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

net/publication/325176512

Comparative study of various strip-theory seakeeping codes in predicting

heave and pitch motions of fast displacement ships in head seas

Chapter · April 2018

DOI: 10.1201/9780429505294-67

CITATION READS

1 830

3 authors:

Filipe Belga Serge Sutulo

INSTITUTO DE ENGENHARIA MECÂNICA E GESTÃO INDUSTRIAL University of Lisbon
6 PUBLICATIONS   40 CITATIONS    125 PUBLICATIONS   859 CITATIONS   

SEE PROFILE SEE PROFILE

Carlos Guedes Soares

University of Lisbon
2,337 PUBLICATIONS   35,318 CITATIONS   

SEE PROFILE

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

MARSTRUCT – Network of Excellence on Marine Structures" (http://www.mar.ist.utl.pt/marstruct/) View project

Wecanet (Wave farm) View project

All content following this page was uploaded by Filipe Belga on 01 April 2020.

The user has requested enhancement of the downloaded file.

 Progress in Maritime Engineering and Technology – Guedes Soares & Santos (Eds.)
© 2018 Taylor & Francis Group, London, ISBN 978-1-138-58539-3

Comparative study of various strip-theory seakeeping codes

in predicting heave and pitch motions of fast displacement
ships in head seas

F. Belga, S. Sutulo & C. Guedes Soares

Centre for Marine Technology and Ocean Engineering (CENTEC), Instituto Superior Técnico,
Universidade de Lisboa, Lisbon, Portugal

ABSTRACT:  The study focuses on a comparative investigation of three seakeeping codes based on the
ordinary strip method: the open-source code PDstrip, the commercial code MaxSurf and an in-house code
earlier developed at CENTEC. All programs were applied to two fast displacement ship forms upon head
seas: a mono-hull and a catamaran, for which experimental data from model testing were available. In the
latter case, an assumption about negligible hull-to-hull interaction, acceptable in head and following seas,
was exploited. The peculiarity of these vessels is reaching Froude numbers up to 1.14, which increases
possible uncertainties caused by the strip method assumptions. All codes are linear and formulated in the
frequency domain. Heave and pitch RAOs have been compared, computed whenever possible both with
and without transom terms activated. The observed differences between the three codes and from the
experimental data motivated an analysis of the computed added masses and damping coefficients.

1  INTRODUCTION The first strip model was devised by Korvin-

Kroukovsky and Jacobs in 1957, see (Korvin-
Despite a definite shift to 3D seakeeping codes Kroukovsky 1961), (Lugovsky 1999), as an alter-
observed nowadays, see e.g., (Söding & Volker native to the thin ship theory earlier developed, or
2009), strip-theory based codes remain rather around the same time, by Haskind, Peters & Stoker
valuable as they allow reasonably accurate and rel- and Newman. Although initially viewed as promis-
atively fast predictions of ship motions in most sit- ing and as a natural extension of the thin ship wave
uations, which becomes particularly useful within resistance theory, it suffered from heavy overpredic-
optimization contexts where multiple calculations tion of the wave damping. The early strip model
must be repeated or in any other cases when multi- exploited, still on an intuitive level, slenderness of
ple computations are inevitable, see e.g., (Davis & the ship hull, which made possible neglecting the
Holloway 2003). longitudinal induced velocities and treating the ship
Another cause for continuing popularity of hull as a set of transverse sections, each subjected
strip methods is their perfect compatibility with to a simpler 2D hydrodynamic problem. While this
the problem of predicting ship global hull strength model provided satisfactory results for zero or small
in a seaway on the level of beam theory: a relatively speed of advance, typical for most merchant dis-
simple modification and extension of the code is placement ships, it was criticised for certain theo-
required for computing the shear forces and bend- retical deficiencies. Remarkably, 5  independent (!)
ing moments for every transverse section of the groups of ship hydrodynamicists from different
hull, which can be viewed as rigid or flexible. In the countries developed, almost simultaneously (years
later case, the study of sea-induced longitudinal 1968–70), an advanced variant of the strip method
vibrations become possible. An example of such (Beck et  al. 1989), (Lugovsky 1999). The most
extended algorithm can be found in (Fonseca & known, cited and comprehensive related publica-
Guedes Soares 1998). tion (Salvesen, Tuck, & Faltinsen 1970) describes
A wide number of strip-theory variants applied the most complete version of the method, developed
to the seakeeping of ships has been developed and for a ship with five degrees of freedom (5DOF), as
proposed since the 50s. A rather comprehensive the surge motion was ignored due to incompat-
review of such variations can be found in the books ibility with the strip model. It comprised the case
(Beck et al. 1989), (Newman 1977), (Faltinsen 2005) of relatively fast ships with the transom becoming
and (Bertram 2012) and only the most important dry at certain speeds. The initials of the authors
and relevant conclusions will be exposed below. of this article inspired the popular designation of

