You are on page 1of 13

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

net/publication/259977129

Multi-objective optimization of container ship design

Article  in  Mediterranean Marine Science · October 2013


DOI: 10.1201/b15813-58

CITATIONS READS
19 4,806

5 authors, including:

Apostolos Papanikolaou Lampros Nikolopoulos


National Technical University of Athens University of Strathclyde
438 PUBLICATIONS   4,606 CITATIONS    15 PUBLICATIONS   88 CITATIONS   

SEE PROFILE SEE PROFILE

Pierre C. Sames
DNV AS
36 PUBLICATIONS   767 CITATIONS   

SEE PROFILE

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

COMPUTER AIDED TENDERING DESIGN - ELEFSIS SHIPYARD View project

LOGBASED View project

All content following this page was uploaded by Apostolos Papanikolaou on 13 May 2014.

The user has requested enhancement of the downloaded file.


Multi-objective optimization of container ship design
G. Koutroukis, A. Papanikolaou, L. Nikolopoulos
National Tech. Univ. of Athens, Ship Design Laboratory, Athens/Greece
P. Sames, M. Köpke
Germanischer Lloyd SE, Hamburg/Germany

ABSTRACT: High fuel prices and environmental concerns have influenced the design and operational charac-
teristics of all types of ships, but particularly of container ships in recent years with considerable engineering
as well as commercial effects. Such ECO-ship designs demonstrate lower operational costs and are more
competitive compared to traditional tonnage. This paper presents a holistic, multi-objective optimization pro-
cedure for the design of containerships, encompassing the development of parametric models for the optimi-
zation of medium size (between 3,500 and 4,000 TEU) containerships serving on a given schedule the Intra-
Asian trade. Fully parametric geometric modeling techniques are employed in the frame of the CAD/CAE en-
vironment of FRIENDSHIP-Framework, combined with sophisticated assessment tools for the evaluation of
critical ship design attributes, such as ship’s weight, stability, resistance and powering for ,common opera-
tional conditions. The developed multi-criteria optimization approach enables the effective exploration of an
extended design space (with the utilization of genetic algorithms-NSGA II) targeting to a reduced Required
Freight Rates (RFR), favorable Energy Efficiency Design Indices (EEDI) and minimum ballast water carriage
for common operational conditions, while also disposing enhanced port efficiency.

1 INTRODUCTION This paper will introduce the development of a


highly optimized ship design with emphasis on the
International shipping is the back bone of global Intra-Asian market. At first it analysis the Intra-
trade. The vast majority of transportation of goods is Asian trade and derives demands to ship design. It
carried out by ships. In parallel to intercontinental describes the multi-objective optimization, an ap-
trade, regional trade is also increasing, whereas vari- proach to enhance a simulate port efficiency and
ous regions are developing differently around the concludes with the presentation of a highly competi-
world. One of the most rapidly growing regions for tive ship design for Intra-Asian trade.
sea transport is Intra-Asia. The strong growing Asian
industry is opening markets closer to their origin.
The demands of regional markets may differ from 2 MARKET POTENTIAL AND NEEDS – INTRA
others in regard to the specifications for ship design ASIAN MARKET
and hence ship to ship competiveness. At the same
time shipping is facing different challenges, both Within the last decades Asia transformed into the
economically and regulatory. Economically shipping “global factory”. Initially industry in Japan pros-
faces a long term high of fuel oil prices, over capaci- pered as the first tier of industrializing economies in
ties and low freight rates. On the regulatory side new Asia, followed by the Republic of Korea; Taiwan;
environmental regulations will be or are likely to be Hong Kong; and Singapore; and then spread to the
introduced in the coming years. Increasing energy ef- second-tier economies from the Association of
ficiency will decrease fuel cost and emissions. Re- Southeast Asian Nations (ASEAN), including Ma-
duced operational speed will reduce fuel oil con- laysia, Thailand, and the Philippines, followed by the
sumption significantly due to the non linear relation People’s Republic China (PRC), Indonesia, and
between ship speed and power demand. However, as Vietnam. The majority of the resulting trade is
itineraries and port slots are the major constrains for linked to the production of parts and components,
short sea shipping measures need to identify to re- with final goods trade mainly to North America and
duce speed while keeping schedules at the same Europe. The intra-Asian regional trade share has
time. Reducing average port time enables speed de- grown rapidly in the recent decades from 18.8 % in
crease on the legs between ports. 1990 to 26.7 % in 2008 (WTO 2012).
In the context of this study intra-Asia comprises (Fig. 2). Obviously these cascaded vessels are not
North and South East Asia (Drewry 2008), (GMT operating on their design points and hence may not
2030). The Asian intra regional container trade is the utilize their efficiency potential.
most growing trade compared to other regional Analysis of different Intra-Asian round trips
trades (Fig. 1). Intra-Asian trade is expected to grow showed that multiple port calls and relatively low
about 8.3 % in 2013 (Clarksons 2013). Long term average speeds are characteristic for theses trade. On
predictions for the Intra-Asian trade growth are tre- basis of the exemplarily used itineraries a typical
mendous with up to 150 % in 2020 compared to trade route scenario is derived, see Table 1. The low
2006 figures, which gives about 53 m TEU traded average transit speed necessary to meet the pub-
(Drewry 2008). lished sailing schedules indicates that sufficient
Freight rates as well as charter rates are deter- speed reserve is available to make up for delays. It is
mined by the supply and demand on individual trade also noted that published port stays vary from a few
lanes. The most significant counter measure to over hours to more than one day for the routes investi-
supply is slow steaming, which is widely adopted gated.
within the container fleet. It is estimated that this
measure absorbed about 1.6 m TEU nominal capac- Table 1. Characteristic of typical round trip scenario.
ity (Clarksons 2013). The trend to slower speeds can Round trip length 6,900 nm
also be observed for recent ship new buildings. Port calls per round trip 13
There is a discrepancy between regional trade Average port stay 15 h
growth and the ship size segments scheduled to be Average transit speed 15.5 kn
delivered. The segment of ships with a nominal ca-
pacity of 8000 TEU and more grows over propor-
tional with a share of some 73 % of the total order Current speeds of mid-size container vessels were
book. As a response to the pressure from larger ton- checked using AIS data gathered for the Taiwan
nage delivered to the market a cascading of existing Strait in April 2012. About two thirds of the re-
tonnage can be witnessed. On the strong growing In- corded vessels steamed at 16 to 20 knots (Fig. 3),
tra-Asian trade container vessels have been relatively which means below design speeds but faster than the
small until recently. In 2010, about 5 % of the re- average transit speed derived from public sailing
ported fixtures were for vessels larger than 3,000 schedules.
TEU (Drewry 2011). However, looking at container
vessel operators’ public sailing schedules in 2012 25
24
and 2013 larger vessels could be found, e.g. 4,252
23
TEU (Interasia Lines 2013), 4,583 TEU.
Avg. speed in kn

