Plenary Talk Presented at the 8th IFAC Conference on Control Applications in Marine Systems,

Sept 15th-17th, Rostock, Germany

Tristan Perez*, *** and Mogens Blanke**,***

* School of Engineering, The University of Newcastle, AUSTRALIA
** Dept. of Elec. Eng. Automation and Control, Danish Technical University, DENMARK
*** CeSOS, Norwegian University of Science and Technology (NTNU), Trondheim NORWAY

8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Motivation
The effects of roll on ship performance became more
noticeable in the mid-19th century when:

  Sails were replaced by engines

  Broad arranges changed to turrets
Many devices have been proposed leading
W. Froude (1819-1879)
to interesting control problems

Performance can easily fall short of expectations because of

deficiencies in control system design due to
  Fundamental limitations due to system dynamics
  Limited actuator authority
  Disturbances with large changes in spectral characteristics

2 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Outline
  Ship Roll Control Devices
  Working Principles,
  Performance,
  Historical aspects
  Key aspects of ship roll dynamics for control design
  Models,
  Wave-induced motion
  Simulation and control design models
  Control System Design
  Objectives,
  Fundamental limitations,
  Control strategies
  Research Outlook
  Unsolved problems, new devices

3 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

I - Ship Roll Control Devices
Performance,
Working Principles,
and Historical Aspects

8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Ship Performance and Roll Motion
Ship roll motion

  Affects crew performance

  Can damage cargo
  May prevent the use of on-board equipment

From a ship operability point of view is necessary to

reduce not only roll angle but also roll accelerations.
5 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
Bilge Keels
  10to 20% roll reduction (RMS)
  Low maintenance
  No control
  No occupied space
  Low price easy to install
  Increase hull resistance when damping is not needed
  Not every vessel can be fitted with them (ice-breakers)

6 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

U-tanks
  40 to 50% roll reduction (RMS)
 Active/Passive
  Independent of the vessel speed
 Anti heeling
  Heavy
  Occupy large spaces
 Affects stability due free-surface
Intering Rolls-Royce

7 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Gyro-stabilisers
  60% to 90% roll reduction (RMS)
  Performance independent of the vessel speed
1906 1914 to 1925 1914,1924
Torpedo Boat active control Twin Gyros Today

Schlick Sperry Fieux Halcyon Time

(Germany) (USA) (France) Ferretti
Sea Gyro
Shipdynamics
Seakeeper

Japanese aircraft carrier ‘Hosho’ 95% RR with

Sperry Gyro

Conte di Savoia 1932

8 8th IFAC CAMS, Sept 15th-17th, Rostock,

Germany
Fin-stabilisers
  60% to 90% roll reduction (RMS)
  Performance depends on speed
  Control is important for performance
Rolls-Royce
  Easy to damage
  Most Expensive stabiliser
  Can produce noise affecting sonar
  Dynamic stall

9 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Rudder Roll Stabilisers (RRS)
  40% to 70% roll reduction (RMS)
  Performance depends on speed
  Control is important for performance
  Special rudder machinery
  Fundamental limitations in control due to
non-minimum phase dynamics (NMP)

10 8th IFAC CAMS, Sept 15th-17th, Rostock,

Germany
Historical Aspects

T. Perez (2005) Ship Motion Control, Springer

11 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

II - Key Aspects of Ship Roll Dynamics
for Control Design
Models
Ocean Environment
Wave-induced motion

8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Dynamics of Roll Motion
General model form to describe ship motion:
! n "
pb/n
η! = [N, E, D, φ, θ, ψ]T .
Θ
! "
ṗb/n
n b
ν! = [u, v, w, p, q, r]T .
ω bb/n

Kinematic model: η̇ = J(η)ν

Kinetic model: Mν̇ + C(ν)ν + D(ν)ν + g(η) = τ

13 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Models for Control Design
For control design, output disturbance models are usually adopted:

Motion Time
series

Control
Forces
1DOF, 4DOF
Force to motion

For roll motion control problems this model is simplified to

  1DOF: roll
  4DOF: surge, roll, sway, and yaw

14 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Dynamics of Roll Motion—1DOF Model
Consider only roll motion,

Kinematic model: φ̇ = p,
Kinetic model:
Ixx ṗ = Kh + Kc + Kd
Hydrodynamic Control Disturbance
moments moments moments

Hydrodynamic Moments:

Kh ≈ Kṗ ṗ + Kp p + Kp|p| p|p| + K(φ)

Potential Viscous Hydrostatic

Effects Effects Effects

15 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Dynamics of Roll Motion—4DOF
Motion Variables: η = [φ ψ] , T

