Professional Documents
Culture Documents
6, 1996
Translated hrn Denki Wai Ronbunshi, Vol. 1 1 5 4 , No. 8, August 1995, pp. 984-991
SETSUO SAGARA
Fukuoka Institute of Technology
SUMMARY
This paper deals with the problem of determination
of feedback gains in order to achieve required performances under linear time invariant perturbations. Based
upon the Lagrange multiplier method, the proposed
method determines feedback gains that ensure both robust
stability and robust performance for a linear uncertain
system. It is applied to a nonlinear crane system control
design. The nonlinear system is first partitioned into some
uncertain linear systems. Then the proposed method is
applied to each uncertain linear system. Application is
easily demonstrated and, consequently, a nonlinear crane
control system can be easily designed. Finally, it is shown
in a numerical example that the proposed method is
effective.
Key words: Robust control; pole placement; feedback gain; optimization; crane; nonlinear control.
1. Introduction
ISSN0424-7760/96/~-0104
0 1996 Scripta Technica, Inc.
Perturbed system:
i=l
i=l
Here Ei, Fi are matrices with elements of 0 or 1 representing structure of perturbations, and a,,bi are scalars
representing the magnitude of perturbations, their
maximum values are assumed to be known. Control rules
are as follows:
U ( t ) = -Kzn(t),
up(t)= - K a p ( t )
(4)
Q := {Z
+ j v : f ( ~v ,) = C C fijzd < 01 ( 5 )
i=Q j = O
More specifically, a nonlinear system was considered as a set of piecewise linear systems related to
parameter perturbations, and applying the aforementioned
method to these linear systems allowed for the design of
a control system to ensure definite responses for the
original nonlinear system. Furthermore, efficiency of this
method was validated using the specific example of crane
control.
C Y ~ ~ A ~ X ( =A -Q
* ) ~
*: conjugate
i,j=O
transform
c&o
cj=;;
Nominal system:
105
n:
- BnK) C Q
X(Ap - B p K ) C Q
X(A"
4. Optimization Problems
(7)
(1 1).
Specification 1
Lemma 1 suggests the following bounds for poles
before and after perturbation, respectively,
Specification 2
Here, arrows represent transfers of Pi and closedloop poles. By "big" is meant that the left upper minor
determinant's sum is big. Basically, Pi is positive definite
if the closed-loop pole falls within a specified region.
However, gain K is also a function of Pi, and the simple
existence of Pi inside a region is not sufficient while the
magnitude of positive definite value may be optimized as
P ~+
A ATPI
P ~+
A ATPZ
+ K T R K - Cz=orj,jAiP3(AT)J= o
+ K T R K - C$=ori,jA'P4(AT)J= o
(1 1)
106
tr(P1) ( 1 - p) . tr(P2)
+PC;=~
Cy=l{detPj(i)}2 (12)
Since this additional term does not belong essentially to evaluation function, it is necessary to reduce as
much as possible its influence on optimality of gain K
determined by setting p(0 < p << p).
p,
Motion of bogey
Bogey
...............
$ Subsystem 1
...............
Subsystem 2
2............
...............
$ Subsystem n
...............
Fig. 1. Nonlinear crane system.
An =
0 -mg/M
- Z1/M
0
0
*A=
ro
0
0
0
0
B, =
,AB=
0
0
0
0
0
0
-GAi/(li2M)
108
0O
Fig. 2. Compensator.
Assumptions were made that 1: - A: = I:, 2, = D + G,
2, = M + m. Here, M is the mass of bogey (kg); G is the
transformation coefficient for torque/driving force (N/V);
D is the transverse friction coefficient (kgls); li is the
rope's length (m); m is the mass of load (kg); and g is
gravity (ds').
+
+
= Jr[XT(Q KTRK)X]dt
= JT[XT(QK T R K ) X ] d t
(24)
5.2 Compensator
J =
P . tr(DP1)+ (1 - P ) . tr(DP2)
+P C4=3CZ1{detPj(i)}2
(25)
P ~ A + A ~ P ~ + K T R K=+ Q
o
~A+A~P~+KTRK+Q
=
(26)
= AX+BU
= [o
1]X
(22)
A = [ An .