22
21
20
19
18
17
98

00

02

04

06

08

10

12

14
19

20

20

20

20

20

20

20

20

Figure 2. Average design speed for 3-5,000 TEU segment


[1057 container vessels (IHS 2012)].

35%

30%

25%

20%

15%

10%

Figure 1. Container traffic per trade route (million TEU), 5%


(Drewry 2011), GL extrapolation for 2011 to 2016. 0%
14

24
16

18

20

22

24

(OOCL 2013) and 4,250 TEU (Yang Ming 2013). At


<

>
to

to

to

to

to
14

16

18

20

22

Speed in kn
the same time, 22% of the mid-size container vessel
fleet (3,000–5,000 TEU) is older than 15 years and Figure 3. Measured speed over ground for mid-size container
14% are on order (HIS 2012). For this size segment vessels.
the average design speed changes significantly, de-
creasing from almost 24 kn in the year 1998 to about For ship operators fuel efficiency is the most impor-
19 kn for vessels delivered after 2010 or on order tant item when choosing a vessel. As fuel prices are
on high level and may stay there in future and char- 3.3 Lackenby Variation
ter rates remain volatile there is high demand for en-
In order to control the desired geometrical proper-
ergy efficient ships. IMO’s new energy efficiency
ties of generated designs, namely the block coeffi-
standard the Energy Efficiency Design Index (EEDI)
cient and the longitudinal centre of buoyancy, a
reflects the transport efficiency of a ship, by dividing
Lackenby transformation embedded in the FFW is
its theoretic CO2 emissions and installed power, re-
applied (Abt 2007). This variation allows to shift
spectively, by the ship’s transport work, i.e. speed
sections aft and fore, while fairness optimized B-
multiplied by capacity. A small EEDI means low
Splines are used to ensure smooth transitions and
specific CO2 emissions in regard to conducted trans-
faired hull forms. The required input for the trans-
port work and hence high energy efficiency. Ships
formation is the extent of the transformation, which
contracted after 2013 must not exceed a certain re-
in this case is from the propeller position to the fore
quired EEDI value. This required EEDI value will be
peak, and the difference of the existing and desired
decreased from 2015 on successively each five years
block coefficient and LCB.
by another 10 % referred to 2013 values until 2025
(IMO 2011). Ships which can prove low attained Port Efficiency
EEDI value and can hence prove their energy effi- Calculation

ciency may have a competitive advantage. Due to the Geometric Model

non-linear relation of speed and power, the ship EEDI Calculation

speed has a vast impact on the attained EEDI value. Initial Hydrostatic Cal-
Small speed reductions yield to larger power reduc- Required Freight
tion. Combined with a high optimised low resistance Rate Calculation

design future EEDI thresholds can be met. Lackenby Variation

Required Water Ballast


Calculation
Cargo Hold Modeling
3 METHODOLOGY, WORKFLOW AND
COMPUTATIONAL MODEL DEVELOPED Trim Check for de-
TEU Capacity Calcula- tailed loading conditions

Holism (from ὂλος holos, a Greek word meaning all,


whole, entire, total), is the idea that natural systems Resistance Prediction Stability Check for de-
tailed loading conditions ac-
(physical, biological, chemical, social, economic, cording to IMO criteria