599
this model as STF. Although at zero ship speed this practice certain differences may be observed,
model becomes identical to that of Korvin-Krouko- see e.g., (Sutulo & Guedes Soares 2004).
vsky, for many low-speed cases (Froude number Fn 5. Some uncertainties can be related to peculi-
≤ 0.2), the older model actually demonstrated bet- arities of interpolation of the sectional char-
ter agreement with experimental data. However, its acteristics between different frequencies and to
performance becomes definitely inferior for higher careless treatment of very small encounter fre-
values of the Froude number when compared to quencies: the STF model has a singularity at the
the STF. As the STF method was also not free of zero frequency and thus, a special asymptotic
certain imperfections, more sophisticated variants form must be developed for small frequencies.
of the slender body theory were later developed 6. Uncertainties related to the “transom effects”.
(Newman 1977), (Bertram 2012). Yet, not only these For relatively high-speed vessels the transom
methods are substantially more complicated for terms accounting for the dry transom should
practical use but they failed to demonstrate definite be accounted for, as they are responsible for
superiority in terms of agreement with experiments the socalled hull lift damping (Faltinsen 2005).
as well. The latest progress focus mainly on the However, it was discovered that, in some cases,
development of 3D models with increased practical the transom terms worsen the agreement with
applications, driven by the growth of the available the experimental data (Beck et al. 1989) and, as
computing power. Meanwhile, when a less accurate result, this option has been suppressed in some
but faster method is required, the STF method is of the codes.
viewed as the most preferable and a winner in the
natural selection process. As the STF method keeps its practical signifi-
In general, it is recognized that the STF model cance, investigations focusing on uncertainties
is not suitable for high-speed crafts with Fn ≥ caused by the use of different implementations are
0.6–0.7. In this case, a more precise, though more of considerable value. An example of such stud-
complicated 2D+t theory is recommended (Faltin- ies is described in the present contribution. Three
sen 2005). However, (Sokolov & Sutulo 2005) easily available codes were chosen for comparative
indicates that at least at Fn ≤ 0.8, the plain STF computations: (1) PDstrip which is a public-domain
method can provide quite reasonable predictions code (Söding & Bertram 2009); (2) an in-house code
of longitudinal motions. developed by Fonseca (Fonseca & Guedes Soares
The STF model is relatively simple and easy to 1998); and (3) the commercially available program
implement and thus, the large number of in-house MaxSurf Motions (Bentley Systems, Inc. 2013).
computer codes written in various programming The code PDstrip is a Fortran 90 remake of the
languages is hard to estimate. In theory, such codes earlier Fortran 77 code STRIP. The main algorithm
must provide identical results. Yet, this does not is the same in both cases, corresponding to the STF
always happen in practice, see e.g., (Guedes Soares model and being characterised by careful treatment
1990), (Guedes Soares 1991). The following reasons of the low end frequencies and by a sophisticated
for the observed discrepancies can be imagined: interpolation method between frequencies, per-
formed separately for the modules and arguments
1. Human errors in coding combined with insuf- of the complex variables of interest. The sectional
ficient debugging and testing. Nowadays such complex added masses are estimated by the proce-
flaws are not likely, especially when it goes dure HMASSE and the sectional excitation forces
about commercial codes. by the procedure WERREG (the Haskind–New-
2. Compiler and run-time library errors. Must be man relations are not used). Both procedures are
extremely unlikely but the second author of the based on a variant of the boundary integral equa-
present article had twice (!), in 1975 and 1999, tion (BIE) method described in (Bertram 2012).
definitely faced erroneous functioning of the The code STRIP was using another BIE method
complex division operation with 2 different earlier developed by Yeung (1973), see also (Sutulo
Fortran compilers on two absolutely different & Guedes Soares 2004). Both are based on the the-
types of hardware. Ironically, in the both cases ory developed in (Jaswon 1953).
the problem occurred inside seakeeping codes The Fonseca linear code was written in For-
and was bypassed by coding definition of the tran 77 and is using the Frank boundary singu-
malfunctioning operation. larities method (Frank 1967). This method is not
3. Different length of the real variables and con- free from singular frequencies but in most cases
stants (typically 4 vs 8 bytes). However, hardly this phenomenon did not emerge. The transom
significant for STF algorithms. terms were suppressed in the used version of the
4. Various codes may use different methods for program. Since the source code was not available
estimating hydrodynamic characteristics of the to the authors, this deficiency could not be fixed.
hull sections. Again, all correct methods are The code MaxSurf Motions is claimed
supposed to produce identical results but in to be applicable for Fn ≤ 0.7 and it uses the

