Вы находитесь на странице: 1из 8

Resistance Prediction for Fast Displacement

Sverre Steenl,

Hans

J@-gen

Rambech],

Rong

Zhaol

and Knut

Catamarans

J. Minsaas2

Resistance prediction is usually carried out at several stages in the design process. At an early stage, a quick
estimate with a minimum of required input is needed, and when the design proceeds towards completion,
more exact methods are required to catch the effect of the important details. In this paper methods for early
estimates, as well as more detailed numerical estimates and the final prediction based on model tests are
presented. A novel form factor concept deriving the form factorfiom the resistance at high speed is presented.
It is found that the empirical method gives a good estimate of the resistance, and that the numerical method is
a valuable supplement for investigating the effect of smaller hull line changes that is not catched by the
empirical method.
INTRODUCTION
During the last two or three decades, fast displacement
catamarans have become an increasingly popular means
of passenger transport world-wide. However, fast ferries
have huge power consumption.
Thus, the optimisation
with respect to resistance and propulsion, as well as exact
knowledge of the required power is very important to any
fast ferry project. Through our work for numerous
Norwegian
and international
builders of high-speed
catamarans, MARINTEK has collected a very valuable
database of results and experience
in the field of
prediction
and optimisation
of resistance
for fast
displacement catamarans. In this paper we will present
our methods in the three stages of power prediction and
optimisation:

Empirical methods in the very early stage

Advanced numerical methods as a tool for hull shape


optimisation
.
Model tests for final optimisation and verification of
resistance and power consumption.
We will start with the experimental method, since that is
the reference and the basis for the empirical method.

tested without the disturbance of the air speed caused by


the carriage. Thus, when the models are tested with
modelled super-structure, they will get the correct wind
resistance and wind-induced trim moment. The models
are tested free to heave and trim, but fixed in sway, yaw
and surge. In addition to resistance test, propulsion tests,
both with propellers and waterjets are carried out on a
routine basis. Propelled models are usually tested free to
surge, with application of tow rope to correct for the
larger viscous resistance in model scale. We will now
explain the method for prediction of full scale resistance
from measured model resistance.
The total resistance coefficient for model is given as:

=
TM
= cFm+CR tC&m
P.
m ~.v:.
sm

From this, the residual

resistance

coefficient

is found:
(2)

for full scale

is then:

CR= cTm- cFm- cAAm


The total resistance

coefficient

(1)

+ AcF)+CA + CAAS
CT,= cR + (cF,

(3)

Both CF~ and CF$are calculated according to the ITTC57

co~dation line: CF =

0.075

(4)

(logRn - 2)2
EXPERIMENTAL

PREDICTION

OF RESISTANCE

MARINTEK operates a 260 meter towing


wave
maker.
Since
1980,
about
40
displacement catarrwans have been tested.
carriage has a free-to-surge rig in front, This
slender structure to which the high-speed
connected. The free-to-surge rig enables the

tank with a
high-speed
The towing
is a light and
models are
model to be

If the model are tested with representative superstructure,


which is the standard, then CM.=CM,, so that both can be
eliminated from the above formulas. ACF is an empirical
correction for the roughness
of the Ml scale hull,
expressed as:
ACF = ~ 10.31. (H . V$)O2403.33]. C:.
This method for scaling resistance from model
scale has been found to be reliable and satisfactory

~ MARINTEK P.O,Box 4125 Valentirdyst, N-7450 Trondt@m, Norway


Norwegian University of Science and Technology, N-7034 Trondheim, Nonvay

915

(5)
to

full
for a

long time. However, one drawback with the method is


that the absence of a form factor leads to large values of
the correlation coefficient CA. Depending on the vessel
size, Froude number and type of propulsion, C,4 is in the
range from -0.15.103--0.38-10-3.
The uncertainty of CA
is significant, since it is based on full scale measurement
results, which are known to be prone to uncertainties due
to lack of control of the test conditions. The well-known
form factors that are derived from the resistance at low
speed, seems not quite suitable, due to the effect of the
transom stern and the large difference in flow pattern
between the low speed regime used to derive the form
factor and the high speed regime of interest. Thus, an
alternative formulation has been derived that uses the fact
that the wave resistance will vanish (relative to the
viscous resistance) at very high speeds. The form factor is
found by assuming that CW is proportional to l% at high
Froude numbers. The exponent x is found as a leastsquares fit of the experimental data to the line given by:
CT.

