You are on page 1of 22

Conceptual Design and Hydrodynamic Analysis of a

High-Speed Sealift Adjustable-Length Trimaran

K. J. Maki (AM), L. J. Doctors (M), R. M. Scher (M), W. M. Wilson (V), S. H. Rhee (AM),
A. W. Troesch (FL), R. F. Beck (FL)

This paper describes a novel adjustable-length trimaran design for service as a high-speed sealift vessel. Different
numerical hydrodynamic tools were exercised to assess the performance of the vessel. An extensive model test program
that included powering and seakeeping experiments was conducted to elucidate utility of the numerical predictions.

KEYWORDS: calmwater resistance; seakeeping; multihull; have not yet converged to a single view. However, several inde-
model testing; thin-ship theory; computational fluid dynamics; pendent visions of the future, paralleled by recent concept stud-
strip theory ies and designs, have included a doubling of transit speed, from
the conventional 20 to 24 knot regime of present sealift ships, to
INTRODUCTION speeds exceeding 40 knots; cargo payloads from several hundred
For ship designers the increasing emphasis on rapid and flex-
to several thousand tons, and stage lengths varying from about
ible support of logistic missions represents several distinct tech-
2500 to about 6000 nautical miles.
nical challenges. Among these, one of the most difficult is the
desire for significantly higher speeds than previous sealift ship Ship length constraints for austere port operations (quite
classes, capable of carrying substantial military payloads over apart from limited acquisition budgets and operating costs) de-
trans-oceanic stage lengths, but without prohibitive increases in mand solutions that differ markedly from the conventional, large,
ship size, power, and fuel consumption (especially in view of re- sealift ship (with whatever increased power, weight saving, and
cent trends in oil price). The need for operational flexibility in variation of form coefficients go with it). While designers strive to
logistics in unconventional warfare places a premium on the abil- generate concepts that will fulfill future logistic needs, including
ity to use less developed, “austere,” ports, with inherent limita- some hull-form types that are quite unprecedented, the state-of-
tions on ship size, including stringent constraints on length and the-art hydrodynamic analysis tools that they rely on are under-
draft. Finally, the developing doctrine of Sea Basing demands the developed and unvalidated in these new operating regimes. The
ability to transfer a variety of cargoes, especially rolling stock, development of new ship concepts, and new design tools, must
between ships in open water. If these challenges are to be suc- accordingly advance together.
cessfully addressed, innovative ship configurations are clearly re- Recently, US Office of Naval Research (ONR) supported
quired: accordingly, certain fundamental changes in design meth- concept studies and hydrodynamic tool development for High-
ods and philosophy may have to occur. What these changes will Speed Sealift (HSSL) to Austere Ports. The design ground rules
be, and specifically what impacts on design criteria and ship op- called for the following major capabilities:
erating methods will result, will undoubtedly be a field of contro-
versy for some time.
• Unrefueled range: 5000 nautical miles at average speed of
Evolving mission needs for logistics in expeditionary warfare
43 knots

1
• Full mobility, including 43 knot speed, through Sea State 4

• Military payload of 4000 short tons (representing approxi-


mately the weight and stowage area for vehicles of a Stryker
battalion task force)

• Austere port access: maximum over-all length 170 m; max-


imum draft 6.5 m

• At-sea transfer of vehicle cargo (nominally through Sea


State 4)
(a) Sidehull (b) Centerhull
This paper summarizes the design and hydrodynamic charac-
terization of a vessel to meet these mission requirements, and is Figure 1: Vessel Geometry
organized as follows. First, our most desirable design candidate
is described with details about the unique challenges that accom-
pany its unconventional form. Next, the experimental measure-
located forward of the two sidehulls. They demonstrated both ex-
ments and numerical predictions used during the design process
perimentally and theoretically (as we shall here) that the (total)
are introduced and discussed. Then, the calm-water resistance
resistance of a trimaran is lower over most of the speed range in
and seakeeping motion results are presented. Finally, the paper
this configuration.
concludes with a summary of the merit of the different numerical
tools in predicting vessel performance. Their work was both theoretical and experimental in nature;
it was based on earlier work by Tuck and Lazauskas (1998),
DESIGN CONCEPT whose theoretical and numerical investigations were very thor-
The Adjustable-Length Trimaran ough. We must also emphasize that trimarans are, by no means,
Clearly, the speed, range, and payload requirements of high- a new concept. For example, at one of the historic meetings
speed sealift present a naval architectural challenge of a high on wave resistance, Narita (1976) considered trimaran layouts
order, even without the addition of severely constrained length in which the two sidehulls were staggered both forward and aft.
and draft for austere port access. With these dimensional con- They demonstrated that both positive and negative staggers lead
straints, especially the length constraint, the problem has re- to similar reductions in wave resistance.
peatedly shown itself to be beyond the reach of conventional Returning now to our mission requirements, the over-all
ship designs with presently available material, machinery, and length of the trimaran with sidehull stagger unfortunately exceeds
fuel technologies. Credible concept-level designs, whether they the austere-port-length constraint. That is the dilemma.
have been of “conventional” (long centerhull) trimaran, catama-
ran, or monohull configuration, displacement, heavy-planing, or An unconventional solution can be proposed, however. Sup-
cushion-borne, do not converge (carrying payload and fuel for the pose that the hulls are structurally connected and locked in a stag-
desired range) except by exceeding the 170 m austere port length gered configuration during transit, even though the over-all length
constraint by a considerable margin, and in some cases breaking exceeds the austere port constraint. On arrival then, the vessel
the draft constraint as well. mechanically alters its configuration for port access by retracting
to a more compact length. On departure, the vessel extends and
However, suppose we envision a trimaran in which the locks into its transit configuration again. This is not a trivial de-
three hulls are of approximately equal length (the austere-port- sign problem, but on the other hand it does not appear to be out
constrained length) and of nearly equal displacements. As the of reach with present technologies.
results presented subsequently will demonstrate, a trimaran of
this type, with the hulls staggered (instead of directly alongside Body plans of sidehull and centerhull for our notional
one another), can achieve substantial wave resistance advantages adjustable-length trimaran are shown in Figure 1; principal char-
compared with other configurations. With an appropriate stagger acteristics of the hulls are listed in Table 1. Transom configura-
for a given transverse separation of the hulls, wave cancellation tions permit the installation of a single waterjet propulsor in each
can be achieved and the wave resistance of the trimaran can be hull, an arrangement considered most suitable for model testing.
less than that of a slender monohull of equivalent displacement However, propulsion-machinery selections (for an actual vessel)
and over-all length. might ultimately favor single or twin waterjets in one, two, or all
three subhulls.
At this juncture, we must acknowledge the paper by Day,
Clelland, and Nixon (2003). This paper covered an extensive Design Challenges of the Adjustable-Length
discussion and investigation into eight different configurations of Trimaran
monohulls and multihulls. In particular, the three authors studied Locking, Retraction and Extension The principal design chal-
a so-called Arrow configuration for a trimaran, which is similar lenge is apparent: mechanical systems required for structural con-
to the subject of the present work, in that the central subhull was nection of the retractable hull must be capable of supporting the

2
Table 1: Nominal Particulars of HSSL ALT Subhulls
Center- Side-
Item Symbol Units
hull hull
Displacement mass ∆ t 5827 5964
Waterline length L m 168.1 168.6
Waterline beam B m 10.03 10.04
Draft T m 6.510 6.510
Waterplane-area coefficient CW P 0.7770 0.8091
Maximum section coefficient CM 0.7860 0.7863
Block coefficient CB 0.5174 0.5279
Prismatic coefficient CP 0.6583 0.6714
Slenderness coefficient L/∇1/3 9.424 9.375