600
multiparametric conformal mapping method (up to following set of complex linear algebraic equations
15 parameters but 2-parameter Lewis forms are used for heave and pitch complex amplitudes ζ * and θ*:
by default) to evaluate the characteristics of the 2D
sections, see e.g., (Ramos & Guedes Soares 1997). [− ω e2 ( m + A33 ) + C33 + iω e B33 ]ζ *
The heave and pitch motions are modelled with the + [ω e2 A35 + C35 + iω e B35 ]θ * = Fζ*,
strip method with somewhat simplified formulae (1)
[− ω e2 A53 + C53 + iω e B53 ]ζ *
for excitation forces. The program has a catamaran + [ω e2 ( I yy + A55 ) + C55 + iω e B55 ]θ * = Fθ*,
option but hull-to-hull interactions are neglected,
i.e., the catamaran is assumed to be hydrodynami-
where ωe  =  ω  +  kU is the encounter frequency
cally equivalent to its single hull. The code allows
in head seas, ω is the absolute wave frequency,
empiric tuning of the model by introducing addi-
k = ω2/g is the wave number, g is the acceleration
tional user-defined damping in heave and pitch,
of gravity, m is the ship mass and Iyy its moment
which is uniformly distributed between sections.
of inertia in pitch; Aij, i, j = 3, 5 are the hydrody-
The latter option reflects the influence of a
namic inertial coefficients and Bij are the damping
rather popular view that ignoring viscous damp-
coefficients; Cij the restoring coefficients; Fζ* and
ing in heave and pitch may result in overestimated
Fθ* are the heave force and pitch moment complex
responses. However, it has been long ago established
amplitudes, respectively (see Table 1).
with indisputable certainty that possible relative
Here, all the integrals are along the length of the
contribution of viscosity is negligible in the entire
ship L. a33 and b33 represent the sectional added
frequency range of interest and, of course, no such
mass and damping coefficients in ζ-direction. A330
attempts were carried out in the present study.
and B330 refer to the speed independent compo-
An attempt to compare the three mentioned codes
nents of A33 and B33. ξtr is the ξ-coordinate of the
from the viewpoint of their applicability to the design
aftermost cross section of the ship and a33 tr
and
optimization problem is undertaken in the present
b33tr are the added mass and damping coefficients
article. For this purpose, a number of comparative
evaluated at that section. ρ is the fluid density,
computations of ship responses in head regular waves
Awp is the static waterplane area, Myy is the first
have been carried out. Two ship forms were selected:
the fast monohull “Model 5” (Blok & Beukelman
1984) and the fast displacement catamaran described Table 1.  STF model: hydrodynamic coefficients.
in (Guedes Soares et al. 1999), for which experimental
data from model testing were available. In the second Main part Transom terms
case, any hydrodynamic interaction between the hulls
A33 U tr
was neglected, which is in accordance with the com- ∫L
a33d ξ − b
ω e2 33
mon practice and recommendations (Bertram 2012).
Again, this presumes that the catamaran is hydrody- A35 U 0 U U 2 tr
− ∫ ξ a33d ξ − B + 2 ξtr b33tr − 2 a33
namically equivalent to its single hull. The conclu- L ω e2 33 ωe ωe
sions drawn in this study helped in the selection of A53 U 0 U
the most suitable program for the seakeeping optimi- − ∫ ξ a33d ξ + B + ξ btr
L ω e2 33 ω e2 tr 33
zation of a fast displacement catamaran to operate as
A55 U2 0 U U2
crew supplier for an offshore platform at the Alentejo − 2 ξtr2 b33tr + 2 ξtr a33
tr

basin, in Portugal (Belga et al. 2018). ∫ L

ξ 2a33d ξ + A
ω e2 33 ωe ωe
B33 +Ua33
tr
∫L 33
b dξ
B35 − ∫ ξ b33d ξ + U A330 U 2 tr
2  BRIEF OVERVIEW OF THE LINEAR −U ξtr a33
tr
− b
SEAKEEPING MATHEMATICAL
L
ω e2 33
MODEL B53 − ∫ ξ b33d ξ − U A330 −U ξtr a33
tr

L
B55 U2 0 U2
The seakeeping inertial Cartesian frames, advanc- ∫L
ξ 2 b33dξ + B
ω e2 33
+U ξtr2a33
tr
+ ξ btr
ω e2 tr 33
ing with the ship speed U but not involved in the
C33 rgAwp –
ship motions excited by waves can be somewhat
C35 -rgMyy –
different. In the present study the primary frame
C53 -rgMyy –
Oξηζ used in (Beck et al. 1989) was considered, i.e.
C55 gmGML –
the frame with the ζ-axis direct vertically upward,
the η-axis to the port side and the ξ-axis pointing Fζ* ρaw ∫
L
( f3D + f3K ) dξ + ρaw
U Dtr
f
forward. The origin O is located at the waterplane iω e
area but on the same vertical as the ship centre of Fθ* − ρaw ∫ ξ ( f3D + f3K ) − ρaw
U
ξ fD
tr

mass (gravity), CG.