- C.4A.

=~+(l+k)

(6)

CF.

When x is determined, k can be found. Thk is illustrated


in Figure 1. Then, the wave resistance coefficient is found
from model resistance as:
(7)
Cw = Crm(l+k). Cr. -CAAm
The form factor k is assumed to be equal in model and
full scale. The total resistance
coefficient
is then
expressed as:
CT. = Cw + (CE, + ACF). (1+ k)+ CA +Cfi,

(8)

The correlation
coefficient
CA to be applied in this
method will be significantly smaller than the CA of the
formerly referred method. Thus, the same percentage
error in CA will give smaller impact on the accuracy of
the final prediction.
This might be attractive if the
correlation material is scarce. However, this concept
should be further verified before the existing method is
replaced. A problem with the form factor method is that
because the exponent x for the Froude number is
determined by a least squares method, the form factor
will be quite dependant on the curvature of the (C~~CMJ-curve,
not only on the level at high Froude
numbers.

EMPIRICAL

y=

The first stage in the development was collection and


systematisation of the results from previous model tests
into a suitable format, and interpolating the results so that
all results gave data for the same five discrete Froude
numbers: 0.6, 0.7, 0.8, 0.9, and 1.0. The second stage was
an investigation of which parameters were giving the best
correlation. Three non-dimensional resistance parameters
were applied initially: C~$, CR, and CW, where the relation
between CR and Cw is expressed as: CR= CW+CFk, where
k is the high-speed form factor. These three resistance
coefficients were plotted as functions of length Froude
number

2.

125

and

the

non-dimensionalised

~n
and Froude number
Correlation on CR (where CR is computed in the
sense of the MARINTEK standard method for
high speed craft, as described in the previous
chapter).
CR is represented
as an empirical
function of L/V and Froude number
Correlation on CR with correction for Bfl-ratio,
s/L ratio and the ratio S/Vm
Correlation on CW. Cw = CT - (1+ k). C~ where

k is the form factor and C~ is the fictional


resistance coefllcient.
Correlation on Cw with correction for B/T-ratio,
5.
sIL ratio and the ratio SIVm
interpolation
and
The program
also incorporates
extrapolation functions, as well as calculation of total
resistance and power consumption
based on specified
efilciency,

1.3

0.5

3.
4.

0.1 OIX+l.32

1
0

Fn = V/@

geometrical parameters length-displacement


ratio L/Vn
and wetted surface-displacement
ratio S/V2n. It was found
that UVB gave much better correlation than S/V2n. Thus,
Fn and L/Via are the primary variables of the method,
while S/Vm as well as WT-ratio (B is demi-hull beam)
and hull separation-length s/L-ratio have been included as
correction terms. The results were implemented
in a
computer program named Catres, with the following five
methods available:
1.
Correlation on total resistance coefficient Cr.,.
C~J is represented as an empirical function of

1,55

12

PREDICTION

An empirical resistance prediction method is useful for


early design stage resistance
predictions
and as a
representation
of the average level of resistance for a
vessel of given dimensions.
This latter aspect is of
importance for an operator of an experimental facility.
The empirical resistance prediction method presented
here was developed as a part of the strategic research
programme SKIPRO 2001. It is based on our database of
results from model tests with a large number of high
speed dk.placement and semi-displacement
catamarans.

1.6

RESISTANCE

1.5

1
2

From the derivation


of the above five resistance
prediction methods, it is clear that the most important
parameter for the resistance is the length-displacement
ratio. In fact, it is possible to give a resistance prediction
accurate
to a few percent
based on length and
disr)lacement
only,
.- as lorw as the hull has a reasonably
.

Fn==YcFm
Figure 1 Determination of form factor from experimental
results on a fast displacement catamaran

916

conventional round-bilge hull form. This is a noteworthy


conclusion, since it shows that without going to exh-eme
measures like dynamic lift, foils or air cushions, there is
not very much that can be achieved in terms of reduced
resistance when length and displacement is fixed.
Figure
2 shows
the empirical
residual
resistance
coefficient of method 2 of the empirical method. It is seen
that there is a huge advantage of high slenderness,
especially at Froude numbers close to the main resistance
hump at Fn=O.5.