design sea loads, and still be able to retract and extend under some
other (presumably less severe) set of loads. This is not completely
without precedent. Locking and bearing mechanisms sized to
carry loads (of magnitudes appropriate for primary ship struc-
tures) have been used in commercial marine applications such
as integrated and articulated tug-barge (ATB) systems. Simi-
larly, mechanical systems for retraction and extension of large ob-
jects, under substantial forces, have been developed for other ma-
rine and offshore applications, such as jack-up drilling rigs, off-
shore structure launching appliances, and heavy-lift derricks and
cranes. However, the additional challenge for a high-speed ship
application is to provide the required system capabilities (lock-
ing, retraction, and extension) at an acceptable weight penalty.
Figure 2 shows a concept for an adjustable-length trimaran
with a retractable centerhull section. The centerhull is locked
in the extended position for transit, with locking mechanisms,
broadly similar to the devices that are used on integrated tug
barges: that is, sliding wedges or rams, sized to take the structural
loads required for open ocean operation at high speed. For access
to an austere port, the locks would be disengaged, and the center-
hull would be retracted along a system of guides which need only Figure 2: Adjustable-Length Trimaran Concept with Retractable
take the loads required in port, sheltered waters at relatively low Centerhull Section Extended and Retracted
speeds. Based on current practice and commercial systems used
in integrated tug barges, a structural locking system can be de-
veloped at a reasonable weight, provided that appropriate design In fact, such an arrangement would permit at least one geomet-
loads can be established. ric advantage in terms of freeboard constraint on the centerhull.
In seakeeping model tests that will be described below, forces However, in other important respects, especially structural conti-
and moments were measured at the connections between the port nuity and arrangements for cargo stowage and access, a concept
and starboard cross structures and the sides of the centerhull. The with the centerhull connected underneath the cross structure of a
highest irregular wave case tested was a representative Sea State 7 “catamaran section” seems more reasonable.)
spectrum (Bretschneider, 7.5 m significant wave height, 15 sec- A concept for locking the centerhull is shown schematically
ond modal period), in head seas, at 25 knots. Based on a pre- in Figure 3. Notionally, the locking system consists of opposed
liminary analysis of those results it is considered likely that ver- pairs of horizontal sliding wedges driven (for example) by a jack
tical bending moments and shear forces will govern the design screw. To lock the centerhull, the wedges are driven outboard and
of the locking system, at least in a configuration where the cen- engage corresponding “buckets” fixed in the cross structure. Two
terhull which retracts under the cross structure between the two assemblies, separated longitudinally, transmit the vertical bend-
side hulls. (Conceptually, there are other ways of performing this ing moment and shear, and also resist longitudinal forces and
retraction. For example, the port and starboard cross-structures torsion of the centerhull about the longitudinal axis. One pair
and hulls could slide fore and aft along the sides of the centerhull. of locking wedges is located near the aft end of the centerhull,

3
to move the centerhull could include rack and pinion (similar to
systems used for jack-up drill rigs), traction wheels, winches and
cables, ball screws, or other linear actuators. Some care is obvi-
ously needed with regard to buoyancy and trim in an adjustable-
length vessel. Just because the entire trimaran is in static equi-
librium with the sections locked in a certain position, it does not
follow that when the locks are released each section will still be in
equilibrium individually. Thus, the guide system may be carrying
both a force and a moment when the locks are released, and there
would be situations when the centerhull would tend to move along
the guide simply because of an excess of buoyancy or a trimming
moment in one section or the other.
Figure 3: Adjustable-Length Trimaran Concept; Section Showing
a Possible Location of Locking and Retraction Systems Weights for the locking and retraction systems were included
in estimates for the adjustable-length trimaran, and amount to
somewhat over 250 tonnes, including additional foundations.
This is not an insignificant weight. However, the penalty has to
and engages with buckets located near the mid-length of the cross be compared with the added propulsion machinery and fuel that
structure. The other pair is located near the mid-length of the would be associated attempting to meet speed and range require-
centerhull, and engages the cross-structure as far forward as prac- ments on an over-all length limited to 170 m, instead of 50 percent
ticable. Thus, the lever arm between the two sets of locks is ap- longer.
proximately half the length of the centerhull.
Arrangement In the arrangement implied by Figures 2 and 3,
In oblique seas and during maneuvers, of course, there are the centerhull contains fuel tankage and propulsion machinery
also significant lateral bending moments and side forces acting spaces, but for reasons that should be apparent no cargo or habit-
on the centerhull. These forces also have to be transmitted by the ability spaces. The cross-structure of the catamaran section con-
locking wedges, although primarily as loads along the axis of the tains all the cargo stowage areas, cargo access arrangements, ac-
actuating screw, rather than as shear in the wedges. Model test re- commodations, and ship control spaces, in addition to propulsion
sults in oblique seas have only been obtained in irregular waves up machinery rooms and most of the auxiliary machinery. There is
to 4 m significant height. Consequently, firm conclusions on the no superstructure on the centerhull section, of course, because
appropriate design loads cannot be drawn yet. However, if ver- of the need to retract it under the cross-structure. This limita-
tical bending and shear loads do indeed govern the design of the tion, in effect, restricts the freeboard of the centerhull to the wet
locking members, as is expected, it appears that an opposed pair deck height of the catamaran section. This in turn can have a
of locking wedges of the type shown in Figure 3 would have to considerable influence on performance and seakeeping, as dis-
be capable of transmitting an extreme load of about 8300 tonnes. cussed in subsequent sections. The solution to excessive wetness
This is hardly a small load, and it reflects to a great extent the of the centerhull forward, without excessive height of the catama-
influence of high speed. However, it is within the shear capacity ran section, remains a practical challenge for design trade-offs.
of practical forgings and weldments, not much different in size [In this connection it is noted that according to Michell’s theory,
from those currently used on large articulated tug-barges. In ad- staggering the centerhull aft offers the same magnitude of wave-
dition, while ATB coupling systems must incorporate bearings to resistance advantage as does forward stagger. Aft stagger results
permit rotation under load, the locking wedges of the trimaran in a rather strange looking planform (Figure 4) and possibly cer-
do not, and this simplifies the design of the system. The dimen- tain additional machinery-arrangement and maneuvering issues
sions of the wedge (diameter or largest dimension of the cross- as well, because of the unusual longitudinal location of the out-
section) would be on the order of 1500 mm, and the material board propulsors. From the purely seakeeping perspective, how-
thickness around the perimeter of the section about 120 mm. In- ever, it does allow the higher-freeboard catamaran section to be
ternal diaphragms could also be required inside the wedge to help forward, even if this may only mean that a bow-wetness concern
maintain the desired distribution of contact forces, and to transfer has been exchanged for wet-deck slamming.]
forces to the jack screw.
Other Design Issues In principle, the adjustable-length tri-
Apart from locking the centerhull in its extended position at maran may (and probably will) have propulsion machinery lo-
sea, the system must also provide a means of retracting and ex- cated in all three hulls. Intakes and exhausts from centerhull
tending the centerhull. Guides for this motion may take the form propulsion engine(s) present an unusual arrangement and gas-
of rails with wheeled carriages or slides. The loads on the guides plume problem. It can be assumed that retraction and extension
are expected to be modest by comparison with the loads on the operations will be performed with the center engines secured, as
locking system, but certainly not negligible. The effects of weight this would only be done at low speeds in sheltered water. Even so,
distribution and trim, which can vary widely depending on fuel on arrival the center engines and machinery spaces would still be
and cargo loads, will be at least one possible driver of load re- rather hot, and the machinery and enclosures would still be radiat-
quirements on the guides, even in calm water. Types of actuators

4
stagger with the centerhull forward) is a leading concern. While
wetness of the centerhull is not necessarily an intractable design
issue, it will require careful attention. The relatively small reserve
buoyancy and breadth forward has been observed (during model
tests) to permit wave impacts to occur on the forward surfaces
of the cross structures. Also, asymmetric pitch responses (bow
down peaks substantially greater than bow up) of the kind noted
for tumblehome wave-piercing forms in general, may become a
limitation in severe seas states. Further hydrodynamic research
and the development of appropriate analytic tools, as discussed
Figure 4: Adjustable-Length Trimaran with Aft Stagger of the below, will be extremely helpful if this concept is to be further
Centerhull developed.

