Lecture 15 - Distributed Control

• Spatially distributed systems

• Motivation
• Paper machine application
• Feedback control with regularization
• Optical network application
• Few words on good stuff that was left out

EE392m - Winter 2003 Control Engineering 15-1

 Distributed Array Control
• Sensors and actuators are organized in large arrays distributed
in space.
• Controlling spatial distributions of physical variables
• Problem simplification: the process and the arrays are
uniform in spatial coordinate
• Problems:
– modeling
– identification
– control
Sensors

Actuators

EE392m - Winter 2003 Control Engineering 15-2

 Distributed Control Motivation
• Sensors and actuators are becoming cheaper
– electronics almost free
• Integration density increases
• MEMS sensors and actuators
• Control of spatially distributed systems increasingly common
• Applications:
– paper machines
– fiberoptic networks
– adaptive and active optics
– semiconductor processes
– flow control
– image processing
EE392m - Winter 2003 Control Engineering 15-3
 Paper Machine Process
MD CD
machine
cross
direction
direction

© Honeywell

Scanning gauge
• Control objective: flat profiles in the cross-direction
• The same control technology for different actuator types: flow
uniformity control, thermal control of deformations, and others
EE392m - Winter 2003 Control Engineering
Headbox with Slice Lip CD Actuators

EE392m - Winter 2003 Control Engineering 15-5

© Honeywell
 Profile Control System

© Honeywell

EE392m - Winter 2003 Control Engineering

 Biaxial Plastic Line Control

© Honeywell

EE392m - Winter 2003 Control Engineering

 Model Structure
• Process-independent model structure
me a sure me nt g j
∆Y = G ∆U
Y ∈ ℜ m , U ∈ ℜ n , G ∈ ℜ m ,n a ctua tor s e tpoint

• G - spatial response matrix with

columns gj
CD da ta box/a ctua tor numbe r
• Known parametric form of the spatial
response (noncausal FIR)
g j , k = gϕ ( x k − c j )
• Green Function of the distributed
system

EE392m - Winter 2003 Control Engineering 15-8

 Process
CD
Model Identification
Actuator setpoint
array, U(t)

MD

Measured profile
response, Y(t)

• Extract noncausal FIR model

• Fit parameterized response shape

EE392m - Winter 2003 Control Engineering 15-9

EE392m - Winter 2003 Control Engineering 15-10
 Simple I control
• Compare to Lecture 4, Slide 5 I control

• Step to step update:

Y (t ) = G ⋅ U (t ) + D (t )
U (t ) = U (t − 1) − k [Y (t − 1) − Yd ]
• Closed-loop dynamics
Y = (( z − 1) I + kG ) [kGYd + ( z − 1) D ]
−1

• Steady state: z = 1
Y = Yd , U = G −1 (Yd − D )

EE392m - Winter 2003 Control Engineering 15-11

 Simple I control
. ERROR NORM
Issues with simple I control 4

2
• G not square positive definite
– use GT as a spatial pre-filter 0
0 50 ERROR100
P ROFILE 150 200
0.5
YG (t ) = G T G ⋅ U (t ) + DG (t )
YG = G T Y , DG = G T D
0

-0.5
0 20 40 60 80 100
• For ill-conditioned G get very 20
CONTROL PROFILE

large control, picketing 0

– use regularized inverse

-20
0 20 40 60 80 100
• Slowly growing instability
– control not robust
– regularization helps again
EE392m - Winter 2003 Control Engineering 15-12
 Frequency Domain - Time
• LTI system is a convenient engineering model
• LTI system as an input/output operator
• Causal
• Can be diagonalized by harmonic functions
• For each frequency, the response is defined by amplitude
and phase

LTI
Plant

EE392m - Winter 2003 Control Engineering 15-13

 Frequency Domain - Space
• Linear Spatially Invariant (LSI) system
• LSI system is a convenient engineering model
• LSI system as an input/output operator
• Noncausal
• Can be diagonalized by harmonic functions
• Diagonalization = modal analysis; spatial
modes are harmonic functions LSI
Plant

