You are on page 1of 16

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

net/publication/225521097

Multi-objective optimization of oil tanker design

Article in Journal of Marine Science and Technology · December 2010


DOI: 10.1007/s00773-010-0097-7

CITATIONS READS

40 13,783

6 authors, including:

Apostolos Papanikolaou George Zaraphonitis


National Technical University of Athens National Technical University of Athens
450 PUBLICATIONS 5,209 CITATIONS 122 PUBLICATIONS 1,297 CITATIONS

SEE PROFILE SEE PROFILE

Evangelos Boulougouris Uwe Langbecker


University of Strathclyde DNV GL
230 PUBLICATIONS 2,154 CITATIONS 16 PUBLICATIONS 81 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Apostolos Papanikolaou on 01 June 2014.

The user has requested enhancement of the downloaded file.


J Mar Sci Technol (2010) 15:359–373
DOI 10.1007/s00773-010-0097-7

ORIGINAL ARTICLE

Multi-objective optimization of oil tanker design


Apostolos Papanikolaou • George Zaraphonitis •

Evangelos Boulougouris • Uwe Langbecker •


Sven Matho • Pierre Sames

Received: 12 November 2009 / Accepted: 26 June 2010 / Published online: 22 July 2010
 JASNAOE 2010

Abstract Parametric optimization was applied to a 1 Background


double-hull AFRAMAX tanker design in order to reduce
oil-outflow probability and increase cargo carrying capac- 1.1 Project outline
ity, and the results are presented here. A multi-criteria
optimization procedure was set up in modeFrontier using Following a series of catastrophic single-hull tanker acci-
the cargo volume, the mean oil-outflow parameter and the dents, current IMO regulations (and long before that, US
steel weight of the cargo block as the objective functions. OPA90) state that double-hull tanker designs are the only
Calculations are based on a parametric geometric model of acceptable solution for the safe carriage of oil in tanker
the ship created in NAPA, and on a structural model ships. According to current MARPOL regulations, the tank
created in POSEIDON. Integration of the above software arrangement of the cargo block of an oil tanker should be
packages leads to an automated optimization procedure properly designed to provide adequate protection against
that provides improved feedback to the designer regarding accidental oil outflow, as expressed by the so-called mean
the trade-off between the various design parameters and outflow parameter. The present paper outlines the risk-
optimization criteria involved. The results obtained suggest based parametric optimization of a double-hull AFRA-
notable improvements in transport capacity and oil-outflow MAX tanker in order to achieve innovative designs with
performance for known, well-established yard designs. The increased cargo carrying capacities, reduced steel weights
presented work derives from a joint industrial project and improved environmental protection.
between Germanischer Lloyd (GL) and the Ship Design The research presented here is based on the results of a
Laboratory of the National Technical University of Athens joint industrial project between Germanischer Lloyd (GL)
(NTUA-SDL), which continues the work done and coor- and the Ship Design Laboratory of the National Technical
dinated by NTUA-SDL within the SAFEDOR project on University of Athens (NTUA-SDL). This work is a further
the same subject. elaboration of an innovative risk-based oil tanker design
procedure that was initiated in the framework of the EU
Keywords Design optimization  Risk-based design  project SAFEDOR. Building on the work presented earlier,
Genetic algorithms  Multi-criteria decision making  the integration of the structural design software POSEIDON
Accidental oil outflow [1] into the multi-criteria optimization procedure allows the
realistic estimation of the steel weight of the alternative
designs, and the latest MARPOL regulations for accidental oil
outflow (applicable to all newbuildings after 1 January 2010)
A. Papanikolaou (&)  G. Zaraphonitis  E. Boulougouris
Ship Design Laboratory, School of Naval Architecture have been implemented [2]. The fully automated optimiza-
and Marine Engineering, National Technical University tion procedure developed here provides improved feedback to
of Athens, Athens, Greece the designer regarding the trade-off between the various
e-mail: papa@deslab.ntua.gr
design parameters and the optimization criteria involved.
U. Langbecker  S. Matho  P. Sames The present study focuses on the optimization of the
Germanischer Lloyd AG, Hamburg, Germany arrangement of the cargo area of an AFRAMAX class