mental, linguistic, etc.) and their properties, should


be viewed as wholes, not as collections of parts. This Machinery Calcula- Deadweight Analysis
often includes the view that systems somehow func-
tion as wholes and that their functioning cannot be Lightship Calculation
fully understood solely in terms of their component
parts. Within this context the Ship Design Labora- Figure 4. Workflow in the developed Optimization methodol-
tory at NTUA has developed methodologies in the ogy.
Friendship Framework (FFW) that can simulate ship
design as a process in a holistic way and applied in
tanker design optimization (Nikolopoulos, 2012) as 3.4 Cargo Hold Modeling and TEU capacity
well as for containership design (Koutroukis, 2012). calculation
The next step in the design simulation workflow is
3.1 Geometric Mode the geometric definition and modeling of the cargo
The core of this methodology and any similar devel- spaces.
oped in a CAD/CAE system is the geometrical In the case of containerships, two distinct design
model (geometrical core). The original surface is features are considered: the in-hold container storage
produced as group of parametric sub-surfaces mod- and the stowage of containers on deck. The in-hold
eled in the FFW. arrangement is bounded by vertical and horizontal
surfaces, such as the deck, double bottom ceiling and
stringers. These form steps on which containers are
3.2 Initial Hydrostatic Properties placed and secured. Every two TEUs (or one FEU)
The Hydrostatic Module is checking the displace- define a cargo bay, while two bays constitute a cargo
ment volume, block coefficient and centre of buoy- hold. These surfaces are parametric and depend on
ancy of each generated design. The calculations are the user defined clearance from the outer hull shell.
conducted within FFW and for the hull form gener- The width of the upper part of each bay is equal to
ated by the Geometry Module. the also by user defined double hull width.
The user (or in an automated procedure the design
scheduler) provides the hull form, its main dimen-
sions, e.g. length, beam, height, in terms of bays, 3.6 Main Engine Calculation
rows and tiers, accompanied by a number of relevant The main engine is dimensioned according to the
regulatory requirements. The Cargo Hold module MAN B&W marine engine program (MAN 2012).
calculates the number of containers per bay and sev- The required power calculated by the Holtrop and
eral other container properties, like centers of grav- Mennen methodology assumes a 20 % weather mar-
ity, volume, moments, etc. gin added and another 5-10 % margin that is used for
Furthermore, on deck stowed containers are also derating purposes. Considering this margin it is en-
modeled and calculated. The stowage plan is defined sured that the design speed is achieved for a load of
for each individual bay both in the vertical (by the the engine in the range of 75 to 80 % MCR, where
number of container tiers) and transverse direction the Specific Fuel Oil Consumption (SFOC) is mini-
mal. Based on the selected engines SFOC, the fuel
(by the number of container rows). The number of
and consumables for a range of 8,000 nautical miles
tiers on deck for each bay is limited by the IMO are calculated.
visibility line coordinates in that particular longitu-
dinal position. The latter is a function of the deck-
house top height and the requirements of SOLAS 3.7 Lightship Weight Estimation
(Chapter V, Reg. 22 – Navigation bridge visibility). The lightship calculation follows through the estima-
The number of bays defines the actual length of the tion of the machinery weight, the outfitting weight
containership. Bays are distributed parametrically aft and the steel weight and the basis of semi-empirical
and fore of the engine room, whereas the position of methods, which were calibrated against known data
the engine room may be also optimized. The snap- of similar ships.
shots (Figs. 5, 6) illustrate the outcome of the above
3.7.1 Machinery Weight
described modeling procedure.
The machinery weight estimation is based on the
Watson-Gilfillan formula (Papanikolaou 1988):
Wm = C md × Pb 0.89 (1)
The latter was recalibrated on the basis of statisti-
cal data for medium size containerships to the fol-
lowing:
1.0241
Wm = 0.0541× Pb (2)

