See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/237586367

A Critical Analysis of Resistance Prediction

Using Regression Methods for High Speed Hull
Forms

Article January 1995

CITATION READS

1 284

4 authors, including:

Prasanta Sahoo
Florida Institute of Technology
47 PUBLICATIONS 85 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

An oblique 2D+T approach for hydrodynamic modeling of yawed planing boats in calm water View
project

All content following this page was uploaded by Prasanta Sahoo on 14 October 2015.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
A Critical Analysis of Resistance Prediction Using Regression Methods for
High Speed Hull Forms

by

Michel de Vos, Brett Murrie & Viliami Siale

Faculty of Maritime Transport & Engineering
Australian Maritime College
Launceston, TAS 7250, Australia

and

Prasanta Sahoo
Australian Maritime Engineering CRC Ltd.
Launceston, TAS 7250, Australia

Summary

The aim of this work has been to look closely at the various regression methods published to-
date and propose a suitable regression method to predict the resistance of high speed crafts in
the preliminary design stage.

Literature review indicates a considerable amount of work in the field of resistance prediction
through equations and formulas that represent model test results mathematically. There are
numerous articles which provide useful information to the naval architect by which resistance
of high speed craft can be predicted by use of a handful of hull form parameters. The methods
illustrated in this article have been carefully selected and reviewed and their range of
applicability have been presented.

This work, undertaken as part of a sub-program on "Resistance of High-Speed Marine

Vehicles" of AMECRC, falls into two sections:

a) To develop and validate software to use some of the more popular regression methods
available from the literature.
b) To evaluate the towing tank results of different hull forms which have been carried out at
the Ship Hydrodynamic Centre of Australian Maritime College, against the various regression
methods.

An extensive study of different methodologies has been provided by MacPherson (1993)

clearly illustrating the advantages and disadvantages of using various methods. The above
article states that all of these methodologies have been limited in one way or another, by a
physical reality where the relationship of dimensional volume to water viscosity cannot be
simultaneously scaled from a model's size to the full scale vessel. Attempts have been made
to apply corrections and correlating factors so that full scale resistance can be predicted from
models. No universal procedure has been developed, so these corrections are often applied
inappropriately between methods and the accuracy of the results has suffered as a result.

Models of single chine vessels have been tested over a range of speeds considered adequate
for semi-planing and planing craft and some of their results have been evaluated. A critical
analysis of the main hull form parameters which have the most significant influence on the
resistance has been conducted. A comparative study has been carried out to evaluate and
establish the suitability of the published regression methods against the towing tank results.

1 Introduction

There are numerous methods available today by which a ship designer can obtain an estimate
of the resistance of high speed craft. Studies made by various authors and the present study
show that no single method is accurate enough to predict the resistance over a wide range of
speeds. Investigations made over the last few years show that some of the modern regression
methods are sufficiently accurate over the speed range for which they had been developed
while some of the other methods have been less than satisfactory. While using regression
methods designers generally tend to satisfy the non-dimensional range for their particular
hullform. The most important aspect i.e., the hull shape needs to be considered bearing in
mind the limitations applicable to a particular method. To ensure confidence in accuracy it is
imperative to investigate the regression methods in detail and compare results of a number of
vessels for whom models have been previously tested.

It is the intention of this paper to provide the designer with a reference to select the most
suitable regression method applicable to their particular vessel keeping in mind the limitations
of the method and the hull shape for which it had been developed. One should remember that
such numerical regression methods are not attempting to eliminate model testing but provide a

2
cost effective means of predicting resistance and are a powerful tool in the hands of the
designer of what to expect in the future.

2 Influence of Hull Form Parameters

Before making a statistical analysis of data, it is necessary to decide upon the hull form
parameters that should be used. The selection of the parameters or variables is quite arbitrary
as stated by Fairlie-Clarke (1975). Fung (1991) points out that this selection was quite often
subjected to the judgment of specific researchers' suggestions. Almeter (1993) pointed out that
there would be a problem when trying to incorporate all the variables defining the hull into the
equation. On the other hand, he also warned the danger of omitting too many terms which is
quite preferable for the equation but cannot fully define the hull considered. However, Fairlie-
Clarke (1975) suggested that the regression equation should not contain more than 10
independent variables, because the accuracy of the prediction becomes stagnant with a large
number of variables. Fung (1991) also suggested the error in the prediction decreases with the
increase in number of terms but after 11 to 17 terms, the errors become stagnant and
additional terms produces little or no improvement at all on the error.

