Sie sind auf Seite 1von 78

Linear Motor Feed Drives

for Grinding Machines

谢秋林 博士研究生
梁越升 教授
7/24/2006
美国佐治亚理工学院机械工程系
Outline of Presentation

 Introduction
 Review of Feed-drive Technologies
 Research Road Map
 System Modeling and Simulations
 Modeling
 Simulations
 Controller Design and Simulations
 Controller Design
 Controller Simulations
 Conclusions and Future Works
Introduction

 Grinding-Abrasive machining process


 Challenges to grinding machines
 Motivations
 Key factors influencing machining accuracy
 Machine tool feed drives
Crucial part for ultra-precision grinding machine
 Objectives
 Grinding-Abrasive machining process
Employ very hard granular refractory particle
Grinding wheel moves at a high surface speed
Applications: Hard or brittle workpiece material, fine
requirements on both finish and dimensional accuracy

Timken precision Sumitomo precision Camshaft Edge grinding of


forging die silicon wafers
bearing
Challenges
 The trend towards miniaturization is continuing
Parts get smaller and smaller. And even the small parts get
more intricate features
PZT actuators 20 gram push force
Finish 60 nm
microsliders and submicron
Ra
resolution

1.5 mm x 1.5 mm x
6 mm- Stroke10 mm

AIST Microlathe World’s smallest linear motor on a


Machined ‘‘microhat’’
fingertip - The SQL-1.5 SQUIGGLE
motor
Challenges
 Even larger work piece have stringent
requirements on accuracies (in micron
tolerance or less) and surface finish
The surface
finish may
be below Ra
50nm

Precision optical glass Intel 0.13m wafer TAMARON17-50/F2.8


moulding mirror
Motivations
 Key factors influencing machining accuracy
Machining center Tooling Materials to be machined
Clamping system Aluminum
Machine frame
Grinding wheel Brass
Slideways
(Aluminum Oxide Copper
Feed Drives ,CBN, Diamond)
Titanium
CNC
Inconel
Spindle-Grinding wheel
Steel up to 62 HRC
In-process tool and
workpiece measuring
capability
Conditioning abilities
Cooling/lubrication system
Motivation
 Machine tool feed drives
 Crucial part for ultra-precision grinding machine

Requirements
Function
High positioning precision
Control the positions
Ability to withstand machining
and velocities of
loads while maintaining
machine tool slides
accuracy of position control
or axes according to
Rapid response of drive
commands issued
system to command inputs from
by CNC interpolator
the machine tool CNC system
analysis and
Precise coordination of the
applications.
control of multiple axes of the
machine tool in contouring
operations
Motivations & Objectives

 Justify the selection of linear motor feed drive


Evaluations of current feed drive technologies
 Identify and model functional critical factors
which will influence the implementation of
linear motor
 Propose model-based servo control algorithm
which will guarantee the performance of linear
motor feed drive
Review of Feed-drive Technologies

 PZT
 Hydraulic Drive
 Voice Coil Motor
 Friction Drive
 Leadscrew-Ballscrew
 Hydrostatic Leadscrew
 Linear Motor
PZT

 Principle

Principle of PZT

Physik Instrumente (PI)


modular PZT nano-positioning stage
PZT
 Advantages
 No backlash nor friction
 large forces
 high bandwidth
 0.1nm accuracy
 Disadvantages
 Open loop hysteretic behavior
 The amount of deflection or deformation is limited-
0.1mm typically
 Low strength in other directions
PZT

 Applications

Precitech’s FTS system for diamond


tuning of non rotational-symmetrical
surface-micro- prisms, off-axis aspheres

Fast tool servo for turning non-rotation


Principle of AFM
Symmetrical components
Hydraulic Drive

 Principle
Hydraulic motor with
servo-walve and
tachogenerator

Lead screw
and nut ,and
slider

Component diagram of an electrohydraulic feed drive


Hydraulic Drive
 Advantages
 High force(100N/mm2) at moderate speed
 High power density at point of action
 High bandwidth motion control at high forces
 Disadvantages
 Possible leakage
 Susceptibility of working fluid to contamination
Hydraulic Drive
 Applications

OKAMOTO ACC-DX series


Johndeere 650D LC Excavator Hydraulically operated table
surface grinding machine
Voice Coil Motor (VCM)
 Principle
 It provides the motive force to the cone by the reaction of a
magnetic field to the current passing through it

