Process Modeling and Adaptive Control for a Grinding...
Transcript of Process Modeling and Adaptive Control for a Grinding...
![Page 1: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/1.jpg)
Process Modeling and Adaptive Control for a Grinding System
Hodge Jenkins
![Page 2: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/2.jpg)
Introduction: Grinding Process
Grinding is the final machining step in much of today’s precision component manufacturing.
We must control position, force, and velocity for:• achieving a desired shape
• eliminating surface and sub-surface damage
• minimal surface variation
![Page 3: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/3.jpg)
Typical Grinding Cycle
Infeed
Time
spark
out
roughing
retract
residual errordesired
trajectory
t1 t2
Current grinding
control is primarily
position and velocity
control.
Force control is
needed to prevent
material damage and
to allow accurate
surface following in
fine finishing, and to
provide dimensional
repeatability.
![Page 4: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/4.jpg)
Methodology and Approach
Select & validate a grinding process model
Develop real-time model parameter estimators
• Develop understanding of grinding process variation
• Investigate performance of estimation algorithms
• Multiple sensor inputs
Develop adaptive control scheme
• Cancel grinding process dynamics
• Assess stability
Perform benchmark experiments
• Plunge grinding force control (assess force fidelity)
• Traverse grinding force control (assess surface following)
![Page 5: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/5.jpg)
Theoretical Grinding Models
Chip Thickness
• Based on undeformed chip thickness theory and power relations(Merchant, 1945).
• Difficult to be deterministic because of the large number of cutting points and variation (average properties used).
• Very general form and transportable applicability
• Chip size effect for specific material power
Contact Length• Relation to other rolling bodies
• Hertzian contact (affected by loading)
hmax
A
A
A-A
lc
(Shaw, 1956, Kalker, 1968, Malkin ,1989)
hCr
V
V
a
wheel diameterMAX
w c41 2/
Q Va Cr hl
P g MAXC2
4
![Page 6: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/6.jpg)
Models for Real-time Implementation
Q V A x Af fmaterial removal rate
Q F FN TH( ) Hahn and Lindsay (1971)
Preston (1927)Q K F VP N
Q K F F VP N TH( ) Combined Model
F KV
Va d uN T o
feed
C
e
e
e
e
e e or F
1
2 3 4General form (Konig 1989)
QF VA
k u
N C
C
,
1
Malkin (1989)
q qKF Vw
a w
N Coes (1972)
![Page 7: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/7.jpg)
Adaptive Force Controller
Parameters
FN desired
FT,
FN,V, A
Estimator
Controller
Grinding
Model
Plant
![Page 8: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/8.jpg)
Grinding Force Control Approach
Traditional fixed-gain PID type force control systems are limited because of the grinding process variations.
Adaptive control grinding• Internal diameter grinding with a fixed reference model force controller
with a changing process parameters
• Real-time process variations considered in controlling the normal grinding force.
![Page 9: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/9.jpg)
Experimental Grinding System with Force and Position Sensors
Z AxisServo Motor
Force Transducer
Tool End
PC
Controller
workpiece
Y Axis
Servo Motor
Motor Control
PMAC
X Axis
Servo Motor
position sensor
sampled at 2.26 kHz
![Page 10: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/10.jpg)
System Dynamic Identification:Shaker Tests
![Page 11: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/11.jpg)
System Close-up
![Page 12: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/12.jpg)
Detailed View of Test ApparatusPlunge Grinding
Grinder
Part
Wheel
xmeasured
xsurf FN
FT1
FT2
![Page 13: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/13.jpg)
Force Controller
G (s)FC
G (s)PC
G (s)TWP
Force
Error
_
+
Commanded
Position
Actual
Position
Fd
Desired
Force
FN
E XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
Desired Result: Stable Force Control
Increased Accuracy and Productivity in Fine Finishing
![Page 14: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/14.jpg)
Position Controller Model
+ E U
-
LdX
C(s) 1/sM(s)
compensator motor
Xlead screw
X
encoder
eX
1
.