L
iω e tr 3
U D
The linear mathematical model for ship motions + f dξ
iω e 3 
in regular head seas can be described by the

601
area moment of the static waterplane and GML Table 2.  Seakeeping codes: general data.
the longitudinal metacentric height. f3K and f3D
represent, respectively, the sectional Froude– Code, Designa­-
Krylov and diffraction forces in the vertical direc- Version tion Main author Reference
tion (ζ-axis) for unit amplitude incident waves. NA Fonseca
Nuno Fonseca (Fonseca &
Regarding the ormer component, associated with (CENTEC) Guedes
the field of incident waves, the classical theory of Soares 1998)
linear gravity-waves defines, within deep water PDSTRIP, PDstrip Heinrich (Söding &
assumptions, the potential of a progressive inci- rev. 33 Söding Bertram
dent wave with an arbitrary direction, which 2009)
makes f3K easy to compute by evaluating it at Maxsurf MaxSurf Bentley (Bentley
the mean wetted cross section. Alternatively, it Motions, Systems Systems,
can be computed at each time step, an approach V20 Inc. 2013)
commonly followed by time domain formulations.
As for f3D , which measures the perturbation of
the field of incident waves due to the presence of Table 3.  Seakeeping codes: main characteristics.
the ship, instead of determining directly the 2-D
Features Fonseca PDstrip MaxSurf
diffraction potential, Green theorem is usually
applied in order to solve it as function of the 2-D Transom terms – + +
radiation potential (Salvesen et al. 1970). Finally, Maximum number 40 100 200
f3Dtr is the diffraction force evaluated at the after- of strips
most section. Minimum recommended 21 30–40 15–30
The set (1) is solved straightforward and the number of strips
solution is typically represented in the form of Offsets/parameters1 per 20 100 15
frequency transfer functions (FTF): strip, max.
Offsets/parameters per 8–10 10 3
ζ* θ* strip, min.
Φζ (ω ) = ; Φθ (ω ) = , (2) Same number of offsets/ + – +
aw ikaw parameters for all strips
Equidistant offsets + + –
where aw is the wave amplitude. Maximum number 30 200 500
The FTF are complex and each of them can be wavelengths
represented in the form: User defined wavelength + + –
range
Φζ ,θ (ω ) =| Φζ ,θ (ω ) | e − i εζ ,θ (ω ) , (3) Computational speed 100% 95% 30%
relative to Fonseca
using its max. settings
where |Φζ,θ(ω)| is the amplitude FTF, traditionally
called the Response Amplitude Operator (RAO) and In MaxSurf the sections are described by conformal
1

mapping parameters rather than offset points.

εζ ,θ (ω ) = − arg Φζ ,θ (ω ) ∈ [ 0, 2π ], (4)
Table 4.  Reference systems used by the codes.
is the response phase lag which is always
positive for any physically meaningful causal sys- Code Origin (ξ, ζ) ξ to: η to: ζ to:
tem, although serious incorrect definitions of this
Hull Fonseca fwdPP, WL aft port up
parameter are rather common in the literature on
Geometry PDstrip L/2, baseline fwd port up
seakeeping and vibrations.
MaxSurf non applicable fwd stbd up
Inputs (e.g. Fonseca L/2, WL fwd port up
3  OVERVIEW OF THE CODES remote PDstrip L/2, baseline fwd port up
locations) MaxSurf L/2, baseline fwd stbd up
The main features of the three strip method sea- RAOs Fonseca CG, baseline fwd port up
keeping programs mentioned in Introduction are PDstrip L/2, baseline fwd stbd down
summarized in Tables 2–3. MaxSurf CG, baseline fwd stbd up
It must be noted that various codes are using
different frames. Of course, this must not affect the
results but, as it is often important to know exactly The Fonseca code is, in fact, a linear counter-
which frame is used, the summary of the used part of a more complex time-domain semi-linear
frames is given in Table 4. program described in (Fonseca & Guedes Soares