DESIGN ANALYSIS
In this work, several different numerical methods were
ing. Several alternative approaches for air management have been
used to predict resistance and seakeeping performance of the
considered, of which the most promising is to route intake and ex-
adjustable-length trimaran. Resistance was predicted with a
haust for the centerhull engines through vertical trunks incorpo-
program that uses thin-ship theory, and a code that solves
rated in the catamaran section. This allows a conventional exhaust
the unsteady Reynolds-Averaged Navier-Stokes (RANS) equa-
location with respect to the catamaran superstructure. However it
tions. The seakeeping performance is simulated using a velocity-
demands the design of a watertight seal detail (at or near the level
potential formulation and strip theory. The adjustable-length tri-
of the wet deck of the catamaran section). These seals must be
maran represents a unique case in terms of the importance of hull
made effective when the centerhull section is extended, and then
interactions, forces and moments, and the geometry and propor-
can be “broken” before disengaging the locks and retracting the
tions of both the below water and above water portions of the
centerhull section. The development of seals for these arrange-
hull and cross structure. In order to make a quantitative assess-
ments is not a trivial design problem, but it is considered feasible.
ment of the accuracy of the performance predictions, a suitable
Distributive systems and personnel access to the centerhull benchmark was needed. Accordingly, a large model test program
present another peculiar challenge to the designer. Important sys- was conducted at the Maritime Research Institute Netherlands
tems that have to be able to connect the centerhull and catamaran (MARIN).
sections are electric power, fuel fill and transfer, and internal com-
munications, control and monitoring. The centerhull and catama- Experiments
ran sections can (and probably will) have independent ballast and A 1-to-34 scale model was used for experiments in the pow-
bilge systems, firemain, seawater cooling, and other auxiliary sys- ering and seakeeping basins. Calm-water resistance was mea-
tems such as lubrication oil, fire extinguishing, compressed air, sured with the center and sidehulls arranged in different con-
and ventilation. However, there are sound reasons to keep the figurations, and the motion and inter-subhull force components
electric generating plant entirely in the catamaran section, and in were measured with the model self-propelled in a range of sea
this case electric power to the centerhull will have to be trans- states. The resulting dataset comprehensively describes the major
ferred by internal “festoon” or cable and reel systems. Signal elements of hydrodynamic performance of an adjustable-length
connections that must be hard-wired may also use a similar fes- trimaran. It also provides unique guidance for the design of
toon or cable and reel, or possibly some wireless connections can unique structural and mechanical systems required to make an
be incorporated. Fuel fill and transfer systems, due to large hose adjustable-length trimaran feasible.
sizes, may be somewhat more difficult, however, systems can be Resistance Test Program The three individual hulls were con-
connected after the hull is extended, and disconnected prior to re- structed out of wood and connected together using aluminum
tracting. Personnel and material access arrangements presumably beams which allowed for easy reconfiguration. The model was
will require a system of double watertight hatches to trunks, with operated in six different loading configurations of varying beam,
locations provided for use in the extended and retracted hull posi- length, and displacement, as listed in Table 2 and viewed in Fig-
tions. Details of these systems have not been developed, but they ure 5. Essentially, this consisted of three different staggers of
are considered feasible. the sidehulls and two different offsets s2 /2 of the centerplane of
The adjustable-length trimaran is an interesting vehicle for the sidehulls from the centerplane of the centerhull. During the
the extremely challenging HSSL speed, range, payload, and aus- planning phase of the experiments, we had intended that the draft
tere port access constraints. At least to the level of a concept should be identical for all test configurations. Unfortunately, the
design weight estimate, it meets the basic HSSL mission require- freeboard proved to be too small for the reduced dimensions of
ments. In exchange for this, however, the concept presents sev- Configuration 4 and Configuration 5; as a result, an excessive
eral unusual design challenges. Apart from primary structural amount of spray impacted on the bridging cross structure con-
loads and locking arrangements, which are obviously key areas necting the three subhulls. This deficiency in the design of the
for further analysis, the low freeboard of the centerhull (assuming trimaran could easily be corrected in any planned extension to the

5
project. The final test matrix operated Configurations 1 to 3 and 6
in the heavier departure-displacement, and Configurations 4 and 5
in the lighter arrival-displacement.
The model was attached to the carriage and towed with a
vertical heave staff and gimbal so that the model was free to
heave and pitch. The longitudinal location of the tow point cor-
(a) Configuration 1 responded to the longitudinal center of buoyancy of the trimaran
(which varied depending on the stagger of the sidehulls). Sand
strips were applied to the bow of each hull to stimulate turbulence.
An image of the model, arranged in Configuration 4, is shown in
Figure 6 where details of the bow geometry and the location of
the sand strips can be seen.

(b) Configuration 2 The same models were used for the seakeeping tests which
required the installation of a water jet in each hull. The water jet
openings were covered for the resistance tests so that the hulls
were effectively bare.
Seakeeping Test Program The seakeeping test program was
conducted in MARIN’s Seakeeping and Manoeuvring Basin. We
concentrated our interests regarding motions on Configuration 6,
(c) Configuration 3 because this is the heavy-load long-range fuel-efficient mode for
which the adjustable-length trimaran was specifically designed. It
is this configuration that would pertain to the vast majority of the
operational time and that would present the principal concern for
the operators of the vessel. The three hulls were connected us-
(d) Configuration 4 ing force frames so that the six degree-of-freedom inter-subhull
forces could be measured, and a portion of the hull structure was
attached in the bow region to facilitate and measure under-deck
slamming. The inertial properties of the seakeeping model are
summarized in Table 3.
(e) Configuration 5 A water jet was installed in each hull so that the model could
be tested in a self-propelled condition. The nozzles on the outer
hull jets were steerable and controlled by an autopilot to keep the
model on course.
The wave test conditions for the model tests with respect to
motions fell into two broad categories: regular waves and irreg-
(f) Configuration 6
ular waves. For each category, the model was operated in dif-
Figure 5: Bottom View of the 6 Hull Configurations ferent combinations of two forward speeds, 25 and 43 kn, wave-
heading angles between 0 and 180 degrees, and wave steepness
or sea state. The wave conditions are noted in Table 4. Because
of the vast quantity of data that was collected, we shall constrain
our discussion in this paper to just the two cases of a 90-degree
heading and a 150-degree heading. The reason for this choice is
that we firstly desired to compare the regular-wave analysis with
the irregular-wave analysis. This eliminates the cases of the 0,
120, 180, and 285-degree headings. Secondly, the cases of 0 and
180-degree heading are less interesting, because some modes of
motion are (ideally) zero.
The regular seas tests at the 150 degree heading used wave-
lengths spanning one-half to six times the overall ship length. The
beam seas tests were conducted at wavelengths between one to
seven times the ship beam. All regular seas tests were done with
Figure 6: Image of Model in Configuration 4 a wave amplitude of 1 m, though repeat tests were done near fre-
quencies of large response. For the 150 degree heading, addi-

6
Table 2: Details of Model Experiments for Resistance

Overall Overall Sidehull Sidehull Displace- Speed


Config- Draft
Length Beam Stagger Offset ment Range
uration T (m)
L (m) B (m) r2 (m) s2 /2 (m) ∆ (t) U (kn)
1 168.6 56.0 0.0 23.0 6.51 17,760 20–56
2 210.6 56.0 −42.5 23.0 6.51 17,760 20–60
3 253.1 56.0 −85.0 23.0 6.51 17,760 20–54
4 168.2 40.0 0.0 15.0 5.19 12,340 20–54
5 210.2 40.0 −42.5 15.0 5.19 12,340 20–48
6 253.1 40.0 −85.0 15.0 6.51 17,760 20–58

Table 3: Details of Seakeeping Model

Item Symbol Units Value


Configuration 6
Longitudinal center-of-gravity from transom LCG m 109.68
Vertical center-of-gravity above keel KG m 10.00
Mass radius of gyration about longitudinal axis kxx m 12.92
Mass radius of gyration about lateral axis kyy m 51.00
Mass radius of gyration about vertical axis kzz m 54.90

Table 4: Details of Model Experiments for Seakeeping

Displace-
Config- Speed Heading Wave
ment
uration U (kn) β (degrees) Types∗
∆ (t)
6 17,760 43 0 I
6 17,760 43 90 R&I
6 17,760 43 120 I
6 17,760 43 150 R&I
6 17,760 43 180 R&I
6 17,760 25 180 I
6 17,760 43 285 R
6 17,760 43 300 R&I

∗ R (Regular), I (Irregular)