![Page 15: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/15.jpg)
UsSystem Identification Techniques
y(t) G(q)u(t) H(q)e(t)
e(t) H1(q) y(t) G(q)u(t) error
ˆ G N , ˆ H N min e2(t)
t 1
N
V s
U sL M s
397
s 2.95
mm / s
volt
The resulting first order system L M(s) is given as
Use a least squares fit for parameters of velocity
as a first order transfer function of control voltage input
![Page 16: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/16.jpg)
Anti-Aliasing Filter for sensor input Cut-off frequency is 500 Hz
Butterworth analog filter
![Page 17: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/17.jpg)
Identified Model of Position Control Loop
G sX s
X s sPC
e
d
( )( )
( )
203
224
Actual and Model Displacement Corresponding Measured Force
![Page 18: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/18.jpg)
Dynamic System Model
Bs ~ 0
Ks
Grinding
Process
Grinding
Wheel
Servo
EncoderSurface
M
xe
xf
_
F N
K PVQ
1A
1s
+ xf
F TH
_
+xf
.
F N
ex
F FA
K VxN TH
P
fEquivalent
Damping:
x s
x s
K K V A
s K K V A
f
e
P S
P S
( )
( )
Position
Transfer
Function:
Ks
![Page 19: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/19.jpg)
Adaptive Force Controller with Real-Time Estimation
G (s)FC
G (s)PC
G (s)TWP
Force
Error
_
+
Commanded
Position
Encoder
Position
Fd
Desired
Force
FN
E XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
Process
Model
Parameter
Estimator
Adaptation
Control
Other
Sensed
ParametersEstimated
Process
Parameters
Controller
Gains
Desired Result: Stable Force Control, Increased Fidelity, Productivity,
Greater Surface Following for Fine Finishing
![Page 20: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/20.jpg)
Fixed PI Force Controller
G s K sK
K sFC prop
prop
( ) int 1
G (s)FC
G (s)PC
G (s)TWP
Force
Error
_
+
Commanded
Position
Actual
Position
Fd
Desired
Force
FN
E XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
![Page 21: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/21.jpg)
Fixed PI Force Controller: Root Locus
Zero at -30 Zero at -20
![Page 22: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/22.jpg)
-6
-5
-4
-3
-2
-1
0
1
10-5 10-4 10-3 10-2 10-1 100 101 102 103
Gain, Kprop
z-d
om
ain
roo
t v
alu
e
Fixed Gain PI Control Digital Design(Ignoring Grinding Process Model)
KPROP
(for KI/KPROP=1,50,100,150,
200,250)
Magnitude r
oot
(z) stable
unstable
G z G z G z Kz
z
K
zFC PC TWP PROP
s( ) ( ) ( )( ) .
.
1 1
1
0 3543
0 6091
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 5 10 15 20 25 30 35
Rise Time vs Gain
Kint/Kprop
90
% R
ise
Tim
e (
s)90%
Ris
e T
ime (
s)
![Page 23: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/23.jpg)
Fixed-Gain Force Loop Block Diagram
G (s)FC
G (s)PC
G (s)TWP
Force
Error
G (s)FS
_
+
Commanded
Position
Actual
Position
Actual
Force
Fd
Desired
Force
Measured
Force
Fa
Fm
e XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
Force Sensor
![Page 24: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/24.jpg)
Fixed PI Force Controller: Step Force input
-1
0
1
2
3
4
5
6
0 0.5 1 1.5 2 2.5 3 3.5 4
Norm
al F
orc
e (N
)
time(s)
SimulationExperimental Results
![Page 25: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/25.jpg)
Position Control Experiment: Identified Parametric Model vs. Actual Data
x s
x s
f
e
( )
( ) s + ; where =
K K V
A
P S
Actual and Model Displacements
FN
xf
![Page 26: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/26.jpg)
Add Grinding Process To System
Grinding
Process
Process
Model
Parameter
Estimator
+-
Ra
Rd
RmSensors
Controller Tool
Dynamics
Adaptation
Control
Process
Variables
Rm=[F, z, v,..]T
![Page 27: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/27.jpg)
Force Control Experiment:Typical Force and Displacement
0
1 0
2 0
3 0
0 1 2 3 4 5 6
For
ce
(N)
Time (s )
-30 0
-20 0
-10 0
0
0 1 2 3 4 5 6
Dis
p (m
icro
n)
Time (s )
Normal
Force
Measured
Infeed
Displacement
KAx
V F FP
f
N TH( )
tanx cons tf
FN
xf
Time (s)
![Page 28: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/28.jpg)
Force Control Experiment:Experimental Results
KAx
V F FP
f
N TH
( )
F FN TH
B B B BB B BB B
B B
J
J
J
J
J
J
JJ
JJJ H
H
H
H
H
HH
H
H
H
H
H
H
H
0
500
1000
1500
2000
2500
3000
0 5 10 15 20 25 30 35 40
Normal Force (N)
![Page 29: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/29.jpg)
Multi-Sensor Data May Be Used toImprove Model Parameter Estimation
Multi-sensor approaches have yielded encouraging results for improving on-line estimation of wear.