Equipment Solutions VCS-10


Voice Coil Linear Stage Principle of voice coil
VCM

 Advantages
 Free from backlash
 Velocity is easily adjusted and controlled
 High accuracy- 0.1 micron
 Disadvantages
 Short strokes-amplitude within 1mm
VCM
 Applications

VCM

Western Digital Caviar SE


WD800JB 80GB 7200 RPM IDE UIUC Three-axis MMT prorotype
Ultra ATA100 Hard Drive
Leadscrew-Ballscrew
 Principle

Hiwin precision ground ballscrew Scheme of ballscrew


Leadscrew-Ballscrew

 Advantages
 Very adaptable to varying machine sizes, feed rates and
process forces
 Disadvantages
 Dead zone and backlash
preloaded ballcrew can decrease backlash at the expense
of increasing friction
 Transmission errors (e.g. pitch error)
 Position, velocity and acceleration limitations (Structural
flexibility)
 and Wear
Friction Drives
 Advantages
 Minimal backlash
 Uncomplicated design
 Disadvantages
 Low drive force
capability
 Low to moderate
stiffness and damping Schematic diagram of
 High sensitivity to drive friction drive
bar cleanliness
 Frictional polymers can
form on dry-running
systems
Lead Screw-
Hydrostatic Leadscrew

 Principle

Schönfeld
hydrostatic lead
screw
Principle of hydrostatic lead
screw
Lead Screw-
Hydrostatic Leadscrew

 Advantages
 No backlash ,minimal friction and no wear
 High nut stiffness
 High positioning accuracy
 Disadvantages
 Oil contamination
 Bulky
Linear Motor
 Principle

Principle of linear motor

430公里/小时

Schematic drawing of linear motor


Shanghai's magnetic levitation train
This flat assembly of coils
 Ironless-core moves between two rows of
magnets creating the multi-pole
magnetic flux.
Forcer The MAGNETIC WAY (normally
attached to the machine frame)
is made up of two rows of
magnets glued to a steel plates
assembled in a U-form .
The COIL ASSEMBLY (moving
part, connected to the
component to be positioned)
includes flat coils assembled in
a composite structure.
this design is free of both magnetic
attraction and detent forces.
 Iron-core
The MAGNETIC WAY (normally attached
to the machine frame) is made up of a row
of magnets glued to a steel plate.
The COIL ASSEMBLY (moving part,
connected to the component to be
positioned) includes the lamination stacks
and windings.
Advantages of Linear Motor
Drive Systems
 Mechanical simplicity no mechanical transmissions
mechanisms), higher reliability, and longer lifetime
 No backlash and less friction, resulting in the potential of high
positioning accuracy
 The problem of stick-slip is reduced with the direct linear drive
 No mechanical limitations on achievable velocity and
acceleration
 Linear motor acceleration rates are limited by the linear bearings, most
of which will tolerate 2 or 3Gs, and those that will take 5Gs are now
available.
 A the linear motors regularly travel up to 5m/s where a lead-screw
typically limits velocity to less than 1.5m/s.
 Bandwidth only limited by encoder resolutions, measurement
noises, calculation time and frame stiffness
 The scales used with the direct linear drives are often resolved to 1nm
Applications

With
hydrostatic
slide way
system

Sodick MC430L milling Okamoto UPG-31Li


machine: Vertical machining profile grinder with an Landis LT2
center with linear motor drive accuracy of 0.3 µm crankshaft grinder
Comparisons among Feed Drive
Technologies

Voice coil Linear motor PZT Ball screw Hydraulic

Higher inertia
Amplitude 1mm No limitation 0.1mm 1000mm
with added
length
0.1 micron 0.1 micron 0.1nm 1 micron 10 micron
Accuracy

Force Up to
300N Up to 30 N/mm2 100N /mm2
/pressure 100,000N
20,000N

Response time 1 ms 10-20ms 0.1ms 100ms 1s


Research Road Map

 Challenges
 Research Objective
 Research Flow Chart
Challenges to Control Linear Motor

 Model uncertainties
 Parametric uncertainties (e.g., load inertia)
 Discontinuous disturbance
 Friction induced stick-slip at low velocity at which grinding machine table is
typically operated
 Pre-sliding behavior which is critical when motion is down to submicron
region
 External disturbances (grinding force)
 Uncompensated grinding force deteriorates position performance
 Significant uncertainties due to position dependent force ripple
 detrimental to fast and accurate position control
 Drawback of without mechanical transmissions
 Gear reduction reduces the effect of model uncertainties and external
disturbances
 Implementation issues( e.g., measurement noise)
Research Objective