123
360 J Mar Sci Technol (2010) 15:359–373

tanker, with the aim being to identify the best-performing For crude oil tankers of C20,000 tonnes DWT and
designs in terms of both reduced accidental oil outflow and product carriers of C30,000 tonnes DWT delivered after
improved economic competitiveness. However, the pro- 1982-06-01, Regulation 18 requires a sufficient capacity of
posed methodology can be extended to include additional segregated ballast tanks. Under ballast conditions, includ-
objectives or design aspects, such as the ship’s hull form ing conditions consisting of lightweight plus segregated
and internal arrangements, and can be easily extended to ballast only, the ship’s draughts and trim should meet the
other oil tanker classes. following requirements:
• Molded draught amidships, dm C 2.0 ? 0.02 L
1.2 Reference design
• Trim by stern B0.015 L
• Draught aft (Taft) should always lead to full immersion
An existing AFRAMAX tanker was selected as the basic
of the propeller(s).
reference design. Its main particulars and general arrange-
ment are presented in Table 1 and Fig. 1, respectively. It is For oil tankers of C5,000 tonnes deadweight delivered
a typical modern AFRAMAX tanker with six tanks along on or after 1996-07-06, Regulation 19 requires ballast tanks
the cargo space and two cargo tanks across, already or spaces other than tanks carrying oil along their entire
adequately optimized by the shipbuilder. cargo tank length to effectively protect the cargo space, and
these tanks or spaces must have the following minimum
1.3 Regulatory framework dimensions:
• Wing tanks or spaces, w = min {0.5 ? DWT/20,000;
Chapter 4 of MARPOL 73/78 [2], which specifies the
2.0 m} [1.0 m
requirements for the arrangement of the cargo areas of oil
• Double-bottom tanks or spaces, h = min {B/15;
tankers constructed after 2010-01-01, was used as the
2.0 m} [1.0 m.
regulatory basis in the present work. In particular, the
following regulations were implemented: It should be noted that the requirements of Reg. 19
regarding the minimum spacing of wing and double bottom
• Regulation 18—requirements for the minimum capac-
from the outer shell (2.0 m for AFRAMAX) are challenged
ity of segregated ballast tanks (SBT)
herein; namely, they are kept flexible during the optimi-
• Regulation 19—requirements for the double-hull
zation runs and set equal to a minimum of 1.7 m for
arrangement
AFRAMAX-sized tankers.
• Regulation 23—requirements for ‘‘accidental oil out-
Regulation 23 applies to oil tankers delivered on or
flow,’’ along with the procedure for its calculation
after 1 January 2010. For oil tankers of 5,000 tonnes DWT
• Regulation 27—criteria for intact stability
and above, it sets the limits for the mean oil outflow
• Regulation 28—criteria for damage stability.
parameter (OM), along with the procedure for its calcula-
tion. For the vessel used in this particular study, with a
Table 1 Main particulars of the reference design
total volume of cargo oil \200,000 m3, an OM value not
Length, oa (m) 250.10 exceeding 0.015 is required. The mean oil outflow
Length, bp (m) 239.00 parameter is calculated independently for side damage and
Breadth, molded (m) 44.00 bottom damage and then combined in nondimensionalized
Depth, molded (main deck) (m) 21.00 form as follows:
Width of double skin sides (m) 2.50
Width of double skin bottom (m) 2.50 OM ¼ ð0:4OMS þ 0:6OMB Þ=C; ð1Þ
Draught scantling (m) 14.60
where OMS and OMB are the mean outflows for the side
Deadweight, scantling draught (tonnes) 112,700
damage and bottom damage, respectively, and C is the total
Cargo capacity (cbm) 127,271
volume of cargo oil in m3 for a 98% full tank. The mean
Slops (cbm) 2,890
outflow due to bottom damage is calculated independently
HFO (cbm) 3,380
for tide conditions of zero and minus 2.5 m, and averaged
DO (cbm) 260
as follows:
Water ballast (cbm) 41,065
Peaks (cbm) 3,500 OMB ¼ 0:7OMBð0Þ þ 0:3OMBð2:5Þ : ð2Þ
Classification Lloyds register
The calculation of the mean outflows for side damage
Propeller diameter (mm) 7,200
and bottom damage is based on a probabilistic approach.
Number of cargo tanks (6 9 2) 12 plus 2 slop tanks
The side damage outflow is calculated by the following
Cargo block length (m) 181.44
formula:

123
J Mar Sci Technol (2010) 15:359–373 361

Fig. 1 General arrangement of the reference design

X
n
OMS ¼ C3 PSðiÞ OSðiÞ ðm3 Þ; ð3Þ
1

where PS(i) is the probability of penetrating cargo tank


i through side damage, OS(i) is the corresponding outflow in
m3, while C3 is an appropriate coefficient. Accordingly, the
bottom damage outflow for either zero or minus 2.5 m tide
conditions is calculated by the following formula:
Xn  3
OMB ¼ PBðiÞ OBðiÞ CDBðiÞ m : ð4Þ
1