It is a common occurrence in the design spiral that changes are made to the shape of a hull
after the initial lines have been drawn and some performance calculations made. Usually these
changes will be relatively minor with the changes made due to some new information being
received. These changes could be estimated displacement, waterline length, beam, transom
area or any of a number of relevant hull parameters. The naval architect may also wish to
explore slightly different variations of some parameters to see their effect on performance. It
would be a fair assumption to say that the vast majority of designers would have a good
knowledge of the effect that a change in the basic parameters would have, but to what extent
would not be as common.

It would therefore be beneficial to explore and quantify the effects that a change in parameters
would have on the resistance of planing hulls over a range of speeds. This would in turn lead
to findings as to which hull parameters have the most significant effect on the resistance of
such vessels. By investigating the results of model resistance tests of a large number of
unrelated commercially operating craft and comparing them to their respective non-
dimensional hull form parameters it is hoped that general trends will be recognised that will
be applicable across the entire spectrum of high speed craft. Results obtained should then be
able to not only identify the effect on performance of changing the design variables but also
indicate their optimum. These trends or patterns if and when identified should also enable the

3
development of a regression equation using the most important hull form parameters. A
previous effort made using this concept for high speed round bilge displacement hulls has
been published by Ping-zhong et al (1980) but as yet no known effort has been made with
regard to planing hulls.

In the past most of the work that attempts to identify the effect of certain hull parameters has
usually been done with the use of a systematic series. These systematic series would usually
involve the extensive testing of a small number of vessels whose forms have been derived by
stretching or compressing the parent model. The resulting forms thus only had a number of
parameters that were varied whilst others were kept strictly constant. This enabled for easy
cross plotting and interpretation of results between models. Unfortunately, some of these
series such as Series 50 have been based on a parent model that has been rather outdated and
as such their applicability to modern vessels is somewhat limited as has been pointed out by
Keuning et al (1993). There have however been a few series that represent modern practices
for hull shapes and have provided some valuable information. The series data currently
available has more than likely reflected a particular viewpoint of the ranges in service and are
therefore sometimes further limited in their application. There is one further consideration that
should be made with regard to systematic series. The effects on resistance of changing a
particular parameter are normally attributed to a direct cause by the parameter in question.
This may not always be the case and it is probable that it is the result of changes in a number
of parameters due to the variation of the single parameter. It is normally true to say that by
varying one dimension of a vessel it will change a number of others and so they are all
interconnected.

To date particular interest has been centred on parameters such as L/B and LCG. For these
parameters quite a lot of data exists. Clement and Blount (1963) presented a paper regarding
the testing of a number of hard chine models to explore the influence of L/B and LCG
positions. The L/B ratios ranged between 2 and 7. It showed that the value of R/ was lower
for higher values of L/B. This occurred up until the volumetric Froude number was
approximately 3.5-4.0 where specific resistance (R/) of the higher L/B values gave higher
results but the difference was much less. It also showed that the lowest value of L/B would be
impractical since the resistance curve would have a rather extreme hump in the drag curve.
The analysis of LCG positions was interesting but showed little that was not already known. It
indicated that as the centre of gravity is moved further aft resistance decreased. There comes
a point however when by moving the LCG position further aft it produces a rather undesirable
hump in the resistance curve. This indicates that there is an optimum range for the position of
the longitudinal centre of gravity. In fact it was stated that the optimum position for the centre

4
of gravity for these tests was between 4 and 8 per cent aft of the centroid of the projected
planing area (Ap). For high speeds the aftermost LCG position gave the best results for all
models. In a practical sense these results appear to indicate that by careful arrangement of
items upon a heavily loaded vessel it is possible to obtain a significant decrease in the
resistance.

