Professional Documents
Culture Documents
Marine Science
and Technology
9 SNAJ 1997
Abstract: An activated antiroll tank system design for ship P-(k) a priori error covariance
roll reduction was investigated. Considering the dynamics of qi generalized coordinates of the system
the ship motion, tank flow, and variable-pitch impeller, a con- Qr covariance of the wave slope
trol law is derived based on optimal control and estimation Qi nonconservative forces or moments acting on the
theory. Finally, a series of numerical computations for a 1174 coordinates q~
tonne ship without an antiroll tank system, with a passive tank Rr covariance of the measurement noise
system, and with the proposed activated antiroll tank system is R~ resistance coefficient of the ship and tank fluid
performed. The results show that the proposed activated tank motion
system has superior antiroll properties for free rolling, regular, T total kinetic energy of the ship and tank fluid
and irregular sea loads to the passive tank system. In addition, L sampling period
this design has a very low sensitivity to variations in ship and V potential energy of the ship and tank fluid
tank dynamics. It shows that the proposed system is reliable XG surge displacement of the ship
and valid for practical use. ~(k) the ith state value of X(k)
X(k) estimated value of X(k)
Key words: activated antiroll tanks, roll stabilization system, X-(k) a priori estimate value of X(k)
optimal control Y(k) measurable states of the system
Yc sway displacement of the ship
List of symbols Zc; heave displacement of the ship
ai weighting coefficient for the state x~(k)
A, local cross-sectional area normal to the tank water
weighting coefficient for the input pressure
centerline
A displacement of the ship
A v cross-sectional area of the vertical tank
b, distance between the centerline of both vertical
or.o~, roll, pitch, and yaw angle of the ship
and 0v
tanks
restoring moment coefficient
0, rotation angle of the tank motion
C 0. wave slope in the y direction
E viscous energy of the ship and tank fluid
v, absolute velocity of the tank fluid
Fy hydrodynamic force in the y direction
VOr~ VOt , measurement noise of Or, 0,, Or, and O,
hGM metacentric arm of ship roll
VOr,
L virtual moment of inertia of the ship without tank
and v0,
fluid
P, density of the tank fluid
l girthlike coordinate along the center line of the tank
r,, time constant of the impeller system
L estimator gain
Mr hydrodynamic mome0t about the x direction
ms virtual mass of the ship without tank fluid Subscripts
Pd demanded control command h heave motion of the ship
Pp pressure difference between the face and the back of p pitch motion of the ship
the impeller r roll motion of the ship
s sway motion of the ship
t rotation motion of the tank
y yaw motion of the ship
Address correspondence to: W.-J. H s u e h
R e c e i v e d for p u b l i c a t i o n o n M a y 7, 1996; a c c e p t e d on
Jan. 17, 1997
78 W.-J. Hsueh and Y.-J. Lee: Activated antiroll tanks
y
--
-: ,
I ZSZ-0, motions of the ship, respectively, and the subscript t
refers to the tank. By coordinate transformation, the
last item of the kinetic energy expression can be de-
scribed as a function of the heave, sway, and roll mo-
tions. The dynamic equations of ship and tank can then
J
be derived from the Lagrangian equations
d(ar~ 0T aV aE
Fig. 1. Coordinate system of the activated antiroll tank . . . . +__ ~/
dt~cgOi ) ~gqi Oqi + =Qi (6)
where xi(k) is the ith state value of the system, ct~is the
weight of the state x~(k), and fl is the weight of the input
( >T
pressure.
Bp= 0000 1
By the theory of calculus of variation, the optimal
control input leads to
_ ,
a.haM A ar~2Ptgb~A~ + Bp PcBp) BTp~X(k) (14)
=
J Kalmangain I L(k+O
calculation
Stabilizing
RollAngle
Moment
e-(k+l) 2(k+l) Tank Angle
Time ] Measurement
update update J Impeller J
l Pc(k),2(k) f
i co=,,e I- Estimator
where
Step 1: time update
= 0, Or 0,/ (23)
(24)
"0,
VOr,VO,,9o; and 90r are measurement noise. Step 2: Kalman gain calculation
The mathematical model of this observer can be de-
scribed by (25)
ArpiZ - Py + -BwQIBJ
S i m u l a t i o n and results
--ArPICT(CPICr + RI)-'CPIAr =O (22)
For real-time control and estimation, the calculation A 1174 tonne ship was studies to illustrate the efficiency
of the Kalman filter can be expressed by a recursive of rolling reduction by the proposed method. The prin-
formulation. The recursive algorithm consists of three cipal parameters of the sample ship are given in Table 1.
