You are on page 1of 12

J Mar Sci Technol (2002) 7:3142

Finite-volume simulation method to predict the performance of


a sailing boat
Hiromichi Akimoto1 and Hideaki Miyata2
1
2

Department of Applied Mathematics and Physics, Tottori University, Minami 4-101, Koyama, Tottori 680-8552, Japan
Department of Environmental and Ocean Engineering, The University of Tokyo, Tokyo, Japan

Abstract A flow-simulation method was developed to predict


the performance of a sailing boat in unsteady motion on a free
surface. The method is based on the time-marching, finitevolume method and the moving grid technique, including
consideration of the free surface and the deformation of the
under-water shape of the boat due to its arbitrary motion. The
equation of motion with six degrees of freedom is solved by
the use of the fluid-dynamic forces and moments obtained
from the flow simulation. The sailing conditions of the boat
are virtually realized by combining the simulations of waterflow and the motion of the boat. The availability is demonstrated by calculations of the steady advancing, rolling, and
maneuvering motions of International Americas Cup Class
(IACC) sailing boats.
Key words Free surface Moving boundary Unsteady
motion Sailing boat

Introduction
Recent advances in computational fluid dynamics
(CFD) enables us to predict the performance of a ship
in steady advancing motion.1 There have also been
some attempts to evaluate the maneuvering abilities of
a ship by CFD techniques.2,3 However, most of these
motions are restricted within steady two-dimensional
motion, e.g., steady circling or obliquely advancing
motions.
In the case of high-speed ships, the attitude of a ship
depends on its forward speed because of the large
dynamic pressure acting on its hull. Experimental studies are difficult owing to the large amplitude motion of
the ship. In the course of maneuvering, a ship makes

Address correspondence to: H. Akimoto


(e-mail: akimoto@damp.tottori-u.ac.jp)
Received: December 25, 2001 / Accepted: March 26, 2002

roll, yaw, pitch, sway, and heave motions, which are


often neglected in the case of a low-speed ship.
Predicting the performance of a sailing boat is an
interesting topic in ship hydrodynamics owing to their
complicated performance, which is different from that
of commercial ships.4 In upwind sailing conditions, the
boat cruises with a large heel angle to obtain thrust from
the wind and a small leeway angle to generate lateral
force to cancel the unwanted component of the sail
force. The attitude of the boat depends on the hydroand aerodynamic forces and moments acting on its hull,
sails, keel, and rudder. These forces must be evaluated
in order to solve the equation of motion of the boat
(Fig. 1). It is difficult to investigate the performance of a
sailing boat in an experimental facility because of the
complexities of the sailing dynamics and wind conditions. In a towing tank, we can only postulate the wind
action and the resultant balance of the boat by adjusting
the towing position and ballast weights on a model ship.
A method of overcoming these difficulties is to
extend the CFD techniques so that the freely moving
properties of the boat can be incorporated. Some
studies have attempted to realize self-propelling and
steady maneuvering motions by the CFD technique.
Most of these have been numerical realizations of a
towing test, where a ship is fixed in a computational grid
system. We can remove this restriction by solving the
equation of motion of the ship simultaneously with the
CFD simulation, including the changes in geometry of
the free surface and the body boundary.
In this work, a flow simulation code WISDAM-VII5 is
developed for the flow around a freely moving ship, in
particular for the large motions of a sailing boat (Fig. 2).
It employs the finite-volume method in the framework
of an O-O-type structured grid system that is fitted to
both the free surface and the hull surface. The grid
system is generated at each time step. Deformations of
the computational domain are treated by a waterline
search procedure and a moving grid method that is

32

H. Akimoto and H. Miyata: Predicting sailing boat performance

Governing equations
The governing equations are the conservation laws of
momentum and mass in control volumes which deform
time-dependently. They are expressed as

d
udV = T dS
S (t )
dt V (t )