7
tional waves of varying amplitude, namely 0.5 and 0.25 m, were and Doctors (2007). In this way, the effective hydrodynamic
tested at the most interesting ship-to-wavelength ratio of unity. length of the subhulls grows with increasing vessel speed in
Additional beam seas tests were conducted at the large-response a physically realistic manner that has been substantiated by
frequency corresponding to the wavelength-to-beam ratio of 1.6. numerous towing-tank experiments.
For this repeat frequency, an amplitude of 0.5 m was used.
The irregular seas results that are analyzed in this paper • The transom-stern model includes a procedure for estimat-
were generated using Pierson-Moskowitz-type spectra. The wave ing the progressive unwetting or ventilation of the transom.
power spectrum of a given test is defined by the significant wave This feature permits a practical estimate of the transom-stern
height, and the peak period. We thoroughly tested the condition or hydrostatic drag suffered by such vessels. This proce-
that simulates Sea State 4, because the mission requirements man- dure has been detailed by Doctors, Macfarlane, and Young
date that the vessel operate fully in this environment. Accord- (2007). In that later publication, the previous very exten-
ingly, the largest portion of the test matrix was attributed to a sig- sive experiments on transom-stern ventilation were repeated
nificant wave height and peak-wave period corresponding of 2 m in a more accurate manner, through the use of specially de-
and 8.8 s, values that emulate Sea State 4. Additional irregular signed “wetting probes” mounted around the transom stern.
seas conditions were tested using peak periods that ranged from We now possess much more reliable data for this purpose
12 to 31 s, and significant wave heights up to 7.5 m. and have, in the process of the investigation, verified that the
critical value of the transom-stern Froude number to obtain
Numerical Tools full ventilation is around 2.5 — at least in the case of a tran-
Resistance Using Thin-Ship Theory In the current work, we som stern with a rectangular cross section at model scale.
describe an exhaustive investigation into the resistance of a can-
didate for a high-speed sealift trimaran. The purpose of the effort • As a special task inspired by the model experiments to be
was twofold: to consider a suitable trimaran design and to test our described here, a further procedure was incorporated in the
available computational tools. Computations were performed in a computer code in order to take into account the viscous in-
“blind fashion”, in that the experimental results for the resistance teraction between the subhulls. This simplistic procedure
were not initially made available to those persons performing the makes an allowance that the subhulls are in the “shadow”
calculations. Of course, the opportunity was later provided to of each other to some degree, as seen in a profile or a side
learn from the comparisons of theory and experiment and, conse- view. This shadow will be greatest when the subhulls are
quently, to improve the software. abreast of each other and will be particularly strong when
The software developed by Doctors and Day (1997), Doctors they are of similar size. In the present theory, the local gap
(1999), and Doctors and Scrace (2003), based on extensions to between the subhulls below the loaded waterplane was com-
the Michell (1898) theory, represented one focus of our computa- puted. From this, an estimate of the increase in local water
tional work. The other principal computational focus was based speed was obtained. This increase in speed was applied to
on computational-fluid dynamics (CFD) as described later in this the usual formula for frictional resistance, thus leading to an
paper. approximate technique for determining the frictional form
factor.
The current version of our thin-ship software contains a num-
ber of enhancements, as follows:
Of course, it is clearly understood that a number of physical
• The program can handle any number of subhulls up to six, by phenomena is neglected here. Neglected effects include the de-
employing additional centerplane source distributions. The formed shape of the free surface and the fact that the flow would
transverse velocities induced by one subhull should be cor- tend to skirt around the outside of the vessel, in cases where the
rected for by the use of laterally-directed centerplane dipole intersubhull gap is very small. However, we wished to preserve
distributions; however, this is ignored for the sake of con- the characteristically simple approach of the software as well as
siderable simplification to the coding of the software. It can its fast computational behavior.
be shown that our simplification is reasonable, provided the The reader is referred to the work of Maki, Doctors, Rhee,
relative lateral spacing between the subhulls of the multihull Wilson, Beck, and Troesch (2007), for our first publication on
is sufficiently large. We admit that in the current project, this research and development project.
we have been burdened with a stringent design requirement
Seakeeping Using Strip Theory The computer program
on the overall beam of the vessel. This implies that we are
KORVIN consists, broadly speaking, of two main modules.
testing this simplifying assumption to the limit.
The purpose of the first main module is to compute the two-
• An enhanced model for the transom-stern flow has been un- dimensional (sectional) hydrodynamic coefficients. The approach
der development for 11 years now. The current model allows utilized here is a direct enhancement of that developed by Doc-
for a realistic extension to the vessel by means of a virtual ex- tors (1986) for the case of an infinite fluid domain, without a free
tension to the vessel in order to represent the presence of the surface. In the current case, we are concerned with a ship section
transom-stern hollow. This has been described by Doctors oscillation at or near the free surface; the mathematical analysis
and Beck (2005), Maki, Doctors, Beck, and Troesch (2006), was presented in detail by Doctors (1988). It is effected through

8
the use of the boundary-element method (BEM), also called the sponding sections of the other subhull and this must be taken
panel method. into account in the theory. In the second method, the subhulls
The surface (more precisely, the line boundary) of the sec- are assumed to behave independently, from the hydrodynamic
tion is subdivided into a series of approximately equal straight viewpoint. This second approach is assumed to be more appro-
segments, typically 20 to 40 in number. A source strength, pul- priate for the vessel when traveling at a high forward speed and
sating harmonically in time, is distributed uniformly along each the waves generated by one subhull are swept downstream before
line segment. The influence of the source panel on the field point they can interact with the other subhull.
is derived analytically. The resulting simultaneous equations for These ideas were studied by Doctors and Scrace (2004) on a
the source strengths are solved in the usual way, thus giving the trimaran hull model similar to the RV Triton Trimaran Demonstra-
source strengths. tor. It was shown that, at typical forward speeds, corresponding
The source strengths are complex with respect to the time. to Froude numbers up to 0.3112, using the independent-subhull
One can then compute the velocity field and, in particular, the analysis yielded superior predictions.
induced pressures on the field panels. The in-phase pressures Further confirming this speculation, a second and more re-
are next integrated around the section to yield the added mass. cent application of the software to a trimaran by Hebblewhite,
The integrals of the out-of-phase pressures are used to obtain the Sahoo, and Doctors (2007) demonstrated that, indeed, better
damping coefficients. Two particularly effective enhancements correlation with theory appeared to be achieved through the
are also employed in the computer program. The first enhance- independent-subhull analysis for both low and high speeds.
ment is to integrate the induced velocities over the field panel as This satisfactory outcome was confirmed even more power-
well as the source panel, when computing the matrix of influ- fully by Thomas, Doctors, Couser, and Hackett (2007). This third
ence coefficients. This approach is sometimes referred to as the example was for a catamaran at zero speed. It is true that gener-
Galerkin technique. The Galerkin technique provides more accu- ally very good agreement with experimental measurements for
rate results. That is, fewer panels are required in order to obtain the heave, yaw, and pitch motion responses was achieved with
a prescribed level of accuracy, compared to the original work of both approaches, even in this very extreme limiting case. Never-
Frank (1967). theless, the analysis assuming hydrodynamic interaction between
It should be mentioned here that other methods exist for en- the two demihulls was markedly more accurate, as anticipated.
hancing the convergence rate of the basic panel method. These in- We note that there is very recent work that focuses precisely
clude the use of higher-order singularities by Hess (1975), curved on improving conventional strip theory for multihull vessel sea-
panels by Bristow (1977 and 1978), and a spline fit by Inoue, keeping. The Ph.D. thesis of Belknap (2008) developed a theory
Kuroumaru, and Yamaguchi (1977). that includes hull-interaction effects between hulls for the lower
We also claim that our approach is simpler to apply than these ship speeds for which the effects are likely. In his work, a far-
approaches. Moreover, it is more elegant and robust than that of field approximation is invoked, whereby the radiated waves from
Barnell (1984), for example, who effected the field-panel inte- one demihull section appear as incident waves to another demi-
gration numerically. It is clearly better to do this algebraically if hull section. The interacting sections are determined by the vec-
feasible. tor combination of the group velocity of demihull radiated waves
The second enhancement in the software is to utilize a “lid” and forward speed of the ship. For very high speeds, all the ra-
over the internal free surface of the section. This feature pos- diated waves are swept downstream and there is no interaction.
sesses the considerable advantage of eliminating the problem of Comparisons of the theory of Belknap to model test data and 3D
the “irregular frequencies”, without any other special treatment. computations have shown that it captures the correct trends and is
The irregular-frequency problem was described, for example, by an improvement over conventional strip theories.
Adachi and Ohmatsu (1979), and Haraguchi and Ohmatsu (1983). In the present investigation, however, the first two theories
The second main module of the software is based on the strip will be plotted and compared with the experimental data in order
theory of Salvesen, Tuck and Faltinsen (1970). This paper has to test this hypothesis of hull-interaction.
become the core reference for the strip theory, in which the sec- Resistance Using RANS Computational tools continue to im-
tional hydrodynamic properties of the ship are integrated along prove in both speed and accuracy, and in particular, Reynolds-
the length of the ship in order to obtain an approximation to the Averaged Navier-Stokes (RANS) methods have demonstrated in-
three-dimensional hydrodynamic coefficients of the ship. creased capabilities in predicting the flow around ships and sub-
In the case of a multihull vessel, as in the present study of marines in recent years. Much of this is due to recent advances
the motions of a trimaran, the program can be executed using two in computing technologies, particularly parallel processing tech-
different methods. In the first method, the abovementioned two- niques, and the application of complex unstructured meshing al-
dimensional sections are taken right through all through subhulls. gorithms that can more efficiently produce suitable computational
It is considered that this should be the appropriate form of anal- grids, even for complicated geometries. What is still required,
ysis for a stationary or low-speed vessel. That is, the sections however, is to continue to study predictions to more clearly un-
of one subhull would interact hydrodynamically with the corre- derstand how these expensive simulations can be best utilized in

