Advanced Multivariable Fiber Orientation Control and Twist ... ppt.pdfFO Btm 0.66 0.30 54% 2.33 1.85...
Transcript of Advanced Multivariable Fiber Orientation Control and Twist ... ppt.pdfFO Btm 0.66 0.30 54% 2.33 1.85...
Advanced Multivariable Fiber Orientation Control and Twist
OptimizationDanlei Chu, Ph.D., P.Eng.Cristian Gheorghe, P.Eng.Cristian Gheorghe, P.Eng.Johan Backstrom, P.Eng.
PaperCon 2011 Page 2412
Outline
• Introduction to Fiber Orientation (FO)t oduct o to be O e tat o ( O)
- FO angle, curl and twist
• Process Model• Process Model
- Band pass characteristics in the spatial frequency domain
• Controller Design
- Actuator edge padding (AEP) and gain retune
• Mill trial results
• Conclusions
PaperCon 2011 Page 2413
Introduction: Fiber Orientation
• Fiber orientation
- Fiber orientation (FO) refers to the dominant alignment
direction for fibers in a paper web. It can be expressed by a
fiber orientation angle and a fiber ratio.
α: Fiber Orientation Angleg
Fiber Ratio:
)( AxisShort )( Axis Long
sl
A cut and dry test
PaperCon 2011 Page 2414
Introduction: Practical Issues
• Several paper quality properties are highly related to fiber orientation of whichorientation, of which
- The web strength, and- The dimensional stability are the most importantThe dimensional stability are the most important
(The ability of paper samples to retain their shape when subjected to varying degrees of temperature, moisture, stress, etc)
• Poor FO can ca se m ltiple practical iss es incl ding• Poor FO can cause multiple practical issues, including
- Paper jam in sheet-fed device,
- Mis-register in color printing,
- Twist in multilayer board, and
- Weakened corrugated containerboard.
PaperCon 2011 Page 2415
Introduction: Curl & Twist
• Curl and twist are two measurements to indicate the deflection of a paper sheet
Rapid drying
of a paper sheet
xykykxkyxd xyyx 22),(non-uniform FO properties
• x: CD coordinate
non uniform FO properties
• y: MD coordinate• kx : CD curl coefficient• ky : MD curl coefficienty• kxy: Twist coefficient (diagonal curl coefficient)• d: deflection index
PaperCon 2011 Page 2416
Introduction: Curl & Twist1
150600
0.5
d(x,
y)
1
0
50
100
0
20
40
Cross DirectionMachine Direction
500
0.5
1
d(x,
y)
MD Curl
10
2040
6080
100120
010
2030
4050
Cross DirectionMachine Direction
CD Curl
-0.5
0
0.5
d(x,
y)
btmtop
MDCDbtmtopfTwist
),,,,(
010
2030
4050
0
50
100
150-1
C Di tiMachine Direction
Twist
p Cross DirectionMachine Direction
PaperCon 2011 Page 2417
Process Model: Spatial Frequency Characteristics
• Compares CDFO Processes to Standard CD Processes
0
0.5
1Spatial Impulse Response
eigt
h (g
sm)
0
1Spatial Impulse Response
(deg
)
-800 -600 -400 -200 0 200 400 600 800-0.5
0
CD Coordinate (mm)
Dry
We
Gain Spectrum
)
-800 -600 -400 -200 0 200 400 600 800-1
CD coordinate (mm)
FO
Gain Spectrum
0
0.1
0.2
Dry
Wei
gth
(gsm
)0
0.1
0.2
FO (d
eg)
Standard CD Processes:• Headbox Slice Lip Dry Weight
CD Fiber Orientation Processes• Headbox Slice Lip Fiber Orientation Angle
0 1 2 3 4 5 6 7