Figure 5. Snapshot of the in-hold arrangement. 3.7.2 Outfitting Weight


For this study, a simplified procedure was followed,
which was calibrated against know date of similar
ships before use. According to Watson's semi-
empirical method (Papanikolaou 1988) outfitting
weight of containerships can be calculated from the
formula:
Wot = 0.32 × L × B (3)
Figure 6. Profile view of ship arrangements. By analysis of available statistical data for the light-
ship, machinery and steel weight, the initial formula
3.5 Resistance Prediction was recalibrated to the following:
Wot = 0.309 × L × B (4)
During the exploratory optimization phase, a great
number of variants need to be evaluated, thus high
processing speed is required. For this reason the ap- 3.7.3 Steel Weight
proximate powering estimation method of Holtrop is During the initial design stages, and the optimization
used that derives from regressed statistical data of the main dimensions stage, it is necessary to iden-
(Holtrop & Mennen 1982). At the second stage of tify the effect of the change of the principal dimen-
optimization, integrated CFD codes are used in order sions of a reference ship on the structural steel
to fine tune and afterwards optimize the bow shape weight. Thus, at first, an accurate calculation of the
of the parametric hull shape, including the fore bulb. steel weight of the reference ship is conducted. Fol-
The stern and transom can also undergo this process lowing this the "Differential Steel Weight" predic-
(particularly in terms of the immersion of the latter) tion method was applied (Papanikolaou 1988).
since the wave making resistance constitutes to a The steel weight of the reference ship was calcu-
significant percentage of ship’s total resistance for lated in different ways, including a detail calculation
containerships. with the ship structural design software of GL PO-
SEIDON, with the resulting value being confirmed weight per TEU for the in hold TEUs (volume) and
by available other data. The structural steel weight of its vertical center of gravity (also of volume). In con-
each variant of the reference ship is then calculated trast to that, the payload weight of the containers
according to the following: stowed on deck depends on each loading condition.
For homogenous loading, a parametric curve (func-
WST 1 = WST 0 × (1 + C1 + C 2 + ... + C 6 ) × (1 + C 7 ) (5) tion) has been developed correlating the number of
C1 = 1.0 × δL / L0 (6) TEUs on deck to the corresponding centers of grav-
C 2 = 0.7 × δB / B0 (7) ity.
C 3 = 0.4 × δD / D0 (8) In the context of this project, three loading condi-
C 4 = 0.25 × C1 (9) tions were addressed and mathematically solved,
C 5 = 0.35 × C 2 (10) namely that of the maximum TEU capacity (nominal
C 6 = 0.65 × C 3 (11) capacity), that of the Zero Ballast Loading Condition
C 7 = 0.3 × C B (12) (zero ballast capacity) and that of a pre-specified av-
erage weight/TEU condition (here the 14tons/TEU
where δL, δB, δD, δCΒ, are the differences from the capacity). The first corresponds to the condition,
equivalent main dimensions of the reference ship where the maximum allowable number of contain-
and C4, C5, C6 are corrections for local structural ers, in terms of IMO visibility, is loaded. The second
elements w.r.t. L, B, D respectively. is a condition where it is assumed that there is no
ballast with the exception of some limited ballast in
the fore and/or aft peak tanks for trim balance. A fi-
3.8 Deadweight Analysis nally the third assumes a homogeneous loading of
The deadweight analysis includes the prediction of 14tons/TEU for all containers onboard and calcu-
the payload of the vessel based on the calculation of lates the maximum number of carried containers.
the consumables. Following the estimation of the
fuel and other consumables weight for the specified 3.10 Required Freight Rate Calculation
range, the payload weight is determined for given
The Required Freight Rate is indicative of the eco-
deadweight capacity. The payload enables to check
nomic performance and competitiveness of the ves-
the stowage factor of the cargo (and average weight sel. The RFR is the second objective of the optimiza-
of carried containers) in relation to the available tion. It represents the minimum freight rate (for
cargo volume. maximum cargo capacity) that can even the ship ex-
penses. As one can understand this index depends
3.9 Stability, Trim and Effect of Water Ballast primarily on two factors: the operating and capital
expenditure from the one hand and the ship's earn-
The stability, trim and required water ballast for the ings (function of the annual transport work) from the
various loading conditions is one of the most critical other hand.
stages in the design of a containership. For a con- The expenditure is separated in two parts: the
tainership that is inherently stability limited, useable capital and operating costs. For the capital costs, the
slot capacity will be a function of the average con- building cost as a function of the ship's main dimen-
tainer weight, the range in container weights and the sions (length, breadth and height) is magnified by a
potential for the stratification of the stowage, with as factor of about 1.7 in order to correspond to actual,
many as possible containers to be stowed on deck. acquisition vessel values for contemporary sales
The stability may be conservatively evaluated by as- (2013) while an interest rate of 8% is used in order
suming a homogenous stow. The assessment of the to calculate the investment Net Present Value
initial and large angle stability of the vessel was un- (NPV). The building cost is broken into hull and out-
dertaken for common type loading conditions in ac- fitting, machinery as well as machinery and piping.
cordance with latest IMO intact stability criteria The operating costs are calculated from a detailed
(IMO 2008). The implementation of the IMO intact description of the operating profile of the vessel
stability criteria is enabled with the integration of the based on the business scenario examined for each
naval architecture software HYDROMAX in the case study. This profile is consisted by the following
FFW design platform. An iterative procedure regard- conditions: Sea Operation and Port Operation, for
ing the marginal values of LCG and VCG was de- each one of those the main engine, auxiliary and
veloped and implemented, ensuring the satisfaction generator loads are defined by the user for each one
of the IMO intact stability criteria and operational of those. In addition the time of them is calculated in
trim requirements. The ultimate goal of this iterative order to be able to derive with the total number of
procedure is to minimize the amount of carried water trips per year given a constant off hire time. The
ballast and identify ‘zero ballast’ loading conditions. time in port is also crucial for the determination of
From the cargo part, the payload weight in the the port efficiency index which is an objective of the
cargo holds derives from an assumed homogeneous optimization, too. Other operational costs such as
manning and provision, spares and stores, insurance conditions to conduct loading and unloading faster
as well as administration are calculated as functions in average, independent of the terminal and individ-
of the ship's principal dimensions and of the main ual voyage stowage. Hence, it is crucial to identify
engine installed power (kW). relevant ship design parameters which have an im-
Another important factor that can manipulate the pact on port efficiency. For this terminals and ship
optimization results is the cost for fuel. When this is owners and operators were asked for issues to be
taken very high, then designs with big powering re- considered regarding ship side port efficiency, incl.
quirements are heavily penalized and ranked lower typical stowage planning, securing cargo, crane
during the optimization process. This is an external times and terminal operation.
way of providing the genetic algorithms that are used Based on their feedback, a method was developed
for the optimization a direction towards favored de- to quantify port efficiency as a function of ship de-
signs that incorporate smaller propulsion needs. For sign. At the design stage of a container ship, how-
this particular reason, it was chosen to use a fuel ever, it is not trivial to predict the actual time used
price of 1,000$/t in order to achieve this external for loading and unloading. Naturally, the time
manipulation as it would happen in real life ship- needed depends strongly on the current number of
ping, where ships with smaller resistance character- containers to be moved and on their specific distri-
istics would be benefited in comparison to others. bution. In addition to the particulars of the ship and
the loading situation at hand, the land facilities
clearly play an important role, most outstandingly
3.11 Energy Efficiency Design Index Calculation the availability and speed of the cranes at the termi-
The Energy Efficiency Design Index (EEDI) is cal- nal. Consequently, so as to quantify the port effi-
culated according to the proposed at IMO resolution ciency of a container ship at its design stage a statis-
MEPC.212(63), using the 70 % deadweight, 75% of tical approach seems appropriate. Within this
the MCR of the engines and the corresponding refer- method port efficiency, defined as the minimum,
ence speed. Two versions are made, one for HFO maximum, mean and significant time needed to
fuel use and one for the case of LNG fuel use: move a selected number of containers, is simulated
in a multi-stage process:
EEDI =
( )(
M nME
)(
∏fj ∑ PME(i) *CFME(i) *SFCME(i) + PAE*CFAE*SFCAE
j=1 i=1
)
+
1. A randomly chosen number of containers are dis-
fi *Capacity*Vref * fw tributed over the vessel.
2. From these containers a lower number of ran-
{( M