9
the design and analysis of marine vehicles. A review of applying promising accuracy. The polyhedra elements, on the other hand,
RANS techniques to a variety of surface ship configurations was are better utilized in resolving the near-body regions that contain
given in Gorski (2004). An assessment of different computational complex geometry. Figure 7 also shows the predicted thickness
tools was given in Wilson, Fu, Fullerton, and Gorski (2006). of the air-water interface, which is seen to occur over at most
The commercially available software FLUENT is a general two to three cells. While it is acknowledged that the use of the
purpose Reynolds-Averaged Navier-Stokes (RANS) solver. It uti- polyhedra elements does compromise the interface resolution in
lizes a volume-of-fluid (VOF) method to predict the location and the near-body regions, it has been demonstrated that this does not
evolution of the free surface. In the VOF approach, it is assumed significantly impact the accuracy of the force predictions (Maki
that the discrete volume elements that comprise the domain are and Wilson, 2008), which will be presented in the following. The
filled with a combination of separate fluids, in this case air and grids used in this work had a total cell count of approximately
water, and the sum of the volume fractions of each fluid in a con- 1.6M cells, composed of 1.3M polyhedra and 0.3M hexahedra.
trol volume equals unity.According to this constraint, the advec- In the present study, the convective terms were discretized
tion of only a single phase must be computed. using the QUICK scheme (Leonard and Mokhtari, 1990). This
FLUENT uses a cell-centered finite-volume method and can scheme uses a weighted average of a second-order upwind
utilize discrete volume elements that are constructed of an arbi- method and a second-order central interpolation method, which is
trary number of sides, appropriately referred to as polyhedra. The typically more accurate for structured grids aligned with the flow
advantage of allowing an arbitrary number of sides per cell is that direction. For unstructured regions of the domain, the scheme re-
different standard element types such as tetrahedra, hexahedra, verts to the second-order upwind method. Therefore, this scheme
pyramid, and prismatic, can be used to generate the mesh. In the is very appropriate for the present case, which involves a hybrid
latest version of the software, tetrahedral zones can be converted mesh that contains both structured hexahedral elements which are
into polyhedra using an agglomeration process where the tetrahe- mostly aligned with the general flow direction, as well as poly-
dral nodes become the polyhedral cell-centers. This conversion hedra elements for flexibility in resolving the model geometry
has several advantages, the most prominent are the reduced cell and near-body flow. The discretization of the volume fraction
count and greatly improved mesh quality through reduced skew- equation used a modified version of the high-resolution interface
ness. The reduced cell count is an obvious advantage, and skew- capturing (HRIC) method (Muzaferija et al, 1998). The HRIC
ness degrades accuracy and impedes solution convergence. Here scheme has been shown to be more accurate than the QUICK or
we use a cell skewness metric that assesses the deviation of the other second-order discretization schemes, and is less computa-
cell face angles from those of a face belonging to an equiangular tionally expensive than using a complete geometric reconstruc-
cell. A value of 0 indicates a best-case equiangular cell, and a tion of the interface, which is also available in FLUENT. The tur-
value of 1 indicates a completely degenerate cell. In the present bulence closure was accomplished using a variant of the k − ω
simulations, the maximum skewness for the near-field region was model as described in Wilcox (1998). The pressure-velocity cou-
reduced from approximately 0.8 when discretized using tetrahe- pling is based on a SIMPLE type segregated solution algorithm.
dral elements to 0.4 after converting to polyhedra. By reducing The simulations were performed as fully unsteady in time, march-
both the cell count and the skewness, it is possible to reduce so- ing to an equilibrium solution. The time advancement was ac-
lution times and enhance accuracy. complished using first-order Euler-implicit differencing. Alge-
braic multi-grid (AMG) methods using pointwise Gauss-Seidel
The computational grids used in the present study exploit iteration were used to aid in the solution convergence.
the advantage of using different element types in the different
regions of the domain where they excel, while also attempting While the capabilities exist within FLUENT to perform pre-
to reduce the difficulties and time associated with generating a dictions of the motions of the ship, the simulations in this effort
suitable mesh. The near-body mesh consists of boundary-layer utilized experimental measurements for the sinkage and trim to
prisms that have a structured nature in the wall normal direction. fix the attitude and position of the model. This provided a sig-
The use of prisms allows for a fewer number surface cells to cover nificant reduction in the computational expense to perform the
the body while providing the necessary spacing to resolve the vis- calculations, and reduces the number of variables in comparing
cous layer appropriately. The farfield domain consists of hexahe- the simulated and experimental drag forces. An effort is currently
dral elements which are aligned with the direction of the flow and underway to assess the accuracy and computational expense as-
clustered near the free surface. The intermediate area between the sociated with dynamically predicting the sinkage and trim as part
boundary-layer-prism cap and the farfield domain is connected of the solution process.
using tetrahedral elements which are converted into polyhedra us- Results for Calm-Water Resistance
ing the previously explained procedure. We turn to Figure 8, whose six parts show the experimental
Figure 7 shows the surface mesh on the centerhull and the and computed results for the abovementioned six configurations.
vessel symmetry plane, and demonstrates the use of each of the The symbols on these plots are listed in Table 5.
element types. The hexahedral elements are effective in solving The curves for the total resistance (or the resistance compo-
the undisturbed free surface locations due to the ability to allow nents) are all represented in a dimensionless form known as the
significant stretching in the streamwise direction without com- specific resistance. This is the ratio of the total resistance (or the

10
(a) Configuration 1, ∆ = 17,760 t, L = 168.6 m, B = 56 m (d) Configuration 4, ∆ = 12,340 t, L = 168.6 m, B = 40 m

(b) Configuration 2, ∆ = 17,760 t, L = 210.6 m, B = 56 m (e) Configuration 5, ∆ = 12,340 t, L = 210.6 m, B = 40 m

(c) Configuration 3, ∆ = 17,760 t, L = 253.1 m, B = 56 m (f) Configuration 6, ∆ = 17,760 t, L = 253.1 m, B = 40 m

Figure 8: Calm-Water Specific Resistance at Model Scale Plotted at Full-Scale Speed

11
Table 5: Nomenclature for Plots of Resistance
Symbol Meaning
U Speed of vessel
W Displacement weight
fs Surface-velocity limit factor
r2 Stagger of sidehull
s2 /2 Offset of sidehull
∆ Displacement mass
RF Frictional resistance
RH Hydrostatic resistance
RT Total resistance
RW Wave resistance
Rair Air resistance

a study of the resistance formula attributed to Michell (1898).


We first consider Configuration 1 in Figure 8(a). The wave
resistance RW possesses its typical maximum at the “hump” con-
dition. This occurs for our vessel at a speed which varies accord-
ing to the configuration. Thus, for Configuration 1, the hump
speed is about 43 knots. This is coincidentally the design opera-
tional speed. The wave resistance is the resistance experienced by
the vessel due to the creation of the wave system associated with
the forward motion of the vessel through the water.
The hydrostatic resistance RH is seen to be relatively small,
indicating that little of the transom is immersed in the water. The
hydrostatic resistance is due to the presence of the transom stern.
It is defined and computed on the basis of the lack of hydrostatic
pressure on the face of the transom, due to the assumed clean
separation of the flow of water past the hull at the transom station.
Figure 7: CFD Grid
It is clear that the presence of a transom gives rise to an
additional and unwanted resistance component in contrast to a
resistance components) to the vessel weight W . It is important to more traditional vessel with a streamlined stern. However, it is
clarify that the calculations are all executed at model scale, in or- not immediately obvious whether the total resistance increases as
der to be directly comparable with the model experiments. How- a result of the transom. This is because the wave resistance is
ever, the results for resistance are plotted versus the prototype also affected by the presence of the transom. The reader is re-
velocity scale, namely the speed U in knots, but without any cor- minded that the wave resistance is computed here assuming that
rection to the resistance due to the fact that the full-scale Reynolds the effective vessel moving through the water is the actual vessel
number is substantially greater. Thus, for a realistic prototype cal- combined with the added hollow in the water behind the tran-
culation, one must perform the appropriate Froude extrapolation. som. Thus, Doctors (1999), in his extensive comparison between
This would substantially lower the specific frictional drag and, different multihull configurations, demonstrated theoretically that
consequently, the specific total drag. the presence of a modest transom resulted in a lowering of the to-
tal resistance — compared to the resistance of a vessel either with
The plots also indicate the relevant values of the offset s2 /2 no transom or a vessel with a substantial transom.
of the sidehull for the six configurations. This is measured be-
tween the centerplane of the centerhull and the centerplane of the We should also add here that the reason for incorporating a
sidehull. Additionally, the stagger r2 of the sidehull is shown. transom in a marine vessel may, in fact, not be related at all to
This is measured as the location of the sidehull transom ahead of hydrodynamic considerations. The transom, because of its ge-
the centerhull transom. The value of r2 was zero or negative for ometry, lends itself to the most convenient location for siting the
all the configurations, because we believe that sidehulls staggered waterjets, which are often used for the propulsion of such vessels.
aft lead to a more practical design. Curiously, however, the theo- The frictional resistance RF is estimated from the 1957 ITTC
retical resistance of the trimaran is lowered a similar amount for line and is seen to be the major drag component, at least at model
positive and negative staggers. This fact can be demonstrated by scale, for the higher-speeds, namely the 50 to 60 knot range.

