Towards an unified theoretical model of ship

dynamics
Jerzy Matusiak
HUT Ship Laboratory

Introduction
Linear models of ship dynamics in waves are well established. In most cases
they result in a sufficiently accurate prediction of loads and ship motions.
Perhaps the biggest benefit of using the linear models is that prediction of
exceeding certain level of load or response can be easily derived. Analysis is
conveniently conducted in the frequency domain. The biggest shortcoming of
the linearity assumption is that it precludes prediction of certain classes of ship
responses. The linear models can not predict the loss of ship stability in waves,
parametric resonance of roll and asymmetry of sagging and hogging. Ship
steering and maneouvring motion are disregarded.
Simulation of ship maneouvring is usually conducted for the still water
condition. Time-domain simulation of ship motion is restricted to in-plane
motion comprising of surge, yaw and sway motion components. If waves are
encounted for, their effect is taken into account as a steady state one.
The method which evaluates in time-domain ship rigid body motions in waves
and manuoeuvring is presented briefly. The so called two-stage approach
(Matusiak, 2000, Matusiak, 2001) is used when evaluating non-linear responses.
The method preserves best features of the linear seakeeping theory and takes
into account most important non-linearities. As a result the loss of ship stability
in waves and parametric resonance of roll are numerically predicted. Ship
maneouvring in regular waves can be simulated, too.

Six-degrees-of-freedom general model of ship

motion
Ship is regarded as a rigid body possessing in general six degrees of freedom.
Four co-ordinate systems are used for describing general ship motion. These are
presented in Figure 1.

Earth fixed axes

RG
RG

X 0 ,I,U

rG

0, i 0

G
Horizontal
body axes

0 , j0

Y 0 ,J,V

y,j,v

x,i,u

Z 0 ,K, W

0 ,k0

z,k,w

Fig. 1 Co-ordinate systems used in ship dynamics.

Inertial co-ordinate system fixed to Earth is denoted by X0Y0Z0. X0-axis points
in the wave propagation direction. The X0-Y0 plane coincides with the still water
level. Ship is on course with respect to waves. Course or encounter angle is a
time-averaged or initial orientation of ship with respect to the direction of wave
propagation. This time-averaged position defines the co-ordinate system
000. G is the origin of this co-ordinate system and it is the time-averaged
position of the ships center of gravity. Axis 0 points in the direction of ship
velocity vector VS. The average position of ship is given by the position vector
RG = XGI + YGJ.
The origin of two other Cartesian co-ordinate systems is located at the
instantaneous position of ships origin (point G in Fig.1). Co-ordinate system xyz
is fixed to the ship so that the x-axis points towards ship bow. This co-ordinate
system is called the body-fixed co-ordinate system. The so-called horizontal
body axes co-ordinate system (Hamamoto, 1993) denoted as moves with
ship so that the - plane stays horizontal that is it is parallel to the plane X0-Y0
and -axis stays at ship centreplane. Both the body fixed and horizontal axes coordinate systems move with ship with a velocity U.
Instantaneous position of ships center of gravity is given by the following
displacement components: surge (0 or x1), sway (0 or x2) and heave (0 or x3).
These are the motion components of the center of gravity in the moving with
ship velocity Vs inertial co-ordinate system 000. Translational motion is
defined as the motion of ships origin 0 in the inertial co-ordinate system

rG = 0 i0 + 0 j 0 + 0 k 0 .

(1)

The velocity of the origin of ship is given as

U = rG = 0 i0 + 0 j 0 + 0 k0 = ui + vj + wk .

(2)

Angular position of the ship is given by the so-called ship Euler angles denoted
in Fig. 1 as , and . The following matrix relation (Clayton&Bishop 1982;
Fossen, 1994) gives the projection of the velocity expressed in body-fixed coordinate system on the Earth-fixed co-ordinates

cos sin sin cos sin cos

cos cos
sin cos
+ sin sin u

0

sin sin sin sin sin cos

v
0 = sin cos

+ cos cos
cos sin

0

w
sin
cos sin
cos cos

(3)

Angular velocity of ship can be expressed in terms of the time derivatives of

roll, pitch and yaw as follows

= Pi + Qj+ Rk .

