete-time Sliding Mode Controller. Transactions
of ASME Journal of Dynamic Systems Measurement and Control 131(5,
051006):112. 10.1115/1.3155005.
[79] Ostermann T, Weck M (2003) Lineardirektantriebe mit Impulsentkopplung.
Antriebstechnik 42(11):6164. ISSN 0722-8546.
[80] Park CH, Chung JH, Lee H, Kim ST (2004) Motion Error Analysis of the Porous
Air Bearing Stages Using the Transfer Function. Journal of Korean Society of
Precision Engineering 21(7):185194.
[81] Park CH, Oh YJ, Shamoto E, Lee DW (2006) Compensation for Five DOF Motion
Errors of Hydrostatic Feed Table by Utilizing Actively Controlled Capillaries.
Precision Engineering 30(3):299305.
[82] Pislaru C, Ford DG, Holroyd G (2004) Hybrid Modelling and Simulation of a
Computer Numerical Control Machine Tool Feed Drive. Proceedings of the
Institution of Mechanical Engineers Part I Journal of Systems and Control
Engineering 218(2):111120.
[83] Pritschow G (1998) A Comparison of Linear and Conventional Electromecha-
nical Drives. Annals of the CIRP 47(2):541547.
[84] Pritschow G, Altintas Y, Jovane F, Koren Y, Mitsuishi M, Takata S, Van Brussel
H, Weck M, Yamazaki K (2001) Open Controller Architecture Past, Present
and Future. Annals of CIRP 50(2):446463.
[85] Pritschow G, Eppler C, Lehner W-D (2003) Ferraris Sensor The Key for
Advanced Dynamic Drives. CIRP Annals 52(1):289292.
[86] Pritschow G, Philipp W (1990) Direct Drives for High-Dynamic Machine Tool
Axes. Annals of the CIRP 39(1):413416.
[87] Pritschow G, Philipp W (1992) Research on the Efciency of Feedforward
Controllers in Direct Drives. Annals of the CIRP 41(1):411415.
[88] Pritschow G, Ro ck S (2004) Hardware in the Loop Simulation of Machine
Tools. CIRP Annals-Manufacturing Technology 53(1):295298.
[89] Pritschow G (1996) On the inuence of the Velocity Gain Factor on the Path
Deviation. Annals of the CIRP 45(1):367372.
[90] Pritschow G (2001) Antriebssystem sowie Verfahren zur Bestimmung der
Bandbreite eines solchen Antriebssystems, German Patent and Trademark
Ofce, Patent Specication DE 10135220.
[91] Ramesh R, Mannan MA, Poo AN (2000) Error Compensation in Machine
Tools A Review. Part I. Geometric, Cutting-force Induced and Fixture-
dependent Errors. International Journal of Machine Tools and Manufacture
40(9):12351256.
[92] Reinhart G, Za h MF, Baudisch T (2004) Simulation Environment for Designing
the Dynamic Motion Behaviour of the Mechatronic System Machine Tool.
Journal of Production Engineering-Research and Development Jg XI(1):145148.
[93] Renton D, Elbestawi MA (2001) Motion Control for Linear Motor Feed Drives
in Advanced Machine Tools. International Journal of Machine Tools and Man-
ufacture 41:479507.
[94] Ro ck S, Pritschow G (2007) Real-time Capable Finite Element Models with
Closed-Loop Control A Method for Hardware-in-the-Loop Simulation of
exible Systems. WGP Annals-Production Engineering Research & Development
XIV(1):3743.
[95] Sakano T (1986) Acceleration/Deceleration Control System, US Patent
4,603,286, Fanuc Ltd., Assignee.
[96] Shamoto E, Park CH, Moriwaki T (2001) Analysis and Improvement of Motion
Accuracy of Hydrostatic Feed Table. Annals of the CIRP 50(1):285288.
[97] Shamoto E, Suzuki N, Hamaguchi A (2006) A New Fluid Bearing Utilizing
Traveling Waves. Annals of the CIRP 55(1):411414.
