KSME International Journal, VoL 14, No.4, pp.

393 - 400, 2000

An Algorithm for Robust Noninteracting Control of Ship
Propulsion System
Young-Bok Kim*
Department of Marine Engineering, Gyeongsang National University
In this paper, a new algorithm for noninteracting control system design is proposed and
applied to ship propulsion system control. For example, if a ship diesel engine is operated by
consolidated control with controllable pitch propeller (CPP), the minimum fuel consumption is
achieved satisfying the demanded ship speed. For this, it is necessary that the ship is operated
on the ideal operating line which satisfies the minimum fuel consumption, and the both pitch
angle of CPP and throttle valve angle are controlled simultaneously. In this context of view, this
paper gives a controller design method for a ship propulsion system with CPP based on
noninteracting control theory. Where, linear matrix inequality (LMI) approach is introduced
for the control system design to satisfy the given H ~ constraint in the presence of physical
parameter perturbation and disturbance input. To the end, the validity and applicability of this
approach are illustrated by the simulation in the all operating ranges.
Key Words: Noninteracting Control, Controllable Pitch Propeller, Linear Matrix Inequality,
Parameter Perturbation, Operating Ranges
393
1. Introduction
Due to the recent technical development in the
marine industry, diesel engine propulsion system
is a subject of renewed interest. Moreover, it is
shown that the propulsion system control can
have a significant effect on the fuel efficiency
(Hendricks et aI., 1986). In this context of view,
this paper gives a control system design method
for a marine diesel engine propulsion system with
controllable pitch propeller (CPP).
In the ship propulsion system with CPP, there
are many operating points which keep the ship
speed constant. In order to achieve the minimum
fuel consumption demand, the ship propulsion
system needs to operate on the ideal line. That the
two controlled outputs, engine-speed and CPP
pitch angle are needed to be controlled simultane-
E-mail: [email protected]
TEL: +82-557-640-3162; FAX: +82-557-640-3128
Dept. of Marine Engineering, College of Marine Sci-
ence, Gyeongsang National University, Korea (Manu-
script Received June 28, 1999; Revised January 22,
2000)
ously, which is difficult, because each input
affects the outputs. Therefore the system used to
be controlled by the conventional approach in
which the one mode of two is controlled and the
other is fixed. In order to overcome this problem,
it is necessary that the controlled system is
divided into two single-input single-output sub-
systems. Then each input affects only correspond-
ing output. This paper deals with noninteracting
control of the controlled system which has two
independent inputs and outputs, where the inputs
are fuel rack position reference signal and voltage
signal of the CPP actuator and the outputs are
engine-speed and CPP pitch angle. In this study,
LMI approach (Boyd et aI., 1990 ; Gahinet and
Apkarian, 1994 ; Iwasaki and Skelton, 1993) is
used to achieve H ~ constraint in the all operating
ranges. The vessel under consideration is the
training ship of Pukyong National Univ. (GfT
653) .
This paper is organized as follows. In Sec. 2, a
dynamic model is developed for the ship propul-
sion system. In Sec. 3, noninteracting control
system design method is introduced. Section 4
394 Young-Bok Kim
describes the constraints for control system design
problem. In Sec. 5, an algorithm is proposed such
that the constraints are achieved. In Sec. 6, the
validity and applicability of this approach is
demonstrated by the simulation. Finally, some
conclusions are given in Sec. 7.
2. System Modeling and Control
Objectives
2.1 Modeling
The block diagram of the controlled system is
shown in Fig. 1. The inputs are the engine throttle
valve angle R[rad] and the reference input of
Cf'P actuator 8cp[V] to adjust CPP pitch angle.
The controlled outputs are engine-speed n.[I/s]
and CPP pitch angle ecp[radJ.