• Several indirect measurement sources of wear have been successful, indicating that successful integration of several sensors may be more reliable than any single sensor method.
Candidate Methods for Real-time Use of Multiple Sensors• Basis Functions
Require apriori basis structures
• Neural Networks Require bounded training sets
Training times can be long
• Recursive Methods
Recursive Least Square– Forgetting Factors
Kalman Filtering– Windowing
![Page 30: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/30.jpg)
Adaptive Force Controller with Real-Time Estimation
G (s)FC
G (s)PC
G (s)TWP
Force
Error
_
+
Commanded
Position
Encoder
Position
Fd
Desired
Force
FN
E XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
Process
Model
Parameter
Estimator
Adaptation
Control
Other
Sensed
ParametersEstimated
Process
Parameters
Controller
Gains
Desired Result: Stable Force Control, Increased Fidelity, Productivity,
Greater Surface Following for Fine Finishing
![Page 31: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/31.jpg)
Windowing of Sensor Data for Variable Parameter Estimation
Collect
data
Estimate
parameters
Determine
Gains
Time
T1
Sensor Data
Collect
data
T2
Estimate
parameters
Determine
Gains
Collect
data
T2
![Page 32: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/32.jpg)
PMAC (DSP)
Dual-Ported
RAM
Motors
Encoders
Force
SensorsDisplacement
Sensors
PC
-Bu
s
Spindle
Tachometer
PC
Windowed
Data
Controller
Gains
(Servo-control,
data acquisition)
(Estimation calculations,
adaptive controller design)
Dual Processing Structure and Information Flow
![Page 33: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/33.jpg)
Actual Grinding and Model Resultswithout Estimation Improvement
xf
Time (s)
![Page 34: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/34.jpg)
Estimated and Actual Grinding Wheel Displacements
Windowed Kalman
Filter (20 pt. window)RLS with 0.9
Forgetting Factor
RLS is selected for real-time estimation for computational efficiency.
Both Methods yield similar improvements.
xf
Time (s)
xf
Time (s)
![Page 35: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/35.jpg)
Window Size Effect on Kalman FilterEstimation: Model Error
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1.000
5 10 20 25 40 50 100
Sample WIndow Size
Rela
tive
Err
or
1020a
1020b
A2
O1
4142
718
8119
Relative
Error
P/Pmax
Window
![Page 36: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/36.jpg)
Effect of Sensor Inputs on Estimation and Model Error
0.500
0.600
0.700
0.800
0.900
1.000
1.100
1.200
1.300FT
FN
FT
2
FT
,FN
FT
2,F
N2
FT
,FT2
FT
,FN
2
FT
,W
FT
,FT2
,
FN
,FN
2
FT
,FT2
,
FN
,FN
2,W
Sensors Input
Rela
tive
Mod
el E
rror
A2
8119
AVG
Relative
Error
P/PFT-sensor
Sensors used for estimation input
![Page 37: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/37.jpg)
Adaptive Force Controller with Real-Time Estimation
G (s)FC
G (s)PC
G (s)TWP
Force
Error
_
+
Commanded
Position
Encoder
Position
Fd
Desired
Force
FN
E XdX
Force
Control
Position
Loop
Tool-work-process
Impedance
Process
Model
Parameter
Estimator
Adaptation
Control
Other
Sensed
ParametersEstimated
Process
Parameters
Controller
Gains
Desired Result: Stable Force Control, Increased Fidelity, Productivity,
Greater Surface Following for Fine Finishing
![Page 38: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/38.jpg)
Adaptive Control Strategy
Grinding variation effects overall system gain and has slow dynamics, and can lead to oscillations.