j=1
nPTI neff

) }( neff

)
∏fj * ∑ PPTI(i) − ∑ feff (i) *PAEeff (i) *CFAE*SFCAE − ∑ feff (i) *Peff (i) *CFME*SFCME
i=1 i=1 i=1
domly selected containers are identified for load-
fi *Capacity*Vref * fw
ing and unloading.
(12) 3. For this specific load case, i.e., the outcome from
The minimization of this index is one of the pri- steps 1 and 2, the respective time is computed
mary aims of the optimization, which is correlated that one, two or more cranes (here up to six)
both with the deadweight of the design as well as the would need to move the containers the fastest
design speed and subsequently the engine power. (determined by the slowest crane).
The engine power is directly connected to the resis- 4. The simulations, i.e., steps 1 to 3, are repeated un-
tance of the hullform and the deadweight is also til sufficiently large sets of load cases are avail-
connected both to the hullform in terms of displace- able for statistical analysis. (Harries et al. 2013).
ment and to the lightship weight. The simulation delivers an ideal average time for
container moves for a given ship layout. Substantial
differences could be identified for the time spent in
3.12 Port Efficiency Index moving given numbers of containers depending on
On short leg routes and coastal voyages a focal point ship layout. Shorter and beamer design showed
of ship operation lies on time constraints such as higher port efficiency than a longer and more slender
time slots and itineraries. Putting emphasis on en- ship of Panamax size, especially when employing a
hanced port efficiency to allow lower average speeds lower number of cranes (Harries et al. 2013). It was
on voyage legs, while keeping the initial schedule found that a rather even distribution of container
seems to be an appropriate complementary measure. slots longitudinally in longitudinal direction shows
Port efficiency is the reduction of time in port for advantages regarding port efficiency. I.e. if one bay
a constant amount of cargo moved. Most of the ex- is significantly larger than the others this bay will be
isting literature considers port efficiency only on the the bottleneck for total port time, as only one crane
terminal’s side, e.g. (OECD 2012), incl. the effi- is able to work on it. This effect becomes smaller
ciency of port activities and services like pilotage, with decrease of number of available cranes.
towing, tug assistance or cargo handling. In the con- Based on these results an empirical formula was
text of this study port efficiency is considered based developed and implemented in the developed multi-
on the ship layout. Addressing port efficiency al- objective optimisation. The formula estimates the
ready at the design stage the ship will have essential net hourly rate of moved TEU and considers the ra-
tio of on-deck and in-hold container, number of bays 4.1.3 Increased "On Deck" Stowed Containers
and deviation of on-deck containers. It is acknowl- Another design criteria that is of importance when
edged that calculated hourly rates are at present theo- considering the design of a containership is the ratio
retical values, considering ideal conditions and real of the TEUs stowed on deck to those stored in hold
hourly rates may be lower; however, it is believed (or the total number of TEUs). Throughout the fol-
that the used empirical formula is sufficient for a lowing design and optimization work it is evident
qualitative analysis and especially for evaluation of that an effort has been made to increase on deck
competing design variants in the frame of an optimi- stowed containers. The reason for this effort is that
zation procedure. the potential crane moves during port operations,
when the majority of the handled boxes are on deck,
are much less compared to the case of the majority
4 OPTIMIZATION STUDIES being in hold. he uniformity of the container stowage
distribution in ship length direction is identified as
4.1 Design Features important and is herein optimized.
The chosen business scenario refers to a medium
4.1.4 Elliptical Bilge of the Midship Section
size containership that will be involved in the Intra-
Another characteristic of this design concept is its
Asia trade lines. The desired merits of the favored
midship section, which was chosen to feature an el-
designs shall be achieved following a systematic op-
liptic bilge, according to previous work of the Ship
timization that is structured in a rational multi-staged
Design Laboratory on a fully elliptic containership
approach.
(Koutroukis 2012) and a tanker with elliptical bilge
(Nikolopoulos 2012). This extended bilge using the
4.1.1 Wide Beam Design
geometrical properties of the ellipse allows decreas-
The implemented design concept involves wide ves-
ing the wetted surface of the ship while the dis-
sel variants. The primary advantage of such a feature
placement volume is not equally decreased and al-
is the increased stability that results to a smaller bal-
most kept constant. The parameters used to control
last capacity required to satisfy the IMO intact stabil-
this surface are the Flat of Bottom (FOB) extent and
ity criteria. The wide beam design also enables the
the Flat of Side (FOS) extent.
designer to increase very efficiently the TEU capac-
ity and maximize the containers on deck; the design
can lead to smaller ship length, thus reducing the 4.2 Underlying optimization principles.
lightship weight (and increasing the corresponding
The target of any optimization procedure is always
DWT for a given displacement). This decrease of the
to achieve the most desiring values/properties for the
L/B can possibly have a negative effect on the hy-
set optimization objectives. The alteration of the de-
drodynamic performance (powering) of the vessel.
signs and assessed entries is performed through the
The magnitude of this effect, however, is also sub-
systematic variation of their distinctive parameters,
ject to the hull's block coefficient and also the design
while each one of the designs must comply with the
speed and corresponding Froude number. Finally,
set constraints, e.g. stability criteria
the wide beam (and shorter length) design concept
The generic targets or objectives in almost any
generally positively affects ship’s port efficiency.
ship design optimization problem are:
-Safety is herein expressed by ship’s stability as
4.1.2 Design Speed
defined by the IMO stability criteria. Thus, the sta-
The design speed is derived from analyzing different
bility of the vessel may not be only a constraint but
intra-Asian routes, acknowledging speed margins for
also an objective for the optimization routine;
possible delays, etc.. Both design and operating
-Efficiency, expressed by the IMO EEDI index;
speeds depend on the required transit time in a port-
-Competitiveness, expressed by the Required
to-port transport system, thus making the port opera-
Freight Rate.
tions also a vital part of the combined transport sys-
In addition to that, one of the most critical design
tem. Within this context and scope of increased port
issues of future designs in view of the upcoming
efficiency (expressed by the defined empirical index
regulations is the quantity of ballast water carried on
described in previous paragraph) the operating speed
board for trim and stability purposes. Its reduction is
for the business scenario was chosen to be at 19
imperative as several efficiency and fuel costs can be
knots, while the normal transit speed is considered at
linked to the management and treatment procedures
15 kn. When taking into account the speed margin
of the latter. Last, as already explained, in this case
together with the engine margin (for derating pur-
study, port efficiency defined by the mentioned em-
poses), the weather as well as fouling margins en-
pirical relative indicator is one of the optimization
sures that in the normal seagoing condition (with
objectives.
calm seas and clean hull) for the operating speed, the
engine load will be at approximately 60%.
where the same procedure is followed. When it is
Design Generation of Design Variants Design Evaluation:
Variables
SOBOL, NSGAII algorithms
-Required Freight Rate
-EEDI
evident that there little more potential for improve-
-Water Ballast Required
-Port Efficiency Index
ment the best designs are picked using the same
ranking principles with utility functions, and are ex-
ported for analysis. At a later stage, these designs
will be subject to a more detailed, localized optimi-
Design Constraints: zation regarding the hullform and structural design
-Displacement and Deadweight
-IMO Intact Stability Criteria
of the variants, which is not covered in the present
-IMO Visibility Line Regulation
-Trim at Full Load Departure Load. Cond.
paper.