(1)

d
V (t ) = (u - v) dS
S (t )
dt

(2)

where V(t) and S(t) denote the volume and surface area
of the control volume, respectively, u is the fluid velocity vector, and v is the moving velocity of the surface of
the control volume. dS is the product of the infinitesimal
area element dS and the outward normal vector n on the
surface of the control volume. Using the eddy viscosity
model for Reynolds stress, the stress tensor T is
expressed as
Fig. 1. Coordinate system

Fig. 2. Block diagram of the performance prediction system

similar to that of Rosenfeld and Kwak.6 The fluiddynamic forces obtained in the CFD simulation are
introduced into the equation of motion of the boat.
Although the forces and moments acting on the sails,
keel, and rudder are given by empirical equations in
this method, this system provides the basis of a virtual
reality system for all sailing boats.

Fluid-dynamics simulation
The CFD code WISDAM-VII was developed to simulate the incompressible viscous water flow around a ship
in arbitrary motion.5 Assuming that the interactions of
the flow between the hull and other parts of the boat are
small, the CFD simulation is performed for the hull only
in this method. Thus, the target of the system is the
hydrodynamic evaluation of hull configurations under
the influences of all the lifting surfaces.

T
T = -(u - v)u - PI +
+ n t u + (u)

Re

(3)

where P is the kinematic pressure defined by P = p/r z/Fn2, p is the pressure, r is the density of water,

Fn (=U/gL) is the Froude number, and U and L are the


standard speed and length of the boat, respectively.
I is the unit tensor, Re (=UL/n) is the Reynolds number,
and nt is the kinematic eddy viscosity. Equations 1
and 2 are solved numerically by a MAC-type timemarching algorithm. The spatial discretization of the
usual convection flux uu is by the 3rd-order upwind
scheme. The additional flux due to grid motion vu is
treated by the method of Rosenfeld and Kwak.6 To
satisfy the minimum geometric conservation, this uses
the momentum in the volume of a hexahedron which is
swept by the face under consideration due to the motion
of the grids. The pressure and the diffusive fluxes are
evaluated by 2nd- and 1st-order spatial descretization, respectively. Time discretization is 1st-order
Euler-explicit.

Boundary condition for velocity


We assume that the boat is rigid. Then the velocity of
the surface vb on the hull is determined from the motion
of the boat and the relative position of the point of
interest xb from the gravitational center.
v b = vG + w G ( x b - x G )

(4)

where xG is the center of gravity of the boat, vG is the


forward velocity of xG, and wG is the angular velocity
vector of the boat with respect to xG. To impose the noslip boundary condition (u = vb), we use dummy velocity
points located in the body (Fig. 3). Their velocity

H. Akimoto and H. Miyata: Predicting sailing boat performance

33

(P)

body

=-

v b 1 2
+
u
t Re
body

(7)

In Eq. 7, the first term on the right-hand side is the


acceleration of the body surface determined by Eq. 4.
By taking the inner product of Eq. 7 and the outward
normal vector on the body surface nb, the normal pressure gradient on the moving body surface is obtained in
the form

P
v
= - nb b
nb
t

Fig. 3. Body boundary condition for velocity

(8)

Here, the contribution of the diffusion term is neglected


because it is parallel to the body surface in the boundary
layer. Pressure at the dummy point is extrapolated using
the normal pressure gradient P/nb in Eq. 8. When the
acceleration of the body is zero, this is equivalent to the
usual zeronormal-gradient condition for pressure.
Free surface condition
We assume here that there is no turning over or breaking motion of free surface waves. Then the wave height
is expressed by a single-valued function z = h(x, y, t).
Although this assumption is not rigorously valid in the
bow region, this is restricted to a small area and has a
relatively small influence on the dynamics of International Americas Cup Class (IACC) boats. The kinematic and dynamic boundary conditions are given on
the free surface. The kinematic condition is expressed
as