12
The air resistance Rair is the fourth drag component plotted Thus, we also allow the possibility of moderating the intersubhull
on the graph. We computed this drag component assuming that influence, as detailed in the following paragraph.
the drag coefficient was 1.000, based on the frontal area of the It is considered that the prescription explained here overes-
model ship. This drag component is negligible for our design. timates the effect under study, as noted when the vessel speed U
The simple sum of these four components is indicated by the exceeds 44 knots. Thus, yet another curve is shown, indicated by
first of the three curves for total resistance RT . This predicted fs = 1.2. The meaning of this is that the computer coding places
result may be compared with the experimental data as represented an upper limit on the subhull surface velocity of 1.2 times the
by the symbols. vessel speed.
It can be seen that there is a total of four curves for the total The other five parts of Figure 8 pertain to the other five con-
resistance RT . As indicated on the graphs, the first curve (indi- figurations. On the whole, it is confirmed that the surface-velocity
cated by circular symbols) represents the experimental data de- limit factor fs = 1.2 does seem to be a reasonable compromise
rived from the model tests. The other three curves for RT (each for achieving the best correlation between theory and experiment.
with a different value of fs associated with it) were obtained from One may note that the hump-like nature of the wave-resistance
the theoretical work, as explained in the following text. curve for Configuration 1 is diminished for Configuration 4 and it
It can be seen that this first prediction of the total resistance, is essentially non-existent for the remaining four configurations.
indicated by the symbol fs = 1, falls short of the experimental This is to be expected because it is well known that longer vessels
resistance. This discrepancy increases with speed. It is believed (Configurations 2, 3, 5, and 6) possess hump-free wave-resistance
that most of this discrepancy can be traced back to the interactions curves.
between the subhulls increasing the frictional drag, as noted ear- Figure 8(d) and Figure 8(f) also show some calculations
lier in this paper. The symbol fs is a surface-velocity limit factor, based on the FLUENT software. Because computational times
which we employed in our simple resistance-prediction software are much greater for these very extensive and detailed CFD stud-
(not the CFD computations, which requires no such correction ies, a limited number of computed points only is depicted here.
factors). For this reason, we have chosen to use square symbols to indi-
To emphasize this point, one has only to note that the sub- cate the FLUENT predictions, whereas it is conventional to use a
hulls each possess a maximum local beam of approximately 10 m curve for theoretical predictions. It is indeed most encouraging to
at the waterline. With an overall subhull centerplane spacing s2 observe the excellent matching between the FLUENT data points
of 46 m, this leads to a minimum gap of 13 m. This suggests that and the experimental data. This statement is particularly true for
the water has been channeled from an initial width of 23 m down Configuration 6, the laterally-closely-spaced, fully-extended con-
to 13 m, implying a large increase in the local speed of the water figuration, for which we are most interested in achieving high-
over the surface of the hull. On the other hand, it should be noted fidelity predictions.
that this effect will be much less on other parts of the subhull sur- The average relative difference between prediction and ex-
face. The current estimate of the velocity increase does take this periment, when using the speed range of 20 to 50 knots in Con-
effect fully into account in the third theoretical prediction, indi- figuration 6, is 9.9% and 1.8% for the thin-ship theory and CFD
cated by fs = ∞. results respectively. For the less operationally important, and hy-
In the computer program, the local “geometric channel drodynamically more challenging Configuration 4, the average
width”, or gap, between the subhulls is computed. The ratio of differences were 16% and 8% respectively. When comparing the
the subhull centerplane spacing to this local gap, which is always two computational methods, we must recall the extra expense that
greater than unity, provides an upper limit on the ratio of the lo- is associated with achieving the increased accuracy. The time re-
cal water speed to the speed of the vessel. This ratio is termed quired to complete a converged drag result for a single forward
the “subhull surface-velocity ratio” in our analysis. (This ratio speed is approximately 24 to 48 hours using 24 processors for the
is 23/13 = 1.769, at the point where the subhull local beam is RANS method and only about 1 second on a single processor for
greatest, in the above example). the thin-ship theory code.

Frictional resistance essentially depends on the square of ve- The RANS computation is also useful for the prediction of
locity. But there is a weak Reynolds-number effect as well, lead- the free surface elevation. For the adjustable-length trimaran ge-
ing to a drop in the frictional resistance coefficient with speed. ometry, the designer is concerned with wetness of the fore and
Thus, the computer subroutine calculates the average value of the underdecks, the water level near the water jet intakes, and the
subhull-surface velocity ratio to the 1.8 power (rather than the interference between the center- and sidehulls. A sample pre-
second power), over the surfaces of the individual subhulls. diction of free-surface elevation from the RANS computations
for an equivalent forward speed of 40 knots is shown in Fig-
This new factor is used to correct the frictional resistance ure 9. A closer examination of this figure also points out how
based on the standard 1957 ITTC formulation, for each subhull, the contour of the wave elevation appears discontinuous along
on an individual basis. This simple correction procedure is a one- the non-conformal grid interface separating the near-body region
dimensional method. It no doubt ignores ameliorating effects due from the outer domain. At the interface, the vertical dimension
to the true three-dimensional nature of the flow over the subhulls. of the polyhedral cells is much larger than that of the stretched-

13
Table 6: Nomenclature for Plots of Ship Motions

Symbol Meaning
L Length of vessel
U Speed of vessel
g Acceleration due to gravity
ω0 Sea-wave radian frequency
γ Heading of vessel
A0 Wave amplitude
A2 Sway amplitude
A3 Heave amplitude
A4 Roll amplitude
A5 Pitch amplitude
A6 Yaw amplitude
ε2 Sway phase
ε3 Heave phase
ε4 Roll phase
(a)
ε5 Pitch phase
ε6 Yaw phase

hexagonal neighbors. Subsequently, the interpolation performed


during post-processing is more refined in the region of cells with
smaller spacing, thus leaving a visible discontinuity in the free-
surface contour. The consequence of this is only a reminder that
the method is solving an assumed continuous process discretely.
Because the focus of this effort was to examine the ability of the
tools to predict the model resistance, the rapid change in cell
spacing across the interface was deemed an acceptable way to
relax the mesh resolution requirements, while not significantly
impacting the resistance computation, as evidenced by the results
in Figures 8(d) and 8(f). Of course, if one were interested in ac-
(b) curately predicting the wave behavior further away from the hull,
then commensurate mesh resolution would be required. Wilson,
Fu, Fullerton, and Gorski (2006) have shown that increased grid
resolution enables a more accurate prediction of the wave eleva-
tion, but the effect on resistance prediction is still unclear and case
dependent since the wave field contributes to only a portion of the
total force.
It is conjectured that the RANS method more accurately rep-
resents the sidehull interference. The inter-hull wave field can be
seen in Figure 9(c), where the camera is placed in between the
center and starboard hull and faces aft. As pointed out earlier, the
subhull interaction increases with increasing forward speed and
hull closeness; hence, the discrepancy between the two computa-
tional methods grows as these parameters increase.
Results for Seakeeping
The five parts of Figure 10 present the results corresponding
(c) to the second row in Table 4, namely a heading of 90 degrees
(beam seas). Figure 10(a) is a plot of the response-amplitude op-
Figure 9: Free-Surface Contours erator (RAO) for sway, as a function of the dimensionless wave
radian frequency. The nomenclature used in these plots of ship
motions are presented in Table 6.