4.4 Design of Experiment and Sensitivity Analysis


The Design of Experiment (DoE) is the first step of
Stage 1: Design Space Exploration
Design of Experiment Sobol Algorithm
the above described multi staged process and aims at
500 variants exploring the design space and optimization poten-
tial given the defined design variables and con-
straints and at providing a concurrent sensitivity
Stage 2: Global Design Optimization step one
NSGA II Algorithm
analysis (the latter is automatically generated from
1000 variants: 50 generations x 20 population the FFW).
Dominant Variants: I.D 780, 1004, 1014, 1013, 388, 251
The design engine that is used for the simulation,
in other words the algorithm responsible for the
variation of the design variables, is the SOBOL algo-
Stage 3: Global Design Optimization, Baseline: I.D 251 rithm which is a quasi-random number generator.
NSGA II Algorithm
1000 variants: 50 generations x 20 population The quasi-random feature is very beneficial in order
to avoid local concentrations and that the variations
are well spread within the range of the design vari-
ables.
Stage 4: Local Design Optimization (post analysis)
Fine Tuning of bow area (incl. bulb) with CFD tools From the results of this DoE, the boundaries and
Structural Design Optimization (Midship Section)
the design constraints are refined in order to have
faster and generate more "targeted" runs in the next
Figure 7. The various stages of the optimization procedure. stages.

Table 2. The design variables and respective boundaries for the


4.3 A multi-staged, "chain" optimization procedure
DoE.
The optimization procedure described in this paper Design Variables Lower Bound Upper Bound
can be described as a stepped (multi stage) one. At LengthBP (Bays) 11 14
first, it is necessary to explore and fully understand B (Rows) 11 14
both the design space (potential for improvement
Double Skin (m) 2 2.5
with given constraints) as well as the sensitivity of
Height (Tiers) 6 8
the methodology by a Design of Experiments
Double Bottom (m) 1.8 2.5
(SOBOL) procedure. The sensitivity analysis is a
very important, preparatory step in which it is en- BilgeHeight (%T) 0.10 1.00

sured that no major, unreasonable manipulations oc- BilgeWidth (%B/2) 0.10 1.00

cur. In addition to that it is important to see that the Par.Body Length (%LBP) 0.00 0.30
results are realistic both on a quantitative and quali- Par.Body Position (%LBP) 0.40 0.55
tative basis, with the latter in need of particular at- dCB -0.06 0.06
tention since the design ranking and selection is the dXCB -0.02 0.02
essence of optimization (the value of a favored de-
sign is not important than the relationship with all The design constraints are primarily stability con-
the other produced designs). straints (IMO criteria), as well as a trim constraint
The following formal optimization runs utilize (trim to be less than 0.5% of Lbp) and bounda-
genetic algorithm techniques (NSGA II algorithm). ries/margins for the displacement as well as the re-
The formal optimization runs involve the determina- quired ballast water.
tion of the number of generations and the definition From the results of the exploration of the design
of population of each generation to be explored. space (Figs 8, 9) it is clear that the designs generated
Then the generated designs are ranked according to a are adequately scattered around the design space for
number of scenarios regarding the mentality of the the displacement range of the examined ship size
decision maker. One favored design is picked to be (Fig. 9). Naturally, as it is very often in ship design,
the baseline design of the next optimization run, larger vessels are favored in terms of the Required
Freight Rate due to the larger economies of scale. ballast containers it can carry, as the ratio of these
However, in this case, the above described behavior, two capacities appears to have a narrow range.
although visible, can be regarded as less pronounced Following the procedure of the design space ex-
(Fig. 10), since larger vessels in the set design speed ploration, a sensitivity analysis was performed in or-
may suffer comparably a large increase in the re- der to make sure that no major discrepancies are evi-
quired power (and thus of fuel costs).The design dent. Afterwards the variants were ranked using the
with the smallest RFR nevertheless is one of the utility functions technique and one of them was set
largest design variants generated in this stage. as the baseline model for the subsequent global op-
timization runs.

4.5 Global Optimization with the Use of Genetic


Algorithms
This stage can be characterized as the stage of the
formal, global optimization. The optimization is
global since the design variables are ship’s principal
particulars and the objectives generic type, yet very
representative indicators. The target of the optimiza-
tion remains unchanged, while the design variables
and constraints are also the same with the design of
experiment (DoE) exercise, only with fine-tuned
variables (Table 2).
In contrast to the previous stage, the algorithm
Figure 8. Scatter diagram of RFR against displacement for quantifying the above mentioned design variables, is
DoE. the NSGA II, which is a genetic algorithm. Such al-
gorithms, also called "evolutionary algorithms" are
widely used nowadays for optimization purposes and
are based on C. Darwin’s theory of biological evolu-
tion and biological trees (Darwin 1857). From the
scatter diagrams produced during the optimization
one can clearly identify the Pareto fronts that are
generated. As in the case of the Design of Experi-
ment the relationship of the EEDI with the RFR
(Fig. 10) is almost linear and strongly correlated
through the fuel costs and transport capacity. In addi-
tion to that, the selected are all on the same line, in-
dicated that the ranking has been done close to the
optimal Pareto front. The baseline model features a
good RFR, since the margin for improvement, which
is at 5.7%, can be regarded lower than the equivalent
Figure 9. Scatter diagram of TEU nominal capacity against dis-
placement for DoE.
margin for the EEDI at 21.5%. This can be attributed
to the improved hull form characteristics (herein of
Another noticeable behavior of the algorithm, ac- innovative elliptical bilge shape) that can reduce ef-
cording to the results of the first exploratory runs, is fectively the wetted surface and thus ship's frictional
that the EEDI and RFR behavior are strongly corre- resistance.
lated, representing win-win scenarios and that there The optimization effect is also apparent in the
is big potential for improvement in terms of the Zero Ballast capacity, where there is an improve-
EEDI, since all of the generated designs are well be- ment of up to 13.5% compared to the reference
low the baseline set by IMO for the EEDI (for a con- model (Fig. 11). This corresponds however to a lar-
tainer vessel of this deadweight it should be at about ger ship that has also a lower RFR. Also noticeable
22g CO2/t-nm). The potential for optimization is are two distinct areas in the design clouds: one with
also witnessed when looking at the relationship of larger designs and one with smaller. This is due to
the RFR with the allowable number of TEU to be the integer/discrete variation of the rows loaded (in
loaded with zero ballast (Zero Ballast Loading Con- the cargo hold) in the transverse direction (that in
dition). The relationship between the RFR and the turn determine the beam of the ship). At each of the
allowable TEUs can be seen as linear, as those two clouds a Pareto front is formulated (red lines) and it
objectives are correlated via the Maximum Nominal is obvious that the selected designs from the ranking
TEU capacity. The larger the ship is, the more zero process are close to these lines.
4.6 Selection and ranking of dominant variants with
the use of utility functions
One of the most critical steps during optimization
of any system is the selection and the sorting of the
dominant variants. For this particular reason it is
necessary to follow a rational, rather than an intui-
tive, approach in order to consider in an unbiased
way all tradeoffs that exist. One such method is util-
ity functions technique.
The optimum solution in our case would dispose
the minimum EEDI value, the maximum value of
the Zero Ballast Container Capacity, the maximum
value for the Port Efficiency Indicator and the mini-
mum RFR value. Instead of using fixed weights for
Figure 10. Scatter diagram EEDI against RFR for NSGAII run.
the set criteria in the evaluation of the variants, we
rather assume a utility function as following