ν = [u v p r]T ,
τ i = [Xi Yi Ki Ni ] , T

Kinematic model: φ̇ = p, ψ̇ = r cos φ ≈ r

Kinetic model: MRB ν̇ + CRB (ν)ν = τ h + τ c + τ d

τ h ≈ −MA ν̇ − CA (ν)ν − D(ν)ν − K(φ)

16 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
Ocean Environment
Usual assumptions for sea surface elevation ζ(t):
  Zero-mean
  Gaussian (depth dependent)
  Narrow banded
  Stationary (20min to 3 hours)
All necessary information is then in the wave elevation power spectral
density (Sea Spectrum):

Φζζ (ω, χ)

17 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Sailing Condition and Encounter Spectrum
Encounter Frequency
Sailing Condition
  Speed
  Encounter angle

ω2 U
ωe = ω − cos(χ)
g
Φζζ (ω)
Φζζ (ωe ) = !! !
!
!1 − 2ωU
g cos(χ)!

18 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

 Wave-induced Motion Motion
Seakeeping Model
Motion spectrum

Sζζ (ω) F(jω,χ,U) G(jω,U) Sηη (ω)

Wave spectrum Wave to force FRF Force to motion FRF
Motion Time
series

F(jω, χ) = [F1 (jω, χ), . . . , F6 (jω, χ)]T

 
G11 (jω, U ) · · · G16 (jω)

G(jω, U ) =  .. .. 
= (−[M + A(ω)] ω 2
+ jωB(ω) + G)−1
. .  RB
G61 (jω, U ) · · · G66 (jω, U )
6
!
Roll RAO: H4 (jω, χ, U ) = G4k (jω, U )Fk (jω, χ, U )
k=1

19 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Example Roll RAOs Naval Vessel @15kt
6
!
H4 (jω, χ, U ) = G4k (jω, U )Fk (jω, χ, U )
4.2 Seakeeping Theory Models 73

k=1
&!*+,
'#*+,
(!*+,
-#*+,
.!*+,
$!#*+,
$%!*+,
) $&#*+,
|H4 (ω, U, χ)|

$#!*+,
$(#*+,
(

'

%
&

! %"#
%!!
%
$#!
$"#
$!! $ ω
χ
#! !"#

! !

Fig. 4.6. motion RAO Roll of the naval vessel benchmark at 15 kt for different
20 encounter angles.
8th IFAC CAMS,The
Septmotion RAO
15th-17th, are given
Rostock, as a function of the wave frequency.
Germany
III – Motion Control Design

Objectives,
Performance Limitations,
and Control Strategies

8th IFAC CAMS, Sept 15th-17th, Rostock,

Germany
Control Design and Performance Limitations
From the ship performance point of view, the control objectives are
  Reduce roll angle
  Reduce roll accelerations
φol
Kc
Actuators Ship φcl
u
Control

Output sensitivity Roll spectrum relations

φcl (s)
S(s) ! Φcl (ω) = |S(jω)|2 Φol (ω)
φol (s)
22 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
Control Design and Performance Limitations
If the closed-loop system is stable minimum phase and strictly proper,
the following Bode integral constraint applies
! ∞
log |S(jω)| dω = 0
0

Φcl (ω) = |S(jω)|2 Φol (ω)

|φol (jω)| −| φcl (jω)|

RR(ω) = 1 − |S(jω)| =
|φol (jω)|

23 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Example Performance Limitations
Gyrostabiliser (Perez & Steinmann, 2009)

24 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Non-minimum phase Dynamics
In some cases, the location of the actuators may result in NMP
dynamics (with a real zero at s=q.)

Then, we have a Poisson-integral constrain:

! ∞
log |S(jω)| W (q, ω)dω = 0
−∞

q = σq + j0
σq
W (q, ω) = 2
σq + ω 2
25 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
Non-minimum phase Dynamics
::;*4+5.*8.,8323532<*=>?:@*!*AB%CDE+5FD$G#!
#&"

Example Blanke
et al (2000)

#
'())*+,-).*'./0123(,

!&"

!
! " #! #" $! $" %!
4+5.6.'3(/*78.19

26 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Energy of Wave Excitation
The energy of the wave excitation changes with the sea-
state and sailing conditions (speed and heading)

For example a change in encounter angle can shift

the energy significantly:
φ(t)