In the above equation, CDB(i) accounts for the capture of


oil flowing out of a tank in the double bottom.
Fig. 2 Generic optimization framework

2 Design optimization
to optimally achieve these contradictory objectives, a for-
The main objective of this study was to improve the
mal multi-objective optimization procedure was developed
accidental oil-outflow performance of the reference cargo
and applied.
tank arrangement, while at the same time minimizing the
steel weight and maximizing the cargo capacity. Improving
2.1 Optimization framework
the performance of a ship in terms of oil outflow, maxi-
mization of cargo capacity, and minimization of steel
A generic optimization framework for a system S incorpo-
weight are contradictory objectives; for example, the for-
rates the following main elements (see Fig. 2):
mer requires an increased distance of the cargo space from
the outer shell, resulting in a reduction in cargo tank vol- • Input EI
ume; also, a reduction in the mean outflow parameter can • Design variables D
be achieved with more subdivision, by decreasing the • Design parameters P
average size of each cargo tank, and at the same time • Merit functions L
increasing the steel weight (with a corresponding increase • Constraints G
in construction cost and reduction in payload). Therefore, • Output EO.

123
362 J Mar Sci Technol (2010) 15:359–373

In the context of the present work, the difference mutation methods, in contrast to more traditional optimi-
between the design parameters and the design variables is zation methods which use gradient information to move
that the former are kept constant during an optimization between (successively better) points in solution space. This
study, while the latter are systematically varied to facilitate makes them uniquely adaptive to multi-objective problems
the efficient exploration of the design space and to obtain such as finding Pareto frontiers.
the optimum solution(s). With the Pareto set of nondominated designs in hand,
At the core of the developed optimization framework the designer can select an optimal solution according to
there is a ‘‘parametric design tool,’’ developed within the his preferences. This can be done in a number of ways,
well-known ship design software NAPA [3]. It consists of such as:
a set of macros, developed in NAPA Basic, that facilitate
• Using a utility function to rank the different designs
the fully automatic generation of the detailed layout of the
• Using scatter 2D and 3D diagrams to visually identify
cargo block of a vessel, based on the values of a series of
the more attractive designs, comparing them on the
design parameters and design variables. The design pool is
basis of the designer’s preferred criteria and experi-
then created by systematically varying the design variables
ence-based selection
while using predefined (user-selected) values for the design
• Using other visual tools (parallel plots, histograms,
parameters. This procedure evaluates the fulfillment of a
frequency plots, Student plots, etc.), and deciding
set of constraints, while a set of objectives are optimized at
according to the designer’s experience.
the same time. This approach is holistic in nature and
allows the integration of as many objective functions and
constraints as needed for the design problem at hand [4]. 2.3 Implemented optimization procedure
The generic optimization framework developed by
NTUA-SDL was applied previously to a variety of prob- The optimization procedure applied herein is show sche-
lems, including the optimization of the watertight subdi- matically in Fig. 3. It integrates the following software
vision of RoRo passenger ships [5], and the external packages:
hullform optimization of high-speed ships [6]. This generic
• NAPA [3], a naval architectural software package
procedure was adapted to the present optimization problem
• POSEIDON [1], a structural design and analysis
by adding methods and the corresponding software tools
software package developed by GL
for the structural design of the steel structure of the ship
• modeFrontier [9], a general optimization software
and for the probabilistic assessment of oil-outflow
package.
performance.
Within NAPA, a set of macros were developed in
2.2 Multi-objective optimization order to:
• Create the parametric 3D model of the hullform and
Ship design is a typical optimization problem involving
internal compartmentation
multiple and frequently contradictory objective functions
• Calculate loading conditions
and constraints. The easiest way to address such a multi-
• Perform intact and damage stability calculations
objective problem would be to combine the objective
functions into one, assuming that the relative weights and
relationships between the objectives are known. In most
cases, however, these weights and relationships are
unknown, and there is little knowledge regarding the space
of feasible solutions. Hence, a truly multi-objective meth-
odology is required, leading to a set of ‘‘best designs;’’ in
other words, designs in which no one objective can be
improved without sacrificing the performance of another
objective. This set of ‘‘best designs’’ is known as the Pareto
set. It is represented graphically as the Pareto frontier.
For the present problem, multi-objective genetic algo-
rithms (GA) were selected as the most suitable optimiza-
tion method [7]. Genetic algorithms are stochastic,
nonlinear optimization methods that apply the principles of
biological evolution [8]. In particular, they utilize popula-
tions of solutions and apply selection, reproduction and Fig. 3 Implemented optimization procedure