EE392m - Winter 2003 Control Engineering 15-14

 Control with Regularization
• Add integrator leakage term

∆U (t ) = − K (Y (t − 1) − Yd ) − SU (t − 1)
• Feedback operator K
– spatial loopshaping
• KG ≈ 1 at low spatial frequencies
• KG ≈ 0 at high spatial frequencies
• Smoothing operator S
– regularization
• S ≈ 0 at low spatial frequencies
• S ≈ s0 at high spatial frequencies - regularization

EE392m - Winter 2003 Control Engineering 15-15

 Spatial Frequency Analysis
• Matrix G → convolution operator g (noncausal FIR) →
spatial frequency domain (Fourier) g(v)
• Similarly: K → k(v) and S → s(v)
• Each spatial frequency - mode - evolves independently
g (ν )k (ν ) z − 1 + s(ν )
y (ν ) = yd + d
z − 1 + s(ν ) + g (ν )k (ν ) z − 1 + g (ν )k (ν )
• Steady state
g (ν )k (ν ) s(ν )
y (ν ) = yd + d
s(ν ) + g (ν )k (ν ) s (ν ) + g (ν )k (ν )
k (ν )
u(ν ) = ( yd (ν ) − d (ν ) )
s (ν ) + g (ν )k (ν )
EE392m - Winter 2003 Control Engineering 15-16
 Sample Controller Design
• Spatial domain loopshaping is easy - it is noncausal
• Example controller with regularization
ERROR NORM
PLANT
1
3
2
2
1.5
G 0.5
1 1
0.5
0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 0 1 2 3 0 100 200
ERROR 300
P ROFILE 400 500
FEEDBACK
0.5
0.5
0.4
0.5
K 0.3
0.2
0.5
0.5
0

0.1
0.5
0 -0.5
-5 -4 -3 -2 -1 0 1 2 3 4 5 0 1 2 3 0 20 40
CONTROL 60
P ROFILE 80 100
SMOOTHING
0.03 2
0.02 0.04

S 0.01
0
0.03
0.02
0

0.01
-0.01
-2
-5 -4 -3 -2 -1 0 1 2 3 4 5 0 1 2 3 0 20 40 60 80 100

For more depth and references, see: Gorinevsky, Boyd, Stein, ACC 2003
EE392m - Winter 2003 Control Engineering 15-17
 WDM network equalization
• WDM (Wave Division Multiplexing) networks
– multiple (say 40) independent laser signals with closely space
wavelength packed (multiplexed) into a single fiber
– each wavelength is independently modulated
– in the end the signals are unpacked (de-mux) and demodulated
– increases bandwidth 40 times without laying new fiber

EE392m - Winter 2003 Control Engineering 15-18

 WDM network equalization
• Analog optical amplifiers (EDFA) 5
0

amplify all channels -5

-10

• Attenuation and amplification distort -15

-20

carrier intensity profile -25

-30

• The profile can be flattened through -35

-40
1525 1535 1545 1555 1565
active control

See more detail at:

www126.nortelnetworks.com/news/papers_pdf/electronicast_1030011.pdf

EE392m - Winter 2003 Control Engineering 15-19

 WDM network equalization
• Logarithmic (dB)
attenuation for a
sequence of notch
filters
A = A1 ⋅ K ⋅ AN
N
log A = ∑ log Ak
k =1

N
a ( λ ) = ∑ wkϕ (λ − λ0 − ck ) WDM
k =1

Attenuation gain -
Notch filter shape
control handle

EE392m - Winter 2003 Control Engineering 15-20

 Good stuff that was left out
• Estimation and Kalman filtering
– navigation systems
– data fusion and inferential sensing in fault tolerant systems
• Adaptive control
– adaptive feedforward, noise cancellation, LMS
– industrial processes
– thermostats
– bio-med applications, anesthesia control
– flight control
• System-level logic
• Integrated system/vehicle control

EE392m - Winter 2003 Control Engineering 15-21