Control Law Design For Switched Repetitive Processes With A Metal Rolling Example
Control Law Design For Switched Repetitive Processes With A Metal Rolling Example
TuB02.4
I. I NTRODUCTION
The unique characteristic of a repetitive process (also
termed a multipass process in the early literature) can be
illustrated by considering machining operations where the
material or workpiece involved is processed by a series of
sweeps, or passes, of the processing tool. Assuming the
pass length < + to be constant, the output vector,
or pass profile, yk (p), p = 0, 1, . . . , ( 1), (p being the
independent spatial or temporal variable), generated on pass
k acts as a forcing function on, and hence contributes
to, the dynamics of the new pass profile yk+1 (p), p =
0, 1, . . . , ( 1), k = 0, 1, . . . . This, in turn, leads to the
unique control problem for these processes in that the output
sequence of pass profiles generated can contain oscillations
that increase in amplitude in the pass-to-pass direction, i.e.
in the collection of pass profile vectors {yk }k .
Industrial examples of such processes include long-wall
coal-cutting and metal rolling, see the original papers cited
in, for example, [1] for further details. A number of socalled algorithmic examples also exist where adopting a
repetitive process setting for analysis has clear advantages
over alternative approaches to systems related analysis. These
J. Bochniak and K. Galkowski are with Faculty of Electrical Engineering, Computer Science and Telecommunications, University of Zielona Gora, Podgorna 50, 65-246 Zielona Gora, Poland
{kgalkows,jbochnia}@uz.zgora.pl
K. Galkowski is currently the Gerhard Mercator Guest Professor in
University of Wuppertal, Germany
E. Rogers is with Department of Electronics and Computer Science,
University of Southampton, Southampton SO 17 1BJ, United Kingdom
etar@ecs.soton.ac.uk
A. Kummert is with Faculty of Electrical, Information and Media Engineering, University of Wuppertal, Rainer Gruenter Str. 21, 42119 Wuppertal,
Germany kummert@uni-wuppertal.de
The most basic state-space model for discrete linear repetitive processes over p = 0, 1, . . . , ( 1), k = 0, 1, . . ., is
700
TuB02.4
given by
xk+1 (p + 1) = Axk+1 (p) + Buk+1 (p) + B0 yk (p)
yk+1 (p) = Cxk+1 (p) + Duk+1 (p) + D0 yk (p)
(1)
both the existence the limit pass profile and that this is stable
in the usual 1D sense.
Numerous sets of necessary and sufficient conditions for
stability along the pass of processes described by (1) are
known [1] but these have not proved to be a suitable basis
on which to develop control law design algorithms. This has
led to the use of a Lyapunov function based approach which
results in a sufficient condition for stability along the pass
which can be computed using LMI methods. Moreover, there
is a natural extension to solve the basic control law design
problem and this fact offsets the conservativeness resulting
from the fact this approach can only give sufficient, but not
necessary stability conditions.
The starting point for the work reported here is the
following result from [8].
Theorem 1: A discrete linear repetitive process described
by (1) is stable along the pass if there exist matrices W1 > 0
and W2 > 0 such that
T W W
where
< 0
(2)
A B0
.
C D0
Previous work, see, for example [8], has used an LMI
setting to design control laws of the following form over
p = 0, 1, . . . , ( 1), k = 0, 1, . . . for processes described
by (1)
W = diag (W1 , W2 ), =
uk+1 (p) =
=
(3)
< 0
(4)
L1 G2
<0
L22 G2
T
V G2 G2
(5)
W
GT1 L1T
2
GT2 LT1
L12 G1
G1 GT1
GT2 L2T
2
701
TuB02.4
III. PASS - TO - PASS S WITCHING
where
for l = 0, 3, . . .
for l = 2, 5, . . .
(6)
C
x
(p)
+
D
u
(p)
+
D
y
(p)
1
l+1
1
l+1
01
l
for l = 0, 3, . . .
C
xl+1 (p) + D3 ul+1 (p) + D03 yl (p)
for l = 2, 5, . . .
x3k+1 (p)
u3k+1 (p)
Xk+1 (p) = x3k+2 (p) , Uk+1 (p) = u3k+2 (p) ,
x3k+3 (p)
u3k+3 (p)
Yk (p) = y3k (p),
Yk+1 (p) = y3k+3 (p)
b = D03 D02 C1 D03 C2 C3 ,
C
b = D03 D02 D1 D03 D2 D3 ,
D
B01
b0 = B02 D01 , D
b 0 = D03 D02 D01 .
B
B03 D02 D01
b .
b D
b0 , = D
C
b and
b are termed the augmented process and
The matrix
the augmented input matrices, respectively.