123
J Mar Sci Technol (2010) 15:359–373 363

• Calculate the accidental oil outflow


• Prepare the necessary geometric data for the software
tools (POSEIDON) that perform the structural design.
POSEIDON implements GL’s latest rules for classi-
fying a ship’s structure (Edition 2008, [10]). It allows the
automatic calculation of the scantlings for all structural
components based on rule requirements for the particular
vessel parameters, class notation, global bending, cargo
loads, and external sea pressure. Note that an additional
module was developed/implemented to create POSEI-
DON models from a set of parameters. The same set of
parameters was used to define compartments in NAPA
and to create the structural model in POSEIDON, hence
ensuring consistency between the two models.
modeFRONTIER is a general-purpose optimization Fig. 4 Hullform modeled in NAPA
scheduler. It provides several optimization algorithms:
genetic algorithms, conjugate gradient method, quasi- 2.6 Constraints
Newton method, sequential quadratic programming, sim-
plex, etc. The various optional algorithms can be com- The following constraints were employed:
bined, such as genetic algorithms for global search and • MARPOL Regulation 18 for mean draft, trim, propeller
another algorithm for local search (refinement). Software immersion, etc.
modules running on different platforms can be integrated • MARPOL Regulation 23, except for the minimum
via a network. spacing of the wings and double bottom, which was set
here for AFRAMAX tankers equal to 1.7 m2
2.4 Design variables • MARPOL Regulation 27—requirements for intact
stability
The parametric definition of the layout and structural • MARPOL Regulation 28—requirements for damage
arrangement of the cargo area of a ship requires a large stability.
number of parameters, controlling the details of the
arrangement and of the various structural components. In
the present study, some of these parameters were kept
constant during each optimization run, while others were 3 Geometric model
treated as free variables and their values were selected (in a
predefined range) by the optimization scheduler. More The geometry of the reference hullform was modeled in
details on the design parameters employed and variables NAPA using available offsets (see Fig. 4). A series of
are given in the following section describing the geometric NAPA macros were developed to parametrically define the
model. internal compartmentation of the design alternatives. In the
geometric modeling, the external hullform and the length
2.5 Objectives and position of the cargo block area were kept fixed.
Typical examples of the variety of configurations that
The following objectives were used: can be parametrically defined are illustrated in Figs. 5, 6, 7
and 8. The details of the internal layout and the structural
• Maximization of the cargo capacity arrangement of the ship along the cargo area are controlled
• Minimization of the accidental oil-outflow parameter by a series of 41 design parameters. The most important of
according to MARPOL Annex I Regulation 23 these can be summarized as follows:
• Minimization of the structural steel weight in the cargo
area while fulfilling the requirements of GL rules for • Compartmentations with one (central) or two longitu-
the construction of double-hull oil tankers (non-CSR).1 dinal bulkheads over the entire cargo block can be

2
The minimum spacing according to MARPOL is 2.0 m; however in
the research presented here, this semi-empirical MARPOL limit was
not considered a hard constraint, but challenged in the framework of a
1
An optimization with respect to CSR is planned for presentation in risk-based design/regulation and approval procedure, as promoted by
the future. the project SAFEDOR [11, 12].