The effect of displacement upon the resistance of planing craft is widely known as an
important parameter since one of the basic rules for craft of this type is low weight and large
power. Through tests made at the Ship Hydrodynamic Centre of the Australian Maritime
College some patterns are evident. These tests involve the results of some 14 hard chine
models with a total of 36 test conditions. The models themselves are of boats that are
currently in commercial service and as such are considered to be unrelated.

Fn = 0.5
25

20
Cr x 1000

15

10

0
4.000 4.500 5.000 5.500 6.000 6.500 7.000 7.500
Length - Volume Ratio
1
Figure 1: Variation of Residual Resistance with Length-Volume Ratio ( L 3
).

Presented in Figure 1 are values of residual resistance coefficients plotted against length-
volume ratio at a Froude number of 0.5 which corresponds to that of hump speed. This plot is
typical of those that have been found throughout the range of Froude numbers encountered.
There is a distinct band in which the majority of points lie. This is as would be expected with
the lower residual resistance coefficients corresponding to the higher length-volume ratios. It
is also indicated that for a change in displacement the residual resistance coefficient changes
much less at high Froude number than at hump speed which correlates with the fact that a
heavy vessel requires a much higher engine power in order to enter the planing mode.

Deadrise angle is considered as an important parameter characteristic of almost all planing

hulls. Du Cane (1974) has rightly said that the use of deadrise angle stems from seakeeping

5
considerations but only marginally influences resistance. Increasing deadrise decreases the
vertical component of lift and therefore the efficiency of the vessel as a lifting surface. There
is also an inherent increase in wetted surface with deadrise and since wetted area is an
important component of high speed resistance then so is deadrise. There are however, certain
arguments pertaining to the use of a deep deadrise hull to give the best performance in smooth
water at high speed as Levi (1990) has suggested.

16
6.50
14 6.24
12 6.01
Cr x 1000

5.89
10
5.64
8 5.42
6 5.25
4
2
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Fn

1
Figure 2: Influence of L 3
ratio on Cr values at constant 17o Deadrise Angle.

Unfortunately no conclusive evidence is forthcoming from the investigations made regarding

the model test data of the 14 models. Figures 2 and 3 show a large variation of residual
resistance coefficients for some hulls of equal deadrise. The curves appear to be more
1
dependant upon the L 3
ratio than upon deadrise itself. This suggests that deadrise may
have a very subtle effect which could only be investigated through either a much larger
number of independant vessels or through systematic tests with vessels of similar form.

6
 16
6.50
14 6.24

12 6.01
7.19
10
Cr x 1000

6.53
6.06
8
6

4
2

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fn

1
Figure 3: Influence of L 3
ratio on Cr values at constant 15o Deadrise Angle.

The previous statements have been made using data that exists for hard chine vessels in
smooth water only and as such may not correspond to experiences in conditions that are less
than perfect.

3 Analysis of Resistance Prediction Methods

As previously mentioned the accurate prediction of the resistance of a vessel is of extreme

importance, and thus any prediction method should give accurate results over the whole range
of the method's validity. Studies have shown that there are numerous methods which have
been developed encompassing a vast array of ship types and as such the selection of which
method to use for a particular vessel is of fundamental importance. Four methods applicable
to high speed hull forms were developed into computer programs. These programs were then
used to validate the methods by comparing the results to the model results of three completely
independent hull forms corresponding to full scale vessels. The salient features of the four
methods that have been discussed here are as follows:

Prediction method of Keuning et al (1993)

Clement and Blount (1963) presented a now well known systematic series planing hull forms
with a deadrise angle of 12.5. In this paper by Keuning et al (1993) the results have been
presented for a similar 25 deadrise angle series which are coupled with results of a 30
deadrise angle series to develop a regression model for total resistance, trim and rise of centre
of gravity. The total resistance of the vessel is calculated by interpolating the (RT/) fitted to

7
separate datasets containing the model resistance scaled to different volume of displacements.
Table 1 summarises the application boundaries of this method.