Fig. 4. Deformation of grids in a cross section

vectors udummy are defined by the no-slip condition and


liner extrapolation from the fluid velocity of the nearest
point u1.
udummy = 2v b - u1

(5)

Without disturbing the no-slip condition, surface grid


points slip on the hull surface according to the regeneration of the grids at every time step, so that excessive
distortion of the grid system does not occur even when
the boat makes a large rolling motion (Fig. 4).

h
h
h
= v3 = u3 - (u1 - v1 )
- (u2 - v2 )
t
x1
x2

where u = [u1, u2, u3] is the fluid velocity vector on the


free surface, and x = [x1, x2, x3] and v = [v1, v2, v3] are the
position and velocity vectors, respectively, of grid points
on the free surface. With the assumption that the surface tension and viscous stress on the free surface are
negligibly small, the dynamic condition is written as
P=

Boundary condition of pressure


To evaluate the pressure on the moving body surface,
an ALE (arbitrary Lagrangian and Eulerian) form of
the NavierStokes equation is employed on the body
surface.

1 2
u
u
+ (u - v b ) u = -P +
Re
t

(9)

pa
x
x
u
- 32 = - 32 ,
=0
ns
r Fn
Fn
(on the free surface)

(10)

where pa is the atmospheric pressure on the free surface,


and u/ns is the velocity gradient in the direction normal to the free surface. The quantity pa is assumed to be
a constant. We can set pa = 0 without loss of generality.

(6)

Imposing the no-slip condition u = vb with Eq. 6, we


obtain an equation for the pressure gradient on the
moving body surface.

Model of turbulence
A hybrid turbulence model is employed to evaluate the
kinematic eddy viscosity nt. This is a combination of the

34

H. Akimoto and H. Miyata: Predicting sailing boat performance

BaldwinLomax algebraic turbulence model7 (BL) and


the Smagorinsky eddy viscosity8 in the subgrid-scale
turbulence model (SGS). The BL model is applied on
the fore part of the hull where the boundary layer is thin
and locally two-dimensional. However, in the rear part
of the body and in its wake region, the boundary layer
rapidly increases in thickness10 and the BL model tends
to overestimate the eddy viscosity. Therefore, we
estimate nt by the following equation:
0

n tBL
nt =
bn tBL + 1 - b n tSGS
SGS
n t

(x < x )
(x < x < x )
(x < x < x )
( x < x)
FP

FP

MID

MID

(11)

AP

AP

where n BL
and n SGS
are kinematic eddy viscosities
t
t
obtained from the BL and SGS model, respectively.1
xFP, xMID, and xAP denote the x-coordinates of the fore
end point, mid-ship position, and after end point of the
wetted surface of the hull, respectively. The parameter

b is selected as b = s(x)/s(xMID), where s(x) is the section area that is the longitudinal distribution of the hull
volume beneath the still water plane in the upright
position.1 b is 1 at the midship, decreases in the aft
direction, and becomes 0 at xAP.
In the original BL model, small calculated n BL
flows
t
are treated as laminar. However, we assumed that the
turbulent boundary layer starts from xFP even at a relatively low Reynolds number condition. This is because
our target is the predicition of forces on the full-scale
boat.
n SGS
is determined by
t

[ {

}]

n tSGS = Cs D 1 - exp(- y + 25)

Sij

where y is the dimensionless wall distance, D is the


minimum grid spacing of the local control volume,
and |Sij| is the magnitude of the strain-rate tensor. The
Samgorinsky constant Cs is 0.5 in this calculation. The
estimate of the eddy viscocity changes gradually from
the BL model to the SGS model in the aft half of the
hull. In the wake region after xAP, the eddy viscocity is
determined by the SGS model only.
+

Motion of the boat


Equation of motion of the boat
The equation for the transverse motion of the boat is
written as

dv G
= Fhull + Fsail + Frudder + Fkeel + Fdamp
dt

(12)

where m is the mass of the boat, and the right-hand-side


terms are forces acting on the main hull, sails, rudder,
and keel, respectively. These forces include hydrodynamic or aerodynamic forces and gravity. The last term,
Fdamp, is an artificial damping force. This is added to
reduce unnecessary transient oscillations of the boat
when the target is a steady-state solution. The conservation law for the angular momentum has a similar form.