[98] Shinno H, Yoshioka H, Taniguchi K (2007) A Newly Developed Linear Motor-
Driven Aerostatic XY Planar Motion Table System for Nano-Machining.
Annals of the CIRP 56(1):369372.
[99] Shinno H, Hashizume H, Yoshioka H, Komatsu K, Shinshi T, Sato K (2004) X
Yu Nano-Positioning Table System for a Mother Machine. Annals of the CIRP
53(1):337340.
[100] Shinno H, Hashizume H, Sato K (1999) Nanometer Positioning of a Linear
Motor-driven Ultraprecision Aerostatic Table System with Electrorheological
Fluid Dampers. Annals of the CIRP 48(1):289292.
[101] Singer NC, Seering WP, Pasch K (1989) Shaping Command Input to Minimize
Unwanted Dynamics, Europa ische Patentschrift EP0433375B1, Europa isches
Patentamt, Paris.
[102] Sipahi R, Olgac N (2003) Active Vibration Suppression With Time Delayed
Feedback. Journal of Vibration and Acoustics 125(3):384389.
[103] Slotine J-JE, Li W (1988) Adaptive Manipulator Control: A Cast Study. IEEE
Transactions on Automatic Control 33(11):9951003.
[104] Slocum A (1992) Precision Machine Design. Prentice-Hall . 401412
(Chapter 7.7).
[105] Slocum A (2003) Linear Motion Carriage withAerostatic Bearings Preloaded
by Inclined Iron Core Linear Electric Motor. Precision Engineering
27(1):382394.
[106] Slutka M, Novotny L, Sveda J, Strakos P, Hudec J, Smolik J, Vlach P (2008)
Machine Tool Lightweight Design and Advanced Control Techniques. MM
Science Journal 3136. ISSN 1803-1269.
[107] Spath D, Rosum J, Haberkern A (1995) Kinematics, Frictional Characteristics
and Wear Reduction by PVD Coating on Ball Screw Drives. Annals of the CIRP
44(1):349352.
[108] Symens W, Van Brussel H, Swevers J (2004) Gain-scheduling Control of
Machine Tools with Varying Structural Dynamics. Annals of the CIRP
53(1):321324.
[109] Takeuchi Y, Sakaida Y, Sawada K, Sata T (2000) Development of 5-axis Control
Ultraprecision Milling Machine for Micromachining Based on Non-friction
Servomechanism. Annals of the CIRP 49(1):295298.
[110] Tarng YS, Kao JY, Lin YS (1997) Identication of and Compensation for
Backlash on the Contouring Accuracy of CNC Machining Centers. International
Journal of Advanced Manufacturing Technology 13(5):359366.
[111] Tomizuka M (1987) Zero Phase Error Tracking Algorithm for Digital Control.
ASME Journal of Dynamic Systems Measurement and Control 109:6568.
[112] Tonshof HK, Lapp C, Kaak R (1998) Control of Linear Direct Drives and New
Demands on Machine Tool Structures. Annals of the German Academic Society
for Production Engineering V(2):5156.
[113] Trzynadlowski AM (2001) Control of Induction Motors. Academic Press, San
Diego.
[114] Tsai M-C, Cheng M-Y, Lin K-F, Tsai N-C (2005) On Acceleration/Deceleration
Before Interpolation for CNC Motion Control. Proceedings of the 2005 IEEE
International Conference on Mechatronics, Taipei, Taiwan, 10.07.12.07.2005,
382387.
[115] Tsuji T, Hashimoto T, Kobayashi H, Mizuochi M, Ohnishi K (2009) A Wide-
range Velocity Measurement Method for Motion Control. IEEE Transactions on
Industrial Electronics 56(2):510519.
[116] Utkin VI (1977) Variable Structure Systems with Sliding Modes. IEEE Trans-
actions on Automatic Control 22(2):212222.
[117] Van Brussel H, Chen C-H, Swevers J (1994) Accurate Motion Controller Design
Based on an Extended Pole Placement Method and a Disturbance Observer.