The parameters appeared in Fig. I are summar-
ized in the Table. The function e-
L
is a time
delay which is approximated with a first-order
rational function. In this paper, we consider
T, L aTe/aR, J, etvs; aTdan., T.
as the nonlinear terms which are linearized at
each operating point. From these, we can obtain
the following system representation :
x(t) =Ax (t) +Bu (t) : A(nx n),
B(nXm)
y(t)=Cx(t): C(pXn). (I)
L [s]
T [s]
R [rad]
J [kg m"]
t: [Nm]
To [Nm]
n. [ljs]
f [kg m
2
/ s]
Va [m/s]
T. [s]
K [rad/V]
Table 1 Parameters
Delay time of engine
Time constant
Engine throttle valve angle
Total inertia
Load torque
Disturbance
Engine-speed
Friction
Ship speed
Time constant of CPP actuator
Proportional gain
where x (t) (= [ne T; i, ecp) T), U (t) and y
(t) are the state. control input and controlled
output, respectively. The coefficient matrices of
the plant are denoted by
-7{ aT
L
+/) l..
1 I aT
L
] ane J -77 ae
ep
a
2 4
0
T T
A=
I st; I
-7' ane
a
T
0
0 a 0
I
- T.
n.
sr,
r-----------ian;
Fig. 1 Block diagram of controlled system
An Algorithm for Robust Noninteracting Control of Ship Propulsion System 395
B=
0
o 0
_I aT. 0
T aR K
o T
s
[
I 00 OJ
' C= 0 0 I .
(2)
the closed-loop system is stable and RMS gain
(H", norm) from w to y does not exceed r( >0).
3. Noninteracting Control
Consider the system illustrated in Fig. 3. It is
represented by the following linear state equation.
2.2 Control Objectives
2.2.1 Noninteraction between inputs and
outputs
As described in Fig. I, the variation of CPP
pitch angle disturbs the engine-speed. Therefore,
by applying the noninteracting control theory, we
can make the system to be decoupled into two
single input single-output subsystems.
i(t)=Ax(t)+Bu(t) :A(nxn),
B(nXm) (5)
y(t)=Cx(t) : C(pXn)
As shown in Fig. 3, linear state feedback
replaces the plant input u (t) by the following
expression
u(t)=-Fx(t)+Cv(t) : F(mXn),
Gt m x n) (6)
x (t) =Ax (t) +n (t) +Bww(t),
y (t) = ex (t) +Dw (t), (3)
where x(t) = [x (t) v (t)T, ii (t), w (t), v (t)
are new state, control input, disturbance input
and output of integrator in the augmented system,
respectively. The system matrices are denoted by
2.2.2 H", Constraint
In order to reject the steady-state tracking error
for constant reference signals, the integral
compensator is introduced. Consider the integral
type servosystem of Fig. 2(Fujisaki and Ikeda,
1992). Then the servosystem (augmented system)
is represented by
where, v (t) is m X I input signal. Then the
closed-loop state equation is described by
i (t) = (A - BF) x (t) +BGv (t) (7)
y(t) =Cx(t)
(8)
L:=C' (/JA-BF' B G
Noninteacting control problem involves using
linear state feedback to achieve two input-output
objectives. The closed-loop state (7) should be
such that for i =1= j the ph-input component us (t)
has no effect on the ilh-output component Yi (t).
This problem is equivalent to the requirement
that the closed-loop impulse response:
be a diagonal matrix. A closed-loop state equa-
tion with this property can be viewed from the
input-output perspective as a collection of m
independent, single-input, single-output linear
systems.
Here, let
(4)
- [A OJ - [B] -
A= -CO ,B= 0' C=[CO],
- [0000-1 0JT - [IOJ
B
w=
0 0 0 -1 ' D= I .
It is assumed that the parameters of controlled
system are perturbed in the specified ranges.