x 10-3
0
Spatial Frequency (1/mm)0 1 2 3 4 5 6 7
x 10-3
0
Spatial Frequency (1/mm)
• Low pass characteristics in spatial frequency• band pass characteristics in spatial frequency
PaperCon 2011 Page 2418
Process Model: Spatial Model
• Slice Lip FO Angle Model
Th I l RTh I l R
• Process Gain
Defines the peak value of th ti l i l
0.1X: 68.45Y: 0.09996
The Impulse Response
0.1*20% = 0.02 deg
0.1X: 68.45Y: 0.09996
The Impulse Response
0.1*20% = 0.02 deg
the spatial impulse responses of a CDFO process
(gain = 0.1 deg/mil)0
0.05
Deg
)
X: 502.5Y: 0.01949
X: -502.50
0.05
Deg
)
X: 502.5Y: 0.01949
X: -502.5 (g g )
• Response Width
Defines the location where -0.05
0
FO (D
X: 502.5Y: -0.01949
-0.05
0
FO (D
X: 502.5Y: -0.01949
the impulse response drops to 20% of its peak value(width = 1000 mm)
-1000 -800 -600 -400 -200 0 200 400 600 800 1000
-0.1 Response Width = 1000 mm
-1000 -800 -600 -400 -200 0 200 400 600 800 1000
-0.1 Response Width = 1000 mm
CD Coordinates (mm)CD Coordinates (mm)
PaperCon 2011 Page 2419
Process Model: Identification
• Model Test Results on a Three-layer Board Machine
0
0.5
Deg
Top FO Angle
Pl
0.5
1
eg
Bottom FO Angle
PlantModel
0 50 100 150 200 250 300 350 400-1
-0.5
CD Bins
D
1000Top Slice Bumps
PlantModel
0 50 100 150 200 250 300 350 400-0.5
0
CD Bins
De
Bottom Slice Bumps
1000
0
1000
mic
ron
0
1000
mic
ron
Model identification results on the top headbox• Process Gain = ‐0.0025 deg/micron
Model identification results on the bottom headbox• Process Gain = 0.002 deg/micron
0 5 10 15 20 25 30 35 40-1000
Actuator Zones0 5 10 15 20 25 30 35 40
-1000
Actuator Zones
• Response Width = 2000 mm • Response Width = 2000 mm
PaperCon 2011 Page 2420
Controller Design
• The Schematic of a CDFO Control System Capture Nonlinearity
Capture Severe Edge Effect
PaperCon 2011 Page 2421
Control: Actuator Edge Padding
• Actuator Edge Padding Captures Severe Edge Effects
-5
0
5
Set
poin
t
Flat padding
physicalvirtual
uGzhy •1 2 3 4 5 6 7 8 9 10
-5
Actuator zones
10
t
Linear padding
augauguGzhy•
uuu
u
lpad
d
•
1 2 3 4 5 6 7 8 9 10-10
0
Actuator zones
Set
poin
t uuuuhpad
aug mod
•
-5
0
5
Set
poin
t
Reflection padding
uGzhy aug mod)( •
G’1 2 3 4 5 6 7 8 9 10Actuator zones
G’
PaperCon 2011 Page 2422
MPC: Gain Retune
• Gain Retune Captures the nonlinearity of Fiber Orientation ProcessProcess
kqq
lamo
eqq
gg
2)(
0
1
• q: the difference between the Jet Speed and the Wire Speed;
• q0: The crossover point where the paper machine is operated in neither “rush” nor “drag”
• glam: Laminar Gain (Model Parameter)
• k: Jet Turbulence (Model Parameter)
PaperCon 2011 Page 2423
Control: MPC
• The Fiber Orientation CD-MPC Controller
||)(||||(||min 2
1
2
)( 21kUYikY Q
H
iQspkU
p
)()(||)(|| 42
1)(
3kUQkUUkU T
Qtgt
i
Subject toSubject to,
)1()( kCUbkUA
)()()( kUzhGkY
Actuator Physical Limits
Fiber Orientation P M d l)()()( kUzhGkY Process Model
Besides FO angles and twist profiles, other CD measurement, (dry weight, g p , , ( y g ,moisture, caliper) can be optionally added to the above MPC controller.
