Professional Documents
Culture Documents
5865, 2015
SCI14 I
FRIT
*
In this paper, we develop a controller tuning method of the Smith compensator for multi input
multi output (MIMO) linear time-invariant systems with output time delays. Particularly, we apply
fictitious reference iterative tuning (FRIT), which is one of the controller tuning methods with
only one-shot experimental data, to such systems. Since the Smith compensator contains a model
of a plant, it is expected that the model is also obtained together with control parameters by
FRIT. To realize this expectation, we introduce a special structure of the Smith compensator with
tunable parameters. Then, we also show that the application of FRIT to the parameterized Smith
compensator yields not only a desired control parameter but also an appropriate mathematical model
of a plant. To show the validity of the proposed method, we give an illustrative numerical example.
1.
[1,2]
[3]
[47] Fig. 1 The closed loop system with the Smith compen-
Fig. 1 sator
2014 7 14
Graduate School of Natural Science, Kanazawa Univer- PE
sity; Kakumamachi, Kanazawa City, Ishikawa 920-1192,
JAPAN
Institute of Science
and Engineering, Kanazawa University; Kakumamachi,
Kanazawa City, Ishikawa 920-1192, JAPAN
Key Words : controller tuning, time delay, MIMO, Smith
compensator, FRIT.
18
FRIT 59
Fictitious
Reference Iterative TuningFRIT[1012]
[1315]
Fig. 2 The closed loop system with the tunable Smith
compensator
w
N
FRIT w(N,) := i=1 w(i)
2
[16] q
w =: [w1 w2 wq ]
q
[17] w(N,) := 2
i=1 wi (N,)
1
2.
[17] Fig. 2
FRIT P q q
P
P11 eL1 s P1q eL1 s
.. .. ..
P =
. . .
FRIT
Pq1 eLq s Pqq eLq s
eL1 s 0 P11 P1q
. .
= ..
. . . . .. (1)
. .
0 eLq s Pq1 Pqq
FRIT [1820]
D Pm
FRIT
Pm D
P = DPm (2)
FRIT
FRIT P ()
(i,j)
(i,j)
P , , w
19
60 28 2 (2015)
P () (1) P
P11 ()eL1 s P1q ()eL1 s
.. .. ..
.
. .
Lq s Lq s
Pq1 ()e Pqq ()e Fig. 3 The conventional feedback system with tunable
L1 s
e 0 P11 () P1q () parameters
.. .. .. ..
=
. (6)
. . .
3. 1 FRIT
0 eLq s Pq1 () Pqq ()
[1012] 1 FRIT
D() Pm () Fig. 3
P C()
Pm ()
D()
P () = D()Pm () (7) yd
Td
Cs
y()
yd
0 C(0 )
u(0 )y(0 )
Td () = D()Tdm (8)
[23]
Tdm Td ()
D() r() = C()1 u(0 ) + y(0 ) (9)
P P
r() C()
1 [21,22,17]
(1) Gry ()r() = y(0 ) (10)
y(0 )
r()
JF ()
0 u(0 ) y(0 )
( ) 2
Td
JF () = 1 y(0 ) (12)
Gry ()
(N,)
,
JF ()
N
Td Gry ()
20
FRIT 61
4.
4.1
Fig. 2 Pm
Fig. 4 IMC[16]
FRIT 1 IMC
Fig. 4 The IMC typed Smith compensator
[2426] Fig. 4
IMC
IMC ([4,16]) Tdm ([17,25])
Fig. 5 The conventional feedback systems equivalent to
Fig. 4 Fig. 5
MIMO smith compensator
4.2 FRIT
1 (14) CF ()
CF () = Pm () {I Tdm D()}1 Tdm (14)
FRIT
0 u(0 ) y(0 )
P = P () 0
1. u(0 ) y(0 )
1
Gry () = Td () (15) r() = CF () u(0 ) + y(0 )
1
( )
= Tdm I Tdm D() Pm ()u(0 )
+y(0 ) (20)
Gry () = Td ()
2. (14)
Gry () ( )
( )1 Td ()r() = I Td () P ()u(0 )
Gry () = I + P P ()1 (I Td ())1 Td () +Td ()y(0 ) (21)
P P ()1 (I Td ())1 Td () (16) y(0 )
Gry () = Td () Td ()r()
( )1 JF () = y(0 ) Td ()r()2(N,) (22)
Td () = I + P P ()1 (I Td ())1 Td ()
P P ()1 (I Td ())1 Td () (17) 1 FRIT
(17)
( )
I P P () 1
Td () = 0 (18)
JF ()
Td () Gry ()
y(0 ) r()
P P ()1 = I (19) Gry ()r() = (I + P CF ())1 P CF ()
( )
1
generic (P CF ()) y(0 ) + y(0 )
= y(0 ) (23)
(14)
r()
CF () Fig. 5
1
r() = Gry () y(0 ) (24)
(22)
21
62 62
28
28
2
2
(2015)
(2015)
( ( 1 )1 )
JF () () =I ITd()G
JF = Td ()G
ry ()
ry () y(y(
2 2
0 )(N,)
0 )(N,)
(25)(25)
3
3
()
PmP
m ()
2
2
1
1
C()
C()
= P=
mP
() TdmT
1 1
m () dm
JF ()
JF ()
Td
Td
GryG ()
ry ()
10[s]10[s]
20[s]
20[s]
FRIT
FRIT
Td ()r
Td ()r 0)
y(y( Fig.