Theorem 2: A discrete linear repetitive process described
by (7) is stable along the pass if there exist matrices W1 > 0
and W2 > 0 such that
bT W
b W
< 0
(9)
where
for k = 0, 1, . . . Note that the variable l is not equal to k.
The variable l is used to denote the pass number and k to
indicate each of three successive (i.e. {3k+1, 3k+2, 3k+3})
pass numbers.
Then the equivalent model of the form (1) for processes
described by (6) is
(
b k+1 (p) + BU
b k+1 (p) + B
b0 Yk (p)
Xk+1 (p + 1) = AX
b
b
b 0 Yk (p)
Yk+1 (p) = CXk+1 (p) + DUk+1 (p) + D
A1
0
0
b = B02 C1
A2
0 ,
A
B03 D02 C1 B03 C2 A3
B1
0
0
b = B02 D1
B2
0 ,
B
B03 D02 D1 B03 D2 B3
(7)
702
b 1 Xk+1 (p) +
Uk+1 (p) = K
b1 =
K
b 2 Yk (p)
K
K11
+ D1 K11 )
3
K2 (D02 + D2 K22 )(C1 + D1 K11 )
0
0
K12
0 ,
3
2
K2 (C2 + D2 K1 ) K13
K22 (C1
K21
b2 =
.
K22 (D01 + D1 K21 )
K
K23 (D02 + D2 K22 )(D01 + D1 K21 )
(11)
TuB02.4
u 3k+1(t)
u 3k+2(t)
spring
11
A12 =
"
0
B02
0
0
A22 =
"
C1
0
0
0
A13 =
A23 =
"
"
0
A2
0
0
0
0
A3
C3
0
0
B03
D03
0
C2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
B03
D03
0
D02
0
0
0
0
0
0
0
0
0
0
0
0
0
0
B01
0
0
0
0
B02
0
0
0
0
0
D01
0
0
0
0
0
D02
0
0
=
,B
#
"
12
21 =
,B
22 =
,B
"
31 =
, B
"
"
32 =
,B
"
0
B2
0
0
0
0
B3
D3
0
D2
0
0
D1
0
0
0
0
0
0
0
0
D2
0
0
0
0
B3
D3
0
0
B3
D3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1 = diag K 1 , K 2 , K 3 , K 1 ,
K
1
1
1
2
2 = diag K22 , K23 , K22 , K22 .
K
B1
0
0
0
0
0
0
0
0
B2
0
0
0
0
0
D1
0
0
0
0
0
D2
0
0
spring
zero compression
separation
y 3k+1(t)
Fig. 1.
y 3k+2(t)
y 3k+3(t)
where
L,
1N
12 Z = A12 Z + B
1 V = AV + B
2 ,
2
2
2
1
1
2 L,
31 N,
2 V = A2 V + B
3 Z = A3 Z + B
2N
23 Z = A23 Z + B
3
(16)
and
X = diag (X1 , X2 ) ,
V = diag (V1 , V2 , V3 , V4 ) , Z = diag (Z1 , Z2 , Z1 , Z1 ) ,
= diag (L1 , L2 , L3 , L4 ) , N
= diag (N1 , N2 , N1 , N1 ) .
L
If this condition holds, the control law matrices are given by
1 = L
V 1 = diag K 1 , K 2 , K 3 , K 1
K
1
1
1
2
(17)
2 = N
Z 1 = diag K22 , K23 , K22 , K22
K
Proof: This follows immediately from interpreting Theorem 2 in terms of the controlled process state-space model
in this particular case and use of Lemma 1. The details are
omitted here due to space limitations.
,
#
M3
13
(14)
0
B2
0
0
B1
0
0
0
z3k+3(t)
zero compression
separation
metal strip
where
22 K
1,
K
1 , 12 = A12 + B
21 K
2 , 22 = A22 + B
1 = A + B
1
1
1
2
2
2
3 = A3 + B3 K2 , 3 = A3 + B3 K2 ,
A1
0
0
0
M2
spring
y 3k(t)
"
zero compression
separation
A =
M3
z3k+2(t)
M1
b0new = B
b0 + B
bK
b 2,
B
b 0new = D
b0 + D
bK
b2
D
K
=
K1 K2
b
b D
b0
D
C
with
M2
z3k+1(t)
where
bnew = A
b+B
bK
b1,
A
bnew = C
b+D
bK
b1,
C
u 3k+3(t)
M1
1 V
12 Z
X
13 Z
Z T 1T
0
Z Z T
23 Z
3
T 1T
T
2T
T
2
Z 2
Z 3
Z Z
2 V
V T
V T T1
0
V T 2T
X
V
2
< 0 (15)
The first results on metal rolling process have been developed in [9], [10] and the results here follow on these
previous works, but in the case of switched process and for
the minimal realization of the state-space representation.