Method

3 Correlation

on CR with correction

This method is equal to method 2, except that CRW has


got an additional correction
term that includes the
empirically calculated effect of draught, beam and wetted
surface. Thus, the total resistance coefficient is expressed
as:

cn$=c;m(
+d(k3+k4H
ki
-I- (CF$ +

ACF ) + CA

+ CAAS

where kl to /Q are empirically calculated coefficients that


are independent of Froude number, s is hull separation
and B is demi-hull beam. The expression for CRcmpis
equal to the one in method 2. This method give better
prediction when values of the values of S/Vm, B/T-ratio
or s/L-ratio are far from the mean values of the
regression.
The mean values of the regression
are
implemented in the program as default values, so that it is
easy to know when you are far from the centre of the
regression.
Method 4 Correlation

on CW

Here, Cw means the residual resistance with the viscous


pressure resistance
subtracted.
The viscous pressure
resistance is expressed by the form factor k. In this case,
the form factor is found by assuming that Cw is
proportional to Fnxat high Froude numbers, as explained
in the experimental method above. The total resistance
coefficient is expressed as:
CT. = C~mp+ (CR, + ACF ). (1+ k)+ C~4,,

(12)

where Cwemp is the empirically calculated value of the


wave resistance coefficient, expressed by different fifth
order polynomials for each of the five Froude numbers
supported by the method. An empirical relation for the
form factor (l+k) is implemented. This empirical relation
is shown in Figure 3. Figure 2 Residual resistance coefficient
slenderness and Froude number.

as jimction

of
1.4.

1.2.

Method

Correlation

The total resistance

on CT,

coefficient

is expressed as:

CT== Cp, + cm,


(9)
where CT$mp is the empirically
calculated
value,
expressed by different fifth order polynomials for each of
the five Froude numbers.
Method

2 Correlation

$r
c

1.

~.~ -

22%
+

*
+

f
E 0.6.
8
- 0.4.

0.2.
Oq

on CR

coeftlcient

y=3.4 275xaW

The total resistance

,,

Jh-

is expressed as:

CT== C:w + (CF, + ACF) + C4 + CM.

Lengthdisplecement

10

11

12

ratio L/Vw

(lo)

where CRempis the empirically calculated value, expressed


by different fifth order polynomials for each of the five
Froude numbers. The frictional resistance coefficient CF,
is calculated by the ITTC 57 correlation line, while the
roughness
correction
ACF is the same as for the
experimental procedure. The resulting regression formula
for CRMPis illustrated in Figure 2.

Figure 3 Empiricrd relation for the form factor ( l+k)


Method

5 Correlation

on Cw with correction

This method is equal to method 4, except that Cwmlhas


got an additional
correction
term that includes the
empirically calculated effect of drau~ht, beam and wetted
su~ace. Thus, the total resistance co~fficient is expressed

917as

Cr.

= a(k+ka(k+
C;mp.

functions do not contribute


predictions.

kl + kz

significantly

to improving

the

+ (CF, + ACF) . (1+ k)+ CM,

NUMERICAL
RESISTANCE

(13)
where kJ to k6 are empirically calculated coefficients that
are independent of Froude number. The values for ki are
different
for correction
of CRP and CWeW. The
expression for Cwew is equal to the one in method 4. In
the same way as method 3, this method is used for vessels
that are not quite typical for the regression material.

method

i CT,
2 CR
3

CR with correction

Cw

A DWL

4.7 %

3.970

3.7

B DWL

~0

4.8

5.0

5 Cw with correction

A WLl

4.470

?404.3%
~0 3.3 %

3.9 %

5.1 Y.

5.2

B WI

9.5%

13.3%
13.4%

5.7 %

6.5 %

5.9 Y.

6.4 k

vessel, ~ = (x,y,z) has been chosen. The surface z


the mean water surface when the forward speed
ship is zero. A velocity potential ~ = $ +
introduced. Here U is the forward speed of the ship
satisfies the 3D Laplace equation

Table 1 Maximum error of Catres prediction relative to


model test based prediction. Include; extrapolated part of
resistance curve.
Computational

method

ADWL

AWLI

BDWL

4.47.