0 )Fig. 6 6
(22)(22)
JF ()
JF ()
(21)(21)
7
Fig.Fig. 7
0 )
u(u( 0 ) 6
Fig.Fig. 6
( ( )( )( ) ) 2 2
JF ()=
JF ()=
ITIT
d ()d ()
IP
I()P 1 1
P ()P 0)(N,)
y(y( 0)(N,)(26)(26)
JF
JF () ()
I ITd()
Td ()
) )y(
y(y( y(
)
) 0 0 0 0
P P
() ()
P
P
(14)(14)
CMA-ES Covariance
CMA-ES Covariance
Matrix
Matrix
Adaptation
Adaptation
FRIT
FRIT Evolution
Evolution
Strategy [27,28]
Strategy[27,28]
Matlab
Matlab
JF
JF () () 8 8
Fig.Fig.
5. 5.
JF
JF () ()
)
P (P (
)
2s 2s
6e 6e 8e2s
8e2s
e2se2s e2se2s
2 2 ++3s 2 +s22 +s4s
2+ 3
P =P=s +s3s +4s
3+ 0.163s
2 + 0.497s
2 + 0.497s 2 + 0.496s
2 + 0.496s
8e4s
8e4s 12e4s 12e4s 0.163s + 0.333
+ 0.333
0.123s
0.123s + 0.375
+ 0.375
2 2+
e4se4s e4se4s
s +s5s2s6 +s62 +s6s
+5s 2+ +6s
8 +8 2 0.623s
2 2 + 0.498s
2 + 0.498s
2s 0.124s
0.124s
+ + 0.623s
+ 0.375
+ 0.375
0.083s
0.083s + 0.667
+0.667
6e 6e 8e2s8e2s 2s 2s 2s 2s
(s+(s 6.08e
6.08e 8.16e
8.16e
= 2)(s
+ 2)(s
+ 1)+ (s
1) +(s
3)(s + 1)+
+ 3)(s 1)
(s+(s
= 8e4s8e 4s 4s 4s
12e 12e (27)(27)
= 2.04)(s
+ 2.04)(s
+ 1.00)
+ 1.00)
(s +(s
3.04)(s
+ 3.04)(s
+ 1.00)
+ 1.00)
(30)
= 8.08e 4s 4s
8.08e 12.12e 4s 4s (30)
12.12e
(s +(s
3)(s
+ 3)(s
+ 2)+ (s
2) +(s
4)(s
+ 4)(s
+ 2)+ 2)
(s +(s
3.05)(s
+ 3.05)(s
+ 1.98)
+ 1.98)
(s +(s
4.04)(s
+ 4.04)(s
+ 2.00)
+ 2.00)
P ()
P ()
Fig.Fig.
9 9
(6)
(6)
)r
Td (
Td (
)r
0
0
)
y(y(
)
Fig.Fig.
6 6
P (P0 )(0 )
Td ()r
Td ()r 0 )
y(y( 0 )
Fig.Fig.
9 9
s s s s
e e e e
0.183s
0.183s
2 + 0.550s
2 + 0.550s
+ 0.367
+ 0.367
0.163s 2
0.163s 2
+ 0.663s
+ 0.663s
+ 0.500
+ 0.500
10
Fig.Fig. 10
)
u(u(
)
es es es es
0.188s 2 0.913s
0.188s 2 2 + 0.525s
2 + 0.525s
+ + 0.913s
+ 0.913
s s
+ 0.913
0.092s
0.092s
s s
+ 0.700
+ 0.700
T Td
d () ()
L
L
5.4e5.4e 5.3e5.3e
(s+(s
= 2.0)(s
+ 2.0)(s
+ 1.0)
+ 1.0)
(s +(s
3.4)(s
+ 3.4)(s
+ 1.4)
+ 1.4)
= 6.2e6.2es s
10.9e s s (28)(28)
10.9e
Fig.Fig.
