Professional Documents
Culture Documents
Adaptive Roll Stabilization of Ships by U-Tanks Based On A Novel Control Strategy
Adaptive Roll Stabilization of Ships by U-Tanks Based On A Novel Control Strategy
strategy
Honghan Zhang, Xinqian Bian, Jiajia Zhou, Tao Chen, Xun Zhang
College of Automation, Harbin Engineering University, Harbin, 150001, China
E-mail: zhanghonghan2008@163.com
Abstract: The application of passively-controlled U-tanks which have frequency adaptive function was
introduced in this paper. The model of air-liquid flow in the U-tanks was established. On the basis of the model,
relationship between the sectional area of air valve and fluid flow characteristics was researched. The optimized
fluid motion was controlled by continuously turning the section area of air duct, which is calculated in
generalized predictive control strategy. In this way, the fluid flow can be tuned to reduce the ship’s roll with the
wave disturbance frequency varying. Simulation studies showed that the novel strategy can achieve more
effective roll reduction compared with traditional intermittent control strategy.
Key Words: ship motion, anti-roll tank, adaptive control, generalized predictive control, passively-control
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
fluid flow was researched. A predictive control strategy was Vi
used to estimate optimized parameters of fluid motion in Where v i = is the specific volume of the air. Vi is the air
mi
real time. And the opening area of the air valve can be
continuously controlled in order to adjust the fluid flow in volume above the water and mi is the mass of the air. C a is
the tank, so as to achieve more effective roll reduction in a constant determined by the initial conditions. γ is the
extended disturbing frequencies. adiabatic air constant. In each side of the tank the rate of
Model of Ship and passively-controlled anti-roll change in pressure can then be written as
tank system ∂Pi ∂Pi ∂Pi
= Vi + m (i=1,2) (3)
∂t ∂Vi ∂m i
2.1 Model of Ship-Tank System According to Eq.(2),
Equipped with U-tanks, the ship’s roll and the tank ∂Pi mγ
= −γC a γi +1 (i=1,2) (4)
fluid flow affect each other. The coupled system can be ∂Vi Vi
described by six degrees of freedom equation [5,7]. For
∂Pi m γ −1
simplification, this work treats the ship and tank as a single = γC a i γ (i=1,2) (5)
coupled system ignoring little impact by other degrees of ∂mi Vi
freedom motion. The differential equations of the ship and
rolling system with tank stabilizer is described as γ P P
⎧ 1 m = A Ph ρ h ( e ) 2 / γ [1 − ( e ) (γ −1) / γ ] (6)
2 γ −1 Ph Ph
⎪φ + 2v s ω sφ + ω s φ + αδ + βδ = J K w
⎪ s (1)
⎨
1
Where Ph and Pe are the high and low pressure in each side
⎪δ + 2v ω δ + ω δ + ηφ + σφ =
2
ΔPS 0 R m
⎪⎩ t t t
Jt of the valve respectively. ρ h = h is the air density of the
Vh
Where, φ is ship roll angle. δ is tank fluid angle. Js is the higher pressure side; A is the sectional area of the valve.
ship roll inertial moment including the liquid masses. Jt is Then the pressure difference between the two sides of the
the liquid motion mass inertial moment. v s , v t express the tank is
ΔP = P1 − P2
damp coefficients respectively. ω s , ω t express the (7)
connatural frequencies of ship roll and the liquid vibration As expressed by Eq.(1), the fluid oscillation motion can
respectively. α , η and β , σ represent the coupled inertial be assumed to a second order motion when the sectional area
of the valve is settled. The ARMA model of fluid oscillation
moment and the coupled resume moment coefficients of the
can be described as
ship and the tank. K w is the disturbance moment of the
δ k = − a 21δ k −1 − a 22 δ k − 2+b21φ k −1 + b22φ k − 2 + b20 (8)
wave roll. R is the horizontal distance from the center of the
The model coefficients can be found by the recursive
side tank to the longitudinal center plane of the ship. S 0 is
least squares algorithm. The damped frequency ( ω̂td ),
the sectional area of the free liquid. ΔP is the pressure
difference of compressible air over both side of the tank. non-damped natural frequency ( ω̂t ) and the damped
Valve Air duct coefficient can be calculated from
1 a
ωˆ td = arccos( 0.5 ln(21− a22 ) ) (9)
Δt 2e
R [ln( − a22 )]2
ωˆ t2 = ωˆ td2 + (10)
4( Δt ) 2
δ
ωˆ td 2 (11)
vˆ t = 1 − ( )
Control System ωˆ t
Where Δt is the sampling time. Using the proposed
method, the fluid oscillation parameters can be off-line
identified with respect to different sectional area of the
Fig.1 Structure of U-tank valve, and the relationship curve between A and ω̂ t as well