U = wEEDI × u(EEDI ) + wRFR × u(RFR) + wPEI × u(PEI )


+wZeroBallastTEU × u(ZeroBallastTEU )

The maximization of this utility function is the


objective now, and the dominant variants of those 10
most favorable with respect to the 5 defined utility
scenarios (Table 3) resulting in the identification and
sorting of 50 designs with best performance accord-
ing to each utility scenario.

Figure 11. Scatter diagram Zero Ballast TEU Capacity against


RFR for NSGAII run.

The same phenomenon, but with more scattered ar-


eas can be seen for the relationship of the Port Effi-
ciency Indicator (PEI) to the RFR. Although some
evidence of Pareto front is visible (red line). There
is a great number of designs that are away from that
dense area, in which the merits of both the port effi-
ciency and RFR reach their optimal values. It should
be noted that the selected designs that are far away
from the front are designs that have a superior per-
formance regarding the Zero Ballast TEU capacity.
This is evident due to the antagonistic nature of
these two objectives, since this capacity can only be Figure 12. Ranking of dominant variants according to U1 Sce-
achieved with increased in hold volumes of contain- nario.
ers (corresponding to designs with increased tiers
stored in hold). This in turn is reducing the PEI since An important point that played a substantial role in
as explained earlier it depends strongly on the ratio the selection process is the repetition of certain de-
of the on deck stowed TEUs to those stored in hold. signs in the top positions of each ranking. This phe-
This created a strong competition between the two nomenon indicates the robustness as well as the
objectives that makes the use of utility functions im- adaptability of the selected designs. Such design
perative in order to assist the decision maker. variants are those with I.D’s 780, 1013 and 388.
Table 3. The weights for each scenario. limited loading the aft and fore peak tanks for trim
1st 2nd 3rd 4th 5th
Objective
Scenario Scenario Scenario Scenario Scenario improvement purposes. The increased Port Effi-
RFR 0.25 0.7 0.2 0.1 0.4 ciency that was defined within the methodology
EEDI 0.25 0.1 0.2 0.1 0.1 based on previous research activity, allowed the de-
Port signers to lower design speed with no implications in
Efficiency 0.25 0.1 0.2 0.4 0.1
Indicator the supply chain of the Intra-Asian trade studied in
Zero
Ballast
this paper. This reduction is beneficial both in terms
0.25 0.1 0.4 0.4 0.4
TEU
Capacity
of fuel costs as well as emissions and efficiency
without sacrificing the competitiveness of the vessel
(in terms of trips per year).
Table 4. Comparison of dominant variants to reference ship.
Baseline
Model I.D I.D I.D
(Result 780 1013 388
Table 5. Principal particulars of dominant variants.
of DoE) Baseline I.D I.D I.D
EEDI 13,862 12,314 -11,17% 11,566 -16,57% 12,153 -12,33% Model 780 1013 388
Required LBP (m) 211.75 213.22 213.24 217.82
Freight 841,49 840,5 -0,11% 793,2 -5,74% 828,1 -1,59%
Rate
Port Breadth (m) 37.3 39.6 40.06 42.09
Efficiency 32,73 31,19 -4,70% 34,412 5,14% 35,738 9,19%
Indicator
Zero
Ballast Depth (m) 19.80 20.12 17.33 17.25
2791 3168 13,51% 1712 38,6% 2222 20,39%
TEU
Capacity
Design Draft (m) 11 11 11 11

When looking at their characteristics, compared to CB 0.558 0.623 0.643 0.591


the reference design, one can see drastic improve-
ments for all four objectives. It is also important to LCB 104.47 105.8 107.6 102.2
notice again the antagonistic relationship of the Zero
Ballast TEU Capacity and the Port Efficiency Indica- No. Bays 10 10 10 10
tor, as the designs that have an improved capacity
have large in hold capacities (additional tiers) that No. Rows 13 14 14 15
decrease the value of the port efficiency. These two
Tiers on Deck 8 8 8 8
extreme designs are I.D’s 780 and 1013, while I.D
388 can serve as a trade-off between them. For this Tiers in Hold 7 7 6 6
particular reason, I.D 388 was selected as the basis
for the next optimization runs, since the decision Bilge Height
73 28.3 30.1 29
(% of Depth)
maker's mentality in this case was to create designs
Bilge Width
with balanced, rather than extreme characteristics. (% of Breadth)
46 77.8 30.8 87