steps: time update, measurement update, and the Three types of ship configurations (ship without tanks,
Kalman gain calculation described as follows: ship with the passive antiroll tank, and ship with the
82 W.-J. Hsueh and Y.-J. Lee: Activated antiroll tanks
activated antiroll tank system) are considered to com- assigned, the higher the efficiency of the rolling reduc-
pare the efficiencies of devices of the activated type and tion. Figure 5 shows that the amplitude of tank angle
the passive type. In the example, the tanks used in the will be enlarged if the value of q~ is increased. F r o m
activated system and the passive system are the same. the figure, we see that the effect of q2 increasing is
The section areas of the vertical and horizontal channels similar to that of q~ decreasing. To know the effect of
are 8 m 2 and 1.8m 2, respectively. The water height is the system for different r values, the system types 1, 2,
3.5m. The dimensions of the antiroll tank are deter- and 3 are modified by changing the r values f r o m 0.1 to
mined by a high efficiency condition in the passive type. 1.0, redefined as activated types 4, 5, and 6. Figures 6
To understand the relationship between the weight of and 7 illustrate that both the efficiency of the roll reduc-
each of the states and the p e r f o r m a n c e of the activated tion and the response of the tank angle will be de-
system, different weight selections in each state are con- creased if r is increased. Figure 8 shows that the
sidered as follows: (1) activated type 1, ql = 100, q2 = 1; pressure difference should be increased if a higher roll
(2) activated type 2, ql = 100, q2 = 10; (3) activated type reduction effect if required. For simplification, only the
3, ql = 1000, q2 = 1 (where ql is the weight for the states type 1 activated system is used in the following com-
of the roll angle, and q2 is that for the tank angle). The parative studies.
weight of the other states is set to zero, and the weight In the free decay test, if the initial conditions of the
for the input signal r is set to 0.1. ship are a roll angle of 10 ~ and zero roll velocity, the roll
Figure 4 is the roll response in a regular b e a m wave. response for the different system types is given in Fig. 9.
W e see that the resonance can be suppressed, but two The rolling of the ship without tanks has a significant
minor peaks are still generated at frequencies near oscillation, with a peak overshoot of 7.5 ~ and a settling
0.04Hz and 0.11Hz in the passive type. However, the time of 60 for the 1 ~ criterion. This result is decreased to
frequency responses of rolling in all activated type sys- a peak overshoot of 3 ~ and a settling time of 25 s by the
tems are decreased to an equivalent value over a wide passive tank. If the type 1 activated antiroll tank is used,
frequency range. Moreover, the larger the value of ql the peak overshoot will be reduced to 1.5 ~ and settling
time is less than 6s. The largest value of the tank angle
of the activated type is 15 ~, which is about 150% of the
Table 1. Principle parameters of the sample ship initial listing.
In the case of the ship in the irregular seaway, a
Lpp 66.7 m
Breadth 9.7 m significant wave height, hw, of 7 m in b e a m sea is first
Draft 3.2 m considered. Figure 10a is the wave elevation history
Depth 6.1 m obtained f r o m the International Towing T a n k Confer-
cb 0.56 ence ( I T T C ) spectrum formulation. 15In the period from
GM 1.0m 0 to 200 s, the m a x i m u m p e a k value of the roll response
Displacement 1174 tonne
in the activated type 1 case is 7 ~ This is less than the
6.00
without tanks
5.00
& passive type //
0 t ~
-- --- activated1 / /
-~ 4.00
. --s ac',Jvated2 / /
.~ 3.00
2.00
1.00
0.00 I I I I I I I I I I I I I I I I I
5.00
i'~,
passive type
/
4.00 - /
....... activated 1 /
0
activated 2 /
> /
nn -- -- . . . . activated 3 /~ ,.-,
m 3.00
__o
A/'-'" t
t'-
2.00
\ , '\
t- ....... kk , 4\
I-- 1.00 - - . ....................
..... \ "..", ,,
......... 7 k --~, '\
0.00 I I I I I I I I I I I I I I I I I
o_ 3.00
0
> activated 4
m 2.00 - -
....... activated 5
activated 6
1.00
(1~ __
0 0.00 i i i i I I I I i I i I I I I I
n."
0.01 0.10 1.00 Fig. 6. Roll r e s p o n s e of t h e a c t i v a t e d sys-
W a v e f r e q u e n c y (Hz) tems in a regular b e a m wave
5.00
activated 4
Q. 4.00 - -
....... activated 5
0 ,/\,
o~ ..... activated 6 / \
(D
> / \
3.00
/ i
O~
C
2.00 -
./. ...... \ ',,
.................. -J" '\\
1.00
.-- '\
0.00 I I I I I I I I I I I I I I I I I
7.00
6.00
activated 1 / / ~
....... activated 2 /
5.00
4.00
(JO
0
3.00
2.00 m
U)
1.00 - -
13.
0.00
10.0 "', Fig. 11a-d. Most significantly, the rolling response will
be magnified in the passive type. The reason for the
magnification is that the dominant part of the wave
0.0--
spectrum is distributed near the frequency 0.15Hz,
o
n,"
-5.0 -- which is close to one of the resonance peaks in the
-10.0 i i i i i
frequency response function.