dhG
= Mhull + Msail + M rudder + M keel + Mdamp
dt

(13)

where hG is the angular momentum vector of the boat


with respect to its center of gravity. The right-hand-side
terms are the moment vectors of all parts and the artificial damping moment.
A ship has natural damping mechanisms due to the
viscous effect and the dynamic forces of its appendages.
Since their magnitude of damping is usually small, occasional transient oscillating motions of the boat waste
considerable CPU time. Therefore, Fdamp and Mdamp are
added to reduce the CPU time for unnecessary transient
motions. These are set at zero if the unsteady motion of
the boat must be simulated rigorously.
Fhull is expressed as

z
mhull
Fhull = - P e
dS - S D dS 2
S
Fn
Fn 2 3

(14)

where S dS means the surface integration on the wetted


surface of the hull, mhull is the mass of the hull, D is the
viscous stress tensor, and e3 is a unit vector orienting
vertically upward. The right-hand-side terms are the
pressure force, viscous force, and gravitational force,
respectively. The hydrodynamic and gravitational moments on the hull are express as

z
Mhull = - P x - x G n dS
S
Fn 2

{(

) (

) }

+ x - x G D n dS S

mhull
x - x G e3
Fn 2 hull

(15)
where (x - xG) is the position vector originating from xG,
and xhull is the center of gravity of the main hull. It
should be noted that the buoyancy of the hull is included in the surface pressure integration that includes
the gravity potential. The effect of added mass around
the hull is also included in this integration.
Correction of the viscous force
It is not practical to perform a simulation of a full-scale
boat with our limited CPU power because the Reynolds
number of such a boat is about 108. Therefore, we have
to estimate the motion of a full-size boat from the com-

H. Akimoto and H. Miyata: Predicting sailing boat performance

putational results with a lower Reynolds number. For


this purpose, only the frictional component of force
calculated by CFD is extrapolated to that of the full-size
boat as

fship =

C ship
f
C

CFD
f

fCFD

(16)

where fship is the estimated frictional force of the fullscale ship, and fCFD is the computed frictional force, i.e.,
integrated tangential stress on the wetted surface of the
hull. Cfship and CfCFD are the frictional force coefficients
of a flat plate at the Reynolds number of the full-scale
ship and the CFD simulation, respectively. Although
the range of Reynolds numbers in the CFD simulations
is from 105 to 106, the calculated pressure force can be
used in the equation of motion of the full-scale boat,
with a small error only, because the scale effect is believed not to be large for this component.

The fluid-dynamic forces acting on sails, rudder, and


keel are evaluated by a wing theory and empirical
equations.
For example, we describe here the model of the rudder. The advancing velocity vector of the rudder vrudder is
expressed as
v rudder = vG + w G ( x rudder - x G )

(17)

where xrudder is the position vector of the rudder. Then


the fluid velocity relative to the rudder is
(18)

va = u - v rudder

A local coordinate system is used to decompose the


hydrodynamic force into the drag and lift components
(Fig. 5). The new coordinate system is chosen as
Yr = nr X r ,

a = p 2 - cos -1 (X r nr )

Zr = X r Yr

(19)

where nr is the unit normal vector of the rudder. The


angle of attack a is

(20)

where a is positive when the lift force is in the Zr direction. The hydrodynamic force on the rudder is
2
1
Frudder = (C L Zr + CD X r ) r va Sr
2