Table 1: Parameters and range of validity as per Keuning et al (1993)

Lp/Bpx, LCG, Ap / 3 &

2
Parameters
Constraints =12.5-30
Speed Range Fn =0.75-3.00

Prediction method of Almeter (1989,1993)

This paper uses the virtually ignored large Soviet BK and MBK planing hull series to develop
a resistance prediction technique for low deadrise, hard chine stepless planing hulls. The
prediction method is based on test results of 29 systematic models of the BK and MBK series
over a range of loading conditions and longitudinal centres of gravity. The resistance of any
hull is predicted by calculating the resistance of an equivalent BK or MBK model that has the
same m and dimensionless values C, LCG/Lp and Lp/Bm as the ship and then scaling the
total resistance to full scale. The BK series is based on planing vessels such as patrol boats,
supply craft and high speed motor yachts, whilst the MBK series is oriented towards smaller
planing craft. The main features of the two resistance prediction methods are shown in Table
2.

Table 2: Parameters and range of validity as per Almeter (1989,1993)

BK MBK
Parameters Lp/Bm, LCG/Lp, C & m Lp/Bm, LCG/Lp, C & m
Constraints Lp/Bm=3.75-7.00 Lp/Bm=2.50-3.75
LCG/Lp=0.35-0.45 LCG/Lp=0.25-0.45
C=0.427-0.854 C=0.158-0.352
m =12-21 m =7-18
Speed Range Fn =1.00-4.50 Fn =1.00-4.50

Prediction method of Radojcic (1984,1985,1990)

In attempting to minimise the hydrodynamic resistance of a vessel the quasi propulsive

efficiency (QPE) may be decreased by a reduction of hull efficiency. Radojcic (1984,1985 &
1990) aimed to reduce the power rather than minimise the resistance or maximise propeller

8
efficiency as is normally done. A resulting equation for resistance for a 100,000 lb vessel was
produced accompanied by a similar equation, but with different coefficients for dynamic trim.
In addition equations for wetted surface and mean length were developed to facilitate the
scaling of the 100,000 lb to that of the actual vessel. Table 3 gives a summary of the method.

Table 3: Parameters and range of validity as per Radojcic (1984,1985,1990)

Ap / 3 , Lp/Bpa, LCG/Lp,
2
Parameters
& m
Constraints Lp/Bm=2.36-6.73
LCG/Lp=.30-.448
Ap / 3 =4.25-9.5
2

m =13-37.4
Speed Range Fn =1.00-3.50

Prediction method of Lahtiharju et al (1991)

This paper investigates the resistance and seakeeping characteristics of fast transom stern
hulls. The author found that there was a requirement for a systematic series of shallow draft
and wide transom hull forms to accommodate the modern use of waterjet propulsion. The two
resistance equations for a 100,000 lb vessel were developed, one for hard chine and the other
for round bilge forms, enabling the comparison of the effect of changing from round bilge to
hard chine whilst keeping the other parameters equal. In order to scale the results to the actual
vessel in question correction factors have been provided. A summary of the method is shown
below in Table 4.

Table 4: Parameters and range of validity as per Lahtiharju et al (1991)

Round Bilge Hard Chine

Parameters LWL/Bx, Bx/T, Cx, At/Ax, LWL/Bx, Bx/T, At/Ax &
LWL / 3 & B3 / LWL / 3
1 1

Constraints LWL/Bx=3.33-8.21 LWL/Bx=2.73-5.43

Bx/T=1.72-10.21 Bx/T=3.75-7.54
Cx=0.567-0.888 At/Ax=0.43-0.995
LWL / 3 =4.49-6.81
1
At/Ax=0.16-0.82
LWL / 3 =4.47-8.3
1

B3 / =0.68-7.76

9
 Speed Range Fn =1.80-3.20 Fn =1.80-3.20

Description of Vessels

Three models, all hard chine were used to validate the various methods. As previously stated
these models are completely independent and are models of full sized ships. When expanding
the model results to full scale, full use was made of the ITTC '57 friction line as well as
applying a ship-model correlation allowance of 0.0004. All the numerical methods thus were
extended to include the facility of choosing a correlation allowance ensuring continuity of
model/numerical results. A summary of each vessel's features is shown below in Table 5.