 To develop a systematic methodology to achieve


submicron positioning performance for linear motor
direct feed drive performance for grinding machines
under the effects of functionality critical force ripple,
friction and grinding force by systematic experiments,
simulation, and sophisticated controller design
Research Road Map
Literature survey on direct drive system

Force ripple modeling Friction modeling Grinding force modeling

Direct drive system simulation

Experimental characterization

Adaptive sliding mode position control

Experimental validation
System Modeling and Simulations

 System Modeling
 Rigid-body motion equation
 Force ripple-How to characterize ripple using
analytical ,or empirical approaches
 Grinding force
 Friction -How to quantitize the friction force
 Modeling Validation
 Open-loop Simulation Results
 Summary on Modeling
System Modeling
Wheel surface speed

Normal grinding force


Linear motor
Table primary part (coils)
Tangential grinding force

Friction Linear motor thrust force


force

Linear motor
secondary part
(Magnet track)

Schematic diagram of linear motor feed drive for cylindrical grinding machine
Newton’s Law

 Rigid-body motion equation

F (t ) = mx + f friction + f ripple + f grinding

F (t ) = K f u (t )
m: Moving thrust mass
F (t ) : Thrust force

Kf : Force constant determined by the amount of motor poles


and magnetic flux, etc
u (t ) : Control signal in terms of voltage
Force Ripple Modeling
 Cogging
Cogging is a magnetic disturbance force
Due to end effect and mutual attraction between the magnet
and steel for iron-core motor
Affect the smoothness of velocity profile that can be
achieved
 Reluctance
The reluctance is an electro-magnetic effect
Due to the change of windings inductance
 Commutation error
Due to the non-perfect sine distribution of the magnetic field
Both will affect the fast positioning capability and
low speed performance
Force Ripple Modeling

 Force ripple [Braembussche et al. 1996 ]


Fcogging = ac sin(ω x x) + bc sin(3ω x x + φc )
Freluctance = Fdesired [a f sin(ω x x + φ11 ) + b f sin(3ω x x + φ12 )]
the position of the motor carriage as it is measured with the optical linear encoder

2π the average pitch of the magnets for the considered stroke


ωx = p
p

Variation of force
ripple w.r.t. position
Grinding Force Modeling
 Grinding process-Three phases
 Rubbing
the grain rubs on the work causing elastic
and some plastic deformation, no material
removal
 Ploughing
plastic flow of work material in the direction of
sliding, extruded material being throw up and
broken off along the sides of the groove,
resulting low rate of stock removal
Rubbing, ploughing,
 Cutting and cutting regimes
fracture takes place in the plastically stressed of deformation in
zone just ahead of the rubbing grain, causing abrasive machining
the formation of chips and resulting in fairy
rapid stock removal rates
Grinding Force Modeling
 Grinding force (Hahn and Lindsay's model)
 Assume that rubbing and ploughing energy can be ignored( for easy-
to-grind material)
 Assume that normal grinding force is proportional to workpiece
material removal rate (MMR)
Fn = cπ d w bsVreff
Vs
Kw Kc Ks d F
Vr − Vreff = ( n )
Vw dt K eq
1 1 1 1
= + +
K eq K c K w K s
Fn Fn ( s ) cπ d wbs
=
Vr ( s ) 1 + (cπ d wbs / keq ) s
Feed slide

Vr C: the proportionality
constant describing the
dullness of the grinding
Bed wheel
Friction Modeling

 Two different regimes


 Sticking regime (Pre-sliding regime)
 The adhesive force (at asperity contacts)
are dominant
 Friction force appears to be a function of
displacement because asperity junction
deforms elasto-plastically
 As displacement increases more and more
junctions will break resulting eventually in
gross sliding
 Sliding regime friction interface btw two
surface (contact between
 All asperity junctions have broken apart (or, two bristles)
all asperity contact)
 Friction is more pronounced function of
velocity, due for example to the build-up of
lubricating film
Friction Modeling

 Static friction model  Dynamic friction model


 Dahl model (1968). Coulomb and
 Coulomb model
pre-sliding.
 Coulomb, viscous model  LuGre model [ De Wit et al.

 Static, Coulomb, viscous 1995. Pre-sliding displacement,


model varying breakaway force, and a
 Static, Coulomb, viscous, frictional lag
Stribeck model

Stribeck Curve vs. Velocity [Armstrong-helouvry, 1994].