123
364 J Mar Sci Technol (2010) 15:359–373

developed. The number of longitudinal bulkheads is


controlled by the corresponding parameter.
• The number of transverse bulkheads in the cargo area
can be controlled by the user by assigning the value of
the corresponding parameter.
• A set of parameters is used to define the position of the
transverse bulkheads in the cargo area.
• A set of parameters is used to define the double bottom
height within each main transverse zone.
• An additional set of parameters is used to define the
inner hull clearance within each main transverse zone.
Fig. 5 Arrangement with 6 9 2 tanks, corrugated bulkheads, con- • A set of parameters is introduced to control the type of
stant double-bottom height, and inner side clearance
inner hull and double bottom. Depending on the values
of the corresponding parameters, the side of the inner
hull and the double bottom may be:
• Parallel to the center plane and bottom
• Inclined (see Fig. 6)
• Stepped (see Fig. 7).
• The transverse and longitudinal bulkheads can be either
flat or corrugated. The type of bulkhead is controlled by
the corresponding design parameter.
• A set of parameters is used to control the details of the
geometry of the hopper plates of the inner hull.
Fig. 6 Arrangement with inclined double bottom and constant inner- • In the case of two longitudinal bulkheads, the width of
side clearance the central tank as a percentage of the ship’s breadth is
specified by the corresponding design parameter.
• A set of parameters is introduced to control the details
of the geometries of the upper and lower stools in the
case of corrugated bulkheads.
• A set of parameters is used to define the various
structural details, such as the number and positions of
the stringer decks, the stiffener spacing on the shell,
inner bottom, strength deck, transverse members, and
longitudinal bulkheads, etc.
For practical purposes, and considering the importance
of the various design parameters, it was decided to select a
Fig. 7 Arrangement with stepped double bottom and inner hull
subset of them and treat them as free variables, while the
others were assigned constant values. The values of the free
variables (herein called the ‘‘design variables’’) were sys-
tematically varied by the optimization scheduler while
searching for the optimal solutions; the most important
design variables are those that define:
• The position of the transverse bulkheads in the cargo
area
• The double bottom height within each main transverse
zone
• The inner hull clearance within each main transverse
zone
• The number of longitudinal bulkheads (either one or
Fig. 8 Arrangements of the reference design two)

123
J Mar Sci Technol (2010) 15:359–373 365

• The width of the central tank as a percentage of the


ship’s breadth in the case of two longitudinal bulkheads
• The distance between transverse frames
• The distance between longitudinal stiffeners
• The inclination of the hopper plate that connects the
double bottom with the inner hull.
For example, for the typical case of a vessel with 6 9 2
or 6 9 3 tanks (i.e., with five transverse bulkheads inside
the cargo block and one or two longitudinal bulkheads),
this results in a total of 26–27 design variables.

4 Structural model

4.1 Typical AFRAMAX structure Fig. 9 Sample POSEIDON model with outer shell

AFRAMAX-sized oil tankers (80,000 tonnes DWT to


119,999 tonnes DWT) are commonly longitudinally
framed ships over the full length of the cargo block. They
usually include a large number of continuous, longitudinal,
closely spaced stiffeners and a small number of web frames
that are spaced more sparsely. A centerline bulkhead sep-
arates across the two cargo tanks. A hopper sloping plate at
the lower part connects the longitudinal girder with the first
stringer and provides strength and rigidity at the double-
bottom wing space interface. There are typically three
stringers in the wing space that connect the inner hull with
the side shell. Floors, vertical webs in the wing tanks and at
the longitudinal bulkheads, and deck transverses are
arranged at every web frame.
A structural model was created within POSEIDON for
the reference design based on available structural infor-
mation [1]. The model was more detailed in the cargo area
and limited in the bow and stern region; see Fig. 9. The Fig. 10 Sample POSEIDON model without outer shell
structural model was created in such a way that all layouts
and topologies addressed in the previous section on
developed/implemented on top of POSEIDON. The first
geometry modeling could easily be built up in an automatic
one creates a POSEIDON model from a set of parameters,
way. As well as the 15 (16) design variables necessary for
while the latter invokes POSEIDON to determine mini-
geometry modeling, an additional 21 structural design
mum scantlings for plates and stiffeners according to GL
parameters were introduced for the parametric structural
rules. This allows the calculation of the structural weight of
model; see for example Fig. 10.
longitudinal and transverse members. The following sim-
plifications were made for the POSEIDON model:
4.2 Classification rules
• Local structural details required for structural continu-
Germanischer Lloyd rules [10] were applied to calculate ity (i.e., brackets, etc.) were not included in the model
the minimum scantlings for the structural arrangements of • Holes and cut-outs were not considered
the design according to the class notation ‘‘GL ?100A5 Oil • The material for the whole structure was Grade A
Tanker.’’ Common structural rules (CSR) were not imple- (mild) steel
mented here, as the reference ship was not designed under • Scantlings were calculated from a longitudinal strength
CSR rules and the optimized designs should remain com- assessment without taking into account global FE
parable to the reference design. Two modules were calculations, local buckling, or a fatigue assessment.