602
1998). Despite the substantial number of research coefficients shown bellow. Yet, these contributions
papers successfully validating the results of the are, according to Söding, much smaller than errors
semi-linear code, mainly for moderately slow ships, of the strip-method.
e.g., (Fonseca & Guedes Soares 2002), (Fonseca &
Guedes Soares 2004a), (Fonseca & Guedes Soares δ A55 = a11ζ 02 , (5)
2004b), (Fonseca & Guedes Soares 2005), the lin-
ear Fonseca code tested here lacks documentation where δA55 denotes the additional contribution to
and validation, as it was not planned to be used A55, ζ0 refers to the waterline level and a11 yields the
on its own in view of the availability of more generalized longitudinal force acting on the ship
advanced non-linear codes at the time. computed using the empirical formula (6), as func-
The commercial code MaxSurf was presum- tion of the ship length L and the displacement m.
ably validated on a wide range of vessel types and
some of these results can be found in the appen-
a11 = m / π ρ L3 / m − 14  (6)
dices of the manual (Bentley Systems, Inc. 2013)
or in independent publications such as (Ghassemi dA
et al. 2015). As could be expected from a commer- δ Fθ*K ∫ ζ p dξ + ζ tr ptr Atr , (7)
L dξ
cial code, it possesses a rather developed graphic
interface facilitating its use by less qualified spe-
cialists. At the same time, this code runs, due to where δ Fθ*K is the additional Froude–Krylov con-
unclear reasons, substantially slower than its coun- tribution to the pitch moment Fθ*, p refers to the
terparts. Although PDstrip is a linear frequency pressure calculated at the centre of each cross sec-
domain program, it can account for some non- tion (ξ, 0, ζ), in head seas, A is the correspond-
linear effects described in (Söding & Bertram 2009). ing immersed area and the index tr denotes values
However, these options were not activated in the evaluated at the transom stern. Therefore, the term
present study. As stated by Palladino et al. (2006), ζtr ptr Atr vanishes if the transom corrections are to
the validation of PDstrip continues to be a work be neglected from the calculations. At the centre of
in progress. Although it has been used by many a cross section, the pressure is given by:
researchers for a wide variety of purposes includ-
ing manoeuvring simulations (Schoop-Zipfel & p = − ρ ge − k (ζ + T)aw e ikξ , (8)
Abdel-Maksoud 2011), analysis of wave-induced
dynamic effects on sailing yachts (Bordogna 2013), where T is the mean draft of the ship cross section.
computation of drift forces on offshore wind farm Regarding the restoring coefficients, they are
installation vessels (Augener & Krüger 2014) or effectively computed has stated in Table  1, con-
even integrated into a global design tool (Salio sidering the mean position of the wetted surface.
et al. 2013), a thorough work of comparison with However, as a result of programming decisions
both experimental results and other seakeeping from the developers, PDstrip computes the first
codes is still to be done (Palladino et  al. 2006), area moment of the static waterplane with respect
(Gourlay et al. 2015). to the midship section, resulting in different coef-
In terms of the method, PDstrip is the only code ficients for C35 and C53, which apparently does not
which the formulation presents noteworthy varia- affect the motion results. Furthermore, there is an
tions from the previously presented STF method. addition in C53 if transom terms are to be included
Recalling Table  4, note that the inertial reference in the calculations:
frame Oξηζ is, according to Siöding & Bertram
(2009), positioned at amidships, rather than at the δ C53 = ζ tr Atr , (9)
centre of gravity with respect to the motion results.
The method computes hydrodynamic coefficients where ζtr is the ζ-coordinate of the centroid of the
in a somewhat different way as compared to the area of wetted transom Atr.
formulae shown in Table  1. Namely, the terms The version of PDstrip that is being used here
related to the longitudinal derivative of the added has been revised by the authors (rev. 33). The most
masses and damping coefficients, which were han- significant modifications relate with the handling
dled directly, without their elimination through of transom terms. In PDstrip rev.32, the latest
integration by parts and application of the Stokes– publicly available version, these terms were always
Tuck transformation that results in the original kept for added masses, damping coefficients and
STF formulae. In addition, this code handles the exciting forces and moments, while the option to
surge motion using a simplified surge equation include or exclude them only affected equations
accounting for only Froude–Krylov and inertial (7) and (9). As this seemed to lack consistency, the
forces. This equation is, however, coupled with source-code was improved in order to enable that
the pitch motion through the modifications of the option for the coefficients of Table 1 as well.

603
4  RESULTS OF COMPARATIVE for the heave and pitch RAOs computed, when-
COMPUTATIONS ever possible, both with and without transom
terms activated. These are shown bellow, divided
The three codes briefly described above were by the wave amplitude (aw) for heave and by the
applied to two ship forms showed in Figure  1: wave slope (kaw) for pitch, as function of the wave
the fast monohull “Model 5” (Blok & Beukelman frequency ω and each plot corresponds to a differ-
1984) and a fast displacement river-going catama- ent Froude number. Results for “Model 5” can be
ran (Guedes Soares et al. 1999), for which experi- seen in Figure 2 for heave and Figure 3 for pitch,
mental data from model testing were available. while for the catamaran case one should refer to
In the latter case, an assumption about negligible Figures 6 and 7 for heave and pitch, respectively.
interaction between hulls, acceptable in head and Figures  4 and 5  summarise the absolute dif-
following seas, was exploited, which presumes that ference (in the same units as the RAOs) between
the catamaran is hydrodynamically equivalent to
its single hull. The main particulars of both mod-
els are shown in Table 5.
In order to obtain comparable results between
the different codes, both hull forms were discre-
tized using 40  strips, imposed by the maximum
settings allowed by Fonseca which has the lowest
capacity. For the same reason, 20 equally spaced
offsets per strip were used, as none of the cross sec-
tions suffered from abrupt changes in curvature, Figure  2.  Heave RAOs as function of the wave fre-
which would require a higher density of points in quency, “Model 5”.
those regions. Regarding wavelengths it was used
the maximum number allowed by each code, in
order to obtain the best definition of the frequency
range possible.
In the case of the catamaran, four Froude num-
bers ranging between 0 and 0.6 were analysed,
while for “Model 5” only the Froude number val-
ues 0.57 and 1.14 have been considered. The pecu-
liarity of the studied vessels is the high Froude Figure  3.  Pitch RAOs as function of the wave fre-
numbers at which they were tested, which increases quency, “Model 5”.
possible uncertainties caused by the strip method
assumptions. Comparisons have been performed

Figure  4.  Root mean squared absolute differences

between numerical and experimental results, “Model 5”.

Figure 1.  Underwater body lines of the full-scale vessels.

Table 5.  Main dimensions of the full-scale vessels.

LWL BWL T
[m] [m] [m]

“Model 5” (Blok & 50 5.83 1.57

Beukelman 1984)
Catamaran (Guedes Soares 43 2.7 1.35
et al. 1999) Figure  5.  Root mean squared absolute differences
between numerical and experimental results, Catamaran.