(21)

where CL and CD are the lift and drag coefficients, respectively, and Sr is the area of the rudder. There are
interactions among the appendages and the hull
throughout the flow field. Because of the complexities
of these interactions, we incorporate here only the induced velocity of the forward-mounted keel. The fluid
dynamic coefficients in Eq. 21 are determined by the
following empirical equations5:
CL =

dC L
C Lkeel
a - C1 keel C 3D
l
da

CD = (1 + C 2C L )CD0 +

Models of appendages

X r = va va ,

35

C L2
plrudder

here lrudder and lkeel are aspect ratios of the rudder and
keel, respectively, and Ckeel
is the lift coefficient of the
L
keel. dCL/da, CD0, and C3D are coefficients of the liftcurve slope, the base drag, and the three-dimensional
correction factor, respectively. C1 and C2 are factors
of the interaction between the keel and the rudder.
These coefficients are obtained from experiments with a
model rudder. The forces acting on the sails and the
keel are evaluated in a similar manner.
Although the effect of added mass around the hull is
included in the CFD part, the added mass of some of the
appendages have not yet been fully considered.
As shown in Eq. 21, the treatments of wing-like appendages are steady-state approximations. Because the
chord length c of these appendages is relatively small in
relation to the boat length L, the reduced frequency of
the boat motion with a wing, (U/L)/(U/c), is about 0.02
0.04. This means that a quasi-steady approximation
which includes the added mass of the wings is not required in calm water conditions. The added mass
around the sail and the bulb is not explicitly included in
the present simulation, because a rough estimate of
their effect was smaller than the uncertainty of the total
inertial moment of the boat.
A more precise estimation of the inertia and the
quasi-steady approximation of the wings will be required when the predicted performance of a boat in
waves is considered.
Grid generation

Fig. 5. Hydrodynamic forces acting on the rudder

Because of the arbitrary motion of the boat and the


deformation of the free surface, the wetted part of the

36

H. Akimoto and H. Miyata: Predicting sailing boat performance

Fig. 7. Local cylindrical coordinates representing the hulls


surface

Fig. 6. a OO-type grid system, and b a close-up around the


hull

hull changes its shape in a time-dependent manner. To


cope with the hull configuration of a sailing boat in large
changes of attitude, an OO-type boundary-fitted structured grid system was selected, as shown in Fig. 6. This
spherical grid system is regenerated at each time step
after the moving boundaries of the free surface and the
hull have settled, so that the distortion of the grid remains small even for a large-amplitude motion.
Surface modeling of the hull
For ease in handling the hull configuration, a cylindrical
coordinate system was defined, as shown in Fig. 7. The
position vector on the hull surface X = [X, Y, Z] in
the hull-fixed coordinate system is expressed by two
parameters of Xs and qs:

X = Xs ,

Y = r ( X s ,q s ) cos q s ,

Z = -r ( X s ,q s ) sin q s
(22)

where Xs is the longitudinal position on the fictitious


hull axis, qs is the angle from the center plane to the
point, and r(Xs, qs) is the distance from the axis to the
surface point. The function r(Xs, qs) is designed to calcu-

late r from Xs and qs using second-order interpolation


from the given data points of the hulls shape. The vector X is converted to the point of the ground-fixed coordinate system x according to the position and attitude of
the boat. For concise notation, we express x as x = r(Xs,
qs) E(X(Xs, qs)), where E(X) means the conversion
from the body-fixed coordinate to the ground-fixed coordinate system using transverse and rotational
transformations.
The configuration of the hull is given in the row
of position vectors generated by a CAD application.
These are loaded and then converted to the cylindrical
coordinate system in CFD code. This implementation
conceals the discreteness of the geometrical data in the
function r(Xs, qs). It then become possible to handle the
complicated geometric changes in the boundaries in
the simulation.
Search procedure for the waterline
In the first step of the grid generation, the position of
the waterline, i.e., the intersection of the free surface
and the hull surface, is determined. Because these two
surfaces deform owing to the wave motion and the
change in the boats attitude, there is no easy way to
determine the waterline.
The determination of the waterline is based on a
bisection search on the hull surface. If there are two
points on the hull surface where one is above the free
surface and the other is under the free surface, the