(4)

The dependence of the derivatives of the Euler angles and angular velocity
components expressed in the moving frame is as follows (Clayton&Bishop,
1982)

1 sin tan cos tan P

= 0
cos
sin Q .

0 sin /cos cos / cos R

(5)

Equations of motion are given by the set of six non-linear 2nd order ordinary
differential equations (Fossen, 1994)

X g mgsin = m(u + Qw Rv)

Yg + mgcos sin = m(v + Ru Pw )

Zg + mgcos cos = m(w + Pv Qu)

K g = Ix P I xyQ Ixz R + (Iz R I zxP I zyQ ) (Iy Q I yzR Iyx P )R

Q

M g = Iyx P + I yQ Iyz R + (Ix P I xyQ Ixz R)R (I zR Izx P Izy Q)P (6)

N g = Izx P I zyQ + Iz R + (Iy Q I yzR Iyx P)P (Ix P I xyQ Ixz R)Q.

In equations (6), Xg, Yg, Zg, Kg, Mg and Ng depict the components of global
reaction force and moment vectors acting on the ship. These are given in thebody fixed co-ordinate system xyz. In general these forces are non-linear. Mass
of ship and the components of the mass moment of inertia are depicted by m and
Iij.
Apart non-linearities of the left-hand-side of equations (6) also the body
dynamics model comprises of non-linear cross-coupling terms.

Approximate models of maneuvering and seakeeping

In the maneuvring simulation normally 1st, 2nd and 6th equation of the equations
set (6) are considered only. Non-linear cross-coupling terms are preserved. The
hydrodynamic forces acting on a hull are represented by the experimentally
derived quantities. The actions of propeller and rudder are taken into account by
relatively simple semiempirical models.
Linear seakeeping theory is based on the full linearization of equations (6). The
dynamics model is based on the small oscillation assumption. Non-linear crosscoupling terms are dropped out. The action of propeller and rudder is
disregarded. Hydrostatic and hydrodynamic forces are represented by the linear
approximations. Relatively sophisticated analytical models of wave and
radiation forces are used. The linear seakeeping approximation to the equations
of motion can be expressed as follows

m uL = X L = X rad + X diff + X F.K,L

m v L = YL = Yrad + Ydiff + YF.K,L

mw L = ZL = Zrestoring,L + Zrad + Zdiff + ZF.K,L

Ix PL Ixz RL = KL = K restoring,L + Krad + Kdiff + K F.K,L

IQ =M =M
+M +M +M
y

restoring,L

rad

diff

(7)

F.K,L

Iz RL I zx PL = N L = N rad + N diff + N F.K,L .

The indices rad, diff, F.K and restoring stand for radiation, diffraction, the socalled Froude-Krylov and restoring forces and moments. Index L depicts linear
approximation to the forces and moments. In the linear approximation wave
excitation is assumed to comprise of the diffraction and Froude-Krylov forces
and moments. The latter are evaluated from the pressures in and undisturbed
oncoming wave. In the integration ship hull is assumed to have a constant

velocity VS pointing in the x-direction and integration is conducted up to the still

water level.
Mainly due to the linearity properties the solution of equations (7) is sought in
the frequency domain. As a result the responses are obtained in a form of
transfer functions. Thus for instance linear x-response is given by

x L = x L 0 (, )cos[t k (X G cos YG sin ) + x ] ,

(8)

where xL0 is motion amplitude linear in respect to wave amplitude aW, wave
frequency, k = 2/g wave number and x phase angle.