123
366 J Mar Sci Technol (2010) 15:359–373

Table 2 Alternative configurations 5.1.3 Configuration 3


Arrangement Bulkhead Number
of cargo tanks type of designs The third configuration was created by introducing an
additional longitudinal bulkhead (flat) in the cargo area.
Configuration 1 692 Flat 7,287 The results are shown in Figs. 17, 18 and 19.
Configuration 2 692 Corrugated 1,738
Configuration 3 693 Flat 6,147 5.1.4 Configuration 4
Configuration 4 693 Corrugated 3,270
Configuration 5 792 Flat 3,043 Configuration 4 was derived from configuration 3 by
replacing the flat bulkheads with corrugated ones. The
results are given in Figs. 20, 21 and 22.
5 Case studies
5.1.5 Configuration 5
5.1 Alternative configurations
Finally, for configuration 5, an additional transverse bulk-
Five different configurations were considered, with six or
head (flat) was introduced, leading to the results shown in
seven tanks in the longitudinal direction, two or three tanks
Figs. 23, 24 and 25.
in the transverse direction, and flat or corrugated bulk-
heads. The five different combinations are summarized in
5.2 Discussion of results
Table 2. A total of 21,500 designs were examined in the
present study. In the following figures, only the feasible
The five alternative configurations were selected to allow
designs are shown. The open circles correspond to domi-
the characteristics of the reference design to be validated,
nated designs, while the full circles correspond to designs
as well as to identify possible improvements through an
on the Pareto front. For comparison, the reference design3
analysis of the respective Pareto frontiers. Putting all of the
is also included, and is marked by a full triangle. It should
Pareto frontiers into a single diagram provides better
be noted that the steel weight of the reference vessel is not
insight into the relationships between design objectives,
its actual weight as built, but the weight calculated by the
design parameters and alternative configurations.
POSEIDON software. This ensures full comparability with
Figure 26 clearly shows that the ‘‘6 9 3 flat’’ Pareto
the generated optimal designs.
designs dominate over all other designs. Furthermore, there
are several Pareto designs that have significantly better oil
5.1.1 Configuration 1
outflow and cargo volume performances than the reference
design. This is very interesting result, considering that the
This configuration corresponds to the tank arrangement of
steel weights associated with the following graphs are com-
the reference design. This is the standard configuration for
parable to or even lower than that of the reference design.4
most AFRAMAX vessels. By comparing the obtained
As expected, Fig. 27 shows that, for the same cargo
designs with the reference design, we can identify whether
volume, most of the generated ‘‘6 9 2 flat’’ Pareto designs
the reference design is already on the Pareto front and
have lower steel weights than the other configurations; note
whether improvements are still needed. The results for the
that the structural weights of the generated Pareto designs
three selected objective functions (cargo volume, structural
and the reference ship were calculated using the same
weight and oil-outflow index) are shown in Figs. 11, 12
model, namely POSEIDON. The reference design is again
and 13.
dominated by several ‘‘6 9 2 flat’’ and ‘‘6 9 3 flat’’ designs.
In Fig. 28, the ‘‘6 9 3 flat’’ designs as well as the
5.1.2 Configuration 2
‘‘6 9 2 flat’’ designs dominate over all other designs. The
reference design is again clearly dominated by several
The second configuration considers a change in the struc-
‘‘6 9 3 flat’’ designs. At the same time, practically all of
tural design from flat to corrugated bulkheads. The results
the ‘‘6 9 2 flat’’ Pareto designs have lower steel weights
are given in Figs. 14, 15 and 16.
than the reference design with acceptable oil-outflow
performances.
3
With a 2.5 m side clearance/double-bottom height and an oil- In addition to the above, the following observations can
outflow index of about 0.010 (compared to the corresponding be made:
MARPOL limits of 2.0 m and 0.015, respectively), the reference
design is very environmentally friendly; however, the design shows
4
room for improvement with respect to both cargo carrying capacity Which is a successful practical design, implemented by a major
and steel weight. shipbuilder.

123
J Mar Sci Technol (2010) 15:359–373 367

Fig. 11 Oil outflow versus


cargo volume for
configuration 1

Fig. 12 Oil outflow versus steel


weight in cargo area for
configuration 1

Fig. 13 Cargo volume versus


steel weight in cargo area for
configuration 1

Fig. 14 Outflow versus cargo


volume for configuration 2

123
368 J Mar Sci Technol (2010) 15:359–373

Fig. 15 Outflow versus steel


weight in cargo area for
configuration 2

Fig. 16 Cargo volume versus


steel weight in cargo area for
configuration 2

Fig. 17 Outflow versus cargo


volume for configuration 3

Fig. 18 Outflow versus steel


weight in cargo area for
configuration 3

• None of the corrugated arrangements proved to be better alternative configurations in general. They have impor-
than the flat bulkhead designs. This does not mean that tant advantages in terms of ease of production and
the corrugated geometries should be disregarded as maintenance that have not been considered in this study.