For this system, it is needed to guarantee that
w
-
y
for each
Fig. 2 An integral type servosystem Fig. 3 Structure of linear state feedback
396 Young-Bok Kim
Lemma 2 (Morse and Wenham, 1971) Ifa pair
(F, G) is given by
F=L1-
1
[A * + FS
a
] , G=L1-
1
(12)
Lemma 1 (Choi, 1998; Falb and Wolovich.,
1967; Freund, 1971; Porter, 1969) Suppose the
time-invariant linear state equation (5) with p=
m has relative degree Xh ... , Xm' Then there exist
constant feedback gains F and G that achieve
non interacting control if and only if
( 14)
state equation does not equal to that of transfer
function, then the matrix Sa is singular. There-
fore, the following new matrix is introduced
In this paper, it is considered that system
parameters are varying in the specified ranges.
For this system, robust stability and H.. perfor-
mance problems of the servosystem are consid-
ered. These problems are given by following
Theorem.
Theorem For the system described by Eq. (3),
the H: norm of T yW which is the closed-loop
transfer function of the system (3) via state feed-
back (6) is smaller than y (>0), if and only if
there exist X (>0) and parameter Y satisfying
the following LMI:
[
Ax +xAT EY+ yTB
T
Ew X!T]
Bw
T
-yI D
T
<0
ex 15 -yI
( 16)
4. H"" Constraint
such that !Sbl'*O, where W is a matrix satisfying
WBp=O. (15)
(9)
(II)
( 10)
where
[

L1= :
cmAXm-1B
is invertible.
Then a pair of (F, G) is given by
F'=L1-
1A
*, G=L1-
1
r
c1A
X,
1
A*=
cmA
Xm
Even though the aforementioned condition is
satisfied such that the noninteracting control is
achieved, the system stability is not guaranteed.
For this, the following lemma is considered.
Suppose that the noninteracting condition illus-
trated in Sec. 3 is satisfied for the controlled
system considered in this study. If the degree of
then the noninteracting control system is stable.
Where F is a feedback gain such that the closed-
loop system is asymptotically stable, where Sa is
given by the following nonsingular matrix:
_ _ _ _ k _
A(a) =Am+oA(a)=A
m
+
i::=l
(17)
Then the feedback gain K
FG
is given by
K
FC
=YX-
1
=[F G]
Proof See reference (Gahinet and Apkarian,
1994)
This result is easily extended to uncertain sys-
tems described by polytopic state-space model.
Let us denote matrices A, E as
A=Am+oA. E=Em+oB, (18)
where Am, Em are nominal parts and oA, oE
denote uncertain parts.
Here, a standard numerical method is consid-
ered to check whether (16) holds or not. If the set
of uncertain plant is polytopic, that is. the set
described as
( 13) (nX n),
cm
cmA
An Algorithm for Robust Noninteracting Control of Ship Propulsion System 397
- - - _..-
B(a) =Bm+oB(a) =Bm+l},aBi'
i=1
..
l},ai= I. a20. i= I ..... k (19)
i=1
then we can easily find the minimum r- a positive
definite matrix X and a matrix Y satisfying the
condition (12) and (16) at all extreme points.
simultaneously. <.iL. Bi). i = 1. 2, ... k denote
the extreme points.
5. Algorithm
The controller design problem considered in
this paper is equivalent to finding a state feedback
gain K
FC
(F, e) which satisfies the following
specifications simultaneously.
AI) Find K
FG
satisfying the noninteracting
control constraints (Lemma 1. Lemma
2) for a nominal system.
A2) The inequality (16) of the Theorem
holds.
Also, this problem can be described by the
following algorithm.
[Algorithm]
Consider the system (I). For given sufficiently
large positive number Yn' set the numerical toler-
ance e>O and n=O.
Step O. Find a feedback gain F and G satis-
fying Lemmas I and 2.
Step I. Using the relation
Kn=[F
n
en] = YnX;l,
Y
n
= [F
n
en] x; and
solve the inequality (16) in Theorem,
to get (Xl!' Yn+I)'
Step 2. If IYn-Yn+ll>e, then let n=n+I.
and go to step O.