27 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

 τ2rudder ≈ L ing
b
τ4rudder ≈ zCP L (8) the
τ6rudder ≈ xbCP L, is a

Fin Stabilisers (minimum phase case) •

•
where xbCP and zCPb
are the coordinates of the cen- •
ter of pressure of the rudder (CP ) with respect to
IDOF model including fin forces:the b-frame. The center of pressure is assumed to The
be located at the rudder stock and half the rudder with
and
φ̇ = p,
span. The rudder data is shown in Table A.4.
bas
For the stabilizer fins, the center of pressure is (Pe
[Ixx + Kṗ ] ṗ + (Kp + 2 rf Klocated
α U ) p + K
halfway φ φ =
along K
the −
span
w 2U
of
2
theKfin.
αα The
coordinates of the center of pressure with respect
to the b-frame are given by the vector rbCP —see
Control issues: Figure 1.
for
•  Parametric uncertainty
CG
•  Sensitivity-integral constraints
ob yb
CP
rbCP y!
Control strategies: y !! τ̄in
zb
•  PID, Hinf z !! z ! whe
and
%
Fig. 1. Reference frames used to compute fin
28 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
ωn
forces.
For
Fin Stabilisers and NMP Dynamics
If the response from the fins is NMP, then the IDOF cannot be used, in
this case roll-sway-yaw interactions need to be considered to avoid
large roll amplifications at low frequencies.

Observations about fin NMP dynamics were made by Lloyd (1989).

29 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Fin Stabilisers – Dynamic Stall
Experimental results of Galliarde (2002) (MARIN, Ned)

CL [-]
CL [-]

αe [deg] αe [deg]
φ [deg]

Time [s]
30 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany
Fin Stabilisers
Perez & Goodwin (2003): MPC constraints effective angle of attack.

31 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Rudder Roll Damping
  Potential discovered from autopilot without wave filter (Taggart 1970)
  Results reported by van Gunsteren in 1972 (the Netherlands)
  Cowley & Lambert (1972) used roll fbk, reported yaw interference.
  Carley & Duberley (1972) integrated rudder-fin control
  Carley (1975) & Lloyd (1975) recognise limitations of NMP dynamics
  Baitis et al.(1983) highlighted need of adaptation
  Advent of Computers in 1980 resulted in several successful results
  Netherlands, Denmark, Sweden, United Kingdom
  Hearns & Blanke (1998) limitations using the Poisson Integral
  Perez (2003) analysed min variance and RR vs yaw interference

32 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations (Hearns & Blanke, 1998)

! " Θσ (ω1 ,ω2 )

1 π−Θσ (ω1 ,ω2 )
!S(jω)!∞ ≥
α1

If the controller does not

adapt, we can easily
have roll amplification

33 33
Performance Limitations – Perez et al. (2003)
Limiting Optimal Control with full knowledge of wave spectrum:

J = E[λφ2 + (1 − λ)(ψ − ψd )2 ]

Minimum variance case (λ=1): E[φ ] ≥ 2 q Φφφ (q)

2

34 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Actuator limitations on RRD
  Actuator rate limits may affect performance

Maximum required here

Maximum attained
here

35 35
AGC – Van Amerongen et al (1982)

36 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations – Perez et al. (2003)
IVC OCP: Rudder angle RMS limited to 15deg

37 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations – Perez et al (2003)
IVC OCP: Rudder angle RMS limited to 15deg

38 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations

  At low encounter frequencies,

  RR vs yaw interference can be large
  MNP dynamics limits RR achievable performance.

  At high encounter frequencies,

  Limitations of rudder machinery usually dominate RR
achievable performance.

39 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations (quatering seas RR 40%)

40 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Performance Limitations (beam seas RR 64%)

41 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Control Strategies

  PID,
  Hinf
  Loop shaping
  Adaptive LQG
  Model Predictive Control
  Switched control
  Nonlinear control

42 42
Changes with Speed (Blanke & Christiansen 1993)
cφδ (1 + sτz1 )(1 − qs )
  Rudder to Roll TF: Gφδ (s) = s2
(1 + sτp1 )(1 + sτp2 )( ω2 + 2ζp ωsp + 1)
p
*+,-./01 ;+,-./01
&'(
!&'(
&'%#

!&'(:
&'%!

Changes in dynamic &'%) !&'9

response due to speed &'%(

!&'9:
coupled with changes in &'%
! " # $ %& %% ! " # $ %& %%
disturbance spectrum 23//4-56728 23//4-56728

results in a strong need ;+,-31</%

!&'9
;+,-31</(

!&'&(:
for adaptation. !&'&9
!&'9:

!&')
!&'&9:
!&'):

!&'&) !&':

!&'&): !&'::

! " # $ %& %% ! " # $ %& %%

23//4-56728 23//4-56728

43 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

H2 – Optimal Design
Φφφ (ω) = Hd∗ (jω)Hd (jω)
E(φ2 ) = !Hd − QV !22 V = Hd Gδφ

−1 −
− −1 Hd Hd
Cδφ (s) = (Gδφ ) s−q −1 −
1 − s+q Hd Hd
s+q s+q
G−
φδ (s) = Gφδ (s) Hd (s) = Hd− (s) + Hd+ (s)
s−q s−q
This shows the strong dependency on sea state, sailing conditions and
changes in the vessel model --- adaptation is required.