H. Akimoto and H. Miyata: Predicting sailing boat performance

37

Fig. 8. Bisection search for the waterline

position of the waterline between them is determined as


described below (Fig. 8).
1. Set initial points xdry = r(Xdry, qdry) and xwet = r(Xwet,
qwet), where Xdry, qdry and Xwet, qwet are parametric
expressions of given dry and wet points.
2. Calculate the midpoint xm between xdry and xwet on the
hull surface, where xm = [xm, ym, zm] = r(Xm, qm), as

X m = X dry + X wet

2,

q m = q dry + q wet 2

3. Replace xwet or xdry with xm according to the local free


surface.
If zm < h( xm , ym ), then replace x wet with x m
If zm > h( xm , ym ), then replace x dry with x m

where z = h(x, y) is the local free surface plane.


Procedures 2 and 3 are repeated until wet and dry
points converge into a single point. The result gives the
local waterline position.
Search procedure for the fore and the aft end points
In the case of a sailing boat, the fore end point xFP and
the aft end point xAP of the wetted surface move a lot
owing to the special hull form configuration. They also
move time-dependently owing to the motion of the boat
and the free surface. We must determine these two
points at the first step of the grid generation because
they are pole points of the OO grids.
xFP is determined by the repeated use of the waterline
search (Fig. 9).
1. Set the initial section X = Xs, i.e., on the downstream
side of xFP.
2. Search the waterline positions of both the starboard
side xL and the port side xR of the section.
3. Obtain the midpoint xM between xL and xR on the
hulls surface.

Fig. 9. Search procedure for the fore end point (FP) by the
bisection method

4. Search the waterline position xFP on the forward


stretching line from xM.
5. Replace Xs with XFP, and then go to point 2.
Procedures 25 are repeated until the three points xL,
xR, and xFP converge into one point. That point becomes
xFP at that moment. The search procedure for xAP is
performed in the same manner. xFP and xAP are connected by two waterlines on the port and starboard
sides by the repeated use of the bisection search for the
free surface around the hull.
Distribution of the surface and volume grids
After the determination of the two waterlines, body
boundary grids are distributed on the wetted surface
between the waterlines. Then the free surface grids are
generated from the waterlines to the outer boundary.
The height of each grid point from the still water plane
is obtained from Eq. 9. The inner-volume grids are algebraically distributed between the hull surface and the
outer boundary. Whole grids are regenerated at every
time step.

Numerical results
Simulation procedure
The simulation procedure using this method is very
similar to that of a model test in a towing tank. It includes setting a boat afloat, adjusting the still waterline
with ballast weights, setting the attitude, accelerating

38

H. Akimoto and H. Miyata: Predicting sailing boat performance

Fig. 10. Distribution of the wave height divided by L along


the hull, Fraude number (Fn) = 0.34, NI is the number of
control volumes in the longitudinal direction along the hull

the boat (by towing), and taking the measurements. All


of these are performed numerically.
In a typical acceleration procedure, all degrees of
freedom except heave are fixed to prevent unnecessary
oscillations of the boat. When the boat reaches a steady
forward speed, one can select suitable conditions of
binding or release and the magnitudes of damping for
the test, in all six degrees of freedom. For example, in
the leewayheel test, the angles of leeway and heel are
gradually changed to the target position to be measured. In this case, all degrees of freedom except heave
and pitch are fixed. Artificial damping terms are added
to reduce unnecessary oscillation. This is effective if
time-averaged properties are the main concern.
Steady forward motion