2.2 Model of the Compressible Air Dynamic as to v̂ t can be obtained.
When the liquid has a oscillation motion, the air
between the liquid surface and the top wall of the tank can
3. Control strategy for ship’s stabilization by
be compressed or expanded, which is known as a changeful passively-controlled anti-roll tank
compressibility process. Subscript number 1 and 2 was used The block diagram of the passively-controlled anti-roll
to represent the parameters variates of the left and the right tank system is shown in figure 2. The ship-tank dynamics
rank respectively, and the air state in both side of the tank system is included in the dash dot line square. The control
can be obtained by the air state equation of a changeful signals are produced by the tank controller according to the
process, ship roll motion state as well as the liquid flow.
Pi v i γ = C a (i=1,2) (2)
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
where: A( z −1 ) = 1 + a11 z −1 + a12 z −2 + a13 z −3 + a14 z −4 ;
B( z −1 ) = b10 + b11 z −1 + b12 z −2 + b13 z −3 + b14 z −4 ;
C ( z −1 ) = c10 + c11 z −1 + c12 z −2 + c13 z −3 + c14 z −4 ;
Δδ (k ) = δ ( k ) − δ (k − 1) ;
A ( z −1 )
;
= (1 − z −1 ) A( z −1 )
= 1 + a11 z −1 + a12 z −2 + a13 z −3 + a14 z − 4 + a15 z −5
a i = a i − a i −1 ,( 1 ≤ i ≤ 4 ); a15 = −a14 。
Fig.2 Block diagram of passively-controlled tank stabilizer
{w(k )} is a white noise series, and
E{w( k )} = 0 , E{w 2 ( k )} = 1
3.1 Intermittent flow control strategy
The Diophantine equation can be introduced after the
The success of the intermittent flow control strategy is
dependent on the correct timing of the opening of the CARIMA model of the system was built:
valves, and this will be determined by the flow C ( z −1 ) = A ( z −1 ) E j ( z −1 ) + z − j G j ( z −1 ) (16)
characteristics of the tank. For an oscillating U-tank with a Where
well-defined natural frequency the timing is relatively E j ( z −1 ) = 1 + e j ,1 z −1 + " + e j , j −1 z − ( j −1)
straightforward if the next ship roll period, and hence the G j ( z −1 ) = g j , 0 + g j ,1 z −1 + " + g j , 4 z −4
time of the next peak, can be adequately estimated [1].
According to the adaptive control theory, the
In order to ensure the flow of tank ‘straddle’ the ship’s
relationship of j step forward predictive ship roll angle can
rolling peak giving an effective phase lag of 90 degrees,
passively PD control strategy can be used for intermittent be expressed as
flow control strategy of the tank stabilizer system. G j ( z −1 ) F j ( z −1 )
φ (k + j ) = φ (k ) + Δδ (k + j − 1) + E j ( z −1 ) w(k + j ) (17)
β = k1φ + k 2φ is the open signal of the valve, which is C ( z −1 ) C ( z −1 )
Eq.(16) can be converted to the other Diophantine
consisted of ship’s rolling velocity and acceleration. The
equation, shown as follow
fluid can move from the left side of tank to the right side
F j ( z −1 ) = B( z −1 ) E j ( z −1 ) = L j ( z −1 )C ( z −1 ) + z − j H j ( z −1 ) (18)
when β > 0 and it is contrary when β < 0 . When the sign
Where
of β changed, open the valve and the direction of fluid
L j ( z −1 ) = l j , 0 + l j ,1 z −1 + " + l j , j −1 z −( j −1)
flow will be reversed. In order to keep the liquid stay in one
H j ( z −1 ) = h j ,0 + h j ,1 z −1 + " + h j ,3 z −3
side of the tank which the ship moves upwards, the valve is
closed when the flow-rate drops zero or the column being The polynomial F j ( z −1 ) was divided into two parts in
filled is almost full. Eq.(18), the frontal part represents the future influence to the
ship by the tank, and the latter one is for past.