44 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Direct Sensitivity Specification (Blanke et al. 2000)
Targeted (Desired) Sensitivity:

s2
ωd2
+ 2ζd ωsd + 1
Sd (s) = βs S(s) = (1 + Cδφ (s)Gφδ (s))−1
(1 + βωd )(1
s
+ ωd )

1− s
q
s
Cδφ (s) = (Sd (s)−1 − 1)G−1 P (s)−1
φδ(s) 1+ s
q

45 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Direct Sensitivity Specification (Blanke et al. 2000)
s2
k1 s ωp2 + 2ζp ωsp + 1 (1 + sτ )(1 + sτ )
p1 p2
s
Cδφ (s) = s2
+ 2ζd ωs + 1 (1 + sτz1 )(1 + q )
s
cφδ ωd ωd2 d

::;*4+5.*8.,8323532<*=>?:@*!*AB%CDE+5FD$G#!
#&"
Successful performance
was demonstrated for the
SF300 patrol boat of the
Danish Navy #

'())*+,-).*'./0123(,
!&"

!
! " #! #" $! $" %!
4+5.6.'3(/*78.19

46 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Gyro Stabilisation
!ωs
Spin angular momentum

!τm !˙
α Kg = Is ωs
!τp

−!τg !τw , φ̇

I44 φ̈ + B44
l
φ̇ + B44
n
|φ̇|φ̇ + C44 φ = τw − Kg α̇ cos α,
! "# $
τg

Ig α̈ + Bg α̇ + Cg sin α = Kg φ̇ cos α +τp

! "# $
τm

47 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Gyro as a Damper

τp : α̇ ≈ q φ̇

I44 φ̈ + (B44
e
+ nKg q)φ̇ + C44 φ = τw

Increase of roll damping

n- number of gyros due to the gyrostabiliser
q-constant

48 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Gyro-control Design
Gyro full state control: τp = −Ka α − Kr α̇

Roll to precession Transfer function:

α(s) α̇(s) Kg s Bg! = Bg + Kr ,

G(s) = = = ,
φ(s) φ̇(s) Ig s + Bg s + Cg
2 ! !
Cg! = Cg + Ka
The control design objective becomes:

Ship
τp : α̇ ≈ q φ̇ → |G(ω)| ≈ q, τg
φ̇
α̇
∀ω ∈ Ω arg G(ω) ≈ 0 gyro

This can be achieved by forcing two real poles on G(s) (Perez Steinman, 2009)

49 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Irregular Sea Performance (Perez & Steinamnn 2009)
Roll Reduction in RMS: 82%
Vessel displacement: 360 ton
Gyro unit weight: 13 ton

50 MCMC Conference, Sept 2009 50

IV - Research Outlook

New Devices
Unsolved Problems

8th IFAC CAMS, Sept 15th-17th, Rostock,

Germany
New Devices
Flapping fins (Fang et al. 2009 Ocean Engineering 36)

Added mass Form and Vortex drag

52 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

New Devices
Rotor Stabiliser (Quantumhydraulic.com)

53 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Parametric roll

Parametric roll is an auto-parametric resonance phenomenon

whose onset causes a sudden rise in roll oscillations.

54 Two-day Tutorial, Ancona, Italy, Sept 2010

Conditions for PR

The following conditions can trigger the effect:

  Hull designs with significant bow flare and hanged stern

  Sailing in longitudinal waves
  Wave length close to the length of the vessel
  Encounter frequency is twice the roll natural frequency
  Low roll damping

Parametric roll is particular of modern container ships,

cruise ships, car ferries, and fishing vessels.

55 Two-day Tutorial, Ancona, Italy, Sept 2010

Restoring in Waves – Dynamic Stability
K(φ) = ρ g ∇ GZ(φ)

56 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Videos

57 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Conclusion

  For over 100 years different devices have been proposed to control
roll motion (reduction)

  Most of these devices require control systems to work

  Control design is not trivial due to

  Fundamental limitations
  Widely-varying disturbance characteristics
  Actuator limited authority

58 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

Current State and Research Outlook
  New developments are being put forward by yacht industry
  Gyros, Flap fins, rotor stabilisers (need for zero speed performance)

  Navies are still considering RRD

  Submarines need roll control at low speeds when in the surface

Research outlook

  Adaptation to sea state and sailing conditions still remains an issue

  New devices with interesting hydrodynamics
  Integrated vessel and roll control design (performance prediction)
  Parametric roll

59 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany

 Thank You

Death Roll of Sailing Vessels

60 8th IFAC CAMS, Sept 15th-17th, Rostock, Germany