14
Four sets of data are shown on each plot. The first set of data, longitudinal direction, and results in much smaller wave build-
indicated as symbols (and denoted by “Exp reg”), show the exper- up. This issue can be clearly seen in figures 10 and 11 where the
imental results from the regular-wave tests. The second data (de- combined subhull predictions become very large and have wild
noted by “Exp irreg”) is derived from the irregular-wave tests and oscillations for certain frequencies.
was obtained through a fast-Fourier transform (FFT); it is plot- On the other hand, the separate-subhull concept assumes that
ted as a short-dashed curve. In this example, there were, in fact, the waves generated by a subhull section are swept downstream
three separate irregular-wave runs. So, these are graphed as three before they can impinge on a neighbor-subhull section. This
separate curves, joining the analysis points with straight line seg- method is more suitable for higher speeds. Indeed, it would be
ments, without any smoothing. One can observe relatively close entirely appropriate at speeds beyond the critical speed, when the
repeatability for these three irregular-wave tests. Perfect repeata- generated waves fail entirely to impinge on the neighbor subhull.
bility is not expected, even if the instrumentation were perfect,
because the irregular-wave system would have been different on For the low-frequency part of the range, the predictions are
the three occasions, thus altering the outcome of the FFT analysis. similar but this is not the case for the high-frequency part of the
range. For this wave-heading angle, the combined-subhull ap-
There is also close agreement between the irregular-wave proach is perhaps slightly more faithful.
data (as a group) and the regular-wave data.
Heave motion for the 90-degree heading case is presented in
Finally, the last two curves show the multihull strip-theory Figure 10(b). The correlation between the regular-wave tests and
predictions, using either the combined-subhull concept (denoted irregular-wave tests is seen to be better for heave. We also note
by “Theory comb”) or the separate-subhull concept (denoted by that the separate-subhull predictions are superior to those from the
“Theory sep”). combined-subhull analysis. Roll motion is shown in Figure 10(c).
By way of clarification of the theoretical analysis, we point Similar comments can be made for this mode of motion, with the
out that we have tested two different approaches for handling the exception that there are some differences between the experimen-
critical matter of the wave interactions between the subhulls. It tal data towards the low-frequency end of the curve. We may dis-
is clear that these interactions will influence the motion behavior cern undesired oscillations in the combined-theory calculations
of the vessel. In a more sophisticated strip-theory analysis, one at the higher frequency. It is assumed that these oscillations are
could account for waves created by one subhull impinging on its a result of the trapped-mode phenomena alluded to above. Pitch
neighbor or neighbors. is shown in Figure 10(d). All the data is of low magnitude be-
Utilizing the standard strip-theory simplification, one should cause, as is well known for beam seas, there is little excitation of
argue that the interaction due to the wave from a subhull section pitch motion for most marine vessels. Finally, yaw is presented
must occur at a downstream location on the neighbor-subhull sec- in Figure 10(e). The quality of the results is high as for the pre-
tion. The interaction would be further downstream for a greater vious subfigure. In both cases, it appears that the analysis of the
separation between the two subhulls under consideration and also irregular data breaks down, to some extent, for low frequencies.
for greater forward speeds of the vessel. This was amplified upon We now turn to the case of the 150-degree heading in the
by Lloyd (1989), where he published a relevant formula for this five parts of Figure 11. Once again, the five parts of the figure
relative downstream location. correspond to the standard modes of motion, numbered from 2
In strip theory used for the present calculations, the through 6. While there are general similarities for the results with
combined-subhull concept assumes simply that these inter- the case of the 90-degree heading, there are also some notable
subhull interactions take place at the same ship section. Thus, differences. In Figure 11 for the sway, there are some unnatu-
one can use the same ship-motion analysis, provided that the hy- ral responses at the low frequencies, not replicated in the much
drodynamic coefficients are computed on the basis of a multihull smoother separate-subhull theory. Also, at a dimensionless wave
cross section. The method of Doctors (1988), which is effec- frequency of 2.50, there is a relatively large variation in the re-
tively an enhanced version of the original approach developed by sponse from the repeated regular-wave experiments. Heave in
Frank (1967), was employed. It is certainly understood that this Figure 11 is well predicted. Disturbingly for roll, in Figure 11(c),
approach will be suitable at low speeds and would be most appro- we observe similar misbehavior for the experimental data that we
priate at zero forward speed. already referred to for the sway; it is even more exaggerated here.
Pitch in Figure 11(d) is well predicted. However, the experiments,
It should be pointed out that the problem with the combined again, appear to be a little misbehaved, because the proper low-
subhull concept is that because strip theory has divided the prob- frequency limit for the RAO of cos(30◦ ) = 0.8660 does not occur.
lem into a series of two-dimensional problems it is susceptible to Lastly, the yaw in Figure 11(e), again, seems high for the experi-
creating a “trapped wave” system between the subhulls. In two- ments.
dimensions there are certain resonant frequencies at which waves
can build-up between the hulls. The build-up leads to unrealisti- In summary for the motion responses, it can be stated that in
cally large hydrodynamic forces and in turn motion predictions. almost all cases, the separate-subhull theory provides a smoother
For actual three-dimensional bodies, the energy build up between and more realistic prediction of the experimental responses. This
the hulls does not happen because the energy can escape in the was anticipated in the introductory comments on the theory, based
on our previous experience.

15
(a) Sway (b) Heave

(c) Roll (d) Pitch

(e) Yaw

Figure 10: Seakeeping RAO, Beam Seas (γ = 90 deg), Configuration 6, U = 43 kn

16
(a) Sway (b) Heave

(c) Roll (d) Pitch

(e) Yaw

Figure 11: Seakeeping RAO, Bow-Quartering Seas (γ = 150 deg), Configuration 6, U = 43 kn

17
(a) Sway (b) Heave

(c) Roll (d) Pitch

(e) Yaw

Figure 12: Seakeeping RAO Phase, Beam Seas (γ = 90 deg), Configuration 6, U = 43 kn

18
(a) Sway (b) Heave

(c) Roll (d) Pitch

(e) Yaw

Figure 13: Seakeeping RAO Phase, Bow-Quartering Seas (γ = 150 deg), Configuration 6, U = 43 kn

19
We shall now discuss the phase angles corresponding to the hull forms was demonstrated in the comparison with the experi-
amplitudes in the previous two figures. The phases for the 90- ments.
degree heading case are plotted in the five parts of Figure 12. On The CFD predictions were conducted on computational grids
the whole, the agreement between the separate-subhull calcula- that utilize polyhedral elements near the body, and hexagonal fi-
tions (definitely not the combined-subhull calculations) and the nite volumes clustered near the free surface in the far field. Two
regular-wave experimental data is seen to be very good. It should of the six hull configurations were simulated, and the comparison
be borne in mind, with plots showing angles, that an angle of 180◦ to experiment was encouraging. The average difference over the
is equivalent to an angle of −180◦. Hence, some apparent gross tested speed range was approximately 5%. The difference relative
discrepancies (note the sole experimental point in Figure 12(d)) to experiment of the CFD results was generally smaller than that
are actually areas of very close correlation. of the thin-ship theory. It is important to emphasize the difference
Finally, the case of phase angles for the 150-degree head- in computational expense between the two methods: the CFD run
ing are shown in Figure 13. There is reasonable correlation for time is about one million times longer than the thin-ship theory
the predictions emanating from the separate-subhull analysis. On for each simulated forward speed.
the other hand, the combined-subhull analysis is seen to contain Strip theory was used to predict the motion RAOs for the tri-
considerable oscillations due to undesired and unphysical (non- maran. The added mass and damping were determined in one
existing) hydrodynamic interactions between sections of the dif- of two ways, either solving the boundary-value problem with all
ferent subhulls at the same station. subhull sections present at each station, or with a proper summa-
tion of the contributions to the forces of each individual subhull.
CONCLUSIONS The comparison to experiment is encouraging, especially with the
In this paper a novel design for a high-speed sealift vessel
predictions utilizing the separate subhull analysis. The designer
has been described. Initially several candidates that utilized the
must accurately predict loads on the centerhull in order to design
high-performance technologies of air-cushion assistance, multi-
the structure and articulation mechanism, and further analysis is
hulls, and adaptive geometry were considered. The demanding
necessary.
speed and range requirement of the vessel eliminated all of the
original prototypes. Subsequently, a standard trimaran necessar- The efficient velocity-potential based codes enabled the au-
ily evolved into an adjustable-length trimaran that achieved dras- thors to explore design tradeoffs, and even examine geometrical
tically reduced wave resistance by extending the centerhull. variants that fell outside the originally defined parameter bound-
aries. In fact it was this capability that led to the novel adjustable-
The hydrodynamic performance of the adjustable-length tri-
length trimaran. Regardless of the advancement in the speed and
maran was studied using three different numerical techniques to-
power of parallel computers, the velocity-potential based meth-
gether with a large model test program. Calm-water resistance
ods will always be orders of magnitude (in this study six) more
was predicted using a code based on thin-ship theory and the CFD
rapid. Conversely, for applications that can afford the extra ex-
software FLUENT. Seakeeping motions were predicted using a
pense, CFD is proving to deliver on its promise of higher accu-
program based on classical strip theory. The powering experi-
racy. Furthermore, CFD provides additional details of the entire
ments tested the trimaran in six different configurations of vary-
flow domain while robustly handling breaking waves and viscous
ing overall length, beam, and draft. The seakeeping model tests
effects.
were done in a wide range of sea states, heading angles, and two
forward speeds, with the model possessing the lowest-resistance
ACKNOWLEDGMENTS
large-length narrow-beam configuration. The authors would like to gratefully acknowledge the sup-
The adjustable-length trimaran geometry has subhulls that port of the US Office of Naval Research through the High-Speed
are located relatively closely to each other, and the subhulls are Sealift (HSSL) Program, under the project title, “ONR BAA 05-
heavily loaded, particularly upon departure. These attributes chal- 007: Architectural Concepts and Hydrodynamic Technologies for
lenge the viscous prediction that is used for calculating total drag High Speed Sealift to Austere Ports: Subtopic B: Computational
with the thin-ship theory code. Consequently, the viscous predic- Approach and Hydrodynamic Tools”.
tion was sensitized to the closeness of the hulls by estimating the The authors would also particularly like to express their ap-
increase in velocity realized by the intersubhull fluid. preciation to Dr Ir Frans van Walree, from the Maritime Research
The accuracy of thin-ship theory resistance predictions was Institute Netherlands (MARIN) in Wageningen, Netherlands, for
satisfactory when compared to the experiments; the average nor- his considerable assistance with the execution of the seakeeping
malized difference was approximately 10%. The most distin- tests and the subsequent analysis of the experimental data.
guishing attribute of this code in the present endeavor is the com- The CFD computations have been performed using resources
putational efficiency of generating predictions. The efficiency of available at the Ship Engineering and Analysis TECHnology
this code enabled the user to investigate new designs of greater (SEATECH) center, located at the Naval Surface Warfare Center
overall length in an effort to reduce fuel consumption. Further- Carderock Division (NSWCCD) and the Air Force Research Lab-
more, the ability of the code to accurately predict the effect of oratory (AFRL) Major Shared Resource Center (MSRC) High
change in length and beam on the resistance of the six different Performance Computing (HPC) facility.