Illustration of the two-stage approach using a

single-degree-of-freedom non-linear system
When forming a mathematical model of ship, the vessel is regarded as a rigid
body possessing in general six degrees of freedom. However, for the sake of
simplicity the two-stage approach is explained in this section using a singledegree-of-freedom model. The model is extended to a multi-degree-of-freedom
general model later on.
Let us consider a single-degree-of-freedom system given by a non-linear
equation

mX + g( X ) + h(X ) = F(X;t) ,

(9)

where m is system mass, t is time. Dots denote time derivatives. The functions g

and h are in general non-linear functions of response velocity X and

displacement X. Function F is also a non-linear function of X describing external
excitation to the system.
The linear version of the equation (9) is given by

m + cx L + kxL = FL (t) ,
xL

(10)

where FL is a linear, independent of the response forcing function. Total respone

is decomposed into a linear part xL and a non-linear portion x as follows
X = xL + x.

(11)

Subtracing linear approximation (10) from the general equation (9) yields
equation for the non-linear part x of the response

m+ [g(x L + x ) c xL ] + [h(x L + x) kx L ] = f ,
x

(12)

where f = F(X;t) FL(t) is a non-linear part of the forcing function.

Why to use two-stage approach?

The question arises why to use a two-stage approach? Why not to solve a nonlinear equation of motion (6) directly in time domain by an appropriate
numerical integration routine? Linear methods of the seakeeping theory are very
well established. In particular hydrodynamic forces (radiation and diffraction)
are well represented by the linear approximation. As a result ship motions and
loads are very well established and they are given in a form of reliable transfer
functions provided ship is on a straight course and motion fulfills the linearty
assumptions. Direct evaluation of ship motions with an aid of a non-linear strip
theory model involves certain compromises. In particular diffraction forces are
often disregarded and the radiation forces are usually represented by a simplified
constant added mass model. These drawbacks of the non-linear strip theory
method are avoided when using a two-stage approach.
In the two-stage approach the main part of the first order, fast response in waves
is given by a linear approximation. These are evaluated for an actual heading
and actual position in wave. Non-linear parts of hydrostatic and hydrodynamic
load, rudder and propeller forces and non-linearities of ship rigid body dynamics
yield non-linear part of the first order motions and slow maneouvring motion.

Perturbed non-linear portion of ship motion

Having linear approximation to ship motions in waves obtained with the aid of
linear seakeeping theory, non-linear part of ship motions is evaluated in the time
domain. This motion takes into account non-linearities of ship hydrostatics and
non-linearities of wave loads at large amplitudes of motion. The only motion
component that is not decomposed into the linear and non-linear part, is surge.
Total surge motion is evaluated using the 1st of equations (6). The effect of
added wave resistance, propulsor action and rudder forces are included in this
equation. Total ship motion, or other type of response, being a sum of linear
approximation and a non-linear part is thus obtained. In other words total
responses in terms of velocities are written in the following form

U = ui + (v L + v ) j + (wL + w )k

= (PL + P )i + (QL + Q)j+ (RL + R)k,

(13)

where variables without subscripts depict non-linear part of the response. Linear
approximation is evaluated with an aid of formula (8) for an actual ship position
(XG,,YG) in wave and for an actual heading .
Subtracting the equations (7) of the linear approximation model from equations
(6) yields the equations for the non-linear part of response

m[u + (QL + Q)(wL + w) (RL + R )(vL + v ) + gsin(L + )] = X

m[v + (RL + R )u (PL + P )(w L + w)gcos( L + )sin( L + )] = Y

m[w + (PL + P )(v L + v) (QL + Q)ugcos( L + )cos(L + )] = Z

(14)

Ix P + [I z (RL + R ) Izx (PL + P ) Izy (QL + Q)](QL + Q) I xyQ

Ixz R [I y (QL + Q) I yz (RL + R ) Iyx (PL + P )](RL + R ) = K

Iy Q I yx P + [I x (PL + P ) I xy (QL + Q) I xz (RL + R )](RL + R)

Iyz R [I z (RL + R ) Izx (PL + P ) Izy (QL + Q)](PL + P ) = M

Iz R I zxP + [Iy (QL + Q) I yz (RL + R ) Iyx (PL + P )](PL + P )

Izy Q [I x (PL + P ) I xy (QL + Q) I xz (RL + R )](QL + Q) = N .