3.5 Y.

5.2 ZO 3.8 %

2 CR

4.7 %

0.9%

6.1 %

7.1 %

3 CR with correction

3.7 %

1.1 k

5.4%

7.2%

3.3 k

5.7 ~o

2.5%

5.9%

2.2 %

Cw

5 Cw with correction

5.0

~0

3.970

5.1

YO

v~=o

BWL1

1 cr.

WAVE

Theory
A potential theory is used to solve the wave resistance
problem since the viscous effects are neglected here. The
problem is solved as a steady problem.
A right-handed coordinate system that is fixed in the

yOt0.4%

9.9 v.

OF

Introduction
Since the viscous resistance of high-speed displacement
catamarans is quite well predicted by friction coefficients
or boundary layer theory, we are looking for a numerical
method to predict m
resistance of displacement ships
at high forward speed. The method should also be robust
and easy for an engineer to use. Since the demi-hulls are
slender for high-speed displacement vessels, the 2% D
solution may be used. The 21%D method means that one
use two-dimensional
Laplace
equation
and threedimensional free-surface conditions. The 2V2 D methods
have been much used in the seakeeping problem, see for
instance Faltinsen and Zhao (199 la, 1991 b). The 2Y2 D
solutions are valid for slender body at high Froude
number. The methods can be applied for predicting wave
resistance of high-speed displacement
ships. Since the
hull is slender and the linear solutions predict quite well
the wave resistance for relatively slender ships, our idea
is to introduce non-linear corrections to improve the wave
resistance prediction. We start with the linear problem
with linear free-surface conditions satisfied on the mean
water surface. The non-linear terms will be included in a
similar way as for the second order problem of mean
wave drift force. That means that the quadratic term in
the Bernoullis equation and a water line integral are
included.

Verifhmtion
Verification
of an empirical method is never really
completed. It will always be of interest to see how good it
is at predicting the resistance of another design. Thus, the
verification carried out in this study is only an example of
the accuracy of the prediction you get.
The verification example included here is for two vessels.
Both are passenger catamarans in the 30-40 m range.
Model A is driven by waterjets, while Model B is driven
by propellers. Both are run at two draughts, DWL and a
deeper draught (IVL 1). The model test results of these
vessels is not a part of the experimental material used to
make the empirical formulas in Catres. Note that the
values above Fn=l.0
in the Catres predictions
are
extrapolations.
The tables below show maximum deviation between
model test based predictions and Catres predictions. In
table 1, the entire speed range that has been model tested
is included, while in Table 2, only the speed range up to
Fn= 1.0 is included. The values are in % of total
resistance.
Computational

PREDICTION

= O is
of the
Ux is
and @
(14)

in the fluid domain.


Following Newman (1976) the non-linear dynamic free
surface condition on the exact free surface can be written
as

Tabk 2 Maximum error of Catres prediction relative to


Does
not include
model
test based
prediction.
extrapolated part of resistance curve.

on z = <(x,y)
The kinematic free surface condition is

The verification material is too small to make definitive


conclusions, but on basis on the existing material one can
conclude that correlation on CW gives better results than
correlation on CR. The error of a prediction based on Cw
is less than 7% of total resistance for both test vessels on
both waterlines. For these two vessels the correction

(15)

Here ~(x,y) is the free surface elevation.


The body boundary condition
SB can be written as

918

on the exact body surface

(17)

where

r = [(y~)z

+(z~)z~s,

ij = (q,;),

SB the body

surface, SF the free surface, & the control surface far


where

; = (nl, n2, n3) is the normal vector on the body

surface. Positive direction of Z is into the fluid domain.