604
Figure 6.  Heave RAOs as function of the wave frequency, Catamaran.

Figure 7.  Pitch RAOs as function of the wave frequency, Catamaran.

the numerical results obtained with each code and the prediction of MaxSurf improve with the inclu-
the experimental data, averaged over all measured sion of transom terms while with PDstrip this
frequencies. In order to average these results, the decreases accuracy. For Fn  =  0.57, the inclusion
square root of the average squared differences of transom terms generally makes pitch predic-
was used. Since the errors are squared before they tions worse. On the contrary, at the highest speeds
are averaged, a relatively higher weight is given to (Fn = 0.6, 1.14), it is advantageous both for heave
larger errors. This is beneficial for this study where and pitch motions, to include such terms in the
big deviations from the experimental results are equations.
particularly undesirable. In a general sense, at higher Froude numbers,
It can be observed that with increasing vessel the inclusion of transom terms is beneficial, as
speed, the accuracy in predicting heave and pitch they add damping to the system. It is interesting
motions with strip-theory decreases, a well-known to note that the semi-linear version of Fonseca
limitation of such theories and a general principle (Fonseca & Guedes Soares 1998) has been mainly
that must be taken into account. Figures 4 and 5 used to predict motions and wave loads of slower
illustrate this effect particularly well. large displacement vessels (e.g., containerships)
Given the nature of the coefficients of the and, in fact, neglects transom terms. The conclu-
coupled heave and pitch equations as shown sions of (Marón et al. 2004) and (Fonseca & Guedes
in Table  1, transom terms have no effect at zero Soares 2004c), where that code was applied to a fast
speed. Such terms are only meaningful as the ship mono-hull, showed precisely that at higher Froude
starts moving. However, since PDstrip considers numbers its accuracy decreases significantly. This
speed-independent transom corrections for the seems to indicate that if the transom terms are
restoring coefficient C53 as shown in equation (9) neglected the range of applications should focus on
and for the Froude–Krylov force as in (7), there is slower vessels. The same applies to Fonseca (linear
at Fn = 0 a small visible difference between its pre- version), which also excludes such terms.
dictions both in Heave, Figure 6(a), and in Pitch, Table 6 summarises the drawn conclusions, sup-
Figure 7(a). This seems to be adverse for pitching ported by Figures  4 and 5. A few comments on
motion. With MaxSurf, this difference does not the capacity of the codes to accurately predict the
exist. In any case, at zero speed, Fonseca seems resonance peaks and the respective frequencies
to make the most accurate predictions. PDstrip, have been added as well. In this regard, note that
without transom terms in particular, also performs for the case of the catamaran, it is possible that
reasonably well. they might not have been accurately captured by
Following Figures 4 and 5, for Fn = 0.2, heave the experiments because of the frequency separa-
predictions seem to have been adversely affected by tion between experimental points (Guedes Soares
the inclusion of transom terms in the equations. et al. 1999).
At Fn = 0.4, the opposite happened. At Fn = 0.57, It seems that at lower speeds, using a strip-
the inclusion of transom terms improve the heave theory code without transom terms is the best
predictions of PDstrip but that is not the case with approach for optimal results. Table 6 suggests that
MaxSurf. Regarding pitch RAOs, at Fn = 0.2, 0.4, Fonseca is, overall, the most suited for Fn  =  0,

605
Table 6.  Summary of code preferences.

Fn Heave Pitch

0 Fonseca1 Fonseca
0.2 Fonseca PDstrip (w/o t. terms)
0.4 Fonseca2 Fonseca3
0.57 MaxSurf (w/o t. terms) PDstrip (w/o t. terms)
0.6 Fonseca4 MaxSurf (w/t. terms)
1.14 MaxSurf (w/t. terms)5 PDstrip (w/t. terms)
1
Differences between the codes are negligible
2
Resonance peak more accurately predicted with PDstrip
(w/t. terms)
3
Resonance frequency more accurately predicted with
MaxSurf
4
Similar results with PDstrip (w/t. terms). Resonance fre-
quency overestimated with both methods.
5
Similar results with PDstrip (w/t. terms). Amplitudes at Figure 8.  Global added masses for Fn = 1.14, “Model 5”.
the middle frequency range poorly estimated with both
methods.