123
J Mar Sci Technol (2010) 15:359–373 369

Fig. 19 Cargo volume versus


steel weight in cargo area for
configuration 3

Fig. 20 Outflow versus cargo


volume for configuration 4

Fig. 21 Outflow versus steel


weight in cargo area for
configuration 4

Fig. 22 Cargo volume versus


steel weight in cargo area for
configuration 4

• The ‘‘7 9 2 flat’’ arrangement performed poorly, as the • The reference design appeared to be on the Pareto front
steel weight increased without any significant gains in of the ‘‘6 9 2 flat’’ designs. It has already been noted
the outflow or the capacity. that the reference design is a proven design in practice,

123
370 J Mar Sci Technol (2010) 15:359–373

Fig. 23 Outflow versus cargo


volume for configuration 5

Fig. 24 Outflow versus steel


weight in cargo area for
configuration 5

Fig. 25 Cargo volume versus


steel weight in cargo area for
configuration 5

Fig. 26 Outflow versus cargo


volume—Pareto designs from
different configurations

123
J Mar Sci Technol (2010) 15:359–373 371

Fig. 27 Cargo volume versus


steel weight in cargo area—
Pareto designs from different
configurations

Fig. 28 Outflow versus steel


weight in cargo area—Pareto
designs from different
configurations

Fig. 29 Design ranking


according to scenario #1

which was optimized with respect to steel weight (by 5.3 Multi-criteria decision-making (MCDM) problem
the yard designer, most likely using FEM). and optimal design selection
• The proof of the dominance of the ‘‘6 9 3 flat’’ designs
holds for this particular AFRAMAX vessel size, which Two different MCDM scenarios were examined using the
is on the border with that of SUEZMAX. utility functions technique [9]:

123
372 J Mar Sci Technol (2010) 15:359–373

1. Scenario #1: the same preference for all objectives is scenario; namely, the increase in cargo volume was con-
assumed; see Fig. 29, Eq. 5, Table 3. sidered to be much more important than the other objec-
2. Scenario #2: the cargo volume (corresponding to the tives (relative weights of 0.75: 0.125: 0.125). In that
revenue) is considered more important than the initial scenario, design #2069, with the characteristics shown in
cost (steel weight) and the environmental impact Table 4, becomes the optimal one. This is due to a sig-
(outflow); see Fig. 30, Eq. 6, Table 4. nificant increase in the cargo capacity (?8%). For this
1 design, the accidental oil outflow is increased by 10% (but
wcv ¼ wsw ¼ wout ¼ ð5Þ
3 it still remains well below the regulatory requirements),
3 1 while the steel weight is reduced by 2%. This design is also
wcv ¼ and wsw ¼ wout ¼ ; ð6Þ better than the reference design. Design #2122 is again
4 8
ranked second here due to a smaller increase in cargo
where wcv, wsw, and wout are the utility functions at satu- volume (?7%) and reduction in steel weight (1%). Its
ration for the cargo volume, steel weight and oil outflow, arrangement is shown in Fig. 31.
respectively.
When all three objectives are considered to be equally
important (scenario #1), design #1710 (with the charac- 6 Conclusions
teristics shown in Table 3) is found to be the optimal one.
This is due to the significant reduction in the oil outflow A multi-objective optimization procedure for the devel-
during collisions or grounding accidents (-23%). At the opment of efficient and environmentally friendly tanker
same time, the cargo volume is also increased (?2%) and designs has been developed. The implemented procedure is
the steel weight is reduced by 2%. This design is in every largely automated and combines the use of the naval
respect better than the reference design. It is interesting to architectural software package NAPA, the optimization
note that design #2122, which is ranked second, achieves software FRONTIER, and the structural design software
less of a reduction in oil outflow (-6%), but a greater POSEIDON. The application of the optimization proce-
increase in cargo volume (?7%) and a small reduction in dure to an AFRAMAX design—already optimized by the
steel weight of 1%. yard—showed that:
Based on the results of the first assessment scenario
assumed, the preferences were modified in the second Table 4 Comparison of optimum and reference designs according to
scenario #2

Table 3 Comparison of optimum and reference designs according to Ref. 6 9 3 Flat 6 9 3 Flat
scenario #1 design

Ref. design 6 9 3 Flat 6 9 3 Flat ID 2069 2122


Rank 1 2
ID 1710 2122
Cargo 126765 137494 (?8%) 135950 (?7%)
Rank 1 2 volume
Cargo vol 126765 129804 (?2%) 135950 (?7%) Oil outflow 0.01006 0.0111 (?10%) 0.00942 (-6%)
Oil outflow 0.01006 0.00777 (-23%) 0.00942 (-6%) Wst cargo 11077 10894 (-2%) 11013 (-1%)
Wst cargo area 11077 10908 (-2%) 11013 (-1%) area