The problem is solved by using a 2% D method. The
solution is based on the linear classical free surface
conditions
on the mean water surface. When one
calculates wave resistance,
non-linear
terms will be
included in a similar way as for the second order problem
of mean wave drift force,

away from the body,

Linear and second order 2%D solution


Ogilvie (1977) showed that the linear 2Y2 D solution
represents the diverging wave systems caused by the ship
in the far-field. That means that the transverse wave
systems are neglected in the 2% D solutions. This is a
good approximation
for high-speed displacement ships
with Froude numbers in the order of one.
Here one assumes that the ship is fairly slender and
introduces a slenderness parameter e, which is the ratio
between the beam (or draught) and the ship length. One
assumes

that

~=

O(fEx),

$=

O(f&-*)

and

$ = O(fc-] ) , where f is any flow variable caused by the


body in the region close to the body. Further one assumes
that nl = O(c) where nl is the x-component of a unit
normal vector to the wetted part of the body surface. The
assumptions follow the similar approach of Faltinsen and
Zhao (1991 a,199 lb) in solving the problem of ship
motion for high speed vessels. Based on the assumptions
the three-dimensional
Laplace equation (eq.(14)) will turn
into a two-dimensional
Laplace equation for each cross
section.
When neglecting the non-linear terms in the equation (15)
and (16), the linear dynamic and kinematic he surface
conditions to be satisfied on the mean water surface can
then be written as

(18)
I afb
&=__onz=o
ax

(19)

u dz

The potential @is solved by using Greens second identity


for each section. The velocity potential @ at a point (y,z)
in the fluid can be written as
2?r@(y,z) = ~ ,$,+~c logFn
[

alogr
f#(ti)~

is the derivative

+C* (x)

(y +Z)

919

(21)

(y+2)

which can be

where Cl(x) and C2(X) are coefficients


obtained in the solution of equation (20).

A solution can be found bv starting at the bow,


equation (18) and (19) to st~p the solutions of the
surface elevation ~ and the velocity potential $ For
cross section a two dimensional problem is solved.
start conditions are ~ = O and $ = O.

use
free
each
The

The 2!/2 D methods are valid when the local Froude


number is in the order of O(l). Thus, this method can be
used for the whole ship for high speed vessels. It was
shown by matched asymptotic expansion by Faltinsen
(1983) that the linear 2% D solution is an approximate
solution for the linear 3D solution in the bow region of
conventional slender ships.
The pressure P on the wetted body
obtained by the following equation
P=-pu

~f

surface

can

be

+[[$J+HIPV
22)

The wave resistance R. can be written as

RW=

(23)

~~Pn, dS

Here SB is the wetted body surface.

Total Resistance Prediction


The total resistance is predicted from the numerically
computed wave resistance by addition of empirically
based viscous resistance, according to the formula:

Cr, =$ +(C,, +icF).

(l+k)+cA

(24)

where Cwmis the wave resistance coefficient computed


by Waveres, according to the method described above. A
correct determination
of the form factor (1+/c) is the
critical point for a good total resistance prediction. In
order to compare numerical calculations to model test
results presented as residual resistance, the following
expression is used to compute CR from CWW

1O!s(ij)
(20)

in the cross-

plane. The positive direction of fi is into the fluid. For


each section the contribution from the free surface far
away from the body can be rewritten. The velocity
potential $ for Iyl>b(xl), where b(xl) is large relative to
the cross-dimensions
of the hull can be expressed as a
vertical dipole (symmetrical
part) and a multipole
(asymmetrical part). t) can be written as (for Iyl>b(xl))
@= C,(x)

The method has been implemented


in a computer
program called Waveres, written in Fortran 90 as part of
the strategic research programme SKIPRO 2001.

CR =C;m+C~m.

(25)

3.OE-037

2.5E-03 .

3
Q

2.OE-03 .

~
s

C.R, ModelTest
c-w

C_~, k=l .324

...0...
QR kyrl.og

3.oE-03 -

~
al
o

2.5E-03 - h

v
1.5E-03 -

j
= 5.OE-04 %
K
O.OE+OO+
0.7

C.R, ModelTest

C.w

C_R: k=l .4239

. ..0..- C_U k=l .2


~

1.5E-03 -

:
=
#

1.OE-03 -

5.OE-04 -

8
W O.OE+OO0.8

0.9
Froude nutier

1.1

1.2

0.7

0.9

0.8

Fn [-]

1.1

1.2

Froude number Fn [-]

Figure 4 Comparison

of residual resistance from model


test and numerical calculations for Model A

Figure 5 Comparison of residual resistance from model test


and numerical calculations for Model B

By use of this expression, a correlation fork can be built


by comparison between a number of experiments and
calculations.

is most probably caused by an error in the sinkage at that


particular speed.