0.2, 0.4. For  Fn  =  0.57, it appears that although

not using transom terms proves to be more effec-
tive, MaxSurf and PDstrip work the best. For the
highest speeds, i.e., Fn = 0.6, 1.14, it becomes clear
that the results improve by including transom cor-
rections in the equations. Overall, PDstrip seems
to be the most consistent and reasonably accu-
rate, even though MaxSurf produces somewhat
similar results. However, from the perspective of
embedding the code into optimization procedures,
computational speed plays a crucial role and thus,
PDstrip becomes the most natural choice (refer to
Table  3). In addition, PDstrip allows: a substan-
tial number of ship sections and offset points for
geometry discretization, which improves accuracy Figure  9.  Global damping coefficients for Fn  =  1.14,
during seakeeping computations; a large number “Model 5”.
of wavelengths to be used for motion results, being
advantageous for a proper definition of the whole
frequency range. Also, the method used to handle to solve the 2D problem (Frank 1967). As hypoth-
the 2D problem at each cross sections is allegedly esized in Introduction, there is a number of pos-
superior (Söding & Bertram 2009), when com- sible reasons for the observed differences. In order
pared to conformal mapping techniques. to fully understand this problem, an in-depth study
The differences between the numerical predic- of the specifics of the methods applied by each
tions and the experimental data motivated an code is required, which stands as suggestion for
analysis of the global added masses and damp- future work.
ing coefficients. As an example, these are shown
below for Fn  =  1.14 (“Model 5” case). Note that
care was taken into displaying the same frequency 5  CONCLUSIONS
range that was captured by the experiments, as in
Figures 2 and 3. A comparative study of three strip-theory based
It has been verified that the discrepancies codes is presented here, comprising the open-source
between the codes increase with speed, as expected. code PDstrip, the commercial software MaxSurf
The best results were obtained for the coefficients Motions and an in-house code earlier developed at
A33 and B33, while for the remainders the results CENTEC referred to as Fonseca. Experimental data
tend to differ more. Also, at higher speeds, the from model testing of two ship hull forms was used
existence of irregular frequencies when using Fon- to verify the numerical predictions of the three codes
seca becomes clear, as a result of the method used in terms of heave and pitch RAOs upon head seas.

606
 With increasing vessel speed, the accuracy in Bertram, V. (2012). Practical Ship Hydrodynamics.
predicting heave and pitch motions with strip- Oxford, UK: Butterworth–Heinemann.
theory decreases, a well-known limitation of such Blok, J.J. & W. Beukelman (1984). The high-speed
theories and a general principle that must be taken displacement ship systematic series hull forms –
seakeeping characteristics. Transactions of the Society
into account. At lower speeds, not using transom
of Naval Architects and Marine Engineers, SNAME
terms is preferable, Fonseca in particular, as the 92, 125–150.
version of the code used here neglects them. In any Bordogna, G. (2013). The aero-hydrodynamic character-
case, both PDstrip and MaxSurf (without transom istics of yachts sailing upwind in waves. Master’s the-
terms) also performed reasonably well in specific sis, TU Delft, Delft University of Technology.
situations. At higher speeds (Fn = 0.6, 1.14), heave Davis, M.R. & D.S. Holloway (2003). The influence of
and pitch response peaks were generally overes- hull form on the motions of high speed vessels in head
timated but the inclusion of transom terms adds seas. Ocean Engineering 30, 2091–2115.
damping to the system, being advantageous for Faltinsen, O.M. (2005). Hydrodynamics of High-Speed
predicting heave and pitch motions. PDstrip and Marine Vehicles. Cambridge, UK: Cambridge Univer-
sity Press.
MaxSurf with transom terms seem the most suited
Fonseca, N. & C. Guedes Soares (1998). Time-domain
codes in these cases. analysis of large-amplitude vertical ship motions and
Considering the high computational speed of wave loads. Journal of Ship Research 42(2), 39–153.
PDstrip, as well as other useful features, it seems Fonseca, N. & C. Guedes Soares (2002). Comparison of
natural to select it as seakeeping tool for design numerical and experimental results of nonlinear wave-
procedures intended to optimize fast displacement induced vertical ship motions and loads. Journal of
vessels in head seas. Also, as an open-source For- Marine Science and Technology 6(4), 193–204.
tran code, it is possible to edit and further improve Fonseca, N. & C. Guedes Soares (2004a). Experimental
PDstrip if necessary, a useful characteristic bear- investigation of the nonlinear effects on the statis-
ing future work in mind. For an example of practi- tics of vertical motions and loads of a containership
in irregular waves. Journal of Ship Research 48(2),
cal applications of the drawn conclusions refer to 148–167.
(Belga et al. 2018), where the hull form of a crew Fonseca, N. & C. Guedes Soares (2004b). Experimental
supply catamaran was optimized with respect to investigation of the nonlinear effects on the verti-
seakeeping. cal motions and loads of a containership in regular
waves. Journal of Ship Research 48(2), 118–147.
Fonseca, N. & C. Guedes Soares (2004c). Validation of
ACKNOWLEDGEMENTS a time-domain strip method to calculate the motions
and loads on a fast monohull. Applied Ocean Research
The study was performed within the project PTDC/ 26(6), 256–273.
Fonseca, N. & C. Guedes Soares (2005). Comparison
EMSTRA/5628/2014 “Manoeuvring and moored
between experimental and numerical results of the
ships in ports–physical and numerical modelling”, nonlinear vertical ship motions and loads on a con-
funded by the Portuguese Foundation for Science tainership in regular waves. International Shipbuilding
and Technology (FCT). Progress 52(1), 57–89.
Frank, W. (1967). Oscillation of cylinders in or below
the free surface of deep fluids. Technical Report 2375,
REFERENCES Naval Ship Research and Development Centre, Wash-
ington DC, USA.
Augener, P.H. & S. Krüger (2014). Computation of drift Ghassemi, H., S. Majdfar, & V. Gill (2015). Calcula-
forces for dynamic positioning within the very early tions of the heave and pitch RAO’s for three different
design stage of offshore wind farm installation vessels. ship’s hull forms. Journal of Ocean, Mechanical and
In Proceedings of the ASME 2014, 33rd International Aerospace – Science and Engineering 22, 1–8.
Conference on Ocean, Offshore and Arctic Engineering, Gourlay, T., A. von Graefe, V. Shigunov, & E. Lataire
San Francisco, California, USA. American Society of (2015). Comparison of aqwa, gl rankine, moses, octo-
Mechanical Engineers. pus, pdstrip and wamit with model test results for
Beck, R.F., W.E. Cummins, J.F. Dalzell, & P. Mandel cargo ship wave-induced motions in shallow water.
(1989). Motions in waves. In E.V. Lewis (Ed.), Princi- In Proceedings of the ASME 2015, 34th International
ples of Naval Architecture, Volume 3, pp. 1–190. Jersey Conference on Ocean, Offshore and Arctic Engineering,
City, NJ: SNAME. St. John’s, Newfoundland, Canada. American Society
Belga, F., M. Ventura, & C. Guedes Soares (2018). Sea- of Mechanical Engineers.
keeping optimization of a catamaran to operate as Guedes Soares, C. (1990). Comparison of measurements
fast crew supplier at the Alentejo basin. In C. Guedes and calculations of wave induced vertical bending
Soares and T.A. Santos (Eds.), Progress in Maritime moments in ship models. International Shipbuilding
Technology and Engineering, London, UK: Taylor & Progress 37(412), 353–374.
Francis. Guedes Soares, C. (1991). Effect of transfer function
Bentley Systems, Inc. (2013). Maxsurf Motions, Windows uncertainty on short term ship responses. Ocean Engi-
Version 20, User Manuals. neering 18(4), 329–362.