Double Bottom
1.8 2.084 2.25 2.19
Height (m)
5 SUMMARY AND CONCLUDING REMARKS Double Hull
2.0 2.45 2.31 2.065
Width (m)
The paper presented a holistic ship design method- Parallel Mid-
body Length 0 5.3 11.3 2.32
ology, which was applied to the optimization of a
(%LBP)
medium size, wide-beam containership, using ge- Parallel Mid-
netic algorithms. The presented method has been in- body Position 50 47.8 51.1 48
tegrated into the FFW design software platform and (%LBP)
allows the fully automated generation of valuable Design Speed
19 19 19 19
containership designs with superior design character- (knots)
istics. The obtained results indicate significant im- Lightship
13356 14769 13958 14984
Weight (tonnes)
provements regarding the IMO EEDI, major reduc- Nominal TEU
tions of the RFR and an improvement of the herein Capacity
3667 3983 3968 3945
defined Port Efficiency. Another step forward is the Installed Power
15491 16878 17068 17262
reduction of the required ballast water by almost (kW)
40% and the increase of carried containers (nominal
capacity and containers on deck) that can be loaded
without the need for using ballast for stability pur- 5.1 Future Work-Perspectives
poses. The homogenous weight for this condition is Future work of the presented R&D project includes a
close to the statistically observed homogenous second, more "exhaustive" global optimization
weight for containers (approximately 12-15 tons). analysis, which will be followed by a local optimiza-
This means that in most real-life loading cases the tion regarding two separate aspects of the dominant
ship can waive ballasting with the exception of some designs, namely hydrodynamic performance and
structural design. By use of CFD software, the hull Merk O. & Dang, T. 2012. Efficiency of world ports in con-
form of the vessel will be optimized for a minimiza- tainer and bulk cargo (oil, coal, ores and grain). OECD Re-
gional Development Working Papers.OECD Publishing.
tion of ship’s powering requirements, including as- Nikolopoulos L. 2012. A Holistic Methodology for the Optimi-
pects of added resistance and powering in waves, zation of Tanker Design and Operation and it's applications.
while the structural design will be assessed by the Diploma Thesis. Athens: NTUA.
structural design software POSEIDON using GL OOCL 2013. Schedule OOCL KTX3.
rules for containerships and ship’s midship section http://www.oocl.com/SiteCollectionDocuments/OOCL/eSer
will be optimized for reduced structural weight. vies/Sailing%20Schedule%20by%20Service/KTX3_LT.pdf
Papanikolaou A.2009. Holistic ship design optimization. Com-
puter-Aided Design. Elsevier.
Papanikolaou A. 1988. Ship Design - Volume Α: Preliminary
6 ACKNOWLEDGEMENTS Design Methodology(2nd ed.). Athens: Symeon.
VISIONS-OLYMPICS COMPETITION, 2011, www.visions-
The presented work derives from a joint industrial olympics.eu.
project between Germanischer Lloyd (GL) and the World Trade Organisation [WTO] 2012. Aid for Trade in Asia
and the Pacific – Its Role in Trade-Driven Growth. WTO 3rd
Ship Design Laboratory of the National Technical Global Review of Aid for Trade. Geneva.
University of Athens (NTUA-SDL) and derives from Yang Ming 2013. Schedule Yang Ming PA2.
the earlier introduced concept of the “E4 Container- http://www.yangming.com/yml_services/ymlservice_route_
ship”, which won the 3rd place award in the Euro- map.asp?QT=bySvc&RV=PA2
pean Academic Competition “VISIONS-
OLYMPICS 2011”.
The authors would like to express their sincere
gratitude to the following people:
Dr. Eleftheria Eliopoulou (NTUA-SDL), Ass. Prof
George Zaraphonitis (NTUA-SDL), Markus Ihms
(GL), F. Rohde (GL), P. Securius (GL), S. Harries
(Friendship Systems), V. Shigonovhde (GL), E.
Dölerud (Friendship Systems).

REFERENCES

Abt C. & Harries S. 2007. Hull Variation and Improvement us-


ing the Generalised Lackenby Method of the FRIEND-
SHIP-Framework. The Naval Architect, RINA.
Clarksons 2013. Container Intelligence Quarterly First Quarter
2013- A comprehensive quarterly review and outlook for
the container shipping market. Clarkson Research Services.
Darwin C. 1859. On the origin of species by means of natural
selection, or the preservation of favoured races in the strug-
gle for life. (1st ed.) London: John Murray.
Drewry 2011. Container Market 2011/2012 – Annual review
and forecast.
Drewry 2008. Intra-Asia Container Trades - Demystifying the
Market.
FRIENDSHIP-SYSTEMS: FRIENDSHIP- Framework, www.
friendship-systems.com
Harries, S. & Dölerud, E. & Sames, P.C. 2013. Port Efficiency
Simulations for the Design of Container Carriers. COMPIT
2013.
Holtrop J. & Mennen G. 1982. An approximate power predic-
tion method. International Shipbuilding Progress.
IMO MEPC 62/24/Add.1 2011. Report of the Marine Envi-
ronment Protection Committee on its sixty-second session.
IMO 2008. Intact Stability Code.
IHS Fairplay 2012. Maritime Statistics.
Interasia Lines 2013. Schedule JAPAN/STRAITS/WEST
ASIA.http://www.ial.co.jp/ecsv/ezcatfiles/ial/img/img/207/
OUTBOUND09.pdf.
Koutroukis G. 2012. Parametric Design And Multiobjective
Optimization-Study Of An Ellipsoidal Containership. Di-
ploma Thesis. Athens:NTUA.
Lloyd’s Register 2013. Global Marine Trend 2030”. QinetiQ,
University of Strathclyde
MAN B&W 2012. Marine Engine Program.

View publication stats

You might also like