Equations (14) govern non-linear part of the rigid body motion in six degrees of
freedom. In order to solve them we need to specify the non-linear part of the
external (fluid) forces X,Y,Z and moments K,M,N acting on a body. These are
presented in bigger detail in reference (Matusiak, 2002). Moreover we use
equations (3) and (5) to express body velocities in the inertial co-ordinate
system. Numerical integration of these equations together with the division of
responses given by equations (13) yields the instantaneous position of ship in the
inertial co-ordinate system X0Y0Z0. Additional, thirteenth ordinary differential
equation of a first order representing the action of auto-pilot is used to control
the rudder angle. Integration is conducted using the 4th order Runge-Kutta
scheme with an integration step being t = 100 ms. Computation is conducted
for a full-scale ship. Linear approximation of responses and forces is related to
ships actual position in waves. It takes into account instantaneous heading
angle. The zero initial conditions are used for all equations with an exception of
surge velocity, which is set initially to a prescribed ship velocity in calm water.
In order to dampen the spurious transients, wave amplitude is gradually
increased from zero to the prescribed final value aw,final using the expression
2

t
for t < Tf
aw (t) = aw, final 1 cos
2Tf

aw (t) = aw, final for t Tf ,

where t is time and with Tf= 50 seconds in full scale being used.

(15)

Combined model of maneuvering hull and radiation

forces in time domain
A quasilinear model making use of the added mass and damping concept
approximates radiation forces. These forces can be expressed as

X rad = a11u b11(u VS ) a15Q b15Q

Yrad = a22v b22v a24 P b24 P a26R b26R

Z = a w b w a Q b Q
rad

33

33

35

35

K rad = a44 P b44 P a46 R b46 R a42v b42v

M = a Q b Q a w b w a u b (u V
rad

55

55

53

53

51

51

(16)

N rad = a66 R b66 R a64 P b64 P a62v b62v.

In equations (16) aij and bij depict added masses and damping coefficients
referred to the origin located in the center of gravity (G in Fig. 1). These are
frequency dependent values. In the present method these coefficients are
evaluated by a standard linear seakeeping theory based computer program
(Journee, 1992). Note that radiation forces are oriented in the body-fixed coordinate system.

Memory effect included using the retardation function concept

Time domain approach requires the so-called convolution integral representation
of the radiation forces (Cummins, 1962). In this time approach radiation forces
vector Xrad is represented by an expression:

Xrad (t) = a
x(t)

k(t )x( )d ,

(17)

where a is the matrix comprising of the added masses coefficients for an

infinite frequency and x is the response vector. Matrix function k is the so-called
retardation function which takes into account the memory effect of the radiation
forces. This function can be evaluated as follows

k(t) =

b( )cos(t)d ,

(18)

where b is the frequency dependent added damping matrix. In order to take into
account the maneouvring hull forces four components of the b matrix are

modified as follows. Terms of sway, yaw and their coupling terms of the linear
hull forces model are subtracted from the corresponding elements of matrix b, ie
b22() = b22() YV
b66() = b66()

b62() = b62()

(19)

b26() = b26() YR..

The k(t) functions have to be evaluated only once before the simulation. The
Fast Fourier Transform algorithm is used when evaluating discrete values of the
retardation functions as follows (Matusiak, 2001)

K k,ij (kt) =

N
FFT(gij (x)) ,

(20)

where the original added damping discrete functions are substituted by a

double-sided function g(x) as follows:

gij (x) = bij (x) for x = ,.....N /2

(21)

gij (N x) = bij (x) for x = 0,..... (N /2 + 1).

Note that as a result the retardation function 16 is obtained at N/2 discrete time
instants with a time step t. FFT analysis is conducted with N = 2048. As a
result the retardation functions are represented by 1024 discrete values covering
the period of 102.4 seconds. An example of the retardation function for heave is
given below.