Otherwise, output (Xl!' Y
n
, Yn+1)'
Step 3. The controller gain is
K
FC
=YnX;I=[Fn en].
This problem is easily solved by LMI
approach.
6. Simulations
In this section. we present simulation results to
illustrate the validity and applicability of the
approach studied in this paper.
First, a type of controller gain for noninteract-
ing control is calculated based on Lemmas I and
2. Next, check the H ~ constraint for the servosys-
tern with noninteracting controller in the presence
of parameter perturbation. For this, consider that
the parameters are varying in the following
ranges:
39.49:::;:0 T
e/i1R:::;:
I 14.39, 0.03:::;: T:::;:O.IO
0.05:::;:L:::;:0.20,147.33:::;:J:::;:199.55.
167.35:::;:0 T
e/ane:::;:262.04,
0.05:::;: T.:::;:O.23.
(20)
12585.13:::;:a tu0 8et>:::;:62931Al
32.62:::;:0 TL!an
e:::;:944.67,
The nominal part Am, s:of (19) are described
by
-3.289 0.006 -0.006 25.080 00
0 - 25.000 50.000 0 00
Am=
-5204.085 0 -21.665 0 00
0 0 0 -12.17500
0 0 0 0 00
0 0 0 0 00
- _[002103.760 0 0 or
(21)
B
m
- 00 0 12.175 .
These are mean values of the system matrices set.
Then, A, jj of (3), the system matrices with
uncertainties are represented by
_ _ 32 _
A(a) =A
m
+I:aAi (22)
i=l
_ _ 32 _
B(a) =B
m
+I:aBi (23)
i=l
k
I:ai= I, a;20 (i= I, "', 32). (24)
i=1
Based on Lemmas I and 2, a gain F illustrated
in (12) is obtained by
F = [ ~ ~ ~ ~ J . (25)
where W=[O -I OOJ(in Eqs. (14) and (15)),
such that the noninteracting control system is
stable. Here, A of equation (9) and Ll of equa-
tion (II) are calculated from nominal val ues.
Even though the gain (F. e) is obtained so
that the noninteracting control system is stable,
the system stability may not be guaranteed for the
398 Young-Bok Kim
(26)
(a)
(b)
Fig. 4 Step responses without noninteracting con-
trol[ (a) i!e[rpsJ, (b) 8
cP[radJ)J
uncertainty. Therefore the robust stability condi-
tion represented in the Theorem for the uncertain
system is considered. If we use the LMI approach,
we can obtain a gain (F, G) based on the
Algorithm proposed in Section 5.
A state feedback gain F and an integral
compensator gain G are calculated as follows:
=[-0.08020.0004 -0.0015 0.1369 J
F 0 0 0 -0.0416
G
= [ 3.2111 -2.6163J (26)
o 0.5310
where 11=1.300, 12=10.000, 13=3.000 of F in
(25) and the bound r=21.3.
Using the gain illustrated in (26), the simula-
tion results, which are obtained in the 16 extreme
points when the parameter perturbation is consid-
ered, are given in the Figs. 4-7. Especially, in the
cases of Figs. 4 and 5, engine-speed and Cf'P
pitch angle reference signals are changing at the
same time as shown in the simulation results.
Figure 4 shows the controlled output to the step
type reference signals when only the robust con-
(a)
(b)
Fig. 5 Step responses with noninteracting controll
[(a) n.[rpsJ, (b) 8cp[radJ)J
trol constraint is considered without noninteract-
ing control. In this case, following gains are used
as the feedback gain.