Verifications and Validations

COMPARISON OF THE DIFFERENT


TION METHODS

The theory and the computer program implementation


have been verified
by comparing
with analytical
solutions. The method has been verified by thin ship
theory of Michell integral (Tuck (1988)) for the wave
resistance of a parabolic strut.

PREDIC-

This section will compare results from model tests,


with
Catres
and
resistance
predictions
empirical
numerical predictions
by Waveres. On basis of the
examples, the use of the different methods is discussed.

We have also done a number of comparisons between


model test results and calculations
of real slender
(semi)-displacement
catamarans. Results in terms of CR
and Cw are shown in Figure 4 and Figure 5 for model A
and model B respectively. It is seen that except for one
bad point, the trend is very well represented. To get the
absolute value correctly, the form factor must be correctly
estimated. In Figure 4 and Figure 5, the highest form
factor value is found from the experimental
data
explained in the first section of this paper. The lower
form factor value is found as the one that gives least
discrepancy between CR from model tests and CR from
the numerical calculations, using equation (25) to convert
from Cw to CR. The form factor derived from the model
test agrees well with the empirical relation given in
Figure 3. The form factor found by minimizing the
difference between computations and model tests is much
smaller, and agrees better with conventional form factor
formulas, like the one found from the Holtrop 84
resistance prediction method. The reason why model B
has larger form factors than model A is that model B is a
propeller
driven boat, tested with propeller
shafts,
brackets and rudders. When corrected for appendix
resistance, the form factor is similar to the one found for
model A, which is a waterjet propelled boat, model tested
without any appendices.
It is found from comparisons between model tests and
numericrd calculation that the sinkage, especially at the
stern, is very important for the results. Thus, it is
necessary to be able to predict the change of trim with
speed that is observed in the model tests in order to get
the slope of the resistance curve correct. It is found that
the hump in the computed resistance curves in Figure 4
920

Figure 6 shows the wave resistance coefficient Cw for


three different models, derived from model tests, and
calculated with the empirical and numerical methods
described previously in this paper. The CW from model
test is co,mputed using equation (25) and a form factor
derived from the model test data as described in the first
section of this paper. The empirical predictions are made
with method 4 for models A and B, and method 5 for
model C. The method with corrections are used for model
C since that vessel is extremely slender, thus having
parameters
away from the centre of the empirical
material.

0.7

0.8

0.9

Froude nu~r

1.1

!A)del TW -.=.-.

/% Empirical

-Q-

R Numerical

B;

Mel

B; Emplrkel

-*-

B Numerical

~G,

Ws3elteet .- A--- C; Empirical

-*-

C, Num*al

teat ..-*--

1.2

Fn [-]

lgure 6 Comparison of wave resistance from model tests,


numericrd and empirical calculations

that is presented here shows good results with respect


predicting the relative differences in resistance between
different
designs.
In order to get good absolute
resistance predictions from a wave resistance calculation
it is necessary to get a reliable estimate of the form
resistance. This can be done empirically, by boundary
layer calculations or by use of a RANS code. The results
presented here indicate that the form factor (l+k) to be
added to the calculated wave resistance is in the order of
1.0-1.1, when effects of appendices are excluded.

30~

The novel form factor method presented here still has


obvious weaknesses. The form resistance obtained by
this method
seems
to contain
other
resistance
components than viscous resistance. Furthermore,
the
estimation of the form factor from model tests is
sensitive to the curvature of the (C~~-CM~)-curve, IIOt
only the level at high speed. This seems not to be in
accordance with the principles on which the method is
based. One must be able to subtract more resistance
components,
like for instance
the transom
stern
resistance, in order to make this method practical for use
together with numerical methods.

O.OE+OO~

0.7

0.8

0.9

1.1

1.2

FroudenumberFn[-]
AA
~
C;

F&Id Test.. Q..