20
REFERENCES D OCTORS , L.J. AND DAY, A.H. 1997, “Resistance Prediction
for Transom-Stern Vessels,” Proc. Fourth International Con-
A DACHI , H. AND O HMATSU , S. 1979, “On the Irregular Fre-
ference on Fast Sea Transportation (FAST ’97), Sydney, Aus-
quencies in the Integral Equation Solutions of the Time-
tralia, Vol. 2, pp 743–750, July.
Dependent Free Surface Problems,” Proc. Society of Naval
Architects of Japan, Fall Meeting, pp 127–136, November. D OCTORS , L.J., M ACFARLANE , G.J., AND YOUNG , R. 2007,
“A Study of Transom-Stern Ventilation,” International Ship-
BARNELL , A. 1984, “Boundary-Integral Methods for Solving
building Progress, Vol. 54, No. 2, 3, pp 145–163.
Potential Flows,” Massachusetts Institute of Technology, De-
partment of Ocean Engineering, Master thesis, 72 pp, Febru- D OCTORS , L.J. AND S CRACE , R.J. 2003, “The Optimisation of
ary. Trimaran Sidehull Position for Minimum Resistance,” Proc.
Seventh International Conference on Fast Sea Transporta-
B ELKNAP, W.F. 2008, “A Computationally Efficient Method for
tion (FAST ’03), Ischia, Italy, Vol. 1, pp KL.1–KL.12, Octo-
Nonlinear Multihull Seakeeping”, PhD thesis, Department
ber.
of Naval Architecture and Marine Engineering, Ann Arbor,
Michigan, USA. D OCTORS , L.J. AND S CRACE , R.J. 2004, “Hydrodynamic In-
teractions between the Subhulls of a Trimaran during Rolling
B RISTOW, D.R. 1977, “Recent Improvements in Surface Sin-
Motion,” Proc. International Conference on Design and Op-
gularity Methods for the Flow Field Analysis about Two-
eration of Trimaran Ships, Royal Institution of Naval Archi-
Dimensional Airfoils,” Proc. American Institute of Aeronau-
tects, London, England, pp 31–39, April.
tics and Astronautics Conference on Computational Fluid
Dynamics, Albuquerque, New Mexico, 11+i pp, June. F RANK , W. 1967, “The Heave Damping Coefficients of Bulbous
Cylinders, Partially Immersed in Deep Water,” J. Ship Re-
B RISTOW, D.R. 1978, “Improvements in Surface Singular-
search, Vol. 11, No. 3, pp 151–153, September.
ity Analysis and Design Methods,” Proc. Conference on
Advanced Technology Airfoil Research, NASA, CP 2045, G ORSKI , J. 2004, “Evolving Computational Capability for Ship
ATAR, pp 221–236. Hydrodynamics,” Naval Surface Warfare Center, Carderock
Division, Hydromechanics Department Technical Report,
DAY, A., C LELLAND , D., N IXON , E. 2003, “Experimental and
NSWCCD-50-TR-2004/058, December.
Numerical Investigation of ‘Arrow’ Trimarans”, Proc. Sev-
enth International Conference on Fast Sea Transportation H ARAGUCHI , T. AND O HMATSU , S. 1983, “On an Improved So-
(FAST ’03), Ischia, Italy, Vol. 3, pp D2.23–D2.30, October. lution of the Oscillation Problem on Non-Wall Sided Float-
ing Bodies and a New Method for Eliminating the Irregular
D OCTORS , L.J. 1986, “Application of the Collocation and
Frequencies,” Trans. West Japan Society of Naval Architects,
Galerkin Approaches to Potential-Flow Solutions by the
Vol. 58, No. 8, pp 9–23, August.
Boundary Element Method,” Proc. Ninth Australasian Fluid
Mechanics Conference, The University of Auckland, Auck- H EBBLEWHITE , K., S AHOO , P.K., AND D OCTORS , L.J. 2007,
land, New Zealand, pp 400–403, December. “A Case Study: Theoretical and Experimental Analysis of
Motion Characteristics of a Trimaran Hull Form,” Ships and
D OCTORS , L.J. 1988, “Application of the Boundary-Element
Offshore Structures, Vol. 2, No. 2, pp 149–156, June.
Method to Bodies Oscillating near a Free Surface,” Compu-
tational Fluid Dynamics — Proc. International Symposium H ESS , J.L. 1975, “The Use of Higher-Order Surface Singularity
on Computational Fluid Dynamics ISCFD-Sydney, Elsevier Distributions to Obtain Improved Potential Flow Solutions
Science Publishers B.V., Amsterdam, pp 377–386. for Two-Dimensional Lifting Airfoils,” Computer Methods
in Applied Mechanics and Engineering, Vol. 5, No. 1, pp 11–
D OCTORS , L.J. 1999, “On the Great Trimaran-Catamaran De-
35, January.
bate,” Proc. Fifth International Conference on Fast Sea
Transportation (FAST ’99), Seattle, Washington, pp 283– I NOUE , M., K UROUMARU , M., AND YAMAGUCHI , S. 1977, “A
296, August–September. Method of Singularity with Spline Fit Distribution for Poten-
tial Flow Problems,” Memoirs of the Faculty of Engineering,
D OCTORS , L.J. 2007, “A Numerical Study of the Resistance
Kyushu University, Vol. 37, No. 2, pp 89–106, June.
of Transom-Stern Monohulls,” Ship Technology Research:
Schiffstechnik, Vol. 54, No. 3, pp 134–144, July. L EONARD , B. P., AND M OKHTARI , S. 1990, “ULTRA-SHARP
Nonoscillatory Convection Schemes for High-Speed Steady
D OCTORS , L.J. AND B ECK , R.F. 2005, “The Separation of
Multidimensional Flow,” NASA TM 1-2568 (ICOMP-90-
the Flow past a Transom Stern,” Proc. First Interna-
12), NASA Lewis Research Center.
tional Conference on Marine Research and Transportation
(ICMRT ’05), Ischia, Italy, 14 pp, September. L LOYD , A.R.J.M. 1989, Seakeeping: Ship Behaviour in Rough
Weather, Ellis Horwood Limited Publishers, Chichester,
England, 486 pp.

21
M AKI , K.J., D OCTORS , L.J., B ECK , R.F., AND T ROESCH , S ALVESEN , N., T UCK , E.O., AND FALTINSEN , O. 1979, “Ship
A.W. 2006, “Transom-Stern Flow for High-Speed Craft,” Motions and Sea Loads,” Trans. Society of Naval Architects
Australian Journal of Mechanical Engineering, Vol. 3, No. 2, and Marine Engineers, Vol. 78, pp 250–279, Discussion:
pp 191–199. 279–287, December.
M AKI , K.J., D OCTORS , L.J., R HEE , S.H., W ILSON , W.M., T HOMAS , G., D OCTORS , L.J., C OUSER , P., AND H ACKETT,
B ECK , R.F., AND T ROESCH , A.W. 2007, “Resistance Pre- M. 2007, “Catamaran Motions in Beam and Oblique Seas,”
diction of a High-Speed Sealift Trimaran,” Proc. Ninth In- Proc. Ninth International Conference on Fast Sea Trans-
ternational Conference on Numerical Ship Hydrodynamics, portation (FAST ’07), Shanghai, China, Paper 53, pp 426–
Ann Arbor, Michigan, pp 64–78, August. 433, September.
M AKI , K.J. AND W ILSON , W.M. 2008, “Steady Drift Force Cal- T UCK , E.O. AND L AZAUSKAS , L. 1998, “Optimum Hull Spac-
culation on the Naval Destroyer Hull 5415,” Workshop on ing of a Family of Multihulls”, Ship Technology Research:
Verification and Validation of Ship Manoeuvering Simulation Schiffstechnik, Vol. 45, No. 4, pp 180–195, October.
Methods (SIMMAN ’08), Copenhagen, Denmark, April.
W ILCOX , D. C. 1998, Turbulence Modeling for CFD, DCW In-
M ICHELL , J.H. 1898, “The Wave Resistance of a Ship,” Philo- dustries, Inc., La Canada, California.
sophical Magazine, London, Series 5, Vol. 45, pp 106–123.
W ILSON , W., F U , T. C., F ULLERTON , A. AND G ORSKI ,
M UZAFERIJA , S., P ERIC , M., S AMES , P., AND S CHELLIN , J. 2006, “The Measured and Predicted Wave Field of Model
T. 1998, “A Two-Fluid Navier-Stokes Solver to Simulate Wa- 5365: An Evaluation of Current CFD Capability,” Proc.
ter Entry,” Proc. Twenty-Second Symposium on Naval Hy- Twenty-Sixth Symposium on Naval Hydrodynamics, Rome,
drodynamics, pp 277–289, Washington, DC. Italy, September.
NARITA , S. 1976, “Some Research on the Wave Resistance of a
Trimaran”, Proc. International Seminar on Wave Resistance,
Society of Naval Architects of Japan, Tokyo, Japan, pp 381–
388, February.

22

You might also like