Fig. 30 Design ranking


according to scenario #2

123
J Mar Sci Technol (2010) 15:359–373 373

• Optimization of the fit of the hullform to the cargo


block, along with the minimization of fuel con-
sumption and emissions
• Extending to other ship sizes such as ULCC, VLCC,
SUEZMAX, PANMAX, etc.
• Refining the probabilistic assessment of oil outflow by
introducing probabilities derived from more recent
damage statistics (beyond MARPOL) in the framework
of a risk-based design procedure, and challenging
existing regulations [12].
Fig. 31 Tank arrangement for design #2122. This design, for which
the tank length was varied lengthwise, demonstrates that finding the
best cargo tank sizes to optimize oil outflow, cargo capacity and steel
weight is not a trivial task, though equally sized tanks are preferable
from a production point of view. In fact, the likely additional References
production cost for this flexibility in tank sizes should also be
considered in an update of the optimization procedure presented here, 1. Germanischer Lloyd (2008) POSEIDON ND v.8.119. Germani-
in which additional costs and benefits should be rationally assessed. scher Lloyd, Hamburg (see http://www.gl-group.com/)
Note that the procedure presented here allows also the easy 2. Marine Environment Protection Committee (2004) Resolution
identification of generated Pareto designs where the tanks are equal MEPC.117(52): Amendments to the annex of the protocol of
in size (as an additional design constraint) 1978 relating to the international convention for the prevention of
pollution from ships, 1973 (MEPC 52nd Session, Agenda Item
24, Annex 2, adopted on October 15). International Maritime
• The reference design was close to the Pareto designs Organization, London
(optimal solutions) generated, which confirms the 3. NAPA Ltd. (2010) NAPA software. NAPA Ltd., Helsinki (see
validity of the modeling set-up http://www.napa.fi/)
4. Papanikolaou A (2009) Holistic ship design optimization. Com-
• Several Pareto designs exhibited improved oil-outflow put Aided Des. doi:10.1016/j.cad.2009.07.002
performances and comparable steel weights and capac- 5. Boulougouris EK, Papanikolaou A, Zaraphonitis G (2004) Opti-
ities to the reference design, whereas other designs misation of arrangements of Ro-Ro passenger ships with genetic
showed improved capacities but slightly worse oil- algorithms. Ship Technol Res 51(3):99–105
6. Zaraphonitis G, Papanikolaou A, Mourkoyiannis D (2003) Hull-
outflow performances form optimization of high speed vessels with respect to wash and
• Particular design features of optimal designs observed powering. In: Proceedings of IMDC 03, Athens, Greece, 5–8 May
in [13] with respect to an increase in double bottom 2003, pp 43–54
height and a decrease in tank size towards the bow were 7. Sen P, Yang J-B (1998) Multiple criteria decision support in
engineering design. Springer, London
confirmed 8. Goldberg D (1998) Genetic algorithms in search, optimization,
• Fine-tuning the hullform around the cargo block is and machine learning. Addison-Wesley, Upper Saddle River
expected to further improve the oil-outflow and cargo 9. ESTECO (2010) modeFRONTIER software. ESTECO, Trieste
carrying capacity performances of the generated (see http://www.esteco.it/)
10. Germanischer Lloyd (2008) Rules for classification and
designs. construction, ship technology, seagoing ships, hull structures.
The way ahead may include: Germanischer Lloyd, Hamburg
11. SAFEDOR (2005) Integrated project on design, operation and
• Enhancing the optimization procedure by including: regulation for safety (EU-funded project, contract TIP4-CT-
2005-516278). Germanischer Lloyd, Hamburg (see http://www.
• Optimization of the local structural design for the safedor.org)
least structural weight 12. Papanikolaou A (ed) (2009) Risk-based ship design—methods,
tools and applications. Springer, New York (ISBN 978-3-540-
• The implementation of common structural rules 89041-6)
(CSR) for the structural design 13. Papanikolaou A, Tuzcu C, Tsichlis P, Eliopoulou E (2007) Risk-
• Other design criteria (e.g., ease of production and based optimization of tanker design. In: Proceedings of 3rd
maintenance, etc.) international maritime conference on design for safety, Berkeley,
CA, USA, 26–29 Sept 2007
• Economic criteria (building, maintenance and oper-
ating costs, RFR, NPV)

123

View publication stats

You might also like