_ [0.0045 X 10-
5
-0.0047 X 10-
5
0.2448 X 10-
5
F= -0.0268 X10-
5
0.OO12x 10-
5
-0.4899 X10-
5
-0.0345 X 10-
5
]
0.0086 X 10-
5
- = [3.3606 2.2650J
G -0.01260.2250
And, Fig. 5 shows the controlled output when
both the nointeracting and robust control are
considered. Figures 6 and 7 illustrate the
controlled output and control input. In the cases
of Figs. 6 and 7, the engine- speed is fixed and the
CPP pitch angle is varied. From these results, it is
clear that we can achieve good suppression of the
interaction between inputs and outputs, and satis-
fy H ~ constraint simultaneously.
7. Concluding Remarks
In this paper, a controller design method for a
An Algorithm for Robust Noninteracting Control of Ship Propulsion System 399
&. _. - - - - - -- - - -- -----.- - - -
---t-1-------+-------
i
-------f-------
, . , .
. . , ,
.... _------.. ------- ..---_ ------- -_._ .. - ----
j l :
:------T------r-------r-------j__---- ... ----
(a) (a)
.
r:::::::t.l:::::::1::::::::)::::::::;:::::::
: : : : '
r .-- -.. -:--- _. --- -_. ..... -
[_-E:EJ:t:;
j ; i ;
(b) (b)
. , . ,

r: ..
..::::::;:::::::
[g!-if
-_ ..--- .i--. ----- -. _...- ...-- --. J. --_._ ..i...- ..
!-l_fJ-::
(e)
Fig. 6 Time responses without noninteraeting con-
trol when the engine-speed is fixed and the
pitch angle is changing
[(a) ne[rps], (b) Bcp[rad].
(c) control inputs]
(e)
Fig. 7 Time responses with noninteracting control
when the engine-speed is fixed and the pitch
angle is changing
[(a) ne[rps]. (b) Bcp[rad],
(e) control inputs]
noninteracting control system design has been
presented and it is applied to ship propulsion
system control problem. In order to reject the
interaction between inputs and outputs. the
noninteracting control theory is used. Linear
matrix inequality(LMI) approach is introduced
so that the control system satisfies the given H;
constraint in the presence of physical parameter
perturbation. We have shown the validity and
applicability of this approach by acheving two
given objectives simultaneously in the simulation.
References
Boyd, S. , et aI., Linear Matrix Inequalities in
System and Control Theory. SIAM Book, 1990.
Chi a J. W., Left Eigenstructure Assignment via
Sylvester Equation, KSME International Jour-
400 Young -Bok Kim
nal, Vol. 12, No.6, pp. 1034-1040, 1998.
Falb, P. L. and Wolovich W. A., Noninteract-
ing in the Design and Synthesis of Multivariable
Control Systems, IEEE Trans. on AC, Vol. 12,
No.6, pp. 651-659, 1967.
Freund E., Design of time-variable rnultivaria-
ble systems by noninteracting and by the inverse,
IEEE Trans. on AC, Vol. 16, No.2, pp. 183
-185, 1971.
Fujisaki, Y. and Ikeda M., Synthesis of Two-
Degree-of-Freedom Design of Optimal Ser-
vosystems, Proc. 31st IEEE CDC, pp. 3588
-3589, 1992.
Gahinet, P. and Apkarian P., A Linear Matrix
Inequality Approach to H"" Control, Int. J.
Robust and Nonlinear Control, Vol. 4, pp. 421
-448, 1994.
Hendricks, E., J. Holst, N. K. Poulson and H.
Joensen, Adaptive minimum Energy Control of
Ship Diesel Engines, IMSOR, The Technical
University of Denmark, Lyngby, Denmark, 1986.
Iwasaki, T. and Skelton R. E., A Complete
Solution to the General H"" Control Problem
LMI Existence Conditions and State Space For-
mulas. ACC, pp. 605-609, 1993.
Morse, A. S. and Wonham W. M., Status of
noninteracting Control. IEEE Trans. on AC,
Vol. 16, No.6, pp. 568-581, 1971.
Porter W. A., Noninteracting of and inverse for
time-varying linear systems. IEEE Trans. on
AC, Vol. 14, No.4, pp. 378-380, 1969.