0.0 10.0 20.0 30.0 40.0 50.0
Since the damping coefficient ~, of the rolling dy-
T i m e (sec) namic equation is the dominant factor influencing the
rolling action, three cases of the damping coefficient for
20.0 the ship roll model (model 1, ~ = 0.4; model 2, ~ = 0.2;
~ 10.0 4 ;~ model 3, ~, = 0.05) are considered to compare the sensi-
tivity property of this controller. F r o m the results shown
is Fig. 12a-c, we find that the rolling responses of the
-• -10.0 ship without an antiroll tank system of with a passive
I--
-2o.0 I ' I
tank system are very sensitive to damping coefficient
' I '
0.0 10.0 20.0 30.0 40.0 50.0 changes. However, a very low sensitivity to the changes
T i m e (sec) in the ship damping are observed in the activated sys-
tem. In addition, three cases of the tank model with
Fig. 9. Free decay during rolling test. For roll angle: solid line,
without tanks; short dashes, passive type; long dashes, acti- different damping coefficients ~, (model 1, ~', = 0.4;
vated type. For tank angle: solid line, passive type; dashes, model 2, ~, = 0.1; model 3, ~, = 0.05) with respect to the
activated type designed model of ~, = 0.2 are analyzed to determine the
sensitivity of the system to tank modeling errors. Figure
13a,b shows that the performance of the passive antiroll
maximum peak value in the passive type of 13 ~ and in tank system is affected significantly by the dynamic
the unstabilized type of 27 ~ If the percentage roll re- characteristics of the tank. However, the performance
duction is defined as the difference between one and the of the activated tank in this scheme is almost indepen-
ratio of the mean value of one-third of the highest roll dent of the tank dynamics.
angle, the roll reduction of the activated type system of
70% is better than the 38% of the passive system, as
shown in Fig. 10b-d. From Fig. 10e,f, we see that the Discussion
peak value of the tank angle response in the passive
system is 24 ~ but this value increases to 40 ~ in the An optimal controller with a Kalman estimator for an
activated system. When a mean sea condition with a activated antiroll tank system was studied. The dynam-
significant wave height of 2 m is considered, the roll ics of the ship motion, tank flow, and variable-pitch
reduction by the activated type is limited, as shown in impeller were determined, and the control law was de-
W.-J. Hsueh and Y.-J. Lee: Activated antiroll tanks 85
5,0 40.0
~D 0.0--
-2.5 --
o
n~
-5o [ ' I ' I ' -40.0
0.0 40.0 80.0 120.0 160.0 2000 0.0 40.0 80.0 120.0 160.0 200.0
Time (sec) Time (sec)
40.0 .
40.0
;= ~ 0.0
is=
o
It"
I-.-
-40,0
-40.0
0.0 40,0 80.0 120.0 160.0 200.0
0.0 40,0 80.0 120.0 160.0 200.0
Time (sec)
Time (sec)
40.0 40.0
~ 0.0 0.0
o
F-
-40.0 -40.0
0.0 40,0 80,0 120.0 160.0 200.0 0.0 40.0 80.0 120.0 160.0 200.0 f
Time (sec) Time (sec)
Fig. 10. Time history of the test in irregular waves of h,, = 7.0m. a wave elevation, b roll angle without tank, c roll angle with
passive tank, d roll angle with activated tank, e tank angle (p~/ssive), f tank angle (activated)
~
~
2.0
0.0
m
N
0.0
10.0
-10.0
0.0 40.0 80.0 120.0 160.0 200.0 0.0 40.0 80.0 120.0 160.0 200.0
Time (sec) Time (sec)
10_0 10.0
O,D
~ 0.0 ~ 0.0
m
o 0
re"
-10.0 -10.0
0.0 40.0 60.0 120.0 180.0 200.0 0.0 40.0 80.0 120.0 160.0 200.0
Time ( s e c ) Time (sec)
Fig. 11. Time history of the test in irregular waves of h w = 2.0m. a wave elevation, b roll angle without tanks, e roll angle with
passive tank, d roll angle with activated tank
rived. F r o m the derived results, w e can select the weight dure. F r o m the analysis results, w e s e e that the rolling
of each of the states to obtain a suitable control law for r e s p o n s e of a ship with an activated stabilizer can be
the controller according to the required efficiency for r e d u c e d to a l o w level o v e r a w i d e f r e q u e n c y range
roll reduction. Since the o p t i m a l control law is stable, it under a regular w a v e load. T h e efficiency of the acti-
is u n n e c e s s a r y to c h e c k the stability in the design proce- v a t e d s y s t e m is also superior for ship roll r e d u c t i o n in an
86 W.-J. H s u e h a n d Y.-J. Lee: A c t i v a t e d antiroll t a n k s
40.0 . 40,0
0l
0.0
IS
o
nr
-40.0
-4o.o I ' I ' I ' I ' I '
0.0 40.0 80,0 120.0 160.0 200.0 0.0 40.0 80.0 120.0 160.0 200.0
Time (sec) Time (sec)
40.0 40.0
c~
-~ o.o 0.0
o o
-40.0 -40.0
0.0 40.0 80.0 120.0 160.0 200.0 0.0 40.0 80.0 120.0 160.0 200.0
T i m e (sec) T i m e (sec)
40.0
Fig. 13. T h e effect of the antiroll t a n k s for different t a n k
models, a passive tank, b activated tank. Solid line, short
dashes, a n d long dashes r e p r e s e n t t a n k m o d e l s 1, 2, a n d 3,
0.0
respectively
o
n,
-40.0