3.2 Continuous flow control strategy
Set
The equation shown in Eq.(1) represents the ship roll Δδ ( k − 1) , φ (k )
Δδ f ( k − 1) = φ f (k ) = (19)
motion with a U-tank, which can be known as a second order C ( z −1 ) C ( z −1 )
motion model with two freedom of degree. The wave
According to Eq.(17) to Eq.(19), the output as ship
disturbance K w is coloured noise which can be obtained by
rolling of the “ship-tank” system predictive model has a
a second order molding filter G w (s) : matrix format as follow:
K w (s) k1 s Y (k + j ) = LΔU (k ) + HΔU f (k − j ) + GY f (k ) + EW (k + j ) (20)
= G w (s) = 2 (12)
w( s ) s + k 2s + k3 Where
Where, k1,k2and k3 are the coefficients being correlative Y (k + j ) = [φ (k + 1), φ (k + 2), " , φ (k + N )]T is the future
with the height of wave, the heading and velocity of the ship. output vector of the ship roll;
w(t ) is a white noise. U (k ) = [δ (k ), δ (k + 1), " , δ (k + N − 1)]T is the future
Transform the first equation shown of Eq.(1) to a Laplace output vector of the desired liquid flow;
form: Y ( k ) = [φ ( k ), φ ( k − 1), " , φ (k − 4)]T is the past output
( )
( s 2 + 2ν s ω s s + ω s2 )φ ( s ) + αs 2 + β δ ( s ) =
1
Js
K w (s) (13) vector of the ship roll;
Y f (k ) = [φ f (k ), φ f (k − 1), " , φ f (k − 4)]T ;
Combining Eq.(12) to Eq.(13) ,The CARIMA model
of the ship roll motion which controlled by the tank can be ΔU (k ) = [Δδ (k ), Δδ (k + 1), " , Δδ (k + N − 1)]T is the
obtained by a z-transform operation excursion increment of the future liquid flow;
A( z −1 )φ ( k ) = B( z −1 )δ ( k ) + C ( z −1 ) w( k ) /(1 − z −1 ) (14) ΔU (k − j ) = [Δδ (k − 1), Δδ (k − 2), " , Δδ (k − 4)]T is the
That is excursion increment of the past liquid flow;
(15) ΔU f (k − j ) = [Δδ f (k − 1), Δδ f (k − 2), " , Δδ f (k − 4)]T ;
A ( z −1 )φ (k ) = B ( z −1 )Δδ (k ) + C ( z −1 ) w( k )
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
W (k + j ) = [ w(k + 1), w( k + 2), " , w( k + N )]T is random ⎧ˆ ˆ [ ˆ
⎪θ ( k ) = θ ( k − 1) + K (k ) Δφ ( k ) − ϕˆ (k )θ (k − 1) ]
white noise vector;
⎡ l1,0 0 " 0 ⎤ ⎡ h1, 0 h1,1 h1, 2 h1,3 ⎤
⎪
[ T
⎨ K ( k ) = P (k − 1)ϕˆ ( k ) λ + ϕˆ ( k ) P (k − 1)ϕˆ ( k )
−1
] (25)
⎢ l ⎥, ⎢h ⎪
l 2, 0 " 0 h2,1 h 2, 2 h 2, 3 ⎥ , 1
⎪ P ( k ) = [I − K ( k )ϕˆ (k ) ]P (k − 1)
L = ⎢ 2,1 ⎥ H = ⎢ 2, 0 ⎥
⎢ # # # ⎥ ⎢ # # # # ⎥ ⎩ λ
⎢⎣l N , N −1 l N , N −2 " l N ,0 ⎥⎦ ⎢⎣h N ,0 h N ,1 h N ,2 h N ,3 ⎥⎦
Where
⎡ g 1,0 g 1,1 " g 1,4 ⎤ ⎡ 1 0 " 0⎤ ϕˆ T (k ) = [−Δφ ( k − 1),−Δφ (k − 2), " ,−Δφ (k − 4), Δδ (k − 1), Δδ ( k − 2), " ,
⎢g g 2,1 " g 2, 4 ⎥ , E = ⎢ e 2,1 1 " 0⎥
Δδ (k − 5), wˆ ( k − 1), wˆ ( k − 2), " , wˆ ( k − 4)]
G = ⎢ 2, 0 ⎥ ⎢ # # #⎥
⎢ # # # ⎥ ⎢ ⎥
⎣⎢ g N ,0 g N ,1 " g N , 4 ⎦⎥ ⎣e N , N −1 e N , N −2 " 1⎦ θˆ = [aˆ11 , aˆ12 , aˆ13 , aˆ14 , bˆ10 , bˆ11 , bˆ12 , bˆ13 , bˆ14 , cˆ11 , cˆ12 , cˆ13 , cˆ14 ] T
Where, L ∈ R N × N , H ∈ R N ×4 , G ∈ R N ×5 , wˆ (k ) can be shown as a filter as follow:
E ∈ R N ×N ; wˆ (k ) = φ (k ) − φˆ(k k ) = φ (k ) − ϕˆ T (k )θˆ(k )
0.6
lagrangian multiplier method from Eq.24. The natural
0.5
frequency ω̂ t and damp coefficient v̂ t can be calculated
0.4
according to Eq.(9) to Eq.(11), which then can be used as a 0.3
set-point in a position controller to adjust the opening of 0.2
valve. 0.1
Recursive least square method was used to estimate the 0
parameters in Eq.(15) for achieving an adaptive adjustment 0 0.1 0.2 0.3 0.4
A (m2)
0.5
to the control system when the ship’s sailing state changed. Fig.3 Natural frequency of tank varying with
opening area of valve
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
high frequency, the continuous flow tank achieved fair
vt
2.5
effectiveness.
unstabilized ship rolling
2 φ /° 5
stabilized ship rolling
0
1.5
-5
δ /° 5 fluid moving
0.5
0
0
0.1 0.2 0.3 0.4 0.5
A (m2) -5
0 50 100 150 200 250 300
0
In order to test the roll reduction performance of the 0 50 100 150 200 250 300
t/s
tank stabilizer with different wave disturbance frequency, Fig.5 Control characteristics of intermittent flow control strategy
choosing the ship’s heading angle to wave as 30°, 90° and
(wave height=6m,encounter angle=45°,ship velocity=18kn)
145° respectively. The simulation was carried out with 10
unstabilized ship rolling
φ / °5
irregular wave height of 6m and 3m. And the velocity of stabilized ship rolling
0
the ship is 18kn. The wave disturbance signal were made -5
V = ( A0 − A) / A0
(26) 0
consisting with only two states which are closed and open. -10
0 50 100 150 200 250 300
10
The strategy achieved good roll reduction performance for δ /°
5
fluid moving
-10
for the wave height of 3m. The results illuminate that the 0 50 100 150 200 250 300
frequency. 0
The results for the continuous flow control strategy 0 50 100 150 200 250
t/s
300
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
φ /° stabilized ship rolling Table2 Simulation Results under wave height of 3m
5 unstablilized ship rolling
No roll
Intermittent flow strategy Continuous flow strategy
wave reduction
0 Encounter
height Standard Standard Anti-rolling Standard Anti-rolling
angle(°)
-5 (m) deviation deviation performance deviation performance
0 50 100 150 200 250 300 (°) (°) (°)
V
1
valve opening
Conclusion
0.8
The model of air-liquid flow in the U-tanks was
0.6
established in the paper. The coefficient such as damp and
0.4
0 50 100 150 200 250 t/s 300 natural frequency of the tank is affected by the open size of
the air valve. The relationship between the open size of the
valve and the characteristics of fluid flow in tank is obtained
Fig.8 Control characteristics of continuous flow control strategy
based on the model established. The fluid flow parameters
(wave height=6m,encounter angle=45°, ship velocity=18kn) of the tank can be controlled by this relationship, which can
be utilized to reduce the wave disturbance for ship.