607
Guedes Soares, C., N. Fonseca, P. Santos, & A. Maron Salvesen, N., E.O. Tuck, & O. Faltinsen (1970). Ship
(1999). Model tests of the motions of a catamaran motions and sea loads. In Transactions of the Society
hull in waves. In Proceedings of the International of Naval Architects and Marine Engineers, SNAME,
Conference on Hydrodynamics of High-Speed Craft, Volume 78, pp. 250–287.
London, UK, pp. 1–10. Royal Institution of Naval Schoop-Zipfel, J. & M. Abdel-Maksoud (2011).
Architects (RINA). A numerical model to determine ship manoeuvring
Jaswon, M. (1953). Integral equation method in potential motion in regular waves. In L. Eça, E. Oñate, J.
theory. Proceedings of Royal Society A275/360, 23–32. García, T. Kvamsdal, and P. Bergan (Eds.), 4th Inter-
Korvin-Kroukovsky, B. (1961). Theory of Seakeeping. national Conference on Computational Methods in
New York, NY: SNAME. Marine Engineering, MARINE 2011, Lisbon, Portu-
Lugovsky, V.V. (1999). Ship Motions (in Russian). St. gal. Springer.
Petersburg: St. Petersburg State Marine Technical Söding, H. & V. Bertram (2009). Program PDSTRIP:
University Publishing Centre. public domain strip method. https://sourceforge.net/
Marón, A., J. Ponce, N. Fonseca, & C. Guedes Soares projects/pdstrip/.
(2004). Experimental investigation of a fast mono- Söding, H. & B. Volker (2009). A 3-d rankine source
hull in forced harmonic motions. Applied Ocean seakeeping method. Ship Technology Research 56(2),
Research 26(6), 241–255. 50–58.
Newman, J.N. (1977). Marine Hydrodynamics. Cam- Sokolov, V. & S. Sutulo (2005, 27–29 June). A practical
bridge, MA: MIT Press. approach to a fast displacement ship’s stabilization in
Palladino, F., B. Bouscasse, C. Lugni, & V. Bertram head seas. In Proceedings of International Conference
(2006, October). Validation of ship motion functions on Fast Sea Transportation FAST-2005, St. Peters-
of pdstrip for some standard test cases. In 9th Numeri- burg, Russia.
cal Towing Tank Symposium, Le Croisic, France. Sutulo, S. & C. Guedes Soares (2004). A boundary
Ramos, J. & C. Guedes Soares (1997). On the assessment integral equation method for computing inertial and
of hydrodynamic coefficients of cylinders in heaving. damping characteristics of arbitrary contours in deep
Ocean Engineering 24(8), 743–763. fluid. Ship Technology Research/Schiffstechnik 51,
Salio, M.P., F. Taddei, P. Gualeni, A. Guagnano, & F. 69–93.
Perra (2013). Ship performance and sea state condi- Yeung, R. (1973). A Singularity Method for Free-Surface
tion: an assessment methodology integrated in an Flow Problems with an Oscillatory Body. Ph. D. the-
early design stage tool. In Proceedings of tthe 10th sis, University of California, College of Engineering,
International Conference on Maritime Systems and Berkeley.
Technologies, MAST 2013, Gdansk, Poland.

608

View publication stats