Table 5: Hull form parameters of the vessels.

Vessel 1-A Vessel 2-A Vessel 3-A

Lp/Bpx 3.63 4.32 3.98
Lp/Bm 3.78 4.66 3.75
Ap / 3
2
0.28 0.71 0.25
LCG from Cap (%Lp) -2.28 -2.74 -3.71
m 20.00 18.5 14.00
LCG/Lp from transom 0.41 0.40 0.427
LWL/Bx 4.02 3.78 3.54
Bx/T 4.59 4.89 6.36
At/Ax 0.95 0.62 0.611
LWL /
1
3
4.93 5.38 5.66

4 Discussion of Results

Figures 4 through 6 are graphical representations of model results compared against the results
obtained from the four prediction methods. The results have only been presented for
volumetric Froude numbers against RT/ corresponding to the available model data.

10
 Almeter Keuning et al
Radojcic Lahtiharju(Hard Chine)
Model Results
0.2
0.18
0.16
0.14
0.12
0.1
R/

0.08
0.06
0.04
0.02
0
0.5 1 1.5 2 2.5 3 3.5 4
Volumetric Froude number - Fnv

Figure 4: Predicted versus Model Results Vessel 1-A

Almeter Keuning et al
Radojcic Lahtiharju(Hard Chine)
Model Results
0.25

0.2

0.15
R/

0.1

0.05

0
0 0.5 1 1.5 2 2.5 3
Volumetric Froude number - Fnv

Figure 5: Predicted versus Model Results Vessel 2-A

11
 Almeter Keuning et al
Radojcic Lahtiharju(Hard Chine)
Model Results
0.2
0.18
0.16
0.14
0.12
0.1
R/

0.08
0.06
0.04
0.02
0
0.5 1 1.5 2 2.5
Volumetric Froude number - Fnv

Figure 6: Predicted versus Model Results Vessel 3-A

It is obvious that in some cases each method provides results of sufficient accuracy to be
useful during the preliminary stages of ship design. Observing the prediction method by
Almeter(1993), it is immediately evident that for vessel 2-A some very erratic results occur in
the range Fn =1.00-2.5. Although this vessel is within the range specified for BK series
unsatisfactory results were obtained and further comparisons of model test results against the
prediction method of BK series need to be carried out before any conclusive remarks can be
made. The results for the other two vessels provide reasonable correlation with model results.
No general trends in over or underprediction are evident from such a small range of results
and further testing is to be performed before any additional conclusions can be made.
Focusing on the method of Keuning et al (1993), it is evident that although this method does
give very good results it seems to underpredict over the entire speed range for all vessels
especially at higher values of Fn . It has been suggested by Keuning et al (1993) that by
utilizing a worm screw function , a dramatic improvement in accuracy could be achieved. It
is therefore concluded that this method is of sound nature and could make a suitable tool for
resistance prediction not only in the initial stages of ship design but also in the more
demanding final stages. The method of Radojcic (1984,1985 & 1990) provides good
correlation with model results over most of the speed range. There is some underprediction at
and above Fn =2.5 for vessels 1-A and 2-A. These errors are of such a small magnitude that
they could almost be ignored, resulting in the conclusion that this method is sufficiently
accurate to be used as prediction of resistance for all phases of the ship design. The method
developed by Lahtiharju et al (1991) shows the best correlation against model results for the
speed range over which this method is valid. There seems to be no under or overprediction at
any speed. It is hereby emphasized that this method along with Radojcic's method are

12
extremely reliable. Conclusions just made are based on just three independent vessels and
many more vessels must be analysed before any firm conclusions can be drawn regarding
under or overprediction for any Fn range.

Conclusions

This paper endeavours to present the validity range of some of the regression methods
developed within the last few years. An attempt has been made to establish the degree of
accuracy and confidence one would encounter while using any one of the methods illustrated
in the preceding paragraphs. One should bear in mind that confidence level would be reduced
if the validity range is violated. The hull shape, which is of equal importance, should be
carefully considered while selecting a particular regression method. It is the view of the
authors that the methods developed by Lahtiharju et al (1991) and Radojcic (1984,1985 &
1990) provide reasonable degree of accuracy not only for the initial stages of ship design but
also for more rigorous analysis in the later stages.