PaperCon 2011 Page 2424
Mill Test Results
• Mill Configuration - A multi-ply carton board machine in Finland.p y- Control Scenario
ControllerType
Actuators Measurements(420 CD Bins)Type (420 CD Bins)
Top Headbox Slice Lip(42 zones)
Top Fiber Orientation Angle
MPC ( ) g
Bottom Fiber Orientation AngleBottom Headbox Slice Lipp
(42 zones)Twist Proxy
(Top – Bottom Angle)
PaperCon 2011 Page 2425
Mill Trial Results - Profiles
4
6
8
e (d
eg)
The profiles of top FO angle
2
4
e (d
eg)
The profiles of bottom FO angle
2
4
6
e (d
eg)
The profiles of twist
0 2000 4000 6000-2
0
2
FO A
ngle
0 2000 4000 6000
-4
-2
0
FO A
ngle
0 2000 4000 6000
-4
-2
0
FO A
ngle
0 2000 4000 6000
CD Position (mm)0 2000 4000 6000
CD Position (mm)0 2000 4000 6000
CD Position (mm)
500
1000
The setpoints of top slice lip
rons
)
500
1000The setpoints of bottom slice lip
rons
)
1000
-500
0
Set
poin
t (M
icr
1000
-500
0
Set
poin
t (M
icr
Before MPCAfter MPC
Fiber Orientation Angles and Twist Proxy Profiles• All controllable disturbances are effectively removed by the MPC controller
0 10 20 30 40-1000
Actuator Zones
0 10 20 30 40
-1000
Actuator Zones
• All controllable disturbances are effectively removed by the MPC controller.
PaperCon 2011 Page 2426
Mill Trial Results – Power Spectrum
1The Power Spectrum of FO Angle Top
0 25The Power Spectrum of FO Angle Btm
0 8The Power Spectrum of Twist
0 6
0.8
1
r (de
g)
0 15
0.2
0.25
r (de
g)
0.5
0.6
0.7
0.8
r (de
g)
Before MPCAfter MPC
0.2
0.4
0.6
Abs
olut
e po
wer
2Xa-3db Xc
0.05
0.1
0.15
Abs
olut
e po
wer
2Xa-3db Xc
0.2
0.3
0.4
Abs
olut
e po
wer
2Xa
inf 694 347 2310
Spatial wavelength (mm)
inf 694 347 231
0
Spatial wavelength (mm)
inf 694 347 231
0
0.1
Spatial wavelength (mm)
Fiber Orientation Angles and Twist Proxy Profiles• All controllable disturbances are effectively removed by the MPC controller.
PaperCon 2011 Page 2427
Mill Test Results – Average & 2-Sigma
• Comparisons Before and After the Fiber Orientation Control
Statistics Average Two-Sigma
Before After Improve Before After ImproveMeas. Before After Improvement
Before After Improvement
FO Top 2.72 1.26 54% 2.26 1.54 32%
FO Btm 0.66 0.30 54% 2.33 1.85 20%
Twist 1.65 0.96 42% 2.50 2.20 12%
Both the averages and the 2-sigma spreads (two times of the standard derivation) ofBoth the averages and the 2 sigma spreads (two times of the standard derivation) of FO angle profiles and twist profiles are improved significantly.
PaperCon 2011 Page 2428
Mill Test Results – Lab Measurements
• Comparisons Before and After the Fiber Orientation Control
Here, the twist average values are measured in the lab.
PaperCon 2011 Page 2429
Conclusions
• The CDFO spatial model has only two parameters. The model is very
intuitive and easy to identify.
• Actuator edge padding algorithm captures the severe edge effects of the
fiber orientation process.
• Gain retune algorithm captures the nonlinearity of the fiber orientation
process with respect to Jet/Wire ratio and wire speed.
• Automatic controller retune algorithms make the CDFO controller adaptive
to the processes variation
• The resulting MPC controller is able to significantly reduce the variance of
twist and CD fiber orientation angle profiles.
PaperCon 2011 Page 2430