Friction Modeling
 LuGre Model, De Wit et al. 1995
 Combination of static and dynamic friction model

dz Symbol Meaning
F f = σ 0 z + σ 1 + σ 2υ
dt Coulomb friction
fc
d z | υ |
= υ − σ 0 z Static friction force
d t g (υ ) fs

δVs The state variable ,


−υ/Vs
g(υ) = fc +( fs − fc)e z represents the average
deflection of the
contacting asperities
σ0
Gaussian model δV
s
= 2 asperity stiffness
σ1
Vs Stribeck velocity the micro-viscous friction
coefficient
σ2
the viscous friction
Modeling Validation
 Friction model validation
u = 0.01sin(40π t )

Linerized system FRF


cm
H ( jω ) =
− J ω 2 + (σ 1 + σ 2 ) jω + σ 0
J 0.018kg2/rad
c 16Nm/V
σ0 1.21*104 Nm/rad

A rotating arm system J α = − F + cmu σ 1 + σ 2 4.47Nm/(rad/s)

Sinusoidal validation
The amplitude corresponds to 90% of static friction

u = 0.035sin(t )

Response is larger than simulation


 Plasic deformation or creep
of the system is not
incorporated in the LuGre
model
 The Lugre model structure is
too limited to describe the pre-
sliding behavior accurately,
e.g., the stiffness function
might be a nonlinear function

Sinusoidal validation
Modeling Validation
Simulation parameters
 Force ripple validation

parameter
f ripple = Avalue
r sin(ω x +units
φ)
fs 20 N
fc 10 N
vs 0.1 m/s
Ar 8.5 N/m
φ 0.05 π Ns/m
Open velocity response of a linear motor system
f ripple = Ar sin(ω x + φ )
σ0

Simulation Results (Open loop)

 Set 1-only friction Simulation parameters

 Pre-sliding parameter value units

 Breakway fs 120 N
 Set 2-friction and force fc 80 N
υs
ripple 0.1 m/s
σ0 2*106 N/m
 Only cogging
σ1 2*104 Ns/m
 Set 3-friction,force σ2 0.4 Ns/m
ripple and grinding
force
Step response


u = 12N u = 130 N

pre-sliding regime sliding regime


 Excitation amplitude=12N
 Excitation amplitude=120N
 Breakaway phenomenon

Excitation rate Breakway point


 Set 2- Force ripple effect

Amplitude=130N

Step response
 Force ripple effect Shift

Sinusoidal response
Sinusoidal response-force ripple
curve
 Set 3-Combination effects of three factors
(friction, force ripple and grinding force)

u=180N

Step response-forces curve Step response-


displacement curve
Modeling Summary
 Studied friction behaviors which are important
for high precision applications
 Investigated how force ripple influences motion
behavior
 Studied the combination effects of three critical
factor for linear motor feed drives
Controller Design
 Motivations
 Sliding Mode Controller Design
 Simulation Results
Controller Design-Motivations

 Inherent process nonlinearities


 Stringent performance requirements demand
explicit compensation of process nonlinearities
 Modeling uncertainties
 Robust nonlinear controller
Sliding mode control
Sliding Mode Controller
 Design a control law to effectively account for
 parameter uncertainty, e.g., imprecision on the mass
proprieties or loads, inaccuracies on the force/torque
constants of the actuator, friction, and so on
 the presence of unmodeled dynamics, such as structural
resonant modes, neglected time-delays( in the actuator) ,
or finite sampling rate
 Quantify the resulting modeling/performance
trade-offs, and in particular, the effect on tracking
performance of discarding any particular term in the
dynamic model
Controller Structure
 Controller Diagram
+ Output
Sliding Mode
LM
Controller
Reference Input
_

Voltage Current Thrust force


Axis
u + Position
Ka Kf Mass
Control
Signal _
Current Amplifier Motor
_ _ Friction

Grinding Force

_ Force Ripple
Design Procedure
 Controller derivation

F (t ) = mx + f friction + f ripple + f grinding F (t ) = K f u (t )


e = xd − x s = e + λe
Reaching
law

s = −η sgn( s ) − ks
m Ff + Fg + Fr
u= ( xd + λ ( x d − x ) + η sgn( s ) + ks ) +
Ka K f Ka K f
1 2  2
V= s V = ss = −η s − ks

2
Observer Design
 Observability analysis
 x 
x = f ( x ) + g ( x ) u  x1   x 
Χ =  x2  =  x   =  
Χ x 
 
y = h(x )  x3   z   z 

 0
  
 L f y = x  1 0 0 
O = ∇  = 0 1 
1
L f y = x 0
   
x

 L 2 y = − F ( x , z )  0 * σ − σ 
 f   0 1
h(υ ) 
m 
 Observer design
σ 0 | x |  σ 0 | x | z = − σ 0 | 
x |
z = x − z zˆ = x − zˆ z
g (υ ) g (υ ) g (υ )