Principal Nomenclature

Ap / 3 = Load Coefficient
2
1
L 3
= Length volume ratio
Ap= Projected planing bottom area LCG= Longitudinal Centre of Gravity
Ax= Maximum section area Lp= Projected chine length
At= Transom area RT= Total resistance
B= Breadth V= Speed
Bm= Midship chine beam = Worm screw function
Bpx= Maximum breadth over chines = Deadrise
Bx= Maximum beam m = Midship deadrise
C = Load coefficient, (gBm 3 )
= Displacement
Cr = Residual resistance coefficient = Mass density of fluid
Fn= Froude Number = Volume of displacement
Fnv, Fn = Volumetric Froude number
L= Length

Acknowledgements

This paper describes the work carried out as part of a sub-program of AMECRC. The
authors are greatly indebted to AMECRC for providing the facilities and resources to
publish the results. Last but not the least the authors would like to thank all their

13
colleagues for their valuable suggestions and help during the course of their work
without which this paper would not have materialised.

References

1. MacPherson, D.M. (1993): " Reliable Performance Prediction: Techniques Using

A Personal Computer", Marine Technology, Vol 30, No. 4, pp. 243-257,
October.

2. Fairlie - Clarke, A.C (1975): " Regression Analysis of Ship Data", International
Shipbuilding Progress, Vol 22 , No. 251, pp 227 - 250.

3. Fung, Siu . C . (Am) (1991): " Resistance and Powering for Transom Stern Hull
forms during Early stage Ship Design", Society of Naval Architects and Marine
Engineers, Transaction Vol 99 , pp 29 - 84.

4. Almeter, John .M (1993): " Resistance Prediction of Planing Hulls ; State of the
Art", Marine Technology, Vol 30 , No.4 , pp 297 - 307,October

5. Ping-zhong, J., Ba-ying, S. & Zhong-kai, T.(1980): "A Parametric Study On

High-Speed Round Bilge Displacement Hulls", High Speed Surface Craft, pp 6 -
13, September.

6. Keuning, J.A., Gerritsma, J. & Van Terwisga, P.F(1993): " Resistance Tests of a
Series Planing Hull Forms With 30 Degree Deadrise Angle, and a Calculation
Model Based on this and Similar Systematic Series", International Shipbuilding
Progress.,Vol 40, No.424 , pp.333-385

7. Clement, E.P. & Blount D.L.(1963): "Resistance Tests of a Systematic Series of

Planing Hull Forms", Society of Naval Arcthitectures and Marine Engineers, pp
491-561

8. Du Cane, P.(1974): " High Speed Small Craft", David & Charles (Holdings)
Limited, Abbot, England

9. Levi, S.(1990): "Dhows to Deltas", 2nd Ed., A&A, Milano, Italy.

14
 10. Almeter, M.(1989):"Resistance Prediction and Optimization of Low Deadrise,
Hard Chine, Stepless Planing Hulls",Society of Naval Arcthitectures and Marine
Engineers, Spring Meeting/STAR Symposium, pp S1-4-1 to S1-4-23, April.

11. Radojcic, D.(1984):"A Statistical Method for Calculation of Resistance of the

Stepless Planing Hulls",International Shipping Progress.,Vol 31, No.363,
pp.272-276, November

12. Radojcic, D.(1985):"An Approximate Method for Calculation of Resistance and

Trim of Planing Hulls",Ship Science Report, No 23, University of
Southhampton, Dept of Ship Science, pp 1-28, December.

13. Radojcic, D.(1990):"An Engineering Approach to Predicting the Hydrodynamic

Performane of Planing Craft using Computer Techniques",RINA Transaction, pp
251-267.

14. Lahtiharju, E., Karpinen, T., Hellevaara, M and Aitta, T.,(1991) "Resistance and
Seakeeping Characteristics of Fast Transom Stern Hulls with Systematically
Varied Form", SNAME Transaction, Vol 99, pp 85-111

15

View publication stats