9 9
(s +(s
3.1)(s
+ 3.1)(s
+ 1)+ 1)
(s +(s
3.6)(s
+ 3.6)(s
+ 2.1)
+ 2.1)
)r
Td (
Td (
)r
)
y(y(
)T )r
Td (
d ( )r
1 1
0 0
(2s
+ 3 1)3
1)+
(2s
TdmT=
dm= 1 1 (29)(29)
0 0 3 1)3
)
y(y(
)
(2s
(2s + 1)+
22 22
FRIT 63
(27) P (30)
P ( ) (28)
P (0 )
6.
23
64 28 2 (2015)
[17] , , : FRIT
;
C, Vol. 131, No. 4, pp. 742750 (2011)
[18] : PID ;
C, Vol. 133, No. 12, pp. 2229-2235 (2013)
[19] , :
; 13
(C)25420432 PS-013 (2013)
(B)23360183 [20] :
;
Vol. 43, No. 5, pp. 391399 (2007)
[21] , :
[1] : (1987) FRIT ; 37
[2] , : , , pp. 235238 (2008)
(2007) [22] , , , , :
[3] : ; PID ;
Vol. 45, No. 9, pp. 827-833 (2006) , Vol. 53, No. 4, pp. 137144
[4] 2003 (2009)
[5] , , , : [23] M. G. Safonov and T. C. Tsao: The unfalsied
; , control concept and learning; IEEE Transactions on
Vol. 33, No. 6, pp. 494501 (1997) Automatic and Control, Vol. 42, No. 6, pp. 843
[6] S. Wijaya, K. Shimizu and M. Nakamoto: LQR 847 (1997)
scheme for SCR process in combined-cycle thermal [24] O. Kaneko, M. Miyachi and T. Fujii: Simultane-
power plants; Proceedings of the 5th International ous updating of a model and a controller based on
Symposium on Advanced Control and Industrial Pro- the ctitious controller; Proceedins of the 47th IEEE
cess, pp. 1924, (2014) Conference on Decision and Control, pp. 13581363
[7] : PD (2008)
MD-PID ; [25] O. Kaneko, M. Miyachi and T. Fujii : Simultaneous
Vol. 49, No. 9, pp. 875879 (2013) updating of model and controller based on ctitious
[8] O. J. M. Smith: A controller to overcome dead time; reference iterative tuning; SICE Journal of Control,
ISA J., Vol. 6, No. 2, pp. 2833 (1959) Measurement, and System Integration, Vol. 4, No. 1,
[9] : , pp. 6370 (2011)
(1993) [26] O. Kaneko, H. T. Nguyen, Y. Wadagaki and S. Ya-
[10] : mamoto : Fictitious reference iterative tuning for
Fictitious non-minimum phase systems in he IMC architecture:
Reference Iterative Tuning ; Simultaneous attainment of controllers and models;
, Vol. 17, No. 12, pp. 528536 (2004) SICE Journal of Control, Measurement, and System
[11] : Integration, Vol. 5, No. 2, pp. 101108 (2012)
; Vol. 47, No. 11, pp. 903908 (2008) [27] N. Hansen: The CMA Evolution Strategy: a tutorial;
[12] O. Kaneko : Data-driven controller tuning: FRIT https://www.lri.fr/ hansen/cmatutorial.pdf (2011)
approach (Tutorial); Proceedings of 11th IFAC Work- [28] Y. Wakasa, S. Kanagawa, K. Tanaka and Y.
shop on Adaptation and Learning in Control and Sig- Nishimura: FRIT for systems with deadzone and
nal Processing (ALCOSP13), pp. 326336 (2013) its applications to ultrasonic motors;
[13] W. S. Lee, B. D. O. Anderson, I. M. Y. Mareels and C, Vol. 131, No. 6, pp. 12091216 (2011)
R. L. Kosut : On some key issues in the windsurfer
approach; Automatica, Vol. 31, No. 11, pp. 15771594
(1995)
[14] : IMC 1. (15)
; (15) (7) (8)
Vol. 36, No. 7, pp. 563568 (2000)
24
FRIT 65
1
Gry () = (I + P CF ()) P CF ()
( )1
= I + D()(I Tdm D())1 Tdm ()
1 2014 3
D(){I Tdm D()} Tdm
( )1 4
= I + (I D()Tdm )1 D()Tdm
(I D()Tdm )1 D()Tdm
= D()Tdm = Td () (A1)
2. (14) Gry ()
(16) (7) (8)
()
1994
1
Gry () = (I + P CF ()) P CF () 4 1996
( ( )1 )1 3
1
= I + P Pm () I Tdm D() Tdm 1999
( )1
1
P Pm () I Tdm D() Tdm 2009
( ( )1 )1
= I + P P ()1 I D()Tdm D()Tdm 2008
( )1 2012
P P ()1 I D()Tdm D()Tdm
( )1
= I + P P ()1 (I Td ())1 Td ()
()
P P ()1 (I Td ())1 Td () (A2)
1964 5 1989
2007
2008
25