A Empirical

El thdel test - .-e-.. B; Empirical


!Axtai teat -. +--

C; Empirical

---~

Numarkal

-+-B;

Numerical

-+--C;

Numerical

Figure 7 Comparison of residual resistance from model


tests, numerical and empirical calculations
Figure 7 shows the residual resistance coefficient CR for
the same three models as in Figure 6. CR is computed
from the numerically calculated Cw by use of equation
(25) and the form factor found from the model test using
the method described previously. The empirical CR is
computed using method 2 for models A and B, and
method 3 for model C.
From Figure 6 and from Figure 7 it is seen that the
numerical method distinguish well between the different
models, predicting quite well the relative differences.
From Figure 7 it is seen that the empirical methods
predict the resistance level very well, but they have
difficulties dktinguishing
clearly between the models A
and B, since they have very similar main dimensions. The
reason why the prediction
of the wave resistance
coefficient
is not similarly
good seems to be the
uncertainty of the form factor. The form factor derived
with the method described in the first section of this paper
is too huge to give good predictions from the wave
resistance calculation. It seems that the form resistance
derived this way contains resistance components that are
already taken into account by the numerical method, like
for instance the transom stern resistance. The problem of
selecting a proper form factor is also believed to be the
reason for the less good prediction of the resistance level
by the numerical method.

REFERENCES
Faltinsen, O. Bow flow and added resistance of slender
ships at high Froude number and low wave lengths, J.
Ship Res., V01,27, pp. 160-171, 1983.
Faltinsen, O. and Zhao, R. Numerical predictions of ship
at
high
forward
speed,
Philosophical
motions
Transactions of the Royal Society, series A, 1991.
Faltinsen, O. and Zhao, R.Flow predictions
around
high-speed
ships in waves, M. P.Tulins festschrift
Mathematical Approaches in Hydrodynamics,
SIAM,
1991.
Hohrop, J., and Mennen, G. G. J., An approximate power
prediction method, International Shipbuilding Progress,
Vol. 29, No. 335, July 1982
Holtrop, J., A statistical re-analysis of resistance and
Shipbuilding
Progress,
propulsion data, International
Vol. 31, No. 363, November 1984
Newman,
J.N. Linearized
wave resistance
theory,
International seminar on wave resistance, Japan, 1976.
Ogilvie, T.F. Singular-perturbation
problems
hydrodynamics, Adv. Appl. Mech. 17, 1977.

CONCLUSIONS
Model tests are still the most reliable way of determining
the resistance of a high speed catamaran. The empirical
method presented here shows good results as long as the
design is not too far from the ones used to make the
resistance regression. But since the number of parameters
that go into the regression is limited in order to make it
useful in an early design stage, it is not that well suited to
study changes in hull lines and details in the hull design.
The numerical method for prediction of wave resistance

921

in ship

Rambech, H.J. Emptilcal Resistance Prediction of Fast


Displacement Catamarans Thesis, Inst. Marine
Hydrodynamics, Norwegian Institute of Science and
Technology, 1998 (in norwegian)

LIST OF SYMBOLS
SW!!X!
Explanation
B
Bdemi

CA
Cu
cj7
cR
cRev

Cv

Beam of hull
Beam of demi-hull
Correlation coefficient
Air resistance coeftlcient
Frictional resistance coefficient
Residual resistance coefficient
Empirically calculated residual resistance
coefficient
Total resistance coefficient
Empirically calculated total resistance
coefficient
Viscous resistance coefficient

Cv=(cF+dcF)
*(I+k)
CUJ Wave resistance coefficient
CwempEmpirically calculated wave

g
H
k
L
P
RAA

resistance
coefficient
Numerically calculated wave resistance
coefficient
Froude number. Fn=V/sqrt(g*L)
Acceleration of gravity
Hull roughness (in p)
Form factor (for high Fn)
Length of hull (in waterline)
Pressure
Air resistance

Rn
RT,

Reynolds number
Total resistance (full scale)

RW

Wave resistance

T
s

Draught
Hull separation (distance from centerline to
center of demi-hull)
Wetted surface
Water velocity
Speed
Increase of frictional resistance due to hull
roughness
Water density

Fn

u
v
ACF

Displacement
Velocity potential
Free surface elevation
Subscript m means model scale value
Subscripts means full scale value

922