Fig. 11. Comparison of the measured and computed surface


pressure (Cp) distribution: heel = 21.3, leeway = 2.0, contour
interval = 0.02

The accuracy of the WISDAM-VII code was tested in a


steady forward motion case. Figure 10 shows the distribution of the wave profile along a typical IACC boat,
JN35, without appendages and in an upright position.
The measurements were performed in a towing tank
at the University of Tokyo with a one-seventh scale
model, and using a still camera. The target speed of the
full-size boat was 9.0 knots (Froude number 0.35,
Reynolds number for this experiment 4 106). The
number of control volumes was 45 000. The results show
that the accuracy of the simulation was satisfactory.
Although this simulation did not include local overturning and breaking waves around the bow, the peak
position of the profile showed good agreement.
Figure 11 shows a comparison of the pressure distributions on the wetted surface area. In this case an IACC
boat, JN32, was in steady sailing motion with a heel
angle of 21.3 and a leeway angle of 2.0. The pressure

was measured at 180 pressure holes located on the hull


beneath the still-water plane. The area of the contour
map measured was narrower than that of the computation. This is because the probes were not set on the hull
surface above the still-water plane where elevated free
surface reaches only when the model is in forward motion. However, the level of agreement was satisfactory.
The asymmetrical pattern of the pressure is well
realized in the computation.
Figure 12 shows a comparison of wave heights on a
contour map. The wave height close to the hull was not
measured in the experiment because we could not set
the wave height probes in the path of the towed model.
Although the main wave patterns are captured in this
simulation, crests of small wavelength are dissipated
rapidly in the propagating process. This is because of

H. Akimoto and H. Miyata: Predicting sailing boat performance

39

ity of 30 per nondimensional time unit. This is about


twice as fast as in the nominal tacking motion of IACC
racing boats. Only heave and roll motions are solved,
and the other four modes of motion are fixed for simplicity. The axis of roll is set at the center of gravity of
the boat. Figure 13 shows the pressure and velocity
distributions in the transverse sections at two different
times. It shows the occurrence of a high-pressure region
due to the acceleration of the hulls surface, and the
large deformation of the free surface around the boat in
section (a). Although comparative experiments have
not yet been performed, the results show the potential
of this method to treat the unsteady rotational motions
on the free surface.
Comparison of course-changing maneuvers
This section considers an example of a typical maneuvering motion in upwind sailing. An IACC boat (JN35)
is in sailing motion at a constant speed of 9 knots with a
true wind angle of 45 and a leeway angle of 3. Initially,
the angles of its rudder and trim tab are adjusted to
balance force and moment. The boat is then ordered to
change its heading by 4, so that it obtains a larger sail
force according to the increased true wind angle.
This maneuver is performed by automatic control of the
rudder. The control routine of the rudder is a combination of proportional and differential control methods
df
dj
= Gpj + Gd
dt
dt

Fig. 12. Wave-height contours. Positive values are shown as


bold lines and negative values as thin lines. The contour interval is 1 10-3 of L, heel = 25, leeway = 3.0, Fn = 0.34, Re = 105
(computed) and 4.2 106 (measured)

insufficient resolution by the coarse free surface grids


and their rapid outward expansion. However, the agreement of the wave profile along the hull suggests that the
main part of the flow aound the hull is qualitatively well
reproduced.
Forced roll motion
To show the applicability of this method to the unsteady
motions of a boat, a simulation of a forced rolling
motion was conducted. In this simulation, an IACC
boat sailing at a constant speed (Froude number 0.35) in
an upright position starts rolling with an angular veloc-

(23)

where f is the rudder angle, j is the error of the heading


angle, and Gp and Gd are gain parameters. In this case,
pitch, roll, and heave motions are fixed to simplify
the situation. Two cases of simulation were performed
for different gain values. In case 1, Gp and Gd were set
at 3.75 and 3.0, respectively, and in case 2 they were
1.88 and 3.0, respectively. The maneuver in case 2 was
expected to be gentle with a smaller proportional
gain.
Figure 14 shows the pressure distributions on the hull
in different times. At the beginning of the maneuver,
there is a high-pressure region in the bow due to the
yawing motion. Then the pressure magnitude decreases
when the angular velocity is approximately constant.
The time-historical variations of the boats speed in Fig.
15 show that in case 2, the boat is accelerating more
smoothly and reaches a higher speed. The result implies
that the maneuver in case 2 is superior to that in case 1
in this situation because it steers the boat relatively
gently. Figure 16 shows a series of pictures of the
maneuvering motion.
This simulation of changes in course shows that the
present method can be used to predict and to optimize
the performance of sailing boats. The simulation shows