2.5
K33*
2
1.5
1
0.5
0
0
-0.5

t*

Fig. 2

Non-dimensional heave memory (retardation) function

K * = K 33 / m g/ L ) as a function of non-dimensional time

33
t = t / g /L , where L is waterline length of ship.
*

Examples of ship motion simulation and method

validation
Two ship cases where investigated within the benchmark study of ITTC. First
one, ship A1 (see Fig. 4), is a containership of waterline length of 150 m and low
megacentric height (GM0 = 0.15 m). Second vessel (see Fig. 5) is a model of
seiner of the length Lpp = 35.68 [m]. Both models were run in regular waves,
different headings and Froude numbers.

Fig. 4 Three dimensional panel model of ship A1.

Fig. 5 Three dimensional panel model of ship A2.

The summary of model test results and simulation results are given in Tables 1
and 2.

Table 1 Summary of the results for containership (Ship A1).

Case

/Lpp

2aW/

Fn

Heading
[deg]

Experiment

Computed

Parametric roll
Parametric roll resonance, nonresonance, capsize
capsizing

1.5

1/25

0.2

1.5

1/25

0.2

45

no- capsizing

no- capsizing

1.5

1/25

0.3

30

no- capsizing

no- capsizing

1.5

1/25

0.4

30

capsize

capsize

Experiment

Computed

Table 2 Summary of the results for Seiner (Ship A2).

Case

/Lpp

2aW/

Fn

Heading
[deg]

1.637

0.1

0.3

-30

no-capsizing

no-capsizing

1.637

0.1

0.43

-10

surfing, capsize

surfing

1.127

0.115

0.3

-30

no- capsizing

no- capsizing

1.127

0.115

0.43

-30

capsize

capsize

Selected time-histories of the angular motions of both vessels are presented in

Figs. 6-13.

n.3 Heading=30 deg, co ai ers

=, 0
n n
t

h case p 3 ;
i

15
10
5
0
-5

100

200

300

400

500

-10
-15
-20
-25

Roll
[deg] [deg]
Pitch
Yaw [deg]
Rudder

-30
-35

Tim [s]
e

Fig. 6. Simulated angular motions of ship in regular quarterirng waves; Case 3

(heading = 30 [deg]). Ship speed is Fn = 0.3.

roll [deg]
pitch [deg]
yaw [deg]

25
20
15
10
5
0
-5 0
-10

50

100

150

200

250

time [s]

-15
-20
-25
-30
-35

Fig. 7 Angular motions of ship in regular quartering waves of containership;

Case 3, (heading = 30 [deg]). Ship speed is Fn = 0.3. Model test result
scaled to full-scale and yaw defined as a deviation from the initial
course.

Fn=0.4 Headi g=30 deg, conta nershi ; cap 4

,
n
i
se

50
25
0
0

50

100

150

-25

Pi ch [deg
t

250

300

350

Ro l [deg]
l

-50

200

Yaw [deg
-75

]
]

Rudder deg]

-100

Time [ ]

[
s

Fig. 8 Containership running at Fn = 0.4 capsizes in regular quartering

regular waves (heading 30 [deg]; case 4). Result of simulation.

Fn=0.4,Heading=30 deg, experiment

50
25
0
0
-25
-50
-75

10

20

30

40

50

60

70

80

90

Roll [deg]
Pitch [deg]
Yaw [deg]
Rudder [deg]

-100
Time [s]
Fig. 9 Model of containership running at Fn = 0.4 capsizes in regular
quartering regular waves (heading 30 [deg]). Model test result scaled to
full-scale. [5].

Ship A2, Case A; computed

20
15
10
5
0
-5

10

20

30

-10

40

50

time [s]

-15

Roll [deg]
Pitch [deg]
Yaw [deg]
Rudder [deg]

-20
-25
-30

Fig. 10 Angular motions of Seiner in regular quartering waves of; Case A,

(heading = 30 [deg]). Ship speed is Fn = 0.3. Simulated result.