Annals of CIRP 43(1):367372.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 795
[118] Van Brussel H, Sas P, Istvan N, De Fonseca P, Van den Braembussche P (2001)
Towards a Mechatronic Compiler. IEEE/ASME Transactions on Mechatronics
6(1):90105.
[119] Van Brussel H, Van den Braembussche P (1998) Robust Control of Feed Drives
with Linear Motors. Annals of the CIRP 47(1):325328.
[120] Van den Braembussche P, Swevers J, Van Brussel H (2001) Design and
Experimental Validation of Robust Controllers for Machine Tool Drives with
Linear Motor. Mechatronics 11:545561.
[121] Varanasi KK, Nayfeh SA (2004) The Dynamics of Lead-screw Drives: Low-
order Modeling and Experiments. Transactions of ASME Journal of Dynamic
Systems Measurement and Control 126:388396.
[122] Verl A, Bodson M (1998) Torque Maximization for Permanent Magnet Syn-
chronous Motors. IEEE Transactions on Control Systems Technology 6(6):740745.
[123] ISG-Virtuos A Powerful Tool for Simulation and Visualization, ISG Indus-
trielle Steuerungstechnik GmbH, Stuttgart, www.isg-stuttgart.de.
[124] Weck M, Ye G (1990) Sharp Corner Tracking Using the IKF Control Strategy.
Annals of the CIRP 39(1):437441.
[125] Weck M, Kru ger P, Brecher C (2001) Limits for Controller Settings with
Electric Linear Direct Drives. International Journal of Machine Tools and
Manufacture 41(1):6588.
[126] Weck M, Brecher C, Schulz A, Keiser R (2003) Coupled Simulation of Control
Loop and Structural Dynamics. Annals of the German Academic Society for
Production Engineering X(2):105110.
[127] Weck M, Brecher C (2005) Werkzeugmaschinen Band 2- Konstruktion und
Berechnung, Springer-VDI, Berlin/Heidelberg/New York, 8. Auage.
[128] Weule H, Golz U (1988) Dynamisches Verhalten von Kugelgewindetrieben.
wt Werkstattstechnik 78, Springer. 623626.
[129] Weck M, Wahner U (1998) Linear Magnetic Bearing and Levitation System for
Machine Tools. Annals of the CIRP 47(1):311314.
[130] Whalley R, Ebrahimi M, Abdul-Ameer AA (2006) Machine Tool Axis
Dynamics. Proceedings of the IMechE Part C Journal of Mechanical Engineer
Science 220:403419.
[131] Zaeh MF, Oertli T, Milberg J (2004) Finite Element Modelling of Ball screw
Feed Drive Systems. CIRP Annals 53(1):289294.
[132] Zaeh MF, Englberger G (2005) Robust Speed Control of Feed Drive Systems
with Ball Screw for Machine Tools. Journal of Production Engineering-Research
and Development XII 1:209212.
[133] Zhang Y, Alleyne A (2003) A Simple Novel Approach to Active Vibration
Isolation With Electrohydraulic Actuation. Transactions of ASME Journal of
Dynamic Systems Measurement and Control 125(1):125128.
[134] Zatarain M, Ruiz de Argandona I, Illarramendi A, Azpeitia JL, Bueno R (2005)
New Control Techniques Based on State Space Observers for Improving the
Precision and Dynamic Behaviour of Machine Tools. Annals of the CIRP
54(1):393396.
[135] Zhu W-H, Lamarche T (2007) Velocity Estimation by Using Position and
Acceleration Sensors. IEEE Transactions on Industrial Electronics 54(5):
27062715.
[136] Zirn O (2008) Machine Tool Analysis Modelling, Simulation and Control of
Machine Tool Manipulators, Habilitation Thesis, Department of Mechanical &
Process Engineering, ETH Zu rich (http://e-collection.ethbib.ethz.ch/view/
eth:41862).
[137] Zirn O, Weikert S, Rehsteiner F (1996) Design and Optimization of Fast
Axis Feed Drives Using Nonlinear Stability Analysis. Annals of CIRP 45(1):
363366.
Y. Altintas et al. / CIRP Annals - Manufacturing Technology 60 (2011) 779796 796