Controller can be unstable for some fixed-gains• Reduce gain before on-set of oscillation
Control strategies• Pole-zero cancellation of grind process via PI-controller
(Windowed update of gains)
K K V AP S
Grinding Process
PI Controller
G s K sK
K sFC PROP
I
PROP
( )1
G sF s
x sK
s
sTPW
N
e
S( )( )
( )
![Page 39: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/39.jpg)
Force Controller Development
G sX s
X s sPC
D
( )( )
( )
203
224
G s K sK
K sFC PROP
I
PROP
( )1
Thus pole-zero cancellation via PI Control
K
K
I
PROP
G sF s
x sK
s
sTPW
N
e
S( )( )
( )
K K V AP S
G s G s G sF s
E sK
s
s
K s
s sFC PC TWP
N
F
SPROP( ) ( ) ( )
( )
( )
( )
( )203
224
G s G s G sF s
E s
K K
sFC PC TWP
N
F
S PROP( ) ( ) ( )( )
( ) ( )
203
224
Position loop
Tool-work & Process
PI Force Control
Combined Plant
Therefore let Typically 224
![Page 40: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/40.jpg)
Digital Control Implementation
Using a digital controller
G G G zK f
z
z f z CFC PC TWP
PROP
o
( )( )
( )
2
11
1
1
f1( ) e- TC eo
T224
Let f1
1
1( ) for cancellation of grinding process
(stable zero near unity)
![Page 41: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/41.jpg)
Benchmark Tests: Combined Real-Time Estimation and Adaptive Control
Two tests• Vertical plunge grinding
Control grinding normal force
Assess force fidelity (mean, range, and variance)
• Traverse grinding
Constant horizontal force on “straight” surface
Control vertical force
Assess surface following ability
• Compare adaptive pole-zero cancellation controlwith fixed-gain PI force controller
![Page 42: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/42.jpg)
Surface Profile Measurement
![Page 43: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/43.jpg)
Surface Grinding Profiles
Fixed-Gain Force Control Adaptive Force Control
Xf
( m)
Length (mm) Length (mm)
Xf
( m)
![Page 44: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/44.jpg)
Plunge Grinding Normal Forces
Fixed-Gain Control Adaptive Control
FN
Time (s)
FN
Time (s)
![Page 45: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/45.jpg)
Normal Grinding Force (Plunge Grind)Experimental Force Statistics
Case Material Ave. Std. Dev. Max. Min. Range Rise
Force Time
(N) (sec)
Fixed Gain 4142 24.34 1.060 27.51 20.43 7.09 0.145
Adaptive 4142 25.18 0.962 27.88 21.67 6.21 0.133
Fixed Gain A2 26.06 1.017 29.48 22.94 6.54 0.142
Adaptive A2 24.78 0.975 27.48 21.94 5.54 0.142
Fixed Gain O1 25.23 0.90 27.72 21.11 6.62 0.164
Adaptive O1 25.52 0.800 27.66 22.41 5.25 0.149
Fixed Gain 8119 25.57 2.193 31.11 20.25 10.86 0.135
Adaptive 8119 25.92 1.979 30.97 21.17 9.80 0.124
![Page 46: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/46.jpg)
Conclusions
• Grinding is time-varying process, and may be thought of as a damping equivalent.
• A two-parameter grinding model is a valid process representation
FTH can be thought of as constant for a particular setup– Function of the wheel grit and part elastic modulus
KP is best modeled in real-time– Average values have a linear correlation to the material hardness
• Grinding process variations are generally at significantly lower frequencies than machine vibrations.
Low-pass filtering may be used to separate the process dynamics from the machine vibrations
• Applying recursive estimation techniques to multiple sensor data provides an improved estimate of the model coefficient.
RLS-FF and the Kalman filter methods yield similar results
Both produce a time-varying “optimal” filter for KP
![Page 47: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/47.jpg)
Conclusions (continued)
• There is a trade-off between the timeliness of sensor data and a meaningful statistical representation of process, leading to an optimum window of sampled data (or forgetting factor).
• Additional correlated sensors have diminishing improvements
An additional sensor data for filtering KP has an order of magnitude improvement on model error covariance.
An average a small (7%) improvement was seen by adding correlated additional sensor.
Industrial application should not require extensive sensoring
• Adaptive pole-zero cancellation of the grinding process provides stable control under process variations.
This yielded greater force control fidelity.
An improved ability to track surfaces in fine finishing has been observed.
![Page 48: Process Modeling and Adaptive Control for a Grinding Systemfaculty.mercer.edu/jenkins_he/documents/GrindingSystemModel.pdf · Introduction: Grinding Process Grinding is the final](https://reader031.fdocuments.us/reader031/viewer/2022030415/5aa1add97f8b9a1f6d8c40f7/html5/thumbnails/48.jpg)
The End
Thank you