Ship A2, case A experiment

40
Roll [deg]
Pitch [deg]
Yaw [deg]
Rudder [deg]

30
20
10
0
0
-10

10

20

30

40

time [s]

-20
-30
-40
-50

Fig. 11 Angular motions of Seiner in regular quartering waves; Case A,

(heading = 30 [deg]). Ship speed is Fn = 0.3. Measured result.

50

Ship A2; case D, computed

90
80
70

Roll [deg]
Pitch [deg]
Yaw [deg]
Rudder [deg]

60
50
40
30
20
10
0
-10

10

15

20

25

30

-20
time [s]

Fig. 12 Ship A2 (Seiner) capsizes in regular quartering waves; Case D,

(heading = 30 [deg]). Ship speed is Fn = 0.43. Simulated result.

Ship A2; Case D, experiment

90
80
70
60
50
40
30
20
10
0
-10 0
-20
-30
-40
-50

Roll [deg]
Pitch [deg]
Yaw [deg]
Rudder [deg]

10

15

20

25

30

35

time [s]

Fig. 13 Ship A2 (Seiner) capsizes in regular quartering waves; Case D,

(heading = -30 [deg]). Ship speed is Fn = 0.43. Measured in model
scale.

Conclusions
The combined model of maneouvring and non-linear seakeeping yields ship
motions which at least qualitatively agree with the model test experiments.
There are several possible reasons for a certain disagreement of simulation and
model test results.
Initial conditions are set to zero in the simulations. Measurements of ship
motions in model tests are started at a certain instant with the initial conditions
which are not very well known. Usually both waves and ship motions are
already well developed at the begining of an experiment. As the initial
conditions have a big influence on the response on a non-linear system, this may
affect the comparison of results.
At high Froude numbers (Fn > 0.4) an effect of dynamic lift may be important
on ship static stability. For the time being this effect is not taken into account in
the presented method.
Maneouvring hull forces are regarded as linear ones and as independent of the
ship first order motions in waves. This assumption may be not good when
considering a lose of dynamic stability in waves where a change of ships course
may be very rapid.
The above mentioned simplifications of the presented method have to further
studied. Neverless the method, already at the present stage of development, may
be used as a tool to evaluate ship dynamic stability and to simulate ship
maneouvring in waves.
The method can be easilly further developed to consider other matters relevant
to ship design. A possibility to evaluate non-linear sectional loads in terms of
total shear forces, bending and torsional moments is being developed. The aim
of this study is to reveal unsymmetry of sagging and hogging of total sectional
loads in waves. Modelling the action of the turnable Azipod-type z-drive unit on
maneouvring is considered, too.

References
Clayton B.R.& Bishop R.E.D 1982 Mechanics of marine
vehicles, ISBN 0 419 12110-2.
Cummins, W.E. 1962 The Impulse Response Function and
Ship Motions,Schiffstechnik 9 (1962 Nr. 47 S101/109.
Fossen, T.,I. 1994 Guidance and control of ocean vehicles, J.
Wiley&SonsISBN 0 471 94113 1.
Hamamoto, M. and Kim, Y.S., 1993 A New Coordinate
System and the Equations Describing Manoeuvring Motion of a
Ship in Waves, J. Soc. Naval Arch., Vol 173.
Journee J. M. 1992 Strip Theory Algorithms, report MEMT 24,
Delft University of Technology, Ship Hydrodynamics
Laboratory.
Matusiak, J. 2000 Two-stage approach to determination of nonlinear motions of ship in waves. 4th Osaka Collouqium on
Seakeeping Performance of Ships, Osaka, Japan, 17-21st
October, 2000
Matusiak, J. 2001 Importance Of Memory Effect For Capsizing
Prediction, to be presented at the Stability Workshop, Trieste
September 2001
Matusiak, J. 2002 Two-stage approach to determination of nonlinear motions of ship in irregular waves, Helsinki University of
Technology, Ship Laboratoory, to be published.