− (υ / υs )2
g (υ ) = f c + ( f s − f c )e
1 2   x 2
V z = zˆ Vz = zz
 = −
g (υ )
z ≤ 0
2
Error dynamics will asymptotically converge if x ≠ 0, ~
z ≠0
Simulation Results
 Case 1-Sinusoidal response

Desired
trajectory
Position error Speed error

Control signal Friction force

Phase portrait
Simulation Results
 Case 2- Step response

Speed
error

Control Friction
signal force
Simulation Results
 Case 3- Acceleration-limited trajectory

Desired Position
trajectory error

Control
signal

Friction
force
Simulation Results

 Case 4- Jerk-limited trajectory


t
τ=
t f − t0
xd (τ ) = x0 + ( x0 − x f )(15τ 4 − 6τ 5 −10τ 3 )
t0 = 0, t f = 5
x0 = 0, x f = 100

vd (τ ) = ( x0 − x f )(60τ 3 − 30τ 4 − 30τ 2 )

ad (τ ) = ( x0 − x f ) (120τ 2 −120τ 3 − 60τ )

jd (τ ) = ( x0 − x f )(240τ − 240τ 2 − 60)

Jerk-limited trajectory
Simulation Results

 Position tracking error

max e(t ) =0.815µ m


RMS=0.0846µ m
Simulation Results

 Case 5- Ramp response

Position Speed Control


error error signal

Friction Force ripple


force Grinding
force
Conclusions and Future Work
 Reviewed the current feed-drive technologies
 Investigated three functional-critical factors
closely related to the implementation of linear
motors in grinding machines
 Presented robust controller design which can
handle external disturbances and model
uncertainties without the loss of system
performance
 The implementation of controller is coming
soon
Appendix Friction Terms

 Static friction: The force (torque) needed to initiate motion from rest.
 Coulomb friction: A friction component which does not depend on the
magnitude of velocity.
 Viscous friction: Velocity dependent friction.
 Break-away force: The amount of force needed to overcome static
friction.
 Dahl effect: Elastic deformation of asperity junction that behaves like a
linear spring for small displacements.
 Stribeck: Decreasing friction with increasing velocity at low velocities.
 Frictional memory: The lag between changes in velocity and changes in
friction force.
Appendix Magnetic circuit terms

H = ∇F
 Magnetic potential F = ∫ H idl

 Magnetic flux φ = ∫ B ids


s

 Reluctance φ=
F

 Permanence ℜ=
F
φ
=
lm
µ Am
 Inductance ρ=
1

λ Nφ
L= =
i i
Appendix-Friction Compensation
 Non-model based compensation
 Stiff PD control [Armstrong-helouvry 1992]
 Integral control [Tung et al. 1993]
 Dither [Cebuhar 1988]
 Impulse control [Yang and Tomizuka 1988; Suzuki and
Tomizuka 1991]
 Dual mode control [Futami et al. 1990 ; Ro and Hubbel 1993]
 Disturbance-observer-based compensation [Lampaert et al.
2004]
 Cerebellar Model Articulation Controller (CMAC) Neural
network control [Larsen et al.1995]
 Model based compensation
 Feed forward compensation [Tung 1996; Lampaert et al. 2004]
 Adaptive control –Model Reference adaptive control [Shafer
and Brandenburg 1993] and adaptive robust control [Yao and
Xu 2002]
Appendix-Force Ripple Modeling and
Compensation
 Modeling
 Cogging
Cogging is a magnetic disturbance force
Due to end effect and mutual attraction betwwn the magnet and steel for iron-
core motor
it will affect the smoothness of velocity profile that can be achived
 Reluctance
The reluctance is an electro-magnetic effect

 Commutation error
non-perfect sine distribution of the magnetic field
It will affect the fast positioning capabilitiy and low speed performance
 Compensation
 Model based feedforward compensation
Braembussche et al. 1996
 Learning control
Otten et al. 1997, Hu et al. 1999 and Lee et al. 2000
 Adaptive robust and robust adaptive control
Yao and Xu,2000 and 2002, Huang et al. 2002 and Tan et al. 2002
Why Nonlinear Control

 Improvement of existing control systems


 Analysis of hard nonlinearities
 Dealing with model uncertainties
 Design simplicity
T
1
RMS = ∫ e(t ) dt
2

T 0

Das könnte Ihnen auch gefallen