0]
..=[:],X=[:]
hx
= - K,x,
+ K,e
=-
X = [A-BK]X
K=[K1
-K2]
(23)
109
Design method
-24.7474
AqAT
AP4AT
-9.2766
-5.5278
- j3.7565
+
+
(25).
(5) Numerical example
6::::;
I -0.4166
Proposed method
Nominal poles
111
-0.5804
-3.9696
(I
1
I1
-24.8420
-1.0234
+ j2.1397
-1.0234
- j2.1397
/I +
11
poles
-1.1000
j0.5735
-1.1000
- j0.5735
+ j2.4526
+ j0.5139
-1.0486
- j0.5139
-12.0633
Max.perturbation
11
-6.7514
,.---I,
I
_
_.__----I
I1
\I
-- - -
(s)
The following matrices A, B were used in the equation of state for load's lift, with state variable vector
being [rope's length, rope's velocity]:
A=[:
assignment
As may
after
be the
seenproposed
from Tables
method,
2 and
irrespective
3, the pole
of
system perturbations, falls within the desired region, i.e.,
within the circle with center -7 and radius of 6.5.
-0.7 ] . B = [ o . ( L ]
111
---
- Proposed method (2
partitions)
partitions)
0.1
-2L2
- - - _---
-.
6 ---.-_.___
8
(2 partitions)
-e- ;
L
Fig. 6. Input.
---..- - - _ ,
(s)
6)
i
-
Proposed method (2
*
II
-0.02
I
*,
I1
,*
I.
00
II
I S
6. Conclusions
APPENDIX
Hence,
Inserting x,, u, in Eq. (8) the following can be derived
Acknowledgment
The authors would like to thank Mr. H.Kakiuchi
and Mr.H. Kobayashi for their help with the numerical
example. This research has been performed with financial
aid from the Ministry of Education (1993).
Here, instead of the problem of determination of K to
minimize J, for all x,,(O), another problem is considered:
determination of feedback gain Kin average terms. Then
.[6]:
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
= trJ7[eATi(Q K*RK)eA']dt
Jn
(A3)
+ KTRK)
or
;Iie
d At = AeAt = ATP,+P,A
Therefore,
P~A+ATP~+Q+K~RK
P~+
A ATPl
+ KTRK - Cc=ori,jAip3A" j T =
0
(1 1)
= tr(P1)
113
Masanao Obayashi graduated in 1975 from Kyushu Institute of Technology, Electric Eng. Dept., and received his M.S.
degree from Kyushu University in 1979 and joined Hitachi, Ltd.In 1988 he was Research Associate at Kyushu University.
He is a member of Society of Instrument and Control Engineering; and Institute of Systems, Control and Infomation
Engineers.
Kotaro Hirasawa graduated in 1964 from Kyushu University. In 1966, he received his M.S.degree and later his Ph.D.
degree from Kyushu University. In 1966 he joined Hitachi, Ltd. Since 1992 he has been a Professor at Kyushu University.
He is a member of the Society of Instrument and Control Engineering; Information Processing Society of Japan; and IEEE
USA.
Junichi Murata received his Ph.D. degree in 1986 from Kyushu University. In 1986 he was a Research Associate at
the same university and Associate Professor in 1988. He is a member of the Society of Instrument and Control Engineers;
and Institute of Systems, Control and Information Engineers. He is engaged in research on system simulation, power load
prediction.
Akio Kajiwara graduated in 1965 from Fukuoka Prefectural Technical High School. In 1965, he became an engineer
at Kyushu University. He is a member of the Society of Instrument and Control Engineering.
Setsuo Sagara graduated in 1943 from Kyushu University, Electrical Eng. Dept., and received his Ph.D. degree later.
In 1956 he was a Lecturer at Hiroshima University, in 1960 an Associate Professor, and in 1963 a Professor, in 1968
Professor at Kyushu University, and in 1992 a Professor at Fukuoka Institute of Technology. He is a member of the Society
of Instrument and Control Engineering; and Institute of Systems, Control and Infomation Engineers.
114