40

H. Akimoto and H. Miyata: Predicting sailing boat performance

Fig. 13. Pressure and velocity distributions


in rolling motion at two different times.
Contour interval = 0.04, Re = 106. Positive
(negative) values are shown by solid
(dotted) lines

fluid-dynamic data of the unsteady motions of a ship.


It is difficult to obtain these data from experiments
because of the limitations of tank facilities and the
complexity of boat motions. Thus, these examples show
the potential of this simulation method in the unsteady
motion of vessels on a free surface.

Conclusion
A new simulation method for the flow around a sailing
boat in motion has been developed. The method is

based on the time-marching, finite-volume method and


the moving grid technique, with consideration of the
unsteady motion of the free surface and the deformation of the under-water geometry due to the motion.
The equation of motion of the boat was solved simultaneously with that of the flow field around the hull using
the fluid-dynamic forces and moments obtained from
the flow solver. The sailing condition of the boat was
virtually realized in the simulation. The simulation results given show the potential of this method to evaluate
unsteady large-amplitude motions of moving vessels on
a free surface.

H. Akimoto and H. Miyata: Predicting sailing boat performance

41

Fig. 15. Changes in forward speed Vs = |vG|/U in time

Fig. 16. Visualization of the maneuvering motion in 4 of


course change, case 2. a t = 4.0 (maneuvering start); b t = 4.6
(elapse time 0.6); c t = 5.2 (elapse time 1.2)

Fig. 14. Pressure (Cp) distribution on the hulls surface in


maneuvering motion at different times. Contour interval =
0.04, time interval = 0.25, Re = 105, positive/negative values are
shown by bold/thin lines

42

Acknowledgments. This research was partly supported


by a Grant-in-Aid for Scientific Research of the
Ministry of Education, Science and Culture and by
the AC Technical Committee of the Ship and Ocean
Foundation.

References
1. Mitsutake H, Miyata H, Zhu M (1995) 3D structure of vortical
flow about a stern of a full ship. J Soc Nav Archit Jpn 177:111
2. Kawamura K, Miyata H, Mashimo K (1997) Numerical simulation of the flow about self-propelling tanker models. J Mar Sci
Technol 2:245256
3. Ohmori T, Fujino M, Miyata H (1998) A study on flow field
around full ship forms in maneuvering motion. J Mar Sci Technol
3:2229

H. Akimoto and H. Miyata: Predicting sailing boat performance


4. Milgram JH (1993) Naval architecture technology used in winning
the 1992 Americas Cup match. SNAME Trans 101:399436
5. Akimoto H (1996) Development and application of a CFD simulation technique for a hull in 3D motion (in Japanese). PhD thesis,
University of Tokyo
6. Rosenfeld M, Kwak D (1991) Time-dependent solution of viscous
incompressible flows in moving coordinates. J Numer Methods
Fluids 13:13111328
7. Baldwin B, Lomax H (1978) Thin-layer approximation and algebraic model for separated turbulent flows. AIAA-Paper 78257
8. Smagorinsky J (1963) General circulation experiments with
primitive equations. Part 1. Basic experiments. Mon Weather
Rev 91:99164
9. Kawamura T, Miyata H (1995) Simulation of nonlinear ship flows
by density-function method (2nd Report). J Soc Nav Archit Jpn
178:17
10. Sung CH, Tsai JF, Huang TT et al (1993) Effects of turbulence
models on axisymmetric stern flows computed by an incompressible viscous flow solver. Proceedings of the 6th International
Conference on Numerical Ship Hydrodynamics, pp 387405

You might also like