Consider a multi-roll roll system (Fig. 1) consists of three
separate pairs of rolls which are controlled by separate input
signals, i.e. different rolling forces. The deformation of the
workpiece takes place between these pairs of rolls with
parallel axes revolving in opposite directions. The metal
strip to be rolled to a pre-specified thickness (also termed
the gauge or shape) through a series of rolls for successive
reductions.
In practice, a number of models of this process can be
developed depending on the assumptions made about the
underlying dynamics and the particular mode of operation
under consideration. The particular task is to develop a
simplified (but practically feasible) model relating the gauge
on the passes through the rolls. The current pass is denoted
by y3k+3 (t), the previous passes by y3k+2 (t), y3k+1 (t) and
y3k (t). The other process variables and physical constants are
defined as follows: u3k+1 (t), u3k+2 (t) and u3k+3 (t) are the
forces developed by the motors, M1 , M2 and M3 are the
lumped masses of the roll-gap adjusting mechanisms, 11 ,
12 and 13 are the stiffnesses of the adjustment mechanism
springs, 2 is the hardness of the metal strip.
703
TuB02.4
We can written the model of the multi-roll rolling machine
with three pairs of rolls in the form of three differentialdifference equations
1i 2
,
Mi (1i + 2 )
b2i =
2
,
1i + 2
b0i =
1i 2
,
Mi (1i + 2 )
c0i =
1i
.
Mi (1i + 2 )
Applying approximation to (18) by backward Euler discretization method with sampling period T yields the
discrete(-discrete) state-space model (6), where for i =
1, 2, 3,
1
T
T
c0i T
1
, Bi = 1+a0i T 2
,
Ai = 1+a0i T 2
a0i T 1
1
B0i =
Di =
T
1
D0i =
Ci =
1
1+a0i T 2
b2i b0i T 2
1+a0i T 2 ,
1 T
with
y3k+1 (p) + b21 y3k (p),
3k+2 (p) =
3k+3 (p) =
V. C ONCLUSIONS
A1 B01 B1
C1 D01 D1
A2 B02 B2
C2 D02 D2
A3 B03 B3
C3 D03
D3
972.222 97.222
7.936 0.278
277.778 972.222 79.365 2.778
981.595 98.159
6.902 0.184
= 103
184.049 981.595 69.018 1.840 .
981.595 98.159 631.902 0.184
985.401 98.540
6.488 0.146
704
TuB02.4
l = 0,...,21;
p = 0,...,100
PASS
PROFILE
5
0
5
10
15
20
100
80
0
60
5
40
10
15
20
20
pass number
PASS
PROFILE
l = 0,...,21;
p = 0,...,100
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
100
80
0
60
5
40
10
15
20
20
pass number
Fig. 2.
R EFERENCES
[1] E. Rogers and D.H. Owens, Stability Analysis for Linear Repetitive
Processes, ser. Lecture Notes in Control and Information Sciences.
Berlin, Germany: Springer-Verlag, 1992, vol. 175.
[2] K.L. Moore, Y. Chen, and V. Bahl, Monotonically convergent iterative
learning control for linear discrete-time systems, Automatica, vol. 41,
no. 9, pp. 15291537, 2005.
[3] P.D. Roberts, Two-dimensional analysis of an iterative nonlinear optimal control algorithm, IEEE Transactions on Circuits and Systems I:
Fundamental Theory and Applications, vol. 49, no. 6, pp. 872878,
2002.
[4] D.H. Owens, N. Amann, E. Rogers, and M. French, Analysis of linear
iterative learning control schemes a 2D systems/repetitive processes
approach, Multidimensional Systems and Signal Processing, vol. 11,
no. 1-2, pp. 125177, 2000.
[5] R. DAndrea and G.E. Dullerud, Distributed control design for
spatially interconnected systems, IEEE Transactions on Automatic
Control, vol. 48, no. 9, pp. 14781495, 2003.
[6] J. Bochniak, K. Galkowski, E. Rogers, D. Mehdi, O. Bachelier, and
A. Kummert, Stabilization of discrete linear repetitive processes with
switched dynamics, Multidimensional Systems and Signal Processing,
vol. 17, no. 2-3, pp. 271293, 2006.
[7] C. Du and L. Xie, H Control and Filtering of Two-dimensional
Systems, ser. Lecture Notes in Control and Information Sciences.
Berlin, Germany: Springer-Verlag, 2002, vol. 278.
[8] K. Galkowski, E. Rogers, S. Xu, J. Lam, and D.H. Owens, LMIs
a fundamental tool in analysis and controller design for discrete linear
repetitive processes, IEEE Transactions on Circuits and Systems I:
Fundamental Theory and Applications, vol. 49, no. 6, pp. 768778,
2002.
705