10 stabilized ship rolling A continuous flow control strategy to the valve of the
φ /° unstabilized ship rolling
air duct was achieved. The optimized parameters of fluid
0
flow in tank were calculated by utilizing GPC strategy in
-10
0 50 100 150 200 250 300
real time. Simulation results show that the proposed
10 continuous control strategy achieve more effective roll
reduction in extended frequencies compared with
fluid moving
δ /°
0 intermittent flow control strategy. And the operational
frequency of the valve, which is adjustable to the wave
-10
1
0 50 100 150 200 250 300 disturbance, is reduced obviously compared with
V valve opening intermittent control strategy.
0.8
0.6
References
0 50 100 150 200 250
t/s
300 [1] E.B.Jones, A.P.Roskilly, B.Webster. Adaptive roll
stabilization of fishing vessels. IMechE 2006,Vol.220 Part
M.
Fig.9 Control characteristics of continuous flow control strategy
[2] Christian Holden, Thor I. Fossen. A nonlinear 7-DOF model
(wave height=6m,encounter angle=90°,ship
for U-tanks of arbitrary shape. Ocean Engineering, 22-37,
velocity=18kn)
10 2012.
stabilized ship rolling
φ /° unstabilized ship rolling [3] Perez, T. Ship Motion Control. Advances in Industrial
0 Control. Springer Verlag, London, 2005.
[4] Khaled S. Youssef, Saad A. Ragab, Ali H. Nayfeh, Dean T.
-10
0 50 100 150 200 250 300 Mook. Design of passive anti-roll tanks for roll stabilization
10
fluid moving in the nonlinear range. Ocean Engineering, 177-192, 2002.
δ /° [5] Reza M., Alistair R.G. On the development of ship anti-roll
0
tanks. Ocean Engineering, 1-19, 2007
-10
[6] FENG Tiecheng; TAO Yaosen. A sudy on passive controlled
0 50 100 150 200 250 300 antiroll tank .Ship Building of China, 2(1):8-14, 1997
V
1 valve opening [7] Youssef, K. S., Mook, D. T., Nayfeh, A.H. Roll stabilization
by passive anti-roll tanks using improved model of the
0.8
tank-liquid motion. Journal of Vibration and Control, 9(1),
0.6
832-862, 2003
0 50 100 150 200 250
t/s
300
[8] Fossen, T. I., Handbook of Marine Craft Hydrodynamics and
Fig.10 Control characteristics of continuous flow control strategy Motion Control. Jon Wiley & sons, Ltd, 2011.
[9] Lai Zhichang. Research on U-tube Anti-rolling Tank and
(wave height=6m,encounter angle=135°, ship velocity=18kn) Test Rolling Table Experimentation. Harbin: Harbin
Table 1 Simulation Results under wave height of 6m Engineering University, 2001.
No roll
Intermittent flow strategy Continuous flow strategy [10] PHAN Anh Tuan, Kuniaki SHOJI. Responses of roll
Wave reduction
height
Encounter
Standard Standard Anti-rolling Standard Anti-rolling damping on anti-rolling tank devices. OCEANS'08
angle(°)
(m) deviation deviation performance deviation performance MTS/IEEE Kobe Techno Ocean'08 Voyage toward the
(°) (°) (°)
45 2.8075 1.6814 42.07% 1.4970 46.68% Future, OTO'08, 2008.
6 90 3.3713 1.4063 58.29% 1.2285 63.55% [11] Osama A. M., Ali H. N. Control of ship roll using passive and
135 2.3775 1.0016 57.54% 1.0825 54.47% active anti-roll tanks. Ocean Engineering, 1-27P, 2009.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on March 19,2022 at 05:43:52 UTC from IEEE Xplore. Restrictions apply.